Nature Immunology

© 2009 Nature America, Inc. All rights reserved.
www.nature.com/natureimmunology
EDITORIAL OFFICE immunology@natureny.com
75 Varick Street, Fl 9, New York, NY 10013-1917
Tel: (212) 726 9207, Fax: (212) 696 9752
Editor: Jamie D. K. Wilson
Senior Editors: Christine Borowski, Laurie A. Dempsey
Copy Editor: Jennifer Fosmire
Production Editors: Marina Corral, Matt Hansen
Senior Illustrator: Katie Vicari
Illustrator: Kimberly Caesar
Senior Editorial Assistant: Adam Lipkin
MANAGEMENT OFFICES
NPG New York
75 Varick Street, Fl 9, New York, NY 10013-1917
Tel: (212) 726 9200, Fax: (212) 696 9006
Executive Editor: Linda Miller
Chief Technology Officer: Howard Ratner
Head of Nature Research & Reviews Marketing: Sara Girard
Marketing Manager: Leah Rodriguez
Assistant Production Coordinator: Karen Wilson
Head of Web Services: Anthony Barrera
Web Production Editor: Colleen Diez
NPG London
The Macmillan Building, 4 Crinan Street, London N1 9XW
Tel: 44 207 833 4000, Fax: 44 207 843 4996
Managing Director: Steven Inchcoombe
Publishing Director: Alison Mitchell
Publisher: Stephanie Diment
Editor-in-Chief, Nature Publications: Philip Campbell
Marketing Director: Della Sar
Director of Web Publishing: Timo Hannay
NPG Nature Asia-Pacific
Chiyoda Building, 2-37 Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843
Tel: 81 3 3267 8751, Fax: 81 3 3267 8746
Publishing Director — Asia-Pacific: David Swinbanks
Associate Director: Antoine E. Bocquet
Manager: Koichi Nakamura
Senior Marketing Manager: Peter Yoshihara
Asia-Pacific Sales Director: Kate Yoneyama
Asia-Pacific Sales Manager: Ken Mikami
DISPLAY ADVERTISING display@natureny.com (US/Canada)
display@nature.com (Europe)
nature@natureasia.com (Asia)
Global Head of Display Advertising Sales: John Michael, Tel: 44 207 843 4960, Fax: 44 207
843 4996
Asia-Pacific Sales Manager: Ken Mikami, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746
Display Account Managers:
New England: Sheila Reardon, Tel: (617) 399 4098, Fax: (617) 426 3717
New York/Mid-Atlantic/Southeast: Jim Breault, Tel: (212) 726 9334, Fax: (212) 696 9481
Midwest: Mike Rossi, Tel: (212) 726 9255, Fax: (212) 696 9481
West Coast South: George Lui, Tel: (415) 781 3804, Fax: (415) 781 3805
West Coast North: Bruce Shaver, Tel: (415) 781 6422, Fax: (415) 781 3805
Germany/Switzerland/Austria: Sabine Hugi-Fürst, Tel: 41 52761 3386, Fax: 41 52761 3419
United Kingdom/Ireland/France/Belgium/Eastern Europe: Jeremy Betts, Tel: 44 207 843 4968,
Fax: 44 207 843 4749
Scandinavia/The Netherlands/Italy/Spain/Portugal/Israel/Iceland: Graham Combe,
Tel: 44 207 843 4914, Fax: 44 207 843 4749
Greater China/Singapore: Gloria To, Tel: 852 2811 7191, Fax: 852 2811 0743
NATUREJOBS naturejobs@natureny.com (US/Canada)
naturejobs@nature.com (Europe)
nature@natureasia.com (Asia)
US Sales Manager: Peter Bless, Tel: (212) 726 9248, Fax: (212) 696 9482
European Sales Manager: Andrew Douglas, Tel: 44 207 843 4975, Fax: 44 207 843 4996
Asia-Pacific Sales Manager: Ayako Watanabe, Tel: 81 3 3267 8765, Fax: 81 3 3267 8746
SITE LICENSE BUSINESS UNIT
Americas: Tel: (888) 331 6288 institutions@natureny.com
Asia/Pacific: Tel: 81 3 3267 8751 institutions@natureasia.com
Australia/New Zealand: Tel: 61 3 9825 1160 nature@macmillan.com.au
India: Tel: 91 124 2881054/55 npgindia@nature.com
ROW: Tel: 44 207 843 4759 institutions@nature.com
CUSTOMER SERVICE www.nature.com/help
Senior Global Customer Service Manager: Gerald Coppin
For all print and online assistance, please visit www.nature.com/help
Purchase subscriptions:
Americas: Nature Immunology, Subscription Dept., 342 Broadway, PMB 301, New York, NY
10013-3910. Tel: (866) 363 7860, Fax: (212) 689 9108
Europe/ROW: Nature Immunology, Subscription Dept., Macmillan Magazines Ltd., Brunel
Road, Houndmills, Basingstoke, RG21 6XS, United Kingdom. Tel: 44 1256 329 242, Fax: 44
1256 812 358
Japan: Nature Immunology, NPG Nature Asia-Pacific, Chiyoda Building, 2-37
Ichigayatamachi, Shinjuku-ku, Tokyo 162-0843. Tel: 81 3 3267 8751, Fax: 81 3 3267 8746
India: Nature Immunology, NPG India, 3A, 4th Floor, DLF Corporate Park, Gurgaon 122002,
India. Tel: 91 124 2881054/55, Fax: 91 124 2881052
REPRINTS reprint@natureny.com
Nature Immunology, Reprint Department, Nature Publishing Group, 75 Varick Street, Fl 9,
New York, NY 10013-1917, USA.
For commercial reprint orders of 600 or more, please contact:
UK Reprints: Tel: 44 1256 302 923, Fax: 44 1256 321 531
US Reprints: Tel: (212) 726 9278, Fax: (212) 679 0843

experimental autoimmune
encephalomyelitis requires the entry of
disease-inducing T cells into the brain.
reboldi and colleagues (p 514; see also
news and views by Steinman,
p 453) find that T H -17 cells initiate
this disease by entering the brain
through the choroid plexus. The original
image shows human brain tissue in
which choroid plexus epithelial cells
are stained with antibody to CCl20
(fuchsia) and astrocytes are ‘decorated’
by antibody to glial fibrillary acidic
protein (brown). original image by
andrew elston (lifeSpan bioSciences).
artwork by lewis long.
brahms and inteferon (p 447)
Editorial
445 The final push?
Essay
447 Aimez-vous Brahms? A story capriccioso from the discovery of a cytokine family
and its regulators
tadatsugu taniguchi
nEws and viEws
451 Local advantage: skin DCs prime; skin memory T cells protect
akiko iwasaki see also pp 488 & 524
453 Gaining entry to an uninflamed brain
robert C axtell & lawrence steinman see also p 514
455 Crohn’s disease–associated Nod2 mutants reduce IL10 transcription
dana J Philpott & stephen E Girardin see also p 471
457 The Foxo and the hound: chasing the in vivo regulation of T cell populations
during infection
Elia d tait & Christopher a Hunter see also p 504
459 rEsEarCH HiGHliGHts
rEviEw
461 Autophagy genes in immunity
Herbert w virgin & Beth levine
artiClEs
volume 10 number 5 may 2009
471 A Crohn’s disease–associated NOD2 mutation suppresses transcription of human
IL10 by inhibiting activity of the nuclear ribonucleoprotein hnRNP-A1
Eiichiro noguchi, yoichiro Homma, Xiaoyan Kang, Mihai G netea & Xiaojing Ma
see also p 455
480 Autophagy enhances the presentation of endogenous viral antigens on MHC class I
molecules during HSV-1 infection
luc English, Magali Chemali, Johanne duron, Christiane rondeau, annie laplante,
diane Gingras, diane alexander, david leib, Christopher norbury, roger lippé &
Michel desjardins
Nature Immunology (issn 1529-2908) is published monthly by nature Publishing Group, a trading name of nature america inc. located at 75 varick
street, Fl 9, new york, ny 10013-1917. Periodicals postage paid at new york, ny and additional mailing post offices. Editorial Office: 75 varick street,
Fl 9, new york, ny 10013-1917. tel: (212) 726 9207, Fax: (212) 696 9752. Annual subscription rates: Usa/Canada: Us$225 (personal), Us$3,060
(institution). Canada add 7% Gst #104911595rt001; Euro-zone: €287 (personal), €2,430 (institution); rest of world (excluding China, Japan, Korea):
£185 (personal), £1,570 (institution); Japan: Contact nPG nature asia-Pacific, Chiyoda Building, 2-37 ichigayatamachi, shinjuku-ku, tokyo 162-0843.
tel: 81 (03) 3267 8751, Fax: 81 (03) 3267 8746. POSTMASTER: send address changes to Nature Immunology, subscriptions department, 342 Broadway,
PMB 301, new york, ny 10013-3910. Authorization to photocopy material for internal or personal use, or internal or personal use of specific clients,
is granted by nature Publishing Group to libraries and others registered with the Copyright Clearance Center (CCC) transactional reporting service,
provided the relevant copyright fee is paid direct to CCC, 222 rosewood drive, danvers, Ma 01923, Usa. identification code for Nature Immunology:
1529-2908/04. Back issues: Us$45, Canada add 7% for Gst. CPC PUB aGrEEMEnt #40032744. Printed on acid-free paper by dartmouth Journal
services, Hanover, nH, Usa. Copyright © 2009 nature Publishing Group. Printed in Usa.
i

nature immunology
Getting aID
(p 540)
The Foxo factor
(pp 457 and 504)
Skin dendritic cells
(pp 451 and 488)
488 Cross-presentation of viral and self antigens by skin-derived CD103 + dendritic
cells
sammy Bedoui, Paul G whitney, Jason waithman, liv Eidsmo, linda wakim,
irina Caminschi, rhys s allan, Magdalena wojtasiak, Ken shortman,
Francis r Carbone, andrew G Brooks & william r Heath see also pp 451 & 524
496 Divergent functions for airway epithelial matrix metalloproteinase 7 and retinoic
acid in experimental asthma
sangeeta Goswami, Pornpimon angkasekwinai, Ming shan, Kendra J Greenlee,
wade t Barranco, sumanth Polikepahad, alexander seryshev, li-zhen song,
david redding, Bhupinder singh, sanjiv sur, Prescott woodruff, Chen dong,
david B Corry & Farrah Kheradmand
504 Transcription factor Foxo3 controls the magnitude of T cell immune responses by
modulating the function of dendritic cells
anne s dejean, daniel r Beisner, irene l Ch’en, yann M Kerdiles, anna Babour,
Karen C arden, diego H Castrillon, ronald a dePinho & stephen M Hedrick
see also p 457
514 C-C chemokine receptor 6–regulated entry of T H-17 cells into the CNS through
the choroid plexus is required for the initiation of EAE
andrea reboldi, Caroline Coisne, dirk Baumjohann, Federica Benvenuto,
denise Bottinelli, sergio lira, antonio Uccelli, antonio lanzavecchia,
Britta Engelhardt & Federica sallusto see also p 453
524 Memory T cells in nonlymphoid tissue that provide enhanced local immunity
during infection with herpes simplex virus
thomas Gebhardt, linda M wakim, liv Eidsmo, Patrick C reading, william r Heath &
Francis r Carbone see also pp 451 & 488
531 T cell antigen receptor signaling and immunological synapse stability require
myosin IIA
tal ilani, Gaia vasiliver-shamis, santosh vardhana, anthony Bretscher &
Michael l dustin
540 HoxC4 binds to the promoter of the cytidine deaminase AID gene to induce AID
expression, class-switch DNA recombination and somatic hypermutation
seok-rae Park, Hong Zan, Zsuzsanna Pal, Jinsong Zhang, ahmed al-Qahtani,
Egest J Pone, Zhenming Xu, thach Mai & Paolo Casali
551 CorriGEnda
natUrE iMMUnoloGy ClassiFiEd
See back pages.
volume 10 number 5 may 2009
iii

© 2009 Nature America, Inc. All rights reserved.
The final push?
Over 20 years ago, the Global Polio Eradication Initiative was launched. Today, polio is still endemic in four countries.
With great fanfare, the World Health Assembly launched the
Global Polio Eradication Initiative (GPEI) just over 20 years
ago. In what was described as a “magnificent gift from the 20th
century to future generations of children,” public health officials and
volunteers committed themselves to ridding the planet of polio by the
year 2000. Unfortunately, there are still new cases of infection in 2009.
Although worldwide cases of polio have fallen from 350,000 in 1988 to
under 2,000 in 2008, some experts question whether eradication will
ever be possible. Many elements are conspiring against the efforts to
eliminate disease, including the present political and economic crisis.
By 2001, it almost seemed that the battle against polio had been won—
worldwide cases fell to an all-time low of 791. Poliomyelitis type II had
been completely eradicated by 1999. However, since 2003 the initiative
has suffered several setbacks. In Nigeria, rumors began circulating that
the polio vaccine contained the AIDS virus and was part of a Western plot
to sterilize Muslim girls. The vaccine was described as “tainted by evildoers
from America” and was “America’s revenge for September 11th.”
As a result, the Nigerian government halted the vaccination campaign
until the vaccine could be proven safe. As a consequence, 18 formerly
polio-free countries suffered outbreaks traceable to Nigeria. A few countries,
such as Sudan, are still struggling to regain their polio-free status.
At present, polio is endemic in four countries: Nigeria, Afghanistan,
Pakistan and India. In 2007, an intensified eradication effort was
launched by the GPEI to circumvent the remaining technical, financial
and operational barriers that were preventing polio eradication in
these countries (http://www.polioeradication.org/content/publications/
PolioStrategicPlan09-13_Framework.pdf). By mid-2008, two independent
World Health Organization (WHO) advisory bodies concluded
that the intensified efforts could overcome these challenges, and consequently
a new Strategic Plan 2009–2013 has been endorsed. This 5-year
plan combines proven eradication strategies with recently developed
tools and tactics, including improving the efficacy of the vaccine. But
even with a new strategic plan in place, is it realistic to expect a poliofree
world by 2013?
The GPEI will need to overcome many obstacles, not least the political
instability of many regions in the polio-endemic countries. Afghanistan
and Pakistan will probably pose the greatest challenge. Because of
successful vaccine drives, no cases of polio have been reported in the
relatively peaceful northern provinces of Afghanistan, but the conflict–
ridden south and southeastern provinces still suffer recurrent outbreaks.
Access of aid workers to vulnerable communities in these areas has been
increasingly limited. Many aid workers have been killed, abducted or
threatened by criminal gangs and Taliban insurgents in recent years,
according to the Afghanistan NGO Safety Office. Only a few months
ago, a day of tranquility organized by the United Nations Assistance
Mission in Afghanistan to enable immunizations in remote areas was
ediTorial
cancelled by the WHO after two Afghan doctors were killed by a suicide
car bombing caused by the Taliban.
Pakistan has also become increasingly unstable. The Taliban and
Al-Qaeda resurgence in the North-West Frontier Province of Pakistan
is allowing the virus to cross borders between the two countries unimpeded.
Cases of polio have exploded in Pakistan, causing outbreaks
in previously polio-free areas. In 2007, the cleric Mufti Khalid Shah
declared a fatwa on employees of the United Nations, WHO and all
other foreign organizations in 2007, and aid workers in Bannu were
sent a 500-rupee note with a letter stating they could either stop or buy
their own coffin.
Cultural differences also present problems for the vaccine program
in these regions. Finding sufficient numbers of women to join the vaccination
team has been difficult. However, the participation of women
is essential, as by tradition only women can enter a Muslim household
when the husband is away, and women with children are usually better
at persuading other mothers to vaccinate. The GPEI must also persuade
people that the vaccine is safe, improve poor sanitation that can interfere
with uptake of the oral vaccine, and offset fatigue that can set in among
volunteers, donors and the general populace. It only takes one unvaccinated
child to trigger a new epidemic.
Fortunately, fatigue has not affected the organizers of the eradication
movement. In January 2009, the Bill & Melinda Gates Foundation,
Rotary International and the governments of the UK and Germany
pledged US$630 million over 5 years for a massive final push to rid
humanity of this scourge. But this may not be enough. Heidemarie
Wieczorek-Zeul, the German Federal Minister for Economic Cooperation
and Development, estimated that even with the new injection of funds,
the global initiative is still some US$340 million short of its budget for
2009–2010. Given that funding agencies and charities, including the
Gates Foundation, are having to prune grant growth because of the recession,
these are uncertain times for the eradication program.
Nevertheless, there are many reasons to be optimistic. The government
in Nigeria is now firmly behind the GPEI, and Muslim clerics who
initially shunned the vaccine program are now actively campaigning for
acceptance of the polio vaccine. In March, President Barack Obama
announced the mobilization of more troops in Afghanistan to “disrupt,
dismantle and defeat” the terrorist Al-Qaeda network in Afghanistan
and neighboring Pakistan. Although the final push for eradication will
be costly and difficult, it certainly will not compare with the costs the
world would face if polio were allowed to reemerge. Since 1988, more
than 2 million children have been immunized, and it is estimated that
as a result, 5 million fewer people have been paralyzed by polio. It is
a testament to the success of the GPEI that worldwide cases of polio
have fallen by over 99%. Now is not the time to give up, even when the
going seems to be getting tough.
nature immunology volume 10 number 5 may 2009 445

© 2009 Nature America, Inc. All rights reserved.
Aimez-vous Brahms? A story capriccioso
from the discovery of a cytokine family and
its regulators
Tadatsugu Taniguchi
Do you delight in Brahms? Do you delight in immunology? Tada Taniguchi recounts the story of Type 1 interferon and
its downstream regulators.
“The Master said ‘Those who know it are not
comparable to those who love it; those who love
it are not compatible to those who delight in
it.’” —Analects of Confucius
Music aficionados may recognize the title of
this article as a phrase originating from
Françoise Sagan’s romantic novel of the same
name (published in English as Goodbye Again).
A typical laboratory scene on a Sunday afternoon
in the 1970s was one in which I would
whistle the first movement of the Brahms second
piano concerto and then the next portion
of the melody would be whistled back from
the nearby office of my graduate school mentor
Charles Weissmann. In this environment, I
began the scientific achievements that would
shape my career. To the credit of Massimo
Libonati, my mentor from two prior years
in Napoli, and my very rudimentary Italian,
I had made my debut at a scientific meeting
held in Roma in 1972 (ref. 1) before meeting
Charles, my subsequent friend and mentor. I
had left Tokyo for Naples without even a master’s
degree, partly because I was appassionato
about classic music. Indeed, it was through my
desire to do science in Italy that I was able to
derive tremendous enjoyment from my personal
meetings with many artists, including the
tenor Mario del Monaco and a retired German
prima donna who lived in a friend’s villa in
Tadatsugu Taniguchi is in the Department of
Immunology, Graduate School of Medicine and
Faculty of Medicine, at Tokyo University, Tokyo,
Japan.
e-mail: tada@m.u-tokyo.ac.jp
Capri and recounted to me many episodes
of her—sometimes romantic—experiences
with the conductors Bruno Walter, Wilhelm
Furtwängler, Herbert von Karajan and others.
In Charles’ lab at the Institut für
Molekularbiologie I der Universität Zürich,
the project that was to become my PhD thesis
involved site-directed mutagenesis of the RNA
phage Qβ, a technique invented by Charles and
one of my brothers in science, Richard Flavell 2 ,
thereby spawning the concept of ‘reverse
genetics’ that was only later applied to DNA.
Although it was not very easy to catch up with
the other members of the lab—who came to
Zurich from top institutes in the world—in
terms of either language or science, their
exceptional kindness helped me to adjust in a
relatively short time. I was fortunate enough
to publish several papers on Qβ, including a
description of the use of recombinant DNA
technology to generate a plasmid encoding
the genome of this virus as a means to produce
an infectious RNA virus 3 . On weekends,
I was invited to Charles’s home to write papers:
as I recall well, writing was accompanied by
music, often the chamber music of Brahms
and Beethoven but sometimes an opera such
as Berlioz’s Les Troyens which, as many readers
may have experienced, was quite beneficial to
scientific writing.
It was in 1978 that the hitherto unheard-of
word “interferon” (IFN) came to my ears
at a lecture given by Peter Lengyel of Yale
University, a good friend of Charles and a
pioneer of IFN research. It was a stormy experience,
and it made me very curious about the
phenomenon by which this molecule with
Silhouette of Brahms.
ESSAy
antiviral activity is produced in mammalian
cells infected by viruses; in my mind, at the
time, I envisioned IFN as a ‘Sleeping Beauty’
gene(s) awakened upon ‘un bacio’ (a kiss) of
a virus. By this point, I had finished my PhD,
and I was about to return to Tokyo for a new
position when Charles asked me to begin
work on the initial isolation of human leukocyte
IFN (afterward renamed IFN-α) mRNA
and the development of strategies for its eventual
gene cloning, in collaboration with Kari
Cantell in Helsinki.
My curiosity about the nature of IFN was
strongly sostenuto after my return to the Cancer
Institute in Tokyo at the end of 1978, and so I
decided to work on the fibroblast IFN (afterward
renamed IFN-β, or IFNB) gene, which
was presumed to be distinct from leukocyte
IFN 4 , mainly in order to avoid unnecessary
overlap and competition with the project in
Charles’s lab. As it turned out, this decision led
to the identification of IFN-α and IFN-β by
Charles’s and my group, respectively, resulting
in the first recognition of a cytokine gene fam-
nature immunology volume 10 number 5 may 2009 447

© 2009 Nature America, Inc. All rights reserved.
eSSay
Tada Taniguchi, his mentor Charles Weissmann
and colleagues in Zurich (1978). Charles (second
from right) is wearing a uniform because he had
just returned from military service.
ily. My initial inclination was to search around
for someone working on fibroblast IFN; the
gene was extremely ricercato by numerous
pharmaceutical companies, and there were
many aspirations for its clinical use, so I was
not very optimistic about getting support
from them because of their desire to protect
trade secrets and potential patents. Although
IFNs and other soluble mediators of immune
responses, later collectively termed ‘cytokines’,
had already garnered considerable attention
in the biological sciences and medicine at that
time, the structure and function of these molecules
remained elusive, along with the underlying
mechanisms of signal transmission and
the regulation of their expression. Indeed,
addressing these issues was hampered by the
fact that these molecules have pleiotropic, or
multiple, biological activities, are usually produced
simultaneously and are expressed only
weakly by a variety of different cell types, making
it difficult to obtain a single cytokine in
pure form.
My initial efforts to obtain a consistent
source of starting material were in vain until
I read a book about IFN, written in Japanese
with great scientific enthusiasm and dedication
by Shigeyasu Kobayashi of Toray, Inc. I
visited Dr. Kobayashi in Toray’s Basic Research
Institute and explained my enthusiasm for
elucidating the structure of fibroblast IFN and
the prospect that such work could lead to the
discovery of the gene-switching mechanism in
mammalian cells, an idea that was beginning
to be addressed worldwide. He listened to me
attentively, and then promptly told me that he
would be willing to provide large amounts of
his fibroblast cell line for the gene cloning. I was
honestly surprised not only by the speed with
which he made his decision but also because he
placed no conditions on my receiving his help. I
then realized that this was a kind of gentlemen’s
agreement, although nowadays such an agreement
is difficult to imagine: perhaps it was our
shared scientific enthusiasm for the discovery
of IFN, which was almost totally unknown at
that time, that helped forge our collaboration.
After all, “Curiosity has its own reason for
existence,” as Albert Einstein (a fellow Zurich
alumnus) once wrote.
There was another big obstacle to my work:
human gene cloning had not yet been done in
Japan and the regulation of recombinant DNA
technology was so strict that all work had to
be done in a P3 facility using only the safest
bacteria, that is, those that are the most difficult
to grow (for example, Escherichia coli strain
χ1776). Fortunately for me, the Institute had
such a facility almost ready to use, though interestingly
I was required, before the start of my
experiments, to spray Neurospora crassa spores
into the cabinet and monitor their capture so
as to demonstrate the facility’s suitability for
P3-level experiments. Finally, with the strong
support and leadership of the Institute’s magnanimous
director Haruo Sugano, I was soon
allowed to proceed with the gene cloning.
I was aware at the time that the competition
for this project worldwide was very strong.
Moreover, because I had to work solo, I thought
it necessary to devise a strategy that would minimize
my time and effort as much as possible.
The most notable difficulty at that time was
to identify cDNA clones for mRNAs that were
very weakly expressed, of which IFN cDNA
was a typical example. As a result, I came upon
a two-step approach by which I would first
search for all suspected clones before proceeding
to identify the desired clone among them.
I first prepared ‘hot’ ( 32 P-labeled) cDNA from
mRNA isolated from poly(I:C)-stimulated
fibroblast cells, and then hybridized the cDNA
with mRNA from unstimulated cells. After
separation of unhybridized cDNA (probe A)
from hybridized cDNA (probe B), I subjected
the unhybridized cDNA to Grunstein-Hogness
in situ colony hybridization with two identical
sets of filters harboring ~3,600 χ1776 colonies.
I then assayed the four clones that preferentially
hybridized to probe A by a ‘hybridizationtranslation
assay’ to search more rigorously for
the cDNA clone that contained the IFN mRNA
sequence. This strategy worked well for me,
and by the end of the summer of 1979, I was
done 5 . Although I can no longer recall this, my
wife Yoko says that I seldom came back home
before midnight, even on holidays.
I sequenced the cDNA (again solo) by the
Maxam-Gilbert method, though the information
I could obtain from one piece of endlabeled
DNA was very limited at that time.
Finally, after a few months, I elucidated the
coding sequence of fibroblast IFN; it was on 2
February 1980, while I was listening to one of
the Beethoven’s Rasumovsky quartets at home
with my son, who celebrated his second birth-
day that day. In the interim, Kathy Zoon and
Ernest Knight, together with Mike Hunkapiller,
Lee Hood and colleagues, had determined the
N-terminal sequences of human IFN-α and
IFN-β, respectively, and Shigekazu Nagata,
another brother in science in Charles’s lab,
and his colleagues had cloned and expressed
a human IFN-α (IFNA) cDNA 6–8 .
I then received two international phone calls,
one from Charles and the other from Mark
Ptashne, already one of the most renowned
molecular biologists around, who was then
working at Harvard; Mark is also a renowned
violinist and so might perhaps be reclassified
as a ‘molecular violinist’. Charles and I talked
about the structures of IFN-α and IFN-β, and
we eventually published papers back to back
on their sequences 9,10 . Naturally, we were very
interested in comparing their sequences. Mark,
meanwhile, was very keen to exploit the new
technique developed by Lenny Guarente and
Tom Roberts of expressing cDNA in E. coli and
so invited me to Harvard so that we could work
together. At the end of February, Yoko and I
left Tokyo for Zurich with a final destination of
Boston. Charles and I, sitting together again in
his room where we had written so many papers,
began the comparison between our IFN-α and
IFN-β sequences by hand. I remember that it
was he who noticed the first portion of the
sequence similarities, which got us very excited,
though only later would we become aware that
this was just the beginning of the discovery of
many cytokine families.
Charles and I then decided to write a paper
together. He kindly said something to the effect
of, “Well, Tada, you are still young and need further
development.” So, after carefully discussing
the issue further with his colleagues Marco
Schwarzstein and Ned Mantei, who sequenced
the IFN-α cDNA, Charles generously gave me
the first authorship. Our paper was eventually
published as an Article in Nature back to back
with the paper by Rik Derynck, Walter Fiers
and colleagues, who also cloned IFN-β 11,12 .
This was for us a truly singular experience
with Nature in that we originally submitted the
manuscript as a Letter, which the editors then
converted into the longer Article format. Yoko
and I then continued on to Boston, where we
received a warm welcome from Mark and his
colleagues. Thanks to the kind help provided
by Jan Vilcek (IFN assay), Alice Wong (virus),
Vicki Sato (human cells) and many others, the
work was done in a relatively short period of
time 13 . During our sojourn, we were given
many opportunities to interact socially with
many scientists and artists, including Wally
Gilbert, Lew Cantley, Hidde Ploegh, (the late)
pianist Patricia Zander and cellist Yo-Yo Ma.
Of the many concerts we enjoyed, perhaps the
448 volume 10 number 5 may 2009 nature immunology

© 2009 Nature America, Inc. All rights reserved.
most unforgettable was Sergey Rachmaninov’s
second sonata for piano played by his friend
Vladimir Horowitz at Boston Symphony Hall.
Overall, my friendships and experiences from
this time have endured ever since then and are
a great treasure of my life.
After returning to Tokyo, I started working
on another class of cytokines, now known as
the interleukins (ILs), and specifically on T
cell growth factor 14 (since renamed IL-2), with
Junji Hamuro, Hiroshi Matsui and colleagues
at Ajinomoto Inc. IL-2 was also the subject
of considerable hopes for its use in basic and
clinical immunology, and we succeeded in
elucidating the structure of human IL-2 and
producing recombinant IL-2 in 1982 (ref. 15).
It is remarkable that, since then, more than
30 ILs have been cloned and characterized. In
the interim, my colleague Shigeo Ohno joined
my lab and discovered that the 5′ region of the
human IFN-β gene confers virus infection–
dependent activation of a reporter gene—the
inception of the virus-inducible promoter
analysis 16 . In parallel, both the IFN-α and
IFN-β promoters were subsequently analyzed
by many investigators, most notably by Charles
and by Tom Maniatis and colleagues (reviewed
in ref. 17). Indeed, the study of the IFN-β gene
has led to the concept of the ‘enhanceosome’
developed by Tom and his colleagues, which
has become a model for understanding the
mechanisms whereby genes are turned on and
off in mammalian cells 18 .
After I moved my lab and family to Osaka,
another talented colleague, Takashi Fujita,
joined my lab and identified a minimal DNA
consensus sequence that functions as virusinducible
enhancer 19 . Then, working together
with Takashi, a master’s degree graduate student,
Masaaki Miyamoto used an expression-
cloning strategy to screen a cDNA library for
gene products that would bind to Takashi’s
consensus sequence. For several reasons, we
originally proposed to name the newly discovered
factor ‘cytokine regulatory factor-1’, but
Ben Lewin, then editor of Cell, rejected our
proposal. So the gene was instead named ‘interferon
regulatory factor-1’ or IRF1 (ref. 20). Soon
after, another master’s degree student, Hisashi
Harada, discovered a related gene, termed IRF2,
which officially signified the recognition of an
IRF family. The family has since been extended
to nine members in mammals by a number of
other investigators (reviewed in ref. 21). In retrospect,
the term ‘CRF’ gene family might indeed
have been more accurate in the light of what we
know now of this family’s role, but c’est la vie.
The roles of IRFs have since been extensively
studied by a number of groups worldwide.
Indeed, interest in IRFs has only grown, particularly
with the recent discovery of patternrecognition
receptors whose activation evokes
type I IFN and other innate immune responses,
such as Toll-like receptors (pioneered by Jules
Hoffman’s group working on Drosophila Toll
and Ruslan Medzhitov, Charles Janeway and
Bruce Beutler on TLR4 in mammals) and cytosolic
receptors for nucleic acids (by Mitsutoshi
Yoneyama and Takashi Fujita on RIG-I and
MDA5, by my laboratory on DAI/DLM-1/ZBP1
and recently by several groups on AIM2). We
now know that IRF1, IRF3 and IRF7 (and possibly
also IRF5) play critical roles in the regulation
of type I IFN (IFN-α and IFN-β) genes,
wherein the contribution of each depends
on the nature of the stimuli and the cell type
(reviewed in ref. 22). The role of IRFs in other
biological responses, for example, in the regulation
of oncogenesis, has also received much
attention: accordingly, the numbers of articles
describing IRFs are rapidly increasing. I am particularly
grateful to Tak Mak, Shigeru Noguchi
and Nobuaki Yoshida, who so kindly collaborated
with us on the generation and analyses of
a number of IRF knockout mice.
Upon reflection, I find it quite remarkable
that, since its independent discovery by Isaacs
and Lindenman and by Nagano and Kojima
more than 50 years ago, the type I IFN family
of cytokines has become the prototype for
cytokine research over the past three decades.
Indeed, IFN research has led directly to the
discovery of Janus-family (JAK) kinases,
signal transducers and activators of transcription
(STATs), the enhanceosome, IRFs
and IFN-inducible genes; and, as key regulators
of pathological and protective immune
responses, IFNs are likely to lead to new
research discoveries in the years to come.
ACKNOWLEDGMENTS
I thank J. Vilcek and D. Savitsky for their kind
reading of this essay. My particular appreciation
goes to my mentor C. Weissmann, from whom I
learned so much about science and music (and
good jokes), and to the colleagues who spent time
in my laboratory through the many facets of the
studies described above. I also thank my wife Yoko
and my friends, either mentioned herein or not, for
their continuous support. Our studies were mostly
supported by Kakenhi, Grants-in-Aid for Scientific
Research from the Ministry of Education, Culture,
Sports, Science and Technology, Japan. Finally, I
also thank baseball’s Hanshin Tigers for their longstanding
roller coaster of hope and disappointment,
which in addition to music has significantly
enriched my life.
eSSay
1. Taniguchi, T., Libonati, M. & Leone, e. azione della
subtilisina sulla ribonucleasi BS-1. Boll. Soc. Ital. Biol.
Sper. XLVIII, 1115–1119 (1972).
2. Flavell, R.a., Sabo, D.L., Bandle, e.F. & Weissmann,
C. Site-directed mutagenesis: generation of an extracistronic
mutation in bacteriophage Qβ RNa. J. Mol. Biol.
89, 255–272 (1974).
3. Taniguchi, T., Palmieri, M. & Weissmann, C.Q. β DNa–
containing hybrid plasmids giving rise to Qβ phage
formation in the bacterial host. Nature 274, 223–228
(1978).
4. Cavalieri, R.L., Havell, e.a., Vilcek, J. & Pestka, S.
Synthesis of human interferon by Xenopus laevis oocytes:
two structural genes for interferons in human cells. Proc.
Natl. Acad. Sci. USA 74, 3287–3291 (1977).
5. Taniguchi, T. et al. Construction and identification of a
bacterial plasmid containing the human fibroblast interferon
gene sequence. Proc. Jpn. Acad. 55B, 464–469
(1979).
6. Knight, e., Jr, Hunkapiller, M.W., Korant, B.D., Hardy,
R.W. & Hood, L.e. Human fibroblast interferon: amino
acid analysis and amino terminal amino acid sequence.
Science 207, 525–526 (1980).
7. Nagata, S. et al. Synthesis in E. coli of a polypeptide
with human leukocyte interferon activity. Nature 284,
316–320 (1980).
8. Zoon, K.C. et al. amino terminal sequence of the major
component of human lymphoblastoid interferon. Science
207, 527–528 (1980).
9. Mantei, N. et al. The nucleotide sequence of a cloned
human leukocyte interferon cDNa. Gene 10, 1–10
(1980).
10. Taniguchi, T., Ohno, S., Fujii-Kuriyama, y. & Muramatsu,
M. The nucleotide sequence of human fibroblast interferon
cDNa. Gene 10, 11–15 (1980).
11. Derynck, R. et al. Isolation and structure of a human
fibroblast interferon gene. Nature 285, 542–547
(1980).
12. Taniguchi, T. et al. Human leukocyte and fibroblast interferons
are structurally related. Nature 285, 547–549
(1980).
13. Taniguchi, T. et al. expression of the human fibroblast
interferon gene in Escherichia coli. Proc. Natl. Acad. Sci.
USA 77, 5230–5233 (1980).
14. Morgan, D.a., Ruscetti, F.W. & Gallo, R. Selective in
vitro growth of T lymphocytes from normal human bone
marrows. Science 193, 1007–1008 (1976).
15. Taniguchi, T. et al. Structure and expression of a cloned
cDNa for human interleukin-2. Nature 302, 305–310
(1983).
16. Ohno, S. & Taniguchi, T. Inducer-responsive expression
of the cloned human interferon β1 gene introduced into
cultured mouse cells. Nucleic Acids Res. 10, 967–977
(1982).
17. Honda, K., Takaoka, a. & Taniguchi, T. Type I interferon
gene induction by the interferon regulatory factor family
of transcription factors. Immunity 25, 349–360
(2006).
18. Kim, T.K. & Maniatis, T. The mechanism of transcriptional
synergy of an in vitro assembled interferon-β
enhanceosome. Mol. Cell 1, 119–129 (1997).
19. Fujita, T., Shibuya, H., Hotta, H., yamanishi, K. &
Taniguchi, T. Interferon-β gene regulation: tandemly
repeated sequences of a synthetic 6 bp oligomer function
as a virus-inducible enhancer. Cell 49, 357–367
(1987).
20. Miyamoto, M. et al. Regulated expression of a gene
encoding a nuclear factor, IRF-1, that specifically binds
to IFN-β gene regulatory elements. pCell 54, 903–913
(1988).
21. Tamura, T., yanai, H., Savitsky, D. & Taniguchi, T. The
IRF family transcription factors in immunity and oncogenesis.
Annu. Rev. Immunol. 26, 535–584 (2008).
22. Honda, K. & Taniguchi, T. IRFs: master regulators of
signalling by Toll-like receptors and cytosolic patternrecognition
receptors. Nat. Rev. Immunol. 6, 644–658
(2006).
nature immunology volume 10 number 5 may 2009 449

© 2009 Nature America, Inc. All rights reserved.
Local advantage: skin DCs prime; skin memory T cells
protect
Akiko Iwasaki
How the immune system responds to local infection and establishes protective immunity in susceptible tissues
remains unclear. Two new studies show that local tissue-resident dendritic cells prime cytotoxic T lymphocyte
responses and that memory cytotoxic T lymphocytes remain in the tissue to provide antiviral immunity.
Cytotoxic T lymphocytes (CTLs) are key
effector cells that provide protection from
viral infection. CD8 + T cells bearing T cell antigen
receptors specific for a given viral antigen
are primed in the secondary lymphoid organs
by dendritic cells (DCs). Exactly which subset
of DCs primes CTLs in response to viral infection
has been an area of intense debate. Many
pathogens enter the host via a specific niche,
most commonly the mucosal surfaces, where
they establish local acute and chronic infection.
Herpes virus family members represent
a good example of such a pathogen. In particular,
herpes simplex virus type 1 (HSV-1) enters
the human host through the oral mucosa and
establishes latency in the trigeminal ganglion.
Reactivation of latent HSV-1 in the ganglion
leads to anterograde transport of virus back
to the skin, causing ‘cold sore’ lesions around
the mouth. CTLs are important both in controlling
HSV-1 reactivation1 and in limiting
viral replication in the peripheral site of replication2
. Given the restricted nature of HSV-1
replication in the epithelial cells and latency
in the innervating ganglia, which cells prime
CD8 + T cells in the draining lymph node and
how protection is afforded by memory T cells
are important unresolved issues. In this issue
of Nature Immunology, two papers show that
local tissue-resident cells do both: Heath and
colleagues demonstrate that CTL priming is
accomplished mainly by langerin-positive
CD103 + dermal DCs3 , whereas Carbone
and colleagues report that memory CTLs in
the skin remain in the tissue and maximize
Akiko Iwasaki is in the Department of
Immunobiology, Yale University School of Medicine,
New Haven, Connecticut, USA.
e-mail: akiko.iwasaki@yale.edu
NewS AND vIewS
protective immunity to subsequent challenge
with HSV-1 (ref. 4).
There are three DC subsets in the skin:
Langerhans cells in the epidermis, and two
subsets of dermal DCs, consisting of langerinnegative
dermal DCs and the newly described
langerin-positive CD103 + dermal DCs 5–7
(Fig. 1). Traditionally, it has been thought
that local tissue-resident DCs ingest microbial
antigens by phagocytosis and migrate to the
draining lymph node to prime naive T cells.
However, that paradigm has been challenged
by many studies showing that lymph noderesident
DCs are the only cells that present
antigens to T cells. As for the DCs involved in
CTL priming, the importance of the lymph
node–resident CD8α + DCs has become well
accepted 8 . CD8α + DCs are the main antigenpresenting
cells for CTLs that are generated
after infection by HSV-1, influenza virus,
vaccinia virus, lymphocytic choriomeningitis
virus and Listeria monocytogenes. CD8α +
DCs can exclusively prime CTL responses
whether HSV-1 is injected by needle into the
footpad, by the intravenous route or by dermal
abrasion 9 (Fig. 1a). In their study presented
here, Bedoui et al. make the intriguing
observation that reactivating HSV-1 causes a
second phase of infection of the entire skin
dermatome innervated by the infected ganglion
3 (Fig. 1b), which results in a second
wave of antigen presentation in the lymph
nodes draining the new site of viral replication.
Reactivating HSV-1 replicates in the
epithelium, notably in the absence of any
artificial manipulation of the skin, allowing
the authors to study the course of natural
infection. Taking advantage of this system,
they assess DC subsets for their ability to
stimulate CTLs. They find that whereas cross-
presentation of viral antigen during primary
infection after scarification is mediated by the
lymph node–resident CD8α + DCs 3,9 , antigen
presentation after natural infection with HSV-1
in the skin during recrudescence is handled
almost exclusively by the CD103 + dermal DCs
(Fig. 1b). These results are consistent with the
fact that dermal DCs are the main antigen-
presenting cells after natural infection of
vaginal mucosa with HSV-2 (ref. 10) and are also
supported by a study showing differences in the
participation of migrant versus lymph node–
resident DCs in CTL priming after natural
mucosal infection versus skin abrasion with
HSV-1, respectively 11 .
Several intriguing questions arise from this
study 3 . First, why are different DCs involved
in CTL priming during the primary and secondary
infection? Is it possible that scarification
allows HSV-1 to be carried by the lymph,
circumventing the requirement for presentation
by migrant DCs? This is unlikely, as
migrant DCs are still needed for the CD8α +
DCs to prime CTL immunity after scarification
9 , which indicates that even if direct entry
of the virus into the lymph node does occur,
it is insufficient for priming by CD8α + DCs.
Because scarification causes considerable tissue
damage, it is conceivable that the CD103 +
dermal DC functions may be suppressed by
tissue-derived factors, rendering them unable
to prime CD8 + T cells. Langerin-positive dermal
DCs are shown to be responsible for crosspresenting
epidermally expressed self antigen 3
and are required for contact-hypersensitivity
responses 7 , which suggests that these cells are
able to present a diverse set of antigens. In
contrast, langerin-negative dermal DCs are
the main antigen-presenting cell for CD4 +
T cells after scarification-induced HSV-1
nature immunology volume 10 number 5 may 2009 451

© 2009 Nature America, Inc. All rights reserved.
nEwS AnD ViEwS
a Primary infection
b Secondary infection
c
Scarification
Basement
membrane
Spinal column
Dorsal root
ganglion
HSV-1
8
4
Skin
Epidermis
Dermis
Draining lymph
node
infection and after natural reactivation
(Fig. 1a,b). Thus, it will also be important to
understand the cellular and molecular mechanisms
by which CD103 + dermal DCs versus
CD103 – dermal DCs prime CD8 + T and CD4 +
T cells. Second, which wave of T cell priming
results in the establishment of protective
immunity? Are the effector and memory CTLs
induced by the first and second waves of HSV-1
infection quantitatively and qualitatively similar?
The answers to these questions will not
only be important for development of vaccines
against HSV-1 but also provide key clues to the
findings reported by Carbone et al. 4 .
Once primed by DCs, the virus-specific
CD8 + T cells differentiate into at least two
types of memory cells: central memory T cells
home to lymphoid organs and have limited
effector functions, whereas effector memory
T cells home to peripheral tissues and rapidly
secrete cytokines 12 . In their study presented
here, Carbone and colleagues propose the
existence of another type of memory T cells:
tissue-resident memory T cells 4 . The authors
carry out an elegant set of transplantation
studies to demonstrate that these cells reside
both near the latently infected ganglia and
in the skin near the primary site of HSV-1
infection (Fig. 1c). Unlike central and effec-
4
Langerhans cells CD103 4
– dDC Langerin-positive CD103 + dDC CD8α + lymph node DC CD4 + T cell 8 CD8 + T cell
tor memory T cells, tissue-resident memory
T cells do not readily enter circulation once
they establish residency in a given tissue and
can proliferate locally after secondary viral
challenge. Notably, tissue-resident memory T
cells enter and remain in the tissue even in the
absence of virus and, presumably, viral antigens.
This is demonstrated by the finding that
scarification alone induces the recruitment
of these cells to damaged tissue and that skin
containing these cells, when transplanted into
a naive recipient, retains these cells for over 3
weeks even though it is separated from the
latently infected ganglia. Most importantly,
when secondary HSV-1 challenge is applied to
previously infected flank (containing tissueresident
memory T cells) or to the opposite
flank (able to recruit only effector memory
T cells), the tissue containing tissue-resident
memory T cells has 1% as much virus as is
present in the site in which only the effector
memory T cells are newly recruited. However,
effector memory T cells still provide some
protection relative to that afforded by unimmunized
control by diminishing viral load to
1% as much as naive control. These data show
that complete protection from a viral challenge
requires not only systemic CTL memory
but also that tissue-resident memory T cells
be mobilized to the site of potential viral
encounter before infection.
Tissue-resident memory T cells could
represent a distinct lineage of memory cells
that arise from the effector CTL pool, or they
may differentiate from effector memory T
cells once they arrive in the infected and/or
damaged tissue. Tissue-resident memory T
cells have high expression of integrin α 1 β 1
(VLA-1) and CD69 but are CD62L lo and
CD122 lo . However, this expression pattern
is also shared by effector memory T cells.
If tissue-resident memory T cells are a distinct
lineage of memory cells, what factors
influence their development? Or if effector
memory T cells do differentiate into
tissue-resident memory T cells, this would
indicate that the chemokines that recruit
effector memory T cells are responsible for
the eventual existence of these cells in a given
tissue. The cues that are responsible for the
conversion of effector memory T cells into
tissue-resident memory T cells and how long
this takes will be important areas of research.
Another question that needs an answer is
how tissue-resident memory T cells are
retained in a particular tissue, given that viral
antigen is not required. This is particularly
relevant for autoimmune diseases in which
452 volume 10 number 5 may 2009 nature immunology
8
4
4
T RM
T RM
T RM
T RM
Memory
T RM T RM
T EM
T RM
T CM
Blood vessel
Figure 1 The induction and execution of immune responses to HSV-1 are coordinated by local DCs and memory T cells. (a) Primary infection by HSV-1
is initiated by mechanical scarification of the superficial layer of the epidermis, which allows the virus to enter and replicate locally. HSV-1 infects the
innervating ganglion by retrograde transport from the nerve endings in the skin and establishes latency. Virus introduced by this route is taken up by local
skin-resident DCs, which, after migrating, present antigens to CD4 + T cells. However, cross-priming of CD8 + T cells is done uniquely by the lymph node–
resident CD8α + DCs. (b) Reactivation of latent virus in the ganglion results in anterograde migration of infectious virions to the skin and infection of epithelial
cells throughout the dermatome innervated by the ganglion. After this natural route of reinfection, the viral antigens are cross-presented to CD8 + T cells by
langerin-positive CD103 + dermal DCs, not CD8α + DCs. (c) CTLs induced by DCs differentiate into three kinds of memory cells: central memory T cells (T CM ),
effector memory T cells (T EM ) and tissue-resident memory T cells (T RM ). Only tissue-resident memory T cells take up residency in the skin (at the previous
site of virus infection) and near the latently infected ganglion. After tertiary infection by HSV-1, tissue-resident memory T cells provide bulk of the protection,
although effector memory T cells can also be recruited to the site from systemic circulation.
Kim Caesar

© 2009 Nature America, Inc. All rights reserved.
tissue-resident memory T cells may cause
chronic tissue destruction.
Those questions aside, the findings
reported by Gebhardt et al. 4 have important
implications for vaccine development.
Most pathogens gain a foothold in the host
through specific portals of entry. For example,
HIV-1 enters through the genital and
rectal mucosa, whereas Mycobacterium tuberculosis
gains access through the respiratory
mucosa. Thus, provision of effective protection
requires mobilization of tissue-resident
memory T cells to the appropriate mucosal
surfaces. Once again, the mechanism of the
recruitment and retention of these cells
in a given tissue needs to be investigated.
One unconventional proposal would be to
‘scratch and save (a life)’: creating minor
scarification to recruit these cells to the
potential site of pathogen encounter after
conventional parenteral immunization. Of
course, not all surfaces are conducive to this
type of manipulation. A more universal and
powerful approach would be to immunize at
the potential site of pathogen encounter. The
development of safe mucosal vaccines able to
establish tissue-resident memory T cells at
the site of pathogen entry might be the way
to prevent the transmission of deadly virus
infection in humans.
In conclusion, these studies by the groups
of Heath 3 and Carbone 4 provide important
insight into the priming and execution of
antiviral immunity to a local viral infection
and open new avenues of investigation
and, possibly, therapeutic approaches. With
the development of a new in vivo approach
for temporally and selectively depleting the
skin of langerin-positive dermal DCs 7 , future
studies should identify the function of these
cells in immune responses to a variety of
antigens. In addition, elucidating the biol-
Gaining entry to an uninflamed brain
Robert C Axtell & Lawrence Steinman
ogy of tissue-resident memory T cells will
provide clues about the generation of tissuespecific
memory that is needed for vaccine
development and for immune intervention
in autoimmune diseases.
1. Divito, S., Cherpes, T.L. & Hendricks, R.L. Immunol.
Res. 36, 119–126 (2006).
2. Zhu, J. et al. J. Exp. Med. 204, 595–603 (2007).
3. Bedoui, S. et al. Nat. Immunol. 10, 488–495
(2009).
4. Gebhardt, T. et al. Nat. Immunol. 10, 524–530
(2009).
5. Poulin, L.F. et al. J. Exp. Med. 204, 3119–3131
(2007).
6. Ginhoux, F. et al. J. Exp. Med. 204, 3133–3146
(2007).
7. Bursch, L.S. et al. J. Exp. Med. 204, 3147–3156
(2007).
8. Heath, w.R. et al. Immunol. Rev. 199, 9–26
(2004).
9. Allan, R.S. et al. Immunity 25, 153–162 (2006).
10. Zhao, X. et al. J. Exp. Med. 197, 153–162 (2003).
11. Lee, H.K. et al. J. Exp. Med. 206, 359–370
(2009).
12. Lanzavecchia, A. & Sallusto, F. Curr. Opin. Immunol.
17, 326–332 (2005).
Little is known about how pathogenic T cells gain access to the uninflamed brain in multiple sclerosis and
experimental autoimmune encephalomyelitis. A new study reports that interleukin 17–producing T helper cells enter
the uninflamed central nervous system through the choroid plexus by a CCR6-CCL20–dependent mechanism.
The blood-brain barrier is a protective
barricade that limits the entry of large
molecules and cells such as erythrocytes,
platelets and lymphoid cells into the central
nervous system (CNS) in normal conditions.
Although there are well known ‘windows’
in this barrier, particularly adjacent to the
hypothalamus, which allow cytokines such as
interleukin 1 (IL-1), IL-6 and tumor necrosis
factor to elicit fever1 , in general, the brain is
very selective about allowing large molecules
and cells to gain entry. Yet, how autoreactive
T cells gain access to the uninflamed brain
to initiate disease has remained unknown.
In this issue of Nature Immunology, Reboldi
and colleagues2 demonstrate that autoreactive
IL-17-producing T helper cells (TH-17 cells) enter the CNS through a ‘chink in
the armor’ of the blood-brain barrier. In a
special location called the ‘choroid plexus’,
these TH-17 cells initiate the inflammatory
cascade, causing demyelination through an
Robert C. Axtell and Lawrence Steinman are in the
Department of Neurological Sciences and Neurology,
Stanford University, Stanford, California, USA.
e-mail: steinman@stanford.edu
interaction between the chemokine CCL20
and its receptor, CCR6.
For years, the brain has been considered a
site of ‘immune privilege’, a place that tends to
exclude members, both cells and molecules, of
the immune community. ‘Privilege’ has its price,
and of course, the immune system is any case
skilled at outwitting protective ‘covenants’. In
multiple sclerosis, the blood-brain barrier is
breached by a variety of immune cells, including
macrophages, dendritic cells, B cells and autoreactive
T cells 3 . T cells are thought to be the
earliest sentinels that penetrate the blood-brain
barrier, but data to support this idea are scant. So
far, much has been reported on the cellular and
molecular processes involved in the migration of
lymphocytes to the brain through inflamed vessels.
Research on homing through the inflamed
blood-brain barrier has shown that the integrin
α 4 β 1 is critical for this 3 . Such studies of the
blood-brain barrier in inflammation have led
to the development of the most potent drug so
far approved for treatment relapsing remitting
multiple sclerosis, natalizumab, a humanized
monoclonal antibody specific for α 4 β 1 .
The migration of T H-17 cells to the CNS has
been linked to the induction of inflammation
nEwS AnD ViEwS
in multiple sclerosis and in the animal model
of experimental autoimmune encephalomyelitis
(EAE) 4,5 . It has been shown in humans
that T H -17 cells ‘preferentially’ express CCR6
(ref. 6). Reboldi and colleagues now confirm
that mice have a similar expression pattern
in which CCR6 expression is restricted to
T H -17 cells and is not found on the surface
of T helper type 1 (T H 1) or T H 2 cells 2 . Given
those data, the authors next explore the contribution
of this chemokine receptor to EAE.
The authors find that Ccr6 –/– mice are completely
resistant to EAE 2 . This effect is not due
to a block in the differentiation of T H -17 or
T H 1 cells in the lymph nodes after induction
of EAE but is associated with a lower capacity
of T H 1 and T H -17 cells to migrate to the CNS.
EAE disease signs are restored in the Ccr6 –/–
mice by the transfer of myelin-specific T cells
from Ccr6 +/+ 2D2 mice. Notably, in this transfer
system, most of the T cells in the spinal
cords and brain at the peak of disease are not
of the Ccr6 +/+ donor origin but are almost all
from the Ccr6 –/– recipient. This finding indicates
that the initial trigger of inflammation is
caused by CCR6-dependent infiltration of the
uninflamed CNS by autoreactive T H-17 cells
nature immunology volume 10 number 5 may 2009 453

© 2009 Nature America, Inc. All rights reserved.
nEwS AnD ViEwS
Subarachnoid
space
Dendritic
cell
CCR6 + T H -17
cell
that then causes a ‘second wave’ of infiltration
by a CCR6-independent mechanism.
The choroid plexus is a highly vascularized
region in the brain where the cerebrospinal fluid
is formed. It is also a chief site for the entry of
lymphocytes into the CNS, where they perform
their normal immune surveillance of the fluid
that surrounds the brain in the subarachnoid
space 7 . Furthermore, there is speculation that
the choroid plexus is an area in which autoreactive
lymphocytes gain access to the CNS to cause
multiple sclerosis 8 . Notably, the authors establish
that CCL20, one of the ligands for CCR6, has
higher expression in the choroid plexus than in
other regions of the brain in both healthy mice
and mice with EAE 2 . Additionally, they find that
Ccr6 –/– cells of the immune system are unable to
pass through the epithelial barrier of the choroid
plexus to gain access to the CNS during EAE.
These elegant mouse experiments are successfully
‘translated’ to the human arena.
The investigators find that this mechanism
may actually occur during the early stages of
multiple sclerosis. In patients experiencing
their first neurological episode, memory T
cells from the cerebrospinal fluid have much
higher expression of CCR6 on their surface
than do cells from peripheral blood 2 . In
addition, CCL20 protein is present in greater
abundance in the choroid plexus in brains of
both healthy subjects and patients with multiple
sclerosis than in the parenchyma of the
brain. Such ‘translation’ of experiments in the
EAE model to the human disease of multiple
Choroid plexus
Venule
CCL20
Choroid plexus
epithelial cell
Ventricular
epithelial cell
Figure 1 CCR6-dependent entry into the CnS. Epithelial cells of the uninflamed choroid plexus
constitutively express CCL20, which attracts T H -17 cells and facilitates the crossing of T cells into
the subarachnoid space of the CnS. in the subarachnoid space, the autoreactive T cells engage their
cognate peptide–major histocompatibility complex expressed on dendritic cells to initiate inflammation.
sclerosis is to be applauded and represents an
important message for immunologists: findings
in mice beg for confirmation in man.
When results can be compared and found to
be concordant from mouse to man, it is reassuring,
and the importance of any such study
is elevated above those far more numerous
investigations restricted to the mouse alone.
‘Translation’ to man is critical; the most
potent therapy so far for the treatment of
relapsing-remitting multiple sclerosis, natalizumab,
came from initial experiments on the
adhesion of human lymphocytes to inflamed
brains of rodents with EAE 8 .
These data are compelling and the evidence
suggests that the earliest event of inflammation
of the CNS in multiple sclerosis occurs
through the interaction of CCR6 + T H -17 cells
with CCL20 expressed by the epithelium of
the choroid plexus 2 (Fig. 1). After this initial
insult, the vasculature in the deep white
matter becomes inflamed and recruits other
immune cells to the CNS by the classic integrin
VLA-4 (α 4 β 1 )–VCAM adhesion molecule–
dependent mechanism 3,8,9 .
There has been an active debate about
which subset of helper T cells is most critical
for the pathogenesis of multiple sclerosis and
EAE. Originally, multiple sclerosis and EAE
were regarded as T H 1 diseases. Interferon-γ
(IFN-γ) is found in considerable abundance
in the cerebrospinal fluid of patients with
multiple sclerosis and mice with EAE 10 . Mice
deficient in the ‘master’ transcription factor
critical for T H 1 differentiation, T-bet, are
resistant to the EAE 11 . Moreover, an early
clinical trial reported that IFN-γ treatment
exacerbated symptoms in multiple sclerosis
patients 12 . However, subsequent studies
have shown that the T H 1 cytokines IFN-γ
and IL-12 have anti-inflammatory effects in
EAE 13,14 . Mice deficient in these cytokines
develop an aggressive atypical form of EAE
that is indicative of greater infiltration of cells
of the immune system into the brain. Those
data catalyzed the discovery of T H -17 cells,
and now, this population has gained notoriety
as the pathogenic population in EAE and
multiple sclerosis. IL-17 is also found in the
inflamed CNS, and altering T H -17 differentiation
can inhibit signs of EAE in mice 4,5 .
But to complicate the issue, two provocative
papers have provided data that question
the pathogenic potential of T H -17 cells in
neuroinflammation: first, mice with conditional
deletion of IL-17 in T cells develop EAE
normally 15 ; second, experimental uveitis is
cured in rats by treatment with recombinant
IL-17 (ref. 16). Assessing the data reported
by Reboldi and colleagues 2 in the context
of those other observations illuminates the
problems of placing functional importance
on a particular cytokine that might define a
particular subset of autoreactive effector T
cells. With this in mind, the implications of
the data from Reboldi et al. 2 suggest that the
important autoreactive effector T cells are not
those that express IL-17 or IFN-γ but actually
those T cells with an autoreactive brain
tissue–specific T cell antigen receptor that also
concomitantly express CCR6. In other words,
the pathogenic function of these cells is due
to their ability to infiltrate the target tissue
and engage major histocompatibility complex
to generate an immune response against the
nervous system. This feature is independent
of whether the pathogenic T cells are T H 1,
T H -17, T H 2 or ‘T H -9’. These studies, illuminating
how pathogenic T cells gain entry into
the normal brain, are marvelous both for their
implications for understanding immune surveillance
of the nervous system and for the
elegant ‘translation’ of experimental work all
the way from mice to man.
1. Conti, B., Tabarean, i., Andrei, C. & Bartfai, T. Front.
Biosci. 9, 1433–1449 (2004).
2. Reboldi, A. et al. Nat. Immunol. 10, 514–523
(2009).
3. Steinman, L. Nat. Rev. Drug Discov. 4, 510–519
(2005).
4. Tzartos, J.S. et al. Am. J. Pathol. 172, 146–155
(2008).
5. Langrish, C.L. et al. J. Exp. Med. 201, 233–240
(2005).
6. Acosta-Rodriguez, E.V. et al. Nat. Immunol. 8, 639–
646 (2007).
7. Ransohoff, R.M., Kivisakk, P. & Kidd, G. Nat. Rev.
454 volume 10 number 5 may 2009 nature immunology
Kim Caesar

© 2009 Nature America, Inc. All rights reserved.
Immunol. 3, 569–581 (2003).
8. Yednock, T. et al. Nature 356, 63–66 (1992).
9. Engelhardt, B., wolburg-Buchholz, K. & wolburg, H.
Microsc. Res. Tech. 52, 112–129 (2001).
10. Sospedra, M. & Martin, R. Annu. Rev. Immunol. 23,
683–747 (2005).
11. Bettelli, E. et al. J. Exp. Med. 200, 79–87 (2004).
12. Panitch, H.S., Hirsch, R.L., Haley, A.S. & Johnson, K.P.
Lancet 1, 893–895 (1987).
13. willenborg, D.O., Fordham, S., Bernard, C.C., Cowden,
w.B. & Ramshaw, i.A. J. Immunol. 157, 3223–3227
(1996).
14. Gran, B. et al. J. Immunol. 169, 7104–7110 (2002).
15. Haak, S. et al. J. Clin. Invest. 119, 61–69 (2009).
16. Ke, Y. et al. J. Immunol. 182, 3183–3190 (2009).
Crohn’s disease-associated Nod2 mutants reduce IL10
transcription
Dana J Philpott & Stephen E Girardin
The 3020insC mutation in Nod2 is associated with Crohn’s disease, but how it influences disease pathogenesis is
unknown. A new study shows that the 3020insC mutant protein fails to activate a key transcription factor that drives
interleukin 10 expression, resulting in reduced production of this anti-inflammatory cytokine.
Crohn’s disease, a chronic inflammatory
disorder that can affect the entire gastrointestinal
tract1 , is generally believed to
result from an inappropriate immune reaction
to the normal commensal flora of the
intestine. The most persuasive argument
for this conclusion is that in many mouse
models of inflammatory bowel disease,
regardless of the underlying cause, intestinal
inflammation does not develop if the animals
are housed under germ-free conditions.
Moreover, there is some improvement in
patients who take antibiotics or have intestinal
shunts, which divert intestinal contents
away from the inflamed areas. The molecular
basis for this overexuberant response
to commensal flora is not understood, but
the gene encoding nucleotide oligomerization
domain 2 (Nod2) is present within a
previously described Crohn’s ‘hot-spot’ on
chromosome 16 called IBD1, and three Nod2
coding region polymorphisms are associated
with increased susceptibility to disease2,3 . In
this issue of Nature Immunology, Noguchi
et al. find that some of these mutations—
specifically, those in the leucine-rich repeat
(LRR) domain of Nod2—result in suppression
of the production of the immunosuppressive
cytokine interleukin 10 (IL-10).
One of these polymorphisms is a frameshift
mutation resulting from an insertion
of a cytosine at nucleotide position 3020 in
Nod2 (3020insC). The nucleotide insertion
generates a premature stop codon, resulting
in a truncated protein lacking part of the last
Dana J. Philpott is in the Department of
Immunology and Stephen e. Girardin is in
the Department of Laboratory Medicine and
Pathobiology at the University of Toronto, Toronto,
Ontario, Canada.
e-mail: dana.philpott@utoronto.ca
LRR near the C terminus of the protein. The
relative risk for developing Crohn’s disease
in individuals bearing a homozygous frameshift
mutation in Nod2 has been calculated
to be approximately 30 times that of normal
individuals. Moreover, these individuals
show a more severe disease phenotype, with
increased risk of ileal stenosis and the need
for surgical intervention 4 .
Since the discovery of the link between
Nod2 and Crohn’s disease, many studies
have focused on trying to understand how
mutations in Nod2 influence Crohn’s disease
development. Nod2 is an Apaf-like
molecule with a distinct domain organization
comprising two N-terminal caspase-activating
and recruitment domains
(CARDs), a central nucleotide binding site
(NBS) and a C-terminal series of LRRs 5 .
The domain architecture of Nod2 is similar
to plant R (for resistance) proteins, which
mediate defense against invading pathogens.
Indeed, Nod2, a member of a family
of cytosolic innate immune molecules called
NBS-LRR receptors (NLRs), is also involved
in host defense. Comprising 22 members
so far, the NLR family includes proteins
that are involved in sensing both microbe-
associated and danger-associated molecular
patterns within cells and commencing innate
immune responses 6 .
Nod2 responds to a motif called muramyl
dipeptide (MDP), which is found in the cell
wall of Gram-positive and Gram-negative
bacteria 7,8 and is a component of peptidoglycan.
It is not yet known how Nod2 senses
MDP; indeed, no study so far has been able
to show whether MDP binds directly to
Nod2. On the basis of studies in the Tolllike
receptor field and plant R proteins, it has
been postulated that the LRR region is likely
to represent the site at which ligand sens-
nEwS AnD ViEwS
ing occurs, either directly or through a coreceptor.
MDP sensing is thought to unfold
Nod2, leading to oligomerization through
the NBS domain. This process exposes the
CARDs, which then provide a platform for
the recruitment of Rip2, a CARD-containing
signaling kinase that triggers the NF-κB
pathway 5 . Rip2 also activates the p38 and Erk
MAP kinases upon MDP stimulation 9 . These
inflammatory pathways lead to the transcription
and expression of several genes, many of
which are proinflammatory or provide some
defensive function.
In relation to Crohn’s disease, the 3020insC
mutation in Nod2 results in a protein product
that, when expressed in vitro, no longer
reacts to MDP to drive NF-κB signaling 7,8 .
These findings are supported by studies
showing that when cells are isolated from the
blood of humans with Crohn’s disease who
are homozygous for the 3020insC mutation
and treated in vitro with MDP, they do not
produce certain cytokines in response to this
bacterial product 10 . A key point, however, is
that no study so far has proven that a mutated
Nod2 protein is actually present in tissues
from subjects with Crohn’s disease. It is possible
that this mutation results in an unstable
protein product. If this is the case, then the
Nod2-deficient mouse is an acceptable model
for the disease. However, if the protein is
present, it might have ligand-independent
functions. Indeed, the paper by Noguchi et
al. supports the latter possibility 11 .
Over the years, much research has gone
into trying to unravel the apparent paradox:
the 3020insC mutation in Nod2 results in a
protein product incapable of responding to
a bacterial ligand, yet Crohn’s disease results
in overt inflammation that probably is triggered
by bacterial products. One explanation
that might help resolve this paradox
nature immunology volume 10 number 5 may 2009 455

© 2009 Nature America, Inc. All rights reserved.
nEwS AnD ViEwS
MDP stimulated
MDP
IκB-α
Nod2
Rip2
IKK
Nod2
NF-κB
Proinflammatory
gene transcription
P
Healthy human:
normal Nod2
Basal Basal MDP stimulated
P
p38
Nod2
P
Nod2
hnRNP-A1 hnRNP-A1
P
p38
P P
IL10
transcription
relies on the observation that cells or tissues
isolated from Crohn’s disease patients show
depressed IL-10 production 12 . IL-10 is produced
by various hematopoietic cells, including
macrophages, dendritic cells and T cells,
and acts to contain and suppress inflammatory
responses and thereby to downmodulate
adaptive immune responses and minimize
tissue damage in response to microbes. It
seems that enhancing IL-10 production,
which can occur after treatment with certain
probiotic bacteria, helps to calm inflammation
in Crohn’s disease 12 . Moreover, reduced
intestinal colonization with an IL-10promoting
member of the commensal flora,
Fecalibacterium prausnitzii, is associated with
enhanced risk of recurrence of Crohn’s disease
after surgical resection 13 . Noguchi et al.
shed light on why IL-10 expression is affected
in Crohn’s disease.
Noguchi et al. begin with the observation
that bacterially induced IL-10 production
is dampened in primary human monocytes
from subjects with Crohn’s disease having
the 3020insC Nod2 mutation. Although
?
?
Mouse:
normal Nod2
P
MDP
3020insC
Nod2
No
signaling
P
p38
Crohn’s disease:
mutant Nod2
Basal Basal
hnRNP-A1
No basal Il10
transcription
IL10
transcription
actual suppression of IL-10 production is
not clear from their studies with these cells,
synergy between Toll-like receptors (TLRs)
and Nod2 is clearly lost. Consistent with this
finding, several previous studies have shown
a synergistic production of cytokines upon
coactivation of TLR2 and Nod2, which is lost
in the 3020insC mutants. Transduction of
human monocytes with a cDNA encoding the
3020insC mutant Nod2 protein blocks basal
transcription and expression of IL-10, but
leaves other cytokines, such as IL-1β, unaffected.
Expression of the 3020insC form of
Nod2 blocks IL10 promoter activity as measured
from a luciferase reporter. Notably, two
other Crohn’s disease–associated Nod2 variants,
R702W and G908R, similarly inhibit
IL10 promoter activity. The 3020insC Nod2
mutant has no affect on the activity of mouse
Il10 reporter constructs and, similarly, the
equivalent mutation made in the mouse
Nod2 gene, referred to as 2939insC, does
not affect human IL10 transcription. Taken
together, these findings demonstrate that the
human 3020insC mutant of Nod2 acts spe-
cifically to block the human IL10 promoter
(Fig. 1). Moreover, it may explain why the
2939insC mouse mutation, when knocked
into the mouse Nod2 locus, does not behave
like the human 3020insC mutant with regard
to proinflammatory cytokine production 14 .
Noguchi et al. show that heterogeneous
nuclear ribonucleoprotein A1 (hnRNP A1)
binds constitutively to the IL10 promoter.
When 3020insC is expressed in cells, the
binding of hnRNP A1 to IL10 is reduced.
Exogenously expressed hnRNP A1 stimulates
IL10 promoter activity, and knocking down
hnRNP A1 expression in primary human
monocytes reduces lipopolysaccharide-
stimulated production of IL-10. Convincingly,
through chromatin immunoprecipitation
analysis, the authors show reduced binding
of hnRNP A1 to the IL10 promoter region in
humans with the 3020insC Nod2 genotype,
who are highly deficient in IL-10 production.
But how does the presence of the 3020insC
mutation in Nod2 result in reduced transcription
factor binding at the IL10 promoter?
Noguchi et al. found that wild-type
Nod2 forms a trimeric complex in the cytoplasm
with hnRNP A1 as well as active,
phosphorylated p38, and it appears that p38
is able to serine-phosphorylate hnRNP A1.
This phosphorylation event is required for
cleavage of hnRNP A1 into its active form,
which can then translocate to the nucleus to
act on the IL10 promoter (Fig. 1). However,
the 3020insC Nod2 mutant protein interacts
with p38 but not with p38hnRNP A1. The
lack of this interaction reduces the presence
of phosphorylated hnRNP A1 in the nucleus,
thereby affecting basal transcription at the
IL10 promoter. Interestingly, the mouse
2939insC Nod2 mutant protein resembles
wild-type mouse Nod2 in its influence on
the interaction between phospho-p38 and
hnRNP A1; this observation provides mechanistic
insight into why the mouse mutant
does not affect IL10 transcription.
Taken together, the findings of Noguchi et
al. stress an important point with regard to
the 3020insC mutation in Nod2 associated
with Crohn’s disease: this mutation might be
a loss of function or gain of function, depending
on the response examined. Indeed, it certainly
seems that the mutant protein is unable
to sense MDP. One might argue that the phenotype
caused by the 3020insC mutation in
Nod2 described by Noguchi et al. also represents
a loss of function. Indeed, the 3020insC
mutant Nod2 protein is unable to form a trimolecular
complex that leads to the activation
of hnRNP A1 and thereby affects the ability
of hnRNP A1 to bind to the IL10 promoter.
However, the authors’ point is well taken that
456 volume 10 number 5 may 2009 nature immunology
p38
2939insC
Nod2
P
?
?
Mouse:
mutant Nod2
P
hnRNP-A1
Figure 1 Reduced production of iL-10 in cells expressing Crohn’s disease–associated mutations in
Nod2. in cells from humans and mice expressing wild-type Nod2, muramyl dipeptide (MDP) triggers
nod2 unfolding, allowing its interaction with Rip2 and the subsequent activation of the nF-κB pathway,
which drives proinflammatory cytokine gene transcription. noguchi et al. show that wild-type nod2
forms a trimolecular complex with active, phosphorylated p38 and the transcription factor hnRnP A1.
This association drives IL10 transcription during the steady-state and after bacterial stimulation (left
panel). The Crohn’s disease–associated 3020insC mutant nod2 (right panel) fails to detect MDP and
activate the nF-κB pathway, and does not efficiently interact with hnRnP A1; as a result, hnRnP A1
activation and subsequent transcriptional control of the IL10 promoter is impaired. Reduced iL-10
production is therefore a consequence of this mutation and may contribute to the hyperinflammatory
response in the intestine that is characteristic of Crohn’s disease. The equivalent mutation translated
to mouse nod2 results in a protein that retains its ability to activate Il10 transcription. These findings
provide a possible explanation of why cells isolated from a knock-in mouse with this equivalent
mutation do not behave like their human counterparts.
P

© 2009 Nature America, Inc. All rights reserved.
if the mutant protein is indeed expressed, it
might have acquired a new, proinflammatory
“gain of function” that is distinct from
the activity of the wild-type version of Nod2.
Finally, these studies bring ideas for possible
treatments for Crohn’s disease: targeting
hnRNP A1 directly to influence its ability to
drive IL10 transcription might bypass the
block erected by mutant Nod2.
1. Xavier, R.J. & Podolsky, D.K. Nature 448, 427–434
(2007).
2. Hugot, J.P. et al. Nature 411, 599–603 (2001).
3. Ogura, Y. et al. Nature 411, 603–606 (2001).
4. Seiderer, J. et al. Inflamm. Bowel Dis. 12, 1114–1121
(2006).
5. Ogura, Y. et al. J. Biol. Chem. 276, 4812–4818
(2001).
6. Franchi, L., warner, n., Viani, K. & nunez, G. Immunol.
Rev. 227, 106–128 (2009).
7. Girardin, S.E. et al. J. Biol. Chem. 278, 8869–8872
(2003).
8. inohara, n. et al. J. Biol. Chem. 278, 5509–5512
(2003).
9. Park, J.H. et al. J. Immunol. 178, 2380–2386 (2007).
10. netea, M.G. et al. J. Biol. Chem. 280, 35859–35867
(2005).
11. noguchi, E. et al. Nat. Immunol. 10, 471–479
(2009).
12. Baumgart, D.C. & Sandborn, w.J. Lancet 369, 1641–
1657 (2007).
13. Sokol, H. et al. Proc. Natl. Acad. Sci. USA 105, 16731–
16736 (2008).
14. Maeda, S. et al. Science 307, 734–738 (2005).
The Foxo and the hound: chasing the in vivo regulation
of T cell populations during infection
Elia D Tait & Christopher A Hunter
T cell expansion and contraction during the immune response to pathogens are regulated by a wide variety of cellintrinsic
and cell-extrinsic factors. A new study identifies a role for CTLA-4 signaling and activation of the Foxo3
transcription factor in modulating T cell populations.
In this issue of Nature Immunology, Dejean
et al. provide new insight into the role of the
forkhead box (Fox) transcription factor Foxo3
in limiting the magnitude of T cell responses1 .
The Fox family of transcription factors participate
in a multitude of biological processes,
including the cell cycle, metabolism and the
stress response. The activity of proteins in the
Foxo subfamily is predominantly regulated by
phosphorylation status and subcellular localization,
but these transcription factors are also
subject to a number of post-translational modifications,
including acetylation and ubiquitination.
Growth factors mobilize numerous kinases,
including Akt, that phosphorylate Foxo proteins,
leading to their export from the nucleus and
inactivation. In contrast, stress stimuli mobilize
a different set of kinases, notably Jun N-terminal
kinase (Jnk), that promote nuclear import of
Foxo and the initiation of Foxo-dependent
gene regulation programs2 . These transcription
factors can modulate gene expression by binding
promoters at Foxo-binding sites to recruit
transcriptional machinery, by competing for
promoter binding sites with other factors, and
by dynamically regulating interactions between
cofactors and various elements of the transcriptional
machinery3 .
In the past decade, immunologists have come
to appreciate the importance of the Fox family
in coordinating immune responses, particularly
elia D. Tait and Christopher A. Hunter are in the
Department of Pathobiology, School of veterinary
Medicine, University of Pennsylvania, Philadelphia,
Pennsylvania, USA.
e-mail: chunter@vet.upenn.edu
in the context of Foxp3 expression by T regulatory
(T reg ) cells. The Foxo subfamily members
(in mammals, Foxo1, Foxo3, Foxo4 and Foxo6)
have also been shown to influence T cell function.
T cell receptor (TCR) and co-receptor
ligation modulate the activity of Foxo proteins
that upregulate prosurvival programs in the
presence of growth factors and initiate apoptotic
pathways in the absence of mitogen or
cytokine 2 . In 2004, Pandiyan et al. demonstrated
that signaling through the inhibitory receptor
cytotoxic T lymphocyte–associated antigen-4
(CTLA-4) protected T helper type 2 (T H 2)
cells from apoptosis via Foxo3 inactivation and
subsequent upregulation of the survival protein
Bcl-2 (ref. 4). Consistent with a proapoptotic,
regulatory role for Foxo3, Lin et al. reported
that Foxo3a antagonizes signaling by the transcription
factor NF-κB and that mice carrying a
mutated Foxo3a allele underwent spontaneous
T cell hyperproliferation and inflammation 5 . In
2007, it was shown that Foxo3a participates in
the maintenance of memory CD4 + T cells by
coordinating signals from cytokine receptors
and the TCR 6 . Thus, Foxo proteins appear to
have an intrinsic role in T cell homeostasis and
in limiting T cell activity.
Dejean et al. examined how Foxo3 influences
T cell population expansion and contraction
during infection by lymphocytic choriomeningitis
virus (LCMV). In this model, the absence
of Foxo3 led to an exaggerated antigen-specific
T cell accumulation, with largely normal contraction
1 . Whereas a previous study had suggested
that Foxo3 was “largely dispensable” 5 in
modulating T cell apoptosis and was involved
exclusively in spontaneous T cell population
nEwS AnD ViEwS
expansion, Dejean et al. showed that antigenspecific
T cells in Foxo3-deficient mice had
enhanced expression of the prosurvival factor
Bcl-2 and decreased binding to the apoptosis
marker annexin-5. Notably, the expansion
of T cell populations observed in the Foxo3deficient
mice was not T cell intrinsic and was
instead dependent on increased production
of the cytokine interleukin-6 (IL-6) by Foxo3deficient
dendritic cells (DC) 1 . IL-6 has long
been regarded as proinflammatory, and it also
has a strong prosurvival effect on activated T
cells 7 . Blockade of IL-6R α-chain restored the
magnitude of the LCMV-driven T cell response
in the Foxo3-deficient environment to a level
comparable to that observed in wild-type mice.
Finally, Dejean et al. showed that CTLA-4 ligation
induced nuclear localization and activation
of Foxo3 and inhibited the production of
IL-6 in DCs in vitro 1 . These findings provide a
previously unsuspected mechanism by which
CTLA-4–expressing T cells can limit their own
survival; by initiating Foxo3 activation and
nuclear import in DCs, T cells modulate DC
cytokine production and thereby influence their
own viability (Fig. 1).
These findings illustrate the importance
of the interactions that take place between
DCs and T cells in regulating the expansion
and contraction of T cell populations during
infection. In addition, this work identifies a
new role for Foxo3 in modulating this process.
Understanding how T cell responses are
regulated and maintained during infection
has obvious implications for vaccine development.
To this end, multiple DC–T cell interactions
that result in the activation of these two
nature immunology volume 10 number 5 may 2009 457

© 2009 Nature America, Inc. All rights reserved.
nEwS AnD ViEwS
CTLA-4 signals:
↓ IL-2 production
↓ Activation
↓ Cell cycle
Lack of IL-6 signals:
↓ Bcl-2
↓ Viability
IL-6R
CTLA-4
TCR
CD4
IL-6
populations have been characterized. The
CD40-CD40L interaction is one example of
a mechanism through which T cells influence
their own activation and accumulation
by optimizing DC responses. In contrast,
the interactions between T cells and DCs
that govern the contraction and control of
T cell numbers are less well defined 8 . DCs
may imprint T cells during activation with
apoptotic programs, and T cell population
expansion is certainly limited by T cell–
mediated DC killing 9 . Also, interactions
of CTLA-4 and programmed death-1 with
their respective ligands intriusically modulate
T cell responses and are important in
controlling autoimmunity 8 . Specifically,
mice lacking CTLA-4 develop spontaneous
lymphoproliferative disease (similar to
that of the Foxo3-deficient mice described
previously 5 ), indicating that this molecule
is vital in maintaining T cell homeostasis 10 .
Overall, however, the signals that orchestrate
the fine balance between too much and too
little T cell population expansion during an
immune response are still unclear.
The findings of Dejean et al. suggest a scenario
in which activated T cells that express
CTLA-4 (or T reg cells that constitutively
express CTLA-4; ref. 11) can induce nuclear
localization of Foxo3, and this active Foxo3
then limits the production of IL-6 by DCs
B7
MHC II
Jnk
Foxo3
P
and thus controls T cell population expansion
(Fig. 1) 1 . Certain aspects of this model
need to be tested further, however. Some
previous studies have shown that antibodymediated
blockade of CTLA-4 can lead to
enhanced T cell responses during infection 12
and cancer 13 , and the model outlined in
Figure 1 suggests that antagonizing CTLA-4
in vivo during LCMV infection should
lead to an increase in the expansion of T
cell populations 1 . However, in agreement
with other studies 14 , Dejean et al. found
that this was not the case: LCMV-driven T
cell responses were unaffected by CTLA-4
blockade in vivo 1 . These negative data serve
as a reminder that blocking CTLA-4 has
different effects during different types of
immune responses 1,4,8,12–14 . Thus, the effect
of CTLA-4 signaling on the expansion and
contraction of the T cell population overall
might vary depending on context. The
apparent contradiction between the model
described in Figure 1 and the negative in vivo
data of Dejean et al. 1 also indicates that there
may be ligands other than CTLA-4 that lead
to Foxo3 activation in vivo. The model proposed
by Dejean et al. prompts other questions
regarding the biology and function of
Foxo3 in response to CTLA-4 signaling. The
lack of Foxo3-binding sites in the Il6 promoter
1 implies that Foxo3 may regulate Il6
transcription through indirect means or by
upregulating a broader anti-inflammatory
gene expression program that includes
downregulation of IL-6 production. Overall,
the principles established by Dejean et al. are
important in advancing our understanding
of the role of DC–T cell interactions in controlling
T cell populations, and they highlight
some of the contradictions and future
directions in this field.
In total, the current literature suggests that
CTLA-4 and Foxo3 have important roles in
the control of spontaneous and antigeninduced
T cell population expansion. In this
issue, Dejean et al. make an important contribution
by providing evidence linking these
pathways 1 . In the context of treating human
disease, particularly autoimmune disease,
controlling the immune response using
therapies that mimic natural cell-intrinsic
immunoregulation is an appealing strategy.
CTLA-4–Ig, a fusion between CTLA-4
and IgG that mimics CTLA-4, was initially
thought to act as an antagonist of B7-CD28–
mediated signaling that would downregulate
inappropriate T cell population expansion in
the context of autoimmunity; we now appreciate
that it has additional effects, including
inducing the immunosuppressive indoleamine-2,3-dioxygenase
pathway 8 . This fusion
protein is now used for the management of
rheumatoid arthritis, but the true biological
mechanism behind its clinical success is
not entirely clear. Whether Foxo3 signaling
is also influenced during treatment with
CTLA-4–Ig remains to be tested. Although
CTLA-4 signaling and the in vivo biology
of Foxo3 are still not fully understood, the
work of Dejean et al. 1 prompts exciting new
questions about the basic biology and clinical
relevance of these factors.
1. Dejean, A.S. et al. Nat. Immunol. 10, 504–513
(2009).
2. Peng, S.L. Oncogene 27, 2337–2344 (2008).
3. Fu, Z. & Tindall, D.J. Oncogene 27, 2312–2319
(2008).
4. Pandiyan, P. et al. J. Exp. Med. 199, 831–842
(2004).
5. Lin, L., Hron, J.D. & Peng, S.L. Immunity 21, 203–
213 (2004).
6. Riou, C. et al. J. Exp. Med. 204, 79–91 (2007).
7. Teague, T.K. et al. J. Exp. Med. 191, 915–926
(2000).
8. Fife, B.T. & Bluestone, J.A. Immunol. Rev. 224,
166–182 (2008).
9. Masson, F., Mount, A.M., wilson, n.S. & Belz, G.T.
Immunol. Cell Biol. 86, 333–342 (2008).
10. waterhouse, P. et al. Science 270, 985–988 (1995).
11. wing, K. et al. Science 322, 271–275 (2008).
12. Kaufmann, D.E. et al. Nat. Immunol. 8, 1246–1254
(2007).
13. Fong, L. & Small, E.J. J. Clin. Oncol. 26, 5275–5283
(2008).
14. Bachmann, M.F. et al. J. Immunol. 160, 95–100
(1998).
458 volume 10 number 5 may 2009 nature immunology
Foxo3
IL-6 IL-6
Figure 1 T cells can limit their own accumulation by multiple mechanisms mediated by CTLA-4
signaling and the Foxo3 transcription factor. During antigen presentation, B7 molecules on DCs engage
CTLA-4 expressed on activated T cells or T reg cells. The resulting signals promote direct control of T cell
activation and expansion through downregulation of iL-2 production, decreased T cell activation and
decreased cell cycle progression. Also, the B7–CTLA-4 interaction initiates signaling in the DCs that
mobilizes nuclear import and activation of Foxo3, which indirectly controls T cell population expansion
by limiting DC production of the prosurvival cytokine iL-6. T cells with limited access to iL-6 are more
susceptible to apoptosis and have lower expression of Bcl-2. MHC ii, major histocompatibility class ii.
P
Kim Caesar

© 2009 Nature America, Inc. All rights reserved.
Dimers dampen T cell activity
The transcription factor NFAT influences T cell function by
acting together with other transcription factors, including
AP-1, T-bet GATA-3 and Foxp3. In the Journal of Experimental
Medicine, Macian and colleagues show that NFAT homodimers
are needed for the transcription of some genes associated with
T cell anergy. In T cells stimulated with ionomycin, mutant NFAT
proteins that bind to AP-1 but not to other NFAT proteins induce
activation of some (Tle4 and Dgka) but not other (Rnf128 and
Casp3) anergy-associated genes. Tandem κB-like binding sites
in the Rnf128 promoter—to which NFAT homodimers bind—are
needed for ionomycin-induced transcription of this gene.
T cells expressing the NFAT mutant that does not form dimers
produce more interleukin 2 and have a lower anergy index score
than do T cells transduced with a constitutively active form of
NFAT. Whether NFAT proteins form dimers with other proteins
to facilitate expression of anergy-related genes such as Tle4 and
Dgka remains to be determined. CB
J. Exp. Med. (23 March 2009) doi:10.1084/jem20082731
Facilitating cross-presentation
Although macrophages, neutrophils and dendritic cells (DCs) are known
to have phagosomal pathways that differ in terms of oxidation, pH and
degradation, it is not known whether phagocytic organization differs
among DC subsets. In Immunity, Amigorena and colleagues compare the
phagocytic pathways of splenic CD8 + and CD8 – DCs. The former subset
is known to cross-present antigens, and cross-presentation requires a high
pH in the phagosome. Consistent with that, CD8 + DCs restrict assembly
of the NADPH oxidase complex to the phagosome, which results in the
production of reactive oxygen species and local alkalization. CD8 – DCs,
however, are unable to assemble this complex on the phagosomal membrane.
The ability of CD8 + DCs to assemble the NADPH oxidase complex
depends on the GTPas Rac2, and deficiency of Rac2 results in a lower
phagosomal pH and cross-presentation efficacy. These data show that DC
subpopulations have differences in their endophagocytic pathways that
can explain their different abilities to cross-present antigen. JDKW
Immunity (26 March 2009) doi:10.1016/j.immuni.2009.01.013
Systemic suppression
Retinoic acid (RA) induces regulatory T cells in the intestine,
but whether it is involved in the systemic immune response
is unclear. In Nature Medicine, Pulendran and colleagues
now find that zymosan induces, through TLR2, expression
of RA–metabolizing enzyme in splenic DCs. This Erk kinase–
dependent pathway produces RA that functions in an autocrine
way to upregulate SOCS3, a negative regulator of the kinase
p38 and proinflammatory cytokines. Consistent with that,
RA, in conjunction with IL-10, acts synergistically to induce
regulatory T cells and suppress T helper type 1 (T H 1) and T H -17
autoimmune responses in vivo. In the absence of TLR2, dectin-1,
a C-type lectin receptor that also recognizes zymosan, induces a
proinflammatory cytokine response in splenic DCs, which promotes
T H1 and T H-17 responses. Identification of the physiological
conditions in which zymosan triggers dectin-1 but not TLR2 in
splenic DCs is needed. JDKW
Nat. Med. (1 March 2009) doi:10.1038/nm.1925
Written by Christine Borowski, Laurie A. Dempsey & Jamie D.K. Wilson
Distinguishing danger
How the body distinguishes sterile inflammation from pathogen-induced
inflammatory responses has been unclear. In Science, Chen et al. show
that the recognition of endogenous ‘danger-associated molecular pattern’
(DAMP) proteins by CD24 and the signaling molecule Siglec-10 inhibits
activation of the transcription factor NF-κB, thereby braking inflammation
due to non–microbe-induced tissue damage. CD24, a glycosylphosphoinositol-anchored
protein, in association with Siglec-10, binds DAMP
proteins such as high-mobility group box 1 and heat-shock proteins 70
and 90. Mice lacking either CD24 or Siglec-10 succumb to sublethal doses
of acetaminophen, which is toxic to liver tissue and induces hepatocyte
necrosis. Siglec-10 interacts with the SHP-1 tyrosine phosphatase to block
activation and nuclear translocation of the NF-κB family member p65.
Despite having more NF-κB activation in response to DAMP proteins,
neither CD24- or Siglec-10-deficient dendritic cells show enhanced NF-κB
activation in response to pathogen-associated molecular pattern motifs
such as lipopolysaccharide. Thus, CD24–Siglec-10 regulation acts to limit
tissue damage in response to non-pathogen threats. LAD
Science 323, 1722–1725 (2009)
Regulating stress
ReseARch highlights
Cellular stress triggered by an abundance of unfolded proteins
in the endoplasmic reticulum activates the stress sensor
IREα1. IREα1 contains an endoribonuclease whose splicing
activity produces functional mRNA encoding the transcription
factor XBP-1 (called ‘XBP-1s’), which, among other functions,
is required for immunoglobulin secretion in plasma cells.
In Molecular Cell, Glimcher and colleagues identify Bax
inhibitor 1 (BI-1) as a negative regulator of this stress-induced
activation of IREα1 and XBP-1s. Like IREα1, BI-1 localizes
to the endoplasmic reticulum membrane and physiologically
associates with IREα1 through its conserved carboxy-terminal
cytoplasmic tail and inhibits the endoribonuclease activity of
IREα1. After stress induction, cells that lack BI-1 have higher
expression of XBP-1s and proteins encoded by its target genes.
Lipopolysaccharide stimulation yields much higher titers of
IgM in BI-1-deficient B cells than in wild-type B cells. How
endoplasmic reticulum stress alters the BI-1–IREα1 interaction
remains unknown. LAD
Mol. Cell 33, 679–691 (2009)
Degrading inflammatory mRNA
If left unrestrained, TLR-driven immune responses induce unwanted tissue
damage. In Nature, Akira and colleagues identify Zc3h12a, a CCCHtype
zinc-finger protein that increases in abundance after TLR signaling, as
an essential feedback inhibitor of TLR-induced inflammatory responses.
Zc3h12a –/– mice show impaired survival and rampant inflammation
characterized by excessive accumulation of plasma cells and activated T
cells in several organs, granuloma formation in lymph nodes, and serum
hyperimmunoglobulinemia. Macrophages from Zc3h12a –/– mice contain
more Il6 and Il12p40 mRNA but similar amounts of Tnf and Cxcl1
mRNA after TLR stimulation. Zc3h12a degrades—apparently through
an endonuclease activity—Il6 and Il12p40 mRNA by a mechanism that
depends on the 3′ untranslated regions of these molecules. Additional
work is needed to determine if other ‘proinflammatory’ mRNA transcripts
are degraded by Zc3h12a and contribute to the severely detrimental phenotype
of Zc3h12a –/– mice. CB
Nature (25 March 2009) doi: 10.1038/nature07924
nature immunology volume 10 number 5 may 2009 459

© 2009 Nature America, Inc. All rights reserved.
Autophagy genes in immunity
Herbert W Virgin 1 & Beth Levine 2
In its classical form, autophagy is a pathway by which cytoplasmic constituents, including intracellular pathogens, are
sequestered in a double-membrane–bound autophagosome and delivered to the lysosome for degradation. This pathway has been
linked to diverse aspects of innate and adaptive immunity, including pathogen resistance, production of type I interferon, antigen
presentation, tolerance and lymphocyte development, as well as the negative regulation of cytokine signaling and inflammation.
Most of these links have emerged from studies in which genes encoding molecules involved in autophagy are inactivated in
immune effector cells. However, it is not yet known whether all of the critical functions of such genes in immunity represent
‘classical autophagy’ or possible as-yet-undefined autophagolysosome-independent functions of these genes. This review
summarizes phenotypes that result from the inactivation of autophagy genes in the immune system and discusses the pleiotropic
functions of autophagy genes in immunity.
Classical macroautophagy (called ‘autophagy’ here) involves the sequestration
of cytoplasmic contents in a characteristic double-membraned
vacuole, the autophagosome. Fusion of the outer autophagosomal
membrane with the lysosome and the subsequent breakdown of the
inner membrane results in the exposure of the sequestered cytoplasmic
material to lysosomal hydrolases 1,2 (Fig. 1). A related term, ‘xenophagy’,
refers to the use of the autophagy pathway to digest foreign rather than
self constituents 3 . The autophagy pathway has two main physiological
functions: it rids the cell of unwanted constituents, and it recycles cytoplasmic
material so that cells can maintain macromolecular synthesis
and energy homeostasis during stressful conditions. These functions
probably underlie the well described roles of autophagy in cell survival
and in the prevention of neurodegeneration, cancer and aging and, at
least in part, the protective roles of autophagy in pathogen-infected
cells 1,2 .
The autophagy pathway involves the concerted action of evolutionarily
conserved gene products involved in the initiation of autophagy,
elongation and closure of the autophagosome, and lysosomal fusion
(Fig. 1). Beclin 1, in complex with the class III phosphatidylinositol-
3-OH kinase (PI(3)K) Vps34 and other proteins, is important for
the nucleation of autophagosomes. This macromolecular complex is
critical for integrating multiple signals that positively and negatively
regulate autophagy. The autophagy (‘Atg’) protein Atg7 is required for
two ubiquitin-like conjugation pathways that in concert result in the
covalent conjugation of Atg5 to Atg12 and the conversion of LC3 (the
mammalian ortholog of yeast Atg8) to its phosphatidylethanolamine-
1 Department of Pathology and Immunology, Department of Molecular
Microbiology and Department of Medicine, Washington University School
of Medicine, St. Louis, Missouri, USA. 2Howard Hughes Medical Institute,
Department of Internal Medicine and Department of Microbiology, University
of Texas Southwestern Medical Center, Dallas, Texas, USA. Correspondence
should be addressed to H.W.V. (virgin@wustl.edu) or B.L. (beth.levine@
utsouthwestern.edu).
Published online 20 April 2009; doi:10.1038/ni.1726
reVIeW
conjugated LC3-II form. The Atg5-Atg12 conjugate forms part of a large
complex with Atg16L1 that is responsible for the membrane localization
of the autophagic machinery and the generation of the autophagosome
4 . At present, there are more than 20 recognized Atg proteins 1 , and
it is likely that more will be identified. For example, FIP200, a protein
reported to interact with the kinases FAK, Pyk2 and ASK1, the tumor
suppressor TSC1, the tumor suppressor p53 and tumor necrosis factor
receptor–associated factor 2, has been found to be critical for the initiation
of autophagy through interaction with the kinases ULK1 and ULK2
(mammalian orthologs of yeast Atg1) 5 . Furthermore, four independent
groups have identified a candidate functional mammalian ortholog of
yeast Atg14, known as either human Atg14 or Barkor, that interacts with
Beclin 1 and the Vps34 class III PI(3)K complex to initiate autophagic
vesicle nucleation 6–9 . Of the many proteins involved in autophagy, Beclin
1, Atg5, Atg7, LC3, Atg12 and Atg16L1 have been studied in one or more
aspects of immunity.
The identification of this core molecular machinery has revolutionized
the ability to detect and genetically manipulate the autophagy pathway.
This has led to considerable advances in understanding the function
of genes encoding molecules involved in autophagy (called ‘autophagy
genes’ here) in immunity and other biological processes 1,2 . In terms of
immunity, studies have shown that autophagy is regulated by pathways
critical for the function and differentiation of cells of the immune system,
including Toll-like receptors (TLRs), p47 GTPases, eIF2α kinases,
and cytokines such as interferon-γ (IFN-γ) 10–13 . Furthermore, autophagy
genes are important for thymic selection, lymphocyte development and
survival, antigen presentation, killing of intracellular pathogens, and tissue
homeostasis. There is also increasing evidence that autophagy genes
may regulate innate immune signaling in a cell type–specific way.
Although some of the functions described above, such as the targeting
of microbes for lysosomal degradation (xenophagy), may unequivocally
reflect the involvement of classical autophagy in immunity (Fig.
2), it is not yet clear whether all functions of autophagy genes in immunity
relate to the autophagy pathway itself (Fig. 3). This uncertainty
arises from both the assays commonly used to detect autophagy and
the lack of universal criteria for interpreting reverse-genetic studies.
nature immunology volume 10 number 5 may 2009 461

© 2009 Nature America, Inc. All rights reserved.
REVIEw
For example, the most commonly used
autophagy assays, biochemical detection of
the lipidated form of LC3 (LC3-II) and detection
of the localization of LC3-II to punctate
dots by light microscopy, are subject to many
interpretations. Most investigators now recognize
that detection of increased LC3-II can
indicate either more autophagosome formation
or a block in autophagosomal maturation,
which necessitates the use of ancillary
approaches to distinguish between these
possibilities 14 . Perhaps less well appreciated
is the idea that LC3 dots may represent the
targeting of LC3 to structures other than
autophagosomes, such as phagosomes, double-membraned
scaffolds for the replication
complexes of positive-strand RNA viruses, or
even protein aggregates 15–17 . Similarly, LC3
may become lipidated, forming LC3-II, in
the absence of autophagosome formation 18,19 .
Thus, whereas LC3-II formation and localization
of LC3 punctae are hallmark features of
autophagosome formation (and sensitive
parameters for the detection of autophagy),
they lack complete specificity as markers of
classical autophagy. The direct demonstration
of a function for autophagosomes in a
process is the gold standard for proving that
autophagy is involved. An extensive discussion
of the various assays used in autophagy
research, as well as the criteria for their use and
interpretation, has been provided in a consensus
paper published by many of the authorities
in the field 14 .
The genetic knockout or knockdown of
core autophagy genes is an effective way to
turn off the autophagy pathway, but it is often
Autophagy induction
Starvation
Growth factor deprivation
Immune signals:
IFN-γ
TNF
TLRs
PKR-elF2α kinase
Jnk
FADD
Immunity-related GTPases
Autophagy suppression
Nutrient abundance
Insulin-Akt-TOR signaling
Immune signals:
IL-4
IL-13
FADD
Immunity-related GTPases
unclear whether resulting phenotypes are due to a deficiency of classical
autophagy or autophagy-independent functions of the autophagy genes.
As reviewed elsewhere, autophagy genes are known to have alternative
functions, including those in other membrane-trafficking events, axonal
elongation and cell death 20,21 . Furthermore, investigators often assume
that phenotypes noted in autophagy gene–deficient cells or organisms
are due to lack of classical autophagy without providing direct experimental
proof. Notably, for many studies of the involvement of autophagy
genes in immunity (Fig. 3), it is not clear how classical autophagy, involving
the autophagolysosomal degradation of sequestered contents, could
contribute to the described functions of autophagy genes. Thus, it is
possible that autophagy genes act in other cellular processes that affect
immune effector cell development and function. Here we will discuss
advances related to the function of autophagy genes in the immune
system.
Autophagy genes in antimicrobial defense
More than a decade ago, the first published paper on Beclin 1, a mammalian
autophagy protein, demonstrated an antiviral function for
exogenous neuronal expression of Beclin 1 in mice with alphavirus
encephalitis 22 . Subsequently, endogenous autophagy genes were shown
to be critical for the successful innate immune response to fungal, bacterial
and viral pathogens in plants 23 , and viral evasion of Beclin 1 function
by a herpes simplex virus neurovirulence factor was found to be essential
Vesicle
nucleation
Isolation
membrane
Vesicle
elongation
Docking &
fusion
Autophagosome
AUTOPHAGY
Atg16L1 Atg12 Atg12
Atg10 E2 Atg12
Atg5 Atg5
Atg5
Atg5 Atg16L1 Atg16L1
Atg12
Atg16L1 Atg16L1
Atg7 E1
Atg5
Atg12
Atg5
Atg12
PE
LC3 -Gly LC3
Atg3 E2
Atg4 LC3
Lysosome
Class III PI(3)K complex
Other
regulators
Beclin 1 Atg14
Bcl-2–Bcl-XL Vps34 Vps15
Ubiquitin-like conjugation systems
Vesicle breakdown
& degradation
Autolysosome
Figure 1 The autophagy pathway and its regulation. The autophagy pathway proceeds through a series
of stages, including nucleation of the autophagic vesicle, elongation and closure of the autophagosome
membrane to envelop cytoplasmic constituents, docking of the autophagosome with the lysosome, and
degradation of the cytoplasmic material inside the autophagosome. Vesicle nucleation depends on a
class III PI(3)K complex that contains various proteins (in light yellow box at right), as well as additional
proteins that regulate the activity of this complex (such as rubicon, UVRAG, Ambra-1 and Bif-1). Vesicle
elongation and completion involves the activity of two ubiquitin-like conjugation systems (light blue
box at right). The autophagy pathway is positively regulated (green box at left) and negatively regulated
(red box at left) by diverse environmental and immunological signals. TNF, tumor necrosis factor; PE,
phosphatidylethanolamine; Gly, glycine; E1 and E2, ligases for ubiquitin-like conjugation systems.
for lethal encephalitis in mice 24 . These studies collectively suggested
a likely function for autophagy genes in pathogen defense in vivo. In
parallel, many studies demonstrated a function for autophagy in vitro
in defense against invading pathogens, including group A Streptococcus,
Shigella flexneri, Mycobacterium tuberculosis, Salmonella typhimurium
and Toxoplasma gondii 10,11,13 .
Two recent studies have further confirmed an antimicrobial function
for autophagy genes in host defense in vivo against intracellular
pathogens and have identified previously unknown relationships among
innate immune signaling, autophagy genes and potential autophagyindependent
functions of autophagy genes. An innate microbial sensor,
the peptidoglycan-recognition protein PRGP-LE, which recognizes bacterial
diaminopimelic acid–type peptidoglycan, is crucial for autophagic
control of Listeria monocytogenes infection in fly hemocytes and for host
survival 25 . PRGP-LE-mediated resistance is associated with autophagy
induction, requires the autophagy gene Atg5 and presumably involves
the xenophagic degradation of bacteria (as bacteria are visualized in
autophagosomes in wild-type animals). In this case, xenophagy targets
cytoplasmic bacteria that have escaped from the endosome for
envelopment in an autophagosome and destruction. In other cases
in which a pathogen inside a vesicular structure is targeted (such as
mycobacteria inside phagosomes, or salmonella inside vacuoles), the
membrane dynamics involved are incompletely defined. It may be that
an autophagosome can envelop the entire vesicular structure containing
462 volume 10 number 5 may 2009 nature immunology
P
LC3

© 2009 Nature America, Inc. All rights reserved.
Antigen presentation
Antigen
Autophagosome
Isolation
membrane
MIIC
MHC class II
antigen presentation
Activation of
interferon production
in plasmacytoid
dendritic cells
Isolation
membrane
Viral
nucleic acid
the pathogen or that the autophagic machinery somehow enhances
phagolyosomal maturation. An alternative mechanism is discussed
below for T. gondii.
Although TLR signaling has been linked to autophagy induction (discussed
below), the consequences of this autophagy induction have not
been studied so far; therefore, the finding of for a function for PRGP-LE
in listeria infection in flies is the first to demonstrate involvement of a
cytoplasmic pattern-recognition receptor in the delivery of microbes to
autophagosomes for degradation and in autophagy-mediated pathogen
defense. One speculation is that other pattern-recognition receptors
function similarly in innate immunity to target diverse microbes
or microbial products to the autophagosome. A related issue is whether
autophagosomal targeting motifs, such as monoubiquitination or
polyubiquitination, that can target cellular proteins and organelles to
the autophagosome through the adaptor protein SQSTM1 (p62) 26 also
target microbial products. It will be useful to determine how the selective
delivery of microbes, through cytoplasmic recognition by innate sensors
and/or ubiquitination, may result in the selective removal or killing of
pathogens by xenophagy, the activation of innate immune signaling
and/or the generation of antigenic peptides from these pathogens for
presentation to T cells.
In mice, phagocytic cell–specific deletion of Atg5 results in greater
susceptibility to infection with two different types of intracellular pathogens,
the bacterium L. monocytogenes and the protozoan T. gondii 27 .
Such work has provided important confirmation that autophagy genes
are involved in antimicrobial host defense in mammals. However, in
contrast to studies in the fly, in which autophagy is believed to degrade
bacteria that invade professional phagocytes called hemocytes, Atg5 is
Virus
Isolation
membrane
Autophagosome
Endosome
TLR7
Type 1 IFN
AutophagosomeLysosome
Suppression of
interferon production
in fibroblasts
and macrophages
Damaged
mitochondria
Reactive
oxygen
species
RIG-I-like
receptor signaling
Autolysosome
Reactive
oxygen
species
Autophagosome
Cell
survival and death
‘decisions’
Pathogen degradation
Isolation
membrane
Autophagosome
proposed to control T. gondii by a mechanism
independent of autophagosome formation.
Published work suggests that IFN-γ-mediated
control of T. gondii involves the localization
of immunity-related GTPases to the parasitophorous
vacuole, vacuolar membrane
stripping, and either the induction of green
fluorescent protein–LC3–positive vesicular
structures that localize together with GTPases 28
or the formation of autophagic membranes
near the damaged parasitophorous vacuoles 29 .
In Atg5 –/– macrophages, the GTPase IIGP1
(Irga6) fails to localize to the parasitophorous
vacuole, which results in a lack of disruption
of parasitophorous vacuole membranes. Thus,
the autophagy machinery, or at least Atg5, may
be critical in the destruction of a parasitecontaining
vesicular structure through the
recruitment of immunity-related GTPases to
the parasitophorous vacuole.
As for T. gondii–infected mouse astrocytes
28 , autophagosomal membranes are not
found surrounding the damaged parasitophorous
vacuole in wild-type macrophages,
which suggests that the GTPase-recruitment
function of Atg5, rather than autophagic
entrapment of T. gondii, may be the main
mechanism by which Atg5 functions ‘downstream’
of IFN-γ to mediate intracellular
resistance to this parasite. Studies using
real-time microscopy have documented that
killing of T. gondii is not consistently associated
with localization of LC3 to the parasi-
tophorous vacuole 30 . These data collectively provide support for a
model in which an autophagy gene, Atg5, is involved in host defense
independently of autophagosome formation. An important question
is whether recruitment of GTPase to the parasitophorous vacuole
requires autophagy proteins other than Atg5 and how Atg5, and perhaps
the autophagic machinery, mediates this trafficking event.
Immunity-related GTPases and autophagy
Several different immune signals positively regulate autophagy, including
PKR, TLRs, tumor necrosis factor, the CD40–CD40 ligand interaction,
IFN-γ and the immunity-related GTPases 10,13,27,31,32 , whereas T helper
type 2 cytokines (IL-4 and 1L-13) negatively regulate autophagy 33 . One
of the most rapidly evolving areas of research centers on the relationship
between autophagy and immunity-related GTPases (IRGs or p47
GTPases), a family of proteins crucial for IFN-γ-mediated resistance
to intracellular pathogens 34 . Although only the expression of mouse
p47 GTPases seems to be regulated by IFN-γ, both mouse (LRG47) and
human (Irgm1) p47 GTPases are thought to be required for IFN-γinduced
autophagy and antimycobacterial activity in macrophages 31,32 .
Furthermore, delivery of T. gondii to autophagosomes in macrophages
may occur in an Irgm3 (IGTP)-dependent way 29 . Thus, a consensus
is emerging that p47 GTPases are important in the IFN-γ-dependent
autophagic elimination of intracellular pathogens. The precise
mechanisms by which p47 GTPases exert this effect, however, remain
unknown. One theory is that their localization to pathogen-containing
phagosomes or vacuoles somehow facilitates damage to the pathogencontaining
compartment with subsequent targeting of pathogen to the
autophagosome.
nature immunology volume 10 number 5 may 2009 463
Lysosome
Lymphocyte
development
Figure 2 Immunological processes involving autophagy. The colored boxes present five immunological
processes that, on the basis of data available at present, involve classical autophagy. MIIC, MHC class
II loading compartment.
REVIEw

© 2009 Nature America, Inc. All rights reserved.
REVIEw
Beyond the proposed involvement of p47 GTPases in the autophagic
control of intracellular pathogens, evidence suggests potentially more
complex crosstalk between this family of immune effectors and
autophagy. As noted above, the autophagy protein Atg5 is needed
for recruitment of the GTPase Irga6 to the T. gondii parasitophorous
vacuole 27 . Thus, autophagy proteins can function ‘upstream’ of p47
GTPases to regulate their trafficking to pathogen-containing compartments.
As Irga6 that is associated with the parasitophorous vacuole
is in an active GTP-bound state and Irga6 in uninfected cells is in an
inactive GDP-bound state 35 , one obvious question is whether the
effects of Atg5 on Irga6 localization occur through regulation of the
GDP- or GTP-bound state of the GTPase. Another question is whether
the Atg5-Atg12-Atg16L1 complex enhances the ability of Irga6, and
perhaps other p47 GTPases, to form multimers on intracellular membranes.
Furthermore, it is not yet known whether pathogen-containing
phagosomes, in addition to parasitophorous vacuoles, may be sites of
autophagy gene–dependent recruitment of p47 GTPase. Along these
lines, it will be useful to determine whether Atg5 and other autophagy
genes are required for the recruitment of p47 GTPases to mycobacteriacontaining
phagosomes.
A somewhat paradoxical function for Irgm1 in negatively regulating
IFN-γ-induced autophagy has also been proposed. Mice deficient in
Irgm1 have less proliferation of mature effector CD4 + T cell populations
in the presence of IFN-γ 36 . This diminished expansion seems to be due
to the induction of a caspase-independent cell death program that is
blocked by PI(3)K inhibitors such as wortmanin or LY294002 or small
interfering RNA specific for the autophagy gene encoding Beclin 1. Thus,
Irgm1 may have dual interactions with the autophagy pathway that act
in synergy to promote IFN-γ antimicrobial interactions; Irgm1 may signal
autophagic sequestration of intracellular pathogens in macrophages
while simultaneously limiting IFN-γ-induced autophagic death of effector
T lymphocytes. That suggests the possibility of divergent functions
for Irgm1 and potentially other p47 GTPases in autophagy regulation
in different cell types. At present, it is unknown how Irgm1 might negatively
regulate IFN-γ-induced autophagy to preserve activated CD4 +
T cell populations. Furthermore, the possibility has not yet been formally
excluded that the greater vacuolization in Irgm1 –/– lymphocytes
is a function of aberrant membrane trafficking rather than enhanced
autophagic flux. As reviewed elsewhere, whether autophagy is truly
involved in mediating cell death also remains controversial 21 .
TLRs and autophagy
TLRs and the autophagy pathway intersect at many different levels 37 :
TLRs can regulate autophagy induction 38,39 , the autophagy machinery
can be used to deliver viral genetic material to endosomal TLRs for efficient
induction of type I interferon 40 , and TLRs may act in the recruitment
of autophagy proteins to phagosomal membranes 16 . Initially it
was shown that bacterial lipopolysaccharide (LPS) induces autophagy
through its cognate receptor TLR4 in macrophages by a mechanism
that requires the TLR adaptor TRIF but not the adaptor MyD88 and
requires the ‘downstream’ molecules receptor-interacting protein 1 and
p38 mitogen-activated protein kinase 38 . Subsequently, another group
confirmed induction of autophagy by LPS-TLR4 and also reported
induction of autophagy by single-stranded RNA, poly(I:C) and imiquimod
through TLR3 or TLR7 39 . TLR7-dependent induction of
autophagy was found to be MyD88 dependent, and another study has
proposed that in response to TLR activation, both TRIF and MyD88
trigger autophagy through a direct interaction with Beclin 1 (ref. 41).
However, it is not yet known how this interaction stimulates autophagy,
and further studies are needed to more fully elucidate the molecular
mechanism(s) by which TLR signaling activates autophagy.
It is noteworthy that more robust TLR-dependent stimulation of
autophagy in RAW264.7 mouse macrophages than in primary macrophages
has been reported 39 and that autophagy induction has not
been reported in primary fetal liver–derived macrophages treated with
LPS or several other TLR stimuli 42 . These primary fetal liver–derived
macrophages, however, are able to mount an autophagic response to
starvation or invasive bacteria, which indicates that this lack of responsiveness
to TLR stimuli does not reflect a deficiency in their general
ability to undergo stress-induced autophagy. These data conflict with
the data outlined above showing positive regulation of autophagy by
TLR signaling. One possibility is that the function of TLR signaling in
autophagy induction has been examined over different time frames in
different studies. Alternatively, there may be TLR-independent mechanisms
for the induction of autophagy in phagocytic cells, the regulation
of autophagy by TLRs may be cell type specific or the function of TLRs
may be tightly regulated by the differentiation state of the macrophage.
The physiological importance of TLR regulation of autophagy requires
further studies in primary cells, in cells other than macrophages and
dendritic cells and, perhaps most importantly, in different in vivo models
of microbial infection.
Although the physiological function of TLR-mediated induction of
autophagy is not yet known, it seems reasonable to speculate that it
may function either in the autophagic control of intracellular pathogens
and/or in other autophagy gene–dependent functions in immunity.
As noted above, the autophagic control of L. monocytogenes in flies
requires another type of pattern-recognition receptor, PGRP-LE 25 ; a
critical question is whether such findings represent a general paradigm
that applies to other families of pattern-recognition receptors in mammalian
hosts, including TLRs and Nod-like receptors. At least in vitro,
LPS stimulation of macrophages (which activates TLR4) or TLR7 activation
results in greater localization of M. tuberculosis in autophagosomes
and lower mycobacterial survival 38,39 , which raises the possibility that
TLR induction of autophagy may similarly participate in autophagic
control of intracellular pathogens in vivo. It is also possible that TLR
induction of autophagy represents a type of feed-forward mechanism
for priming the autophagy machinery to deliver viral nucleic acids to
endosomal TLRs to activate innate immune signaling. Furthermore,
another unexplored possibility is that induction of autophagy by TLRs
or pattern-recognition receptors enhances autophagy-mediated presentation
of endogenous antigens to major histocompatibility complex
(MHC) class II loading compartments (Fig. 2).
Studies also suggest that TLR signaling may not only induce classical
autophagy but also recruit autophagy proteins to the phagosomal
membrane and deliver phagocytosed material to the lysosome (Fig. 3).
Engagement of TLRs during the phagocytosis of LPS-coated beads or
zymosan particles induces the rapid recruitment of Beclin 1 and LC3
to phagosomal membranes in an Atg5- and Atg7-dependent way 16 .
Furthermore, proteomic analyses of latex bead–containing phagosomes
from macrophages shows enhanced recruitment of LC3 to membranes
in response to autophagy inducers 43 . The translocation of Beclin
1 and LC3 to phagosomal membranes in macrophages treated with
LPS-coated beads or zymosan particles is associated with enhanced
phagosome fusion with lysosomes but not with detectable doublemembraned
structures characteristic of autophagosomes 16 . Thus, it
is postulated that TLR signaling, through recruitment of autophagy
proteins to the phagosome, may constitute another mechanism for
expediting lysosomal microbial elimination. Similar to the failure to
detect autophagosomes surrounding T. gondii parasitophorous vacuoles
in some studies 27,28,30 , it is not yet known whether the elimination
of pathogen-containing phagosomes is truly a process that requires
autophagy proteins but not the formation of autophagosomes or,
464 volume 10 number 5 may 2009 nature immunology

© 2009 Nature America, Inc. All rights reserved.
Figure 3 Immunological processes involving autophagy genes. The colored boxes present six
immunologic processes that involve various autophagy genes but that are not yet proven to involve
classical autophagy.
alternatively, whether autophagosomes are sufficiently evanescent to
have escaped detection in ultrastructural studies so far. Nevertheless,
the findings obtained with T. gondii and with TLR signaling, phagocytosis
and autophagy proteins do raise the possibility that components
of the autophagy machinery may be involved in membrane-mediated
events that do not involve the classical delivery of cytoplasmic constituents
or pathogens to the lysosome (Fig. 3).
Autophagy genes and cytokine secretion
The first connection between autophagy and innate immune signaling
emerged after the demonstration that Atg5 is essential for the production
of type I interferon in plasmacytoid dendritic cells infected with vesicular
stomatitis virus by a mechanism presumed to involve autophagymediated
delivery of viral genetic material to endosomal TLRs 40 (Fig.
2). Somewhat paradoxically, several studies have shown that absent or
hypomorphic expression of autophagy genes in certain cell types can
result in enhanced production of type I interferon or other cytokines,
including proinflammatory molecules such as IL-1β and IL-18, or adipocytokines,
such as leptin and adiponectin 42,44–46 (Fig. 3). As discussed
below, enhanced secretion of such molecules may be involved in certain
aspects of Crohn’s disease in mice deficient in the expression of the
autophagy gene Atg16l1. Thus, the autophagic machinery may serve a
dual function in innate immune signaling, acting not only to stimulate
antiviral type I interferon responses in dendritic cells but also to ensure
homeostatic balance by preventing excess innate immune activation in
other cell types.
Two distinct models have been proposed to explain the higher production
of type I interferon in autophagy gene–deficient fibroblasts and
primary macrophages (Fig. 3). Enhanced production of type I interferon
in Atg5 –/– and Atg7 –/– mouse embryonic fibroblasts in response
to infection with vesicular stomatitis virus or double-stranded RNA has
REVIEw
been reported 44 . In this model, the Atg5-Atg12
conjugate can bind to cytoplasmic viral sensors,
the retinoic acid–inducible helicases RIG-1 and
Mda5 and their adaptor protein IPS-1, to suppress
the activity of such helicases in stimulating
the production of type I interferon. Thus, it
is possible that the autophagic machinery may
suppress innate immune signaling by direct
inhibitory interactions with these helicases
and their adaptor proteins. Of note, however, a
completely different, but not mutually exclusive,
mechanism for the higher production of type I
interferon in Atg5-deficient mouse embryonic
fibroblasts and primary macrophages has also
been proposed 45 . Dysfunctional mitochondria
have been shown to accumulate in the absence
of Atg5, leading to higher concentrations of
IPS-1 and a reactive oxygen species–dependent
increase in retinoic acid–inducible helicase signaling
and production of type I interferon in
response to infection with vesicular stomatitis
virus (Fig. 3). Therefore, it is also possible
that the homeostatic functions of constitutive
autophagy, especially mitochondrial turnover
and quality control, indirectly regulate the production
of type I interferon.
It will be useful to determine whether this
enhanced response to viral infection is present
in cells deficient in autophagy genes that act
in other steps of the autophagy pathway and
whether the underlying mechanism reflects a deficiency of the cellular
process of autophagy or alternative, as-yet-undefined functions of
autophagy proteins in regulating interferon signaling. Because Atg5deficient
plasmacytoid dendritic cells with impaired type I interferon
responses are more sensitive to infection with vesicular stomatitis virus
and Atg5-deficient mouse embryonic fibroblasts and primary macrophages
with enhanced type I interferon responses are more resistant
to such infection, another critical question is how potential cell type–
specific differences in autophagy gene–dependent regulation of type I
interferon signaling may interact in vivo to determine the fate of viral
infection. It is not yet clear whether the autophagic machinery truly
has a direct immunosuppressive function that increases susceptibility
to viral infection by inhibiting cytokine release or if it merely functions
as a negative feedback mechanism that provides a brake on unneeded
innate antiviral signaling.
Another study has shown an important function for the autophagy
gene Atg16l1 in regulating the secretion of proinflammatory cytokines
(Figs. 3,4). The Nod-like receptor protein cyropryrin (also called NALP
or NLRP3) forms a complex known as the ‘inflammasome’, which contains
the adaptor protein ASC (apoptosis-associated speck-like protein
containing a caspase-activating and caspase-recruitment domain) and
caspase 1 and is responsible for the processing of pro-IL-1β to its mature,
secreted form 47,48 . Macrophages from mice lacking Atg16l1 produce
more of the proinflammatory cytokines IL-1β and IL-18 after endotoxin
stimulation of TLR4 (ref. 42). Similarly, mouse chimeras engrafted
with Atg16l1 –/– fetal liver hematopoietic progenitors have higher serum
concentrations of IL-1β and IL-18 after treatment with dextran sodium
sulfate; this enhanced cytokine secretion probably contributes to pathology,
as treatment with antibody to IL-1β and antibody to IL-18 results
in lower susceptibility of Atg16l1 –/– chimeric mice to dextran sodium
sulfate–induced colitis.
nature immunology volume 10 number 5 may 2009 465

© 2009 Nature America, Inc. All rights reserved.
REVIEw
The precise mechanisms responsible for the enhanced secretion of
IL-1β and IL-18 by Atg16l1-deficient cells are not yet clear, but TRIFdependent
activation of caspase 1 and consequent enhanced processing
of IL-1β in Atg16l1 –/– macrophages has been noted. In addition, similar
to results obtained with virus-infected Atg5-deficient cells 45 , higher production
of reactive oxygen species in LPS-stimulated Atg16l1 –/– macrophages
has been found 42 . Thus, a gain of function in cytokine release
may be obtained with deficiency in both Atg5 and Atg16l1, and in both
cases it may result directly from effects of the autophagy machinery on
signaling pathways that regulate cytokine production and/or indirectly
from effects of autophagy on the cellular redox state.
Another study has emphasized the involvement of Atg16L1 in regulating
cytokine expression in a specialized intestinal epithelial cell that is
important in innate immunity, the Paneth cell (Figs. 3,4). Substantially
enhanced gene expression of acute-phase reactants, PPAR signaling molecules
and adipocytokines such as leptin and adiponectin was observed
after transcriptional profiling by microarray of Paneth cell RNA from
mice with a hypomorphic Atg16l1 allele 46 . Similar alterations have not
been found in thymocytes from these mice, which suggests that Atg16l1
deficiency may upregulate cytokine production in a cell type–specific
way. Furthermore, this study has emphasized the potential importance
of the autophagic machinery in Paneth cells, as specialized cells of the
innate immune system, in regulating the expression of molecules with a
broad range of systemic pro- and anti-inflammatory effects. The mechanism
by which deficiency in an autophagy gene results in these transcriptional
changes in Paneth cells remains unknown. It is also unknown
whether the polymorphism in human ATG16L1 that is associated with
Crohn’s disease (ATG16L1 T300A) results in similar changes and, if so,
whether such changes contribute to the pathophysiology of Crohn’s
diseases and/or other types of inflammatory diseases.
Autophagy, antigen presentation and thymic selection
There is growing evidence that autophagy may be involved in the MHC
class II presentation of certain endogenously synthesized peptides 10,49
(Fig. 2). The involvement of autophagy in MHC class I–restricted antigen
presentation remains more speculative 50 ; one paper has suggested
involvement of Atg5 in MHC class I antigen presentation 51 . Although
it is clear that autophagosomes can deliver peptides to MHC class II
loading compartments for presentation to CD4 + T cells 52,53 , it is not yet
known whether this represents a major pathway for antigen presentation
during the generation of adaptive immune responses in physiological
conditions in primary cells or in vivo. It will therefore be necessary to
assess the involvement of autophagy in the function of professional
antigen-presenting cells during immune responses in vivo. Regardless
of the results, the specific targeting of antigens to autophagosomes by
fusion with the LC3 autophagy protein may represent an effective vaccine
strategy for enhancing CD4 + T cell responses; for example, a 20-fold
enhancement of MHC class II presentation to CD4 + T cell clones has
been noted when influenza virus matrix protein is targeted to autophagosomes
by fusion with LC3 (ref. 53).
One notable physiological context in which autophagy pathway–
dependent endogenous loading of MHC class II may be important
is in shaping the T cell repertoire during thymic selection. Thymic
epithelial cells that function in positive and negative selection show
constitutive autophagy, as demonstrated by the presence of Atg5dependent
green fluorescent protein–LC3 dots 54,55 . Normal thymic T
cell selection requires Atg5 expression in the stromal cells in thymic
allografts 55 . Experiments with thymic transplantation have shown that
the selection of certain MHC class II–dependent TCRs but not MHC
class I–dependent TCRs requires Atg5. Furthermore, mice with Atg5 –/–
thymic implants develop autoreactive CD4 + T cells, as measured both
by the proportion of CD62L lo peripheral cells and the development of
inflammatory infiltrates in many organs, including the intestine. This
last observation supports speculation that abnormal negative selection
might explain the association between a polymorphism in the autophagy
protein Atg16L1 and susceptibility to Crohn’s disease (discussed below;
Fig. 4). The development of autoreactive CD4 + T cells in mice with
Atg5 –/– thymic implants is thought to reflect a function of Atg5 and,
presumably, the autophagy pathway in controlling the repertoire of self
peptides presented by thymic epithelial cells that are responsible for
normal thymic selection and the generation of T cell tolerance. However,
autophagy genes other than Atg5 have not yet been analyzed in similar
studies, and thus the function of the cellular process of autophagy in
thymic selection remains to be defined. It will be useful to determine
whether the mechanism responsible for the function of Atg5 in thymic
selection reflects the generation of peptides or a function for nutritional,
survival or transcriptional responses regulated by Atg5 or other
autophagy genes.
Another important question is whether autophagy, in either the
antigen-presenting cell or the cell donating antigen, is involved in efficient
antigen cross-presentation. Indeed, one study has shown that small
interfering RNA–mediated knockdown of either Beclin 1 or Atg12 in
antigen-donor tumor cells results in less cross-presentation 56 . The
authors postulate a mechanism involving direct involvement of the
autophagosome as a carrier of protein antigens from tumor cells. More
studies are needed to confirm this mechanism, as well as explore its
potential clinical implications for cancer vaccines.
An additional potential link between autophagy and cross-presentation
has emerged from the observations that autophagy is required for
the removal of apoptotic corpses in mouse embryoid body and chick
retinal development 57,58 . In these settings, autophagy is required for
dying cells to have sufficient energy to generate the engulfment signals
necessary for the clearance of such corpses by phagocytes. Therefore, one
interesting, as-yet-unexplored theory is that autophagy in dying cells
may also be required for the generation of signals that induce efficient
uptake of cell corpses for subsequent cross-presentation. This hypothesis
is potentially important for understanding and enhancing the immunogenicity
of cell-based tumor vaccines. The function of autophagy in
the clearance of dead cells during normal tissue turnover might also
contribute to peripheral tolerance.
Autophagy genes and lymphocytes
Lymphocytes undergo extensive rearrangement of cytoplasm and
organelles during the selection, expansion and contraction of antigenspecific
clones and frequently confront life-or-death decisions during
their development and activation. Therefore, it is not unexpected that
autophagy might have critical cell-intrinsic functions in lymphocyte
biology. Indeed, several studies have shown that deletion of autophagy
genes in T lymphocytes and B lymphocytes alters lymphocyte development,
survival and/or function (Fig. 2).
The thymus is a site of constitutive autophagy 54,55 , and the expression
of Beclin 1 is regulated during T cell and B cell development and T cell
activation 59 , which provides indirect evidence of potential links between
autophagy and lymphocyte development. Of note, loss of Atg5 or Atg7
impairs the survival and proliferation of mature T lymphocytes in
vivo 60–62 , and Atg5 is required in B lymphocytes for the survival of developing
pre–B cells in the bone marrow and of mature B-1a cells in the
periphery 63 . On the basis of studies of lethally irradiated mice reconstituted
with Atg5 –/– fetal liver progenitors, it seems that macrophages, plasmacytoid
dendritic cells, neutrophils, erythrocytes and immature T cells
develop and survive normally in the absence of Atg5. Therefore, a critical
question is how deletion of Atg5 and Atg7 confers lymphocyte-specific
466 volume 10 number 5 may 2009 nature immunology

© 2009 Nature America, Inc. All rights reserved.
Abnormal secretion into
intestinal lumen
Abnormal
granules
mRNAs for adipocytokines
and acute phase reactants
Abnormal
small
vesicles
Autophagy gene mutation
(Atg5, Atg7, Atg16l1)
in Paneth cells
Degenerating
mitochondria
Intestinal inflammation
Release of autoimmune T cells
Abnormal
thymus
gland
Abnormal
negative
selection
Autophagy gene mutation
(Atg5) in
thymic epithelial cells
and lymphocyte lineage–specific (for example, T cell and B-1a cell but
not B2 B cell) effects on development and survival.
It is not yet completely clear whether the effects of the deletion
of Atg5 and Atg7 on lymphocyte development are mediated through
autophagy (as proposed in Fig. 2) or other effects of these genes; studies
of mice with mutations in genes whose products act at other stages in
the autophagy pathway may be helpful in elucidating this. One mechanism
postulated for the involvement of Atg7 and Atg5 in the survival of
mature T cells involves a classical autophagy function in the clearance
of mitochondria and consequent prevention of the accumulation of
reactive oxygen species and imbalance of the expression of pro- and
antiapoptotic proteins 60,62 . Notably, the developmental pattern of
Beclin 1 in B lymphocytes, similar to that of the antiapoptotic protein
Bcl-2, shows downregulation at the transition from pro–B cell to
pre–B cell 58 . Thus, it will be useful to determine whether suppression
of autophagy, in addition to apoptosis, is necessary for the physiological
cell death that occurs at this stage in B lymphocyte development
and whether null mutations in autophagy genes result in excessive cell
death that is incompatible with efficient development of pre–B cells
in the bone marrow.
A controversial area is whether autophagy is exclusively a prosurvival
pathway in lymphocytes or whether it also can be used as a death pathway.
As noted before, deletion of Atg5 or Atg7 results in the impaired
survival of certain populations of B cells or T cells 60–63 , which suggests a
major function in lymphocyte survival. However, as discussed earlier, it
has been postulated that IFN-γ may induce Irgm1-regulated autophagic
cell death in CD4 + T cells 36 . In addition, knockdown of the gene encoding
Beclin 1 or Atg7 results in less death of CD4 + T cells induced by
growth-factor withdrawal 64 , and autophagy genes are required for the
death of bystander CD4 + T cells triggered by the human immunodeficiency
virus envelope protein 65 . Thus, a consensus is growing that
autophagy may mediate activation-induced death of CD4 + T cells.
Failure to control
intracellular
bacterial replication
Increased ROS production
Increased LPS-induced
IL-1β and IL-18 secretion
Autophagy gene mutation
(Atg7, Atg16l1)
in macrophages
However, several caveats must be considered when determining
whether autophagy is truly a cause of cell death (death by autophagy)
rather than simply being associated with cell death (death with
autophagy) 21 . An example of some of the complexities in determining
the contribution of autophagy to T cell death is provided by a
study examining the inter-relationships among signaling by the death
domain protein FADD, autophagy and the death of primary CD8 + T
cells 66 . Transgenic expression of dominant negative FADD in CD8 +
T cells leads to enhanced autophagy and cell death in response to
mitogenic signals, which can be prevented by either targeting of the
autophagy pathway (as with pharmacological inhibitors, dominant
negative Vps34 or short hairpin RNA specific for Atg7) or by use of the
necropoptosis inhibitor Nec-1, which blocks the kinase RIPK1. Thus,
FADD signaling may somehow dampen the autophagy response to
prevent RIPK1-dependent necroptotic cell death, but it is as yet unclear
whether autophagy is a true cell death pathway for proliferating CD8 +
T cells or whether the autophagic machinery functions in signaling
necroptosis. If autophagy does prove to be a cell death pathway for T
cells, one possible explanation for the apparently conflicting observations
is that basal autophagy is required for T cell survival, whereas
unrestrained autophagy, which may occur in response to IFN-γ stimulation
or compromise of FADD signaling, may be deleterious.
The stress-activated signaling molecules Jnk1 and Jnk2 may also be
involved in the dual integration of autophagy signaling with T cell differentiation,
activation and function. Jnk1 and Jnk2 both serve important
but nonredundant functions in CD4 + and CD8 + T cells 67–69 . Notably, several
studies have indicated that Jnk kinases are important in autophagic
signaling. Although one study has reported more green fluorescent
protein–LC3 dots in T cells lacking Jnk1 or Jnk2 (ref. 64), another study
has found that Jnk1 positively regulates autophagy by a mechanism that
involves phosphorylation of Bcl-2 and disruption of the Bcl-2–Beclin 1
complex, but Jnk2 does not 70 . The last study showed that Jnk2-deficient
nature immunology volume 10 number 5 may 2009 467
ROS
IL-1β
IL-18
ROS
Altered gut
microbiota
Natural
antibody
B1a
B cell
Atg5 mutation
B cell
precursor
Altered
intestinal
immunity
Altered T cell
homeostasis
Atg5, Atg7,
Irgm1 mutations
T cell
Autophagy gene mutation
(Atg5, Atg7, Irgm1)
in lymphocytes
REVIEw
Figure 4 Possible mechanisms by which alterations in autophagy may be involved in the pathogenesis of human Crohn’s disease. The colored boxes present
four potential mechanisms by which the mutation of an autophagy gene could contribute to Crohn’s disease, proposed on the basis of studies of model
organisms in which various autophagy genes have been mutated in various cell types (boxes at bottom). So far, only ATG16L1 and IRGM1 have been linked to
susceptibility to Crohn’s disease. ROS, reactive oxygen species.

© 2009 Nature America, Inc. All rights reserved.
REVIEw
mouse embryonic fibroblasts have enhanced Jnk1 signaling and hyperactive
autophagy 70 . Therefore, it is plausible that the lack of autophagy
in Jnk1-deficient mice (or perhaps even enhanced autophagy in Jnk2deficient
mice) may contribute to alterations in T cell differentiation.
Further assessment of the function of Jnk signaling in the regulation of
autophagy in T cells and the effects of Jnk signaling–mediated autophagy
regulation in T cell differentiation is clearly needed.
Autophagy genes and Crohn’s disease
One of the most exciting recent developments in the field of autophagy
genes and immunity was catalyzed by genome-wide association studies
identifying strong links between polymorphisms in two human
autophagy genes, IRGM (called ‘IRGM1’ here) and ATG16L1, and susceptibility
to inflammatory bowel disease 71,72 (Fig. 4). A polymorphism
in IRGM1 is associated with both Crohn’s disease and ulcerative colitis;
this association is due to the absence of specific single-nucleotide
polymorphisms associated with a 2.7-kilobase deletion upstream of the
IRGM1 transcriptional start site 71,73,74 . This raises the possibility that
Irgm1-mediated regulation of autophagy may contribute to susceptibility
to inflammatory bowel disease; however, so far there are no studies
that mechanistically link Irgm1 to features of inflammatory bowel disease
in animal models, nor is it known whether the IRGM1 risk allele
affects autophagy regulation.
The Crohn’s disease–associated ATG16L1 risk allele encodes a protein
with a threonine-to-alanine substitution (T300A) in the carboxyterminal
domain containing tryptophan–aspartic acid dipeptide (WD)
repeats. Although this domain of Atg16L1 is not conserved in yeast, is
not required for the formation of multimeric complexes with Atg5-
Atg12 and is not required for starvation-induced autophagy, the Atg16L1
T300A mutant protein may have lower stability and may be defective in
localizing the autophagic machinery to intracellular bacteria 75,76 . Most
notably, two studies of mice with genetically engineered mutations in
Atg16l1 have identified critically important functions of this autophagy
protein in innate immunity 42,46 .
Mice with deletion of the coiled-coil domain of Atg16L1, which
results in the lack of a functional protein, have been generated 42 . As
reported for null mutations of Atg5 and Atg7, these mice die in the first
day of life, presumably because of starvation or a bioenergetic crisis
during the postnatal period. As noted above, studies with macrophages
from these mice have shown enhanced cytokine responsiveness after
stimulation with LPS (TLR4) or stimulation of other TLRs, and chimeric
mice engrafted with Atg16l1 –/– hematopoietic progenitors from
fetal liver have increased mortality, weight loss, colitis and serum concentrations
of IL-1β and IL-18 in response to dextran sodium sulfate.
This Atg16L1-dependent regulation of endotoxin-induced activation
of inflammasomes demonstrates a previously unknown function for
a component of the autophagic machinery in controlling inflammatory
immune responses. A key question is whether the Atg16L1 T300A–
encoding risk allele of ATG16L1 results in similar hyperexpression of
proinflammatory genes and thus represents a potential mechanism by
which this variant protein increases the risk for Crohn’s disease and
potentially other inflammatory disorders (Fig. 4). It will also be useful
to determine if this gain-of-function increase in cytokine release occurs
with deficiency of additional autophagy proteins.
Viable mice have been bred that are hypomorphic for Atg16L1 expression
by means of intronic insertion of a gene-trap vector, which has
allowed direct assessment of the effects of lower Atg16L1 expression
on intestinal pathology in adult mice 46 (Fig. 4). As function is not
completely abolished (but may be lower because of altered stability)
by the T300A substitution of Atg16L1, this model of lowered mouse
Atg16L1 protein expression may be relevant to humans carrying the
allele encoding Atg16L1 T300A. Notably, in addition to the transcriptional
changes in adipocytokines and other inflammatory regulators
in Paneth cells described above, prominent histological abnormalities
in Paneth cells of Atg16L1-hypomorphic mice have been reported that
resemble changes in resected ileal tissue from patients with Crohn’s
disease who are homozygous for the allele encoding Atg16L1 T300A.
In both settings, Paneth cells show morphological defects in secretory
granules and granule exocytosis, and in the Atg16L1-hypomorphic mice,
deficient Paneth cell secretion of the antibacterial protein lysozyme has
been noted in the small intestine.
As Paneth cells function as a barrier to bacterial invasion (in part by
secreting granule contents that contain antimicrobial peptides) and as
regulators of intestinal inflammation 77,78 , the findings reported above
suggest that defects in Atgl6L1 function in Paneth cells may contribute to
the pathogenesis of Crohn’s disease. Studies of mice with intestinal epithelial
cell–specific deletion of other autophagy genes, including Atg5 or
Atg7, have demonstrated similar abnormalities in Paneth cells 46,79 . Thus,
several components of the autophagic machinery that act at the membrane
expansion-completion stage (such as Atg16L1, Atg5 and Atg7)
are involved in the maintenance of Paneth cell function. At present, it is
not yet known whether this function of Atg16L1, Atg5 and Atg7 reflects
involvement of classical autophagy in Paneth cell function or perhaps an
as-yet-uncharacterized function for the autophagy protein–conjugation
machinery in granule exocytosis (Fig. 3).
A model is thus emerging in which dysregulation of autophagy gene
function may contribute to the pathogenesis of Crohn’s disease or other
types of intestinal inflammatory disorders by several possible routes
(Fig. 4). Lack of Atg5 in the thymus results in autoreactive CD4 + T cells
and intestinal inflammatory infiltrates 55 ; lack of Atg16l1 in macrophages
results in enhanced endotoxin-induced inflammatory signaling 42 , and
lower expression of Atg16l1 in Paneth cells results in transcriptional
alterations in molecules that regulate inflammation and in defects in
granule exocytosis 46 . More speculatively, defects in other autophagy
gene–dependent processes, such as the maintenance of normal numbers
of B-1a cells that make natural antibody to bacterial carbohydrate
antigens 62 or the survival of T cells 60–62 , might be involved in the pathogenesis
of Crohn’s disease. Although much more work is needed to elucidate
the precise effect of polymorphisms in IRGM1 and ATG16L1 on
the pathogenesis of Crohn’s disease, these observations are beginning
to delineate previously unknown functions for autophagy genes in the
regulation of innate immunity.
Therapeutic implications
The many functions of autophagy and autophagy genes in immunity
provide both opportunities and risks for manipulating autophagy therapeutically.
For example, there is the opportunity to enhance CD4 + T
cell–dependent vaccines by targeting antigens to the autophagic pathway.
A notable idea that has arisen from the finding that autophagy is
important in host resistance to viral, bacterial and parasitic infection is
that pharmacological enhancers of autophagy may function as broadspectrum
antimicrobial agents. Furthermore, enhancing autophagy
might ameliorate Crohn’s disease if agents that overcome deficiencies
in the function of Atg16L1 or Irgm1 can be identified. The potential
beneficial effects of autophagy in immunity must be taken into consideration
in the clinical development of agents for other diseases, such
as cancer, that are targeted at autophagy inhibition. Although inhibition
of autophagy-dependent cell survival may be beneficial in cancer
therapy, the adverse potential immunological effects of such approaches
include diminished negative selection of thymocytes, with consequent
autoimmunity, enhanced secretion of proinflammatory cytokines by
hyper-responsive innate immune cells, greater susceptibility to infection
468 volume 10 number 5 may 2009 nature immunology

© 2009 Nature America, Inc. All rights reserved.
with intracellular pathogens and lower survival of B-1a cells and T cells.
Perhaps additional understanding of the molecular basis of the diverse
functions of autophagy genes in immunity will open new doors for
rational intervention in a variety of diseases without untoward immunological
consequences.
Conclusion
Studies of autophagy genes and immunity have identified important
potential functions for autophagy and unexpected functions for
autophagy proteins in the regulation of innate immunity and inflammation.
Autophagy proteins are involved in immunity-related processes
that may not involve the formation of autophagosomes, such as the
recruitment of immunity-related GTPases to pathogen-containing compartments,
TLR-mediated phagolysosomal maturation, the regulation
of RNA helicases and the inflammasome, and the exocytosis of Paneth
cell granules (Fig. 3). Converging evidence suggests that autophagy proteins,
although critical for normal immune responses, may also function
to prevent immunological hyper-responsiveness in certain cell types,
such as macrophages, fibroblasts and Paneth cells; this may at least in
part underlie the associations between polymorphisms in autophagy
genes and Crohn’s disease. Another emerging theme is that deletion
of autophagy genes can have different effects in different immune cell
populations, which may reflect an effect of the milieu of differentiated
immune cells on the function of autophagy genes and/or cell type–specific
differences in the complex interaction between autophagy proteins
and the specialized functions of various cells of the immune system.
Distinguishing between the effects of autophagy proteins that are mediated
by classical autophagy and those that are perhaps mediated by other
non–autophagosome-dependent mechanisms may help elucidate how
proteins of a primordial stress-response pathway may have functionally
diversified to shape the immunological responses of complex higher
eukaryotic organisms.
ACKNOWLEDGMENTS
We thank A. Diehl for medical illustration. Supported by the US National Institutes
of Health (R01 CA074730, R01 CA096511, R01 AI054483, R01 AI065982 and U54
AI057160 to H.W.V., and R01 AI151267 and R01 CA109618 to B.L.), the Howard
Hughes Medical Institute (B.L.) and the Ellison Medical Foundation (B.L.).
Published online at http://www.nature.com/natureimmunology/
reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Levine, B. & Kroemer, G. Autophagy in the pathogenesis of disease. Cell 132, 27–42
(2008).
2. Mizushima, N., Levine, B., Cuervo, A.M. & Klionsky, D.J. Autophagy fights disease
through cellular self-digestion. Nature 451, 1069–1075 (2008).
3. Levine, B. Eating oneself and uninvited guests: autophagy-related pathways in cellular
defense. Cell 120, 159–162 (2005).
4. Kuma, A., Mizushima, N., Ishihara, N. & Ohsumi, Y. Formation of the approximately
350-kDa Apg12-Apg5.Apg16 multimeric complex, mediated by Apg16 oligomerization,
is essential for autophagy in yeast. J. Biol. Chem. 277, 18619–18625 (2002).
5. Hara, T. et al. FIP200, a ULK-interacting protein, is required for autophagosome
formation in mammalian cells. J. Cell Biol. 181, 497–510 (2008).
6. Sun, Q. et al. Identification of Barkor as a mammalian autophagy-specific factor for
Beclin 1 and class III phosphatidylinositol 3-kinase. Proc. Natl. Acad. Sci. USA 105,
19211–19216 (2008).
7. Itakura, E., Kishi, C., Inoue, K. & Mizushima, N. Beclin 1 forms two distinct phosphatidylinositol
3-kinase complexes with mammalian Atg14 and UVRAG. Mol. Biol.
Cell 19, 5360–5372 (2008).
8. Zhong, Y. et al. Distinct regulation of autophagic activity by Atg14L and Rubicon
associated with Beclin 1–phosphatidylinositol-3-kinase complex. Nat. Cell. Biol. 11,
468–476 (2009).
9. Matsunaga,K. et al. Two Beclin 1-binding proteins, Atg14L and Rubicon, reciprocally
regulate autophagy at different stages. Nat. Cell Biol. 11, 385–396 (2009).
10. Munz, C. Enhancing immunity through autophagy. Annu. Rev. Immunol. 27, 423–449
(2009).
11. Levine, B. & Deretic, V. Unveiling the roles of autophagy in innate and adaptive immunity.
Nat. Rev. Immunol. 7, 767–777 (2007).
12. Schmid, D. & Munz, C. Innate and adaptive immunity through autophagy. Immunity
27, 11–21 (2007).
REVIEw
13. Orvedahl, A. & Levine, B. Eating the enemy within: autophagy in infectious diseases.
Cell Death Differ. 16, 57–69 (2009).
14. Klionsky, D.J. et al. Guidelines for the use and interpretation of assays for monitoring
autophagy in higher eukaryotes. Autophagy 4, 151–175 (2008).
15. Kuma, A., Matsui, M. & Mizushima, N. LC3, an autophagosome marker, can be incorporated
into protein aggregates independent of autophagy: caution in the interpretation
of LC3 localization. Autophagy 3, 323–328 (2007).
16. Sanjuan, M.A. et al. Toll-like receptor signaling in macrophages links the autophagy
pathway to phagocytosis. Nature 450, 1253–1257 (2007).
17. Jackson, w.T. et al. Subversion of cellular autophagosomal machinery by RNA viruses.
PLoS Biol. 3, 861–871 (2005).
18. Suzuki, K., Noda, T. & Ohsumi, Y. Interrelationships among Atg proteins during
autophagy in Saccharomyces cerevisiae. Yeast 21, 1057–1065 (2004).
19. Matsui, Y. et al. Distinct roles of autophagy in the heart during ischemia and reperfusion:
roles of AMP-activated protein kinase and Beclin 1 in mediating autophagy. Circ.
Res. 100, 914–922 (2007).
20. Codogno, P. & Meijer, A.J. Atg5: more than an autophagy factor. Nat. Cell Biol. 8,
1045–1047 (2006).
21. Kroemer, G. & Levine, B. Autophagic cell death: the story of a misnomer. Nat. Rev.
Mol. Cell Biol. 9, 1004–1010 (2008).
22. Liang, X.H. et al. Protection against fatal Sindbis virus encephalitis by beclin, a novel
Bcl-2-interacting protein. J. Virol. 72, 8586–8596 (1998).
23. Liu, Y. et al. Autophagy regulates programmed cell death during the plant innate
immune response. Cell 121, 567–577 (2005).
24. Orvedahl, A. et al. HSV-1 ICP34.5 Confers neurovirulence by targeting the beclin 1
autophagy protein. Cell Host Microbe 1, 23–35 (2007).
25. Yano, T. et al. Autophagic control of listeria through intracellular innate immune recognition
in Drosophila. Nat. Immunol. 9, 908–916 (2008).
26. Kim, P.K., Hailey, D.w., Mullen, R.T. & Lippincott-Schwartz, J. Ubiquitin signals
autophagic degradation of cytosolic proteins and peroxisomes. Proc. Natl. Acad. Sci.
USA 105, 20567–20574 (2008).
27. Zhao, Z. et al. Autophagosome-independent essential function for the autophagy protein
Atg5 in cellular immunity to intracellular pathogens. Cell Host Microbe 4, 458–469
(2008).
28. Martens, S. et al. Disruption of Toxoplasma gondii parasitophorous vacuoles by the
mouse p47-resistance GTPases. PLoS Pathog. 1, e24 (2005).
29. Ling, Y.M. et al. Vacuolar and plasma membrane stripping and autophagic elimination
of Toxoplasma gondii in primed effector macrophages. J. Exp. Med. 203, 2063–2071
(2006).
30. Zhao, Y.O., Khaminets, A., Hunn, J.P. & Howard, J.C. Disruption of the Toxoplasma
gondii parasitophorous vacuole by IFNγ−inducible immunity-related GTPases (IRG
proteins) triggers necrotic cell death. PLoS Pathog. 5, e1000288 (2009).
31. Gutierrez, M.G. et al. Autophagy is a defense mechanism inhibiting BCG and
Mycobacterium tuberculosis survival in infected macrophages. Cell 119, 753–766
(2004).
32. Singh, S.B., Davis, A.S., Taylor, G.A. & Deretic, V. Human IRGM induces autophagy to
eliminate intracellular mycobacteria. Science 313, 1438–1441 (2006).
33. Harris, J. et al. T helper 2 cytokines inhibit autophagic control of intracellular
Mycobacterium tuberculosis. Immunity 27, 505–517 (2007).
34. Taylor, G.A., Feng, C.G. & Sher, A. p47 GTPases: regulators of immunity to intracellular
pathogens. Nat. Rev. Immunol. 4, 100–109 (2004).
35. Papic, N., Hunn, J.P., Pawlowski, N., Zerrahn, J. & Howard, J.C. Inactive and active
states of the interferon-inducible resistance GTPase, Irga6, in vivo. J. Biol. Chem. 283,
32143–32151 (2008).
36. Feng, C.G. et al. The immunity-related GTPase Irgm1 promotes the expansion of
activated CD4 + T cell populations by preventing interferon-γ-induced cell death. Nat.
Immunol. 9, 1279–1287 (2008).
37. Delgado, M. et al. Autophagy and pattern recognition receptors in innate immunity.
Immunol. Rev. 227, 189–202 (2009).
38. Xu, Y. et al. Toll-like receptor 4 is a sensor for autophagy associated with innate immunity.
Immunity 27, 135–144 (2007).
39. Delgado, M.A., Elmaoued, R.A., Davis, A.S., Kyei, G. & Deretic, V. Toll-like receptors
control autophagy. EMBO J. 27, 1110–1121 (2008).
40. Lee, H.K., Lund, J.M., Ramanathan, B., Mizushima, N. & Iwasaki, A. Autophagydependent
viral recognition by plasmacytoid dendritic cells. Science 315, 1398–1401
(2007).
41. Shi, C.S. & Kehrl, J.H. MyD88 and Trif target Beclin 1 to trigger autophagy in macrophages.
J. Biol. Chem. 283, 33175–33182 (2008).
42. Saitoh, T. et al. Loss of the autophagy protein Atg16L1 enhances endotoxin-induced
IL-1β production. Nature 456, 264–268 (2008).
43. Shui, w. et al. Membrane proteomics of phagosomes suggests a connection to
autophagy. Proc. Natl. Acad. Sci. USA 105, 16952–16957 (2008).
44. Jounai, N. et al. The Atg5 Atg12 conjugate associates with innate antiviral immune
responses. Proc. Natl. Acad. Sci. USA 104, 14050–14055 (2007).
45. Tal, M.C. et al. Absence of autophagy results in reactive oxygen species-dependent
amplification of RLR signaling. Proc. Natl. Acad. Sci. USA 106, 2770–2775
(2009).
46. Cadwell, K. et al. A key role for autophagy and the autophagy gene Atg16l1 in mouse
and human intestinal Paneth cells. Nature 456, 259–263 (2008).
47. Petrilli, V., Dostert, C., Muruve, D.A. & Tschopp, J. The inflammasome: a danger sensing
complex triggering innate immunity. Curr. Opin. Immunol. 19, 615–622 (2007).
48. Kanneganti, T.D., Lamkanfi, M. & Nunez, G. Intracellular NOD-like receptors in host
defense and disease. Immunity 27, 549–559 (2007).
nature immunology volume 10 number 5 may 2009 469

© 2009 Nature America, Inc. All rights reserved.
REVIEw
49. Lunemann, J.D. & Munz, C. Autophagy in CD4 + T-cell immunity and tolerance. Cell
Death Differ. 16, 79–86 (2009).
50. Vyas, J.M. van,d., V, & Ploegh,H.L. The known unknowns of antigen processing and
presentation. Nat. Rev. Immunol. 8, 607–618 (2008).
51. English, D. et al. Autophagy enhances the presentation of endogenous viral antigens
on MHC class I molecules during HSV-1 infection. Nat. Immunol. 10, 480–487
(2009).
52. Paludan, C. et al. Endogenous MHC class II processing of a viral nuclear antigen after
autophagy. Science 307, 593–596 (2005).
53. Schmid, D., Pypaert, M. & Munz, C. Antigen-loading compartments for major histocompatibility
complex class II molecules continuously receive input from autophagosomes.
Immunity 26, 79–92 (2007).
54. Mizushima, N., Yamamoto, A., Matsui, M., Yoshimori, T. & Ohsumi, Y. In vivo analysis
of autophagy in response to nutrient starvation using transgenic mice expressing a
fluorescent autophagosome marker. Mol. Biol. Cell 15, 1101–1111 (2004).
55. Nedjic, J., Aichinger, M., Emmerich, J., Mizushima, N. & Klein, L. Autophagy in thymic
epithelium shapes the T-cell repertoire and is essential for tolerance. Nature 455,
396–400 (2008).
56. Li, Y. et al. Efficient cross-presentation depends on autophagy in tumor cells. Cancer
Res. 68, 6889–6895 (2008).
57. Qu, X. et al. Autophagy gene-dependent clearance of apoptotic cells during embryonic
development. Cell 128, 931–946 (2007).
58. Mellen, M.A., de la Rosa, E.J. & Boya, P. The autophagic machinery is necessary for
removal of cell corpses from the developing retinal neuroepithelium. Cell Death Differ.
15, 1279–1290 (2008).
59. Arsov, I. et al. BAC-mediated transgenic expression of fluorescent autophagic protein
Beclin 1 reveals a role for Beclin 1 in lymphocyte development. Cell Death Differ. 15,
1385–1395 (2008).
60. Stephenson, L.M. et al. Identification of Atg5-dependent transcriptional changes and
increases in mitochondrial mass in Atg5-deficient T lymphocytes. Autophagy (in the
press).
61. Pua, H.H., Dzhagalov, I., Chuck, M., Mizushima, N. & He, Y.w. A critical role for the
autophagy gene Atg5 in T cell survival and proliferation. J. Exp. Med. 204, 25–31
(2007).
62. Pua, H.H., Guo, J., Komatsu, M. & He, Y.w. Autophagy is essential for mitochondrial
clearance in mature T lymphocytes. J. Immunol. 182, 4046–4055 (2009).
63. Miller, B.C. et al. The autophagy gene ATG5 plays an essential role in B lymphocyte
development. Autophagy 4, 309–314 (2007).
64. Li, C. et al. Autophagy is induced in CD4 + T cells and important for the growth factor-
withdrawal cell death. J. Immunol. 177, 5163–5168 (2006).
65. Espert, L. et al. Autophagy is involved in T cell death after binding of HIV-1 envelope
proteins to CXCR4. J. Clin. Invest. 116, 2161–2172 (2006).
66. Bell, B.D. et al. FADD and caspase-8 control the outcome of autophagic signaling in
proliferating T cells. Proc. Natl. Acad. Sci. USA 105, 16677–16682 (2008).
67. Arbour, N. et al. c-Jun NH 2 -terminal kinase (JNK)1 and JNK2 signaling pathways have
divergent roles in CD8 + T cell-mediated antiviral immunity. J. Exp. Med. 195, 801–810
(2002).
68. Conze, D. et al. c-Jun NH 2 -terminal kinase (JNK)1 and JNK2 have distinct roles in
CD8 + T cell activation. J. Exp. Med. 195, 811–823 (2002).
69. Dong, C. et al. Defective T cell differentiation in the absence of Jnk1. Science 282,
2092–2095 (1998).
70. wei, Y., Pattingre, S., Sinha, S., Bassik, M. & Levine, B. JNK1-mediated phosphorylation
of Bcl-2 regulates starvation-induced autophagy. Mol. Cell 30, 678–688
(2008).
71. Barrett, J.C. et al. Genome-wide association defines more than 30 distinct susceptibility
loci for Crohn’s disease. Nat. Genet. 40, 955–962 (2008).
72. Massey, D.C. & Parkes, M. Genome-wide association scanning highlights two autophagy
genes, ATG16L1 and IRGM, as being significantly associated with Crohn’s disease.
Autophagy 3, 649–651 (2007).
73. The wellcome Trust Case Control Consortium. Genome-wide association study of
14,000 cases of seven common diseases and 3,000 shared controls. Nature 447,
661–678 (2007).
74. McCarroll, S.A. et al. Deletion polymorphism upstream of IRGM associated with altered
IRGM expression and Crohn’s disease. Nat. Genet. 40, 1107–1112 (2008).
75. Rioux, J.D. et al. Genome-wide association study identifies new susceptibility loci
for Crohn disease and implicates autophagy in disease pathogenesis. Nat. Genet. 39,
596–604 (2007).
76. Kuballa, P., Huett, A., Rioux, J.D., Daly, M.J. & Xavier, R.J. Impaired autophagy of an
intracellular pathogen induced by a Crohn’s disease associated ATG16L1 variant. PLoS
ONE 3, e3391 (2008).
77. Vaishnava, S., Behrendt, C.L., Ismail, A.S., Eckmann, L. & Hooper, L.V. Paneth cells
directly sense gut commensals and maintain homeostasis at the intestinal host-microbial
interface. Proc. Natl. Acad. Sci. USA 105, 20858–20863 (2008).
78. Kaser, A. et al. XBP1 links ER stress to intestinal inflammation and confers genetic
risk for human inflammatory bowel disease. Cell 134, 743–756 (2008).
79. Cadwell, K., Patel, K.K., Komatsu, M., Virgin, H.w. & Stappenbeck, T.S. A common
role for Atg16L1, Atg5, and Atg7 in small intestinal Paneth cells and Crohn’s disease.
Autophagy 5, 250–252 (2008).
470 volume 10 number 5 may 2009 nature immunology

© 2009 Nature America, Inc. All rights reserved.
A Crohn’s disease–associated NOD2 mutation
suppresses transcription of human IL10 by inhibiting
activity of the nuclear ribonucleoprotein hnRNP-A1
Eiichiro Noguchi 1 , Yoichiro Homma 1 , Xiaoyan Kang 2 , Mihai G Netea 3 & Xiaojing Ma 2
A common mutation in the gene encoding the cytoplasmic sensor Nod2, involving a frameshift insertion at nucleotide 3020
(3020insC), is strongly associated with Crohn’s disease. How 3020insC contributes to this disease is a controversial issue.
Clinical studies have identified defective production of interleukin 10 (IL-10) in patients with Crohn’s disease who bear the
3020insC mutation, which suggests that 3020insC may be a loss-of-function mutation. However, here we found that 3020insC
Nod2 mutant protein actively inhibited IL10 transcription. The 3020insC Nod2 mutant suppressed IL10 transcription by
blocking phosphorylation of the nuclear ribonucleoprotein hnRNP-A1 via the mitogen-activated protein kinase p38. We
confirmed impairment in phosphorylation of hnRNP-A1 and binding of hnRNP-A1 to the IL10 locus in peripheral blood
mononuclear cells from patients with Crohn’s disease who bear the 3020insC mutation and have lower production of IL-10.
Crohn’s disease is a chronic inflammatory bowel disease that affects
mainly the terminal ileum, cecum, colon and perianal area. Inflammatory
bowel disease is thought to result from inappropriate and
continuing activation of the mucosal immune system driven by
normal luminal flora. This aberrant response is probably facilitated
by defects in both the barrier function of the intestinal epithelium and
the mucosal immune system. These defects are in turn caused by
genetic and environmental factors 1 .
Interleukin 10 (IL-10; A001243) is a pleiotropic cytokine produced
by T cells, B cells and macrophages. IL-10-deficient mice show
enhanced development of several inflammatory and autoimmune
diseases, which suggests that IL-10 serves a central function in vivo
in restricting inflammatory responses 2 . Although they are healthy in
germ-free conditions, some IL-10-deficient mouse strains develop
colitis when given a normal, specific pathogen–free bowel flora,
probably as a result of a poorly controlled mucosal immune response.
Innate immune responses are initiated by the detection of microbial
invaders by several distinct host systems collectively called ‘patternrecognition
receptors’. The nucleotide-binding oligomerization
domain (Nod)-like receptors, Toll-like receptors (TLRs), C-type lectin
receptors and RIG-I-like receptors are examples of pattern-recognition
receptors. Two members of the Nod-like receptor family of proteins,
Nod1 and Nod2, are believed to be cytoplasmic sensors of microbial
products 3 . Both Nod1 and Nod2 recognize peptidoglycan, but each
protein senses distinct molecular motifs in peptidoglycan. Nod1
recognizes a naturally occurring muropeptide of peptidoglycan that
presents a unique amino acid at its terminus called ‘diaminopimelic
Received 26 January; accepted 24 February; published online 6 April 2009; doi:10.1038/ni.1722
ARTICLES
acid’. This amino acid is found mainly in the peptidoglycan of
Gram-negative bacteria 4 . In contrast, Nod2 can detect the minimal
bioactive fragment of peptidoglycan called ‘muramyl dipeptide’. Thus,
whereas Nod1 detects mainly Gram-negative bacteria, Nod2 is a more
general sensor of bacterial peptidoglycan 5 .
Nod2 expression is restricted to monocytes 6 andisrecognizedasan
important initiator of inflammation. Models hypothesize that after
binding to muramyl dipeptide, Nod2 forms oligomers and binds to the
caspase-recruitment domain–containing serine-threonine kinase RICK
(also called RIP2, CARDIAK, CCK and Ripk2). RICK then forms
oligomers and facilitates the ubiquitination of a site on the modulator
of transcription factor NF-kB NEMO (also called IKKg) 7 ,whichresults
in activation of the inhibitor of NF-kB kinase complex and NF-kB 8 .
The gene encoding Nod2 on human chromosome 16 (NOD2) was
the first gene identified as being associated with susceptibility to
Crohn’s disease 9 . Between 30% and 50% of patients with Crohn’s
disease in the Western hemisphere carry NOD2 mutations on at least
one allele. Mutations resulting in substitution of tryptophan for
arginine at amino acid residue 702 (R702W) or arginine for glycine
at amino acid residue 908 (G908R) and a frameshift substitution at
amino acid position 1007 are estimated to represent 32%, 18% and
31%, respectively, of all disease-causing mutations in NOD2 and are
independently associated with susceptibility to Crohn’s disease 10 .All
three mutations are located in the leucine-rich repeat (LRR) domain
of NOD2. The frameshift substitution at amino acid 1007 associated
with Crohn’s disease stems from an insertion mutation at nucleotide
3020; this mutation (3020insC) results in partial deletion of the
1 Department of Surgery II, Tokyo Women’s Medical University, Tokyo, Japan. 2 Department of Microbiology and Immunology, Weill Medical College of Cornell University,
New York, New York, USA. 3 Department of Medicine (463), Radboud University Nijmegen Medical Center, Nijmegen, The Netherlands. Correspondence should be sent to
X.M. (xim2002@med.cornell.edu).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 471

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a b c d e
Mouse IL-10 (ng/ml)
3
2
1
0
WT
Nod2-KO
Medium
LPS
PGN
MDP
Pam3Cys Mouse IL-12p40 (ng/ml)
12
10
8
6
4
2
0
WT
Nod2-KO
Medium
LPS
PGN
MDP
Pam3Cys terminal LRR domain of the protein. The genotype relative risk for
developing Crohn’s disease in people heterozygous or homozygous for
this mutation alone is calculated to be 1.5 2.6 or 17.6 42.1,
respectively 11 . Overall, patients homozygous for 3020insC demonstrate
a much more severe disease phenotype than other patients with
Crohn’s disease and have a higher risk for ileal stenosis and surgical
intervention 12 . It is postulated that the LRR domain of Crohn’s
disease–associated Nod variants is impaired in its ability to recognize
microbial components and/or in its ability to inhibit the formation of
Nod2 dimers, which results in lack of activation of NF-kB in
monocytes 9 . However, the higher NF-kB activity in patients with
Crohn’s disease and the clinical effectiveness of NF-kB inhibitors in
ameliorating symptoms of Crohn’s disease 1 suggest that diseasecausing
variants of Nod2 may promote activation of NF-kB.
IL-10-deficient mice develop a chronic enterocolitis that shares
histopathological features with human Crohn’s disease 13 .Colitisin
IL-10-deficient mice is driven by T helper type 1 (TH1) cells but not by
B cells 14 and is dependent on the presence of resident enteric
bacteria 15 . In accordance with those findings, Helicobacter hepaticus
triggers colitis in specific pathogen–free IL-10-deficient mice by
a mechanism that depends on T H1 proinflammatory cytokines 16 .
The relationship between Nod2 and IL-10 is not completely understood.
Peripheral blood mononuclear cells (PBMCs) from patients
with Crohn’s disease who are homozygous for the 3020insC mutation
show defective release of IL-10 after stimulation with enteric microorganisms
and TLR ligands 17 . Those findings contrast with studies of
mice in which an artificially created mouse version (2939insC) of
human 3020insC replaces the endogenous wild-type mouse Nod2 gene
(Nod2 2939insC ) 18 . In these mice, macrophages stimulated by the TLR4
ligand lipopolysaccharide (LPS), peptidoglycan or muramyl dipeptide
release quantities of IL-10 resembling those released by wild-type
macrophages. We hypothesized that the human 3020insC and mouse
Nod2 2939insC alleles may have different functional properties because of
species-specific physiology or evolutionary adaptation. We undertook
this study to explore that possibility and the potential target(s) of the
mutant Nod2 protein produced by the 3020insC mutation (called
‘3020insC Nod2’ here).
RESULTS
Suppression of IL10 production by 3020insC Nod2
To investigate the influence of endogenous Nod2 on cytokine production,
we measured the production of IL-10 and IL-12p40 by mouse
bone marrow–derived macrophages (BMDMs) from wild-type and
Mouse IL-10 (ng/ml)
2.0
1.5
1.0
0.5
0
WT
RIP2-KO
Medium
LPS
PGN
Pam3Cys Human IL-10 (pg/ml)
900
Nod2 / mice 19 . We detected little difference in the amount of IL-10
(Fig. 1a) orIL-12p40(Fig. 1b) released by wild-type and Nod2 /
BMDMs after stimulation with LPS, peptidoglycan or Pam 3Cys (a
synthetic TLR2 ligand). Muramyl dipeptide alone stimulated little
production of IL-10 or IL-12p40 in this experiment, consistent with
results obtained with human PBMCs, in which muramyl dipeptide
augments only TLR2 ligand–mediated induction of IL-12 and IL-10
(ref. 20). The patterns of production of IL-10 and IL-12p40 by
peritoneal macrophages and splenocytes were very similar to those
in BMDMs (data not shown). Furthermore, RICK (RIP2) was also
dispensable for IL-10 production in primary macrophages stimulated
by LPS, peptidoglycan or Pam 3Cys (Fig. 1c).
Next we tested the response of primary monocytes to muramyl
dipeptide, Pam3Cys or a combination of muramyl dipeptide and
Pam 3Cys in cells from patients with Crohn’s disease who were
homozygous for the 3020insC mutation, as well as cells from control
volunteers bearing the wild-type NOD2 allele. In contrast to the data
obtained with mice, IL-10 production was much lower in cells from
patients homozygous for 3020insC (Fig. 1d). Production of tumor
necrosis factor was also significantly lower in cells from patients
homozygous for 3020insC after stimulation with muramyl dipeptide
alone but was similar with that of wild-type cells after stimulation with
Pam 3Cys or muramyl dipeptide plus Pam 3Cys (Fig. 1e).
Thus, we hypothesized that the 3020insC Nod2 mutant may be able
to regulate expression of the gene encoding IL-10 (IL10). To determine
the effect of 3020insC Nod2 on endogenous IL10 expression, we
transduced primary human monocytes with lentivirus expressing
wild-type or 3020insC Nod2 (Fig. 2a). We found that 3020insC
Nod2 but not wild-type Nod2 inhibited both basal as well as LPSinduced
expression of IL-10 mRNA and protein (Fig. 2b,c). In
contrast, we found no difference in IL-1b production by cells overexpressing
wild-type or 3020insC Nod2 (Fig. 2d), and IL-12p40
production was suppressed by high expression of wild-type Nod2
but not 3020insC Nod2 (Fig. 2e). However, it is possible that the lack
of influence of 3020insC Nod2 on IL-12p40 production was secondary
to its suppression of IL-10 production. Thus, wild-type and 3020insC
Nod2 seem to affect IL-10, IL-1b and IL-12p40 expression differently
in human monocytes. We obtained similar results when we used
peptidoglycan and Pam 3Cys instead of LPS (data not shown).
To confirm our finding of an influence of Nod2 on IL10 transcription,
we transfected primary human monocytes with a human IL10
promoter–luciferase reporter construct (containing the sequence
between positions 1044 and +30 relative to the transcription start
Human TNF (pg/ml)
400
600
300
0 *
*
200
0 *
Medium
WT Nod2
3020insC Nod2
MDP
Pam3Cys MDP +
Pam
Medium
WT Nod2
3020insC Nod2
Figure 1 Influence of Nod2 signaling on macrophage cytokine production. (a–c) ELISA of mouse IL-10 (a,c) and IL-12p40 (b) in supernatants of wild-type
(WT), Nod2 –/– (Nod2-KO) and Ripk2 –/– (RIP2-deficient; RIP2-KO) mouse BMDMs stimulated for 24 h in vitro with medium, LPS (1 mg/ml), peptidoglycan
(PGN; 10 mg/ml), muramyl dipeptide (MDP; 10 mg/ml) or Pam 3Cys (1 mg/ml); results are normalized to total cellular protein. Data represent three
independent experiments (error bars, s.e.m.). (d,e) ELISA of human IL-10 (d) and tumor necrosis factor (e) in supernatants of monocytes obtained from
healthy people expressing wild-type Nod2 (n ¼ 9) and patients with Crohn’s disease who were homozygous for 3020insC (n ¼ 5) and then stimulated for
24 h with medium, muramyl dipeptide, Pam 3Cys or a combination of muramyl dipeptide and Pam 3Cys (MDP + Pam); results are normalized to total cellular
protein. *, P o 0.05. Data are representative of one experiment with each donor sample measured in triplicate (error bars, s.d.).
472 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
MDP
Pam3Cys MDP +
Pam

© 2009 Nature America, Inc. All rights reserved.
a b
IL-10 (ng/ml)
IL-12p40 (ng/ml)
Medium
WT CARD CARD Nod
1,040
LRR
IL-10
3020insC CARD CARD Nod
CV WT Nod2
c
3020insC Nod2
4
3
LRR
1,007
WT or
3020insC Nod2
β-actin
CV WT Nod2
3020insC Nod2
d 4
2
1 * *
3
2
1
0
Supernatant
(µl)
5 50
Medium
5 50
LPS
0
Supernatant
(µl)
5 50
Medium
5 50
LPS
e 10
CV WT Nod2
3020insC Nod2 f 1.4
8
6
4
2
*
1
0.6
0.2
*
0
Supernatant
(µl)
5 50
Medium
5 50
LPS
pCDNA3 WT 3020insC
Nod2 Nod2
site) 21 together with vectors expressing wild-type or 3020insC Nod2.
Neither the empty vector (pCDNA3) nor wild-type Nod2 altered
constitutive transcription of IL10; however, 3020insC Nod2 suppressed
constitutive IL10 transcription by about 50% (Fig. 2f). These results
collectively demonstrate that the 3020insC mutation conferred on
Nod2 a new function, the ability to inhibit IL-10 production, and loss
of the ability to repress IL-12p40 expression.
Inhibition of IL10 promoter activity by 3020insC
To further explore the molecular mechanism whereby 3020insC Nod2
inhibited IL10 expression, we transfected the IL10 promoter–luciferase
reporter and the wild-type or 3020insC Nod2 construct into the
RAW264.7 mouse macrophage cell line. We verified expression of
each construct by RT-PCR (Supplementary Fig. 1 online). Although
the IL10 promoter was constitutively active, peptidoglycan and
Pam 3Cys further boosted luciferase activity (Fig. 3a). Expression
of 3020insC Nod2 strongly inhibited IL10 transcription in all conditions,
even in the absence of any microbial stimuli, but wild-type
Nod2 did not.
To determine if wild-type and 3020insC Nod2 could functionally
compete (for example, in the setting of a heterozygous person), we
transfected wild-type and 3020insC Nod2 together, at various
ratios, into RAW264.7 cells, along with the IL10 promoter–luciferase
reporter. When wild-type Nod2 was more abundant than 3020insC
IL-1β (ng/ml)
Stimulation index
LPS
CV
WT
3020insC
CV
WT
3020insC
Figure 3 Inhibition of IL10 transcription by 3020insC Nod2. (a) Luciferase
activity of lysates of RAW264.7 cells transfected for 1 d with the human
IL10 promoter–lucifersase reporter together with empty vector (pCDNA3) or
vector encoding wild-type or 3020insC Nod2 at an effector/reporter molar
ratio of 0.3:1, then stimulated for 24 h with medium (Med), peptidoglycan
or Pam 3Cys. RLU, raw luciferase units. Expression of the transfected Nod2
and 3020insC mutant in RAW264.7 cells was verified by RT-PCR with
primers selective for human but not mouse Nod2 (Supplementary Fig. 1).
(b,c) Luciferase activity in lysates of cells transfected for 1 d with the
human IL10 promoter–luciferase reporter, together with various molar ratios
(horizontal axes) of vectors encoding wild-type and 3020insC Nod2 (b) or
wild-type Nod2 and empty vector (pCDNA3; c), then stimulated for 7 h with
medium or peptidoglycan. Reporter amount remained constant, set as 1.
(d) Immunoassay of extracts of HEK293T cells transfected for 24 h (Input)
with Flag- or Myc-tagged wild-type Nod2 (WT) or 3020insC Nod2 (frameshift
(FS)), then immunoprecipitated (IP) with anti-Flag (a-Flag) or anti-c-Myc
(a-Myc) and analyzed by immunoblot (IB) with anti-Flag. Arrows indicate that
3020insC Nod2 is smaller than wild-type Nod2. Data represent the pooled
results of at least three independent experiments (error bars (a–c), s.d.).
Nod2, it countered the inhibitory effect of 3020insC Nod2 on IL10
transcription; this counter-inhibitory effect was lost as 3020insC Nod2
increased in abundance (Fig. 3b,c). Competition between wild-type
and 3020insC Nod2 could have been due to direct interaction between
the two molecules. To investigate that possibility, we generated wildtype
and 3020insC Nod2 constructs tagged with Flag and c-Myc
epitopes. After being transfected into human embryonic kidney
(HEK293) cells, wild-type and 3020insC Nod2 constructs engaged in
homotypic and heterotypic interactions (Fig. 3d). Thus, wild-type
Nod2 may inhibit 3020insC Nod2–mediated suppression of IL10
transcription through direct physical interference.
Because only a fraction of patients with Crohn’s disease are
homozygous carriers of the 3020insC mutation, we also analyzed the
effect of the other two common Nod2 variants (R702W and G908R)
on IL10 expression. Like the 3020insC Nod2 mutant, the R702W
and G908R Nod2 mutants exerted an inhibitory effect on IL10
promoter activity (data not shown). In summary, these data suggest
that all three LRR mutants of Nod2 are able to inhibit basal and
TLR-induced IL10 transcription.
Gain of function is unique to human Nod2 and IL10
AstudyofNod2 2939insC mice (in which an artificially created mouse
version of human 3020insC replaces endogenous wild-type mouse
Nod2) did not detect alteration in IL-10 expression in macrophages 18 .
a pCDNA3
b
Luciferase activity
(10 3 RLU/s)
c
Luciferase activity
(10 2 RLU/s)
18
14
10
6
2
24
16
8
0
WT Nod2
3020insC Nod2
*
Med
Med
PGN
4:0 3:1 2:2 1:3 0:4
WT:pCDNA3
*
*
PGN Pam 3 Cys
Luciferase activity
(10 2 RLU/s)
d
16
12
8
4
Med
PGN
ARTICLES
Figure 2 Different effects of wild-type and 3020insC Nod2 on endogenous
IL-10 expression in primary human monocytes. (a) Wild-type and 3020insC
Nod2 constructs. CARD, caspase-recruitment domain. (b) RT-PCR analysis
of the expression of mRNA encoding IL-10, Nod2 and b-actin by primary
human monocytes transduced for 4 d with empty control vector (CV) or
lentivirus encoding wild-type or 3020insC Nod2 and then stimulated for 5 d
with medium or LPS. On the basis of lentiviral expression of green fluorescent
protein, an average infection rate of 38% was achieved. ‘Background’
bands represent endogenous Nod2 mRNA in infected cells. (c–e) ELISAof
the secretion of IL-10 (c), IL-1b (d) and IL-12p40 (e) by human monocytes
infected with 5 or 50 ml lentivirus-containing supernatant (key) and treated
with medium or LPS, analyzed on day 5 after infection and normalized to
cytokine produced (pg/ml) per 1 10 5 live cells. *, P o 0.03. (f) Luciferase
activity in lysates of primary human monocytes transfected for 6 h with a
human IL10 promoter–luciferase reporter, together with empty vector
(pCDNA3) or vector encoding wild-type or 3020insC Nod2 constructs
(optimized response); results are presented as the stimulation index, derived
from the ratio of the luciferase activity in each experimental condition to that
of cells transfected with empty vector. Mean transfection efficiency, 35%.
Data are representative of three experiments with one donor each (error
bars (c–f), s.d.).
0
4:0 3:1 2:2 1:3 0:4
WT:3020insC
IP: α-Flag α-Myc
Flag WT WT FS FS WT WT FS FS
Input
Myc WT FS WT FS WT FS WT FS
IB:
α-FLAG
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 473

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a b
Luciferase activity (10 3 RLU/s)
c d
Mouse IL-10 (pg/ml)
Human IL10 reporter Mouse Il10 reporter
30
25
120
100
CV
Human WT Nod2
3020insC Nod2
20
80
15
60
10
40
5
20
0
0
30
25
20
15
10
5
140
120
100
80
60
40
20
CV
Mouse WT Nod2
2939insC Nod2
0
Med MDP LPS PGN
0
Med MDP LPS PGN
1,400
CV
1,200
WT Nod2
1,000
3020insC
800
Nod2
600
400
200
0
Med LPS PGN Pam3Cys Luciferase activity (10 3 RLU/s)
Luciferase activity
(10 3 RLU/s)
That is inconsistent with our experimental data presented above and
with clinical observations of human patients with Crohn’s disease 17,20 .
We hypothesized that the human and mouse Nod2 mutants may have
different functional properties, as the latter was created artificially and
thus was not subject to evolutionary adaptation in mammalian
physiology. We tested our hypothesis by comparing the effects of the
human mutant (3020insC Nod2) and mouse mutant (2939insC Nod2)
on the expression of a human IL10 promoter–luciferase reporter
construct or mouse Il10 promoter–luciferase reporter construct.
Human 3020insC Nod2 suppressed basal and TLR-induced expression
of the human reporter but not of the mouse reporter (Fig. 4a,b). In
contrast, mouse 2939insC Nod2 did not inhibit expression of either
the human or mouse reporter (Fig. 4a,b). Furthermore, retrovirusmediated
expression of 3020insC Nod2 into BMDMs derived from
Nod2 / mice had little effect on either basal or TLR-stimulated IL-10
production (Fig. 4c). These results indicate that the human and
mouse Nod2 mutants have different functional properties and that
only the human Nod2 mutant can inhibit expression of human IL-10.
Blockade of hnRNP-A1–IL10 binding by 3020insC Nod2
Next we searched for IL10 promoter elements (3020insC Nod2–
response elements (NREs)) required for 3020insC Nod2–mediated
inhibition of IL-10 production. Using an extensive and systematic
3
2
1
0
Med
LPS
WT IL10
WT Nod2
3020insC Nod2
PGN
Med
LPS
PGN
Mut IL10
Figure 5 Binding of nuclear proteins to the NRE. (a) Immunoassay of
lysates of RAW264.7 cells stably transfected with empty vector or vector
encoding wild-type or 3020insC Nod2, then immunoprecipitated and
analyzed by immunoblot with anti-Nod2. Results are representative of more
than five separate experiments. (b) Sequences of probes used for EMSA. The
mutant NRE probe bears the same mutations as those between positions
30 and 25 in the IL10 promoter mutant construct in Figure 4d.
(c) EMSA (with the probes in b) of nuclear extracts of the stable RAW264.7
transfectants in a, stimulated with medium, LPS or peptidoglycan. FP, free
probe. Results are representative of four separate experiments. (d) DNA
precipitation of nuclear extracts of the stable RAW264.7 transfectants in a
with biotinylated oligonucleotides encompassing the IL10 NRE, followed by
elution at increasing NaCl concentrations (above gel), separation by 12%
SDS-PAGE and silver staining. Arrows indicated bands that were excised and
analyzed. M, molecular size markers (in kilodaltons (kDa)). Results are
representative of three independent experiments with identical results.
Figure 4 Human and mouse Nod2 mutant proteins have different
transcriptional effects. (a,b) Luciferase activity in lysates of RAW264.7 cells
transfected for 1 d with a human IL10 promoter–luciferase reporter (a) ora
mouse Il10 promoter–luciferase reporter (b) together with empty control
vector or vector expressing human wild-type or 3020insC Nod2 (top) or
mouse wild-type or 2939insC Nod2 (bottom) at a reporter/effector molar
ratio of 1:1, then stimulated for 24 h with medium, muramyl dipeptide, LPS
or peptidoglycan. *, P o 0.03. Data represent three to four independent
experiments (mean and s.d.). (c) ELISA of mouse IL-10 in supernatants
of Nod2 –/– BMDMs infected twice with empty control vector (CV) or
retrovirus encoding wild-type or 3020insC Nod2 and then, on day 5 after
infection, stimulated for 24 h with medium, LPS, peptidoglycan or
Pam 3Cys. Equivalent expression of transduced human wild-type and
3020insC Nod2 mRNA in unstimulated Nod2 –/– BMDMs was ensured
(Supplementary Fig. 3 online). Data are representative of three experiments
with one mouse each (error bars, s.d. of triplicate samples). (d) Luciferase
activity in lysates of RAW264.7 cells transiently transfected with constructs
encoding wild-type or 3020insC Nod2 (key) plus luciferase reporter
constructs driven by the IL10 promoter fragment (positions 782 to +12)
with (Mut IL10) or without (WT IL10) point mutations in the region between
positions 30 and 25 and then stimulated for 7 h with medium, LPS
or peptidoglycan. Data are representative of four individual experiments
(mean and s.d.).
deletion- and mutagenesis-based search strategy, we focused on a
narrow region of the IL10 promoter between positions 30 and 25
upstream of the transcription-initiation site. A version containing point
mutations introduced into the region between positions 30 and 25
was not inhibited by 3020insC Nod2, in contrast to the inhibition of a
construct containing a wild-type IL10 promoter fragment (Fig. 4d).
Notably, the mutant also lost a substantial portion of its basal
transcriptional activity. These results suggest that the TACACA
sequence in the region between positions 30 and 25 may form
the core of an NRE and that a nuclear DNA-binding protein may
constitutively interact with this region. It is noteworthy that the NRE is
predicted to be a positive functional element because its mutation
resulted in less basal and microbe-stimulated IL10 transcription.
We hypothesized that the NRE may bind to transcription factors
involved in IL10 transcription and that this binding may be regulated
by 3020insC Nod2. To test our hypothesis, we established stable clones
of RAW264.7 cells constitutively expressing wild-type or 3020insC
Nod2 (Fig. 5a). We stimulated these clones with LPS or peptidoglycan,
then isolated nuclear extracts and assessed by electrophoretic
mobility-shift assay (EMSA) their binding to a probe containing the
wild-type or mutated human IL10 NRE (Fig. 5b,c). We detected a
nuclear factor bound to the wild-type but not mutant probe in
unstimulated cells; this binding did not increase after stimulation
a CV WT 3020insC b
IP: α-NOD2
IB: α-NOD2
c d
Effector: FP
TF-X
WT NRE Mutant NRE
Med LPS PGN Med
FP
CV
WT3020insC
CV
WT
3020insC
CV
WT
3020insC
CV
WT
3020insC
NaCl (M)
Human WT NRE (–36 to –19)
5′– AAGGTCTACACATCAGGG–3′
Human mutant NRE
5′– AAGGTCCGTGTGTCAGGG–3′
0.1
(kDa)
250 M CV
WT
150
100
75
50
37
25
0.2 0.4
3020insC
WT
3020insC
WT
3020insC
474 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a b c d e
Luciferase activity (10 3 RLU/s)
7
6
5
4
3
*
*
*
10
8
6
4
*
*
*
*
2.0
1.5
1.0
*
2
1
2
0.5
0
1:0 1:0.25 1:0.5 1:0.75 1:1
0
1:0 1:0.25 1:0.5 1:0.75 1:1
0
WT NRE WT NRE Mut NRE
Reporter:effector Reporter:effector
pCDNA3 hnRNPA1
Luciferase activity (10 3 RLU/s)
with LPS or peptidoglycan (Fig. 5c). We designated this binding
factor, which may represent a putative transcription factor involved in
maintaining basal IL10 transcription, ‘TF-X’. Consistent with our
prediction, 3020insC Nod2 resulted in much less binding of TF-X to
the NRE, but wild-type Nod2 did not (Fig. 5c).
We then did a series of biochemical experiments to identify TF-X.
First we used biotinylated oligonucleotides encoding the NRE to
precipitate TF-X from nuclear extracts derived from RAW264.7 clones
expressing wild-type or 3020insC Nod2 (Fig. 5d). Next we excised the
approximately 26-kilodalton TF-X band and analyzed it by nanoflow
liquid chromatography–tandem mass spectrometry. The result with
the highest score (30.98) for bands excised from cells expressing wildtype
Nod2 was heterogeneous nuclear ribonucleoprotein A1 (hnRNP-
A1; A001137); these hnRNP-A1 peptides (1,218.64 and 1,784.91
daltons in size, with intensities of 1.89 10 6 and 5.57 10 5 ,
respectively), which covered 8% of the amino acid residues of the
full-length protein, were undetectable for the bands excised from
extracts of cells expressing 3020insC Nod2.
Activation of IL10 transcription by hnRNP-A1
The hnRNPs are among the most abundant proteins in the eukaryotic
cell nucleus and are directly involved in DNA repair and telomere
elongation; chromatin remodeling and transcription; RNA splicing
and stability; export of mature RNA; and translation. Approximately
30 hnRNPs have been identified. Autoantibodies specific for A and B
proteins of the hnRNP complex have been detected in rheumatoid
arthritis, systemic lupus erythematosus and mixed connective tissue
disease 22,23 . Moreover, hnRNP-A1 is also found as a potential autoantigen
in 38% of psoriasis patients 24 .
To analyze the influence of hnRNP-A1 on IL10 transcription, we
transfected a mouse hnRNP-A1 expression vector (effector) together
with the IL10 promoter–luciferase reporter into RAW264.7 cells. We
found that hnRNP-A1 stimulated Il10 promoter activity in a dosedependent
way at fairly low ratios of reporter to effector (Fig. 6a).
Moreover, hnRNP-A1 also augmented LPS- and peptidoglycan-stimulated
IL10 promoter–luciferase reporter activity (data not shown).
Notably, mouse hnRNP-A1 also stimulated a mouse Il10 promoter–
luciferase reporter construct (Fig. 6b). However, hnRNP-A1 had little
effect on luciferase reporters driven by human IL-12p40 or ‘generic’
NF-kB promoters, even at high ratios of reporter to effector (data not
shown). Furthermore, the stimulatory effect of human hnRNP-A1
Stimulation index
3
2
IL-10
IL-12p40
*
700
500
IL-10 *
IL-12p40
1
300
100
0
0
hnRNP-A1
hnRNP-A1
β-actin
β-actin
siRNA Mock 4 1-3 Vector pCDNA3 hnRNP-A1
Figure 6 Stimulation of IL10 transcription by hnRNP-A1. (a,b) Luciferase activity of RAW264.7 cells transfected for 48 h with a human (a) ormouse
(b) IL-10 reporter together with pCDNA3 and/or mouse hnRNP-A1 at various reporter/effector ratios (horizontal axes). (c) Luciferase activity of RAW264.7
cells transfected with reporter constructs driven by a wild-type (WT NRE) or mutant (Mut NRE) IL10 NRE, together with empty vector (pcDNA3) or vector
encoding mouse hnRNP-A1, presented relative to that in pCDNA3-transfected cells. (d) ELISA of IL-10 and IL-12p40 in supernatants of primary human
monocytes transfected with siRNA specific for hnRNP-A1 (4 and 1–3) or mock transfected (Mock; d) or transfected with with empty vector (pCDNA3) or
expression vector encoding hnRNP-A1 (e), then, 6 h after transfection, stimulated for 12 h with LPS. Below, RT-PCR analysis of the expression of hnRNP-A1
and b-actin mRNA demonstrates the efficiency of siRNA-mediated knockdown (d) or vector transfection (e). *, P o 0.05. Data represent from three to five
independent experiments (mean and s.d.).
Cytokine (ng/ml)
was mediated through the NRE in the IL10 promoter, as hnRNP-A1
had no effect on the NRE-mutant IL10 promoter–luciferase
reporter (Fig. 6c).
To demonstrate the physiological function of hnRNP-A1 in the
regulation of endogenous IL-10 production in primary cells, we transfected
human peripheral blood–derived monocytes with small interfering
RNA (siRNA) specific for human hnRNP-A1. A pool of three
siRNA molecules that diminished expression of hnRNP-A1 inhibited
LPS-stimulated IL-10 production by about 45%; a fourth hnRNP-A1specific
siRNA resembled the mock siRNA, as it failed to suppress
hnRNP-A1 and did not influence IL-10 production (Fig. 6d). These
siRNAs had no effect on LPS-induced production of IL-12p40. The
degree of inhibition of IL-10 production by hnRNP-A1-specific siRNA
was notable, given that fewer than 40% of the monocytes were transfected
with the siRNA (data not shown) and that hnRNP-A1 expression
was only partially ‘knocked down’. Conversely, overexpression of
hnRNP-A1 in these cells strongly enhanced LPS-stimulated production
of IL-10 but had no effect on IL-12p40 secretion (Fig. 6e). These results
suggest that hnRNP-A1 promotes transcription of human IL10.
Suppression of hnRNP-A1–NRE binding by 3020insC Nod2
To confirm the identification of hnRNP-A1 as an NRE-binding
nuclear protein by mass spectrometry, we used EMSA and chromatin
immunoprecipitation (ChIP). As shown by EMSA, overexpression of
hnRNP-A1 in RAW264.7 cells resulted in binding of protein to the
human NRE probe; this binding activity was diminished by an
antibody to hnRNP-A1 but not by an isotype-matched control antibody
(Fig. 7a). ChIP analysis of primary human monocytes showed
strong and specific binding of hnRNP-A1 in the region between
positions 121 and +42 of IL10, which contains the NRE, but not
in an irrelevant upstream region between positions 3158 and 2947
(Fig. 7b). The binding seemed to be constitutive and was not
regulated by stimulation of primary human monocytes with LPS
(Fig. 7c). These data indicate that hnRNP-A1 interacts with the IL10
NRE in human monocytes.
Next we analyzed PBMCs from three patients with Crohn’s
disease who were homozygous for 3020insC and had profoundly
diminished IL-10 production 25 . The binding of hnRNP-A1 to IL10
around the NRE was much lower in these patients than in age- and
sex-matched healthy people and patients with Crohn’s disease
who lacked the 3020insC mutation (Fig. 7d). The diminished
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 475
Cytokine (pg/ml)
ARTICLES

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a c
b
FP
–121 to +42
–3158 to –2947
pCDNA3
hnRNP-A1
Control IgG
α-hnRNP-A1
hnRNP-A1
hnRNP-A1
e
hnRNP-A1 binding (relative)
Input
Control lg
α-hnRNP
d
Input Rat IgG α-hnRNP
Med LPS Med LPS Med LPS
Input Rat IgG α-hnRNP
CD CD CD
Ctrl WT FS Ctrl WT FS Ctrl WT FS
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
Ctrl WT FS Ctrl WT FS Ctrl WT FS
CD
CD
CD
Input rlgG α-hnRNP
hnRNP-A1 binding in the patients with Crohn’s disease who
were homozygous for 3020insC was even more prominent in
experiments using real-time quantitative PCR (Fig. 7e).
We further investigated the molecular basis for the impaired
nuclear binding activity of hnRNP-A1 in patients with Crohn’s
disease who were homozygous for 3020insC. For this, we overexpressed
Flag-tagged wild-type or 3020insC Nod2 and Flag-tagged
hnRNP-A1 in HEK293 cells. Wild-type and 3020insC Nod2 were
expressed in the cytoplasm but not in the nucleus, whereas hnRNP-
A1 was expressed in both compartments (Fig. 8a). Notably, in
Figure 8 Nod2–hnRNP-A1 interaction and
phosphorylation of hnRNP-A1 by p38.
(a) Immunoassay of HEK293 cells transiently
transfected for 1 d with empty vector (C) or
with vector encoding Flag-tagged wild-type (W)
or 3020insC (M) Nod2, together with vector
encoding Flag-tagged hnRNP-A1, followed by
immunoprecipitation of cytoplasmic and nuclear
extracts with anti-Flag and immunoblot analysis
with anti-Flag, anti-hnRNP-A1 (a-hnRNP) and
antibody to phosphorylated serine (a-p-Ser). **,
immunoglobulin heavy chain; *, immunoglobulin
light chain. (b) Immunoassay of HEK293 cells
transiently transfected for 1 d with the Flagtagged
constructs described in a, followedby
immunoprecipitation and immunoblot analysis of
whole-cell lysates with anti-Flag or anti-hnRNP-
A1. (c) Immunoassay of HEK293 cells
transfected with human wild-type Nod2 (W) or
3020insC Nod2 (M; left) or with mouse wild-type
Nod2 (W) or 2939insC Nod2 (M; right), followed
by immunoprecipitation of whole-cell lysates with
antibody to serine-phosphorylated p38 (a-p-p38)
and immunoblot analysis with anti-Flag.
(d) Immunoassay of HEK293 cells treated for
60 min with the p38 inhibitor SB203580,
a
d
IP: α-Flag
Cytoplasm Nucleus
FLAG-NOD2 C W M C W M
IB: α-Flag
IB: α-hnRNP
IB: α-p-Ser
addition to the intact form of hnRNP-A1 (about 36 kilodaltons),
we noted three shorter forms of hnRNP-A1, which have been
described as products of hnRNP-A1 cleavage mediated by caspase
3 activated during apoptosis of the human Burkitt lymphoma cell
line BL60 induced by antibody to immunoglobulin M (anti-IgM) 26 .
There was less cleaved hnRNP-A1 in cells expressing 3020insC
Nod2 than in cells expressing empty vector or wild-type Nod2.
In addition, intact hnRNP-A1 was serine-phosphorylated exclusively
in the nucleus, whereas we detected serine-phosphorylated
forms of cleavage products only in the cytoplasm. Expression
b C W M e
IB: α-Flag
c
*
IP: α-p-p38
Human
Mouse
Flag-Nod2 C W M C W M C W M
Flag–hnRNP-A1
Nod2
– – – + + + + + +
IgH
hnRNP-A1
TPA (µM) 0 1 1 1
SB203580 (µM) 0 0 1 10
IP: α-p-Ser
IB: α-hnRNP
Figure 7 Binding of hnRNP-A1 to the IL10 NRE. (a) EMSAofnuclear
extracts of RAW264.7 cells stably transfected with empty vector (pCDNA3)
or vector encoding hnRNP-A1, analyzed with a probe containing wild-type
IL10 NRE. Right, extracts incubated with anti-hnRNP-A1 or control
immunoglobulin (IgG) before incubation with probe. Results are representative
of three independent experiments. (b) ChIP of primary human monocytes
with anti-hnRNP-A1 or control immunoglobulin (Ig), followed by PCR
amplification of IL10 promoter regions between positions 121 and +42
(top) or positions 3158 and 2947 (bottom) in input DNA or DNA extracted
from immunoprecipitates. Results are representative of three separate
experiments. (c) ChIP of unstimulated (Med) and LPS-stimulated (LPS)
primary human monocytes, followed by PCR amplification of the IL10
promoter region between positions 121 and +42. Rat IgG serves as a
control. Results are representative of three separate experiments. (d) ChIP
analysis of unstimulated human PBMCs from healthy control subjects (Ctrl)
or patients with Crohn’s disease (CD) with wild-type NOD2 (WT) or homozygous
for the 3020insC mutation (FS). Results are from one representative
of three experiments. (e) Real-time quantitative PCR analysis of the binding
of hnRNP-A1 in each sample in d. Results for immunoprecipitated DNA are
presented relative to those for genomic input DNA. Data are representative of
experiments done in triplicate (mean and s.d. of three donors per group).
IB: α-hnRNP
hnRNP-A1
Flag-
IP: α-Flag
IP: α-hnRNP
IP: α-hnRNP
IP: α-Flag
followed by the addition of TPA (12-O-tetradecanoylphorbol-1,3-acetate) for 30 min (concentrations, above lanes) and immunoprecipitation and immunoblot
analysis of nuclear extracts. Results in a–d represent at least three independent experiments. (e) Immunoblot analysis of whole-cell lysates of BMDMs
obtained from wild-type mice (WT; n ¼ 3) and p38a-deficient mice (p38-KO; n ¼ 3) and then stimulated with LPS. Results are representative of two
separate experiments. (f) Luciferase activity of lysates of RAW264.7 cells transfected with the human IL10 promoter–luciferase reporter together with empty
vector (pCDNA3) or vector encoding wild-type, S310312A or S192A hnRNP-A1, then stimulated for 24 h with medium or LPS. *, P o 0.03. Data represent
three independent experiments (mean and s.d.). (g) Immunoprecipitation and immunoblot analysis of cytoplasmic and nuclear extracts of PBMCs isolated
from patients with Crohn’s disease homozygous for 3020insC (M; n ¼ 4) or expressing wild-type Nod2 (W1; n ¼ 4) and healthy people expressing wild-type
Nod2 (W2; n ¼ 6). Results are representative of one experiment.
**
cleaved
f
Luciferase activity
(10 3 RLU/s)
g
IB:
8
6
4
2
*
* *
0 pCDNA3 WT S310–
312A
α-p38
α-hnRNP-A1
IP: α-p-Ser
IB: α-hnRNP
IP: α-p-p38
IB: α-hnRNP
p38
hnRNP-A1
p-hnRNP-A1
*
M W1 W2
WT p38-KO
Med
LPS
S192A
Cytoplasm
Nuclear
Cytoplasm
476 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
of 3020insC Nod2 selectively diminished the presence of
phosphorylated hnRNP-A1 in the nucleus. By coimmunoprecipitation,
we found evidence of direct physical interaction between
endogenous hnRNP-A1 and exogenous wild-type Nod2; hnRNP-A1
interacted less efficiently with 3020insC Nod2 (Fig. 8b). This
interaction seemed to take place exclusively in the cytoplasm
(data not shown), presumably because Nod2 is present only in
the cytoplasm.
Next we investigated whether the mitogen-activated protein kinase
p38a (A001717) was responsible for phosphorylation of hnRNP-A1.
Flag-tagged wild-type and 3020insC Nod2 interacted equally well with
phosphorylated p38 (Fig. 8c). However, we detected more hnRNP-A1
in complexes of serine-phosphorylated p38 and wild-type Nod2 than
in complexes of serine-phosphorylated p38 and 3020insC Nod2.
Notably, the 2939insC Nod2 mouse counterpart of 3020insC Nod2
did not act like the human mutant in that it resembled wild-type
Nod2 in its ability to facilitate interaction between serine-phosphorylated
p38 and hnRNP-A1.
In support of the idea that p38 was responsible for hnRNP-A1
phosphorylation, we found that SB203580, a small chemical inhibitor
of p38, blocked phorbol ester–induced phosphorylation of cleaved
hnRNP-A1 in HEK293 cells in a dose-dependent way (Fig. 8d). As
chemical inhibitors of p38 may have additional unidentified targets,
we confirmed our findings in mice lacking p38a specifically in
myeloid cells 27 . BMDMs from p38a-deficient and wild-type mice
expressed similar amounts of total hnRNP-A1, but LPS-induced
serine phosphorylation of hnRNP-A1 was much lower in p38adeficient
macrophages (Fig. 8e). These data support the idea that
p38, particularly the a-isoform, is critical for phosphorylation
of hnRNP-A1.
It has been shown that hnRNP-A1 is phosphorylated at the serine
residue at position 192 (Ser192), Ser310, Ser311 and Ser312 after T cell
stimulation 28 . To determine if these serine residues are critical for
hnRNP-A1-mediated IL10 transcription, we generated mutants of
hnRNP-A1 in which these serine residues were replaced with alanine
(S192A and S310–312A). We transfected cells expressing the IL10
promoter–luciferase reporter with empty vector or vector encoding
wild-type, S192A or S310–312A hnRNP-A1 constructs. Wild-type
hnRNP-A1 enhanced both basal and LPS-stimulated IL10 promoter
activity relative to the promoter activity resulting from transfection of
empty vector (Fig. 8f). S192A hnRNP-A1 also induced promoter
activity, albeit less efficiently than wild-type hnRNP-A1 did; in
contrast, S310–312A hnRNP-A1 failed to stimulate IL10 reporter
activity (Fig. 8f). These data collectively suggest that phosphorylation
of hnRNP-A1 at Ser310–Ser312, possibly by p38, is important for
its ability to enhance transcription of human IL10; moreover,
3020insC Nod2 may interfere with p38-mediated phosphorylation
of hnRNP-A1.
In agreement with the hypothesis proposed above, we noted
considerable impairment in the phosphorylation of cleaved and fulllength
hnRNP-A1 in the nucleus of cells from patients with Crohn’s
disease who were homozygous for 3020insC, relative to that of cells
from healthy subjects or patients with Crohn’s disease who lacked
3020insC (Fig. 8g). Expression of total p38 and hnRNP-A1 was similar
in all samples. The average extent of phosphorylation of hnRNP-A1 in
control subjects (n ¼ 6) was 3.2-fold higher than that of patients
with Crohn’s disease who were homozygous for 3020insC (n ¼ 6;
P ¼ 0.02467). Consistent with the impaired phosphorylation of
hnRNP-A1, there was less interaction between serine-phosphorylated
p38 and hnRNP-A1 in these patients (Fig. 8g). These data collectively
indicate that p38 phosphorylates hnRNP-A1 and that 3020insC Nod2
ARTICLES
inhibits this event by blocking the interaction between p38 and
hnRNP-A1 (Supplementary Fig. 2 online).
DISCUSSION
In this study we have investigated the controversial issue of the
influence of the NOD2 mutation 3020insC on host defense and
inflammation. We have shown that 3020insC Nod2 acted as an
inhibitor of IL10 expression in human monocytes. Because of the
critical function of IL-10 in controlling mucosal inflammation caused
by intestinal microflora, this finding links 3020insC to a probable
route leading to the development and pathogenesis of Crohn’s disease.
Our observations indicate that 3020insC Nod2 can interfere with the
steady-state intracellular signaling responsible for constitutive IL10
transcription. We have identified hnRNP-A1 as a chief target of
3020insC Nod2 in this pathway and have shown that hnRNP-A1
acted as a physiological inducer of IL10 transcription. Although
3020insC Nod2 did not inhibit the expression or nuclear localization
of hnRNP-A1, it did suppress the DNA-binding activity of hnRNP-A1
by interfering with p38-mediated phosphorylation of hnRNP-A1.
Notably, 3020insC Nod2 blocked transcription of human IL10 but
not mouse Il10, and the engineered mouse counterpart of the
3020insC Nod2 mutant, 2939insC Nod2, did not modulate expression
of either mouse or human IL-10. Thus, the 3020insC-p38-hnRNP
pathway is operative only in humans. This species specificity seems to
be due to the ability of 3020insC Nod2 but not 2939insC Nod2 to
block the interaction between p38 and hnRNP-A1.
We did not find evidence that 3020insC Nod2 could enhance
secretion of IL-1b from primary human monocytes. That finding is
inconsistent with studies of the Nod2 2939insC knock-in mouse, which
shows exacerbated IL-1b production due to hyperactivation of
NF-kB 18 . It suggests that the discrepancy between those results and
our finding of a lack of effect on IL-1b production in primary human
monocytes transfected with 3020insC Nod2 is due to the possibility
that 2939insC Nod2 has a gain of activity that the human mutant does
not have. The data collectively demonstrate that the mouse 2939insC
Nod2 mutant is not equivalent to human 3020insC Nod2 and that
lacking Nod2 is not the same as expressing 3020insC. These findings
call for caution in comparisons of mouse models and human disease.
As 3020insC Nod2blockedsteady-stateaswellasmicrobe-induced
IL10 expression, people with the 3020insC mutation may have chronically
lower IL-10 production. This diminution in IL-10, when combined
with other contributing factors over a lengthy period of time,
could cumulatively cause persistent inflammation and homeostatic
imbalance in the intestinal mucosa, leading to the development and
pathogenesis of Crohn’s disease. In people heterozygous for 3020insC,
wild-type Nod2 can presumably directly bind to 3020insC Nod2 and
may thereby diminish its IL-10-inhibiting and disease-causing activities.
This interaction could explain the much lower susceptibility to
Crohn’s disease of heterozygous patients. Emphasizing the influence of
other contributing factors, not all people with the 3020insC mutation
on both chromosomes suffer from Crohn’s disease 29 . Moreover, none
of the three common mutations in the LRR domain of NOD2 have
been found in Japanese patients with Crohn’s disease 30 . That observation
provides evidence for the presence of genetic heterogeneity among
patients of different ethnic and racial backgrounds.
It should be emphasized that published research documenting lower
defensin expression 31 , dysfunction of dendritic cells 32 and diminished
production of other cytokines 25,33 supports the idea that 3020insC
represents a loss-of-function mutation. It is likely that the cumulative
phenotypic effect of the 3020insC variant in patients with Crohn’s
disease may be attributable to other alterations in addition to
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 477

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
diminished IL-10 production. Whether the 3020insC Nod2 mutant
represents a loss of function or gain of function may depend on the
response being measured. For example, because the 3020insC mutation
is in the LRR domain of NOD2,the3020insC Nod2 mutant may not be
able to bind muramyl dipeptide; thus, all muramyl dipeptide–induced
responses could be lost. In that context, 3020insC Nod2 is a loss-offunction
mutant. In agreement with that conclusion, monocytes
derived from patients with Crohn’s disease who bear other mutations
in the LRR domain, like those from patients with the 3020insC
mutation, show impaired responses to muramyl dipeptide and several
TLR ligands in terms of secretion of IL-8 and tumor necrosis factor,
respectively 33 . However, our evidence suggests that 3020insC Nod2 and
other Nod2 variants bearing mutations in the LRR domain have also
acquired the ability to inhibit IL-10 transcription.
In summary, our study challenges the present paradigms about the
influence of the 3020insC mutation on Crohn’s disease. Notably,
3020insC is also strongly associated with graft-versus-host disease 34 ,
which is similar to Crohn’s disease in that both disorders are thought
to be exacerbated by TH1 responses and suppressed by IL-10. Thus,
our findings might be useful in efforts to identify therapeutic targets
for the treatment of Crohn’s disease and other T H1-mediated autoimmune
diseases associated with the 3020insC mutation.
METHODS
Mice. Nod2-knockout and control littermate mice were from P.J. Murray.
RICK-knockout (RIP2-knockout) and control mice were from E. Pamer. Mice
with conditional knockout of p38a were provided by J.M. Park 27 .
Antibodies and reagents. Antibody to phosphorylated serine (AB1603) was
from Chemicon International; anti-Flag (M2) was from Sigma; anti-p38 (9212)
and antibody to phosphorylated p38 (c9211) were from Cell Signaling; and
other antibodies for immunoblot analysis, EMSA and ChIP (anti-hnRNP-A1
(10030), anti-Nod2 (30199)) were from Santa Cruz Biotechnology. Recombinant
human and mouse interferon-g were from Genzyme. Muramyl dipeptide,
peptidoglycan and LPS from Escherichia coli were from Sigma. The peptidoglycan,
muramyl dipeptide and Pam3Cys (tripalmitoyl cysteinyl seryl tetralysine)
used in all experiments were from the same source and were used at the
concentrations described before 35 . They were free of endotoxin contamination,
as shown by analysis with polymixin B (data not shown).
Expression vectors. The mouse Nod2 expression vector and the mouse
2939insC Nod2 mutant were provided by G. Nunez. The 3020insC and
other human Nod2 mutants were generated by overlapping PCR. Human
hnRNP-A1 cDNA was amplified by RT-PCR of total RNA from human
monocytes. The sequences of all vectors were verified and all plasmids
were isolated with an EndoFree Maxi kit (Qiagen).
DNA affinity binding assay. The DNA affinity binding assay was done
essentially as described 36 with 2 mg biotinylated DNA oligonucleotides encompassing
the IL10 NRE conjugated to 100 ml streptavidin-bound Dynabeads.
Bound proteins were eluted by increasing the strength of the elution buffer and
were separated by 10% SDS-PAGE gel and visualized by silver staining.
Retroviral and lentiviral packaging and macrophage transduction. GP2-293
packaging cells (Clontech Laboratories) at a confluency of about 50 70% in
100-mm culture plates were transfected, using Lipofectamine (Invitorogen),
with 5 mg each of plasmids (pVSV-G; Clontech Laboratories) encoding human
wild-type Nod2 or 3020insC Nod2. Then, 2 d after transfection, supernatants
were cleared by centrifugation for 10 min at 1,000g and were pelleted for
90 min at 50,000g. Pelleted viruses were resuspended at 4 1C inDMEM.Bone
marrow cells were infected with retroviral particles on day 2 and again on day 4
during their maturation in the presence of 20% (vol/vol) conditioned medium
from mouse L cells. For lentivirus generation, HEK293TN cells (System
Biosciences) at a confluency of 50% were transfected, using FuGENE 6 (Roche),
with 3 mg of plasmid encoding viral envelope protein (G glycoprotein of
vesicular stomatitis virus), 5 mg of the pMDLg/pRRE lentivirus packaging
plasmid, 2.5 mg pRSV-REV (expressing human immunodeficiency virus 1
reverse transcriptase under control of the Rous sarcoma virus U3 promoter),
and 10 mg Nod2-MA1 or 3020insC Nod2–MA1 expression plasmid (the
bidirectional vector MA1). Then, 2 d after transfection, supernatants were
cleared by centrifugation for 10 min at 1,000g and were pelleted for 90 min at
50,000g. Pelleted viruses were resuspended at 4 1C in DMEM. Human
monocytes (0.5 10 6 cells per condition) were infected with 5 ml or50ml
of lentivirus-containing supernatant on day 1 or day 3, respectively, before
being stimulated with LPS on day 4.
Patients and genotyping of NOD2 mutations. Blood was collected from
patients with Crohn’s disease (n ¼ 74) and healthy volunteers (n ¼ 20). NOD2
gene fragments containing the 3020insC site were amplified by PCR in 50-ml
reaction volumes containing 100 200 ng genomic DNA and the appropriate
primers in various concentrations (Supplementary Table 1 online) in 10 mM
Tris-HCl pH 9.0, 50 mM KCl, 1.5 mM MgCl 2, 0.01% (wt/vol) gelatin, 0.1%
(vol/vol) Triton X-100, 0.35 mM dNTPs and 2 U Taq DNA polymerase
(Invitrogen). Samples were denatured at 92 1C for 5 min and then were
amplified in a PTC-200 thermal cycler (MJ Research; Biozym) by 35 cycles of
92 1C for1min,59.81C for 1 min and 72 1C for 1 min, followed by a final
extension step of 72 1Cfor3min.The3020insC polymorphism was analyzed by
Genescan on an ABI Prism 3100 Genetic Analyzer according to the manufacturer’s
protocol (Applied Biosystems). Three patients with Crohn’s disease
homozygous for the 3020insC mutation were selected for further studies. Three
patients with Crohn’s disease bearing the wild-type NOD2 allele and three
healthy people, matched for age and sex (23–38 years of age; male), served as
controls. All patients with Crohn’s disease were in remission and were not
treated with steroid drugs or antibiological therapy (such as anti–tumor
necrosis factor) during the 2 months before the study. None of the people
selected had the NOD2 mutations of C to T at position 2104 (R702W) or G to
C at position 2722 (G908R), also known to be associated with Crohn’s disease
(data not shown) 37 .
Additional methods. Information on cell culture, cytokine measurement,
reporter plasmids, preparation of nuclear extracts, immunoblot and immunoprecipitation,
ChIP assay, siRNA, nanoflow liquid chromatography–tandem
mass spectrometry analysis and the database search of tandem mass spectrometry
data for the identification of peptide sequences is available in the
Supplementary Methods online.
Statistical analysis. Student’s t-test (one-tailed) was used for data analysis
where appropriate.
Accession codes. UCSD-Nature Signaling Gateway (http://www.signalinggateway.org):
A001243, A001137 and A001717.
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank P.J. Murray (St. Jude Children’s Research Hospital) for Nod2knockout
and control littermate mice; E. Pamer (Memorial Sloan-Kettering
Cancer Center) for RICK-knockout (RIP2-knockout) and control mice;
J.M. Park (Harvard University School of Medicine) for mice with conditional
knockout of p38a; and G. Nunez (University of Michigan) for the mouse Nod2
expression vector and mouse 2939insC Nod2. Supported by the Broad
Medical Research Program (IBD-210R2 to X.M.).
AUTHOR CONTRIBUTIONS
E.N. contributed to the work in Figures 3,4,6,8; Y.H. contributed to the work in
Figures 1–7; X.K. contributed to Figure 4; M.G.N. contributed to Figures 1,7,8;
and X.M. contributed to the overall project.
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Podolsky, D.K. Inflammatory bowel disease. N. Engl. J. Med. 347, 417–429 (2002).
2. Fuss, I.J., Boirivant, M., Lacy, B. & Strober, W. The interrelated roles of TGF-b and
IL-10 in the regulation of experimental colitis. J. Immunol. 168, 900–908 (2002).
478 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
3. Inohara, N. & Nunez, G. NODs: intracellular proteins involved in inflammation and
apoptosis. Nat. Rev. Immunol. 3, 371–382 (2003).
4. Girardin, S.E. et al. Nod1 detects a unique muropeptide from gram-negative bacterial
peptidoglycan. Science 300, 1584–1587 (2003).
5. Girardin, S.E. et al. Nod2 is a general sensor of peptidoglycan through muramyl
dipeptide (MDP) detection. J. Biol. Chem. 278, 8869–8872 (2003).
6. Ogura, Y. et al. Nod2, a Nod1/Apaf-1 family member that is restricted to monocytes and
activates NF-kB. J. Biol. Chem. 276, 4812–4818 (2001).
7. Abbott, D.W., Wilkins, A., Asara, J.M. & Cantley, L.C. The Crohn’s disease protein,
NOD2, requires RIP2 in order to induce ubiquitinylation of a novel site on NEMO. Curr.
Biol. 14, 2217–2227 (2004).
8. Inohara, N. et al. An induced proximity model for NF-kB activationintheNod1/RICK
and RIP signaling pathways. J. Biol. Chem. 275, 27823–27831 (2000).
9. Hugot, J.P. et al. Association of NOD2 leucine-rich repeat variants with susceptibility to
Crohn’s disease. Nature 411, 599–603 (2001).
10. Lesage, S. et al. CARD15/NOD2 mutational analysis and genotype-phenotype correlation
in 612 patients with inflammatory bowel disease. Am. J. Hum. Genet. 70,
845–857 (2002).
11. Ogura, Y. et al. A frameshift mutation in NOD2 associated with susceptibility to Crohn’s
disease. Nature 411, 603–606 (2001).
12. Seiderer, J. et al. Predictive value of the CARD15 variant 1007fs for the diagnosis
of intestinal stenoses and the need for surgery in Crohn’s disease in clinical
practice: results of a prospective study. Inflamm. Bowel Dis. 12, 1114–1121
(2006).
13. Kuhn, R., Lohler, J., Rennick, D., Rajewsky, K. & Muller, W. Interleukin-10-deficient
mice develop chronic enterocolitis. Cell 75, 263–274 (1993).
14. Davidson, N.J. et al. T helper cell 1-type CD4 + T cells, but not B cells, mediate colitis in
interleukin 10-deficient mice. J. Exp. Med. 184, 241–251 (1996).
15. Sellon, R.K. et al. Resident enteric bacteria are necessary for development of
spontaneous colitis and immune system activation in interleukin-10-deficient mice.
Infect. Immun. 66, 5224–5231 (1998).
16. Kullberg, M.C. et al. Helicobacter hepaticus triggers colitis in specific-pathogen-free
interleukin-10 (IL-10)-deficient mice through an IL-12- and g interferon-dependent
mechanism. Infect. Immun. 66, 5157–5166 (1998).
17. Netea, M.G. et al. NOD2 mediates anti-inflammatory signals induced by TLR2
ligands: implications for Crohn’s disease. Eur. J. Immunol. 34, 2052–2059
(2004).
18. Maeda, S. et al. Nod2 mutation in Crohn’s disease potentiates NF-kB activity and IL-1b
processing. Science 307, 734–738 (2005).
19. Pauleau, A.L. & Murray, P.J. Role of nod2 in the response of macrophages to toll-like
receptor agonists. Mol. Cell. Biol. 23, 7531–7539 (2003).
20. Netea, M.G. et al. Nucleotide-binding oligomerization domain-2 modulates specific
TLR pathways for the induction of cytokine release. J. Immunol. 174, 6518–6523
(2005).
ARTICLES
21. Cao, S., Liu, J., Song, L. & Ma, X. The protooncogene c-Maf is an essential transcription
factor for IL-10 gene expression in macrophages. J. Immunol. 174, 3484–3492
(2005).
22. Steiner, G., Skriner, K., Hassfeld, W. & Smolen, J.S. Clinical and immunological
aspects of autoantibodies to RA33/hnRNP-A/B proteins–a link between RA, SLE and
MCTD. Mol. Biol. Rep. 23, 167–171 (1996).
23. Caporali, R., Bugatti, S., Bruschi, E., Cavagna, L. & Montecucco, C. Autoantibodies to
heterogeneous nuclear ribonucleoproteins. Autoimmunity 38, 25–32 (2005).
24. Jones, D.A. et al. Identification of autoantigens in psoriatic plaques using expression
cloning. J. Invest. Dermatol. 123, 93–100 (2004).
25. Netea, M.G. et al. NOD2 3020insC mutation and the pathogenesis of Crohn’s disease:
impaired IL-1b production points to a loss-of-function phenotype. Neth. J. Med. 63,
305–308 (2005).
26. Brockstedt, E. et al. Identification of apoptosis-associated proteins in a human Burkitt
lymphoma cell line. Cleavage of heterogeneous nuclear ribonucleoprotein A1 by
caspase 3. J. Biol. Chem. 273, 28057–28064 (1998).
27. Kim, C. et al. The kinase p38a serves cell type-specific inflammatory functions in skin
injury and coordinates pro- and anti-inflammatory gene expression. Nat. Immunol. 9,
1019–1027 (2008).
28. Buxade, M. et al. The Mnks are novel components in the control of TNF a biosynthesis
and phosphorylate and regulate hnRNP A1. Immunity 23, 177–189 (2005).
29. Linde, K. et al. Card15 and Crohn’s disease: healthy homozygous carriers of the
3020insC frameshift mutation. Am. J. Gastroenterol. 98, 613–617 (2003).
30. Yamazaki, K., Takazoe, M., Tanaka, T., Kazumori, T. & Nakamura, Y. Absence of
mutation in the NOD2/CARD15 gene among 483 Japanese patients with Crohn’s
disease. J. Hum. Genet. 47, 469–472 (2002).
31. Wehkamp, J. et al. NOD2 (CARD15) mutations in Crohn’s disease are associated with
diminished mucosal a-defensin expression. Gut 53, 1658–1664 (2004).
32. Kramer, M., Netea, M.G., de Jong, D.J., Kullberg, B.J. & Adema, G.J. Impaired
dendritic cell function in Crohn’s disease patients with NOD2 3020insC mutation.
J. Leukoc. Biol. 79, 860–866 (2006).
33. van Heel, D.A. et al. Muramyl dipeptide and toll-like receptor sensitivity in NOD2associated
Crohn’s disease. Lancet 365, 1794–1796 (2005).
34. Holler, E. et al. Both donor and recipient NOD2/CARD15 mutations associate with
transplant-related mortality and GvHD following allogeneic stem cell transplantation.
Blood 104, 889–894 (2004).
35. Watanabe, T., Kitani, A., Murray, P.J. & Strober, W. NOD2 is a negative regulator of Tolllike
receptor 2–mediated T helper type 1 responses. Nat. Immunol. 5, 800–808 (2004).
36. Liu, J., Cao, S., Herman, L.M. & Ma, X. Differential regulation of interleukin (IL)-12
p35 and p40 gene expression and interferon (IFN)-g-primed IL-12 production by IFN
regulatory factor 1. J. Exp. Med. 198, 1265–1276 (2003).
37. Netea, M.G. et al. NOD2 mediates induction of the antiinflammatory signals induced by
TLR2-ligands: implications for Crohn’s disease. Eur. J. Immunol. 34, 2052–2059
(2004).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 479

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Autophagy enhances the presentation of endogenous
viralantigensonMHCclassImoleculesduring
HSV-1 infection
Luc English 1 , Magali Chemali 1 , Johanne Duron 1 , Christiane Rondeau 1 , Annie Laplante 1 , Diane Gingras 1 ,
Diane Alexander 2 , David Leib 2 , Christopher Norbury 3 , Roger Lippé 1 & Michel Desjardins 1,4,5
Viral proteins are usually processed by the ‘classical’ major histocompatibility complex (MHC) class I presentation pathway.
Here we showed that although macrophages infected with herpes simplex virus type 1 (HSV-1) initially stimulated CD8 + T cells
by this pathway, a second pathway involving a vacuolar compartment was triggered later during infection. Morphological and
functional analyses indicated that distinct forms of autophagy facilitated the presentation of HSV-1 antigens on MHC class I
molecules. One form of autophagy involved a previously unknown type of autophagosome that originated from the nuclear
envelope. Whereas interferon-c stimulated classical MHC class I presentation, fever-like hyperthermia and the pyrogenic
cytokine interleukin 1b activated autophagy and the vacuolar processing of viral peptides. Viral peptides in autophagosomes
were further processed by the proteasome, which suggests a complex interaction between the vacuolar and MHC class I
presentation pathways.
The elaboration of an efficient immune response against pathogens
involves complex intracellular antigen-processing events. Endogenous
antigens such as viral proteins synthesized by infected host cells are
degraded in the cytoplasm by the proteasome, and the resulting
peptides are translocated into the endoplasmic reticulum, where
they are loaded onto major histocompatibility complex (MHC) class I
molecules. In contrast, exogenous antigens are processed by hydrolases
in lytic endovacuolar compartments and are loaded on MHC class II
molecules that reach the cell surface using recycling machineries
associated with these organelles. Although initially thought
to be strictly segregated, these pathways are actually functionally
interconnected, as shown by the ability of cells to present exogenous
antigens on MHC class I molecules; this process is referred
to as ‘cross-presentation’ 1 .
Autophagy, a process that allows the transfer of endogenous cellular
components into lytic vacuolar compartments, has been shown to be
essential to both innate and adaptive immunity 2–4 . This process can
be used to eliminate intracellular bacteria and viruses 5–11 . Furthermore,
autophagy can facilitate the presentation of endogenous
antigens on MHC class II molecules, thereby leading to activation of
CD4 + T cells 12,13 . The nature of the endovacuolar membranes
involved in the formation of autophagosomes is still a matter of active
investigation, but the endoplasmic reticulum is one potential source
of autophagosome membrane 14 . As substantial amounts of viral
Received 23 December 2008; accepted 12 February 2009; published online 22 March 2009; doi:10.1038/ni.1720
membrane glycoproteins are synthesized in the endoplasmic reticulum
of infected cells, it is likely that some of these antigens will reach lytic
vacuolar organelles during autophagy. Despite the fact that several
reports have demonstrated that antigens generated in the lumen of
phagosomes can be loaded (or ‘cross-presented’) on MHC class I
molecules and trigger a CD8 + T cell response 15–17 , it is not known
whether a similar process could occur after autophagy. Here we
provide evidence that infection of macrophages with herpes simplex
virus type 1 (HSV-1) triggered a vacuolar response that increased
the presentation of a peptide of HSV-1 glycoprotein B (gB) to
CD8 + T cells on MHC class I molecules. This vacuolar response,
linked to autophagy, could be modulated by various cytokines and
stress conditions.
RESULTS
Two phases of MHC class I presentation
IncubationofBMA3.1A7(called‘BMA’here)macrophageswithHSV-1
expressing a green fluorescent protein (GFP)-tagged capsid protein
(K26-GFP) resulted in the infection of about 35% of the cells, as
determined by flow cytometry (data not shown). Fluorescence microscopy
showed that K26-GFP capsids assembled in the nucleus of
infected cells at about 6–8 h after infection, reaching a maximum at
about 12 h after infection (data not shown). Starting at 6 h after
infection, infected macrophages stimulated a CD8 + T cell hybridoma
1 Département de Pathologie et Biologie Cellulaire, Université de Montréal, Succursale Centre-Ville, Montreal, Quebec, Canada. 2 Department of Ophthalmology and Visual
Sciences, Washington University School of Medicine, St. Louis, Missouri, USA. 3 Department of Microbiology and Immunology, Pennsylvania State University, Milton S.
Hershey College of Medicine, Hershey, Pennsylvania, USA. 4 Département de microbiologie et immunologie, Université de Montréal, Succursale Centre-Ville, Montreal,
Quebec, Canada. 5 Caprion Pharmaceuticals, Montréal, Québec, Canada. Correspondence should be addressed to M.D. (michel.desjardins@umontreal.ca)
480 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Figure 1 A vacuolar pathway participates in the processing of
endogenous viral proteins for presentation on MHC class I molecules.
(a) Activation of the gB-specific CD8 + T cell hybridoma (which
expresses b-galactosidase as an indicator of T cell activation)
by macrophages infected for various times (horizontal axis) with HSV-1,
then incubated for 12 h at 37 1C with the hybridoma. A595, absorbance
at 595 nm. (b) Activation of the hybridoma as described in a, withthe
addition of dimethyl sulfoxide (DMSO; negative control), bafilomycin A
(Baf), brefeldin A (BFA) or MG-132 at 2 h after macrophage infection.
(c) Activation of the hybridoma as described in a, with the addition of
bafilomycin A at 2 h after macrophage infection. (d) Immunofluorescence
microscopy of gB (blue) and LAMP-1 (red) in HSV-1infected
macrophages; pink indicates colocalization. Original
magnification, 40. (e) CD8 + T cell–stimulatory capacity (as described
in a) of uninfected macrophages (Mock), of BMA (BMA-HSV) or J774
(J774-HSV) macrophages infected for 8 h with HSV-1, and of cocultures
of J774 macrophages (H-2 d ) infected for 1 h with HSV-1, then mixed
with uninfected BMA (H-2 b ) macrophages at a ratio of 1:1 and cultured
together for 8 h (J774-HSV + BMA). Results in b,c,e are normalized to
results obtained for CD8 + T cells stimulated with macrophages treated
with DMSO (b,c) or infected BMA macrophages (e) and are presented in
arbitrary units. Data are from one representative of three independent experiments (mean and s.d. of triplicate samples; a), are from three independent
experiments (mean and s.e.m. of triplicate samples error bars; b,c,e) or are representative of three independent experiments (d).
specific for a peptide of amino acids 498–505 of HSV-1 gB 18,19 ,as
measured by the release of b-galactosidase after T cell activation
(Fig. 1a). The capacity of macrophages to stimulate CD8 + T cells
continued to increase up to 12 h after infection, the time at which
cellular mortality induced by the viral infection began to occur (data
not shown). This macrophage cell death probably explains the
decrease in CD8 + T cell stimulation between 12 h and 14 h after
infection (Fig. 1a).
Stimulation of the CD8 + T cell hybridoma was much lower after
treatment of macrophages with the proteasome inhibitor MG-132 or
with brefeldin A, a drug that inhibits the transport of molecules
through the biosynthetic pathway (Fig. 1b). These results indicate that
processing and presentation of the viral antigen gB in infected
a
Time (h)
2
4
6
8
e
β-galactosidase
activity (relative)
2
1
0
LC3 Phase-contrast b
Basal
Basal + Baf
Rapa
Rapa + Baf
HS
HS + Baf
β-galactosidase
activity (relative)
c
β-galactosidase
activity (relative)
f
β-galactosidase activity
(relative)
1.5
1.0
0.5
2
1
0
0
2
1
0
Rapa:
DMEM
DMEM + 3-MA
8 10 12
Time after infection (h)
Ctrl siRNA
Atg5 siRNA
8 12 8 12
Time after
infection (h)
– + – +
d
Ctrl
siRNA Atg5
siRNA
Ctrl siRNA
Atg5 siRNA
Atg5
Tubulin
a b c 2
β-galactosidase activity
(A595 )
1.0
0.8
0.6
0.4
0.2
0
4 6 8 101214
Time after infection (h)
β-galactosidase activity
(relative)
d e
1.0
0.5
0
DMSO
Baf
BFA
MG-132
β-galactosidase activity
(relative)
gB LAMP-1 Merge
ARTICLES
0
8 10 12
Time after infection (h)
macrophages involves the ‘classical’ endogenous pathway of MHC
class I presentation. However, bafilomycin A, a drug that inhibits the
vacuolar proton pump and the acidification of endosomes and
lysosomes, had only a minimal effect during the early period of
infection (up to 8 h after infection) but strongly inhibited the capacity
of infected macrophages to stimulate CD8 + T cells at 10 h and 12 h
after infection (Fig. 1c). These results suggest that the initial processing
of endogenous gB by the classical pathway is followed by the
engagement of a vacuolar pathway that considerably improves the
processing of gB and the activation of CD8 + T cells. The possibility of
a contribution by a vacuolar pathway was supported by immunofluorescence
analyses that indicated that endogenous gB localized
together with the endo-lysosomal marker LAMP-1 during infection
(Fig. 1d). The possibility of the presence of gB in degradative
compartments was further supported by the finding of higher gB
expression in infected macrophages treated with bafilomycin (data not
shown). These results raised the issue of how gB reaches the lysosomal
1
β-galactosidase
activity (relative)
DMSO
Baf
1.0
0.5
0
Mock
BMA-HSV
J774-HSV + BMA
Figure 2 Autophagy induced during HSV-1 infection contributes to the
processing and presentation of endogenous viral antigens on MHC class I
molecules. (a) Immunofluorescence microscopy of LC3 expression in
macrophages infected for 2–8 h (left margin) with HSV-1. Original
magnification, 20. (b) Activation of the gB-specific CD8 + T cell hybridoma
(described in Fig. 1a) by macrophages infected for various times (horizontal
axis) with HSV-1, with (DMEM + 3-MA) or without (DMEM) the addition of
3-methyladenine 2 h after infection, then incubated for 12 h at 37 1C with
the hybridoma. (c) Activation of the hybridoma (as described in b) by
macrophages transfected for 60 h with control (Ctrl) siRNA or Atg5-specific
siRNA, then infected for 8 h or 12 h with HSV-1. (d) Immunoblot analysis
of Atg-5 in siRNA-treated macrophages. (e) Activation of the hybridoma
(as described in b) by macrophages infected with HSV-1 and incubated at
37 1C (Basal), incubated for 12 h at 39 1C before being infected with
HSV-1 (heat shock (HS)), or treated with rapamycin during HSV-1 infection
(Rapa), with (+ Baf) or without the addition of bafilmycin A at 2 h after
infection. (f) Activation of the hybridoma (as described in b) by macrophages
transfected for 60 h with control siRNA or Atg5-specific siRNA, then infected
for 8 h with HSV-1, with (+) or without (–) the addition of rapamycin at
2 h after infection. Results in b,c,e,f are normalized to results obtained for
CD8 + T cells stimulated with macrophages infected for 8 h at 37 1C without
further treatment (b,e) or infected macrophages treated with control siRNA
(c,f) and are presented in arbitrary units. Data are representative of three
independent experiments (a,d) or are from three independent experiments
(mean and s.e.m. of triplicate samples; b,c,e,f).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 481

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
degradative pathway. The possibility of transfer of gB to lysosomes
through phagocytosis of infected cells (cross-presentation) was ruled
out by results showing that incubation of H-2 b BMA macrophages
together with HSV-1-infected H-2 d J774 macrophages did not result
in activation of the H-2 b -specific CD8 + T cell hybridoma (Fig. 1e).
Hence, vacuolar processing of gB in the later period of infection was
more likely to involve membrane-trafficking events that occurred
exclusively in the infected cell.
Data have shown that endogenous viral proteins can be presented
on MHC class II molecules by a process involving autophagy 9 ,which
indicates that trafficking events that enable the transport of endogenous
proteins to vacuolar degradative organelles can occur in virusinfected
cells. To determine if autophagy was involved in the late
processing of endogenous gB and its presentation on MHC class I
molecules, we first monitored the presence of LC3, a marker of
autophagy, in macrophages at various times after infection
(Fig. 2a). Although we did not detect it in uninfected cells, we
found LC3 beginning at 4–6 h after infection, which indicated that
an autophagic response occurred during the late phase of HSV-1
infection in macrophages. Whereas the inhibitory effect of bafilomycin
indicated that a vacuolar response of some kind was triggered during
the late phase of infection in macrophages (Fig. 1c), the possibility
of a contribution of autophagy to this process was suggested by
the substantial inhibition of the CD8 + T cell–stimulatory capacity
of macrophages treated with 3-methyladenine, a commonly used
a
b
c
g
Control
HS
Rapa
Time (h)
IB: α-HSV
d Time (h)
WT ∆34.5
0 6 8 10 6 8 10
4
6
8
10
(kDa)
250
100
75
50
37
25
20
VP26 gB LC3 Merge
h
Time (h)
1,200
Counts
960
620
480
240
6 8
1,200
960
620
480
240
inhibitor of autophagy (Fig. 2b). Confirmation of the involvement of
autophagy in the processing and presentation of gB peptides on MHC
class I molecules was provided by experiments involving small interfering
RNA (siRNA)-mediated silencing of Atg5, a protein involved in
the formation of autophagosomes 20 . Macrophages treated with a
control siRNA had a greater capacity to stimulate CD8 + T cells
between 8 h and 12 h after infection, but macrophages treated with
Atg5-specific siRNA did not (Fig. 2c),whichlinkedthislategainin
stimulation to the induction of autophagy.
Further support for the idea that autophagy contributes to the
vacuolar processing and presentation of gB on MHC class I molecules
was provided by results indicating that treatment of infected macrophages
with rapamycin, an inhibitor of the kinase mTOR that
stimulates autophagy 21 ,considerablyimprovedCD8 + T cell stimulation
(Fig. 2d). We obtained similar results with macrophages exposed
to a mild heat shock before infection (39 1C for 12 h), a condition also
known to induce autophagy 22 (Fig. 2d). The enhanced CD8 + T cell
stimulation induced by mTOR or heat shock was abolished by the
addition of bafilomycin (Fig. 2e). These data further link the vacuolar
processing of gB to autophagy. We obtained similar results with mouse
embryonic fibroblasts isolated from Atg5 –/– mice 20 (Supplementary
Fig. 1a online). Silencing of Atg5 in infected macrophages with siRNA
abolished the effect of rapamycin (Fig. 2f), which confirmed the
specificity of this drug for the autophagic pathway. Similarly, neither
bafilomycin (Supplementary Fig. 1c,d) nor 3-methyladenine (data
e
6 h
8 h
∆34.5
WT
∆34.5
WT
1,200
gB LC3 Overlay
10
Ctrl
WT 17+
∆34.5
0
0
0
100 101 102 103 104 100 101 102 103 104 100 101 102 103 104 gB
960
620
480
240
f
β-galactosidase
activity (relative)
2.5
WT 17+
∆34.5
2.0
1.5
1.0
0.5
0
6 8 10
Time after
infection (h)
i WT 17+
∆34.5
1.00
0.75
0.50
0.25
0
Baf: – + – + – + – + – + – +
6 8 10
Time after infection (h)
Figure 3 Both gB and LC3 accumulate in perinuclear regions during HSV-1 infection. (a–c) Immunofluorescence microscopy of uninfected macrophages
incubated at 37 1C (control; a), subjected to mild heat shock (b) or treated with rapamycin (c), then stained with anti-LC3. (d) Immunofluorescence
microscopy of the expression of LC3, gB and GFP (VP26) by macrophages infected for various times (left margin) with HSV-1. White indicates colocalization.
(e) Immunofluorescence microscopy of macrophages infected for 6 h or 8 h (left margin) with wild-type HSV-1 (WT) or HSV-1 lacking ICP34.5 (D34.5).
Blue, staining of nuclei with DAPI (4,6-diamidino-2-phenylindole). Original magnification, 100 (a–c) or 63 (d,e). Results in a–e are representative of
three independent experiments. (f) Activation of the gB-specific CD8 + T cell hybridoma (as described in Fig. 1a) by macrophages infected for various times
(horizontal axis) with wild-type HSV-1 strain 17+ (WT 17+) or D34.5 HSV-1. Data are from three independent experiments (mean and s.e.m. of triplicate
samples). (g,h) Immunoblot analysis (IB; g) and flow cytometry (h) of the expression of HSV-1 proteins (g) and gB (h) in macrophages infected for various
times (above lanes (g) or plots (h)) with wild-type or D34.5 HSV-1. Ctrl, control (uninfected BMA macrophages; h). Data are representative of two (g) or
three (h) independent experiments. (i) Activation of the gB-specific CD8 + T cell hybridoma (as described in Fig. 1a) by macrophages infected for various
times (below graph) with wild-type or D34.5 HSV-1, with (+) or without (–) the addition of bafilomycin A at 2 h after infection. Results in f,i are normalized
to results obtained for CD8 + T cells stimulated with macrophages infected for 6 h with wild-type virus (f) or with infected macrophages incubated without
bafilomycin (i) and are presented in arbitrary units. Data are from three independent experiments (mean and s.e.m. of triplicate samples).
482 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
β-galactosidase
activity (relative)

© 2009 Nature America, Inc. All rights reserved.
a
b
N
VP
c d
e f
N
not shown) affected the capacity of infected macrophages to stimulate
CD8 + T cells when autophagy was inhibited by silencing of Atg5. Our
results so far indicated that a vacuolar pathway linked to autophagy,
triggered within 8–10 h of infection, enhanced the ability of infected
macrophages to stimulate gB-specific CD8 + T cells.
Late-stage autophagy involving the nuclear envelope
To study the autophagic response associated with HSV-1 infection in
macrophages, we used immunofluorescence and electron microscopy.
We first analyzed the induction of autophagy by assessing the
autophagic marker LC3. We noted a weak signal for LC3 in uninfected
macrophages (Fig. 3a), but uninfected cells submitted to a mild heat
shock (Fig. 3b) or treated with rapamycin (Fig. 3c) had strong
punctate LC3 signals in the cytoplasm. In contrast, there was a strong
LC3 signal in close association with the nuclear envelope in cells 6–8 h
after HSV-1 infection (Fig. 3d,e). In many cases, vesicles strongly
labeled for LC3 seemed to be connected to the nuclear envelope. The
colocalization of LC3 and gB suggested that autophagic structures
containing viral proteins might originate in the vicinity of the nucleus
at a late phase of infection. The difference between the typical labeling
noted when autophagy was triggered by rapamycin and that in
infected cells might be linked to the fact that HSV-1 infection has
been shown to inhibit macroautophagy 23 .
The accumulation of LC3 around the nucleus was possibly associated
with a cellular process distinct from classical macroautophagy
and induced as a late response to infection. To test that hypothesis, we
compared the distribution of gB and LC3 in macrophages infected
with wild type HSV-1 and a mutant HSV-1 lacking the ICP34.5
protein (D34.5); unlike the wild-type virus, this mutant virus is unable
N
N
VP
VP
N
ARTICLES
Figure 4 HSV-1 induces the formation of autophagosome-like structures
from the nuclear envelopes of infected macrophages. Electron microscopy
of macrophages 10 h after infection with HSV-1. (a) Arrows indicate
membrane-coiled structures emerging from the nucleus of an infected cell.
(b–d) Four-layered membrane structures formed by coiling of the nuclear
membrane. (e,f) Glucose-6-phosphatase (black deposits) on autophagosomelike
structures emerging from the nuclear envelope or free in the
cytoplasm, and viral capsids in the cytoplasm engulfed in the lumen of an
autophagosome-like compartment. N, nucleus; VP, viral particles. Scale
bars, 1 mm (a), 0.25 mm (b–d,f) or0.4mm (e). Images are representative
of three independent experiments with at least 100 cell profiles in each.
to inhibit macroautophagy. As expected, the D34.5 virus failed to
express any detectable ICP34.5 protein (data not shown). In addition,
consistent with the ability of ICP34.5 to mediate dephosphorylation of
the translation-initiation factor eIF2a 24 ,theD34.5 virus induced more
phosphorylated eIF2a than did wild-type HSV-1 (data not shown).
Macrophages infected with the D34.5 virus showed considerable
accumulation of LC3 on vesicular structures, which also contained
large amounts of gB and were present throughout the cytoplasm
(Fig. 3e). We found no apparent labeling for LC3 in the vicinity of the
nuclear envelope. In contrast, infection with the corresponding wildtype
virus (strain 17+) led to the accumulation of LC3 near the
nuclear envelope between 6 h and 8 h after infection (Fig. 3e); these
results are in agreement with results obtained with the KOS wild-type
strain (Fig. 3d). These findings collectively suggested that distinct
types of autophagic structures were induced in response to the D34.5
and wild type viruses. Macrophages infected with D34.5 or wild-type
HSV-1 (at an identical multiplicity of infection) were able to stimulate
gB-specific CD8 + T cells to a similar extent (Fig. 3f). However, the
expression of HSV-1 proteins (Fig. 3g) and gB (Fig. 3h) was much
lower in macrophages infected with the D34.5 virus, in agreement with
published studies 25,26 . Thus, macrophages infected with the D34.5
virus stimulated gB-specific CD8 + T cells much more efficiently.
Bafilomycin strongly inhibited the stimulatory ability of macrophages
infected with either D34.5 or wild-type HSV-1 (Fig. 3i), which
emphasizes the involvement of a vacuolar pathway in both cases.
To characterize the autophagosomal structures induced during the
late phase of infection, we used electron microscopy for detailed
morphological analysis (Fig. 4 and Supplementary Fig. 2 online). We
noted two types of prominent structures in infected cells. The first was
characterized by the presence of double-membraned structures
reminiscent of the morphology of autophagosomes found in cells
treated with rapamycin (Supplementary Fig. 2a). These structures
often surrounded viral particles in the cytoplasm (Supplementary
Fig. 2b). The degradation of microbial invaders by autophagy has
been described before and is referred to as ‘xenophagy’ 27 . These
structures showed substantial labeling for LC3 and, to a lesser extent,
for gB (Supplementary Fig. 2c,d), which confirmed the presence of
viral proteins in autophagosomes. The second type of organelle had
multiple membranes either connected to the nuclear envelope or
present in the cytoplasm (Fig. 4a,e). These structures seemed to
emerge through a coiling process of the inner and outer nuclear
membrane, forming four-layered structures that engulfed part of the
nearby cytoplasm (Fig. 4b,c). In several examples, similar structures
containing cytoplasm and unenveloped viral capsids seemed to be
disconnected from the nucleus (Fig. 4d).Thesestructureswerenot
present in uninfected or rapamycin-treated cells (Supplementary
Fig. 2e). However, there were more two- and four-membraned structures
in infected macrophages at the late phase of infection (Supplementary
Fig. 2e,f). To determine whether the four-layered membrane
structures present in the cytoplasm were similar to those that emerged
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 483

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a LC3 b LC3
gB
c d
N
BSA-gold
Figure 5 The four-layered membrane structures that emerge from the
nuclear envelope have autophagosome-like features. Immunoelectron
microscopy of macrophages 10 h after infection with HSV-1.
(a–c) Accumulation of LC3 (a,b) and gB (c). (d) Fusion of four-layered
membrane structures and lysosomes preloaded with bovine serum albumin–
gold (BSA-gold; black dots). Original magnification, 54,800 (a,b),
69,000 (c) or 38,000 (d). Images are representative of three (a–c) or
two (d) independent experiments.
from the nuclear envelope, we used electron microscopy to locate the
product of the enzyme glucose-6-phosphatase, a specific marker of the
endoplasmic reticulum and nuclear envelope. The results indicated
that the product of glucose-6-phosphatase was restricted to the
endoplasmic reticulum and nuclear envelope (Fig. 4e). The fourlayered
membrane structures connected to the nuclear envelope, as
well as those present in the cytoplasm, were also positive for glucose-
6-phosphatase (Fig. 4f), which confirmed their origin in the endoplasmic
reticulum and/or nuclear envelope.
To determine whether the four-layered membrane structures had
features of autophagosomes, we used immunoelectron microscopy to
detect the autophagosome marker LC3. We found accumulation of
LC3 on membrane structures emerging from the nuclear envelope, as
well as on four-layered membrane structures apparently disconnected
from the nucleus and present in the cytoplasm (Fig. 5a,b). The
association between LC3 and the membrane of autophagosomes
Figure 6 Involvement of lytic vacuolar compartments in the processing and
presentation of endogenous antigens on MHC class I molecules after
treatment with proinflammatory cytokines. (a) Activation of the gB-specific
CD8 + T cell hybridoma (as described in Fig. 1a) by macrophages exposed
to DMSO (negative control), mild heat shock, IL-1b or IFN-g.
(b) Dansylcadaverin staining of untreated macrophages (Basal) or
macrophages exposed to mild heat shock, IFN-g or IL-1b and infected
for 8 h with wild-type HSV-1, normalized to the signal obtained in basal
conditions and presented in arbitrary units. (c–f) Activation of the gBspecific
CD8 + T cell hybridoma (as described in Fig. 1a) by macrophages
incubated at 37 1C (c) or exposed to mild heat shock (d), IL-1b (e) or
IFN-g (f) and infected for 8 h with wild-type HSV-1 with the addition of
DMSO (negative control), 3-methyladenine, bafilomycin, brefeldin A or
MG-132 at 2 h after infection. Results in a,c–f are normalized to results
obtained for CD8 + T cells stimulated with macrophages incubated with
DMSO in each condition and are presented in arbitrary units. Data are from
three independent experiments (mean and s.e.m. of triplicate samples).
indicated that our antibody recognized the LC3-II cleaved form of
the protein 28 . Quantitative analysis of the immunolabeling for LC3
showed there was an average (± s.e.m.) of 3.03 ± 0.47 gold particles
per mm membrane on autophagosomes, compared with 0.15 ± 0.02
and 0.70 ± 0.15 gold particles per mm membrane for the plasma
membrane and nuclear membrane, respectively. We also found gB on
the membrane of the nuclear envelope (data not shown), as well as on
the four-layered membrane structures (Fig. 5c). The ability of these
structures to fuse with lytic organelles was confirmed by the presence
of bovine serum albumin–gold particles, transferred from lysosomes,
in these structures (Fig. 5d). These data collectively confirm the
autophagosomal nature of the four-layered membrane structures
originating from the nuclear envelope.
Cytokines engage classical and vacuolar responses
Treatment of macrophages with the proinflammatory cytokine interferon-g
(IFN-g) stimulates the clearance of mycobacteria by a process
involving autophagy 6 . Therefore, we tested whether this cytokine
might promote the vacuolar processing of gB and improve the ability
of HSV-1-infected macrophages to stimulate CD8 + T cells. As heat
treatment of macrophages stimulated autophagy, we also tested the
potential effect of the pyrogenic cytokine interleukin 1b (IL-1b). IFN-g
considerably enhanced the ability of HSV-1-infected macrophages to
stimulate CD8 + T cells (Fig. 6a). IL-1b and mild heat-shock treatment
also augmented the stimulatory capacity of macrophages (Fig. 6a).
Notably, staining with dansylcadaverin, a marker of autophagy, was
greater after exposure to mild heat shock or IL-1b but not after
treatment with IFN-g (Fig. 6b). Electron microscopy also showed that
cells stimulated with heat shock or IL-1b had more autophagosomes
(data not shown). These results suggest that the enhanced ability of
macrophages to stimulate CD8 + T cells after treatment with mild heat
shock or IL-1b was linked to the contribution of a vacuolar pathway
related to autophagy. We confirmed that hypothesis by showing that
treatment of IL-1b- or heat shock–treated macrophages with either
3-methyladenine, the inhibitor of autophagy, or bafilomycin, the
inhibitor of the vacuolar proton pump, resulted in a much lower
capacity of macrophages to stimulate CD8 + T cells than that of control
cells (Fig. 6c–e). In contrast, the very strong stimulation of CD8 +
T cells induced by infected macrophages kept at 37 1C or treated with
a
β-galactosidase activity (relative)
5
4
3
2
1
0
DMSO
HS
1.0
0.5
0
HS
IL-1β
IFN-γ
DMSO
3-MA
Baf
BFA
MG-132
b
Dansylcadaverin intensity
(relative)
1.3
1.2
1.1
1.0
0.9
0.8
0
IL-1β
1.0
0.5
0
Basal
HS
IL-1β
IFN-γ
c
DMSO
3-MA
Baf
BFA
MG-132
β-galactosidase
activity (relative)
d e f
β-galactosidase
activity (relative)
β-galactosidase
activity (relative)
37 °C
1.0
0.5
0
β-galactosidase
activity (relative)
IFN-γ
1.0
0.5
0
DMSO
3-MA
Baf
BFA
MG-132
DMSO
3-MA
Baf
BFA
MG-132
484 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
IFN-g was not strongly affected by 3-methyladenine or bafilomycin
and therefore was not due to vacuolar processing (Fig. 6c,f). Confirmation
that autophagy did not contribute to the stimulatory effect
of IFN-g was provided by experiments showing that this cytokine
enhanced the capacity of infected embryonic fibroblasts isolated from
wild-type and Atg5 –/– mice to stimulate CD8 + T cells (Supplementary
Fig. 2a). However, the stimulatory effect induced by IL-1b and mild
heat shock was completely abolished by brefeldin A and MG-132
(Fig. 6d,e). The effect of these two inhibitors of the ‘classical’ pathway
of MHC class I presentation indicated a close interaction occurring
between this pathway and the vacuolar pathway in the processing of
gB (Supplementary Fig. 3 online).
DISCUSSION
One of the main components of the complex cell-entry ‘machinery’ of
herpes viruses is gB 29 . This transmembrane protein of the viral
envelope is synthesized in the endoplasmic reticulum of infected
cells. In agreement with published work 18 , we found that gB was
expressed within 2 h after infection in BMA macrophages and was
present in the perinuclear region of infected cells. gB can also
accumulate in the inner membrane as well as the outer membrane
of the nuclear envelope 30,31 . Our results indicated that the processing
of gB by infected macrophages has two distinct phases. During the
first phase, which occurs 6–8 h after infection, gB is processed mainly
by the classical pathway of MHC class I presentation, which involves
proteasome-mediated degradation and transport steps through the
biosynthetic apparatus. During the second phase of infection, which
occurs 8–12 h after infection, a vacuolar pathway is triggered and
contributes substantially to the capacity of infected macrophages to
stimulate CD8 + T cells. The cellular trafficking events that enable the
transport of gB in the lysosomal degradative pathway are poorly
understood. We were able to rule out the possibility of a contribution
by a cross-presentation pathway involving the phagocytosis of infected
cells by neighboring macrophages, which emphasized the fact that gB
reaches the vacuolar processing pathway by an endogenous route.
Instead, our results indicated the involvement of autophagy in facilitating
the processing and presentation of endogenous viral peptides on
MHC class I molecules.
Our findings may seem to contradict earlier reports indicating
that HSV-1 inhibits macroautophagy 23 ; this inhibition is key to the
neurovirulence of the virus 11 . However, a closer look at infected
macrophages suggests that a form of autophagy distinct from
macroautophagy is triggered. Macroautophagy can be induced
in a variety of cells by mild heat shock or treatment with the
mTOR inhibitor rapamycin 21,22 . In uninfected macrophages, these
conditions induced throughout the cytoplasm the formation of
autophagosomes with a double-membraned structure. In contrast,
HSV-1-infected macrophages had two types of structures. In addition
to the double-membraned structures, four-layered membrane
structures that emerged from the nuclear envelope and accumulated
in the cytoplasm at around 8 h after infection were present in
most infected macrophages. We did not find such structures in
uninfected cells treated with rapamycin, which suggested that
they arose from a specific host response to HSV-1 infection distinct
from macroautophagy. Although four-layered membrane structures
have been reported before in the cytoplasm of HSV-1-infected
mouse embryonic fibroblasts 25 , the autophagosomal nature of
those structures was not documented. Here we have shown that
these four-layered membrane structures were ‘decorated’ with LC3,
a protein that is key to the formation of autophagosomes 28 ,and
that these structures fused with lysosomes filled with bovine
ARTICLES
serum albumin–gold, thereby generating an environment suitable
for the hydrolytic degradation of gB.
We conclude that the ability of HSV-1 to inhibit macroautophagy
early after infection, and thus to potentially limit the presentation of
viral peptides, is counterbalanced by a host response involving the
induction of a previously unknown autophagy pathway at a later time
after infection. Our conclusion is supported by the results obtained
with the D34.5 mutant virus. Macrophages infected with this mutant
were able to trigger a strong immune response, like macrophages
infected with wild-type viruses. Although the apparent magnitude of
activation of CD8 + T cells induced by wild-type and D34.5 viruses was
similar, the quantity of gB and viral proteins expressed in D34.5infected
cells was much lower than that in macrophages infected
with wild-type HSV-1. Macrophages infected with either D34.5 or
wild-type HSV-1 engaged strong vacuolar responses. The distinct
localization of LC3 to autophagosomes in the cytoplasm in macrophages
infected with D34.5 virus, in contrast to its localization
to the nuclear membrane in macrophages infected with wildtype
virus, indicates that both types of autophagic responses can
participate in the processing of viral proteins for presentation on
MHC class I molecules.
It has been reported that IFN-g stimulates autophagy and the
clearance of mycobacteria in macrophages 3 . In our studies, IFN-g had
no substantial effect on the induction of autophagy, a discrepancy that
might be explained by the cell type and/or concentration of the
cytokine used for stimulation 32 . Nevertheless, IFN-g-treated macrophages
were much more efficient at stimulating CD8 + T cells after
HSV-1 infection. As cotreatment with bafilomycin or 3-methyladenine
had no effect on this IFN-g-induced stimulatory capacity, we conclude
that vacuolar processing and autophagy were not essential to the
response induced by IFN-g. The stimulatory effect of IFN-g on
antigen presentation is well established. This cytokine upregulates
assembly of the immunoproteasome 33 and stimulates expression of
TAP1, the transporter associated with antigen presentation 34 , two
key components of the classical MHC class I presentation pathway.
Therefore, it was not unexpected to find strong inhibition of the
stimulation of CD8 + T cells when we treated IFN-g-stimulated
macrophages with MG-132 and BFA. In contrast, the IL-1b-induced
improvement in the ability of infected macrophages to stimulate
CD8 + T cells was inhibited considerably by 3-methyladenine
and bafilomycin, which confirmed the contribution of autophagy to
this process.
The similarity in the results obtained with IL-1b and mild heatshock
treatment suggests that IL-1b induces a cellular response similar
to the one that occurs during fever-like conditions. Indeed, IL-1b is
well known for its ability to induce fever 35 . Tumor necrosis factor, a
second pyrogenic cytokine, also stimulated autophagy and the vacuolar
processing of gB in our system (data not shown). These results
suggest that the stress induced during fever conditions or after
stimulation with pyrogenic cytokines triggers defense mechanisms,
promoting more-efficient processing of viral antigens in vacuolar
organelles. Notably, cytomegalovirus, a member of the herpesviridae
family, can block signaling by IL-1b and tumor necrosis factor
during the early phase of infection 36 , a process that might protect
the virus by inhibiting the induction of antigen processing through a
vacuolar response.
Our results showing that lytic organelles associated with the
processing of antigens for presentation on MHC class II molecules 12
participated in the presentation of endogenous viral peptides on MHC
class I molecules emphasize the dynamic cooperation between the
‘classical’ and ‘vacuolar’ pathways of antigen presentation. This close
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 485

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
interaction is further emphasized by results showing that even in
conditions in which vacuolar processing contributes to most of the
viral antigen processing, such as after IL-1b or heat-shock stimulation,
MG-132 and BFA still had a strong inhibitory effect. These results
support a model in which the vacuolar processing of viral proteins in
autophagosomes is followed by processing by the proteasome and
peptide loading onto MHC class I molecules in the endoplasmic
reticulum. The molecular mechanisms that enable the transfer of viral
peptides from autophagosomes to proteasomes in the cytoplasm are
unknown. A similar transport step has been shown to take place on
phagosomes 37 . Although the molecular ‘machines’ that enable these
translocation events have not been identified, it has been proposed
that endoplasmic reticulum translocons such as Sec61 and/or derlin-1
could be involved 16,17,38 . Although the nature of the endomembranes
involved in the formation of autophagosomes during macroautophagy
is still a matter of debate, our results have shown that the nuclear
envelope, made of endoplasmic reticulum, is the membrane used to
form the gB-enriched autophagosomes with four-layered membranes
in HSV-1-infected macrophages. The endoplasmic reticulum has been
shown to participate in a form of autophagy in yeasts referred to as
‘ER-phagy’ 14 . Whether this process is homologous with the nuclear
envelope–derived autophagic process documented here remains to be
investigated. Published studies have shown that viruses take advantage
of macroautophagy to find refuge in organelles, where they freely
assemble 39 . Our results indicate that autophagy can also benefit the
host by providing an additional pathway for the degradation of
endogenous viral proteins for antigen presentation.
METHODS
Cells, viruses, antibodies and reagents. The BMA3.1A7 macrophage cell
line was derived from C57BL/6 mice as described 37 and was cultured in
complete DMEM (10% (vol/vol) FCS, penicillin (100 units/ml) and streptomycin
(100 mg/ml)). The mouse macrophage cell line J774 was from American
Type Culture Collection. The b-galactosidase–inducible HSV gB–specific CD8 +
T cell hybridoma HSV-2.3.2E2 (provided by W. Heath, University of
Melbourne) was maintained in RPMI-1640 medium supplemented with 5%
(vol/vol) FCS, glutamine (2 mM), penicillin (100 units/ml), streptomycin
(100 mg/ml), the aminoglycoside G418 (0.5 mg/ml) and hygromycin B
(100 mg/ml). The HSV-1 K26-GFP mutant (strain KOS), carrying a GFPtagged
capsid protein VP26, was provided by P. Desai 40 . The ICP34.5-null virus
D34.5 was constructed by a strategy similar to that used to make the null
mutant 17termA 41 . A 3,333–base pair DpnII fragment containing the gene
encoding ICP34.5 with a nonsense mutation inserted at the sequence encoding
the amino acid at position 30 was transfected into Vero cells together with
infectious DNA from HSV-1 strain 17 (ref. 25). Individual plaques resulting
from this transfection were screened by PCR amplification, followed by
screening by SpeI digestion. After three rounds of plaque purification, viral
stock was generated from which infectious DNA was prepared and ‘marker
rescue’ was done. The ICP34.5-null D34.5 virus was characterized by immunoblot
analysis for ICP34.5 and phosphorylated eIF2a as described 25 . Primary
antibodies used were as follows: rabbit polyclonal antibody to LC3a (anti-LC3a;
AP1801a; Abgent), rabbit polyclonal anti-LC3b (AP1802a; Abgent), rabbit
polyclonal antibody to cleaved LC3b (AP1806a; Abgent), mouse monoclonal
anti-gB (M612449; Fitzgerald), rat anti-LAMP-1 (1D4B; Developmental Studies
Hybridoma Bank), rabbit polyclonal anti-Atg5 (NB-110-53818; Novus Biologicals),
mouse monoclonal anti-tubulin (B-5-1-2; Sigma) and rabbit polyclonal
anti-HSV (RB-1425-A; Neomarkers). The secondary antibodies Alexa Fluor
568–conjugated goat anti-rabbit, Alexa Fluor 633–conjugated goat anti-mouse
and Alexa Fluor 488–conjugated goat anti-mouse were from Invitrogen.
Brefeldin A, bafilomycin A, MG-132, 3-methyladenine and rapamicyn were
from Sigma.
Heat shock, cytokine treatment and infection. For heat-shock treatment,
macrophages (1 10 5 cells per well in 24-well plates) were incubated for 12 h
at 39 1C, followed by a recovery period of 2 h at 37 1C before infection. The
cytokines IL-1b (5 ng/ml; R&D Systems) and IFN-g (200 U/ml; PBL) were
added 18–24 h before infection and were kept in the medium during viral
infection. Macrophages were infected by incubation for 30 min with virus at a
multiplicity of infection of 10. Cells were then washed and were incubated in
fresh medium for a total of 8 h unless indicated otherwise. Drugs were added to
the medium from 2 h after infection until the end of infection at the following
concentrations: brefeldin A, 5 mg/ml; bafilomycin A, 0.5 mM; MG-132, 5 mM;
3-methyladenine, 10 mM; and rapamicyn, 10 mg/ml.
CD8 + T cell hybridoma assay. Mock- or HSV-1-infected macrophages
(2 105 ) were washed in Dulbecco’s PBS and were fixed for 10 min at
23 1C with 1% (wt/vol) paraformadehyde, followed by three washes in
complete DMEM. Antigen-presenting cells were then cultured for 12 h at
37 1C together with 4 105 HSV-2.3.2E2 cells (the b-galactosidase–inducible,
gB-specific CD8 + T cell hybridoma) for analysis of the activation of T cells.
Cells were then washed in Dulbecco’s PBS and lysed (0.125 M Tris base, 0.01 M
cyclohexane diaminotetraacetic acid, 50% (vol/vol) glycerol, 0.025% (vol/vol)
Triton X-100 and 0.003 M dithiothreitol, pH 7.8). A b-galactosidase substrate
buffer (0.001 M MgSO4 7H2O, 0.01 M KCl, 0.39 M NaH2PO4 H2O, 0.6 M
Na2HPO4 7H2O, 100 mM 2-mercaptoethanol and 0.15 mM chlorophenol
red b-D-galactopyranoside, pH 7.8) was added for 2–4 h at 37 1C. Cleavage
of the chromogenic substrate chlorophenol red-b-D-galactopyranoside was
quantified in a spectrophotometer as absorbance at 595 nm.
Immunufluorescence and dansylcadaverin labeling. For immunofluorescence
analysis, control, treated and/or infected macrophages were fixed and made
permeable with a Cytofix/Cytoperm kit according to the manufacturer’s
recommendations (BD Biosciences). Cells were then incubated for 60 min
at 25 1C with anti-LC3, anti-gB or anti-LAMP-1. For analysis of infection with
HSV-1 K26-GFP, infected cells were visualized by detection of GFP fluorescence
at 488 nm. Cells were analyzed with a confocal laser-scanning microscope
(LSM 510Meta Axiovert; Carl Zeiss) or standard Axiophot fluorescent microscope
(Zeiss) or by flow cytometry with a FACSCalibur (BD). For labeling
with dansylcadaverin (Sigma), macrophages left untreated or exposed to IL-1b,
IFN-g or mild heat shock were infected for 8 h and then stained for 15 min at
37 1C with 50 mM dansylcadaverin. Cells were then washed in Dulbecco’s PBS
and were lysed in 200 ml lysis buffer (described above). Total fluorescence was
quantified with a SpectraMax Gemini electron microscopy spectrophotometer
(excitation, 380 nm; emission, 525 nm).
Electron microscopy. For morphological analysis, cells were fixed in 2.5%
(vol/vol) glutaraldehyde and were embedded in Epon (Mecalab). Glucose-
6-phosphatase was detected by electron microscopy cytochemistry as described
42 . For immunocytochemistry after embedding, cells were fixed in 1%
(vol/vol) glutaraldehyde and were embedded at –20 1C in Lowicryl (Canemco).
Lowicryl ultrathin sections were incubated overnight with antibodies and were
visualized by 60 min of incubation with protein A–gold complex (10 nm).
Rabbit anti-LC3 and mouse anti-gB were used at dilution of 1:10.
Analysis with siRNA. Control siRNA (nontargeting; siCONTROL) and siRNA
specific for mouse Atg5 (L-064838-00-0005; ON-TARGETplus SMARTpool)
were from Dharmacon. Cells were transfected with 100 nM siRNA using the
DermaFECT 4 siRNA transfection reagent according to the manufacturer’s
recommendations (Dharmacon). After 24 h, transfection medium was replaced
by complete medium.
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank J. Thibodeau and C. Perreault for critical reading of the manuscript;
K. Rock (University of Massachusetts Medical School) for BMA cells;
W. Heath (University of Melbourne) for the HSV-2.3.2E2 hybridoma; G. Arthur
(University of Manitoba) for the wild-type and Atg5 –/– mouse embryonic
fibroblasts produced by N. Mizushima (Medical and Dental University,
Tokyo); P. Desai (Johns Hopkins University) for the HSV-1 K26-GFP
mutant; and M. Bendayan for assistance with electron microscopy. Supported
by the Canadian Institutes for Health Research (R.L. and M.D.), the Natural
Science and Engineering Research Council of Canada (L.E.), Fonds de la
486 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Recherche en Santé du Québec (L.E.), the US National Institutes of Health
(EY09083) and Research to Prevent Blindness (D.L.)
AUTHOR CONTRIBUTIONS
L.E. planned and did most of the experiments and actively participated in
writing the manuscript; M.C. did the experiments with mouse embryonic
fibroblasts; J.D. maintained viral stocks; C.R. did the technical work for Epon
electron microscopy; A.L. provided technical assistance for immunoblot analysis
and immunofluorescence; D.G. did the immunogold labeling and morphological
quantification; D.A. and D.L. produced the ICP34.5 HSV-1 mutant; C.N.
participated in the planning and development of the antigen-presentation assay
and provided help in writing the manuscript; R.L. provided HSV-1 virus stocks
and expertise with the infection system and helped write the manuscript; and
M.D. planned and directed the work and wrote the manuscript.
COMPETING INTERESTS STATEMENT
The authors declare competing financial interests: details accompany the full-text
HTML version of the paper at http://www.nature.com/natureimmunology/.
Published online at http://www.nature.com/naturegenetics/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Bevan, M.J. Cross-priming for a secondary cytotoxic response to minor H antigens with
H-2 congenic cells which do not cross-react in the cytotoxic assay. J. Exp. Med. 143,
1283–1288 (1976).
2. Deretic, V. Autophagy in innate and adaptive immunity. Trends Immunol. 26, 523–528
(2005).
3. Levine, B. & Deretic, V. Unveiling the roles of autophagy in innate and adaptive
immunity. Nat. Rev. Immunol. 7, 767–777 (2007).
4. Schmid, D. & Munz, C. Innate and adaptive immunity through autophagy. Immunity 27,
11–21 (2007).
5. Andrade, R.M., Wessendarp, M., Gubbels, M.J., Striepen, B. & Subauste, C.S. CD40
induces macrophage anti-Toxoplasma gondii activity by triggering autophagy-dependent
fusion of pathogen-containing vacuoles and lysosomes. J. Clin. Invest. 116,
2366–2377 (2006).
6. Gutierrez, M.G. et al. Autophagy is a defense mechanism inhibiting BCG and Mycobacterium
tuberculosis survival in infected macrophages. Cell 119, 753–766 (2004).
7. Ling, Y.M. et al. Vacuolar and plasma membrane stripping and autophagic elimination
of Toxoplasma gondii in primed effector macrophages. J. Exp. Med. 203, 2063–2071
(2006).
8. Nakagawa, I. et al. Autophagy defends cells against invading group A Streptococcus.
Science 306, 1037–1040 (2004).
9. Ogawa, M. et al. Escape of intracellular Shigella from autophagy. Science 307,
727–731 (2005).
10. Liu, Y. et al. Autophagy regulates programmed cell death during the plant innate
immune response. Cell 121, 567–577 (2005).
11. Orvedahl, A. et al. HSV-1 ICP34.5 confers neurovirulence by targeting the Beclin 1
autophagy protein. Cell Host Microbe 1, 23–35 (2007).
12. Paludan, C. et al. Endogenous MHC class II processing of a viral nuclear antigen after
autophagy. Science 307, 593–596 (2005).
13. Dengjel, J. et al. Autophagy promotes MHC class II presentation of peptides from
intracellular source proteins. Proc. Natl. Acad. Sci. USA 102, 7922–7927
(2005).
14. Bernales, S., Schuck, S. & Walter, P. ER-phagy: selective autophagy of the endoplasmic
reticulum. Autophagy 3, 285–287 (2007).
15. Ackerman, A.L., Kyritsis, C., Tampe, R. & Cresswell, P. Early phagosomes in dendritic
cells form a cellular compartment sufficient for cross presentation of exogenous
antigens. Proc. Natl. Acad. Sci. USA 100, 12889–12894 (2003).
16. Houde, M. et al. Phagosomes are competent organelles for antigen cross-presentation.
Nature 425, 402–406 (2003).
17. Guermonprez, P. et al. ER-phagosome fusion defines an MHC class I cross-presentation
compartment in dendritic cells. Nature 425, 397–402 (2003).
ARTICLES
18. Mueller, S.N. et al. The early expression of glycoprotein B from herpes simplex virus
canbedetectedbyantigen-specificCD8 + Tcells.J. Virol. 77, 2445–2451 (2003).
19. Mueller, S.N., Jones, C.M., Smith, C.M., Heath, W.R. & Carbone, F.R. Rapid cytotoxic
T lymphocyte activation occurs in the draining lymph nodes after cutaneous herpes
simplex virus infection as a result of early antigen presentation and not the presence of
virus. J. Exp. Med. 195, 651–656 (2002).
20. Kuma, A. et al. The role of autophagy during the early neonatal starvation period.
Nature 432, 1032–1036 (2004).
21. Noda, T. & Ohsumi, Y. Tor, a phosphatidylinositol kinase homologue, controls autophagy
in yeast. J. Biol. Chem. 273, 3963–3966 (1998).
22. Komata, T. et al. Mild heat shock induces autophagic growth arrest, but not apoptosis in
U251-MG and U87-MG human malignant glioma cells. J. Neurooncol. 68, 101–111
(2004).
23. Talloczy, Z. et al. Regulation of starvation- and virus-induced autophagy by the eIF2a
kinase signaling pathway. Proc. Natl. Acad. Sci. USA 99, 190–195 (2002).
24. He, B., Gross, M. & Roizman, B. The g134.5 protein of herpes simplex virus 1
complexes with protein phosphatase 1a to dephosphorylate the alpha subunit of the
eukaryotic translation initiation factor 2 and preclude the shutoff of protein synthesis
by double-stranded RNA-activated protein kinase. Proc. Natl. Acad. Sci. USA 94,
843–848 (1997).
25. Alexander, D.E., Ward, S.L., Mizushima, N., Levine, B. & Leib, D.A. Analysis of the role
of autophagy in replication of herpes simplex virus in cell culture. J. Virol. 81,
12128–12134 (2007).
26. Talloczy, Z., Virgin, H.W. IV. & Levine, B. PKR-dependent autophagic degradation of
herpes simplex virus type 1. Autophagy 2, 24–29 (2006).
27. Levine, B. Eating oneself and uninvited guests: autophagy-related pathways in cellular
defense. Cell 120, 159–162 (2005).
28. Kabeya, Y. et al. LC3, a mammalian homologue of yeast Apg8p, is localized in
autophagosome membranes after processing. EMBO J. 19, 5720–5728 (2000).
29. Heldwein, E.E. et al. Crystal structure of glycoprotein B from herpes simplex virus 1.
Science 313, 217–220 (2006).
30. Stannard, L.M., Himmelhoch, S. & Wynchank, S. Intra-nuclear localization of two
envelope proteins, gB and gD, of herpes simplex virus. Arch. Virol. 141, 505–524
(1996).
31. Farnsworth, A. et al. Herpes simplex virus glycoproteins gB and gH function in fusion
between the virion envelope and the outer nuclear membrane. Proc. Natl. Acad. Sci.
USA 104, 10187–10192 (2007).
32. Khalkhali-Ellis, Z. et al. IFN-g regulation of vacuolar pH, cathepsin D processing
and autophagy in mammary epithelial cells. J. Cell. Biochem. 105, 208–218
(2008).
33. Griffin, T.A. et al. Immunoproteasome assembly: cooperative incorporation of interferon
g (IFN-g)-inducible subunits. J. Exp. Med. 187, 97–104(1998).
34. Epperson, D.E. et al. Cytokines increase transporter in antigen processing-1 expression
more rapidly than HLA class I expression in endothelial cells. J. Immunol. 149,
3297–3301 (1992).
35. Conti, B., Tabarean, I., Andrei, C. & Bartfai, T. Cytokines and fever. Front. Biosci. 9,
1433–1449 (2004).
36. Jarvis, M.A. et al. Human cytomegalovirus attenuates interleukin-1b and tumor
necrosis factor a proinflammatory signaling by inhibition of NF-kB activation.
J. Virol. 80, 5588–5598 (2006).
37. Kovacsovics-Bankowski, M. & Rock, K.L. A phagosome-to-cytosol pathway for exogenous
antigens presented on MHC class I molecules. Science 267, 243–246
(1995).
38. Ackerman, A.L., Giodini, A. & Cresswell, P. A role for the endoplasmic reticulum protein
retrotranslocation machinery during crosspresentation by dendritic cells. Immunity 25,
607–617 (2006).
39. Jackson, W.T. et al. Subversion of cellular autophagosomal machinery by RNA viruses.
PLoS Biol. 3, e156 (2005).
40. Desai, P. & Person, S. Incorporation of the green fluorescent protein into the herpes
simplex virus type 1 capsid. J. Virol. 72, 7563–7568 (1998).
41. Bolovan, C.A., Sawtell, N.M. & Thompson, R.L. ICP34.5 mutants of herpes simplex
virus type 1 strain 17syn+ are attenuated for neurovirulence in mice and for replication
in confluent primary mouse embryo cell cultures. J. Virol. 68, 48–55 (1994).
42. Griffiths, G., Quinn, P. & Warren, G. Dissection of the Golgi complex. I. Monensin
inhibits the transport of viral membrane proteins from medial to trans Golgi cisternae in
baby hamster kidney cells infected with Semliki Forest virus. J. Cell Biol. 96, 835–850
(1983).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 487

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Cross-presentation of viral and self antigens by
skin-derived CD103 + dendritic cells
Sammy Bedoui 1–3 , Paul G Whitney 1,3 , Jason Waithman 1–3 , Liv Eidsmo 1 , Linda Wakim 1 , Irina Caminschi 2 ,
Rhys S Allan 1 , Magdalena Wojtasiak 1 , Ken Shortman 2 , Francis R Carbone 1 , Andrew G Brooks 1 &
William R Heath 1,2
Skin-derived dendritic cells (DCs) include Langerhans cells, classical dermal DCs and a langerin-positive CD103 + dermal subset.
We examined their involvement in the presentation of skin-associated viral and self antigens. Only the CD103 + subset efficiently
presented antigens of herpes simplex virus type 1 to naive CD8 + T cells, although all subsets presented these antigens to CD4 +
T cells. This showed that CD103 + DCs were the migratory subset most efficient at processing viral antigens into the major
histocompatibility complex class I pathway, potentially through cross-presentation. This was supported by data showing only
CD103 + DCs efficiently cross-presented skin-derived self antigens. This indicates CD103 + DCs are the main migratory subtype
able to cross-present viral and self antigens, which identifies another level of specialization for skin DCs.
Dendritic cells (DCs) serve an essential antigen-presenting function in
the initiation of T cell responses 1,2 . DCs can be categorized into many
independent subsets that seem to represent end-stage populations 3–5 .
One main subclassification can be made of plasmacytoid and conventional
DCs. In the spleen, both plasmacytoid DCs and at least three
subsets of conventional DCs can be found in the steady state, the latter
being loosely categorized as CD8a + DCs and CD8a – DCs, and the
CD8a – DCs being further classified into CD4 + and CD4 – populations.
These DCs enter the spleen from the blood either as mature plasmacytoid
DCs or as precursors of conventional DCs. The lymph nodes
also contain these blood-derived DC subsets but in addition have a
cohort of DCs that have migrated from peripheral tissues 6 . Depending
on the site of drainage for a particular lymph node, it will contain a
variety of migratory DCs. For skin-draining lymph nodes, the migratory
population consists of both epidermis-derived Langerhans cells
and dermis-derived DCs. This represents an abbreviated list of DC
subsets, with many intricacies associated with particular tissues, but it
can be further extended to include monocyte-derived DCs generated
in inflammatory conditions such as infection 4,7 .
Although a variety of DC subsets have been identified, understanding
of the specific functions of individual groups, particularly in
the skin, where at least three different subsets reside, is limited. Clearly,
migratory DCs are critical for trafficking antigen from peripheral sites
to the lymph nodes 8,9 , and plasmacytoid DCs are a chief source of
interferon-a during viral infection 10,11 . One property that varies
extensively among subsets is the ability to introduce exogenous
antigens into the major histocompatibility complex (MHC) class I
Received 10 December 2008; accepted 9 March 2009; published online 6 April 2009; doi:10.1038/ni.1724
pathway, a process referred to as ‘cross-presentation’ 12–14 . In particular,
CD8a + DCs are the dominant splenic population responsible for
cross-presentation of cell-associated antigens 12,14 or those antigens
targeted by monoclonal antibodies 13 . This same subset has been
strongly suggested to be involved in the presentation of viral antigens
for many different infections 15–19 , with some evidence that this relates
to their ability to cross-present 20 .
The importance of CD8a + DCs in viral immunity has been
emphasized by studies examining the initiation of CD8 + T cell
immunity to infection of the skin with herpes simplex virus type 1
(HSV-1) 9,16 . These studies have shown that CD8a + DCs are the sole
DC subset responsible for antigen presentation in the draining lymph
nodes shortly after infection, which challenges the long-held paradigm
that Langerhans cells control immunity to epidermal antigens. The
limited involvement of Langerhans cells has been emphasized by
examination of cytotoxic T lymphocyte (CTL) immunity in bone
marrow chimeras in which radioresistant Langerhans cells are the only
DCs expressing the correct MHC-restriction elements 16 .Inthiscase,
CTL immunity is abrogated, which further challenges the importance
of Langerhans cells in the initiation of skin immunity. This view has
been reinforced for skin infection with other viruses, such as vaccinia
18,21 and influenza 18 , although there is some evidence for the
involvement of skin-derived DCs in the generation of CTL immunity
to lentivirus infection 21 , the only noncytopathic virus tested.
Several subsequent studies have provided some evidence that
Langerhans cells contribute to immunity induced by contact
sensitization 22 and to self tolerance 23 , although even these conclusions
1 The Department of Microbiology and Immunology, The University of Melbourne, Parkville, Victoria, Australia. 2 The Walter and Eliza Hall Institute of Medical Research,
Parkville, Victoria, Australia. 3 These authors contributed equally to this work. Correspondence should be addressed to W.R.H. (wrheath@unimelb.edu.au), F.R.C.
(fcarbone@unimelb.edu.au) or A.G.B. (agbrooks@unimelb.edu.au).
488 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a
b
c
1° site, day 2
HSV-1 DAPI
2° site, day 2
are not universal 24–26 . Some studies have used transgenic expression of
fluorescent proteins and/or the diphtheria toxin receptor under
control of the promoter of the gene encoding langerin for the
identification or depletion of Langerhans cells 22,24–26 . However, this
approach has led to the identification of a second skin-associated
population of langerin-positive cells that, unlike Langerhans cells,
are radiosensitive 26–28 and, in contrast to other skin-derived DC
populations, express the integrin CD103 (ref. 26). The identification
of this previously unknown population of langerin-positive cells
suggests a need to reevaluate their involvement in immunity to
HSV-1 (ref. 27).
Here we examine the contribution of this migratory CD103 + DC
population to the presentation of epidermis-expressed self antigen or
to the presentation of HSV-1 antigen during infection of the skin. We
provide evidence that despite the presence of three distinct DC
subtypes in the skin, CD103 + DCs dominated the MHC class I–
restricted cross-presentation of both pathogen and self antigens. This
provides new insight into the specialization of DC subtypes in the skin
and suggests that the CD103 + DC subset, present in various tissues,
may be broadly responsible for generating CTL immunity and
tolerance to tissue-associated pathogens and antigens, respectively.
RESULTS
Two phases of acute HSV-1 infection
Acute cutaneous infection with HSV-1 can be achieved by scarification
of mouse flank skin 29,30 . Infection initially results in viral replication
in keratinocytes surrounding the site of scarification (Fig. 1a), before
the virus enters neuronal cells that innervate this area. HSV-1 then
travels by retrograde axonal transport to the cell body, where it
replicates extensively, spreading throughout several local dorsal root
ganglia. By days 3–4 of infection, virions begin to travel by anterograde
transport along the many axons derived from the infected
ganglia (Fig. 1b), reaching the epidermis, where they infect the entire
dermatome innervated by the ganglia. In this site, HSV-1 again
replicates extensively in epidermal cells (Fig. 1c), causing a zosteriform
band beginning at day 6 after infection. Of critical relevance here,
HSV-1 has two phases of acute viral replication in the skin: a primary
infection that is limited to the site of scarification, and a secondary
growth phase involving the entire innervated dermatome. The dose of
virus used to infect mice had little influence on viral titer at either the
ARTICLES
Figure 1 Progress of infection after inoculation of HSV-1 on scarified flank
skin. (a) Confocal microscopy (left) and direct photography (right; box
outlines primary (11) site) of HSV-1 in flank skin of C57BL/6 mice on day 2
after infection with HSV-1 by flank scarification. Left: sections 8 mm in
thickness acquired with a 10 objective; red, viral proteins; blue, nuclei
(stained with the DNA-intercalating dye DAPI). Scale bar, 100 mm.
(b) Confocal microscopy (z-stack images) of nerves expressing viral proteins
on day 4 of infection as described in a, assessing five to seven sections
1 mm in thickness acquired with a 63 oil objective and stacked according
to maximum intensity. Scale bar, 20 mm. (c) Confocal microscopy of flank
skin on day 5 of infection as described in a, assessing sections 8 mm in
thickness acquired with a 10 objective (left), or photography of flank skin
on day 7 of infection as described in a (right; box outlines secondary (21)
site). Scale bar, 100 mm. For these studies, uninfected skin was examined
visually and control serum was used to confirm the validity of positive
staining. No viral staining was detected in healthy skin. Areas on either
side of the primary infection site in a represent uninfected regions that
act as an internal negative control. Data are representative of at least
three experiments.
primary site (day 2) or the secondary site (day 5), but did affect the
chance of being infected (Supplementary Fig. 1 online).
T cell activation occurs in phases
After infection of the flank with HSV-1, antigen presentation occurs
rapidly in the draining brachial lymph nodes, with strong presentation
evident at 48 h (ref. 16). Because viral recrudescence causes a second
phase of infection of the entire skin dermatome beginning about 72 h
after infection 30 , we investigated whether this led to a second wave of
antigen presentation. To monitor antigen presentation after flank
infection, we tracked the proliferation of CD8 + gBT-I T cells specific
for an H-2K b -restricted epitope of glycoprotein B (gB). We infected
C57BL/6 mice with HSV-1 by scarification of the upper flank and
then, on day 2 or day 5 after infection, intravenously injected the mice
with gBT-I cells labeled with the cytosolic dye CFSE to monitor
proliferation 42 h later. We chose the 42-hour time frame to allow
proliferation without sufficient time for T cells to exit the lymph node
in which they responded. In this way, any proliferation detected would
mirror antigen presentation in that lymph node. With this approach,
T cell proliferation was evident in the brachial lymph nodes but not in
the closely associated axillary lymph nodes on day 2 (Fig. 2a). On day
5, however, viral recrudescence led to antigen presentation in both the
brachial and axillary lymph nodes, as measured by proliferation of
gBT-I cells in each site. We found no such proliferation in the
mesenteric lymph nodes (Supplementary Fig. 2 online), which
indicated that the response was not systemic. As the delayed presentation
in the axillary lymph nodes coincided with a greater antigen load
associated with viral recrudescence, antigen-bearing DCs may have
drained from the secondary site directly to the axillary lymph nodes
or, alternatively, may have passed through the brachial lymph node,
overflowing into the axillary lymph node. To distinguish between
those possibilities, we altered the initial site of scarification to the
lower flank. In this case, the pattern of antigen presentation was
reversed: T cell proliferation occurred initially in the axillary but not
the brachial lymph nodes on day 2 and occurred in both lymph nodes
on day 5 (Fig. 2b). These findings indicate that the first phase of
infection initiated antigen presentation in a single lymph node,
whereas the second phase contributed a later wave of presentation
involving at least two lymph nodes.
We confirmed the view that the second phase of T cell activation
required viral recrudescence by using a thymidine kinase–deficient
form of HSV-1 (HSV-TK – ), which does not replicate in the dorsal root
ganglia or form secondary lesions. We found that this thymidine
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 489

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
Ax LN
Br LN
Day 2 Day 5
6.5 ± 2.1 82.8 ± 4.8
48.4 ± 11.0 75.1 ± 5.7
CFSE
Ax LN
Br LN
kinase–deficient form of the virus had a primary growth phase in the
skin similar to that of wild-type virus, with titers diverging only after
the wild-type virus returned from the ganglia during the zosteriform
phase of disease (Supplementary Fig. 3 online). Infection of the upper
flank with HSV-TK – did not result in T cell activation in the axillary
lymph nodes on day 2 or day 5, although activation in the brachial
lymph nodes was evident on days 2 and 5 as a consequence of the
primary infection (Fig. 2c).
Presentation by migratory DCs after viral recrudescence
There are several subsets of DCs; studies have indicated involvement of
a lymph node–resident population of CD8a + DCs in the initiation of
T cell activation after HSV-1 infection 9,16 . Those studies and our
studies here used ex vivo isolation of DCs to show extensive presentation
by CD8a + DCs on day 2 (refs. 9,16; Fig. 3a,b) with no evidence of
presentation by any DC subset on day 4 (Fig. 3a,c). As in the earlier
studies 9,16 , we prepared DCs by first depleting lymph node suspensions
of other cell types and then staining for CD11c, CD205 and
CD8 (Fig. 3). Then, we sorted DCs by flow cytometry on the basis
of expression of CD11c and various combinations of CD8a and
CD205 (Fig. 3a); Langerhans cells were CD205 hi CD8a – ,dermalDCs
were CD205 int CD8a – , lymph node–resident CD8a + DCs were
CD205 int CD8a + , and we called the remaining mixed population
‘double-negative DCs’. These studies initially led us to assume that
presentation had mostly ended by day 4 (Fig. 3c), although it is
b
Day 2 Day 5
53.9 ± 8.9 77.9 ± 7.1
3.5 ± 1.5 76.1 ± 7.7
CFSE
c
Ax LN
Br LN
Day 2 Day 5
6.4 ± 3.2 10.5 ± 5.0
38.4 ± 8.7 57.4 ± 10.2
Figure 2 In vivo antigen presentation to HSV-specific CD8 + T cells after cutaneous HSV-1 infection. (a,b) Proliferation in the axillary lymph nodes
(Ax LN; top) and brachial lymph nodes (Br LN; bottom) at 42 h after transfer of 1 10 6 CFSE-labeled gBT-I cells intravenously into mice infected with
HSV-1 on the upper flank (a) or lower flank (b), 2 d earlier (left) or 5 d earlier (right). Numbers in plots indicate percent proliferated cells (± s.e.m.).
Data are from three independent experiments. (c) Proliferation analysis as described in a, but for mice infected with HSV-TK – . Data are from two
independent experiments.
CD205
a
LC
CD8α +
DC
dDC
DN DC
CD8
b
Divided gBT-I/well (×10 3 )
Day 2, br LN
30
20
10
c
Divided gBT-I/well (×10 3 )
Day 4, br LN
30
20
10
CFSE
worth noting that studies using a T cell hybridoma that produces
b-galactosidase after stimulation suggest that presentation may extend
beyond this time 9 .
Given the evidence presented above for viral recrudescence–
dependent T cell activation on day 5 in the axillary lymph nodes
(Fig. 2a), we questioned whether ex vivo presentation by DC subsets
might be evident on day 5 in the axillary lymph nodes. This was of
particular interest, as such secondary-site-dependent presentation
would not require mechanical scarring of the skin but instead
would be a consequence of the natural movement of virus from the
nerves to the epidermis. We infected mice by scarification of the upper
flank and then, 5 d later, collected axillary lymph nodes and assessed
the isolated DC subsets for their ability to stimulate CD8 + gBT-I cells
in vitro. In this case, extensive presentation was evident in the axillary
lymph nodes on day 5 (Fig. 3d), but it was not apparent when we used
nonrecrudescent HSV-TK – as the infectious agent (Supplementary
Fig. 4 online). In contrast to the presentation on day 2 in the brachial
lymph node, dermal DCs seemed to dominate presentation in the
axillary lymph node at this later time point. Presentation by CD8a +
DCs still occurred at an amount similar to that in the brachial lymph
nodes on day 2, and there was also some stimulation by Langerhans
cells, but it was difficult to exclude the possibility of contamination of
this population by the highly stimulatory dermal DCs that overlapped
in CD205 expression. Consistent with the proposal that a new wave of
DCs entered the lymph nodes from the secondary site of infection,
d
Divided gBT-I/well (×10 3 )
Day 5, ax LN
120
e
Divided gBT-I/well (×10 3 )
LC
dDC
Day 5, br LN
30
CD8α + DC
DN DC
0
0 1.9 3.8 7.5 15 30 0 1.9 3.8 7.5 15 30 0 1.9 3.8 7.5 15 30 0 1.9 3.8 7.5 15 30
DC/well (×10 3 0
0
0
)
DC/well (×10 3 ) DC/well (×10 3 ) DC/well (×10 3 )
Figure 3 Migratory DCs are involved in antigen presentation to CD8 + T cells after secondary viral infection of the skin. (a) Purification of DC subsets from
lymph nodes of infected mice: after DC enrichment, expression of CD205 and CD8a by CD11c + DCs was assessed, and Langerhans cells (LC; CD205 hi CD8a – ),
dermal DCs (dDC; CD205 int CD8a – ), double-negative DCs (DN DC; CD205 – CD8a – )andCD8a + DCs (CD205 int CD8a + ) were isolated with a cell sorter after
gating on CD11c + cells. (b–e) Proliferation of 5 10 4 CFSE-labeled gBT-I cells cultured for 60 h together with serial dilutions of DC subsets (identified
as in a) isolated from brachial lymph nodes (b,c,e) or axillary lymph nodes (d) of mice infected 2 d earlier (b), 4 d earlier (c) or 5 d earlier (d,e). Day-4
assays were done on the same day as day-2 assays. Data are pooled from three to four individual experiments (mean ± s.e.m.).
490 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
90
60
30
20
10

© 2009 Nature America, Inc. All rights reserved.
CD11c
CD205
a
CD8
CD103
dDC
CD326
LC
CD8α +
DC
CD103 +
DC
Divided gBT-I/well (×10 3 )
Divided gBT-I/well (×10 3 )
b
c
25
20
15
10
0
DC/well (×10 3 0 1.9 3.8 7.5 15
)
50
40
20
LC
dDC
CD8α + CD103
DC
+ DC
analysis of presentation in the brachial lymph nodes on day 5 showed
presentation by dermal DCs (Fig. 3e) that was not evident on day 4
(Fig. 3c). In this case, presentation by CD8a + DCs was minimal and
presentation by dermal DCs was somewhat less than in the axillary
lymph nodes, possibly because effector CTLs were already actively
killing antigen-presenting cells in this site.
CD103 + DCs present after viral recrudescence
Several groups have reported a third population of skin-derived DCs
that, like Langerhans cells, express langerin 26–28 but, unlike other skinderived
DCs, express CD103 (ref. 26). These cells are restricted mainly
to the dermis and have therefore been called ‘langerin-positive dermal
DCs’, but for simplicity, we will call them ‘CD103 + DCs’ here. These
DCs are resident in the skin and migrate to the lymph nodes after the
skin is painted with tetramethylrhodamine isothiocyanate plus irritant
26 . Both dermal DCs and Langerhans cells migrate after skin is
painted with fluorescein isothiocyanate (FITC) 9 ; here we used painting
with FITC to show that CD103 + DCs migrated in the context of
irritant or infection with HSV-1 (Supplementary Fig. 5 online). These
studies also confirmed the migratory nature of Langerhans cells and
the classical CD103 – dermal DCs and showed a lack of migration by
lymph node–resident CD8a + DCs.
Given the strong presentation by dermal DCs on day 5 reported
above (Fig. 3d), it became imperative to examine the antigenpresenting
activity of contaminating CD103 + DCs in these groups.
To achieve this, we included CD103 staining in our sorting protocol.
Because of limitations in the number of groups that could be sorted,
in subsequent studies we did not include the double-negative DCs
identified above, which represented mainly lymph node–resident
CD8a – DCs that did not seem to contribute to HSV-1-specific
5
30
10
0
Day 2, br LN
Day 5, ax LN
DC/well (×10 3 0 1.9 3.8 7.5 15
)
Figure 5 Many DC subsets present MHC class II–restricted viral antigen to
gDT-II CD4 + T cells. (a,b) Proliferation of 5 10 4 CFSE-labeled HSV-1specific
CD4 + T cells (gDT-II) after 60 h of culture together with serial
dilutions of DC subsets isolated (as described in Fig. 4a) frombrachial
lymph nodes (a) or axillary lymph nodes (b) of mice infected 2 d earlier.
(c,d) Proliferation analysis as described in a,b for DC subsets isolated
from brachial lymph nodes (c) or axillary lymph nodes (d) of mice infected
5 d earlier. Data are pooled from two to four individual experiments
(mean ± s.e.m.).
Figure 4 CD103 + DCs present HSV-1 antigens to CD8 + T cells after
secondary viral infection of the skin. (a) Gating strategy for isolation of
CD103 + DCs: after DC enrichment, CD8a + DCs were purified on the
basis of expression of CD11c and CD8a (top; right gate); expression of
CD205 and CD103 (middle) was assessed in CD11c + CD8a – cells (top; left
gate) for the isolation of CD103 + DCs (middle; right gate); and CD103 –
CD11c + CD8a – DCs (middle; left gate) were categorized as Langerhans cells
(CD326 hi ) and dermal DCs (CD326 – ) on the basis of CD326 expression
(bottom; sort gates outlined). Cells of DC subsets purified from lymph
nodes of naive and infected mice are counted in Supplementary Figure 8.
(b,c) Proliferation of 5 10 4 CFSE-labeled gBT-I cells cultured for 60 h
together with serial dilutions of DC subsets (identified as in a) sorted from
brachial lymph nodes 2 d after infection (b) or from axillary lymph nodes
5 d after infection (c). Data are pooled from two to four individual
experiments (mean ± s.e.m.).
CD8 + T cell activation. In this sorting scheme (Fig. 4a), we isolated
CD8a + DCs directly from the CD11c + cells, then isolated CD103 + cells
and separated Langerhans cells and dermal DCs on the basis of CD205
expression and differences in expression of CD326 (Ep-CAM;
expressed by Langerhans cells). We isolated CD8a + DCs first, as
they had low expression of CD103 (Supplementary Fig. 6 online)
that would otherwise potentially result in contamination of the
CD103 + population by this subset.
After sorting the cells as described above, we examined these DC
subsets for their presentation of antigen on day 2 in the brachial
lymph nodes (Fig. 4b) and on day 5 in the axillary lymph nodes
(Fig. 4c), which represent the primary and secondary sites, respectively.
This showed that CD103 + DCs were the dominant migratory
DCs that presented viral antigen on day 5 in the axillary lymph nodes,
whereas little activity by any migratory DC was evident on day 2 in the
brachial lymph nodes. In contrast, CD8a + DCs presented viral antigen
during both phases. Some presentation by dermal DCs and, to a lesser
extent, Langerhans cells was also present on day 5 in the axillary
lymph nodes, but it is difficult to exclude the possibility that this was
not due to contaminating CD103 + DCs, which showed detectable
presentation even with as few as 467 cells (data not shown). These
findings suggest that the CD103 + DCs were dominant among migratory
DCs for the ability to cross-present.
To formally address whether CD103 + DCs that had migrated from
the skin presented viral antigen, we examined presentation on day 5 in
the axillary lymph nodes after labeling the flank skin with FITC at 2 d
Divided gDT-II/well (×10 3 )
Divided gDT-II/well (×10 3 )
a
c
Day 2, br LN
25
20
15
10
5
0
0 1.9 3.8 7.5 15
Day 5, br LN
40
30
20
10
DC/well (×10 3 )
LC
dDC
CD8α + CD103
DC
+ DC
0 1.9 3.8 7.5 15
DC/well (×10 3 )
0
0 1.9 3.8 7.5 15 0 1.9 3.8 7.5 15
DC/well (×10 3 ) DC/well (×10 3 0
)
Divided gDT-II/well (×10 3 )
Divided gDT-II/well (×10 3 )
b
d
25
20
15
10
5
40
30
20
10
Day 2, ax LN
Day 5, ax LN
ARTICLES
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 491

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Figure 6 CD103 + DCs present
skin-derived self antigen to
CD8 + T cells. Proliferation of
5 10 4 CFSE-labeled OVAspecific
CD8 + T cells (OT-I)
after 60 h of culture together
with serial dilutions of DC
subsets isolated (as described
in Fig. 4a) from the skindraining
lymph nodes of K5mOVA
mice. Data are pooled
from two individual experiments
(mean ± s.e.m.).
Divided OT-I/well (×10 3 )
LC
dDC
CD8α + CD103
DC
+ DC
after infection (Supplementary Fig. 7 online). In this case,
FITC + CD103 + DCs, which had migrated from the skin after viral
recrudescence, presented viral antigens to CD8 + T cells. Notably,
FITC – CD103 + DCs also efficiently presented viral antigen. Presentation
by this latter population most likely reflected the fact that only a
small proportion of the CD103 + DCs are amenable to labeling
with FITC. That idea was reinforced by the finding that although
the number of CD103 + DCs in the axillary lymph nodes
increased fourfold between day 2 and day 5 (Supplementary Fig. 8
online), only about 6% of the cells were labeled with FITC (Supplementary
Fig. 5e).
Access to viral antigen is not limited to CD103 + DCs
‘Preferential’ MHC class I–restricted presentation of viral antigen by
CD103 + DCs could relate to access to viral antigens rather than a
dominant ability to cross-present. To investigate that possibility, we
explored whether other migratory DCs had access to viral antigen for
MHC class II–restricted presentation. To achieve this, we generated a
new line of T cell antigen receptor–transgenic mice that produce CD4 +
T cells specific for glycoprotein D (gD) of HSV-1 (gDT-II mice;
P.G.W., S.B., G. Davey, W.R.H., F.R.C., A.G.B., unpublished observations).
We used CD4 + gDT-II T cells as responders to examine
presentation in the brachial and axillary lymph nodes on days 2 and
5(Fig. 5). This showed that all DCs could present antigens to virusspecific
CD4 + T cells on day 5 in the axillary lymph nodes (Fig. 5d)
and brachial lymph nodes (Fig. 5c) and all but the Langerhans cells
presented on day 2 in the brachial lymph nodes (Fig. 5a). As expected,
there was no presentation on day 2 in the axillary lymph nodes
(Fig. 5b). These studies support the view that the dominant ability of
CD103 + DCs to present MHC class I–restricted antigens on day 5 is
related to their ability to process viral antigen into this pathway, rather
than their ability to access the antigen. This was especially evident by
comparison of MHC class I– and class II–restricted presentation on
day 5 by dermal DCs and CD103 + DCs. Analysis of cytokines
produced by gDT-II cells on day 5 in response to viral antigens
presented by the various DC subsets showed a bias toward production
of T helper type 1 cytokines (interferon-g, interleukin 2 and tumor
necrosis factor) in amounts that mirrored proliferation but did not
differ among DC subtypes (Supplementary Fig. 9 online).
CD103 + DCs cross-present epidermal antigens
It is difficult to distinguish between cross-presentation and direct
infection of DCs during viral responses. As an alternative approach to
examine the cross-presentation ability of skin-derived migratory DCs,
we used the DC-sorting protocol described above for isolation of
CD103 + DCs to examine the presentation of a membrane-associated
form of ovalbumin (OVA) by DCs from mice expressing OVA in
epidermal keratinocytes under control of the promoter of the gene
10
8
6
4
2
0
DC/well (×10 3 0 1.6 3.1 6.3 12.5 25
)
encoding keratin 5 (K5-mOVA mice). In accordance with our studies
of HSV-1 infection, this showed that CD103 + DCs were the dominant
DC population presenting MHC class I–restricted epidermal skin
antigen in the steady state (Fig. 6). As this antigen was expressed by
keratinocytes but not by DCs 23 , it was apparent that in this case,
CD103 + DCs used cross-presentation.
DISCUSSION
This study has shown that langerin-positive CD103 + dermal DCs are
the dominant migratory DCs of the skin that present MHC class I–
restricted antigens. This dominance most likely relates to their ability
to cross-present, at least for the types of antigens tested here. An
alternative explanation is that presentation of viral antigens by
CD103 + DCs occurs as a consequence of ‘preferential’ infection.
Definitive resolution of this possibility is problematic by most
approaches available, as it is almost impossible to distinguish between
an infected cell and a cell that has captured infected cellular material.
Although some doubt may exist over whether CD103 + DCs use crosspresentation
for HSV-1 antigens, it is likely that this function was
responsible for the presentation of OVA in K5-mOVA mice. The
failure of classical dermal DCs to cross-present OVA in this case might
be explained by limited access to antigen, as OVA is expressed by
epidermal cells. This explanation cannot apply to Langerhans cells,
however, as they reside directly in the region of OVA expression in
K5-mOVA mice 31 . Unfortunately, we have not been able to detect
presentation of OVA on MHC class II by any DC subset in K5-mOVA
mice, which precludes a definitive statement about access by dermal
DCs. However, the ability of all migratory DCs to present viral
antigens on MHC class II during HSV infection of the skin indicates
that all had access to viral antigens, yet only the CD103 + DCs
presented these antigens efficiently on MHC class I.
It is worth considering whether migratory DCs might access viral
antigens in the draining lymph nodes rather than the skin. As mature
DCs are very poor at capturing antigen for presentation 20 , and
migratory DCs have a mature phenotype in the lymph nodes, it is
most likely that viral antigens presented by migratory DCs are
captured in the skin. More importantly, if CD103 + DCs captured
viral antigen in the lymph nodes, then they might be expected to
cross-present it efficiently on both days 2 and 5, as reported for the
CD8a + DCs, yet cross-presentation by CD103 + DCs was absent on day
2. Furthermore, the ability of CD103 + DCs to cross-present skinderived
OVA in the K5-mOVA mice indicated that this is related to
capture in the skin, as capture in the lymph nodes should lead to
similar access and cross-presentation by CD8a + DCs. Although the
possibility that virus was captured in the draining lymph nodes cannot
be completely excluded, the finding that all migratory DCs presented
MHC class II–restricted viral antigens yet only CD103 + DCs presented
these antigens in an MHC class I–restricted way supports the main
point of this report, which is that CD103 + DCs are the dominant
migratory subset that cross-presents viral antigen. Our data are mostly
compatible with published studies of HSV-2 (ref. 32) and HSV-1
(ref. 33) showing that a dermal-like population of DCs provides a
principal antigen-presenting contribution to both CD4 + and CD8 +
T cell antiviral responses. Notably, we specifically separated CD103 +
DCs from classical dermal DCs and identified the CD103 + DCsasthe
key dominant cross-presenting migratory population.
A corollary of the conclusion that CD103 + DCs are the dominant
cross-presenting population of migratory DCs from the skin is that
both Langerhans cells and classical dermal DCs are inefficient at crosspresentation.
Earlier studies have provided evidence that Langerhans
cells do not prime CD8 + T cell responses to HSV 9,16 .Thatdoesnot
492 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
exclude the possibility that they could transport virus or even
present viral antigens to CD4 + T cells, as we have now shown here,
which raises the possibility that these classical epidermal DCs contribute
to priming. Earlier studies linked Langerhans cells to the crosspresentation
of both cellular and soluble OVA expressed or introduced
into the skin, respectively 23,34 . However, the authors of those studies
were unaware of CD103 + DCs, which also express langerin. Thus,
much of the earlier data indicating involvement of Langerhans cells
can be explained by functions associated with contaminating CD103 +
DCs. Notably, however, the use of bone marrow chimeras has shown
that radioresistant cells are able to present antigen and induce deletion
of naive CD8 + T cells in the K5-mOVA model, which suggests some
role for Langerhans cells 23 . Similarly, here we have shown weak but
detectable MHC class I–restricted presentation of either HSV-1 or
OVA by Langerhans cells, although, as in the other studies, it was
difficult to exclude the possibility of residual presentation by contaminating
CD103 + DCs.
The disparate cross-presenting abilities of classical dermal DCs and
CD103 + DCs is reminiscent of differences between lymphoid tissue–
resident CD8a + and CD8a – DC subtypes, in which CD8a + DCs
dominate 13,35 . Notably, CD8a + DCs and CD103 + DCs have many
similarities 26 , including expression of langerin and CD103, relatively
low expression of CD11b, responsiveness to Toll-like receptor 3
ligands 36 and ablation in mice deficient for the transcription factor
Batf3 (ref. 37). Given such similarities, it is plausible that the
precursors of these DCs are related, with differences between
CD103 + DCs and CD8a + DCs being determined by the type of tissue
their precursor seeds (lymphoid versus skin).
Although langerin-positive CD103 + CD11b lo DCs have been identified
in the skin only recently, several studies have identified a similar
DC type in other organs or their draining lymph nodes 19,26,36 .
Examination of the presentation of influenza viral antigens by
those DC subsets after lung infection has shown that both CD8a +
DCs and CD103 + DCs present viral antigens to CD8 + Tcells 18,38 ,
although only the CD103 + subset migrates from the lung 18 .Wehave
defined this migratory subset mainly on the basis of their low
expression of CD11b relative to that of other migratory DCs. Similar
DCs have been found in lymph nodes draining the kidney, liver,
pancreas and lung 18 . One group, using langerin and CD103 to
identify this subset in the lungs and mediastinal lymph nodes, has
found that whereas they present viral antigens to CD8 + T cells, they
are not the main subset containing viral material; this is present in
both plasmacytoid DCs and a migratory CD11b + DC population 38 .
Such findings are consistent with our conclusion that MHC class I–
restricted presentation by CD103 + DCs is not a consequence of
‘preferential’ access to viral material but instead is a consequence of
their dominant cross-presenting ability. Our view is further substantiated
by studies examining cross-presentation in the lung of introduced
innocuous antigen in the form of soluble OVA 39 . Again, these
studies have found CD103 + DCs are the dominant subset crosspresenting
OVA, whereas other CD11b + DCs ‘preferentially’ present it
to CD4 + T cells 39 . Another group extensively examining the presentation
of particulate and soluble OVA introduced into the lung has
concluded that CD103 + DCs are the most effective at cross-presenting
antigens associated with latex beads 40 . However, in that case, ‘preferential’
capture might have been responsible. Our study has
demonstrated the dominant cross-presenting ability of CD103 +
DCs among the migratory populations in the skin but also raises
the possibility that this DC subtype might be broadly represented in
many tissues, mainly for its function in the generation of CD8 + T cell
immunity and tolerance.
ARTICLES
The ability of CD103 + DCs to present viral antigens raises the issue
of whether they are fully able to prime effective viral immunity. It is
difficult to address this issue for HSV-1 infection, as CD8a + DCs also
contribute to presentation. It is evident that CD8a + DCs can prime
CTL immunity on their own when HSV-1 is introduced intravenously
18 or when infection is cut short by excision of the site of
infection at 8 h (ref. 41). Thus, it is unlikely that removal of CD103 +
DCs by approaches such as use of the langerin–diphtheria toxin
receptor system 22,24–26 will substantially ablate priming. To address
this issue directly, a system is needed that will allow removal of CD8a +
DCs while leaving CD103 + DCs intact but able to be depleted.
As a final point of discussion, it is worth revisiting the function of
Langerhans cells in immunity to HSV-1. It has been shown that
Langerhans cells are not able to generate CTL immunity to this
virus 16 . That finding is consistent with their poor ability to present
viral antigen to CD8 + T cells shown here on both day 2 and day 5.
Why Langerhans cells fail to effectively prime HSV-1-specific CTL
responses remains a mystery, but if they are poorly endowed with the
ability to cross-present, then presentation of MHC class I–restricted
viral antigens would depend entirely on infection and access to the
classical MHC class I pathway. As extensive evidence indicates that
DCs infected with HSV-1 respond poorly to chemokines 42 ,failto
upregulate costimulatory molecules 42 , die rapidly and are poorly
stimulatory 43,44 , it is perhaps not unexpected that Langerhans cells,
even if infected, would be ineffective at priming HSV-specific CTL
immunity. Whether infected and dying Langerhans cells represent a
likely source of antigenic material for CD103 + DCs to cross-present as
they attempt to migrate to the draining lymph nodes remains an open
issue. Our data indicate that CD103 + DCs can sample antigens from
the epidermis, despite their proposed residence in the dermis 26–28 ,as
they cross-presented OVA from the epithelia of K5-mOVA mice. That
view is further supported by evidence that CD103 + DCs may be able
to reach their dendrites through to the epidermis 26 .
Our studies have indicated that CD103 + DCs are the dominant
migratory subset involved in MHC class I–restricted presentation of
both self and pathogen antigens. This property probably relates to a
specialized ability to cross-present cell-associated antigens. Such a
‘division of labor’ raises the issue of what alternative specializations
other DC subsets might be endowed with. It also suggests that
vaccination strategies targeting the generation of CTL immunity
might require careful consideration of how to best ‘bait’ vaccine
antigens for the CD103 + subset.
METHODS
Mice. C57BL/6 gBT-I mice 45 on a C57BL/6.SJL-PtprcaPep3b/BoyJ background
(Ly5.1 gBT-I), gDT-II mice, OT-I mice (ovalbumin-specific TCR-transgenic
mice) 46 and K5.mOVA mice 31 were bred and maintained at The Walter and
Eliza Hall Institute of Medical Research animal facility or the Department of
Microbiology and Immunology at the University of Melbourne. The gDT-II.1
mice express transgenes encoding T cell antigen receptors obtained from
a Va3.2 + Vb2 + I-A b -restricted HSV-1-specific T cell hybridoma (P.W., S.B.,
G. Davey, W.R.H., F.R.C., A.G.B., unpublished data). Animal experiments were
approved by the University of Melbourne Animal Ethics Committee or the
Walter and Eliza Hall Institute Animal Ethics Committee.
Viruses and viral infection. The wild-type parental HSV-1 strain KOS (HSV-1)
and thymidine kinase–deficient HSV KOS.Cre (HSV-TK – ) 47 were grown in
minimal essential medium with 10% (vol/vol) heat-inactivated FCS and were
titrated on Vero cells (CSL). Unless stated otherwise, mice were infected with
1 10 6 plaque-forming units of HSV-1 or HSV-TK – as described 30 .Insome
experiments, mice were infected on the upper flank proximal to the dorsal
midline on one side or both sides of the mice or on the lower flank proximal to
the ventral midline.
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 493

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
DC isolation and sorting strategy. Single-cell suspensions were prepared from
brachial or axillary lymph nodes of mice infected 2, 4 or 5 d earlier with HSV-1
(for HSV experiments) or from pooled inguinal, cervical and brachial lymph
nodes of naive K5-mOVA mice (for K5-mOVA experiments). Lymph node
suspensions were enriched for conventional DCs (excluding plasmacytoid DCs)
with antibody to Gr-1 (anti-Gr-1; RB6-8C5), anti-CD3e (KT3), anti-CD19
(ID6), anti-Thy-1 (T24), anti-B220 (RA36B2) and antibody to erythrocytes
(anti-Ter119), followed by depletion with magnetic beads as described 9,48 .
Enrichment of the cell suspension usually yielded between 80% and 85%
CD11c + cells. Cells were stained with phycoerythrin-indotricarbocyanine–
labeled anti-CD11c (N418; BD Biosciences), phycoerythrin-labeled anti-CD8
(53-6.7; BD Biosciences), allophycocyanin-labeled anti-CD205 (NLDC; prepared
‘in house’), biotinylated anti-CD326 (anti-Ep-CAM; G8.8 BioLegend)
and FITC-labeled anti-CD103 (2E7; BD Biosciences). DC subtypes from
propidium iodide–negative CD11c + events were sorted by flow cytometry
(FACSAria; BD Biosciences) according to the following two strategies: in one,
expression of CD205 and CD8 was analyzed in CD11c + cells for the identification
of Langerhans cells, dermal DCs, double-negative DCs and CD8a + DCs
(Fig. 3a); in the other, three sorting steps were used (Fig. 4a), the first to
identify CD8a + DCs (CD11c versus CD8), the second to identify CD103 + DCs
among all non-CD8a + DCs (CD205 versus CD103) and the third to categorize
CD8a – CD103 – DCs as Langerhans cells and dermal DCs on the basis of CD326
expression. After sorting, DC subtypes were washed and resuspended in RPMI
medium supplemented with 10% (vol/vol) FCS, 2-mercaptoethanol (50 mM),
L-glutamine (2 mM), penicillin (100 U/ml) and streptomycin (100 mg/ml) for
culture together with T cells. The purity of sorted subsets was routinely
over 94%.
Preparation of T cells. OT-I, gBT-I and gDT-II T cells were purified as
described 23 . T cells were isolated from lymph nodes and spleens, and samples
were enriched by incubation for 30 min with the following antibodies: anti-
Mac-1 (M1/70), anti-F4/80 (F4/80), antibody to erythrocytes (anti-Ter119),
anti-Gr-1 (RB6-8C5), anti-I-A/E (M5114) and either anti-CD4 (GK1.5)
for enrichment for CD8 + T cells or anti-CD8 (53.6-7) for enrichment for
CD4 + T cells (gDT-II). Cells that bound antibodies were removed with magnetic
beads coupled to goat anti–rat immunoglobulin G (Qiagen, Victoria,
Australia). Cells were routinely 90–95% pure, as determined by flow cytometry.
Proliferation of T cells. T cells were labeled with 2.5 mM CFSE (carboxyfluorescein
diacetate succinimidyl ester; Sigma) as described 48 . Forin vivo
proliferation, 1 10 6 CFSE-labeled CD8 + Ly5.1 gBT-I T cells were
transferred intravenously into C57BL/6 mice on day 2 or day 5 after flank
infection with HSV-1 or HSV-TK – . T cell proliferation in the brachial, axillary
or mesenteric lymph nodes was assessed 42 h later by flow cytometry
(FACSCalibur or LSRII) after staining of single-cell suspensions with phycoerythrin-labeled
anti-Ly5.1 (A20; BD Biosciences) and allophycocyanin-labeled
anti-CD8 (53-6.7; BD Biosciences). For ex vivo antigen-presentation assays, 5
10 4 CFSE-labeled gBT-I, gDT-II or OT-I cells were cultured together for 60 h
with various concentrations of purified DC subtypes from HSV-1-infected or
K5-mOVA mice as described 16,23 . Proliferation was measured by flow cytometry
(FACSCalibur or LSRII) as CFSE dilution by CD8 + (allophycocyaninlabeled
anti-CD8; 53-6.7; BD Biosciences), Va2 + (phycoerythrin-labeled
anti-Va 2; B20.1; BD Biosciences) T cells for gBT-I and OT-I; or as CD4 +
(allophycocyanin-labeled anti-CD4; GK1.5; BD Biosciences), V 3.2 + (biotinylated
RR3-16; BD Biosciences) cells for gDT-II.
Immunofluorescence. Skin samples were collected from primary and secondary
sites of infection and were immediately frozen in Tissue-Tek optimal
cutting temperature compound (Sakura Finetek). Acetone-fixed sections 8 mm
in thickness were stained with polyclonal rabbit anti–HSV immunoglobulin
(N1562; Dako) or control rabbit immunoglobulin (X0903; Dako) and were
visualized with Alexa Fluor 594–conjugated anti–rabbit immunoglobulin
(a21207; Molecular Probes; Invitrogen). Slides were mounted with Vectashield
containing DAPI (4,6-diamidino-2-phenylindole; Vector Laboratories). Confocal
images were obtained with a Meta 5110 confocal microscope (Leica) and
were analyzed with ImageJ software (US National Institutes of Health).
Painting with FITC. For monitoring of the migration of DCs from the skin
during HSV infection, at 2 d after viral inoculation, an area of flank
skin corresponding to the secondary site was painted with 20 ml of a 1%
(vol/vol) solution of FITC (Isomer 1; Sigma-Aldrich) prepared in acetone
(Sigma-Aldrich). FITC was first dissolved as a 10% (wt/vol) solution in
dimethylsulfoxide (Sigma-Aldrich) and then was diluted to 1% (wt/vol) in
acetone. For experiments involving monitoring of DCs after skin irritation,
anaesthetized mice were depilated and were painted with the 1% FITC solution
described above, or they were painted with a 1% (vol/vol) FITC solution
prepared in acetone and dibutyl phthalate (1:1) as described 9 . DCs were
enriched from the axillary lymph nodes as described above and migration
after either HSV infection or skin irritation was assessed by flow cytometry 3 d
after FITC application after cells were stained with Alexa Fluor 700–conjugated
anti-CD11c (N418; BD Biosciences), phycoerythrin-indotricarbocyanine–
labeled anti-CD8 (53-6.7; BD Biosciences), allophycocyanin-labeled anti-
CD205 (NLDC), biotinylated anti-CD326 (G8.8) and phycoerythrin-labeled
anti-CD103 (2E7; BD Biosciences).
Cytokine bead array. The secretion of cytokines into supernatants after culture
of gDT-II T cells together with various DC subsets was analyzed by cytometric
bead array (Mouse Th1/Th2 Cytokine kit (interleukins 2, 4 and 5); BD
Biosciences) according to the manufacturer’s instructions.
Measurement of viral titers. Mice were infected with 1 10 6 plaque-forming
units of HSV-TK – or wild-type HSV-1 (strain KOS) by flank scarification. The
primary inoculation site, defined as a 5-mm 5-mm full-thickness piece of
skin encompassing the area of scarification, was excised and homogenized. Skin
removed from the secondary site was 5 mm in width and extended from 5 mm
below the inoculation site to the ventral midline. Infectious virus in the tissue
was measured by standard assay of plaque-forming units on confluent Vero cell
monolayers (CSL) as described 30 .
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank the flow cytometry facilities and animal facility staff of the Walter
and Eliza Hall Institute of Medical Research and K. Field; we also thank
B. Davies and J. Langley for technical assistance. Supported by Deutsche
Forschungsgemeinschaft (BE 3285-1/2 to S.B.), the National Health and
Medical Research Council of Australia (W.R.H., F.R.C., A.G.B. and P.G.W.)
and the Howard Hughes Medical Institute (W.R.H.).
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Steinman, R.M. Lasker Basic Medical Research Award. Dendritic cells: versatile
controllers of the immune system. Nat. Med. 13, 1155–1159 (2007).
2. Villadangos, J.A. & Heath, W.R. Life cycle, migration and antigen presenting functions
of spleen and lymph node dendritic cells: limitations of the Langerhans cells paradigm.
Semin. Immunol. 17, 262–272 (2005).
3. Shortman, K. & Liu, Y. Mouse and human dendritic cell subtypes. Nat. Rev. Immunol.
2, 151–161 (2002).
4. Shortman, K. & Naik, S.H. Steady-state and inflammatory dendritic-cell development.
Nat. Rev. Immunol. 7, 19–30 (2007).
5. Heath, W.R. et al. Cross-presentation, dendritic cell subsets, and the generation of
immunity to cellular antigens. Immunol. Rev. 199, 9–26(2004).
6. Jakubzick, C. et al. Lymph-migrating, tissue-derived dendritic cells are minor constituents
within steady-state lymph nodes. J. Exp. Med. 205, 2839–2850 (2008).
7. Wakim, L.M., Waithman, J., van Rooijen, N., Heath, W.R. & Carbone, F.R. Dendritic
cell-induced memory T cell activation in nonlymphoid tissues. Science 319, 198–202
(2008).
8. Itano, A.A. et al. Distinct dendritic cell populations sequentially present antigen to
CD4 + T cells and stimulate different aspects of cell-mediated immunity. Immunity 19,
47–57 (2003).
9. Allan, R.S. et al. Migratory dendritic cells transfer antigen to a lymph node-resident
dendritic cell population for efficient CTL priming. Immunity 25, 153–162 (2006).
10. Cella, M. et al. Plasmacytoid monocytes migrate to inflamed lymph nodes and produce
large amounts of type I interferon. Nat. Med. 5, 919–923 (1999).
11. Diebold, S.S. et al. Viral infection switches non-plasmacytoid dendritic cells into high
interferon producers. Nature 424, 324–328 (2003).
12. den Haan, J.M. & Bevan, M.J. Constitutive versus activation-dependent crosspresentation
of immune complexes by CD8 + and CD8 – dendritic cells in vivo. J. Exp.
Med. 196, 817–827 (2002).
494 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
13. Dudziak, D. et al. Differential antigen processing by dendritic cell subsets in vivo.
Science 315, 107–111 (2007).
14. Schnorrer, P. et al. The dominant role of CD8 + dendritic cells in cross-presentation is
not dictated by antigen capture. Proc. Natl. Acad. Sci. USA 103, 10729–10734
(2006).
15. Smith, C.M. et al. Cutting edge: conventional CD8a + dendritic cells are preferentially
involved in CTL priming after footpad infection with herpes simplex virus-1.
J. Immunol. 170, 4437–4440 (2003).
16. Allan, R.S. et al. Epidermal viral immunity induced by CD8a + dendritic cells but not by
Langerhans cells. Science 301, 1925–1928 (2003).
17. Belz, G.T., Shortman, K., Bevan, M.J. & Heath, W.R. CD8a + dendritic cells selectively
present MHC class I-restricted noncytolytic viral and intracellular bacterial antigens in
vivo. J. Immunol. 175, 196–200 (2005).
18. Belz, G.T. et al. Cutting edge: Conventional CD8a + dendritic cells are generally involved
in priming CTL immunity to viruses. J. Immunol. 172, 1996–2000 (2004).
19. Belz, G.T. et al. Distinct migrating and nonmigrating dendritic cell populations are
involved in MHC class I-restricted antigen presentation after lung infection with virus.
Proc. Natl. Acad. Sci. USA 101, 8670–8675 (2004).
20. Wilson, N.S. et al. Systemic activation of dendritic cells by Toll-like receptor ligands or
malaria infection impairs cross-presentation and antiviral immunity. Nat. Immunol. 7,
165–172 (2006).
21. He, Y., Zhang, J., Donahue, C. & Falo, L.D. Jr. Skin-derived dendritic cells induce
potent CD8 + T cell immunity in recombinant lentivector-mediated genetic immunization.
Immunity 24, 643–656 (2006).
22. Bennett, C.L. et al. Inducible ablation of mouse Langerhans cells diminishes but fails
to abrogate contact hypersensitivity. J. Cell Biol. 169, 569–576 (2005).
23. Waithman, J. et al. Skin-derived dendritic cells can mediate deletional tolerance of
class I-restricted self-reactive T cells. J. Immunol. 179, 4535–4541 (2007).
24. Kaplan, D.H., Jenison, M.C., Saeland, S., Shlomchik, W.D. & Shlomchik, M.J.
Epidermal langerhans cell-deficient mice develop enhanced contact hypersensitivity.
Immunity 23, 611–620 (2005).
25. Kissenpfennig, A. et al. Dynamics and function of Langerhans cells in vivo dermal
dendritic cells colonize lymph node areas distinct from slower migrating Langerhans
cells. Immunity 22, 643–654 (2005).
26. Bursch, L.S. et al. Identification of a novel population of Langerin + dendritic cells.
J. Exp. Med. 204, 3147–3156 (2007).
27. Poulin, L.F. et al. The dermis contains langerin + dendritic cells that develop and
function independently of epidermal Langerhans cells. J. Exp. Med. 204, 3119–3131
(2007).
28. Ginhoux, F. et al. Blood-derived dermal langerin+ dendritic cells survey the skin in the
steady state. J. Exp. Med. 204, 3133–3146 (2007).
29. Simmons, A. & Nash, A.A. Zosteriform spread of herpes simplex virus as a model of
recrudescence and its use to investigate the role of immune cells in prevention of
recurrent disease. J. Virol. 52, 816–821 (1984).
30. van Lint, A. et al. Herpes simplex virus-specific CD8 + T cells can clear established lytic
infections from skin and nerves and can partially limit the early spread of virus after
cutaneous inoculation. J. Immunol. 172, 392–397 (2004).
ARTICLES
31. Azukizawa, H. et al. Induction of T-cell-mediated skin disease specific for antigen
transgenically expressed in keratinocytes. Eur. J. Immunol. 33, 1879–1888
(2003).
32. Zhao, X. et al. Vaginal submucosal dendritic cells, but not Langerhans cells, induce
protective Th1 responses to herpes simplex virus-2. J. Exp. Med. 197, 153–162
(2003).
33. Lee, H.K. et al. Differential roles of migratory and resident DCs in T cell priming after
mucosal or skin HSV-1 infection. J. Exp. Med. 206, 359–370 (2009).
34. Stoitzner, P. et al. Langerhans cells cross-present antigen derived from skin. Proc. Natl.
Acad. Sci. USA 103, 7783–7788 (2006).
35. den Haan, J.M., Lehar, S.M. & Bevan, M.J. CD8 + but not CD8 – dendritic cells crossprime
cytotoxic T cells in vivo. J. Exp. Med. 192, 1685–1696 (2000).
36. Sung, S.S. et al. A major lung CD103 (a E)-b 7 integrin-positive epithelial dendritic cell
population expressing Langerin and tight junction proteins. J. Immunol. 176,
2161–2172 (2006).
37. Hildner, K. et al. Batf3 deficiency reveals a critical role for CD8a + dendritic cells in
cytotoxic T cell immunity. Science 322, 1097–1100 (2008).
38. GeurtsvanKessel, C.H. et al. Clearance of influenza virus from the lung depends on
migratory langerin + CD11b – but not plasmacytoid dendritic cells. J. Exp. Med. 205,
1621–1634 (2008).
39. del Rio, M.L., Rodriguez-Barbosa, J.I., Kremmer, E. & Forster, R. CD103 – and CD103 +
bronchial lymph node dendritic cells are specialized in presenting and cross-presenting
innocuous antigen to CD4 + and CD8 + T cells. J. Immunol. 178, 6861–6866
(2007).
40. Jakubzick, C., Helft, J., Kaplan, T.J. & Randolph, G.J. Optimization of methods to study
pulmonary dendritic cell migration reveals distinct capacities of DC subsets to acquire
soluble versus particulate antigen. J. Immunol. Methods 337, 121–131 (2008).
41. Stock, A.T., Mueller, S.N., van Lint, A.L., Heath, W.R. & Carbone, F.R. Cutting edge:
prolonged antigen presentation after herpes simplex virus-1 skin infection. J. Immunol.
173, 2241–2244 (2004).
42. Salio, M., Cella, M., Suter, M. & Lanzavecchia, A. Inhibition of dendritic cell
maturation by herpes simplex virus. Eur. J. Immunol. 29, 3245–3253 (1999).
43. Kruse, M. et al. Mature dendritic cells infected with herpes simplex virus type 1 exhibit
inhibited T-cell stimulatory capacity. J. Virol. 74, 7127–7136 (2000).
44. Jones, C.A. et al. Herpes simplex virus type 2 induces rapid cell death and functional
impairment of murine dendritic cells in vitro. J. Virol. 77, 11139–11149 (2003).
45. Mueller, S.N., Heath, W., McLain, J.D., Carbone, F.R. & Jones, C.M. Characterization of
two TCR transgenic mouse lines specific for herpes simplex virus. Immunol. Cell Biol.
80, 156–163 (2002).
46. Hogquist, K.A. et al. T cell receptor antagonist peptides induce positive selection.
Cell 76, 17–27 (1994).
47. Wakim, L.M., Jones, C.M., Gebhardt, T., Preston, C.M. & Carbone, F.R. CD8 + T-cell
attenuation of cutaneous herpes simplex virus infection reduces the average viral copy
number of the ensuing latent infection. Immunol. Cell Biol. 86, 666–675 (2008).
48. Belz, G.T., Bedoui, S., Kupresanin, F., Carbone, F.R. & Heath, W.R. Minimal activation
of memory CD8 + T cell by tissue-derived dendritic cells favors the stimulation of naive
CD8 + Tcells.Nat. Immunol. 8, 1060–1066 (2007).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 495

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Divergent functions for airway epithelial
matrix metalloproteinase 7 and retinoic acid
in experimental asthma
Sangeeta Goswami 1 , Pornpimon Angkasekwinai 2 , Ming Shan 1 , Kendra J Greenlee 3 , Wade T Barranco 4 ,
Sumanth Polikepahad 4 , Alexander Seryshev 4 , Li-zhen Song 4 , David Redding 5 , Bhupinder Singh 5 , Sanjiv Sur 5 ,
Prescott Woodruff 6 , Chen Dong 2 , David B Corry 1,4 & Farrah Kheradmand 1,4
The innate immune response of airway epithelial cells to airborne allergens initiates the development of T cell responses that
are central to allergic inflammation. Although proteinase allergens induce the expression of interleukin 25, we show here that
epithelial matrix metalloproteinase 7 (MMP7) was expressed during asthma and was required for the maximum activity of
interleukin 25 in promoting the differentiation of T helper type 2 cells. Allergen-challenged Mmp7 –/– mice had less airway
hyper-reactivity and production of allergic inflammatory cytokines and higher expression of retinal dehydrogenase 1. Inhibition of
retinal dehydrogenase 1 restored the asthma phenotype of Mmp7 –/– mice and inhibited the responses of lung regulatory T cells,
whereas exogenous administration of retinoic acid attenuated the asthma phenotype. Thus, MMP7 coordinates allergic lung
inflammation by activating interleukin 25 while simultaneously inhibiting retinoid-dependent development of regulatory T cells.
Proximal and distal airway epithelial cells directly communicate with
the outside environment, and thus their response to inhaled allergens
could be integral to the initiation of allergic lung inflammation.
Identification of the critical epithelium-derived mediators that drive
allergic lung responses is at an early stage but is essential to complete
understanding of the molecular mechanisms that underlie diseases
such as asthma. After being inhaled, proteinase allergens (such as
airborne allergens with active proteinase properties) rapidly induce a
powerful inflammatory response that initiates the recruitment of
allergic immune cells into the lung 1,2 . Although the function of
cytokines expressed in the airway epithelium in asthma has been
studied extensively 3,4 , the mechanism by which allergens initiate the
differentiation of airway T helper type 2 (TH2) cells is poorly understood.
Interleukin 25 (IL-25; also called IL-17E) is a member of the
IL-17 family of cytokines that is produced mainly in the airway
epithelium in response to proteinase allergens 5,6 . Transgenic expression
of human and mouse IL-25 induces T H2 immune responses
marked by considerable lung eosinophilia, mucus hyperproduction
and higher expression of IL-4, IL-5, IL-13 and CCL11 (eotaxin 1) in
the airway 7,8 . Neutralization of IL-25 by blocking antibodies has been
shown to reverse airway hyper-reactivity, a pathognomonic feature of
asthma, in a mouse model of asthma 9 . Moreover, IL-25-deficient mice
are unable to develop a strong TH2-biased immune response and fail
Received 12 January; accepted 12 February; published online 29 March 2009; doi:10.1038/ni.1719
to clear gut parasitic infection with Nippostrongylus brasiliensis or
Trichuris muris 10 . Such studies suggest a critical function for epithelial
cells in providing the IL-25 signals that may be required for effective
T H2 responses. Although the function of IL-25 in allergic lung disease
is well known, a broader understanding of how IL-25 interacts with
other airway epithelial factors to coordinately regulate allergic
responses is needed.
Matrix metalloproteinases (MMPs) are zinc-dependent enzymes
that are induced in response to many stimuli. Although MMPs were
originally shown to mediate the turnover of extracellular matrix
molecules, it is now apparent that they control diverse biological
processes unrelated to matrix degradation 11 . Although the exact
function of MMPs in chronic models of lung inflammation is
debatable, it is clear that several members of this family are acutely
induced either locally or systemically at the onset of allergen challenge
12–16 . MMP2 and MMP9 are not required for the development of
allergic lung disease but serve key functions in the clearance of
inflammatory cells from lung parenchyma by establishing the necessary
trans-epithelial chemokine gradients 15,16 . However, whereas
neither MMP2 nor MMP9 is expressed in the airway epithelium 17 ,
MMP7 (also called matrilysin; A001477), an epithelial cell–specific
MMP, is induced in the lung and gut during inflammation 18,19 .
MMP7 has been shown to modify pro-a-defensin, the ligand for the
1 Department of Immunology, Baylor College of Medicine, Houston, Texas, USA. 2 Department of Immunology, University of Texas MD Anderson Cancer Center, Houston,
Texas, USA. 3 Department of Biological Sciences, North Dakota State University, Fargo, North Dakota, USA. 4 Department of Medicine, Baylor College of Medicine, Houston
Texas, USA. 5 Department of Medicine, University of Texas Medical Branch Galveston, Galveston, Texas, USA. 6 Department of Medicine, University of California San
Francisco, San Francisco, California, USA. Correspondence should be addressed to F.K. (farrahk@bcm.edu) or D.B.C. (dcorry@bcm.edu).
496 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Figure 1 MMP7 is induced in allergic
inflammation and modulates IL-25 function.
(a) Immunohistochemistry of MMP7 in lungs of
PBS-treated (WT PBS) and complete Aspergillus
allergen (CAA)-immunized (WT CAA) C57BL/6
mice. Arrows and arrowheads indicate MMP7
expression; insets, 40 magnification of the
airway. Scale bars, 100 mm (main images) and
25 mm (insets). Data are representative of at
least three independent experiments. (b) Silver
staining of rIL-25 (5 mg), rIL-13 (5 mg) and
rTSLP (5 mg) left untreated or incubated for 2 h
at 37 1C with 4-aminophenylmercuric acetate–
activated MMP7 (0.5 mg; + MMP7) and then
separated by electrophoresis through a 16.5%
tricine gel. Arrowheads indicate MMP7-cleaved
fragments of rIL-25. Results are representative
of at least three independent experiments.
(c) ELISA of IL-4, IL-5, IL-13 and interferon-g
(IFN-g) in supernatants of naive lymph node cells
(left) and spleen cells (right) isolated from
C57BL/6 mice and stimulated for 2 d with platebound
anti-CD3 (2 mg/ml) in the presence of
equal amounts of native rIL-25 (250 ng/ml)
or rIL-25’C (250 ng/ml). Media alone and the
4-aminophenylmercuric acetate–containing
buffer used to activate MMP7 (Vehicle) serve as
controls. *, P o 0.05, versus rIL-25 (one-way
ANOVA). Data are representative of three
different experiments (mean and s.d.;
Lymph
*
n ¼ 3 samples). (d) Flow cytometry of the expression of IL-4 and IL-5 by lymphocytes and spleen cells cultured for 3 d in the conditions described in c,
then restimulated for 5 h with ionomycin (500 ng/ml) and phorbol 12-myristate 13-acetate (50 ng/ml) in the presence of GolgiStop, followed by
intracytoplasmic staining. Numbers adjacent to outlined areas indicate percent IL-5 + Il-4 – cells (top left) or IL-5 – Il-4 + cells (bottom right). Data are
representative of three independent experiments.
cell surface receptor Fas (FasL; CD95L) and tumor necrosis factor in
the gut during defense against enteric pathogens, but its function in
allergic inflammation has yet to be explored 19,20 .
In this study, we examine how airway epithelial IL-25 coordinates
proteinase-dependent allergic lung disease in the context of other
potentially relevant airway epithelium-derived factors, including
MMP7. We show that in response to proteinase allergen or recombinant
IL-25 (rIL-25), mouse airway epithelial cells expressed MMP7.
Activation of MMP7 was critical for IL-25 function because MMP7cleaved
rIL-25 (called ‘rIL-25’C’ here) enhanced T H2 differentiation.
Furthermore, we show that humans with chronic asthma also
expressed MMP7 and IL-25 in their distal airspaces and, in response
to ragweed extract, a potent allergenic stimuli, patients with a history
of allergy had significantly more nasal secretion of MMP7. In the
absence of MMP7, mice showed attenuated allergic responses to
challenge with proteinase allergen and enhanced expression of retinal
dehydrogenase 1 (RALDH-1), a rate-limiting enzyme for the production
of retinoic acid. Moreover, Mmp7 –/– mice had more regulatory
T cells in the lung parenchyma. Our results collectively support the
idea that MMP7 serves as a proinflammatory mediator that specifically
enhances the function of IL-25 that is necessary for robust TH2
responses. Finally, we demonstrate a tolerogenic mechanism initiated
by airway epithelial cells in response to challenge with proteinase
allergen in which RALDH-1 was induced to promote immunosuppressive
regulatory T cells.
RESULTS
MMP7 mediates enhanced function of IL-25
MMP2 and MMP9 are not expressed in airway epithelial cells after
allergen challenge 17 . In contrast, we found that a fungus-derived
proteinase allergen strongly enhanced airway epithelial expression of
a b
IL-4 (ng/ml)
IL-5 (ng/ml)
IL-13 (ng/ml)
IFN-γ (ng/ml)
2
1
0
1.5
1.0
0.5
0.0
150
100
50
0
150
100
50
0
Media
WT PBS
*
*
rIL-25
rIL-25′C
Vehicle
4
3
2
1
0
1.5
1.0
0.5
0.0
200
100
0
400
300
200
100
0
Media
WT CAA
c d
Spleen
*
*
*
rIL-25
rIL-25′C
Vehicle
(kDa) (kDa)
37
20
17.5
15
25
20
10
12.5
10
Media
rIL-25
rIL-25′C
IL-5
MMP7
rIL-25
rIL-25 + MMP7
ARTICLES
rTSLP
rTSLP + MMP7
rIL-13
rIL-13 + MMP7
10
Lymph
Spleen
0.14
0.18
0.27 0.88
0.23 0.43
0.43 2.34
0.55 0.48
0.95
3.5
IL-4
4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 4
10 3
10 3
10 2
10 2
10 1
10 1 10 0
10 0
10 4
10 3
10 2
10 1
10 0
MMP7 (Fig. 1a). Whereas induction of MMP2 and MMP9 is IL-13
dependent 15 , MMP7 induction was independent of IL-13, as there
was equivalent induction of MMP7 in wild-type mice and allergenchallenged
mice deficient in STAT6, the principal mediator of the
effects of IL-13 in the lung (data not shown). Because MMP7 is known
to cleave and activate cytokines in the gut 13 , we next determined if
MMP7 modifies cytokines critical for initiating allergic lung responses.
In addition to IL-25, both IL-13 and thymic stromal lymphopoietin
(TSLP) are critical participants in the molecular pathways that
underlie allergic lung disease 5,21 . We found that although rIL-13 and
rTSLP were not substrates for activated MMP7, in the same conditions,
rIL-25 was cleaved at multiple sites (Fig. 1b and Supplementary
Table 1 and Fig. 1 online). To determine the function of cleaved IL-25,
we examined the responses of naive spleen and lymph node cells after
activation by crosslinking of T cell antigen receptors in vitro.Wefound
that rIL-25’C induced significantly more secretion of T H2 cytokines
(IL-4, IL-5 and IL-13) than did native rIL-25 and it showed enhanced
binding to a fusion protein of IL-17 receptor B and the Fc fragment
in vitro (Fig. 1c,d and Supplementary Fig. 2a,b online). Indeed,
native IL-25 had little effect on IL-4 secretion, but rIL-25’C enhanced
IL-4 secretion tenfold, which suggested that the main active form of
IL-25 is the MMP7-cleaved form. We found no effect of rIL-25 or
rIL-25’C on the induction of interferon-g, a canonical T H1 cytokine
(Fig. 1c). Collectively, these data indicate that IL-25 is a substrate for
MMP7 and that cleavage of IL-25 by MMP7 is needed to drive robust
T H2 responses.
Attenuated asthma phenotype of Mmp7 –/– mice
The greater potency of rIL-25’C in activating T H2 cells in vitro raised
the possibility of an essential function for MMP7 in mediating allergic
inflammation in vivo. To investigate this, we compared the responses
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 497

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
b
WT
60
WT CAA
Mmp7 *
WT PBS
*
**
PBS
**
0 0.03 0.1
Ach (µg/g)
0.3 1.0
WT
**
CAA
*
*
**
*
**
–/– CAA
Mmp7 –/– PBS
Mmp7 –/–
Mmp7 –/–
WT
Mmp7 –/–
WT
Mmp7 –/–
WT
Mmp7 –/–
WT
Mmp7 –/–
WT
Mmp7 –/–
WT
Mmp7 –/–
40
20
0
3
2
1
0
CAA – + – +
60
40
20
f 5
4
3
2
1
*
4
3
2
1
*
0
0
0
CAA
25
20
15
10
– + – + rlL-25
g 8
6
4
– + + rlL-25
200
150
100
– + +
5
0
2
0
ND *
50
0
ND *
CAA – + – +
rlL-25 – + +
rlL-25 – + +
c
R RS (cm H 2 O s/ml)
PC200 (µg/g)
Glycoprotein
(µg/ml)
d
BAL eosinophils
(×10 5 /ml)
BAL total cells
(×10 5 /ml)
of Mmp7 –/– and wild-type mice to the proteinase allergen CAA
(complete aspergillus allergen) extracted from fungi, a powerful
allergen used before to elicit allergic lung disease in mice 1 . This airway
inflammation model is characterized by a predominantly eosinophilic
influx and airway obstruction due to enhanced glycoprotein secretion,
as well as more airway hyper-reactivity to secondary challenge with
acetylcholine 22,23 . Mmp7 –/– mice immunized with CAA had less airway
hyper-reactivity than did wild-type mice, as shown by both their
lower respiratory system resistance in response to many doses of
acetylcholine and higher stimulating concentration of acetylcholine
required to elicit a 200% change from the baseline respiratory system
resistance. Moreover, Mmp7 –/– mice had less secretion of glycoproteins
in bronchoalveolar lavage (BAL) fluid after CAA challenge (Fig. 2a–c).
Airway eosinophilia was also significantly lower in CAA-immunized
Mmp7 –/– mice (Fig. 2d). Because the egress of lung parenchymal
eosinophils into the airways is impaired in Mmp2 –/– and Mmp9 –/–
mice 15,16 , we assessed lung parenchymal eosinophils to rule out the
possibility of a similar defect in Mmp7 –/– mice. Analysis of the total
number of lung parenchymal inflammatory cells showed that the
IL-4 (pg/ml)
Figure 3 Attenuated T H2 cytokines and
chemokines in Mmp7 –/– mice. (a–c) Luminex
assay of IL-4 (a), IL-5 (b) and IL-13 (c) inBAL
fluid from age- and sex-matched wild-type and
Mmp7 –/– mice (n ¼ 5 mice per group)
immunized intranasally with PBS (–) or CAA (+)
every 4 d for a total of five doses and assessed
24 h after the final immunization. (d) ELISA of
CCL11 in BAL fluid of mice treated as described
in a–c. (e) Real-time RT-PCR analysis of the
expression of Il25 mRNA in the lungs of
e
IL-4 (pg/ml)
BAL eosinophils
(×10 5 /ml)
IL-5 (pg/ml)
lungs of CAA-immunized Mmp7 –/– mice had fewer eosinophils than
did those of wild-type mice (Fig. 2e and Supplementary Fig. 3
online), which indicated that the lower cell number in the airway
was not due to less cell trafficking.
We next assessed whether MMP7-mediated modification of IL-25
was influential in determining allergic lung responses. We administered
rIL-25 intranasally to wild-type and Mmp7 –/– mice and assessed
eosinophil recruitment. In this assay, rIL-25 alone readily induced
lung eosinophilia and new expression of MMP7 in airway epithelial
cells of wild-type mice, but we found significantly fewer eosinophils
and lower concentrations of IL-4 and IL-5 in the BAL fluid of
Mmp7 –/– mice (Fig. 2f,g and Supplementary Fig. 4 online). Collectively,
these findings support the idea that induction of MMP7 in
response to allergens is critical for enhancing IL-25 function during
initiation of the allergic immune response.
Diminished TH2 responses of Mmp7 –/– mice
We next determined if the diminished airway hyper-reactivity and
BAL inflammatory cells in Mmp7 –/– mice were secondary to lower
a b c d e
WT
Mmp7
25
20
15
10
5
0
*
ND
**
ND
70
60
50
40
30
20
10
0
*
**
150
100
50
0
*
**
0.3
0.2
0.1
0.0
*
**
25
20
15
10
5
0
*
**
CAA – + – + CAA – + – + CAA – + – + CAA – + – + CAA – + +
–/–
WT
Mmp7 –/–
WT
Mmp7 –/–
WT
Mmp7 –/–
WT
Mmp7 –/–
unimmunized (–) and CAA-immunized (+) wild-type and Mmp7 –/– mice, presented relative to Actb expression. *, P o 0.05, versus PBS-challenged
wild-type; **, P o 0.05, versus CAA-immunized wild-type (one-way ANOVA and t-test). Data are representative of three independent experiments
(mean and s.d.).
IL-5 (pg/ml)
IL-13 (pg/ml)
Figure 2 MmP7 –/– mice have an attenuated
asthma phenotype. (a,b) Airway hyper-reactivity in
age- and sex-matched wild-type mice (WT; n ¼ 4)
and Mmp7 –/– mice (n ¼ 4) immunized
intranasally with PBS (–) or CAA (+) every 4 d
for a total of five doses and assessed 24 h after
the final immunization as respiratory system
resistance (RRS) in response to various doses
of acetylcholine (Ach; a) or concentration of
acetylcholine needed to elicit a 200% change
from baseline respiratory system resistance
(PC200; b). (c,d) ELISA of glycoproteins (c) and
eosinophil cell counts (d) in BAL fluid from the
mice in a,b. Ina–d, *,P o 0.05, versus PBSchallenged
wild-type; **, P o 0.05, versus
CAA-immunized wild-type (one-way ANOVA and
t-test). Data are representative of at least three
independent experiments (mean and s.d.).
(e) Photomicrographs of bronchovascular bundles
stained with hematoxylin and eosin (n ¼ 4 mice
per group). Scale bar, 100 mm. Data are
representative of at least three independent
experiments. (f,g) Total cell and eosinophil counts
(f) and ELISA of IL-4 and IL-5 (g) inBALfluid
of wild-type and MmP7 –/– mice given PBS (–) or
rIL-25 (+; 5 mg) intranasally twice daily for 3 d,
assessed 24 h after the final dose. *, P o 0.05,
versus rIL-25-treated wild-type mice (one-way
ANOVA). Data are representative of three
independent experiments (mean and s.d.;
n ¼ 4 mice per group).
498 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
CCL11 (ng/ml)
IL-25 mRNA
(relative expression)

© 2009 Nature America, Inc. All rights reserved.
a
c
ATRA (AU)
BAL total cells
(×10 5 /ml)
WT CAA Mmp7 Superimposed
–/– CAA
BAL eosinophils
(×10 5 /ml)
IL-25 mRNA
(relative expression)
T H2 responses in Mmp7 –/– mice. As expected, CAA-challenged
Mmp7 –/– mice had less IL-4, IL-5, IL-13, IL-6 and tumor necrosis
factor in BAL fluid than did wild-type mice (Fig. 3a–c and Supplementary
Fig. 5 online), whereas we found no significant differences in
the concentrations of CCL3, IL-9 and IL-1b (data not shown).
Consistent with the diminished lung eosinophilia, we found significantly
lower CCL11 expression in CAA-immunized Mmp7 –/– mice
than in CAA-immunized wild-type mice, but we detected no significant
differences the concentrations of CCL7 and CCL17 (Fig. 3d,
and data not shown). Furthermore, whole-lung analysis of RNA
showed that immunized Mmp7 –/– mice had lower expression of Il25
than did wild-type mice (Fig. 3e), which suggested that in addition to
proteolytic modification of IL-25, activation and function of MMP7 in
the lung is required for the transcriptional expression of Il25 in this
model of allergic airway disease.
More retinoic acid in allergen-challenged Mmp7 –/– mice
To identify additional events that might regulate the airway epithelial
response to allergen, we did proteomic analyses of BAL fluid from
ATRA (AU)
CCL11 mRNA
(relative expression)
RALDH-1 abundance (log)
ATRA (AU)
0.3
0.2
0.1
0
–0.1
–0.2
–0.3
–0.4
–0.5
IL-4 (pg/ml)
WT PBS
WT CAA
Mmp7 –/– CAA
Mmp7 –/– WT PBS
WT CAA
CAA
–0.00000
–0.00002
–0.00004
–0.00006
–0.00008
–0.00010
–0.00012
–0.00000
–0.00002
–0.00004
–0.00006
–0.00008
–0.00010
–0.00012
8.081
–0.00000
–0.00002
–0.00004
–0.00006
–0.00008
–0.00010
–0.00012
8.053
–0.00014
–0.00014
–0.00014
–0.00016
–0.00016
–0.00016
–0.00018
–0.00018
–0.00018
–0.00020
–0.00020
–0.00020
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Time (min) Time (min) Time (min)
d 60
40
20
*
**
e 50
40
30
20
10
*
**
f
25
20
15
10
5
*
**
g 4
3
2
1
*
**
h 30
20
10
*
**
i 0.4
0.3
0.2
0.1
*
**
0
0
0
0
0
0.0
CAA – – + + CAA – – + + CAA – – + + CAA – – + + CAA – – + + CAA – – + +
ATRA – + – + ATRA – + – + ATRA – + – + ATRA – + – + ATRA – + – + ATRA – + – +
Figure 4 Proteomics analysis of BAL fluid from CAA-immunized wild-type and Mmp7 –/– mice. (a) Two-dimensional gel electrophoresis assessing the
processing of proteins in pooled BAL fluid from CAA-immunized wild-type mice (green; indocarbocyanine) and Mmp7 –/– mice (red; indodicarbocyanine;
Supplementary Table 2). Superimposed images (right) show differences in proteins: yellow, no change; red spots, more abundant in Mmp7 –/– ; green spots,
more abundant in wild-type; inset, enlargement of the area with RALDH-1. Arrows indicate RALDH-1 protein. Data are representative of two independent
experiments (n ¼ 3 mice per group). (b) RALDH-1 abundance in BAL fluid from wild-type and Mmp7 –/– mice immunized with PBS or CAA, assessed as spot
volumes standardized and log-transformed with DeCyder software, Biological Variation Analysis. Each point represents the spot volume of a pooled sample
(n ¼ 3 mice). *, P o 0.05, versus saline-challenged wild-type; **, P o 0.05, versus CAA-immunized wild-type mice (t-test). Data are representative of at
least three independent experiments (mean ± s.d.). (c) HPLC of ATRA in BAL fluid from wild-type and Mmp7 –/– mice (n ¼ 3) immunized with PBS or CAA,
presented as absorbance units (AU). Numbers in graphs indicate time of ATRA peak; area under the curve for each peak is 344 (WT CAA) and 755
(Mmp7 –/– CAA). (d,e) Total cells (d) and eosinophils (e) in BAL fluid from wild-type mice treated with liposomal ATRA (5 mg per g body weight) 1 h before
intranasal immunization with CAA. Negative control, liposomal ATRA and PBS (n ¼ 5); positive control, CAA alone (n ¼ 5). (f,g) Real-time RT-PCR analysis
of the expression of Il25 mRNA (f) andCcl11 mRNA (g) inthelungsofthemiceind,e, presented relative to the expression of Actb (f) or 18S RNA (g).
(h,i) Luminex assay of IL-4 (h) and IL-13 (i) in BAL fluid. *, P o 0.05, versus PBS-challenged wild-type; **, P o 0.05, versus CAA-immunized wild-type
(one-way ANOVA). Data are representative of three independent experiments (mean and s.d.).
CAA-immunized wild-type and Mmp7 –/– mice 17 . Among the proteins
that were significantly and differently modulated in CAA-immunized
Mmp7 –/– mice (P o 0.05; Supplementary Table 2 online), we found
that they had more RALDH-1 (Fig. 4a,b and Supplementary Fig. 6
online). RALDH-1 is the rate-limiting enzyme for the conversion of
retinaldehyde to all-trans retinoic acid (ATRA), which is critical for
maintenance of the tolerogenic environment of mucosal surfaces 24 .In
support of our proteomic-based finding of more RALDH-1, analysis
of BAL fluid showed that CAA-immunized Mmp7 –/– mice had a
higher ATRA concentration than did CAA-immunized wild-type
mice (Fig. 4c).
Next, we determined if more production of ATRA in the lungs of
Mmp7 –/– mice might account in part for their attenuated response to
allergen. We delivered liposomal ATRA by means of respiratory
aerosol to wild-type mice 1 h before each immunization with CAA,
then assessed lung allergic responses. Mice that received ATRA had an
attenuated asthma phenotype, as demonstrated by fewer total cells in
BAL fluid, especially eosinophils, and lower expression of Il25 mRNA
than that of mice given aerosol vehicle (empty liposome; Fig. 4d–f).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 499
b
*
IL-13 (ng/ml)
ARTICLES
**

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
WT
10 4
Mmp7 –/–
WT Mmp7 –/–
a b c
PBS
Consistent with lower expression of inflammatory markers, we also
detected less CCL11, IL-4 and IL-13, which act ‘downstream’ of IL-25
signaling (Fig. 4g–i and Supplementary Fig. 7 online). These data
confirm that ATRA has potent anti-inflammatory activity when
administered to the lung and that this activity can be enhanced
through inhibition of Il25 expression.
Epithelial RALDH-1 and ATRA induce regulatory T cells
We found expression of RALDH-1 protein in airway epithelial cells
and alveolar macrophages of CAA-immunized wild-type and Mmp7 –/–
mice (Fig. 5a). Higher expression of RALDH-1 and concentrations of
ATRA have been linked to enhanced immune tolerance and more
production of inducible regulatory T cells 25 . Therefore, we determined
if higher expression of RALDH-1 acted as a negative regulator in
response to allergens and increased the abundance of regulatory T cells
in the lungs of wild-type and Mmp7 –/– mice after immunization with
CAA. CAA-immunized wild-type and Mmp7 –/– mice had more
PBS
Foxp3
10 3
10 2
10 1
10 0
10 4
0.14 0.31
BAL total cells
(×10 5 /ml)
BAL eosinophils
(×10 5 /ml)
CD25 + Foxp3 + regulatory T cells in lungs than did naive mice (Fig. 5b
and Supplementary Fig. 8 online). To further examine the function
of RALDH-1 in inducing the production of regulatory T cells, we
administered citral, an inhibitor of RALDH-1, to mice immunized
with CAA 26 . Inhibition of the synthesis of retinoic acid by citral
resulted in fewer CD25 + Foxp3 + regulatory T cells in the lung (Fig. 5b
and Supplementary Fig. 9 online), which suggested possible involvement
of RALDH-1 and ATRA in the development of this inhibitory
subset of lung T cells. Moreover, wild-type and Mmp7 –/– mice
immunized with CAA and challenged with citral had more eosinophils,
IL-5 and CCL11 in BAL fluid than did mice immunized with
CAA alone (Fig. 5c–e and data not shown).
MMP7 expression in response to allergen in human asthma
To address the relevance of MMP7 expression to human asthma, we
studied bronchial biopsy specimens from human volunteers with and
without asthma. Notably, whereas we detected IL-25 in airway
d
IL-5 (pg/ml)
12.5
10.0
7.5
5.0
2.5
0.0
CAA –
Citral +
*
WT
Mmp7 –/–
**
+ + + +
– + – +
10 3
10 2
10 1
10 0
10 4
10 4
10 3
10 3
10 2
10 2
10 1
10 1
1.26
3.01
0.45
0.7
10
CD25
0
10 0
10 4
10 3
10 2
10 1
10 0
CAA
CAA
10
8
6
4
WT
Mmp7
2
0
CAA – + + + +
Citral + – + – +
CAA
+
citral
200
150
100
50
0
CAA
Citral
–
+
+
–
+
+
+
–
+
+
–/–
WT
Mmp7 –/–
*
**
Figure 5 Epithelial expression of RALDH-1 in response to
allergens initiates a negative regulatory response.
(a) Immunohistochemistry of RALDH-1 (arrowheads)
in lung sections of wild-type and Mmp7
e
*
**
/ mice immunized
intranasally with PBS or CAA. Scale bars, 100 mm. Data are
representative of two independent experiments. (b) Flow
cytometry of regulatory T cells obtained from the lungs of
wild-type and Mmp7 / mice immunized intranasally with
PBS or CAA (same dose for all), with (bottom row) or without
(top and middle rows) the administration of citral 1 h before CAA immunization, stained with anti-CD4, anti-CD25 and anti-Foxp3. Numbers in outlined
areas indicate percent Foxp3 + CD25 + cells. Data are representative of two independent experiments (n ¼ 4 mice per group). (c–e) Total cells (c), eosinophils
(d) and IL-5 (e) in BAL fluid from mice treated as described in b. Intranasal administration of citral or CAA alone serves as a negative or positive control,
respectively. *, P o 0.05, versus CAA-challenged wild-type; **, P o 0.05, versus CAA-challenged Mmp7 / independent experiments (mean and s.d.; n ¼ 4 mice per group).
mice (t-test). Data are representative of three
a Nonasthmatic, MMP7
Nonasthmatic, IL-25 b
Asthmatic, MMP7
Asthmatic, IL-25
MMP7 (pg/ml)
1.5
1.0
0.5
0.0
RW (30 min) RW (5 h)
*
Figure 6 Expression of MMP7 and IL-25 in
humans with asthma. (a) Serial photomicrographs
of bronchial biopsies from a nonasthmatic subject
(top row) and an asthmatic patient (bottom row),
stained with anti-MMP7 (left) or anti-IL-25 (right).
Arrows indicate positive immunoreactivity; arrowhead
indicates lack of MMP7 expression. Data are
representative of at least three independent experiments
(n ¼ 5 subjects per group). (b) ELISA
of MMP7 responses in nasal washings from seven
atopic-allergic volunteers (n ¼ 7) challenged
intranasally with increasing doses of ragweed
(RW) or saline (data not shown). *, P o 0.05,
versus 30-minute ragweed challenge (two-tailed
paired t-test). Data are representative of at least
three independent experiments.
500 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
epithelia of subjects with and without asthma and thus it seemed to be
constitutively present, we found MMP7 expression exclusively in the
airways of asthmatic subjects (Fig. 6a). Consistent with those findings,
analysis of nasal washings collected from volunteers with allergic
rhinitis 30 min and 5 h after intranasal challenge with ragweed (a
potent proteinase allergen) showed significantly more MMP7 secretion
in response to ragweed (Fig. 6b) but not in response to the
normal saline control (data not shown). We did not detect secretion of
IL-25 in nasal washes. Nonetheless, these data suggest that constitutive
IL-25 was activated by inducible expression of MMP7 during allergen
challenge of the human airway.
DISCUSSION
We have shown here how proteinase allergens simultaneously activated
pro- and anti-inflammatory programs in airway epithelial cells.
Allergens enhanced MMP7 expression in airway epithelial cells to
maximize allergic lung inflammation; conversely, in the absence of
MMP7, mice developed an attenuated asthmatic phenotype. The
mechanisms responsible for the diminished allergic inflammation in
Mmp7 –/– mice seemed to be in part related to lower expression of
IL-25, which has been shown to be critical in initiating robust T H2
responses 5,7,27,28 . We found that in addition to modulating Il25
expression, MMP7 mediated IL-25 cleavage that was required for
most of the biological activity of IL-25 in allergic inflammation.
Enhanced expression of MMP family members is a frequent
manifestation of active or chronic inflammatory process, but the
precise function of MMPs in these conditions remains poorly understood
13 . Studies have documented that proteinase allergens are able to
induce an innate immune response that further modulates the adaptive
immune response during allergic inflammation 1,2 .Wehaveshownhere
that airway epithelial expression of MMP7 was critical for the development
of asthma-like disease, a finding in contrast to the functions of
MMP2 and MMP9, two gelatinases that are also upregulated in the
same model, as absence of these enzymes leads to exaggerated lung
allergic inflammation 15–17 . Furthermore, we have shown that MMP7
was expressed in the distal lung parenchyma of untreated asthmatics
and that in response to a proteinase allergen derived from ragweed,
MMP7 expression was significantly induced. Although the secretion of
IL-25 in the same conditions was below the detection limit of our
enzyme-linked immunosorbent assay (ELISA), we found IL-25 expression
in the distal lung space in normal and asthmatic volunteers.
Collectively, these findings demonstrate a temporal relation between
exposure to allergens and induction of MMP7 in human allergic
disease and may indicate involvement of activation of IL-25 by
MMP7 in conditions of allergen exposure. It is important to note
that during allergic inflammation, many endogenous proteinases,
including members of the serine and cysteine family, are abundantly
present in the airway 1,29 . However, our finding that MMP7 was
required for the amplification of T H2 responses emphasizes the
nonredundant functions that MMPs may serve in the course of
inflammation. The specificity of this response was exemplified by the
specific proteolysis of IL-25, but not of other allergy-related cytokines
such as TSLP and IL-13, by MMP7. Similarly, MMP7-mediated
proteolytic modification of defensins, apoptotic ligands and cytokines
has been shown to be important in diverse biological functions, such as
innate immunity in the gut to cancer cell metatasis 19,30–32 . The findings
that rIL-25 induced expression of lung MMP7 and intranasal administration
of rIL-25 to Mmp7 –/– mice only weakly induced allergic
inflammation suggest that the bioactivity of IL-25 depends mainly
on proteolysis by MMP7. These data further suggest the existence of a
feedback loop in which IL-25 upregulates lung MMP7 expression,
ARTICLES
followed by cleavage and activation of IL-25 that further enhances both
MMP7 expression and allergic lung disease.
Although regulation of gene expression has been studied for other
members of the IL-17 family, little is known about IL-25 regulation 12 .
We found that although there was little or no RALDH-1 in the airways
of naive mice, this enzyme had prominent expression in the airway
epithelial cells of both wild-type and Mmp7 –/– mice in response to a
proteinase allergen. By high-throughput proteomic analysis, we found
excess RALDH-1 enzyme in CAA-immunized Mmp7 –/– mice that was
functionally relevant because it resulted in higher BAL fluid concentrations
of ATRA and less allergic inflammation. The function of ATRA
in the regulation of adaptive immunity, especially TH2 responses, is
controversial. For example, IL-4-induced eotaxin production, eosinophil
lineage commitment and IL-5 receptor expression are all inhibited
by the addition of ATRA in vitro 33,34 .Furthermore,retinoicacidinthe
presence of transforming growth factor-b1 inhibits the binding of
STAT6 to the promoter of the gene encoding the transcription factor
Foxp3 and effectively inhibits IL-4 signaling in T cells 35 . In other
models of allergic lung disease, deficiency in vitamin A results in lower
expression of the muscarinic M2 receptor, which results in more airway
hyper-reactivity 36 . In contrast, intranasal administration of high concentrations
of ATRA (over 1,500 mg) has been shown to fail to alter
T H2 responses in another asthma model 37 , and vitamin A deficiency
has been shown to attenuate the production of T H2 cytokines in
mice 38 . In our study, allergen-challenged mice that received 100 mg
ATRA showed much lower allergic inflammation and attenuation of
the asthmatic phenotype. Such divergent findings are not unexpected,
given the known idiosyncrasies of retinoid, which can act as a hormone
and, depending on the physiological or pharmaceutical context, can
mediate diverse effects on cells of the immune system 39 .
Studies have indicated involvement of ATRA in the development
and differentiation of inducible regulatory T cells, an antiinflammatory
subset of T cells that are important in the homeostasis
of lymphocytes in various organs 40 . In particular, dendritic cells in the
lamina propria have been found to provide ATRA that is critical for
the differentiation of regulatory T cells and homing of these cells to
the intestinal mucosa 41,42 . Although we did not identify the cellular
origin of ATRA, we found that RALDH-1, the rate-limiting enzyme
that converts retinaldehyde to ATRA, was upregulated in response to
allergic stimulation in the airway epithelium, which indicates a
possible site for ATRA production in the lung. Consistent with the
requirement for ATRA in the differentiation of regulatory T cells
in vivo 25,41,43 , we found more CD25 + Foxp3 + T cells in the lungs
of Mmp7 –/– mice immunized with CAA than in the lungs of
CAA-immunized wild-type mice. Moreover, we found that inhibition
of the synthesis of retinoic acid by citral inhibited the development of
regulatory T cells. The higher percentage of regulatory T cells after
allergen challenge in both wild-type and Mmp7 –/– mice suggested that
RALDH-1 is a negative regulator of allergic inflammation, a plausible
protective response regulated by airway epithelial cells that we found
to be enhanced in the absence of MMP7. What remains unclear is the
mechanism responsible for the higher expression and function of
RALDH-1 in the lung in the absence of MMP7. Our finding of
RALDH-1 expression in airway epithelial cells and macrophages in
response to allergen challenge is in agreement with published reports
indicating an anti-inflammatory function for these cells in allergic
inflammation 44 . Furthermore, depletion of alveolar macrophages
results in an exaggerated allergic response that can be inhibited by
adoptive transfer of alveolar macrophages in brown Norway rats 45 .
Our data suggest that alveolar macrophages that mediate antiinflammatory
responses during allergic inflammation might be acting
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 501

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
through RALDH-1, an idea that is also consistent with the tolerogenic
function of RALDH-1 in gut lamina propria 41,42 .
In summary, we have described here a proinflammatory function
of MMP7 that was activated in response to proteinase
allergens and involved proteolytic modification of the expression
and function of IL-25. In this model, we found activation of
RALDH-1 was ‘downstream’ of allergen-induced inflammation
and that MMP7 actively inhibited the function of this enzyme
in vivo. Thisuniquemechanisminvolved,atleastinpart,inhibition
of regulatory T cells that inhibit lung inflammation. Our
findings thus substantially expand the understanding of the regulation
of IL-25, a critical innate cytokine in allergic responses,
and further identify pro- and anti-inflammatory targets that can
be explored for therapeutic benefit.
METHODS
Mice. Mmp7 –/– mice backcrossed nine generations with C57BL/6 mice were
provided by P.W. Park and were bred in the transgenic animal facility at Baylor
College of Medicine (accredited by the Association for Assessment and
Accreditation of Laboratory Animal Care). All experiments were in accordance
with protocols approved by the Institutional Animal Care and Use Committee
of Baylor College of Medicine. Information on the genotyping of Mmp7 –/– mice
and description of control mice are in the Supplementary Methods online.
Stat6 –/– mice (BALB/c background) and Il13 –/– mice (C57BL/6 background)
from the Jackson Laboratories were bred for ten to twelve generations onto the
BALB/c background.
Experimental model of asthma. CAA, composed of Aspergillus oryzae proteinase
(1 mg/ml; Sigma) and ovalbumin (0.5 mg/ml; Sigma) in a volume of 50 ml,
was administered intranasally every 4 d for a total of five doses as described
before 1,16 . In some experiments, rIL-25 (5 mg in50ml PBS)wasadministered
intranasally twice daily for 3 d. In some experiments, liposomal ATRA or citral
was given by aerosol administration 1 h before CAA challenge. Additional
information on the liposome formulation and administration of ATRA is in the
Supplementary Methods.
Experimental ragweed allergen challenge in human volunteers. Nasal ‘provocation’
with ragweed pollen extract was done as described 46 . The Institutional
Review Board at the University of Texas Medical Branch reviewed and
approved the protocol and consent forms. The challenge procedure and
analysis of nasal lavage fluid and bronchial biopsies are described in the
Supplementary Methods.
Analysis of asthmatic phenotype. All data were collected 24 h after the final
allergen challenge. Airway hyper-reactivity was assessed by estimation of the
concentration of acetylcholine (in mg per g body weight) that caused a 200%
increase in airway resistance over the baseline, calculated by linear interpolation
of the appropriate dose-response curves as described 15 . Collection and analysis
of BAL fluid and lung tissue is described in the Supplementary Methods.
Immunohistochemistry. A standard published protocol was used for immunohistochemistry
47 . Deparaffinized sections were washed with 3% (vol/vol)
H2O2 and 2% (vol/vol) Triton X-100 in PBS, followed by antigen retrieval
(Target Retrieval Solution; Dako). Slides were incubated overnight at 4 1C with
rabbit polyclonal antibody to mouse ALDH1A1 (24343; Abcam), rat monoclonal
antibody to mouse MMP7 (377314; R&D Systems), mouse monoclonal
antibody to human MMP7 (377314 (R&D Systems) or ID2 (Calbiochem)) and
mouse monoclonal antibody to human IL-17E. Additional information on
staining and analysis is in the Supplementary Methods.
Measurement of cytokines and chemokines in BAL fluid. CCL11, CCL7 and
CCL17 in BAL fluid were measured by standard antibody-based ELISA and/or a
multiplex bead-based cytokine detection kit (BioRad) as described 16 .Reagents
and the LINCOplex assay are described in the Supplementary Methods.
Proteomics analysis of BAL protein. Proteomics analysis of BAL fluid was
done as described 17 . BAL samples from wild-type and Mmp7 –/– mice were
pooled and were concentrated with AMICON filters, and protein concentrations
were measured by BCA protein assay (bicinchoninic acid; Pierce). An
equal amount of protein from each pooled sample (50 mg) was labeled with
400 pmol fluorescent dye (CyDye; Amersham Biosciences) matched by
charge and molecular size and was brought to a final volume of 250 ml with
sample buffer that also contained 0.5% (wt/vol) ampholytes (pH 3–11, nonlinear),
0.1% (wt/vol) bromophenol blue and DeStreak reagent (12 ml/ml;
GE Healthcare). Additional information on proteomics analysis is in the
Supplementary Methods.
Quantitative RT-PCR. The RNeasy mini kit (Qiagen) was used for extraction of
total cellular RNA from mouse lung tissues stabilized with the RNA Later
reagent (Qiagen). One-step real-time quantitative RT-PCR (Real-Time PCR
system 7300; Applied Biosystems) was used to assess the expression of Ccl11
(encoding CCL11 (eotaxin 1)) and Aldh1a1 (encoding RALDH-1). The primer
and probe mixtures of target genes (Ccl11, Mm00441238_m1; Raldh-1,
Mm01194995_mH; Applied Biosystems identifiers) and 18S rRNA were from
Applied Biosystems. Additional information on RT-PCR is in the Supplementary
Methods.
MMP activation and in vitro cleavage assay. Carrier-free mouse rIL-25 (5 mg;
R&D), mouse rIL-13 (Peprotech) and mouse rTSLP (R&D Systems) were
incubated for 2 h at 37 1C with0.5mg 4-aminophenylmercuric acetate–activated
MMP7 in the presence or absence of the MMP inhibitor 1,10-phenanthroline
(10 mM; Sigma). After incubation, equal volumes of each sample (16.5 ml) were
reduced and were resolved by separation through a 16.5% tricine gel. Cleaved
and uncleaved proteins were visualized with the Proteosilver Plus Silver Stain kit
according to the manufacturer’s instructions (Sigma-Aldrich). Additional information
on these assays is in the Supplementary Methods. Amino-terminal
sequencing of IL-25 was done by standard protein-sequencing methods (Procise
492cLC; Applied Biosystems).
Cell culture. Lymph node cells and splenocytes isolated from C57BL/6 mice
were stimulated with plate-bound antibody to CD3 (anti-CD3; 2 mg/ml) and
human IL-2 (50 U/ml) in the presence of mouse rIL-25 (250 ng/ml; R&D
Systems) or rIL-25’C (250 ng/ml). After 2 d, culture supernatants were collected
and analyzed for IL-4, IL-5, IL-13 and interferon-g by ELISA as described 5 .On
day 3, cells were restimulated for 5 h with ionomycin (500 ng/ml) and phorbol
12-myristate 13-acetate (50 ng/ml) in the presence of GolgiStop (BD Biosciences).
Cells were made permeable with a Cytofix/Cytoperm kit (BD
Biosciences) and were analyzed for intracellular expression of IL-4 (with
phycoerythrin-conjugated anti-IL-4; 11B11) and IL-5 (with allophycocyaninconjugated
anti-IL-5; TRFK5; both from BD Biosciences).
Flow cytometry. Total lung eosinophils were quantified by bead-enhanced flow
cytometry as described 48 . Total mouse lung cells were stained with fluorescein
isothiocyanate–conjugated conjugated antibody to major histocompatibility
class II (2G9; 553623; BD Biosciences) and phycoerythrin-conjugated anti-
Siglec-F (1 mg per10 6 cells; E50-2440; BD Biosciences), followed by incubation
with fluorescent beads (Flow-Check Fluorospheres; Beckman Coulter). Samples
were acquired with a flow cytometer (XL2; Beckman Coulter) and cell and bead
counts were measured. Absolute numbers of eosinophils were calculated with
formulas as described 48 . Additional information on flow cytometry is in the
Supplementary Methods.
High-performance liquid chromatography (HPLC). Pooled BAL samples
from wild-type and Mmp7 –/– mice were evaporated under nitrogen and the
dry residue was dissolved in 100 ml acetonitrile (Fisher). Aliquots of 25 ml were
then analyzed with the Waters HPLC system, consisting of a WISP autosampler
(model 717 Plus), dual-absorbance detector (model 2487), pump (model 515)
and Nova-Pak C18 columns (3.9 mm 150 mm). ATRA was separated with a
solvent system of 57.5% (vol/vol) acetonitrile, 25% (vol/vol) acetic acid (2%
solution) at a flow rate of 1.3 ml/min. Data were analyzed with Waters
Millennium software (version 3.2).
Binding assay. Analysis of the binding of IL-25 to the fusion protein of IL-17
receptor B and Fc fragment is described in the Supplementary Methods.
502 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Statistics. All data were analyzed with Prism 4.0a statistical analysis software.
Significant differences (P o 0.05) between treatment groups were identified
with a t-test, one-way analysis of variance (ANOVA) or Bonferroni multiplecomparison
test.
Accession code. UCSD-Nature Signaling Gateway (http://www.signalinggateway.org):
A001477.
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank P. Woo Park (Children’s Hospital, Harvard School of Medicine) for
Mmp7 –/– mice backcrossed to C57BL/6 mice; D.A. Engler, R.K. Matsunami and
K. Gonzalez for technical help with proteomics analysis; N. Barrows for editorial
support; and all members of the Kheradmand and Corry laboratory for
comments and criticisms. Supported by the US National Institutes of Health
(AI070973 and HL082487 to F.K.; AI071130 and AR050772 to C.D.; and
HL075243, AI057696 and AI070973 to D.B.C.), the American Lung Association
(C.D.), the Leukemia and Lymphoma Society (C.D.) and MD Anderson Cancer
Center (C.D.).
AUTHOR CONTRIBUTIONS
S.G. did animal experiments, in vitro cleavage assay, immunohistochemistry,
flow cytometry, RT-PCR, ELISA and Luminex assays; P.A. did in vitro T cell
experiments; W.T.B. and S.P. did airway physiology experiments; K.J.G. did
proteomics analysis; A.S. made liposomal ATRA and helped with HPLC; M.S.,
L.S., D.R., B.S. and S.S. did the ragweed challenges of humans and ELISA of
samples from allergic volunteers; P.W. obtained bronchial biopsies of asthmatic
and control volunteers; S.G., F.K., D.B.C. and C.D. designed experiments;
S.G. and F.K. wrote the manuscript; and P.A., D.B.C. and C.D. critically
reviewed the manuscript.
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Kheradmand, F. et al. A protease-activated pathway underlying Th cell type 2 activation
and allergic lung disease. J. Immunol. 169, 5904–5911 (2002).
2. Kiss, A. et al. A new mechanism regulating the initiation of allergic airway inflammation.
J. Allergy Clin. Immunol. 120, 334–342 (2007).
3. Li, L. et al. Effects of Th2 cytokines on chemokine expression in the lung: IL-13
potently induces eotaxin expression by airway epithelial cells. J. Immunol. 162,
2477–2487 (1999).
4. Zhu, J. et al. Exacerbations of bronchitis: bronchial eosinophilia and gene expression for
interleukin-4, interleukin-5, and eosinophil chemoattractants. Am. J. Respir. Crit. Care
Med. 164, 109–116 (2001).
5. Angkasekwinai, P. et al. Interleukin 25 promotes the initiation of proallergic type 2
responses. J. Exp. Med. 204, 1509–1517 (2007).
6. Moseley, T.A., Haudenschild, D.R., Rose, L. & Reddi, A.H. Interleukin-17 family and
IL-17 receptors. Cytokine Growth Factor Rev. 14, 155–174 (2003).
7. Pan, G. et al. Forced expression of murine IL-17E induces growth retardation, jaundice,
a Th2-biased response, and multiorgan inflammation in mice. J. Immunol. 167,
6559–6567 (2001).
8. Kim, M.R. et al. Transgenic overexpression of human IL-17E results in eosinophilia, Blymphocyte
hyperplasia, and altered antibody production. Blood 100, 2330–2340
(2002).
9. Ballantyne, S.J. et al. Blocking IL-25 prevents airway hyperresponsiveness in allergic
asthma. J. Allergy Clin. Immunol. 120, 1324–1331 (2007).
10. Fallon, P.G. et al. Identification of an interleukin (IL)-25-dependent cell population that
provides IL-4, IL-5, and IL-13 at the onset of helminth expulsion. J. Exp. Med. 203,
1105–1116 (2006).
11. Greenlee, K.J., Werb, Z. & Kheradmand, F. Matrix metalloproteinases in lung: multiple,
multifarious, and multifaceted. Physiol. Rev. 87, 69–98 (2007).
12. Dong, C. TH17 cells in development: an updated view of their molecular identity and
genetic programming. Nat. Rev. Immunol. 8, 337–348 (2008).
13. Parks, W.C., Wilson, C.L. & Lopez-Boado, Y.S. Matrix metalloproteinases as
modulators of inflammation and innate immunity. Nat. Rev. Immunol. 4, 617–629
(2004).
14. McMillan, S.J. et al. Matrix metalloproteinase-9 deficiency results in enhanced
allergen-induced airway inflammation. J. Immunol. 172, 2586–2594 (2004).
15. Corry, D.B. et al. Overlapping and independent contributions of MMP2 and MMP9 to
lung allergic inflammatory cell egression through decreased CC chemokines. FASEB J.
18, 995–997 (2004).
16. Corry, D.B. et al. Decreased allergic lung inflammatory cell egression and increased
susceptibility to asphyxiation in MMP2-deficiency. Nat. Immunol. 3, 347–353
(2002).
ARTICLES
17. Greenlee, K.J. et al. Proteomic identification of in vivo substrates for matrix metalloproteinases
2 and 9 reveals a mechanism for resolution of inflammation. J. Immunol.
177, 7312–7321 (2006).
18. Li, Q., Park, P.W., Wilson, C.L. & Parks, W.C. Matrilysin shedding of syndecan-1
regulates chemokine mobilization and transepithelial efflux of neutrophils in acute lung
injury. Cell 111, 635–646 (2002).
19. Wilson, C.L. et al. Regulation of intestinal a-defensin activation by the metalloproteinase
matrilysin in innate host defense. Science 286, 113–117 (1999).
20. Gearing, A.J. et al. Matrix metalloproteinases and processing of pro-TNF-a. J. Leukoc.
Biol. 57, 774–777 (1995).
21. Corry, D.B. IL-13 in allergy: home at last. Curr. Opin. Immunol. 11, 610–614 (1999).
22. Corry, D.B. et al. Requirements for allergen- induced airway hyperreactivity in T and B
cell-deficient mice. Mol. Med. 4, 344–355 (1998).
23. Grunig, G. et al. Requirement for IL-13 independently of IL-4 in experimental asthma.
Science 282, 2261–2263 (1998).
24. Denning, T.L., Wang, Y.C., Patel, S.R., Williams, I.R. & Pulendran, B. Lamina propria
macrophages and dendritic cells differentially induce regulatory and interleukin 17producing
T cell responses. Nat. Immunol. 8, 1086–1094 (2007).
25. Mucida, D. et al. Reciprocal Th-17 and regulatory T cell differentiation mediated by
retinoic acid. Science 317, 256–260 (2007).
26. Connor, M.J. & Smit, M.H. Terminal-group oxidation of retinol by mouse epidermis.
Inhibition in vitro and in vivo. Biochem. J. 244, 489–492 (1987).
27. Cheung, P.F., Wong, C.K., Ip, W.K. & Lam, C.W. IL-25 regulates the expression of
adhesion molecules on eosinophils: mechanism of eosinophilia in allergic inflammation.
Allergy 61, 878–885 (2006).
28. Fort, M.M. et al. IL-25 induces IL-4, IL-5, and IL-13 and Th2-associated pathologies
in vivo. Immunity 15, 985–995 (2001).
29. Zheng, T. et al. Inducible targeting of IL-13 to the adult lung causes matrix
metalloproteinase- and cathepsin-dependent emphysema. J. Clin. Invest. 106,
1081–1093 (2000).
30. Swee, M., Wilson, C.L., Wang, Y., McGuire, J.K. & Parks, W.C. Matrix metalloproteinase-7
(matrilysin) controls neutrophil egress by generating chemokine gradients.
J. Leukoc. Biol. 83, 1404–1412 (2008).
31. Liu, D. et al. Overexpression of matrix metalloproteinase-7 (MMP-7) correlates with
tumor proliferation, and a poor prognosis in non-small cell lung cancer. Lung Cancer
58, 384–391 (2007).
32. Ito, T.K., Ishii, G., Chiba, H. & Ochiai, A. The VEGF angiogenic switch of
fibroblasts is regulated by MMP-7 from cancer cells. Oncogene 26, 7194–7203
(2007).
33. Upham, J.W. et al. Retinoic acid modulates IL-5 receptor expression and selectively
inhibits eosinophil-basophil differentiation of hemopoietic progenitor cells. J. Allergy
Clin. Immunol. 109, 307–313 (2002).
34. Takamura, K. et al. Retinoic acid inhibits interleukin-4-induced eotaxin production in a
human bronchial epithelial cell line. Am. J. Physiol. Lung Cell. Mol. Physiol. 286,
L777–L785 (2004).
35. Takaki, H. et al. STAT6 Inhibits TGF-b1-mediated Foxp3 induction through direct
binding to the Foxp3 promoter, which is reverted by retinoic acid receptor. J. Biol.
Chem. 283, 14955–14962 (2008).
36. McGowan, S.E., Holmes, A.J. & Smith, J. Retinoic acid reverses the airway hyperresponsiveness
but not the parenchymal defect that is associated with
vitamin A deficiency. Am. J. Physiol. Lung Cell. Mol. Physiol. 286, L437–L444
(2004).
37. Maret, M. et al. Liposomal retinoic acids modulate asthma manifestations in mice.
J. Nutr. 137, 2730–2736 (2007).
38. Schuster, G.U., Kenyon, N.J. & Stephensen, C.B. Vitamin A deficiency decreases and
high dietary vitamin A increases disease severity in the mouse model of asthma.
J. Immunol. 180, 1834–1842 (2008).
39. Mora, J.R., Iwata, M. & von Andrian, U.H. Vitamin effects on the immune system:
vitamins A and D take center stage. Nat. Rev. Immunol. 8, 685–698 (2008).
40. Coombes, J.L. & Powrie, F. Dendritic cells in intestinal immune regulation. Nat. Rev.
Immunol. 8, 435–446 (2008).
41. Sun, C.-M. Small intestine lamina propria dendritic cells promote de novo generation of
Foxp3 T reg cells via retinoic acid. J. Exp. Med. 204, 1775–1785 (2007).
42. Benson, M.J., Pino-Lagos, K., Rosemblatt, M. & Noelle, R.J. All-trans retinoic acid
mediates enhanced T reg cell growth, differentiation, and gut homing in the face of high
levels of co-stimulation. J. Exp. Med. 204, 1765–1774 (2007).
43. Elias, K.M. et al. Retinoic acid inhibits Th17 polarization and enhances FoxP3
expression through a Stat-3/Stat-5 independent signaling pathway. Blood 111,
1013–1020 (2008).
44. Peters-Golden, M. The alveolar macrophage: the forgotten cell in asthma. Am.
J. Respir. Cell Mol. Biol. 31, 3–7 (2004).
45. Careau, E. et al. Antigen sensitization modulates alveolar macrophage functions in
an asthma model. Am. J. Physiol. Lung Cell. Mol. Physiol. 290, L871–L879
(2006).
46. Diaz-Sanchez, D., Tsien, A., Fleming, J. & Saxon, A. Combined diesel exhaust
particulate and ragweed allergen challenge markedly enhances human in vivo nasal
ragweed-specific IgE and skews cytokine production to a T helper cell 2-type pattern.
J. Immunol. 158, 2406–2413 (1997).
47. Grumelli, S. et al. An immune basis for lung parenchymal destruction in chronic
obstructive pulmonary disease and emphysema. PLoS Med. 1, e8(2004).
48. Montes, M., Jaensson, E.A., Orozco, A.F., Lewis, D.E. & Corry, D.B. A general method
for bead-enhanced quantitation by flow cytometry. J. Immunol. Methods 317, 45–55
(2006).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 503

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Transcription factor Foxo3 controls the magnitude of
T cell immune responses by modulating the function
of dendritic cells
Anne S Dejean 1 , Daniel R Beisner 1,4 , Irene L Ch’en 1 , Yann M Kerdiles 1 , Anna Babour 1 , Karen C Arden 2 ,
Diego H Castrillon 3,4 , Ronald A DePinho 3 & Stephen M Hedrick 1
Foxo transcription factors regulate cell cycle progression, cell survival and DNA-repair pathways. Here we demonstrate that
deficiency in Foxo3 resulted in greater expansion of T cell populations after viral infection. This exaggerated expansion was not
T cell intrinsic. Instead, it was caused by the enhanced capacity of Foxo3-deficient dendritic cells to sustain T cell viability by
producing more interleukin 6. Stimulation of dendritic cells mediated by the coinhibitory molecule CTLA-4 induced nuclear
localization of Foxo3, which in turn inhibited the production of interleukin 6 and tumor necrosis factor. Thus, Foxo3 acts to
constrain the production of key inflammatory cytokines by dendritic cells and to control T cell survival.
Foxo transcription factors guide the cellular response to growth factors,
nutrients and stress. This information is ‘encoded’ as post-translational
modifications, referred to as the ‘Foxo code’, that govern Foxo
intracellular localization, cofactor associations and transcriptional
activity. The resulting program of gene expression regulates cell cycle
arrest, repair, apoptosis and autophagy and many other aspects of
cellular homeostasis 1 . In mammals, four Foxo members have been
identified. Foxo1 (FKH1 and FKHR) 2 , Foxo3 (FKHRL1; A000945) 3 ,
and Foxo4 (AFX) 4 are widely expressed and similarly regulated,
whereas expression of Foxo6 (ref. 5) is confined to specific structures
of the brain and is subject to distinct regulatory mechanisms.
Insulin, insulin-like growth factor and other growth factors induce
activation of phosphatidylinositol-3-OH kinase and the kinase Akt,
which phosphorylate three Foxo amino acids 6 . These modifications
result in the association of the Foxo protein with the adaptor protein
14-3-3 and the nuclear exclusion and eventual degradation of Foxo3
(refs. 7–9). This process can be actively opposed by stress-induced
signals that activate the Jnk mitogen-activated protein kinase, which
result in Foxo3 nuclear localization 10 .Foxofactortargetspecificitycan
be further refined by the SIRT1 deacetylase 11,12 .
Given their involvement in coordinating cellular growth, proliferation
and survival, we predicted that Foxo factors would be central to
the highly dynamic, infection-mediated expansion and contraction of
antigen-specific T cell populations. Indeed, Foxo activity is regulated
by signaling by the T cell antigen receptor (TCR) and CD28, as well as
by cytokines such as interleukin 2 (IL-2) 13 , IL-3 (ref. 14) and IL-7
Received 22 September 2008; accepted 19 March 2009; published online 12 April 2009; doi:10.1038/ni.1729
(refs. 15,16). These stimuli result in the phosphorylation of Foxo by
Akt, the kinase Sgk or inhibitor of transcription factor NF-kB (IkB)
kinase, and in Foxo3 nuclear exclusion 17 . Withdrawal of growth factor
causes dephosphorylation of nuclear Foxo3, binding of Foxo3 to the
promoters of the genes encoding the proapoptotic molecules Bim and
Puma, induction of transcription of those genes, and T cell apoptosis
18 . In contrast, enforced expression of a constitutively nuclear
form of Foxo3 in T cell lines causes cell cycle arrest 13 . Furthermore,
Foxo3 is involved in the persistence of CD4 + central memory T cells in
mice and humans 19 .
Published studies have shown spontaneous T cell activation, lymphoproliferative
disease and organ infiltration in ‘Foxo3 Trap ’mice,which
have a mutated Foxo3 allele generated by gene-trap technology 20 .Such
studies indicate an important function for Foxo3 in immune regulation,
although the underlying molecular mechanisms are not understood. In
addition, Foxo3 regulates superoxide dismutase in dendritic cells (DCs);
this signaling pathway may affect the suppressive or stimulating
characteristics of plasmacytoid DCs 21,22 .
In this study, we sought to determine whether Foxo3 is involved in
the quiescence of cells of the immune system and the dynamics of
T cell population expansion and contraction. We found that in
response to LCMV infection, Foxo3 deficiency caused superabundant
expansion of antigen-specific T cell populations. However, contrary to
our expectations, this phenotype was not intrinsic to T cells but
instead arose from altered stimulatory properties of Foxo3-deficient
DCs. Foxo3 deficiency resulted in a DC-specific higher production of
1 Molecular Biology Section, Division of Biological Sciences, and Department of Cellular and Molecular Medicine, University of California, San Diego, San Diego, California,
USA. 2 Ludwig Institute for Cancer Research, University of California, San Diego School of Medicine, La Jolla, California, USA. 3 Center for Applied Cancer Science, Belfer
Institute for Innovative Cancer Science, Departments of Adult Oncology, Medicine and Genetics, Dana-Farber Cancer Institute, Harvard Medical School, Boston,
Massachusetts, USA. 4 Present addresses: Genetics Institute of the Novartis Foundation, San Diego, California, USA (D.R.B.) and Department of Pathology, University of
Texas, Southwestern Medical Center at Dallas, Dallas, Texas, USA (D.H.C.). Correspondence should be addressed to S.M.H. (shedrick@ucsd.edu).
504 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Table 1 Leukocyte subsets
Spleen B6 B6 FVB FVB
Foxo3 +/+ Foxo3Kca Foxo3 +/+ Foxo3 –/–
CD4 + 15.0 ± 2.0 17.3 ± 1.6 29.3 ± 5.1 31.0 ± 5.9
CD8 + 13.4 ± 3.5 9.7 ± 1.3 13.0 ± 2.2 14.9 ± 1.9
B220 + 46.9 ± 6.7 51.7 ± 9.2 58.2 ± 9.7 56.3 ± 12.1
Gr-1hiCD11bhi Lymph node
1.4 ± 0.2 6.0 ± 1.1* 2.0 ± 0.2 5.4 ± 0.3*
CD4 + 10.3 ± 1.4 9.4 ± 1.2 11.0 ± 2.5 12.8 ± 2.4
CD8 + 7.9 ± 1.0 7.3 ± 0.7 5.5 ± 1.2 7.0 ± 1.4
B220 + 13.8 ± 1.4 12.4 ± 1.1 4.0 ± 1.2 3.6 ± 0.7
Results are presented as the number of cells 10 6 (± s.e.m.). *, P o 0.001, wild-type
versus Foxo3-deficient mice (two-sample t-test assuming equal variances). Data are
representative of two experiments with five to six mice per group.
IL-6. Nuclear localization of Foxo3 induced by signals initiated by
cytotoxic T lymphocyte–associated antigen 4 (CTLA-4; A000706)
suppressed Toll-like receptor–induced production of IL-6. Our results
collectively emphasize the importance of Foxo3 in restraining the
production of inflammatory cytokines by DCs and T cell viability.
RESULTS
No spontaneous activation of Foxo3-deficient T cells
To examine the function of Foxo3 in the immune system, we used two
independently derived Foxo3-null strains distinct from the Foxo3 Trap
mutant 20 . One strain, called ‘Foxo3 Kca ’ here, was backcrossed to the
C57BL/6 strain, whereas mice of the other strain, called ‘Foxo3 –/– ’here,
were maintained as congenic FVB mice. Although we detected strainspecific
differences, deletion of Foxo3 had no effect on the proportion
of T cells or B cells in the spleens and lymph nodes of 6- to 10-weekold
mice (Table 1). However, we noted a greater proportion of
Gr-1 hi CD11b hi cells, which include granulocytes and macrophages,
in the spleens of both Foxo3-deficient strains. We detected neither
lymphocytic organ infiltration nor lymphadenopathy 23 (data not
shown) in 6- to 10-week-old Foxo3-deficient mice. However 3- to
6-month-old Foxo3-deficient mice had enlargement of the spleen
a b
CD44
CD69
B6
Foxo3 +/+
B6
Foxo3 Kca
FVB
Foxo3 +/+
FVB
Foxo3 –/–
7.2 ± 0.6 7 ± 1.1 4.1 ± 0.6 4.9 ± 0.5
CD62L
14.5 ± 0.9 14.3 ± 1.1 12 ± 0.9 13.4 ± 1.3
CD4
CD8
CD69
B6
Foxo3 +/+
B6
Foxo3 Kca
FVB
Foxo3 +/+
FVB
Foxo3 –/–
13.2 ± 0.7 14.4 ± 1.5 13.4 ± 1.1 15.8 ± 1.4
CD44
7.4 ± 0.3 7.4 ± 0.3 8 ± 0.8 8.1 ± 1.0
CD8
Figure 1 Foxo3-deficient mice show no spontaneous T cell
activation. (a) Flow cytometry of the expression of CD44, CD62L
and CD69 by T cells from C57BL/6 (B6) Foxo3 Kca and FVB
Foxo –/– mice and their wild-type (Foxo +/+ ) congenic littermates
(n ¼ 5–6 mice per genotype), gated on CD4 + T cells (left) or CD8 +
T cells (right). Numbers adjacent to outlined areas indicate percent
cells in gate (mean ± s.e.m.). (b) Proliferation of CFSE-labeled
CD4 + T cells from Foxo3 Kca and Foxo –/– mice and their wild-type
littermates, activated in the presence of anti-CD3 (a-CD3) alone or
CD4 + c T cells
0.6 ± 0.1
IL-4
10 ± 1
0.7 ± 0.1
IFN-γ
and more erythrocyte progenitors (Ter119 + ; data not shown). That
observation probably relates to the shorter erythrocyte lifespan
and lower rate of erythrocyte maturation in Foxo3-deficient mice
reported before 24 . Next we measured the expression of CD69, CD44
and CD62L (L-selectin) on T cells from Foxo3-deficient mice and
their wild-type littermates. Although we again detected differences
between the C57BL/6 and FVB strains, inactivation of Foxo3 had no
effect on the proportion of activated (CD69 hi ) or memory-effector
(CD44 hi CD62L lo )Tcells(Fig. 1a).
Published studies have shown constitutive activation of the NF-kB
pathway in Foxo3 Trap mice, as reflected by the absence of IkB 20 .
However, T cells purified from Foxo3 Kca and Foxo3 –/– mice showed
no change in the expression of IkBa, IkBb or IkBe protein relative to
that of wild-type T cells (Supplementary Fig. 1 online). To measure
the effect of Foxo3 deficiency on T cell proliferation, we labeled
purified lymph node T cells with the cytosolic dye CFSE and then
cultured the cells with plate-bound antibody to CD3 (anti-CD3) with
or without anti-CD28. Neither the number of cell divisions nor the
accumulation of cells in each division was altered by Foxo3 deficiency
(Fig. 1b). Also, freshly explanted Foxo3 Kca and wild-type CD4 + and
CD8 + T cells produced similar amounts of cytokines (Fig. 1c). As FasL
(CD95L), the ligand for the cell surface receptor Fas, is a known target
of Foxo3, we also analyzed FasL expression on unstimulated and
antibody-stimulated T cells from wild-type and Foxo3 Kca mice; however
we found no differences related to Foxo3 status (data not shown).
These results suggest that loss of Foxo3 alone is not sufficient to elicit
manifestations of T cell activation and spontaneous autoimmunity.
Foxo3 Kca mice show enhanced T cell accumulation
As Foxo3 is involved in cell cycle progression and apoptosis 25 ,we
sought to investigate its function in the dynamics of T cell population
expansion and contraction in response to viral infection. Historically,
the C57BL/6 strain has been used to study the progression of viral
responses, and as the two mutant strains homozygous for Foxo3
deficiency had identical profiles here and in published results 23,26 ,our
further efforts at characterization focused on the C57BL/6 Foxo3 Kca
mice. We infected Foxo3 Kca mice and their wild-type littermates with
12 ± 2
IL-17
0.5 ± 0.1
10 ± 1
0.5 ± 0.1
IFN-γ
12 ± 2
CD4
IL-2
B6
FVB
Cells
Control α-CD3
α-CD3 +
α-CD28
500 400 1,000
1,000 400 2,000
CFSE
Foxo3 +/+
20 ± 2
Foxo3 Kca
22 ± 1
TNF
CD8 + T cells
32 ± 3 22 ± 2
4 ± 1
27 ± 6 22 ± 2
CD8
6 ± 2
IFN-γ IL-2
ARTICLES
12 ± 2
13 ± 3
Foxo3 Kca
Foxo3 +/+
Foxo3 –/–
Foxo3 +/+
with anti-CD28 (a-CD28) and assessed by CFSE dilution after 72 h of stimulation. Numbers in plots indicate number of accumulated CFSE + cells.
(c) Cytokine secretion by splenocytes from Foxo3 Kca mice and their wild-type littermates (n ¼ 6 mice per group), activated for 3 h in the presence of phorbol
12-myristate 13-acetate and ionomycin and analyzed by intracellular staining with gating on CD4 + T cells (left) or CD8 + T cells (right). Numbers in quadrants
and adjacent to outlined areas indicate percent cells in each (mean ± s.e.m.). Similar results were obtained with lymph node T cells. Data are from five
separate experiments (a) or are representative of three independent experiments (b,c).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 505
Foxo3 +/+
Foxo3 Kca

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a b LCMV No infection c
T cells (× 10 6 )
gp61-specific
CD4
4
+ T cells
**
3
2
1
**
*
T cells (× 10 6 )
gp33-specific
CD8 + T cells
30
**
Foxo3 Kca
0
0
0 10 20 50 60 0 10 20 50 60
Time (d) Time (d)
20
10
Foxo3 +/+
Adoptive transfer
P14 Foxo3 +/+
P14 Foxo3 Kca
Adoptive transfer
lymphocytic choriomeningitis virus (LCMV) and assessed LCMVresponsive
T cell populations at various times after infection. Whereas
the Foxo3 genotype had no effect on the kinetics of T cell population
expansion, Foxo3 Kca mice developed a threefold greater accumulation
of LCMV-specific T cells than did their wild-type littermates (Fig. 2).
Measurement of virus in the liver at day 8 after infection showed
complete viral clearance in both strains (data not shown).
The lack of an effect of Foxo3 deficiency on T cells stimulated in
culture prompted us to test whether the enhanced T cell accumulation
during an LCMV response was T cell intrinsic. We transferred T cells
from LCMV-specific TCR-transgenic P14 Foxo3 Kca or P14 wild-type
mice into wild-type C57BL/6 mice and measured the accumulation of
P14 T cells after infection of the recipient mice with LCMV. We
detected no significant difference in the accumulation of Foxo3 Kca and
wild-type P14 T cells (Fig. 2b). To further examine this issue, we
transferred wild-type P14 T cells into either wild-type or Foxo3 Kca
mice and infected recipients with LCMV. P14 T cells transferred into
Foxo3 Kca mice accumulated in greater numbers than did P14 T cells
transferred into wild-type hosts (Fig. 2b). We reproduced that result
with the transfer of ovalbumin (OVA)-specific OT-I T cells into wildtype
and Foxo3 Kca hosts that we subsequently infected with OVAexpressing
vesicular stomatitis virus (Supplementary Fig. 2 online).
To further investigate the cellular cause of the enhanced T cell
population expansion, we reconstituted irradiated wild-type mice with
bone marrow from wild-type or Foxo3 Kca mice. After 8 weeks, we
infected the recipient mice with LCMV and analyzed expansion of the
LCMV-specific T cell population. Excessive T cell accumulation was
apparent in recipients of Foxo3 Kca bone marrow (Fig. 2c), which
indicated involvement of a bone marrow–derived cell type in this
phenotype. Thus, a non–T cell bone marrow–derived cell type was
responsible for the enhanced T cell proliferation and/or survival in
Foxo3-deficient mice.
Stimulatory capacity of DCs from Foxo3 Kca mice
The efficiency of antigen presentation can affect the magnitude
of a T cell response to viral infection 27 . Therefore, we assessed
whether Foxo3 regulates the number, phenotype and/or function of
P14
P14
Foxo3 +/+
Foxo3 Kca
T cells (× 10 6 )
T cells (× 10 6 )
4
3
2
1
0
60
40
20
Foxo3 Kca
LCMV No infection
**
antigen-presenting DCs. Naive Foxo3 Kca mice had significantly more
DCs than did their wild-type littermates (Fig. 3a), and there was a
slightly greater proportion of DCs with high expression of the
costimulatory molecules B7-1 (CD80) and B7-2 (CD86). We found
no difference in the expression of major histocompatibility complex
(MHC) class I, MHC class II or CD40 on DCs from these mice
(Fig. 3b), which suggested that a small number of Foxo3-deficient
DCs had a more mature phenotype.
Analysis of various DC subsets showed that Foxo3-deficient mice
had more CD11c + CD11b + CD8 – , CD11c + CD11b – CD8 + and
CD11c + B220 + DCs (Fig. 3c). All subsets showed a moderate shift
toward high expression of B7-1 and B7-2 (Fig. 3d), although the
biological importance of this shift is unclear. To further characterize
the activation status of DCs from naive mice, we cultured splenic DCs
overnight and then measured secreted cytokines. Foxo3 Kca DCs
produced more IL-6, tumor necrosis factor (TNF) and chemokine
CCL2 (MCP-1) than did wild-type DCs (Fig. 3e). Interferon-g
(IFN-g), IL-10 and IL-12 were undetectable (data not shown).
To determine whether DCs from Foxo3 Kca mice had enhanced
antigen presentation, we infected mice with LCMV and analyzed
DCs 3 d after infection, when virus is still present and T cell
populations are expanding exponentially 28 . Compared with wildtype
DCs, Foxo3 Kca DCs had slightly higher expression of B7-1,
B7-2 and MHC class II (Fig. 4a) and more effectively stimulated
the accumulation of P14 T cells, as measured by CFSE dilution and
staining with annexin V and 7-amino-actinomycin D (7-AAD;
Fig. 4b). This enhanced stimulatory capacity was not due to a
difference in antigen processing, as differences between cultures with
Foxo3 Kca and wild-type DCs remained even after DCs were pulsed
with LCMV glycoprotein 33 (gp33) peptide (Fig. 4b).
To rule out the possibility that Foxo3-deficient DCs from LCMVinfected
mice can cause proliferation of bystander T cells independently
of the presence of antigen, we also cultured DCs from
LCMV-infected mice with OT-I T cells with or without OVA
peptide (amino acids 257–264; OVA (257–264)). In the absence of
OVA (257–264), there was no detectable division of OT-I T cells
(Fig. 4b), which ruled out the possibility of bystander effects. Finally,
T cells (× 10 6 )
Foxo3 +/+
Foxo3 Kca
gp61-specific
CD4 + T cells
2.5
**
2.0
1.5
1.0
0.5
0.0
Radiation chimera
gp33-specific
CD8 + T cells
5 *
4
Figure 2 Foxo3 regulates the magnitude of an LCMV-induced immune
0
response. (a) TheCD4 + and CD8 + T cell responses to LCMV in Foxo3 Kca
mice and their wild-type littermates infected with LCMV, Armstrong strain
(2 105 plaque-forming units), analyzed after restimulation with gp61 or
gp33 peptide, respectively, by intracellular staining for IFN-g. (b) P14
T cells in recipient mice given CD45.2 congenic wild-type or Foxo3 Kca P14 T cells (transferred into (-) CD45.1 wild-type mice; top) and CD45.1 congenic
wild-type P14 T cells (transferred into CD45.2 wild-type or CD45.2 Foxo3Kca mice; bottom) and then left uninfected (right) or infected with LCMV (left) and
assessed with a congenic marker 8 d later. (c) Virus-specific T cells in lethally irradiated (gray ‘lightning bolt’ around) CD45.1 wild-type mice reconstituted
with CD45.2 congenic wild-type or Foxo3Kca bone marrow and then infected with LCMV 8 weeks later, assessed by specific peptide restimulation and
intracellular IFN-g staining. *, P o 0.01, and **, P o 0.005 (unpaired two-tailed Student’s t-test). Data are representative of four (a) ortwo(b,c)
independent experiments with at least three mice per group (error bars, s.e.m.).
506 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
Foxo3 +/+
Foxo3 +/+
Foxo3 +/+
Foxo3 +/+
Foxo3 Kca
Foxo3 +/+
Foxo3 +/+
Foxo3 Kca
Foxo3 Kca
Foxo3 +/+
Foxo3 Kca
3
2
1
0
Foxo3 +/+
Foxo3 +/+
Foxo3 +/+
Foxo3 Kca

© 2009 Nature America, Inc. All rights reserved.
a b
CD11c + cells/spleen (× 10 5 )
40
20
0
Foxo3 +/+
***
Foxo3 Kca
B7-1 hi (%)
B7-1 hi
B7-1 B7-2 MHC II
MHC I CD40
20 ** 20 ** 60
20
20
15
15
40
15
15
10
5
10
5
20
10
5
10
5
0
0
0
0
0
B7-2 hi (%)
B7-2 hi
Figure 3 Foxo3 Kca mice have more DCs and greater
activation of DCs. (a) Total DCs among splenocytes
obtained from Foxo3 Kca mice and their wild-type
littermates (n ¼ 10 mice per group) and stained for
CD11c. (b) Flow cytometry of CD11c + DCs obtained
from naive Foxo3 Kca mice and their wild-type littermates
(n ¼ 5 mice per group) and stained with antibodies
specific for various markers (below plots). Bottom row,
accumulated data. MHCII, MHC class II; MHCI, MHC
class I. (c) Absolute numbers of CD11c + CD11b + CD8 – ,
CD11c + CD11b – CD8 + and CD11c + B220 + DCs in spleens
we measured the capacity of DCs from uninfected mice to stimulate
P14 T cells in the presence of gp33 peptide. We noted slightly greater
accumulation and viability of T cells from cultures containing
Foxo3 Kca DCs than from those containing wild-type DCs (Fig. 4c).
To determine if the enhanced function of Foxo3-deficient DCs was
cell autonomous, we generated DCs in vitro. Forthis,wecultured
bone marrow cells for 8 d with granulocyte-macrophage colonystimulating
factor. The number and proportion of the resulting
CD11c + cells (called ‘BMDCs’ here) were unaffected by Foxo3 status,
and none of the BMDCs had higher expression of B7-1, B7-2, CD40,
MHC class I or MHC class II (Supplementary Fig. 3a online).
Consistent with published reports of cells stimulated with granulocyte-macrophage
colony-stimulating factor, wild-type BMDCs had a
uniformly cytoplasmic Foxo3 localization 29 (Supplementary Fig. 3b).
Next we used BMDCs to stimulate OT-II and P14 T cells at various
ratios of T cells to DCs. Foxo3 Kca BMDCs induced accumulation of
OT-II and P14 T cells more effectively than did wild-type BMDCs
(Fig. 5a). Moreover, the Foxo3-deficient BMDCs also more effectively
sustained T cell viability, regardless of cell division (Fig. 5b,c). To
determine whether the differences in viability correlated with changes
in Bcl-2 family members, we analyzed expression of the prosurvival
factors Bcl-2 and Bcl-x L. Total and naive CD44 lo T cells had higher
expression of both when stimulated with Foxo3-deficient BMDCs
(Fig. 5d).
To assess the capacity of BMDCs to enhance T cell survival in vivo,
we transferred CFSE-labeled OT-II T cells into naive CD45.1 congenic
hosts. Then, 1 d later, we transferred wild-type or Foxo3-deficient
BMDCs, with or without preloaded OVA peptide (amino acids
323–339; OVA(323–339), into the footpads of these mice. After
2 additional days, Foxo3 Kca BMDCs induced a greater accumulation
of OT-II T cells in the draining lymph nodes but a similar number of
OT-II T cell divisions, compared with wild-type DCs (Fig. 5e).
Notably, Foxo3 Kca DCs facilitated greater OT-II T cell survival, even
MHC II hi (%)
d
MHC I hi (%)
CD40 hi (%)
CD11c + CD11b + CD11c + B220 + CD11c + CD8 +
B7-1
B7-2
MHCII hi
MHCI hi
Foxo3 Kca
Foxo3 +/+
CD40 hi
Foxo3 Kca
Foxo3 +/+
c
CD11c + CD11b +
cells/spleen (× 10 5 )
e
IL-6 (pg/ml)
15
10
5
0
Foxo3 +/+
60
40
20
0
*
Foxo3 Kca
CD11c + B220 +
cells/spleen (× 10 5 )
CD11c + CD8 +
cells/spleen (× 10 5 )
in the absence of OVA(323–339). These data collectively establish that
Foxo3-deficient DCs produce greater antigen-induced T cell population
expansion by enhancing survival.
Enhanced IL-6 secretion by Foxo3-deficient DCs
To determine if the enhanced T cell survival induced by Foxo3deficient
DCs was due to altered cytokine secretion, we infected
wild-type and Foxo3 Kca mice with LCMV and measured cytokine
concentrations in blood plasma at various times after infection. The
concentrations of IL-12, IL-17, CCL2 (MCP-1), IL-1b, IL-5, IL-10
were similar in wild-type and Foxo3 Kca mice (Supplementary Fig. 4a
online), but the concentrations of IL-6 and TNF were higher in Foxo3deficient
mice (Fig. 6a and Supplementary Fig. 4a). To determine if
DCs were the source of this abundant IL-6 and TNF, we collected
splenic DCs from mice on day 3 after infection and cultured the cells
for 24 h. Supernatants of Foxo3 Kca DC cultures had significantly higher
concentrations of IL-6, TNF, CCL2 (MCP-1) and IFN-g (Fig. 6b and
Supplementary Fig. 4b), whereas we detected no difference in the
concentration of IL-10 or IL-12. We also found higher expression of
IL-6 mRNA in Foxo3 Kca DCs (Fig. 6b).
To determine whether cells other than DCs could be responsible for
the greater IL-6 production, we collected T cells, B cells and macrophages
from mice 3 d after LCMV infection and cultured the cells for
24 h. Neither T cells nor B cells had detectable production of IL-6
(data not shown). Macrophages did produce IL-6, although there was
no difference related to Foxo3 status (data not shown). However, given
the finding of more macrophages in Foxo3-deficient mice, these cells
may have contributed to the excess IL-6 production in vivo.
We next determined whether BMDCs also produced excess IL-6 in
the absence of Foxo3. We detected significantly more IL-6 in cultures
containing OT-II CD4 + T cells or OT-I CD8 + T cells and Foxo3 Kca
BMDCs than in those containing T cells and wild-type DCs (Fig. 6c).
Confirming the DC-specific nature of this abundant IL-6, RT-PCR
TNF (pg/ml)
25
0
6
3
0
IL-6
**
*** 50 * *
60
Foxo3 Kca
Foxo3 +/+
Foxo3 +/+
Foxo3 Kca
Foxo3 Kca
Foxo3 +/+
ARTICLES
of Foxo3 Kca mice and their wild-type littermates (n ¼ 5 mice per group). (d) Flow cytometry of the expression of B7-1 and B7-2 on DC subsets (above plots)
from Foxo3 Kca mice and their wild-type littermates. (e) Cytokine bead array analysis of the concentration of IL-6, TNF and CCL2 (MCP-1) in supernatants of
splenic DCs (CD11c + ) purified from wild-type or Foxo3 Kca mice and cultured for 24 h (4 10 5 cells). In a,c, each symbol represents an individual mouse;
small horizontal lines indicate the mean. *, P o 0.01; **, P o 0.005; and ***, P o 0.001 (unpaired two-tailed Student’s t-test). Data are representative of
three independent experiments (a,b) or two independent experiments with at least three mice per group (c–e; error bars (b,e), s.e.m.).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 507
MCP-1 (pg/ml)
30
0
4
2
0
Foxo3 +/+
**
Foxo3 Kca
Foxo3 +/+
Foxo3 Kca

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
b
Foxo3 +/+
Foxo3
Annexin V
Kca
Foxo3
OTI, no peptide OTI, OVAp
Kca
Foxo3 +/+
B7-1 B7-2 MHCII MHCI
P14, no peptide P14, gp33p
CFSE
9
23 5 14
24
7-AAD
44
9
Foxo3 Kca
Foxo3 +/+
showed that Foxo3 Kca BMDCs had twofold more IL-6 mRNA than did
wild-type BMDCs; IL-6 mRNA was at least tenfold greater in BMDCs
than in CD4 + T cells, and IL-6 mRNA was undetectable in CD8 +
T cells (Fig. 6d).
In combination with transforming growth factor-b, IL-6induces
the differentiation of IL-17-producing T cells 30 . Given the results
presented above, we analyzed the balance between IL-17-producing
T cells and Foxp3 + regulatory T cells in Foxo3-deficient mice.
OT-II
P14
Cells
24
c
Foxo3 +/+
Foxo3
Uninfected mice
P14, no peptide P14, gp33p
Kca
Foxo3 +/+
Foxo3
Annexin V
Kca
CFSE
a b
d
Cells
30:1 90:1 270:1
200 150 130
1,200 800 400
CFSE
Bcl-2
OT-II
P14
OT-II, CD44 lo
P14, CD44 lo
Bcl-x L
P14
Foxo3 Kca
Foxo3 +/+
2.5
49
6 60
7-AAD
Foxo3 +/+
Foxo3 Kca
7-AAD
OT-II OT-II, CD44 lo
P14, CD44 lo
CFSE
The proportion of Foxp3 + and IL-17 + CD4 + T cells in naive
LCMV-infected Foxo3 Kca mice was similar to that of their wild-type
littermates (Supplementary Fig. 5 online). We did not detect
IL-17-producing cells in the CD8 + lineage (data not shown).
To evaluate whether the excess IL-6 produced by Foxo3-deficient
DCs was required for their ability to induce enhanced T cell survival,
we cultured OT-II and P14 T cells with BMDCs in the presence of an
IL-6-specific blocking antibody or an immunoglobulin G1 (IgG1)
isotype-matched control antibody. IL-6 blockade diminished the
expansion of T cell populations cultured with Foxo3 Kca BMDCs to
an amount resembling that of T cells cultured with wild-type DCs
(Fig. 6e). Counting of dead cells stained with 7-AAD demonstrated
OT-II
P14
OVAp No peptide gp33p No peptide
24 48
16 27
Foxo3 Kca
Foxo3 +/+
Figure 4 Greater immunogenicity of LCMV-infected Foxo3-deficient DCs.
(a) Flow cytometry of the expression of activation markers on wild-type and
Foxo3 Kca DCs on day 3 after LCMV infection. (b) Proliferation and viability
of CFSE-labeled wild-type P14 CD8 + T cells stimulated by DCs isolated from
LCMV-infected Foxo3 Kca mice or their wild-type littermates (day 3) in the
presence or absence of gp33 peptide (gp33p; left), or of CFSE-labeled
wild-type OT-I CD8 + T cells stimulated in the presence or absence of
OVA (323–339) (OVAp; bystander proliferation; right), assessed by CFSE
dilution (proliferation; top row) or staining with annexin V and 7-AAD after
3 d of culture (viability; middle and bottom rows). Numbers in dot plots
indicate percent live CD8 + Tcells.(c) Proliferation and viability of T cells
from P14 mice stimulated by DCs purified from the spleens of uninfected
mice, cultured and analyzed as described in b. Data are representative of
two independent experiments with at least three mice per group (a,c) or
three independent experiments (b).
65 87
23 68
e CD45.1 mice
0 1 2 3
CFSE-labeled
CD45.2
OT-II T cells
CFSE
OVAppulsed
BMDC
footpad
BMDC BMDC + OVAp
CD4 + CD45.2 + cells (× 10 5 )
c
Foxo3 +/+
Foxo3 Kca
5
4
3
2
1
0
Annexin V
Analysis of
T cells in
draining LN
***
BMDC
32
78
7-AAD
OT-II
BMDC +
OVAp
21
46
P14
Foxo3 Kca
Foxo3 +/+
Figure 5 Enhanced T cell responses induced by Foxo3 Kca BMDCs. (a) CFSE-dilution analysis of the accumulation of wild-type OT-II CD4 + T cells or wild-type
P14 CD8 + T cells stimulated for 3 d at various ratios (above plots) with wild-type or Foxo3 Kca BMDCs in the presence of OVA(323–339) (for OT-II T cells)
or gp33 peptide (for P14 T cells). Numbers in plots indicate number of accumulated CSFE + cells. (b) Death of OT-II CD4 + T cells or P14 CD8 + Tcells
(1 10 5 ) cultured with wild-type or Foxo3 Kca BMDCs (3.3 10 2 ) in the presence or absence of the appropriate peptide, assessed by 7-AAD staining.
Numbers adjacent to outlined areas indicate percent 7-AAD + Tcells.(c) Viability of OT-II CD4 + TcellsorP14CD8 + T cells cultured for 3 d with BMDCs in
the presence or absence of the appropriate peptide and stained with annexin V and 7-AAD. Numbers in outlined areas indicate percent viable CD4 + or CD8 +
Tcells.(d) Flow cytometry of the expression of Bcl-2 and Bcl-x L in total T cells or gated CD44 lo T cells among OT-II or P14 T cells cultured as described in
b. Dashed lines, isotype-matched control antibody. (e) Proliferation (left) and number (right) of CD4 + T cells purified from wild-type CD45.2 OT-II mice,
labeled with CFSE and injected intravenously into CD45.1 recipient mice, followed by subcutaneous injection of unpulsed or OVA(323–339)-pulsed
(+ OVAp) BMDCs from Foxo3 Kca mice or their wild-type littermates (n ¼ 4 mice per group) into the footpads of the same recipient mice and analysis 3 d
later. Left, CFSE-dilution analysis; right, CD45.1 cells in draining lymph nodes (LN). Data are representative of at least six independent experiments (a,b) or
two (c,d) orthree(e) independent experiments (error bars (e), s.e.m.).
508 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a b c Foxo3 d
e
+/+
Foxo3 +/+
Foxo3 +/+
Serum IL-6 (pg/ml)
250
200
150
100
50
***
Foxo3 Kca
Foxo3 +/+
0 2 4 6 8 10
IL-6 (pg/ml)
100
75
50
25
0
0 3
that this reversion was due to lower T cell survival (Fig. 6f). In
addition, we determined whether exogenous IL-6 was sufficient to
enhance the survival of T cells cultured with wild-type BMDCs.
Indeed, we noted a direct correlation between the amount of IL-6
added and the proportion of live T cells in each culture 31 (Fig. 6g).
To directly ascertain if IL-6 was responsible for the greater expansion
of T cell populations in response to LCMV, we treated mice on
days –1 and +4 (relative to LCMV infection at day 0) with 100 mg of
blocking antibody to IL-6 receptor a-chain. Blockade of this receptor
resulted in LCMV-specific T cell responses of substantially smaller
magnitude in Foxo3-deficient mice but did not affect the number of
LCMV-specific T cells in wild-type mice (Fig. 6h). These experiments
collectively suggest that wild-type DCs are restrained from abundant
IL-6 production by Foxo3 and thus that the factors that affect the
localization and transcriptional activity of Foxo3 will influence the
magnitude of a T cell–mediated immune response.
Foxo3 acts ‘downstream’ of CTLA-4-induced signals
Stimulation of B7 receptors by CTLA-4 promotes nuclear localization
of Foxo3 and activation of DCs 32 . As cell surface CTLA-4
expression is induced on activated T cells, we analyzed whether
signaling through the B7 receptor inhibits IL-6 production in a
Foxo3-dependent way. We stimulated splenic DCs with loxoribine,
a Toll-like receptor 7 agonist, in the presence of varying amounts
of a fusion of CTLA-4 and immunoglobulin (CTLA-4–Ig).
The production of IL-6 and TNF stimulated by loxoribine was
strongly inhibited by CTLA-4–Ig in wild-type DCs but not in
***
Il6 mRNA
(relative expression)
f
7-AAD
0 3
Time (d) Time (d) Time (d)
Figure 6 IL-6 synthesis by Foxo3-deficient
DCs is involved in enhanced T cell survival.
(a) Cytokine bead array analysis of IL-6 in
plasma from wild-type and Foxo3 Kca mice
(n ¼ 4 mice per group) at various times after
LCMV infection. (b) Cytokine bead array
analysis (left) of IL-6 secreted by DCs
(CD11c + ) purified from wild-type and
Foxo3 Kca mice without infection (0 d) or 3 d
after LCMV infection (n ¼ 4 mice per group)
and then cultured for 24 h (2 10 5 cells).
Right, RT-PCR analysis of Il6 mRNA
2.0
1.5
1.0
0.5
0.0
Foxo3 +/+
Foxo3 Kca
Foxo3 Kca
CFSE
**
IL-6 (pg/ml)
200
150
100
50
Isotype Anti-IL-6
64 66
42 61
0
Foxo3 Kca
**
**
OT-II P14
Il6 mRNA
(relative expression)
g
Live cells (%)
60
40
20
0
2
1
0
**
Foxo3 Kca
DC OT-II P14
0 0.4 4 40 α-IL-6
rIL-6 (ng/ml)
h
T cells × 10 6
OT-II
P14
CFSE
Isotype
gp33-specific
CD8
40
+ T cells
Anti-IL-6
0
0
IgG1 α-IL-6R IgG1 α-IL-6R
Foxo3-deficient DCs (Fig. 7a). Although treatment with CTLA-4–Ig
resulted in a slightly greater proportion of dead cells, its toxic effects
were equivalent in wild-type and Foxo3 Kca splenic DCs showed the
same toxicity (Fig. 7a).
To confirm the involvement of Foxo3 in the inhibition of cytokine
production induced by CTLA-4–Ig, we analyzed the intracellular
localization of Foxo3 after stimulation with loxoribine or CTLA-4–Ig.
Foxo3 remained in the cytosol of unstimulated and loxoribine-treated
BMDCs; however, CTLA-4–Ig stimulation strongly promoted the
nuclear accumulation of Foxo3 (ref. 32; Fig. 7b). Notably, signaling
through B7 was dominant over Toll-like receptor 7 signals, as DCs
treated with both loxoribine and CTLA-4–Ig showed Foxo3 nuclear
localization. These experiments collectively show that CTLA-4–Ig
treatment profoundly suppressed Toll-like receptor–induced production
of IL-6 and TNF in wild-type DCs by promoting the nuclear
accumulation of Foxo3.
If CTLA-4 signals DCs through B7-1 and B7-2, then blocking
CTLA-4 should allow wild-type DCs, like Foxo3-deficient DCs,
to enhance T cell survival. To investigate this issue, we cultured
TCR-transgenic T cells in the presence of wild-type or Foxo3 Kca DCs
and specific peptides with or without the addition of blocking antibody
to CTLA-4. As expected, cultures with Foxo3 Kca BMDCs
included more viable T cells than did cultures with wild-type
BMDCs (Fig. 7c). However, the addition of anti-CTLA-4 essentially
eliminated the difference in T cell accumulation and viability in these
cultures. The results of these experiments are consistent with the idea
that CTLA-4 is involved in promoting the nuclear localization of
30
20
10
ARTICLES
gp61-specific
CD4 + T cells
2
1
Foxo3 Kca
Foxo3 +/+
Foxo3 Kca
Foxo3 +/+
expression, presented relative to the expression in uninfected wild-type cells, set as 1. (c) Enzyme-linked immunosorbent assay of IL-6 in supernatants
of wild-type OT-II or P14 T cells cultured for 3 d in the presence of wild-type or Foxo3 Kca BMDCs loaded with OVA(323–339) or gp33 peptide. (d) RT-PCR
analysis of Il6 mRNA in BMDCs (DC), OT-II T cells and P14 T cells purified after 2 d of culture as described in c, presented relative to Gapdh mRNA
(encoding glyceraldehyde phosphate dehydrogenase). (e,f) Accumulation (e) anddeath(f) of wild-type OT-II or P14 T cells activated by BMDCs generated
from wild-type or Foxo3 Kca mice pulsed with OVA(323–339) (for OT-II CD4 + T cells) or gp33 peptide (for P14 CD8 + T cells), in the presence of an IgG1
isotype-matched control antibody or IL6-specific blocking antibody (10 mg/ml), assessed after 3 d of culture by CFSE dilution (accumulation) or 7-AAD
staining (death). Numbers adjacent to outlined areas (f) indicate percent dead 7-AAD + CFSE + cells. (g) Death of P14 T cells stimulated by wild-type BMDCs
in the presence of increasing amounts of recombinant IL-6 (rIL-6) or IL-6-specific blocking antibody (a-IL-6), assessed after 3 d of culture by 7-AAD
staining. (h) LCMV-specific T cell responses of Foxo3 Kca mice and their wild-type littermates (n ¼ 4 mice per group) treated on days –1 and +4 with 100 mg
of antibody to IL-6 receptor a-chain (a-IL-6R) or isotype-matched control antibody (IgG1) and infected on day 0 with LCMV, Armstrong strain, and then
analyzed on day +8 by counting of P14 CD8 + T cells and OT-II CD4 + T cells producing IFN-g after restimulation with OVA(323–339) (for OT-II CD4 + Tcells)
or gp33 peptide (for P14 CD8 + T cells). Data are representative of three (a–f) ortwo(g,h) independent experiments (error bars (a–d,g,h), s.e.m.).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 509

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
IL-6 (pg/ml)
0
Loxo
CTLA-4–Ig
b
Media
Loxo
600
400
200
**
– + + +
0 0 20 40
TNF (pg/ml)
Control Ig
600
400
200
0
Loxo – + + +
0
Loxo – + + +
CTLA-4–Ig 0 0 20 40 CTLA-4–Ig 0 0 20 40
CTLA-4–Ig
Foxo3 and the resulting inhibition of inflammatory cytokines
produced by DCs. Published experiments have shown that antibody
specific for CTLA-4 can enhance an immune response in vivo 33,34 ,
although, to our knowledge, this has not been demonstrated before in
an immune response to LCMV. Nonetheless, we sought to determine
whether we could detect Foxo3-dependent enhanced T cell population
expansion. We inoculated mice with 100 mg of a CTLA-4-specific
antibody and, 4 h later, infected the mice with LCMV. We further
inoculated mice with a CTLA-4-specific antibody every other day and
assessed the results on day 8. We did not find enhancement of T cell
accumulation in either wild-type or Foxo3-deficient mice (Supplementary
Fig. 6 online).
DISCUSSION
In this report, we have established an essential function for Foxo3 in
modulating the magnitude of antigen-specific T cell immune responses.
Foxo3-deficient mice showed enhanced T cell proliferation in response
to viral infection, an enhancement that was not T cell intrinsic but
instead depended on the augmented capacity of DCs to sustain T cell
viability. Despite the many potential targets of Foxo3, including
regulators of cell cycle, apoptosis and reactive oxygen detoxification,
Foxo3 loss-of-function mutation alone had no effect on the steady-state
homeostasis, survival or proliferation of T cells. That finding is
consistent with studies showing that Foxo1 is essential for normal
T cell homeostasis and self tolerance 16 (data not shown). Instead, we
found a critical DC-intrinsic function for Foxo3 in the control of T cell
responses. Foxo3 acted ‘downstream’ of CTLA-4-induced signals to
constrainIL-6productionbyDCs.The natural conditions that would
override the CTLA-4-mediated nuclear localization of Foxo3 are not
known; however, a new immunostimulatory cancer treatment regimen
that involves blocking CTLA-4 may act in part by this mechanism 35 .
**
7-AAD + cells (%)
40
30
20
10
Foxo3 Kca
Foxo3 +/+
c
P14 OT-II
Foxo3
Anti-CTLA-4
Anti-CTLA-4
900 900
150
150
Kca
Foxo3 +/+
Foxo3 37
+/+ 66
Foxo3 Kca
Annexin V
CFSE
60 67
7-AAD
Our results differ substantially from a published characterization of
Foxo3 Trap mice 20 . Foxo3 Trap mutant mice were derived from an
independent insertional mutant embryonic stem line obtained from
BayGenomics and were backcrossed to the 129 strain for three
generations. Foxo3 Trap mice show spontaneous T cell activation,
lymphoproliferation and organ infiltration associated with constitutive
activation of NF-kB. In contrast, we found none of those
characteristics in Foxo3 –/– or Foxo3 Kca mice. T cells isolated from
these two strains of mice seemed to be indistinguishable from wildtype
T cells in terms of NF-kB activation, expression of activation
markers and response to mitogenic stimulation. Our results are
consistent with published analyses of Foxo3 Kca and Foxo3 –/– mice in
that extensive histological analysis has not shown lymphocytic infiltration
of organs 23,26 .
Differences in autoimmunity are not uncommon in mixed
C57BL/6 and 129 mouse strains. The 129 mice have autoimmunitysusceptibility
loci, one of which (Sle16) induces humoral autoimmunity
in congenic C57BL/6 mice. Several studies have noted modifier
genes in 129 strains that confer enhanced autoimmune disease to 129
C57BL/6 mice 36–38 . Foxo Trap mice are apparently inbred 129 mice
without a contribution from the C57BL/6 strain, but perhaps a 129
susceptibility locus that modifies the Foxo3 deficiency is present.
IL-6 is a pleiotropic cytokine that regulates many aspects of the
immune system, including antibody production, hematopoiesis,
inflammation and, most relevant to our study, T cell survival 39 .IL-6
‘rescues’ resting T cells from apoptosis by inhibiting the downregulation
of Bcl-2 in a dose-dependent way 40 ,andIL-6increasesthesurvival
of antigen-stimulated T cells 31 . Consistent with that observation, we
noted higher expression of Bcl-2 and Bcl-x L in T cells from LCMVinfected
Foxo3 Kca mice than in those from wild-type mice. In addition,
the division rate of T cells stimulated with Foxo3 Kca DCs was not
43
61
73 75
Figure 7 Production of IL-6 and TNF by stimulated DCs is inhibited by CTLA-4–Ig stimulation in a Foxo3-dependent way. (a) Immunoassay of the
concentration of IL-6 (left) and TNF (middle) in supernatants of DCs (2 10 5 ) purified from the spleens of Foxo3 Kca mice and their wild-type littermates
and stimulated for 18 h with (+) or without (–) loxoribine (Loxo) plus increasing amounts of CTLA-4–Ig (below graphs, in mg/ml). Right, DC viability, assessed
by 7-AAD staining. (b) Immunofluorescence analysis of Foxo3 localization (green) after 18 h of stimulation of BMDCs from wild-type and Foxo3 Kca mice with
loxoribine (bottom) or without loxoribine (Media; top) in presence of control immunoglobulin or CTLA-4–Ig. Blue, DAPI staining of nuclei. Original
magnification, 63. (c) Accumulation (top) and death (below) of OT-II CD4 + T cells or P14 CD8 + Tcells(1 10 5 ) cultured for 3 d with wild-type or
Foxo3 Kca BMDCs (3.3 10 2 ) in the presence of the appropriate peptide plus anti-CTLA-4 (50 mg/ml; 9D9), assessed as CFSE dilution (accumulation) and
staining with annexin V and 7-AAD (death). Numbers in plots indicate number of accumulated cells (top) or percent live OT-II CD4 + T cells or P14 CD8 +
T cells (middle and bottom). Data are representative of at least two independent experiments (error bars (a), s.e.m.).
510 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
different from that of T cells stimulated with wild-type DCs; instead,
the effect was manifested as improved T cell survival. Moreover,
Foxo3 Kca DCs sustained T cell viability both in vitro and in vivo even
in the absence of specific peptide, which suggests that the enhanced
T cell survival induced by IL-6 is not restricted to TCR-engaged T cells.
Notably, IL-6-specific blocking antibody did not affect the viability of
wild-type T cells in vitro, nor did it result in lower T cell accumulation
in vivo. That observation indicates that in the conditions tested, the
amount of IL-6 produced by wild-type DCs was not sufficient to affect
T cell viability. It would be useful to determine whether immune
responses to other agents are sensitive to IL-6 concentration and, in
those infections, whether CTLA-4, regulatory T cells and Foxo3 affect
the magnitude and dynamics of the T cell response.
CTLA-4 acts as a counterbalance to costimulation of CD28 by
antagonizing TCR signals in activated T cells and by promoting
tolerance 41 . CTLA-4 is also expressed on regulatory T cells and is
essential for their effectiveness 42 . CTLA-4 can also modify function of
cells of the immune system by triggering ‘reverse signaling’ through B7
receptors; these signals promote the DC production of stimulatory
and/or suppressive mediators 21 . Reverse signals through B7 also
activate the immunosuppressive pathway of tryptophan catabolism
in plasmacytoid DCs involving indoleamine 2,3-dioxygenase (IDO) 43 .
Foxo3 is involved in the IDO pathway in plasmacytoid DCs, as CTLA-
4–Ig leads to nuclear accumulation of Foxo3 and expression of
superoxide dismutase 2 (ref. 32). As IDO induces T cell death through
the generation of kinurenin and tryptophan privation 43 ,wespeculated
that a potential IDO defect in Foxo3 Kca DCs could explain the
improved T cell survival; however, addition of the IDO inhibitor
1-MT did not enhance T cell survival in cultures containing wild-type
or Foxo3 Kca BMDCs (data not shown). Moreover, IDO is expressed
mainly by plasmacytoid and CD8a + DC subsets 22 ,andlowornoIDO
protein is detected in resting DCs derived from bone marrow 44 .
Presumably CTLA-4 expressed on regulatory T cells constitutes the
main B7-stimulatory ligand, but the possibility that activated T cells or
other cells contribute to B7 ligation in vivo has not been investigated.
In addition, the relative contribution of each CTLA-4-mediated
suppressive mechanism to the regulation of immune responses to
various infectious agents is not well understood.
Our experiments collectively indicate that nucleus-localized Foxo3
functions in DCs to inhibit the production of inflammatory cytokines
that would otherwise be activated in response to infection. We deduce
that the IL-6-induced T cell survival in Foxo3 Kca mice resulted from an
inability of Foxo3-deficient DCs to shut down cytokine production in
response to CTLA-4 signals. Conditions that alter Foxo3 activity, such
as stress, growth factors or nutrients, would be expected to substantially
influence the activity of DCs and perhaps macrophages, and such
conditions might be expected to alter the outcome of responses to
infectious agents.
METHODS
Mice. C57BL/6, C57BL/6-CD45.1, P14, OT-I and OT-II mice were maintained
in specific pathogen–free conditions at the University of California, San Diego.
C57BL/6 mice lacking Foxo3 were generated from embryonic stem cell clones
from the OmniBank embryonic stem cell library of randomly targeted cell lines
(Lexicon Genetics) 23 . These mice (Foxo3 Gt(VICTR20)1Kca ; called ‘Foxo3 Kca ’ here)
were backcrossed to the C57BL/6J strain for 12 generations and were then
intercrossed to generate congenic C57BL/6 Foxo3 Kca mice. ‘Foxo L/L ’mice,with
conditional targeting of Foxo3, were produced in FVB embryonic stem cells, and
the loxP-flanked allele was deleted by breeding with male EIIa-Cre–transgenic
mice 26 . Offspring with complete excision were bred to exclude EIIa-Cre; these
mice are called ‘Foxo3 –/– ’ here 26 . TCR-transgenic mice with a Foxo3-null
genotype were produced by breeding of C57BL/6 Foxo3 Kca mice with either
ARTICLES
P14 or OT-II mice. All procedures were approved by the Institutional Animal
Care and Use Committee of the University of California, San Diego.
Virus infection and analysis. Foxo3 Kca mice and their wild-type littermates
6–12 weeks of age were infected by intraperitoneal injection of
2 10 5 plaque-forming units of LCMV, Armstrong strain, in 0.2 ml PBS.
Alternatively, mice were infected with 1 10 5 plaque-forming units of vesicular
stomatitis virus 45 expressing OVA (provided by L. Lefrançois). At various times,
spleens were collected from LCMV-infected mice and uninfected control mice
and splenocytes were stimulated with gp33 peptide, an MHC class I–restricted
LCMV epitope (gp33-41), or gp61 peptide, an MHC class II–restricted LCMV
epitope (gp61-80; Genemed Synthesis) and brefeldin A (1 mg/ml; Golgistop; BD
PharMingen). After 5 h of stimulation, cells were stained for intracellular IFN-g
(XMG1.2; eBioscience). For transfer experiments, CD45.2 congenic wild-type
P14 or Foxo3 Kca P14 T cells (2 10 4 ) were injected intravenously into CD45.1
wild-type hosts, which were infected with LCMV 24 h later. Conversely, CD45.1
congenic wild-type P14 T cells were transferred into congenic CD45.2 wild-type
littermate or Foxo3 Kca mice. P14 T cells in the spleen were counted at day 8 after
LCMV infection. Infection with vesicular stomatitis virus was done in an
identical way except that T cells from CD45.2 OT-I mice were transferred. For
bone marrow–chimera experiments, congenic CD45.1 recipient mice were
lethally irradiated (1,200 rads) and were injected intravenously with 2 10 6
bone marrow cells from Foxo3 Kca or wild-type littermate donor mice (both
CD45.2). After 2 months, the reconstitution of peripheral blood lymphocytes
was over 90%. These radiation chimeras were then infected with LCMV and the
magnitude of the LCMV response was assessed at day 8 after infection. For
antibody blockade of IL-6 receptor a-chain, wild-type or Foxo3-deficient mice
were inoculated intraperitoneally with 100 mg of antibody to IL-6 receptor
a-chain (purified antibody to mouse and rat CD126; D7715A7; BioLegend) on
day –1 and day +4 and were infected with LCMV, Armstrong strain, on day 0.
Cell isolation and purification. Bone marrow cells were isolated by flushing of
the femur and tibia with RPMI medium (containing 10% (vol/vol) FBS, 2 mM
L-glutamine, 100 U/ml of penicillin, 100 mg/ml of streptomycin and 50 mM
b-mercaptoethanol). Spleens were removed and were incubated for 20 min at
37 1C with collagenase D (1 mg/ml; Roche), and splenocytes were collected by
homogenization through a 100-mm tissue strainer. Cells were resuspended in
Tris–ammonium chloride buffer for lysis of red blood cells. For purification
of splenic DCs, cells were first incubated for 15 min with supernatants of
2.4G2 hybridoma cultures (anti–mouse FcgRII-III), then were incubated for
20 min with anti–mouse ‘pan-DC’ microbeads (Miltenyi Biotec), followed
by positive selection. The positive fraction was typically over 95% CD11c + .
For T cell purification, splenocytes from P14 or OT-II mice were incubated
with a mixture of biotinylated anti-B220 (RA3-6B2), anti-CD19, anti-Gr-1
(RB6-8C5), anti–MHC class II, anti-DX5 and anti-CD11b (M1/70; all from
eBioscience) and were negatively selected to over 95% purity with strepavidincoupled
microbeads (Miltenyi Biotec).
BMDC cultures. Bone marrow cells were isolated and were cultured at a
density of 2 10 6 cells/ml in RPMI medium containing recombinant mouse
granulocyte-macrophage colony-stimulating factor (20 ng/ml; Peprotech). On
days 2, 4 and 6, half the culture was removed and centrifuged and the cell pellet
was resuspended in 10 ml fresh media containing recombinant mouse
granulocyte-macrophage colony-stimulating factor (20 ng/ml). Cultures were
collected on day 8 and CD11c + cells were selected with microbeads (Miltenyi
Biotec). The final population was 98% CD11c + CD11b + B220 – DCs.
Flow cytometry. Cells were incubated for 15 min with anti–mouse FcgRII-III
and then were stained with primary antibodies for 20 min on ice. The
following mouse monoclonal antibodies were used (all from eBioscience):
allophycocyanin- or phycoerythrin-conjugated anti-CD11c (N418); fluorescein
isothiocyanate– or phycoerythrin-conjugated anti-B220 (RA3-6B2);
phycoerythrin-indotricarbocyanine– or peridin chlorophyll protein–conjugated
anti-CD11b (M1/70); allophycocyanin-conjugated anti-CD8a (53-6.7); fluorescein
isothiocyanate–conjugated anti–MHC class I and anti-Gr-1 (RB6-8C5);
peridin chlorophyll protein–conjugated anti-CD3; and phycoerythrinconjugated
anti-CD86, anti-CD80, anti–MHC class II, anti-Ly6C, anti-CD19,
anti-CD3 and anti-NK1.1.
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 511

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
T cell proliferation and death assays. For in vitro experiments, DCs were
purified from wild-type or Foxo3 Kca mice on day 3 after LCMV infection. DCs
were also purified from uninfected mice or were derived from the bone marrow
of wild-type or Foxo3 Kca mice. These DCs were used to stimulated P14 CD8 +
T cells, OT-II CD4 + T cells or OT-I CD8 + T cells (1 10 5 cells per well) labeled
with 1 mM CFSE (carboxyfluorescein diacetate succinimidyl ester; Molecular
Probes) in the presence of gp33 peptide (0.1 mg/ml) or OVA(323–339);
(1 mg/ml). Proliferation was analyzed by flow cytometry by measurement of
CFSE dilution, and T cell death was measured by staining with annexin V and
7-AAD. For in vivo transfer experiments, CFSE-labeled CD4 + T cells purified
from OT-II mice were injected retro-orbitally into congenic CD45.1 C57BL/6
mice. Wild-type or Foxo3 Kca BMDCs were pulsed with OVA(323–339) and were
injected intradermally after 24 h. Then, 3 d later, draining lymph nodes
were collected, were stained with allophycocyanin-conjugated anti-CD4
(GK1.5) and phycoerythrin-conjugated anti-CD45.1 (A20; both from
eBioscience) and were analyzed by flow cytometry.
In vitro stimulation of DCs and quantification of cytokine production.
Sorted DCs (2 10 5 ) were cultured in 96-well round-bottomed culture
plates (Nunc) and were stimulated with medium alone or 0.03 mM
loxoribine (InvivoGen) alone or in the presence of a recombinant mouse
fusion protein consisting of the ectodomain of CTLA-4 linked to the
Fc portion of IgG2a (CTLA-4–Ig) at a concentration of 20 or 40 mg/ml
(R&D Systems). After culture for 18 h, supernatants were collected and
cytokine concentrations were determined by immunoassay. Enzyme-linked
immunosorbent assay kits were used for the detection of TNF or IL-6
(Ready-Set-Go; eBioscience) unless otherwise specified. Cytokines were also
measured by cytokine bead array (BD Biosciences) or Luminex technology
(Invitrogen).
Fluorescence microscopy. DCs derived from bone marrow of wild-type or
Foxo3 Kca mice were cultivated for 12 h on sterile poly-L-lysine-coated coverslips
(BioCoat; BD). Cells were fixed in 4% (vol/vol) formaldehyde, then were made
permeable with 0.02% (vol/vol) Triton X-100. Coverslips were blocked for 2 h
in PBS containing 5% (wt/vol) BSA and anti–mouse FcgRII-III. Anti-Foxo3
(FKHRL1; Cell Signaling) was added for 1 h, followed by a secondary antibody
conjugated to fluorescein isothiocyanate (goat anti–rabbit IgG; 65-6111;
Zymex). For staining of nuclei, DAPI (4¢-6-diamidino-2-phenylindole) was
added at a concentration of 0.04 mg/ml. All cells were visualized with a
microscope (Axiovert 200M; Carl Zeiss MicroImaging) with a 63 objective.
Images were captured with an Axiocam monochrome digital camera and were
analyzed with Axiovision software (Carl Zeiss MicroImaging).
Semiquantitative RT –PCR. Total RNA was extracted from DCs with Trizol
reagent according to the manufacturer’s instructions (Invitrogen), then cDNA
was synthesized with SuperScript First-Strand synthesis System for RT-PCR
(Invitrogen). Il6 and the gene encoding cyclophilin A (Ppia) wereamplifiedby
PCR with the following primers: Il6 forward, 5¢-ACCTGGAGTACATGAAGAA
CAACTT-3¢ and reverse, 5¢-GGAAGCACTCACCTCTTGGT-3¢; Ppia forward,
5¢-CACCGTGTTCTTCGACATC-3¢ and reverse, 5¢-ATTCTGTGAAAGGAG
GAACC-3¢.
Immunoblot analysis. Equal amounts of protein from whole-cell extracts were
resolved by 4–12% SDS-PAGE (Invitrogen) and were transferred to a polyvinylidene
difluoride membrane (Millipore) by semidry transfer (Bio-Rad).
Blots were blocked and incubated with primary antibody overnight at 4 1C,
followed by 2 h of incubation at 25 1C with the appropriate horseradish
peroxidase–conjugated secondary antibody. Rabbit polyclonal anti-Foxo3a was
provided by A. Brunet (Stanford University). Antibodies specific for IkB
proteins were from Santa Cruz Biotechnology.
Accession codes. UCSD-Nature Signaling Gateway (http://www.signalinggateway.org):
A000945 and A000706.
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank J.P. Allison (Sloan-Kettering Memorial Hospital) and J.A. Bluestone
(University of California at San Diego) for CTLA-4-specific blocking antibodies;
A. Brunet (Stanford University) for Foxo3-specific antibody; L. Lefrançois
(University of Connecticut Health Center) for OVA-expressing vesicular
stomatitis virus; L. Mack and E. Zuniga for assistance with virus titers; and
M. Niwa for microscope facility use. Supported by the American Cancer Society
(R.A.D.), the Robert A. and Renee E. Belfer Institute for Innovative Cancer
Research (R.A.D.), the US National Cancer Institute (R.A.D.), the Division of
Biological Sciences of the University of California, San Diego (S.M.H.) and the
Fondation pour la Recherche Médicale (A.S.D.).
AUTHOR CONTRIBUTIONS
D.R.B. initiated the project and did lymphocyte characterization and LCMV
infection under the supervision of S.M.H.; A.S.D. designed and did the
remaining experiments in collaboration with D.R.B., I.L.C., Y.M.K. and S.M.H.;
A.B. provided expertise in fluorescence microscopy analysis; R.A.D. and D.H.C.
produced Foxo3 –/– mice and provided intellectual input on the data; K.C.A.
provided Foxo3 Kca mice; S.M.H. initiated the project with input from R.A.D.
and K.C.A. and supervised the experiments; and A.S.D. and S.M.H. wrote the
manuscript with editorial and intellectual contributions from the other authors.
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Calnan, D.R. & Brunet, A. The FoxO code. Oncogene 27, 2276–2288 (2008).
2. Galili, N. et al. Fusion of a fork head domain gene to PAX3 in the solid tumour alveolar
rhabdomyosarcoma. Nat. Genet. 5, 230–235 (1993).
3. Hillion, J., Le Coniat, M., Jonveaux, P., Berger, R. & Bernard, O.A. AF6q21, a novel
partner of the MLL gene in t(6;11)(q21;q23), defines a forkhead transcriptional factor
subfamily. Blood 90, 3714–3719 (1997).
4. Corral, J. et al. Acute leukemias of different lineages have similar MLL gene fusions
encoding related chimeric proteins resulting from chromosomal translocation. Proc.
Natl. Acad. Sci. USA 90, 8538–8542 (1993).
5. Jacobs, F.M. et al. FoxO6, a novel member of the FoxO class of transcription factors
with distinct shuttling dynamics. J. Biol. Chem. 278, 35959–35967 (2003).
6. Barthel, A., Schmoll, D. & Unterman, T.G. FoxO proteins in insulin action and
metabolism. Trends Endocrinol. Metab. 16, 183–189 (2005).
7. Biggs, W.H. III, Meisenhelder, J., Hunter, T., Cavenee, W.K. & Arden, K. C. Protein
kinase B/Akt-mediated phosphorylation promotes nuclear exclusion of the winged helix
transcription factor FKHR1. Proc. Natl. Acad. Sci. USA 96, 7421–7426 (1999).
8. Brunet, A. et al. Akt promotes cell survival by phosphorylating and inhibiting a Forkhead
transcription factor. Cell 96, 857–868 (1999).
9. Kops, G.J. et al. Direct control of the Forkhead transcription factor AFX by protein
kinase B. Nature 398, 630–634 (1999).
10. van der Horst, A. & Burgering, B.M. Stressing the role of FoxO proteins in lifespan and
disease. Nat. Rev. Mol. Cell Biol. 8, 440–450 (2007).
11. Brunet, A. et al. Stress-dependent regulation of FOXO transcription factors by the
SIRT1 deacetylase. Science 303, 2011–2015 (2004).
12. Motta, M.C. et al. Mammalian SIRT1 represses forkhead transcription factors. Cell
116, 551–563 (2004).
13. Stahl, M. et al. The forkhead transcription factor FoxO regulates transcription of
p27Kip1 and Bim in response to IL-2. J. Immunol. 168, 5024–5031 (2002).
14. Dijkers, P.F. et al. Forkhead transcription factor FKHR-L1 modulates cytokine-dependent
transcriptional regulation of p27 KIP1 . Mol. Cell. Biol. 20, 9138–9148 (2000).
15. Barata, J.T. et al. Activation of PI3K is indispensable for interleukin 7-mediated
viability, proliferation, glucose use, and growth of T cell acute lymphoblastic leukemia
cells. J. Exp. Med. 200, 659–669 (2004).
16. Kerdiles, Y.M. et al. Foxo1 links homing and survival of naive T cells by regulating
L-selectin, CCR7 and interleukin 7 receptor. Nat. Immunol. 10, 176–184 (2009).
17. Peng, S.L. Foxo in the immune system. Oncogene 27, 2337–2344 (2008).
18. You, H. et al. FOXO3a-dependent regulation of Puma in response to cytokine/growth
factor withdrawal. J. Exp. Med. 203, 1657–1663 (2006).
19. Riou, C. et al. Convergence of TCR and cytokine signaling leads to FOXO3a phosphorylation
and drives the survival of CD4 + central memory T cells. J. Exp. Med. 204,
79–91 (2007).
20. Lin, L., Hron, J.D. & Peng, S.L. Regulation of NF-kB, Th activation, and autoinflammation
by the forkhead transcription factor Foxo3a. Immunity 21, 203–213 (2004).
21. Orabona, C. et al. CD28 induces immunostimulatory signals in dendritic cells via CD80
and CD86. Nat. Immunol. 5, 1134–1142 (2004).
22. Mellor, A.L. & Munn, D.H. IDO expression by dendritic cells: tolerance and tryptophan
catabolism. Nat. Rev. Immunol. 4, 762–774 (2004).
23. Hosaka, T. et al. Disruption of forkhead transcription factor (FOXO) family members in
mice reveals their functional diversification. Proc. Natl. Acad. Sci. USA 101,
2975–2980 (2004).
24. Marinkovic, D. et al. Foxo3 is required for the regulation of oxidative stress in
erythropoiesis. J. Clin. Invest. 117, 2133–2144 (2007).
25. Greer, E.L. & Brunet, A. FOXO transcription factors at the interface between longevity
and tumor suppression. Oncogene 24, 7410–7425 (2005).
26. Castrillon, D.H., Miao, L., Kollipara, R., Horner, J.W. & DePinho, R.A. Suppression of
ovarian follicle activation in mice by the transcription factor Foxo3a. Science 301,
215–218 (2003).
512 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
27. Villadangos, J.A. & Schnorrer, P. Intrinsic and cooperative antigen-presenting functions
of dendritic-cell subsets in vivo. Nat. Rev. Immunol. 7, 543–555 (2007).
28. Oldstone, M.B. Biology and pathogenesis of lymphocytic choriomeningitis virus infection.
Curr. Top. Microbiol. Immunol. 263, 83–117 (2002).
29. Rosas, M. et al. IL-5-mediated eosinophil survival requires inhibition of GSK-3 and
correlates with b-catenin relocalization. J. Leukoc. Biol. 80, 186–195 (2006).
30. Korn, T., Bettelli, E., Oukka, M. & Kuchroo, V.K. IL-17 and Th17 cells. Annu. Rev.
Immunol. 27, 485–517 (2009).
31. Rochman, I., Paul, W.E. & Ben-Sasson, S.Z. IL-6 increases primed cell expansion and
survival. J. Immunol. 174, 4761–4767 (2005).
32. Fallarino, F. et al. CTLA-4-Ig activates forkhead transcription factors and protects
dendritic cells from oxidative stress in nonobese diabetic mice. J. Exp. Med. 200,
1051–1062 (2004).
33. Leach, D.R., Krummel, M.F. & Allison, J.P. Enhancement of antitumor immunity by
CTLA-4 blockade. Science 271, 1734–1736 (1996).
34. Karandikar, N.J., Vanderlugt, C.L., Walunas, T.L., Miller, S.D. & Bluestone, J.A. CTLA-4:
a negative regulator of autoimmune disease. J. Exp. Med. 184, 783–788 (1996).
35. Zang, X. & Allison, J.P. The B7 family and cancer therapy: costimulation and
coinhibition. Clin. Cancer Res. 13, 5271–5279 (2007).
36. Botto, M. et al. Homozygous C1q deficiency causes glomerulonephritis associated with
multiple apoptotic bodies. Nat. Genet. 19, 56–59 (1998).
ARTICLES
37. Bickerstaff, M.C. et al. Serum amyloid P component controls chromatin
degradation and prevents antinuclear autoimmunity. Nat. Med. 5, 694–697
(1999).
38. Santiago-Raber, M.L. et al. Role of cyclin kinase inhibitor p21 in systemic autoimmunity.
J. Immunol. 167, 4067–4074 (2001).
39. Kishimoto, T. Interleukin-6: from basic science to medicine–40 years in immunology.
Annu. Rev. Immunol. 23, 1–21(2005).
40. Teague, T.K., Marrack, P., Kappler, J.W. & Vella, A.T. IL-6 rescues resting mouse T cells
from apoptosis. J. Immunol. 158, 5791–5796 (1997).
41. Keir, M.E. & Sharpe, A.H. The B7/CD28 costimulatory family in autoimmunity.
Immunol. Rev. 204, 128–143 (2005).
42. Wing, K. et al. CTLA-4 control over Foxp3 + regulatory T cell function. Science 322,
271–275 (2008).
43. Grohmann, U. et al. CTLA-4-Ig regulates tryptophan catabolism in vivo. Nat. Immunol.
3, 1097–1101 (2002).
44. Hara, T. et al. High-affinity uptake of kynurenine and nitric oxide-mediated inhibition of
indoleamine 2,3-dioxygenase in bone marrow-derived myeloid dendritic cells. Immunol.
Lett. 116, 95–102 (2008).
45. van Stipdonk, M.J., Lemmens, E.E. & Schoenberger, S.P. Naive CTLs require a single
brief period of antigenic stimulation for clonal expansion and differentiation. Nat.
Immunol. 2, 423–429 (2001).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 513

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
C-C chemokine receptor 6–regulated entry of TH-17
cells into the CNS through the choroid plexus is
required for the initiation of EAE
Andrea Reboldi 1 , Caroline Coisne 2 , Dirk Baumjohann 1 , Federica Benvenuto 3,4 , Denise Bottinelli 1 , Sergio Lira 5 ,
Antonio Uccelli 3,4,6 , Antonio Lanzavecchia 1 , Britta Engelhardt 2 & Federica Sallusto 1
Interleukin 17–producing T helper cells (TH-17 cells) are important in experimental autoimmune encephalomyelitis, but their
route of entry into the central nervous system (CNS) and their contribution relative to that of other effector T cells remain to
be determined. Here we found that mice lacking CCR6, a chemokine receptor characteristic of T H-17 cells, developed T H-17
responses but were highly resistant to the induction of experimental autoimmune encephalomyelitis. Disease susceptibility
was reconstituted by transfer of wild-type T cells that entered into the CNS before disease onset and triggered massive
CCR6-independent recruitment of effector T cells across activated parenchymal vessels. The CCR6 ligand CCL20 was
constitutively expressed in epithelial cells of choroid plexus in mice and humans. Our results identify distinct molecular
requirements and ports of lymphocyte entry into uninflamed versus inflamed CNS and suggest that the CCR6-CCL20
axis in the choroid plexus controls immune surveillance of the CNS.
The identification of specialized subsets of effector CD4 + T cells has
provided a paradigm for understanding immunity and immunopathology.
Interleukin 17 (IL-17)-producing T cells (T H-17 cells)
have been characterized in the mouse as a distinct lineage of
CD4 + T cells that can differentiate from uncommitted naive T cell
precursors under the aegis of the transcription factors RORgt and
RORa and the polarizing cytokines transforming growth factor-b,
IL-6 and IL-23 (ref. 1–3). IL-17 can mediate protection against
extracellular pathogens by promoting neutrophil recruitment but
has also been shown to cause immunopathology in various models
of autoimmunity 4,5 .
Experimental autoimmune encephalomyelitis (EAE) is a CD4 +
T cell–mediated disease of the central nervous system (CNS) that is
used as a model of multiple sclerosis, a devastating inflammatory
demyelinating disease of the human CNS. Several lines of evidence
indicate that T H-17 cells are involved in the onset and maintenance of
EAE 6 . Thus, mice lacking RORgt, IL-17 or IL-23 as well as mice
treated with IL-17-blocking antibodies are less susceptible to EAE than
are wild-type or untreated mice 7–10 . In addition, IL-17 + T cells have
been found in lesions in brain tissues from patients with multiple
sclerosis 11 . However, it has also been shown that T helper type 1 (T H1)
cells are present in lesions of EAE and in multiple sclerosis during the
active phase of the disease and that mice lacking T-bet, the TH1
‘master’ transcription factor, are resistant to the development of
Received 4 November 2008; accepted 10 February 2009; published online 22 March 2009; doi:10.1038/ni.1716
EAE 12 . EAE can be induced by transfer of either T H-17 or T H1
cells 13,14 and the T H-17/T H1 ratio of infiltrating cells determines
where inflammation occurs in the CNS 15,16 . Together these studies
suggest the possibility that TH-17 and TH1 cells may be involved in
pathogenesis at different times or at different sites.
In the model of active EAE, autoreactive myelin-specific effector
T cells are primed in peripheral lymph nodes and must migrate into
uninflamed CNS to initiate tissue inflammation. The molecular
determinants that control this initial step of cell migration, which is
probably the same used for constitutive immune surveillance in the
brain, remain to be determined. In contrast, the molecular requirements
for lymphocyte rolling and adhesion to activated vessels of the
inflamed blood-brain barrier have been intensively investigated in
both active and passive models of EAE 17 . The integrin a4b1 (VLA-4)
serves a key function in controlling the entry of lymphocytes into
the CNS by interacting with the adhesion molecule VCAM expressed
by inflamed endothelial cells 18 . P-selectin does not seem to be
necessary for the recruitment of inflammatory cells during active
EAE 19 but can be expressed in small amounts on resting brain
endothelium or can be rapidly induced on endothelial cells by
inflammatory stimuli 20,21 . Chemokine receptors are required for the
entry of lymphocytes into the CNS 21,22 ,butthenatureofthereceptors
has not been identified and may vary depending on the inflammatory
conditionorlocation.
1 Institute for Research in Biomedicine, Bellinzona, Switzerland. 2 Theodor Kocher Institute, University of Bern, Bern, Switzerland. 3 Neuroimmunology Unit and 4 Center of
Excellence for Biomedical Research, University of Genoa, Genoa, Italy. 5 Mount Sinai School of Medicine, New York, New York, USA. 6 Advanced Biotechnology Center,
Genoa, Italy. Correspondence should be addressed to F.S. (federica.sallusto@irb.unisi.ch).
514 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Lymphocytes can also enter into the CNS
through the choroid plexus 23,24 . These
plexuses are villous structures extending into
the cerebrospinal fluid–filled ventricular
spaces that establish a blood–cerebrospinal
fluid barrier at the level of apical tight junctions
between epithelial cells of the choroid
plexus. The specific enrichment for memory
T cells in the cerebrospinal fluid of healthy
people 24 and patients with multiple sclerosis
25 , as well as the regulated functional
expression of adhesion molecules on the
choroid plexus epithelium 26 , suggest that
T cells may use these structures to enter the
cerebrospinal fluid and disseminate to the
meningeal and perivascular spaces 27 . However,
the homing determinants that regulate
the entry of lymphocytes through the choroid
plexus and the function of this port of entry
relative to that of the blood-brain barrier
remain to be defined.
It is well established that both in humans
and mice, chemokine receptors have differences
in expression on subsets of effector and
memory T cells and provide specificity to cell
trafficking both in the steady state and
inflammation 28 . For example, CCR7 endows
naive and central memory T cells with the
ability to migrate into peripheral lymph
nodes, whereas CCR9 and CCR4 direct the
migration of memory T cells into the gut and
skin, respectively. In addition, some receptors,
such as CXCR3 and CCR5, are ‘preferentially’
expressed on T H1 cells, whereas
others, such as CCR3, CCR8 and CRTH2,
are ‘preferentially’ expressed on T H2 cells. The finding that in humans,
CCR6 (A000629), the receptor for CCL20 (a chemokine expressed in
the liver, lungs and Peyer’s patches) 29 , is expressed on IL-17-producing
T cells (including some that also produce interferon-g (IFN-g)) 30
prompted us to investigate the function of CCR6 in regulating T H-
17-mediated immune pathology.
Here we report that CCR6-deficient (CCR6-knockout) mice were
highly resistant to EAE induction but became susceptible when given
transfer of small numbers of CCR6-sufficient T cells. CCR6 was
required on the first wave of TH-17 cells that entered the CNS through
epithelial cells of the choroid plexus, which constitutively expressed
CCL20 in both mice and humans. CCR6 + T cells triggered the entry of
a second wave of T cells that migrated in large numbers into the CNS
by crossing activated parenchymal vessels. Our results demonstrate
distinct molecular requirements and anatomical sites for lymphocyte
entry during the development of EAE and suggest that the CCR6-
CCL20 axis controls an evolutionary conserved pathway of immune
surveillance in the brain.
RESULTS
CCR6-knockout mice are highly resistant to EAE
To study the function of CCR6 in T H-17-mediated immunopathology,
we compared the susceptibility of wild-type and CCR6-knockout
mice 31 to EAE induction. When immunized by subcutaneous injection
of a peptide consisting of amino acids 35–55 of myelin oligodendrocyte
glycoprotein (MOG(35–55)) in complete Freund’s
a
c
Clinical score
* * *
3.5
3.0
2.5
2.0
1.5
WT
CCR6-KO
1.0
0.5
0
* * * * * * * *
0 5 10
Time (d)
15 20
WT
CCR6-KO
20 µm
20 µm
20 µm
20 µm
b
WT
CCR6-KO
d
IL-17 (ng/ml)
20 µm 20 µm
20 µm
10
8
6
4
2
0
0
WT KO WT KO WT KO WT KO
Spleen Brain Spleen Brain
adjuvant (CFA) and pertussis toxin, wild-type mice developed a
monophasic disease characterized by ascending paralysis 9–16 d
after immunization and prominent leukocyte infiltration and microglial
activation in the CNS (Fig. 1a–c and data not shown). Notably,
most CCR6-knockout mice were completely resistant to the development
of EAE and did not show leukocyte infiltration in the CNS,
whereas a few (8 of 26) had only minimal disease (clinical score, 0.17
± 0.28 (mean ± s.d.)) that developed with similar kinetics but had a
much lower score (Table 1). On day 20, we detected MOG-specific
T cells able to produce IL-17 and IFN-g in the spleens of both wildtype
and CCR6-knockout mice (Fig. 1d). In contrast, we detected
MOG-specific T cells that produced IL-17 and IFN-g only in the
brains of wild-type diseased mice. These results indicate that in CCR6knockout
mice, MOG-reactive T H-17 and T H1 cells are primed in
lymph nodes and enter the circulation but fail to migrate into the CNS
and induce EAE.
CCR6 is not required for TH-17 priming
To rule out a possible contribution of CCR6 in the priming and
differentiation of T cells, we assessed the ability of T cells from CCR6knockout
mice to polarize into effector T cells in vitro. We stimulated
wild-type and CCR6-knockout CD4 + naive T cells in vitro in T H1-,
TH2- or TH-17-polarizing conditions (Supplementary Methods
online). Both wild-type and CCR6-knockout T cells showed a similar
capacity to upregulate mRNA encoding the transcription factors
T-bet, GATA-3 and RORgt and to produce IFN-g, IL-4 and IL-17,
IFN-γ (ng/ml)
ARTICLES
Figure 1 CCR6-deficient mice are resistant to EAE induction. (a) Clinical scores of wild-type mice
(WT; n ¼ 5) and CCR6-knockout mice (CCR6-KO; n ¼ 5) at various times after immunization with
MOG(33–55) in CFA. *, P o 0.01 (Student’s t-test). Data are representative of six experiments (mean
and s.d.). (b,c) Immunofluorescence and histology of cryosections of brains and spinal cords from wildtype
and CCR6-knockout mice with clinical scores of 2 and 0, respectively, collected on day 13 after
perfusion. (b) Staining with anti-laminin (red) and anti-CD45 (green). (c) Immunoperoxidase staining
for CD45 with a hemalaun counterstain. Scale bars, 20 mm. Results are representative of three
experiments. (d) IL-17 and IFN-g in supernatants of total splenic cells (spleen; 2.5 10 6 ) and CD45 +
cells enriched from brains (brain; 0.5 10 6 to 1 10 6 ) obtained from MOG(35–55) immunized
wild-type mice (clinical score, 3) and CCR6-knockout mice (KO; clinical score, 0), restimulated in vitro
with MOG(33–55) and assessed at 72 h of culture. IL-17 and IFN-g were not detected in supernatants
of unstimulated cultures (data not shown). Data are representative of three independent experiments
with three mice per group in each (mean and s.d.).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 515
20 µm
25
20
15
10
5

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Table 1 EAE in wild-type and CCR6-knockout mice
Mouse genotype Incidence
Day of onset
(mean ± s.d.)
Maximum
clinical score
(mean ± s.d.)
Wild type 26 of 29 (89.65%) 16.67 (±0.81) 2.78 (±0.42)
CCR6-knockout 8 of 26 (30.77%) 16.33 (±0.52) 0.17 (±0.28)
Results are cumulative data from six different experiments.
respectively (Supplementary Fig. 1a online). As shown before in
humans 30 , CCR6 was selectively upregulated on wild-type TH-17 cells
but not T H1 or T H2 cells (Supplementary Fig. 1b). To address
whether the lack of CCR6 affected T cell priming in vivo, for example,
by affecting antigen presentation by dendritic cells that express CCR6
(ref. 32), we immunized wild-type and CCR6-knockout mice by
subcutaneous injection of ovalbumin admixed with lipopolysaccharide–monophosphoryl
lipid A or CFA (Supplementary Methods). The
magnitude and quality of the ovalbumin-specific recall T cell
responses were similar in wild-type and CCR6-knockout mice, with
T H1 responses prevailing in mice primed with lipopolysaccharide–
monophosphoryl lipid A and TH-17 responses prevailing in mice
primed with CFA (Supplementary Fig. 1c). Furthermore, OT-II
T cell antigen receptor (TCR)–transgenic T cells labeled with the
cytosolic dye CFSE (carboxyfluorescein diacetate succinimidyl ester)
and adoptively transferred into wild-type or CCR6-knockout mice
proliferated and differentiated into T H-17 cells to a similar extent in
both types of mice after immunization with ovalbumin in CFA
(Supplementary Fig. 2 online). These results collectively indicate
that CCR6 is selectively upregulated in developing mouse T H-17
cells and that its expression is not required for T H-17 priming and
differentiation in vitro and in vivo.
CCR6-knockout mice given T cells are susceptible to EAE
To determine whether CCR6 was required only on T cells, we
adoptively transferred green fluorescent protein–positive (GFP + )
MOG-specific naive 2D2-transgenic T cells into CCR6-knockout
mice or wild-type control mice. After immunizing mice with
MOG(35–55) in CFA, we found CCR6 + IL-17-producing 2D2
T cells in similar proportions in the spleens of CCR6-knockout and
wild-type mice (Fig. 2a,b), which indicated that in CCR6-knockout
mice, MOG-reactive T cells differentiated into CCR6 + TH-17 effector
Figure 2 Transfer of wild-type 2D2 T cells reconstitutes EAE susceptibility
in CCR6-knockout mice. Analysis of wild-type and CCR6-knockout mice
given sham treatment (PBS) or adoptive transfer of naive GFP + CD4 + 2D2
T cells or naive polyclonal CD4 + T cells from wild-type mice or mice of
various knockout strains, then immunized 16 h later with MOG(33–55)
in CFA for EAE induction. (a) Expression of CCR6 (left) and production of
IL-17 and IFN-g (right) by GFP-gated 2D2 T cells transferred into wild-type
and CCR6-knockout mice and primed 7 d earlier (left). Numbers in
quadrants indicate percent CCR6 + cells on GFP-gated 2D2 T cells (left)
or percent cells in each (right). Data are representative of three different
experiments. (b) Production of IL-17 and IFN-g by sorted CCR6 + and CCR6 –
2D2 T cells primed in wild-type mice. Numbers in quadrants indicate
percent cells in each. (c) Clinical scores of wild-type and CCR6-knockout
mice given sham treatment or adoptive transfer of 2D2 T cells (+ 2D2) or
with 2D2 CXCR6-knockout T cells (+ 2D2 CXCR6-KO). (d) Clinical scores
of CCR6-knockout mice given sham treatment or adoptive transfer of CD4 + T
cells (T) from wild-type, CXCR3-knockout, IFN-g-knockout or CCR7-knockout
mice. Data are representative of at least three different experiments with
groups of four or five mice per condition (b,c; mean and s.d.).
cells. Notably, when given 2D2 T cells, both CCR6-knockout and wildtype
mice developed EAE with same kinetics and with greater severity
relative to that of mice that did not receive 2D2 T cells (Fig. 2c).
Finally, CCR6-knockout mice given CCR6-deficient 2D2 T cells
(obtained by crossing of 2D2 TCR–transgenic mice with CCR6-knockout
mice) did not develop EAE (Fig. 2c). These results indicate that
CCR6 expression on transferred T cells is both necessary and sufficient
to reconstitute disease susceptibility in CCR6-knockout mice.
We further investigated the molecular requirements for the reconstitution
of susceptibility to EAE induction in CCR6-knockout mice.
Adoptive transfer of wild-type polyclonal naive CD4 + T cells reconstituted
disease susceptibility in CCR6-knockout mice, as did the
transfer of naive CD4 + T cells from CXCR3-knockout and IFN-gknockout
mice (Fig. 2d). Transfer of naive CD4 + T cells from CCR7knockout
mice induced the development of EAE, which was delayed
approximately 2 weeks (Fig. 2d), possibly due to inefficient priming of
CCR7-deficient naive T cells in lymph nodes. The results presented
above collectively indicate that T cells able to reconstitute disease
susceptibility in CCR6-knockout mice require CCR6, which is characteristic
of T H-17 cells, but not CXCR3 or IFN-g, which are
characteristic of T H1 cells.
CCR6-knockout T cells enter the CNS during active EAE
We next analyzed the cells that infiltrated the CNS at the peak of the
disease. We isolated CD45 + cells from the perfused brains and spinal
cords of CCR6-knockout and wild-type mice given 2D2 T cells and
counted endogenous and transferred 2D2 T cells on the basis of the
expression of GFP and CD45 congenic markers. Unexpectedly, in
both wild-type and CCR6-knockout mice, GFP + 2D2 T cells were
almost completely undetectable in the brain, and we detected only a
few in the spinal cord on day 20 (Fig. 3a). Notably, the abundant
cellular infiltrate in both CCR6-knockout and wild-type mice was
mainly endogenous CCR6-knockout CD4 + and CD8 + T cells, B cells,
neutrophils and inflammatory monocytes (Fig. 3a,b and data not
shown). We confirmed those findings by immunohistology that
showed only rare GFP + cells in inflammatory clusters and several
inflammatory foci with apparently no infiltrating GFP + cells (Fig. 3c).
After in vitro stimulation with MOG(35–55), brain-infiltrating endogenous
CD4 + T cells produced IL-17 and IFN-g (Fig. 3d), which
demonstrated that they were antigen specific and may have been
T H-17 cells, T H1 cells and T cells producing both IL-17 and
IFN-g. The results reported above suggest that both CCR6-sufficient
a
CD4
CD4
b
IL-17
GFP
12.4 0.7
+ 2D2 in WT
10.4
CCR6
GFP
14.8 1.3
+ IFN-γ
2D2 in CCR6-KO
11.6
84.1
81
2.7
2.9
CCR6
IFN-γ
Sorted
GFP
7.3 0.1 85.3 2.5
+ CCR6 – Sorted
2D2 GFP + CCR6 + 2D2
91.4
IFN-γ
1.2
IL-17
IL-17
IL-17
11.9
IFN-γ
0.2
c
Clinical score
d
Clinical score
WT
WT + 2D2
CCR6-KO
CCR6-KO + 2D2
3.5
3.0
2.5
2.0
1.5
1.0
0.5
CCR6-KO + 2D2 CCR6-KO
0
0 5 10
Time (d)
15 20
4.0
3.0
2.0
1.0
WT T in CCR6-KO
CXCR3-KO T in CCR6-KO
IFN-γ-KO T in CCR6-KO
CCR7-KO T in CCR6-KO
CCR6-KO
0
0 5 10 15
Time (d)
20
25 30
516 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a
CD4
Spleen
WT + 2D2 CCR6-KO + 2D2
70.8 0.05 71.3 0.05
GFP-2D2
Brain
WT + 2D2 CCR6-KO + 2D2
84.1 0 84.6 0
Spinal cord
WT + 2D2 CCR6-KO + 2D2
69.0 0.089 73.9 0.01
0
– + 2D2 – + 2D2
WT CCR6-KO
WT CCR6-KO WT CCR6-KO
c d
20 µm 20 µm 20 µm 10 µm
and CCR6-deficient T cells can enter the CNS in mice with active EAE
and that at least some of these cells are MOG specific.
CCR6-knockout T cells roll and adhere to inflamed CNS venules
To visualize the interaction of effector T cells with CNS postcapillary
venules, we injected in vitro–primed wild-type and CCR6-knockout
TH-17 cells into wild-type mice in which EAE had been induced 12 d
before and measured cell rolling and adhesion by intravital microscopy.
Activated wild-type and CCR6-knockout CD4 + T cells rolled
and adhered to spinal cord postcapillary venules to a similar extent
(Fig. 4a). In addition, the fraction of permanently adhering T cells was
similar and did not change over time (Fig. 4b), which suggested that
the adhering cells began to enter the spinal cord during the time of
observation. In contrast, when we injected T cells into CCR6-knockout
mice that had been primed with MOG(35–55) and CFA 12 d
before but did not develop disease, the cells failed to interact with
endothelial cells (data not shown). Similarly, and consistent with a
published report 21 , in healthy mice, activated T cells of either
wild-type or CCR6-knockout origin failed to show adhesive interaction
with brain endothelium (data not shown). These results
indicate that CCR6 is not sufficient to mediate adhesion of TH-17
cells to uninflamed endothelial cells (even in CCR6-knockout mice
immunized with CFA and pertussis toxin) and that it is not required
for the entry of T cells into the CNS once tissue inflammation
is established.
T cells migrate into the CNS in two waves
On the basis of the results reported above, we hypothesized that a first
wave of T H-17 cells migrating into the uninflamed CNS in a CCR6dependent
way is needed to trigger activation of parenchymal
b Brain
Figure 3 Recruitment of endogenous effector T cells into the CNS of CCR6-knockout mice after
transfer of 2D2 T cells. (a,b) Seven-color staining (with lineage-specific antibodies) of infiltrating
cells isolated from various organs (above plots) of wild-type and CCR6-knockout mice immunized
20 d before. (a) Expression of CD4 and GFP on gated CD3 + T cells from spleens, brains and spinal
cords of mice after transfer of GFP + 2D2 T cells. Numbers in quadrants indicate percent CD4 + GFP –
cells (top left) or CD4 + GFP + cells (top right). Data are representative of three independent
experiments with three mice each. (b) Absolute number of endogenous (GFP – )CD3 + CD4 + Tcells,
CD3 + CD8 + T cells and B220 + B cells in brains of wild-type and CCR6-knockout mice with (+ 2D2)
or without (–) adoptive transfer of GFP + 2D2 T cells. NS, not significant. P values, Student’s t-test.
Data are representative of three independent experiments (mean and s.d. of three mice per group).
CD4 + T cells (×10 4 )
4
3
2
1
NS
P < 0.01
CD8 + T cells (×10 3 )
IL-17 (ng/ml)
IFN-γ (ng/ml)
Brain
10
8
6
4
2
0
– + 2D2 – + 2D2
WT CCR6-KO
Spleen
4
3
2
1
0
– + 2D2 – + 2D2
WT CCR6-KO
Spleen
15
10
5
0
NS
P < 0.01
– + 2D2 – + 2D2
WT CCR6-KO
0
– + 2D2 – + 2D2
WT CCR6-KO
4
Brain
NS
3
P < 0.01
0
– + 2D2 – + 2D2
WT CCR6-KO
Brain
4
NS
P < 0.01
3
endothelial cells and the recruitment of a second wave of effector
cells, composed of T H-17 and T H1 cells, which can migrate in a CCR6independent
way across the activated blood-brain barrier. To test
our hypothesis, we transferred into wild-type mice a mixture of
naive 2D2 T cells from wild-type and CCR6-knockout mice that could
be identified by their expression of GFP and CD45 and, after
immunizing the recipient mice with MOG(35–55), we measured
the relative proportions of transferred cells in the brain during
disease development (Fig. 5a). On day 10, wild-type T cells
predominated over CCR6-knockout T cells. In contrast, on day 16,
more cells of both populations were present in the brain in
similar proportions.
We also adoptively transferred a mixture of in vitro–primed T H-17
cells from wild-type mice (which had more than 60% CCR6 + cells)
and CCR6-knockout mice and, 48 h later, measured their migration
into the spleens and brains of mice that developed EAE (Fig. 5b). We
recovered wild-type and CCR6-knockout T cells in similar proportions
from spleens at all time points tested. In contrast, in the brain,
wild-type T cells were present at higher frequency than were CCR6knockout
T cells on day 7, but the proportion of CCR6-knockout cells
steadily increased at later time points (days 10 and 16; Fig. 5b). These
results are collectively consistent with the hypothesis that early
migration of T cells in the brain is CCR6 dependent, whereas late
migration can occur in a CCR6-independent way.
Epithelial cells of the choroid plexus express CCL20
To identify the initial port of entry of CCR6 + T cells, we analyzed in
the CNS of healthy wild-type and CCR6-knockout mice and diseased
wild-type mice expression of the CCR6 ligand CCL20, which is
expressed in liver and Peyer’s patches 29 . We found CCL20 on scattered
B cells (×10 3 )
IL-17 (ng/ml)
IFN-γ (ng/ml)
ARTICLES
Brain
15
10
5
2
1
2
1
0
NS
P < 0.01
– + 2D2 – + 2D2
WT CCR6-KO
(c) Immunofluorescence staining for CD45 (red) in spinal cords from a wild-type mouse and a CCR6-knockout mouse (two images from each) after adoptive
transfer of GFP + 2D2 T cells at day 20 of EAE, showing inflammatory cuffs with or without GFP + 2D2 T cells. Scale bars, 20 mm. Results are representative
of three independent experiments. (d) IL-17 and IFN-g in supernatants of CD45 + cells enriched from spleens and brains of immunized mice with (+ 2D2) or
without ( ) adoptive transfer of GFP + 2D2 T cells, and stimulated in vitro with MOG(33–55), assessed at 72 h of culture. IL-17 and IFN-g were not
detected in unstimulated cultures (data not shown). P values, Student’s t-test. Data are representative of three separate experiments (mean and s.d.).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 517

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Figure 4 CCR6 is not required for T cell rolling and the adhesion of T cells
to inflamed brain endothelia. Intravital microscopy (epi-illumination) of
in vitro–primed T H-17 cells obtained from wild-type and CCR6-knockout
mice, labeled with CellTracker green and injected into the catheterized right
carotid arteries of wild-type mice (with a clinical score of 0.5–1) that had
undergone laminectomy from C2–C7 and removal of the dura. (a) Initial
contact, rolling and capture of T cells in postcapillary venules, among total
T cells passing through a given venule during a 1-min observation period.
Each dot represents one postcapillary venule; small horizontal lines indicate
the median with the interquartile range. Statistical analysis, Mann-Whitney
U-test. Data are representative of three experiments analyzing 20 venules in
five mice that received wild-type T H-17 cells and 18 venules in four mice that received CCR6-knockout T H-17 cells. (b) Adherent and plugging T cells,
presented as cell number per field of view (Cells/FOV), normalized to the number of fields of view. Data are representative of three experiments analyzing four
to six fields of view per mouse (mean and s.d. of five mice that received wild-type T H-17 cells and four mice that received CCR6-knockout T H-17 cells).
cells in several regions of the brain but not on normal or inflamed
endothelial cells of the brain (data not shown). Notably, however, we
found very high and uniform expression of CCL20 in epithelial cells of
the choroid plexus of healthy wild-type and CCR6-knockout mice and
of wild-type mice that developed EAE (Fig. 5c and data not shown).
There was no staining of the choroid plexus parenchyma (data not
shown), which suggested accumulation of CCL20 in and around
choroid plexus epithelium. We detected CCL20 mRNA in liver and
Peyer’s patches, as expected, as well as comparable expression in
choroid plexuses of healthy and EAE mice, whereas we did not
detect it in the brain parenchyma (Supplementary Methods and
Supplementary Fig. 3 online).
The findings reported above suggested that CCL20 might be
required for the initial entry of CCR6 + cells into uninflamed CNS
through the choroid plexus and cerebrospinal fluid and that the same
pathway might be used for the constitutive migration of lymphocytes
for immunosurveillance of the CNS. That hypothesis is consistent
with two observations. First, the choroid plexus of MOG-immunized
CCR6-knockout mice had an accumulation of CD45 + cells greater
than that of wild-type mice (Fig. 5d), which supports the idea that the
epithelial layer of the choroid plexus is the barrier that EAE-initiating
T cells must pass through. In addition, CD45 + cells accumulated in the
choroid plexus in CCR6-deficient mice in the parenchyma between
Figure 5 CCR6 is required for the migration of T cells into the CNS through
CCL20-expressing epithelial cells of the choroid plexus in the steady state
and at early time points of EAE. (a,b) Recruitment of wild-type and CCR6knockout
T cells in developing EAE. (a) Recovery of naive CD4 + Tcells
(2.5 10 6 ) sorted from wild-type (GFP + ) mice and CCR6-knockout
(CD45.1 + ) mice, mixed at a ratio of 1:1 and transferred into wild-type
CD45.2 + mice, which were then immunized with MOG(33–55) in CFA,
presented as the proportion of transferred wild-type and CCR6-knockout
T cells recovered from the brain 10 d or 16 d after immunization.
(b) Recovery of naive CD4 + T cells sorted from wild-type (GFP + ) and CCR6knockout
(CD45.1 + ) mice, stimulated in vitro in T H-17 conditions, mixed at
a ratio of 1:1 and transferred into CD45.2 + wild-type mice in which EAE
was induced 7, 10 or 16 d earlier, presented as the proportion of transferred
T cells recovered from the spleen and brain at 48 h after transfer. Data are
representative of three separate experiments (mean and s.d.). (c) Immunofluorescence
of CCL20 staining of cryosections of brains from wild-type and
CCR6-knockout mice collected after perfusion. Scale bars, 50 mm. Results
are representative of three experiments. (d) Immunoperoxidase staining and
hemalaun counterstaining of CD45 + cells in the choroid plexus parenchyma
of wild-type and CCR6-knockout mice in which EAE was induced 13 d
earlier. Scale bars, 50 mm (top row and bottom left) or 20 mm (bottom,
right). Results are representative of at least three separate experiments.
(e) Absolute numbers of CD4 + Tcells,CD8 + T cells and CD19 + B cells in
brains of wild-type and CCR6-knockout mice in the steady state. Each symbol
represents an individual mouse; small horizontal lines indicate the median.
P values, Student’s t-test. Data are representative of two experiments.
a 20
b
Event (%)
15
10
5
NS NS NS
Cells/FOV
0
0
WT CCR6-KO WT CCR6-KO WT CCR6-KO WT CCR6-KO WT CCR6-KO WT CCR6-KO
Initial contact Rolling Capture 10 min 30 min 60 min
laminin-positive endothelial and epithelial basement membranes (data
not shown). Finally, in the steady state, CCR6-knockout mice had
significantly fewer CNS-associated CD4 + T cells, CD8 + T cells and
B cells than did wild-type mice (Fig. 5e).
CCR6 and CCL20 in human cerebrospinal fluid and brain
To determine whether the findings reported above obtained with the
mouse model could be extended to the human disease of multiple
sclerosis, we analyzed CCR6 expression in T cells from patients with
multiple sclerosis and analyzed CCL20 expression in the brains of
controls and patients with multiple sclerosis. In eight patients with an
initial demyelinating event (the first clinical episode of multiple
sclerosis), CCR6 + CD25 – CD4 + inflammatory T cells were present at
significantly higher frequencies in the cerebrospinal fluid than in
peripheral blood (Fig. 6a). This finding demonstrates that many
cells detectable in the cerebrospinal fluid at the earliest clinical
demyelinating event express CCR6.
Immunohistology of normal healthy tissues showed that CCL20
was expressed in the liver, mainly in Kupffer cells, and in scattered cells
a
Frequency (%)
WT
CCR6-KO
Brain
90
60
30
Cells (×10 2 )
WT
CCR6-KO
b
c d
e
Frequency (%)
CD4
15 P < 0.05
15 50
40
10
10
30
5
5
20
10
0
WT CCR6-KO
0
WT CCR6-KO
0
WT CCR6-KO
+ T cells CD8 + T cells CD19 + B cells
P < 0.05
P < 0.05
Cells (×10 2 )
25
20
15
10
5
Spleen Brain
60 90
40
20
CCR6-KO WT
60
30
Cells (×10 2 )
WT
CCR6-KO
0
0
0
Day 10 Day 16 Day 7 Day 10 Day 16 Day 7 Day 10 Day 16
Frequency (%)
50 µm 50 µm 50 µm
518 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
50 µm
50 µm
20 µm

© 2009 Nature America, Inc. All rights reserved.
a
CD4 + CCR6 + CD25 –
T cells (%)
50
40
30
20
10
P = 0.004
0
PBMC CSF
b Healthy tissues c
Liver
Ileum
Choroid plexus Choroid plexus
in the crypts of the ileum (Fig. 6b). Notably, there was strong and
uniform staining for CCL20 in epithelial cells of the choroid plexus in
the brain (Fig. 6b), whereas normal brain parenchyma contained
some cells that stained faintly for CCL20, including astrocytes and cells
with neuronal and microglial morphology (data not shown). In tissues
from patients with multiple sclerosis, we detected high CCL20
expression in inflamed areas, in astrocytes positive for glial fibrillary
acidic protein and in the choroid plexus (Fig. 6c and data not shown).
Accordingly, we detected some CD3 + CCR6 + T cells in inflamed
parenchyma near astrocytes that were CCL20 + (data not shown).
These findings suggest that in humans, recruitment of CCR6 + cells
into the CNS may occur initially through epithelial cells of the choroid
plexus that constitutively express CCL20, whereas at a later stage,
CCL20 production by activated astrocytes may contribute to the
recruitment of CCR6 + T cells into the brain parenchyma. Thus,
CCR6 and CCL20 may represent an evolutionary conserved axis
that regulates the CNS entry and dissemination of T cells in the
steady state and during inflammation.
DISCUSSION
We have shown here that the CCR6 serves an essential function in the
initiation of EAE by controlling the migration of a first wave of
autoreactive T H-17 cells in the uninflamed CNS. The entry of CCR6 +
T cells into the CNS probably occurs through the blood–cerebrospinal
fluid barrier, as epithelial cells of the choroid plexus constitutively
expressed the CCR6 ligand CCL20. The first wave of migratory T cells
was required for the recruitment of a second wave of T cells that
entered the CNS parenchyma in a CCR6-independent way through
activated parenchymal postcapillary venules.
The conclusions above were based on three main findings. First,
when immunized by a standard protocol MOG(35–55) in CFA plus
pertussis toxin, CCR6-knockout mice developed TH-17 responses but
failed to develop EAE. In particular, we did not recover MOG-reactive
effector T cells from the CNS of these mice and their parenchymal
venules did not support the extravasation of adoptively transferred
T cell blasts. Second, the transfer of wild-type naive T cells into CCR6knockout
mice was sufficient to reconstitute disease susceptibility.
However, wild-type CCR6 + T cells predominated in the CNS only at
an initial asymptomatic stage of the disease, whereas during active
disease, the cellular infiltrate in the CNS parenchyma was composed of
endogenous CCR6-knockout T cells, including MOG-reactive TH-17
and T H1 cells, that efficiently adhered to
activated CNS endothelial cells. Third, the
CCR6 ligand CCL20 had high and constitutive
expression by epithelial cells of the
choroid plexus but not by parenchymal
endothelial cells. In addition, in CCR6knockout
mice, T cells seemed to be trapped
between the endothelial and epithelial basement
membranes of the choroid plexus.
Our findings provide a molecular and
anatomical basis for distinguishing between
constitutive and inflammatory pathways of
T cell entry into the CNS and support a twostep
model of EAE pathogenesis in which a
first wave of CCR6 + TH-17 cells leads to the
CCR6-independent recruitment in the CNS
of a second wave of T cells, including T H1
cells, and inflammatory leukocytes. The twostep
model of EAE pathogenesis is consistent
with the finding that in the early phases of
EAE, CD4 + T cells accumulate first in the subarachnoid space and
subsequently appear in the CNS parenchyma 33,34 . In those studies, the
nature of the T cells that initially infiltrate the subarachnoid space and
their port of entry were not defined, although it was suggested that
they might enter directly through the meningeal vessels or disseminate
through the cerebrospinal fluid after crossing the epithelial cell layer of
the choroid plexus. The model is also consistent with the published
description of ‘pioneer’ lymphocytes that migrate into uninflamed
CNS through the choroid plexus in a P-selectin-dependent way 20 and
with studies emphasizing the importance of the activation state and
antigenic specificity of T cells that migrate in the CNS in the initial
phases of EAE 35,36 .Atwo-wavemigrationofTH-17 and TH1 cells
into tissues may apply to other autoimmune diseases such as collageninduced
arthritis 37 . In addition, in a model of Mycobacterium
tuberculosis infection, it has been shown that vaccination induces
IL-17-producing CD4 + T cells that populate the lung and, after
challenge, trigger the production of chemokines that recruit CD4 + T
cells that produce IFN-g, which ultimately restrict bacterial growth 38 .
The findings that CCR6 is essential for the entry of T cells into the
CNS in the early phase of EAE and the expression of CCL20 in
epithelial cells of the choroid plexus provide a molecular determinant
and an anatomical site for the first triggering step in EAE pathogenesis.
It remains to be established which adhesion molecules are needed
to T cells to cross the blood–cerebrospinal fluid barrier and whether
chemokines other than CCL20 might be expressed on epithelial cells
of the choroid plexus. Consistent with our finding that CXCR3 was
not needed to mediate the entry of T cells in the initial phase of EAE,
we found that its ligands, CXCL9 and CXC10, were not expressed in
the choroid plexus of C57BL/6 mice (D.B., unpublished data), which
also do not express CXCL11 because of a genetic defect.
An implication of our findings is that one function of the first wave
of CCR6 + TH-17 cell is to activate the postcapillary venules in the CNS
parenchyma, which in normal circumstances are inefficient in sustaining
rolling and adhesion of activated leukocytes and hence the entry of
cells into the CNS parenchyma 21 . It is conceivable that once they have
entered through the choroid plexus into the cerebrospinal fluid, T H-17
cells may disseminate at the pial surface and in the enlarged perivascular
Virchow-Robin spaces, where they may recognize self antigens
displayed on resident antigen-presenting cells. Once activated, T H-17
cells produce cytokines and chemokines that act locally to trigger
activation of the blood-brain barrier and to initiate the influx of large
Brain (MS)
Figure 6 Expression of CCL20 and CCR6 in human normal tissue and multiple sclerosis tissues.
(a) CCR6 + cells among gated CD4 + CD25 – T cells from matched samples of peripheral blood (PBMC)
and cerebrospinal fluid (CSF) obtained from eight patients with multiple sclerosis. P value, Mann-
Whitney U-test. Data are representative of eight experiments. (b,c) CCL20 expression on normal human
tissues (b) and in various areas of the brains of patients with multiple sclerosis (MS; c), assessed by
immunohistochemistry of formalin-fixed, paraffin-embedded sections with a CCL20-specific antibody.
Original magnification, 40 (b, all images; c, top left and bottom row) or 60 (c, top right). Results
are representative of two separate experiments on two different tissue samples.
ARTICLES
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 519

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
numbers of inflammatory cells, including T H-17 and T H1 cells,
neutrophils and inflammatory monocytes, that build up the lesions
characteristic of EAE. Although we have shown that the T cell infiltrate
was composed of MOG-reactive T cells, it is likely that antigennonspecific
bystander T cells are also recruited through the bloodbrain
barrier and may constitute a large fraction of the total cellular
infiltrate, as has been shown in delayed-type hypersensitivity reactions
as well as several autoimmune diseases 39 .
It is worth noting that in the experimental conditions of active EAE
used in our studies, effector T cells were induced in peripheral lymph
nodes by subcutaneous immunization and had to enter uninflamed
CNS, as the inflammatory stimuli were limited to the local effect of
CFA and the systemic effect of pertussis toxin administered at the time
of immunization. It can be anticipated that the requirement for the
CCR6-dependent entry of autoreactive T cells through the choroid
plexus, which seems to be rate-limiting in steady-state conditions, may
be bypassed whenever the CNS vasculature is activated by local or
systemic inflammatory stimuli 40 or in models of passive EAE induced
by the injection of highly activated T cell blasts.
An open issue relates to the nature of the cytokines produced by the
first wave of T H-17 cells needed for the initiation of EAE. T H-17 cells
produce IL-17, IL-22, tumor necrosis factor and, in some cases, IFN-g,
which can act on microglial cells and endothelial cells 41 . We found that
IFN-g expression was not required in the CCR6 + T cells that initiated
EAE, and we have preliminary evidence that IL-17A may also be
dispensable (A.R., unpublished data). In addition, studies have shown
that IL-17A and IL-17F, as well as IL-22, may not be required for the
development of EAE 42,43 ; such findings would be consistent with
either a redundant function of these cytokines in disease development
or with an essential involvement of another functional property
(cytokine or chemokine production; chemokine receptor expression)
of TH-17-lineage cells. Experiments with CCR6-knockout mice given
adoptive transfer of CCR6-sufficient T cells carrying selective genetic
defects will help to resolve this issue.
We have shown here that once EAE was triggered, effector T cells of
both wild-type and CCR6-knockout origin efficiently rolled and
adhered to inflamed CNS postcapillary venules and migrated with
similar efficiency into the CNS. These findings indicate that at a late
stage, the CCR6-CCL20 axis is dispensable and the recruitment of
inflammatory cells into the CNS parenchyma can be mediated by
other chemokine receptors 44 . Several inflammatory chemokines are
upregulated in the CNS during EAE 45–47 , and some chemokine
receptors, such as CCR1, CCR2 and CXCR3, have been shown to be
involved in EAE 48–52 . In addition, a CCR5 receptor antagonist and
antibodies to CCL20 have been reported to diminish disease severity
53,54 . Such findings are consistent with the presence of multiple
redundant mechanisms that regulate the entry of leukocytes into the
CNS once inflammation has been established.
In the context of its function in migration into the CNS, CCR6
expression on activated and memory T cells and other cell types
deserves mention. We have shown that CCR6 was selectively upregulated
in developing mouse TH-17 cells but not TH1 orTH2 cells both
in vitro and in vivo. In addition, we have shown that in mice, CCR6 was
selectively induced by CFA, the adjuvant typically used for EAE
induction. It is notable that in humans, CCR6 is expressed not only
on T H-17 cells, where it is expressed together with CCR4, but also on
cells that produce both IL-17 and IFN-g, aswellasonasubsetofT H1
cells characterized by CXCR3 expression 30 . It remains to be determined
whether CCR6 can be induced on mouse T H1 cells in certain conditions,
a possibility that may explain the finding that T H1 cells can enter
uninflamed brain 40 , although in that case 40 , the port of entry into the
CNS was not identified. CCR6 is also expressed on B cells, which have
been linked to the pathogenesis of multiple sclerosis 55 , and on a subset
of regulatory T cells 56 . Indeed, it has been shown that CCR6 is
important in regulating the recruitment of T H-17 cells as well as
regulatory T cells into inflammatory tissues 57 . Thus, it is possible that
CCR6 is used by a variety of cells to enter the brain through the
choroid plexus and participate in immunoregulatory circuits.
Our findings help address a controversial aspect of the relative
functions of TH1 andTH-17 cells in brain inflammation. Although
studies have demonstrated a requirement for IL-17-producing T cells
in EAE 10 , it was puzzling that T H-17 cells often accounted for a minor
fraction of infiltrating T cells, at least at the peak of the disease. Indeed,
subsequent studies have led to a reevaluation of the function of T H1
cells 58 . The two-step model of EAE pathogenesis provides a way to
reconcile those findings with the following considerations. First, the
proportion of T H1andT H-17 cells may be highly variable depending
on the priming conditions. Second, a few autoreactive CCR6 + T cells
may be able to promote the vascular recruitment of large numbers of
inflammatory cells. Third, CCR6 is expressed in subsets of TH1 cells in
humans and possibly in mice.
It is well appreciated that chemokines and their receptors, together
with adhesion molecules, control the constitutive homing of lymphocytes
to lymphoid and nonlymphoid tissues to accomplish their
immunosurveillance function. For example, CCR9 defines a subset of
gut-homing lymphocytes, whereas CCR4 and CCR10 direct skin-tropic
T-cell trafficking 59 . Similar selective mechanisms that target T cells to
theCNSarethoughttoexist 60 , but they have not been identified. On
the basis of our findings, we propose that CCR6 is a brain-specific
determinant for the constitutive trafficking of patrolling T cells and
B cells in the CNS. The rapid distribution of T cells throughout the
choroid plexus and the cerebrospinal fluid into the meningeal space
seems to be instrumental for broad surveillance over the entire surface
of the CNS, thus maximizing the chance that antigen-bearing antigenpresenting
cells will be detected at superficial sites while at the same
time the entry of inflammatory cells into the parenchyma, where this is
needed, will be elicited. The cytokine-mediated activation of endothelial
cells of the blood-brain barrier executed by patrolling T cells might
provide a means for targeting large numbers of effector cells to a
precise region of the CNS parenchyma.
Some aspects of the mouse model also apply to humans. These
include the fact that in humans, CCR6 is expressed by T H-17 cells 30 ,as
well as the enrichment for CCR6 + cells in the cerebrospinal fluid of
patients presenting with the first clinical symptom of multiple
sclerosis, a condition that may parallel the earliest phase of EAE.
Moreover, in noninflammatory conditions in humans, CCL20 is
expressed almost exclusively by epithelial cells of the choroid plexus,
whereas in multiple sclerosis tissues, CCL20 can also be expressed by
astrocytes positive for glial fibrillary acidic protein (A.U., unpublished
data), which suggests that in multiple sclerosis, selective recruitment of
CCR6 + cells may occur through the choroid plexus in the early phase
of disease, whereas at a later stage, astrocytes may contribute to the
recruitment of CCR6 + T cells in brain parenchyma. These findings
indicate involvement of the CCR6-CCL20 axis in initiating brain
inflammation in humans and possibly an evolutionary conserved
mechanism of immune surveillance in the CNS. The relapsingremitting
form of human multiple sclerosis and the anatomical
distribution of multiple sclerosis lesions may be consistent with either
distinct waves of migration or asynchronous activation of already
resident CCR6 + T cells. Distinguishing between these possibilities is
relevant to understanding the potential therapeutic use of CCR6blocking
drugs in human multiple sclerosis.
520 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
METHODS
Mice. C57BL/6 mice were from Harlan; 2D2 TCR–transgenic (006912), UBC-
GFP (004353) and Ifng –/– (002287) mice were from the Jackson Laboratory.
OT-II TCR–transgenic mice (provided by J. Kirberg) were bred onto backgrounds
of various Cd45 alleles in the animal facility of the Institute for
Research in Biomedicine. Ccr6 –/– mice have been described 31 , Cxcr3 –/– mice
were provided by C. Gerard, and Ccr7 –/– mice were provided by M. Lipp. Mice
were treated in accordance with guidelines of the Swiss Federal Veterinary
Office and experiments were approved by the Dipartimento della Sanità
eSocialità.
EAE model. Groups of female C57BL/6 mice 8–10 weeks of age were
immunized subcutaneously on day 0 with 100 mg MOG(35–55) (MEVG
WYRSPFSRVVHLYRNGK; Servei de Proteòmica, Pompeu Fabra University,
Barcelona) emulsified in CFA (Difco). Pertussis toxin in 100 ml salinewas
injected intravenously twice, on days 0 and 2 or days 1 and 3. Disease severity
was assigned scores on the following scale: 0, no disease; 1, tail weakness; 2,
paraparesis; 3, paraplegia; 4, paraplegia with forelimb weakness or paralysis; 5,
moribund or dead. In some EAE experiments, naive T cells (1 10 5 )from
GFP + 2D2-transgenic mice or 2D2 CCR6-knockout mice or total CD4 +
T cells (10 10 6 ) from wild-type, CXCR3-knockout, IFN-g-knockout or
CCR7-knockout mice were transferred intravenously into mice 16 h before
immunization. In some experiments, a mixture of GFP + 2D2-transgenic T cells
(2.5 10 6 ) and 2D2 CCR6-knockout T cells (at a ratio of 1:1) were
transferred intravenously into mice 16 h before immunization. Alternatively,
effector TH-17 cells were generated in vitro from GFP + 2D2-transgenic T cells
(2.5 10 6 ) and 2D2 CCR6-knockout T cells (at a ratio of 1:1) and were
transferred intravenously in wild-type mice at various times after EAE induction.
For the preparation of CNS lymphocytes, mice were perfused through the
left cardiac ventricle with cold PBS. The forebrain and cerebellum were
dissected, were cut into pieces and were digested for 45 min at 37 1C with
collagenase D (1 mg/ml; Roche Diagnostics) and DNaseI (1 mg/ml; Sigma).
CD45 + cells were isolated by passage of the tissue through a cell strainer
(70 mm), followed by incubation with beads coated with antibody to CD45
(anti-CD45; 130-052-301; Milteny). After passage through the column, CD45 +
cells were washed and resuspended in culture medium for further analysis.
Flow cytometry analysis and in vitro stimulation. For analysis of mouse
phenotypes, the following monoclonal antibodies were used: anti-L-selectin
(MEL14), anti-CD44 (IM7), anti-CD4 (RM4-5), anti-CD8a (53-6.7), anti-CD3
(145-2C11), anti-CD28 (37.51), anti-B220 (RA3-6B2), anti-CD19 (1D3), anti-
CD11b (M1/70), anti-CD45.1 (A20), anti-CD45.2 (104), anti-CD127 (A7R34)
and anti-IFN-g (XMG1.2; all from eBiosciences); and anti-IL-17A (TC11-
18H10) and anti-CCR6 (140706; both from BD Biosciences). For intracellular
cytokine staining, cells were stimulated for 4 h with phorbol 12-myristate
13-acetate (100 nM; Sigma) and ionomycin (1 mg/ml; Sigma), with the final 2 h
of culture in the presence of brefeldin A (10 mg/ml; Sigma). Labeled antibodies
were used after cells were fixed in 4% (wt/vol) paraformaldehyde and made
permeable with 0.5% (wt/vol) saponin (Sigma-Aldrich). Six-color staining of
the cell surface was done with the appropriate combinations of antibodies
conjugated to fluorescein isothiocyanate, phycoerythrin, peridinine chlorophyll
protein complex, phycoerythrin-indotricarbocyanine, allophycocyanin, allophycocyanin-indotricarbocyanine
or biotin, and with streptavidin labeled with
phycoerythrin-indotricarbocyanine or allophycocyanin-indotricarbocyanine
(BD Biosciences). Samples were acquired on a FACSCanto (BD Biosciences)
and were analyzed with FlowJo software (TreeStar).
For studies of human cell phenotypes, peripheral blood and cerebrospinal
fluid were obtained from eight patients (after informed consent was provided)
who presented with a first demyelinating event suggestive of multiple sclerosis
and underwent venipuncture and lumbar puncture for diagnostic purposes.
The study was approved by the Ethical Committee and Board of the Department
of Neurosciences, Ophthalmology and Genetics, University of Genoa.
Peripheral blood mononuclear cells were isolated with Ficoll and cerebrospinal
fluid leukocytes were collected after centrifugation, then these cells were
incubated for 30 min with the following monoclonal antibodies: fluorescein
isothiocyanate–conjugated anti-CD4 (RPA-T4), phycoerythrin-conjugated
anti-CCR6 (11A9), phycoerythrin-indodicarbocyanine–conjugated anti-CD25
ARTICLES
(M-A251) and allophycocyanin-conjugated anti-CD45RO (UCHL1; all from
BD Biosciences). At the end, cells were washed with PBS, were resuspended,
were counterstained with propidium iodide (1 mg/ml; Sigma Aldrich) and were
analyzed by flow cytometry of the propidium iodide–negative population. A
FACSCanto (BD Biosciences) was used for all flow cytometry and data were
analyzed with FACSDiva & FloJo software.
For antigen-specific restimulation of mouse T cells, 5 10 6 splenocytes or
2.5 10 6 lymph node cells were cultured for 3 d in presence of MOG(35–55).
Cells purified from the brain were cultured for 3 d in presence of fixed
splenocytes loaded with MOG(35–55). For fixation, spleens were removed from
naive C57BL/6 mice and erythrocytes were lysed. Splenocytes were incubated
for 90 min at 37 1C with MOG(35–55) (50 mg/ml), then were washed in 1%
(vol/vol) FCS in PBS and were fixed for 30 s at 20 1C in 0.05% (vol/vol)
glutaraldehyde (Merck). An equal volume of a solution of glycine (0.2 M;
Sigma-Aldrich) was added for 30 s and then cells were washed and used for
coculture. Cytokines in culture supernatants were measured by enzyme-linked
immunosorbent assay according to the manufacturer’s protocols (BD Biosciences).
Data were analyzed with the SoftMax program.
Immunohistology and immunofluorescence. Mice were perfused through the
left cardiac ventricle with 1% (vol/vol) formaldehyde (Grogg Chemie) in PBS.
Brains and spinal cords were removed, were embedded in Tissue-Tek optimum
cutting temperature compound (Haslab) and were ‘snap-frozen’ in a bath of
dry ice and isopentane (Grogg Chemie). Cryostat sections 6 mm inthickness
were air-dried overnight, were fixed in acetone and were stained for immunohistology
with a three-step immunoperoxidase staining kit according to the
manufacturer’s protocol (Vectastain; Reactolab). For immunofluorescence
staining, sections were blocked for 20 min with skim milk and then were
incubated for 1 h each with primary and secondary antibody diluted in skim
milk, with washing steps of Tris-buffered saline between incubations. After a
final wash in Tris-buffered saline, sections were mounted in Mowiol solution
(Calbiochem). Antibodies used were as follows: fluorescein isothiocyanate–
conjugated anti-CD45 (30F11; BD Biosciences), anti-CCL20 (AB9829; Abcam),
anti-laminin (Z0097; Dako), anti-PECAM-1 (Mec13.3; BD Biosciences) and
anti–‘pan cytokeratin’ (C2562; Sigma).
LifeSpan Biosciences analyzed CCL20 expression in healthy and diseased
human tissues. Formalin-fixed, paraffin-embedded tissues were stained with
rabbit polyclonal anti-CCL20 (15 mg/ml; AB9829; Abcam). The detection
system consisted of anti-rabbit secondary antibody (BA-1000; Vector) and an
ABC-AP kit (avidin–biotinylated enzyme complex–alkaline phosphatase;
AK-5000; Vector) with a Red substrate kit (SK-5100; Vector), which was used
to produce a fuchsia-colored deposit. Only tissues that were positive for
staining of CD31 and vimentin were used. The negative control consisted
of immunohistochemistry of adjacent sections in the absence of primary
antibody. Slides were imaged with a DVC 1310C digital camera coupled to a
Nikon microscope.
Intravital microscopy. Active EAE was induced by immunization with
MOG(35–55) in CFA. Mice with a score of 0.5 (limp tail) to 1 (hind leg
weakness) were used for intravital fluorescence videomicroscopy experiments.
The spinal cord window was created as described in the Supplementary
Methods. Normal body temperature was maintained throughout the entire
experiment. Epi-illumination techniques were used for intravital fluorescence
videomicroscopy with an IVM500 microscope (Mikron Instruments) coupled
to a 50-Watt mercury lamp (HBO 50 microscope illuminator; Zeiss) and
combined with blue filter blocks (exciter, 455DF70; dichroic, 515DRLP;
emitter, 515ALP) and green filter blocks (exciter, 525DF45; dichroic, 560DRLP;
emitter, 565ALP). Observations were made with 4 , 10 and 20 longdistance
working objectives (Zeiss). All experiments were recorded by means of
a low-light silicon-intensified target video camera with an optional image
intensifier for weak fluorescence (Dage-MTI of Michigan City). Data were
transferred to a digital video system for later offline analysis of the interaction
of cells with CNS microvessels. The injection of 1% (vol/vol) tetramethylrhodamine
isothiocyanate–conjugated dextran in 0.9% (wt/vol) NaCl allowed
visualization of the spinal cord microvasculature. In parallel, mouse T cells were
stained for 45 min at 37 1C with2.5mM CellTracker green (Molecular Probes)
and were injected into the carotid artery (4 10 6 T cells in three
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 521

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
0.1-ml aliquots). Injection into this catheter resulted in direct transport into
observed vessels, where interactions were recorded by digital video for subsequent
offline analysis. Several adjacent areas were scanned in a ‘stepwise’ way
for recording of permanently adhering T cells at 10 min, 30 min and 1 h after
injection. Recorded videos were subsequently analyzed offline as described
(Supplementary Methods).
Statistics. Differences between data sets were analyzed by Student’s t-test or a
Mann-Whitney U-test.
Accession code. UCSD-Nature Signaling Gateway (http://www.signaling-gate
way.org): A000629
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank D. Jarrossay for cell sorting; T. Périnat and S. Minghelli for
immunohistology analysis; E. Mira Catò and L. Perlini for technical assistance;
B. Becher (University Hospital Zurich), C. Gerard (Harvard Medical School),
J. Kirberg (Max-Plank Institute of Immunobiology) and M. Lipp (Max
Delbruck Center) for mouse strains; L. Sallusto for discussions and support;
and A. Almeida, A. Martín-Fontecha, G. Napolitani, U. Schenk and
M. Uguccioni for discussions. Supported by the Swiss National Science
Foundation (31-101962 to F.S.), the European Commission Sixth Framework
Programme (LSB-CT-2005-518167 IINOCHEM and LSHG-CT-2005-005203
MUGEN), the US National Multiple Sclerosis Society (B.E. and C.C.), the Swiss
Multiple Sclerosis Society, the Italian Foundation for Multiple Sclerosis (Stem
Cell Project 2007-2009), the European Union-funded International Graduate
Program in Molecular Medicine (A.R.), Boehringer Ingelheim Fonds (D.B.) and
the Helmut Horten Foundation (to The Institute for Research in Biomedicine).
AUTHOR CONTRIBUTIONS
A.R. did most of the experiments and contributed to experimental design;
C.C., D. Baumjohann, F.B. and D. Bottinelli did experiments; S.L. generated the
CCR6-knockout mice and provided intellectual input; B.E. and A.U. interpreted
data, provided intellectual input and contributed to writing the manuscript;
A.L. provided intellectual input and wrote the manuscript; and F.S. conceived
the study, interpreted the data and wrote the manuscript.
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. McGeachy, M.J. & Cua, D.J. Th17 cell differentiation: the long and winding road.
Immunity 28, 445–453 (2008).
2. Bettelli, E., Korn, T., Oukka, M. & Kuchroo, V.K. Induction and effector functions of
T H17 cells. Nature 453, 1051–1057 (2008).
3. Dong, C. TH17 cells in development: an updated view of their molecular identity and
genetic programming. Nat. Rev. Immunol. 8, 337–348 (2008).
4. Weaver, C.T., Hatton, R.D., Mangan, P.R. & Harrington, L.E. IL-17 family cytokines and
the expanding diversity of effector T cell lineages. Annu. Rev. Immunol. 25, 821–852
(2007).
5. Gaffen, S.L. An overview of IL-17 function and signaling. Cytokine 43, 402–407 (2008).
6. Steinman, L. A brief history of T H17, the first major revision in the T H1/T H2hypothesis
of T cell-mediated tissue damage. Nat. Med. 13, 139–145 (2007).
7. Ivanov, I.I. et al. The orphan nuclear receptor RORgt directs the differentiation program
of proinflammatory IL-17 + T helper cells. Cell 126, 1121–1133 (2006).
8. Komiyama, Y. et al. IL-17 plays an important role in the development of experimental
autoimmune encephalomyelitis. J. Immunol. 177, 566–573 (2006).
9. Hofstetter, H.H. et al. Therapeutic efficacy of IL-17 neutralization in murine experimental
autoimmune encephalomyelitis. Cell. Immunol. 237, 123–130 (2005).
10. Langrish, C.L. et al. IL-23 drives a pathogenic T cell population that induces
autoimmune inflammation. J. Exp. Med. 201, 233–240 (2005).
11. Tzartos, J.S. et al. Interleukin-17 production in central nervous system-infiltrating
T cells and glial cells is associated with active disease in multiple sclerosis.
Am. J. Pathol. 172, 146–155 (2008).
12. Bettelli, E. et al. Loss of T-bet, but not STAT1, prevents the development of experimental
autoimmune encephalomyelitis. J. Exp. Med. 200, 79–87 (2004).
13. Luger, D. et al. Either a Th17 or a Th1 effector response can drive autoimmunity:
conditions of disease induction affect dominant effector category. J. Exp. Med. 205,
799–810 (2008).
14. Kroenke, M.A., Carlson, T.J., Andjelkovic, A.V. & Segal, B.M. IL-12- and IL-23modulated
T cells induce distinct types of EAE based on histology, CNS chemokine
profile, and response to cytokine inhibition. J. Exp. Med. 205, 1535–1541 (2008).
15. Stromnes, I.M., Cerretti, L.M., Liggitt, D., Harris, R.A. & Goverman, J.M. Differential
regulation of central nervous system autoimmunity by T H1andT H17 cells. Nat. Med.
14, 337–342 (2008).
16. Lees, J.R., Golumbek, P.T., Sim, J., Dorsey, D. & Russell, J.H. Regional CNS responses
to IFN-g determine lesion localization patterns during EAE pathogenesis. J. Exp. Med.
205, 2633–2642 (2008).
17. Engelhardt, B. & Ransohoff, R.M. The ins and outs of T-lymphocyte trafficking to the
CNS: anatomical sites and molecular mechanisms. Trends Immunol. 26, 485–495
(2005).
18. Yednock, T.A. et al. Prevention of experimental autoimmune encephalomyelitis by
antibodies against a 4b 1 integrin. Nature 356, 63–66 (1992).
19. Engelhardt, B., Vestweber, D., Hallmann, R. & Schulz, M. E- and P-selectin are not
involved in the recruitment of inflammatory cells across the blood-brain barrier in
experimental autoimmune encephalomyelitis. Blood 90, 4459–4472 (1997).
20. Carrithers, M.D., Visintin, I., Kang, S.J. & Janeway, C.A. Jr. Differential adhesion
molecule requirements for immune surveillance and inflammatory recruitment. Brain
123, 1092–1101 (2000).
21. Piccio, L. et al. Molecular mechanisms involved in lymphocyte recruitment in inflamed
brain microvessels: critical roles for P-selectin glycoprotein ligand-1 and heterotrimeric
Gi-linked receptors. J. Immunol. 168, 1940–1949 (2002).
22. Vajkoczy, P., Laschinger, M. & Engelhardt, B. a4-integrin-VCAM-1 binding mediates
G protein-independent capture of encephalitogenic T cell blasts to CNS white matter
microvessels. J. Clin. Invest. 108, 557–565 (2001).
23. Engelhardt, B., Wolburg-Buchholz, K. & Wolburg, H. Involvement of the choroid
plexus in central nervous system inflammation. Microsc. Res. Tech. 52, 112–129
(2001).
24. Ransohoff, R.M., Kivisakk, P. & Kidd, G. Three or more routes for leukocyte migration
into the central nervous system. Nat. Rev. Immunol. 3, 569–581 (2003).
25. Giunti, D. et al. Phenotypic and functional analysis of T cells homing into the CSF of
subjects with inflammatory diseases of the CNS. J. Leukoc. Biol. 73, 584–590
(2003).
26. Steffen, B.J., Breier, G., Butcher, E.C., Schulz, M. & Engelhardt, B. ICAM-1, VCAM-1,
and MAdCAM-1 are expressed on choroid plexus epithelium but not endothelium and
mediate binding of lymphocytes in vitro. Am. J. Pathol. 148, 1819–1838 (1996).
27. Pedemonte, E. et al. Mechanisms of the adaptive immune response inside the central
nervous system during inflammatory and autoimmune diseases. Pharmacol. Ther. 111,
555–566 (2006).
28. Bromley, S.K., Mempel, T.R. & Luster, A.D. Orchestrating the orchestrators: chemokines
in control of T cell traffic. Nat. Immunol. 9, 970–980 (2008).
29. Rossi, D.L., Vicari, A.P., Franz-Bacon, K., McClanahan, T.K. & Zlotnik, A. Identification
through bioinformatics of two new macrophage proinflammatory human chemokines:
MIP-3a and MIP-3b. J. Immunol. 158, 1033–1036 (1997).
30. Acosta-Rodriguez, E.V. et al. Surface phenotype and antigenic specificity of human
interleukin 17–producing T helper memory cells. Nat. Immunol. 8, 639–646
(2007).
31. Cook, D.N. et al. CCR6 mediates dendritic cell localization, lymphocyte homeostasis,
and immune responses in mucosal tissue. Immunity 12, 495–503 (2000).
32. Greaves, D.R. et al. CCR6, a CC chemokine receptor that interacts with macrophage
inflammatory protein 3a and is highly expressed in human dendritic cells. J. Exp. Med.
186, 837–844 (1997).
33. Brown, D.A. & Sawchenko, P.E. Time course and distribution of inflammatory and
neurodegenerative events suggest structural bases for the pathogenesis of experimental
autoimmune encephalomyelitis. J. Comp. Neurol. 502, 236–260 (2007).
34. Kivisakk, P. et al. Localizing central nervous system immune surveillance: Meningeal
antigen-presenting cells activate T cells during experimental autoimmune encephalomyelitis.
Ann. Neurol. published online, doi:10.1002/ana.21379 (21 May 2008).
35. Hickey, W.F., Hsu, B.L. & Kimura, H. T-lymphocyte entry into the central nervous
system. J. Neurosci. Res. 28, 254–260 (1991).
36. Flugel, A. et al. Migratory activity and functional changes of green fluorescent effector
cells before and during experimental autoimmune encephalomyelitis. Immunity 14,
547–560 (2001).
37. Hirota, K. et al. Preferential recruitment of CCR6-expressing Th17 cells to inflamed
joints via CCL20 in rheumatoid arthritis and its animal model. J. Exp. Med. 204,
2803–2812 (2007).
38. Khader, S.A. et al. IL-23 and IL-17 in the establishment of protective pulmonary CD4 +
T cell responses after vaccination and during Mycobacterium tuberculosis challenge.
Nat. Immunol. 8, 369–377 (2007).
39. Steinman, L. A few autoreactive cells in an autoimmune infiltrate control a vast
population of nonspecific cells: a tale of smart bombs and the infantry. Proc. Natl.
Acad. Sci. USA 93, 2253–2256 (1996).
40. O’Connor, R.A. et al. Cutting Edge: Th1 cells facilitate the entry of Th17 cells
to the central nervous system during experimental autoimmune encephalomyelitis.
J. Immunol. 181, 3750–3754 (2008).
41. Kebir, H. et al. Human T H17 lymphocytes promote blood-brain barrier disruption and
central nervous system inflammation. Nat. Med. 13, 1173–1175 (2007).
42. Kreymborg, K. et al. IL-22 is expressed by Th17 cells in an IL-23-dependent fashion,
but not required for the development of autoimmune encephalomyelitis. J. Immunol.
179, 8098–8104 (2007).
43. Haak, S. et al. IL-17A and IL-17F do not contribute vitally to autoimmune neuroinflammation
in mice. J. Clin. Invest. 119, 61–69 (2009).
44. Charo, I.F. & Ransohoff, R.M. The many roles of chemokines and chemokine receptors
in inflammation. N. Engl. J. Med. 354, 610–621 (2006).
45. Balashov, K.E., Rottman, J.B., Weiner, H.L. & Hancock, W.W. CCR5 + and CXCR3 +
T cells are increased in multiple sclerosis and their ligands MIP-1a and IP-10 are
expressed in demyelinating brain lesions. Proc. Natl. Acad. Sci. USA 96, 6873–6878
(1999).
522 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
46. Ambrosini, E., Columba-Cabezas, S., Serafini, B., Muscella, A. & Aloisi, F. Astrocytes
are the major intracerebral source of macrophage inflammatory protein-3a/CCL20 in
relapsing experimental autoimmune encephalomyelitis and in vitro. Glia 41, 290–300
(2003).
47. Columba-Cabezas, S., Serafini, B., Ambrosini, E. & Aloisi, F. Lymphoid chemokines
CCL19 and CCL21 are expressed in the central nervous system during experimental
autoimmune encephalomyelitis: implications for the maintenance of chronic neuroinflammation.
Brain Pathol. 13, 38–51 (2003).
48. Rottman, J.B. et al. Leukocyte recruitment during onset of experimental allergic
encephalomyelitis is CCR1 dependent. Eur. J. Immunol. 30, 2372–2377
(2000).
49. Muller, M. et al. CXCR3 signaling reduces the severity of experimental autoimmune
encephalomyelitis by controlling the parenchymal distribution of effector and regulatory
T cells in the central nervous system. J. Immunol. 179, 2774–2786
(2007).
50. Liu, L. et al. Severe disease, unaltered leukocyte migration, and reduced IFN-g
production in CXCR3 / mice with experimental autoimmune encephalomyelitis.
J. Immunol. 176, 4399–4409 (2006).
51. Gaupp, S., Pitt, D., Kuziel, W.A., Cannella, B. & Raine, C.S. Experimental autoimmune
encephalomyelitis (EAE) in CCR2 / mice: susceptibility in multiple strains. Am. J.
Pathol. 162, 139–150 (2003).
ARTICLES
52. Eltayeb, S. et al. Temporal expression and cellular origin of CC chemokine receptors
CCR1, CCR2 and CCR5 in the central nervous system: insight into mechanisms of
MOG-induced EAE. J. Neuroinflammation 4, 14(2007).
53. Matsui, M. et al. Treatment of experimental autoimmune encephalomyelitis with the
chemokine receptor antagonist Met-RANTES. J. Neuroimmunol. 128, 16–22 (2002).
54. Kohler, R.E., Caon, A.C., Willenborg, D.O., Clark-Lewis, I. & McColl, S.R. A role for
macrophage inflammatory protein-3a/CC chemokine ligand 20 in immune priming
during T cell-mediated inflammation of the central nervous system. J. Immunol. 170,
6298–6306 (2003).
55. Franciotta, D., Salvetti, M., Lolli, F., Serafini, B. & Aloisi, F. B cells and multiple
sclerosis. Lancet Neurol. 7, 852–858 (2008).
56. Kleinewietfeld, M. et al. CCR6 expression defines regulatory effector/memory-like cells
within the CD25 + CD4 + T-cell subset. Blood 105, 2877–2886 (2005).
57. Yamazaki, T. et al. CCR6 regulates the migration of inflammatory and regulatory T cells.
J. Immunol. 181, 8391–8401 (2008).
58. Steinman, L. A rush to judgment on Th17. J. Exp. Med. 205, 1517–1522 (2008).
59. Sigmundsdottir, H. & Butcher, E.C. Environmental cues, dendritic cells and the
programming of tissue-selective lymphocyte trafficking. Nat. Immunol. 9, 981–987
(2008).
60. Hickey, W.F. Basic principles of immunological surveillance of the normal central
nervous system. Glia 36, 118–124 (2001).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 523

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Memory T cells in nonlymphoid tissue that provide
enhanced local immunity during infection with herpes
simplex virus
Thomas Gebhardt, Linda M Wakim, Liv Eidsmo, Patrick C Reading, William R Heath & Francis R Carbone
Effective immunity is dependent on long-surviving memory T cells. Various memory subsets make distinct contributions to
immune protection, especially in peripheral infection. It has been suggested that T cells in nonlymphoid tissues are important
during local infection, although their relationship with populations in the circulation remains poorly defined. Here we describe a
unique memory T cell subset present after acute infection with herpes simplex virus that remained resident in the skin and in
latently infected sensory ganglia. These T cells were in disequilibrium with the circulating lymphocyte pool and controlled new
infection with this virus. Thus, these cells represent an example of tissue-resident memory T cells that can provide protective
immunity at points of pathogen entry.
Immune memory results from the action of various cellular components
that combine to control infectious pathogens 1 .Tcellsarekey
mediators of the memory response, although uncertainty remains as
to the exact function of individual memory T cell subsets 2–6 . Memory
cells have been categorized as effector memory and central memory
T cells, depending on their respective ability to recirculate between
secondary lymphoid organs or to enter peripheral tissues 5,7,8 . Localized
infection results in rapid population expansion of effector T cells
in draining lymph nodes, and although most are lost, a substantial
population of circulating memory T cells survives this contraction 9 .In
addition, many T cells are also present in nonlymphoid tissues 7,8 ,
where their numbers may actually exceed those in the circulation 10 .
The relationship between peripheral and recirculating memory cells
remains mostly undefined. Peripheral T cells can be replaced from the
circulation in the steady state and can be further supplemented by
effector memory T cells newly recruited during infection 11 . However,
an active mechanism of T cell retention may exist in nonlymphoid
tissues 12–14 , and this could explain the accumulation noted at sites that
have cleared an infectious virus 14–16 . Thus, once in the periphery, at
least a subset of memory cells may be separated from the circulating
Tcellpool.
Herpes simplex virus (HSV) can infect the skin of mice, causing a
primary infection that rapidly moves to the innervating sensory
ganglia 17,18 . There replication is controlled in about a week and
then persists as a tightly controlled latent infection 19 . The acute
infection induces a robust CD8 + T cell effector and memory response
that is able to control virus in the ganglia 20 . The requirement for
memory T cell responses in peripheral tissues has been examined with
Received 5 December 2008; accepted 10 February 2009; published online 22 March 2009; doi:10.1038/ni.1718
latently infected ganglia containing a long-lived population of
virus-specific memory T cells 21,22 . It was found that these memory
cells were not terminally differentiated but could instead mount a
secondary proliferative response after virus reactivation 22 . Notably,
this occurred in the ganglia itself, in the absence of any involvement of
lymphoid tissue.
One issue that arose from those studies was whether the T cells
residing in the ganglia were circulating memory cells or, alternatively,
represented an autonomous memory population separate from the
circulating pool. Although in the case of HSV, such local memory cells
act to control a preexisting latent infection 21,23 , this type of T cell
sequestration could conceivably result in more efficient protection
against reinfection by an environmentally derived pathogenic agent.
Here we provide evidence of a population of peripheral memory T cells
that seemed to be in disequilibrium with the circulation. Notably, we
show that such tissue-resident memory cells contributed to peripheral
HSV control, thus effectively limiting the extent of renewed infection.
RESULTS
Nonmigrating peripheral tissue–resident memory T cells
Stimulation of T cells occurs after transplantation of sensory ganglia
containing both a persisting latent form of HSV and a population of
CD8 + T cells specific for this pathogen 22 . In this model, the local
T cells undergo secondary restimulation exclusively in the grafts
during virus reactivation, with no apparent involvement of lymph
nodes. This suggests that either the ganglionic cells form part of a true
nonmigrating memory population separate from the recirculating
memory pool, or the excision and transplantation of the ganglia
Department of Microbiology and Immunology, The University of Melbourne, Melbourne Victoria, Australia. Correspondence should be addressed to W.R.H. (wrheath@
unimelb.edu.au) or F.R.C. (fcarbone@unimelb.edu.au).
524 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Figure 1 Tissue-resident memory T cells fail to
recirculate after localized or systemic viral
infection. (a) Flow cytometry of short-term BrdU
uptake 7 d after transplantation of latently
infected dorsal root ganglia containing gBT-I.GFP
memory CD8 + T cells (resident) into recipient
mice containing circulating OT-I.CD45.1 (Control)
and gBT-I.CD45.1 (recruited) CD8 + memory
T cells. Numbers above lines indicate percent
BrdU + cells (mean ± s.e.m.). Data are from two
independent experiments with three to four mice.
(b–d) Analysis of gBT-I populations in grafts and
spleens after transplantation of latently infected
ganglia containing gBT-I.GFP memory CD8 +
T cells (resident (res)) into mice containing
gBT-I.CD45.1 memory CD8 + T cells (recruited
(rec)), followed 6 d later by retransplantation into naive recipients and, 20 d later, intravenous challenge of these recipient mice with HSV. (b) Numbers
above outlined areas indicate percent recruited cells (left) and resident cells (right) among all CD8 + cells. (c) Proportion of gBT-I populations among all CD8 +
cells in the spleen. (d) Recovery of gBT-I cells from grafts. Statistical analyses, paired Student’s t-test. Data are from two independent experiments with five
to six mice per group (mean and s.e.m. in c,d). (e) Recruited and tissue-resident gBT-I memory CD8 + T cells 9 d after retransplantation of latent ganglia,
manipulated as described in b–d (original), together with a fresh latently infected ganglia graft containing no gBT-I cells (fresh). Statistical analyses, paired
Student’s t-test. Data are from two independent experiments with six to eight mice per group (mean and s.e.m.).
under the kidney capsule of recipient mice affects the normal egress of
T cells from this tissue. To show that recirculating memory T cells can
leave the transplanted tissues, we first recruited such a migrating
population into the grafts by transplanting latent ganglia into mice
that had circulating HSV-specific CD8 + memory cells. We generated
these primary recipients by infecting them with a recombinant
influenza virus carrying the immunodominant major histocompatibility
complex class I–restricted determinant from HSV glycoprotein B
(gB; Supplementary Fig. 1 online). In these experiments, we ‘genetically
marked’ the graft-resident T cells with green fluorescent protein
(GFP) by using cells from gBT-I.GFP-transgenic mice, whose T cells
are all specific for the immunodominant HSV gB determinant 24 .
T cells recruited from the circulation were from gBT-I.CD45.1transgenic
mice, which meant that we could use differences in the
expression of CD45 and GFP to track the respective recruited and
resident memory T cell populations. It should be noted that ganglia
transplantation resulted in recruitment of the circulating gBT-I cells as
CD45.1
Acute
2.0
L 8.1
R 1.2 R 0.1
CD8
0.5 0.3 P < 0.03
Original Fresh
8 d
30 d > 100 d 8 d 30 d > 100 d
P < 0.01 P < 0.005 P < 0.01
P < 0.01 P > 0.1 P > 0.1
100
60 4,000
800
800
50
30
Endogenous
CD8 + /cm 2
0
0
0
0
0
0
L R L R L R
L R L R L R
8 d
30 d
> 100 d
75
P < 0.001
P < 0.05
30
P < 0.001
P < 0.001
30
P < 0.001
P < 0.001
75
50
*
**
50
20
20
25
25
10
10
0
0
0
0
8 30 50
L R Spleen L R Spleen L R Spleen Time after HSV (d)
Skin Skin Skin
2,000
d e
gBT-l of TCRtg (%)
400
ARTICLES
400
P > 0.7
GFP + CD45.1 +
Figure 2 Long-term retention of virus-specific CD8 + T cells at the site of previous viral infection. (a–d) Analysis of naive gBT-I.CD45.1 CD8 + T cells
transferred into C57BL/6 mice 1 d before HSV-1 infection of the left flank skin at various times after inoculation (above graphs). L, left flank (skin); R, right
flank (skin); C, control skin from mock-infected mice. Statistical analyses, paired Student’s t-test (b,c) or analysis of variance followed by Tukey’s post-test
comparison (d). Data are from at least three independent experiments with two to four mice per group (mean and s.e.m. in b,c). (a) Numbers adjacent to
outlined areas indicate percent gBT-I.CD45.1 CD8 + T cells among all events. (b,c) Isolation of gBT-I.CD45.1 and endogenous CD8 + T cells. The number
of gBT-I.CD45.1 T cells in skin from mock-infected mice was below the limit of detection. (d) Frequency of gBT-I.CD45.1 cells among all CD8 + Tcells.
(e) Proportion of gBT-I.CD45.1 T cells among all T cell antigen receptor–transgenic T cells (TCRtg; gBT-I plus OT-I) after transfer of in vitro–activated gBT-
I.CD45.1 and OT-I.CD45.1 CD8 + T cells together into C57BL/6 mice 2–4 d after HSV skin infection. *, P o 0.01; **, P o 0.001 (paired Student’s t-test).
Data are from two to three independent experiments with three mice per group (mean and s.e.m.).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 525
Rec
Res
Skin
Spleen

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Figure 3 HSV-specific CD8 + Tcells
‘preferentially’ localize to the epidermal skin
layer covering the previously infected site and
express the intraepithelial cell marker CD103.
(a,b) Microscopy of skin after transfer of naive
gBT-I.GFP CD8 + T cells into C57BL/6 mice
1 d before skin infection with HSV-1.
(a) Immunohistochemistry of the distribution
of virus-specific gBT-I.GFP CD8 + T cells in skin
obtained from previously infected areas 13 d after
infection, then stained with DAPI and antibody to
keratin 1 expressed in keratinocytes in the upper
layer of the epidermis. Control, tissue from
contralateral flanks. Arrowheads indicate gBT-
I.GFP CD8 + T cells in the dermal-epidermal
region. Scale bars, 100 mm. (b) Epidermal sheets
from previously infected flank skin 25 d after
infection, when the epithelium had reformed. The
density of gBT-I.GFP CD8 + T cells is shown in the
central area of lesion resolution (lesion center), on the edge of the lesion (lesion periphery) and in an unaffected area (control). Virus-specific T cells could not
be detected by this method in unaffected areas in infected mice (control). Scale bars, 50 mm. (c) CD103 expression by memory gBT-I.CD45.1 CD8 + T cells
isolated from spleen and epidermal sheets 13 weeks after infection of the skin with HSV. Numbers above lines (right) indicate percent CD103 + Infected
Control
Spleen
1.5
b Lesion center
gBT-l (GFP) Keratin 1 DAPl
Lesion periphery Control
Vα2 CD103
Epidermis
98.1
Vα2 CD103
gBT-l (GFP) DAPl
cells. Data are
from one experiment representative of three.
HSV-infected recipients of primary grafts confirmed that only the
recruited T cells underwent considerable population expansion after
systemic challenge (Supplementary Fig. 2 online).
To demonstrate that the resident cells had a migration defect and
did not simply lack the ability to mount a systemic response, we again
recruited memory T cells into reactivating ganglia and, on day 6 after
transplantation, collected the graft and retransplanted it into a naive
recipient together with ‘fresh’ latent ganglia in a separate site under the
same kidney capsule (Supplementary Fig. 3 online). On day 9 after
retransplantation, we collected each graft and assessed if the infiltrating
memory T cells (CD45.1 + ) and tissue-residing T cells (GFP + ) were
able to enter the neighboring ganglia, which represented a fresh site of
HSV infection. We detected a substantial population of recruited
memory CD8 + T cells in the fresh graft, in contrast to those resident
memory T cells carried by the original graft (Fig. 1e). Thus, the
resident memory T cells did not migrate to a nearby yet anatomically
distinct site, even though it was equivalent to the tissue of origin for
these cells.
Enrichment for ‘biased’ T cells after skin infection
The data reported above suggested that HSV-specific T cells resident
in the ganglia seemed functionally different from those found in
Figure 4 Skin-resident, virus-specific CD8 + T cells are phenotypically and
functionally different from their circulating counterparts. Analysis of naive
gBT-I.CD45.1 CD8 + T cells transferred into C57BL/6 mice 1 d before
infection of the skin with HSV-1, followed by a period of rest for the
establishment of immunological memory. (a) Flow cytometry of memory gBT-
I.CD45.1 CD8 + T cells isolated from brachial lymph nodes (bLN), spleen
and skin 30–386 d after infection (median, 51 d). Data represent four
independent experiments with pooled cell preparations from six to seventeen
mice per group (mean and s.e.m.). (b) Expression of CD49a (VLA-1) on
gBT-I.CD45.1 CD8 + T cells isolated 50 d after infection. Numbers above
outlined areas indicate percent CD49a + cells. Isotype, isotype-matched
control antibody. Data are from three independent experiments with two to
four mice per group (mean ± s.e.m.). (c) Flow cytometry of the homeostatic
proliferation of cells from ‘memory mice’ (43–428 d after infection;
median, 72 d) treated for 7 d with BrdU. Data are from six individual
experiments with three to eight mice per group (mean and s.e.m.).
Statistical analyses, repeated measures analysis of variance followed
by Tukey’s post-test comparison.
a c
CD45.1
CD45.1
the circulation. We reasoned that an analogous population of tissueresident
memory T cells would also be found in other peripheral sites
of infection, notably the skin. Evidence for this took the form of
‘preferential’ retention of virus-specific T cells in flank skin involved in
overt HSV disease relative to that in the uninvolved contralateral
flanks. There was ‘preferential’ enrichment for gBT-I T cells in
flank skin ipsilateral to the origin of infection at all times assessed
(Fig. 2a,b). In contrast, there was no significant difference in the
number of endogenous CD8 + T cells in opposing flanks after clearance
of replicating virus, which occurs by day 8 after infection 18 (Fig. 2c). It
should be noted that gBT-I cells represented a relatively small
proportion of all CD8 + T cells in the skin, especially at these later
times (day 30 or beyond; Fig. 2d). The accumulation of HSV-specific
T cells was not restricted to transgenic cells, because repeat experiments
with tetramer staining to detect HSV-specific T cells showed
similar biases (Supplementary Fig. 4 online). Retention of T cells in
the skin had a nonspecific component, as demonstrated by the use of
attenuated recombinant viruses (HSV strains rgB and rgB-L8A) 25
(Supplementary Fig. 5 online). The number of long-term gB-specific
T cells was the same in flanks infected with gB-sufficient rgB virus and
control gB-deficient rgB-L8A in dual-inoculated mice. Even scarification
alone led to equivalent infiltration into rgB-infected and control
a
CD62L hi gBT-l (%)
b
Isotype CD49a
100 P < 0.01
75
50
25
0
bLN Spleen Skin
bLN
35.2 ± 16.2
CD45.1
P < 0.001 P < 0.05
Spleen
CD122 hi gBT-l (%)
100
75
50
25
0
Skin
42.5 ± 9.8 85.1 ± 1.7
Cells
Cells
P < 0.01
P > 0.05 100
75
50
25
0
P < 0.001
bLN Spleen Skin bLN Spleen Skin
CD69 + gBT-l (%)
c
BrdU + gBT-l (%)
20
10
0
P < 0.01
P < 0.05
bLN Spleen Skin
526 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a
CD45.1
b
gBT-l/cm 2
Recipient
G 0.02 C 0 L 0.6
0.1
750 6
6
500
4
4
250
0
2
2
L R C C C L R
C C L R L R
1° HSV 1° HSV: – – +
1° HSV: – – + +
gBT-l: – + +
gBT-l: – + + +
α-CD8: – – – +
described in d or also with depleting anti-CD8 before reinfection with HSV. Data are pooled from three independent experiments with two to five mice per
group. Each symbol represents an individual mouse; small horizontal lines indicate the mean (a,b,d,e). Statistical analyses, one-way analysis of variance
followed by Tukey’s post-test comparison.
6
4
2
50 d
P < 0.001
P < 0.001 P < 0.001
P < 0.001
6
4
2
P < 0.001
PFU/sample (log)
>100 d
P < 0.001
c d e
b
PFU/sample (log)
PFU/sample (log)
6
4
2
ARTICLES
P < 0.001
P < 0.001
P > 0.05
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 527

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
later times (day 35; Supplementary Fig. 7b,c) showed that T cells were
present only in the transplanted tissue. This was consistent with the
possibility that tissue-resident memory cells are an autonomous
population. Moreover, severing of the nerves that link the skin to
the infected ganglia also indicated that the persistence of memory
T cells was not due to continuous feeding of virus or antigen to the
skin from latently infected sensory nerve bodies undergoing spontaneous
bouts of reactivation, which in any case is not a property of
mouse infection with HSV 19 .
Immune protection by tissue-resident memory T cells
As well having as ‘preferential’ accumulation of memory T cells,
ipsilateral flanks had enhanced protection against subsequent infection.
To demonstrate this without antibody interference, we infected B cell–
deficient mMT mice in the flank with HSV and challenged them at
various times afterward. The ipsilateral flank showed 100-fold better
control of infection than that of the control contralateral flank up to
50 d after the first infection (Fig. 6a). We noted a bias toward ipsilateral
protection even at more than 100 d after infection, although this was
less than that present at the earlier times. The greater protection of
ipsilateral flanks was virus specific, as it did not extend to heterologous
infection with vaccinia virus (Supplementary Fig. 8 online).
Flank immunity was T cell dependent, because elimination of CD4 +
and CD8 + cells by in vivo antibody depletion diminished antiviral
protection (Fig. 6b). Antibody to CD4 (anti-CD4) had a much greater
effect than did depletion with anti-CD8, which suggested that CD4 +
T cells dominated control of skin infection 33 or that help was critical
to the contribution of CD8 + T cells to virus elimination 34,35 . To show
that local CD8 + T cells could provide antiviral protection, we injected
mice deficient in recombination-activating gene 1 (Rag1 –/– mice) with
in vitro–activated gBT-I CD8 + T cells 2–4 d after flank infection with
HSV. Previously infected skin retained more T cells than did the
contralateral flank (Fig. 6c), and this was associated with ‘preferential’
protection against subsequent infection with HSV (Fig. 6d). Such
biased protection was eliminated by treatment with anti-CD8
(Fig. 6e), which showed that it was indeed mediated by the transferred
T cells. In the Rag1 –/– mice, CD8 + T cells act in the complete absence
of CD4 + T cell help, most likely because their numbers are in excess of
those that normally persist long term after infection of wild-type
hosts 36 . Nevertheless, these data collectively show enhanced local
protection by T cells that accumulate in tissues after resolution of a
previous infection.
DISCUSSION
There are obvious advantages provided by the existence of a sequestered
T cell population for persisting, reactivating infections such as
the one we used here. CD8 + T cells are known to maintain HSV in a
latent state in the ganglia by constant expression of effector molecules
such as granzyme B and interferon-g 37,38 . Whether such populations
exist in acute infection systems, especially after the virus has been
completely cleared, remains to be determined. There are reports of
enrichment for memory T cells 7,8 at sites of initial pathogen infection
well after clearance has been achieved 14–16 . That is similar to what we
found after infection of the skin with HSV, with a notable imbalance
of antigen-specific memory T cells in regions ipsilateral or contralateral
to the original site of inoculation. Published studies using
rapidly cleared viruses have excluded the possibility of prolonged
infection as the mechanism of persistent accumulation 13,14,16,30 .An
argument could conceivably be mounted for the possibility that
ongoing gB-specific stimulation impedes the release of circulating
memory T cells, given the nature of HSV latency. However, experiments
with bone marrow–chimeric mice have shown that even in the sensory
ganglia, T cells persist in the absence of specific recognition of infected
neurons 39 . Retention in the skin by ongoing virus reactivation seems
even less likely, given our demonstration that resident T cells survived
transplantation and thus separation from the neuronal source of
infection. Unlike humans, mice do not show spontaneous reactivation
of HSV 19 , and skin-resident T cells show no signs of continuous
antigen stimulation, such as the upregulation of granzymes, which is
evident in T cells persisting in the ganglia 40 . Thus, although ongoing
antigen presentation remains a formal possibility, it is unlikely to fully
explain the retention of HSV-specific peripheral T cells, especially in
the skin.
Although localized infection can ‘imprint’ ‘preferential’ T cell
migration 41,42 , and this could explain other cases of local accumulation
of T cells, it is difficult to see how such tissue-tropic infiltration
would contribute to the ‘side-biased’ T cell accumulation in the flank
skin shown here. Moreover, the lack of T cell migration between the
two ganglia transplanted under the same kidney capsule indicates that
these tissue-resident T cells simply did not migrate between anatomically
distinct but otherwise equivalent tissues. As a consequence, we
propose these HSV-specific T cells represent a sessile or nonmigratory
memory subset that we define as ‘tissue-resident memory T cells’.
These cells are either separate from the recirculating pool or retained
long enough in peripheral tissues to infrequently exchange with this
population. Therefore, they are different from other memory T cells,
including the central and effector memory T cells found in the
wider circulation 4,5 .
CD8 + T cells in the ganglia can mount a stand-alone new immune
response that includes helper T cell–dependent proliferation and
stimulation by dendritic cells recruited from the circulation 22 .
Together with our data here, this suggests that tissue-resident memory
T cells are not terminally differentiated but can self-renew either
through homeostatic or antigen-stimulatory mechanisms. Furthermore,
they suggest that studies confined to circulating effector
memory and central memory T cell populations may underestimate
the totality of immunological memory in a given person.
Persisting T cells have been found in human skin that has cleared
HSV infection 43 , and embedded T cells are associated with recurring
allergic responses in humans, known as ‘fixed drug eruptions’, driven
by CD8 + T cells in the epidermis that do not disseminate throughout
the skin 44,45 . In addition, pathogenic T cells persist in pre-psoriatic
human skin, and these cells can be transferred during mouse xenotransplantation
46 . In those last two examples, the residing T cells
express markers common to the tissue-resident memory T cells in our
study, such as VLA-1 and CD103; the latter are important for
epithelial localization 26 . Finally, we have shown that these memory
T cells can efficiently inhibit HSV replication in peripheral nonlymphoid
tissues, including those at the body surface. Thus, our
study raises the possibility that resident memory cells can be exploited
as a means of controlling infection at body surfaces that act as portals
of entry for invading pathogens.
METHODS
Mice. C57BL/6, B6.SJL-PtprcaPep3b/BoyJ (B6.CD45.1), gBT-I, gBT-I
B6.CD45.1 (gBT-I.CD45.1), gBT-I.GFP, OT-I B6.CD45.1 (OT-I.CD45.1),
mMT and Rag1 –/– /Je (Rag1 –/– ) mice were bred in the Department of Microbiology
and Immunology of The University of Melbourne in Parkville,
Australia. The gBT-I and OT-I mice are CD8 + T cell receptor-transgenic mice
that recognize the H-2K b -restricted HSV-1 gB epitope of amino acids 498–505
(gB(498–505)) and the ovalbumin-derived epitope of amino acids 257–264
(OVA(257–264)), respectively. B6.CD45.1 mice express the congenic marker
528 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
CD45.1, in contrast to C57BL/6, mMT and Rag1 –/– mice, which express CD45.2.
The gBT-I.CD45.1 and OT-I.CD45.1 mice were F1 offspring expressing both
CD45.1 and CD45.2. Animal experiments were approved by The University of
Melbourne Animal Ethics Committee.
Virus infection. Viruses used were HSV-1 KOS, KOS.CreTK – , KOS.rgB 25 (rgB)
and KOS.rgB-L8A (rgB-L8A) 25 , as well as WSN/NA/gB (flu-gB; from
S. Tevethia), WSN/NA/OVA (flu-OVA; from P. Doherty) and vaccinia-NP
(from J. Yewdell). The recombinant influenza viruses express gB(498–505)
(flu-gB) or OVA(257–264) (flu-OVA) in their neuraminidase stalk. Routes
of infections were epidermal (skin; 1 10 6 plaque-forming units (PFU)),
subcutaneous (5 10 2 PFU influenza virus and 1 10 6 PFU HSV) and
intravenous (1 10 5 PFU). Skin was infected with HSV-1 KOS and
KOS.CreTK – by scarification as described 18 . The HSV-1 KOS strain caused
typical zosteriform lesions spreading from the primary inoculation site along
the affected dermatomes on the same flank. The other HSV strains used had
attenuated skin replication and did not cause any visible zosteriform lesions.
Skin was infected with vaccinia-NP (5 10 7 PFU) by inoculation of flank skin
as for HSV infection. For subcutaneous infection, 25 ml of a virus suspension
was injected into foot hocks.
In vitro activation and adoptive transfer of transgenic CD8 + T cells. All
adoptive transfers of gBT-I and OT-I cells were done intravenously with
lymph node suspensions or in vitro–generated effector splenocytes activated
by peptide-pulsed targets as described 18 . First, gBT-I or OT-I splenocytes
were cultured for 4 d together with gB(498–505)- or OVA(257–264)-pulsed
(1 mg/ml) C57BL/6 splenocytes, respectively. On days 2 and 3, cultures were split
1: 2 and interleukin 2 (10 U/ml; PeproTech) was added. Then, 1 10 6 to 1.5
10 6 effector cells derived from those cultures were transferred intravenously
into mice whose skin was infected with HSV-1. Adoptive transfer of naive
gBT-I.CD45.1 cells (5 10 4 ) resulted in a frequency of gBT-I cells roughly ten
times higher than that of their endogenous HSV-1-specific counterparts 47 .
Flow cytometry of tissue cells. Mice were killed by carbon dioxide administration
and were perfused with 10 ml Hank’s buffered-salt solution (Media
Preparation Unit, The University of Melbourne). Lymph nodes and spleens
were collected and disrupted by passage through a wire mesh. Skin tissue
(1–2 cm 2 ) and ganglia were removed, were chopped into small fragments and
were incubated for 90 min at 37 1C in Eagle’s minimum essential medium
containing 2% (vol/vol) FCS (Gibco BRL), collagenase (3 mg/ml; Worthington)
and DNase (5 mg/ml; Sigma). For minimization of the digestion of certain
surface receptors, this incubation period was decreased to 30 min in some
phenotyping experiments. Thereafter, cell suspensions were filtered twice
through nylon meshes (pore size, 70 mm and 20 mm) and were stained for
30 min with antibodies for flow cytometry. The following antibodies were from
BD Pharmingen: allophycocyanin- and phycoerythrin-indotricarbocyanine–
conjugated anti-CD8a (53-6.7), fluorescein isothiocyanate– and phycoerythrin-conjugated
anti-CD45.1 (A20), fluorescein isothiocyanate–conjugated
anti-CD45.2 (104), phycoerythrin-conjugated anti-CD4 (GK1.5), anti-CD49a
(Ha31/8), anti–keyhole limpet hemocyanin hamster IgG2 553962, fluorescein
isothiocyanate–conjugated anti-CD44 (IM7), phycoerythrin-conjugated anti-
Va2 (B20.1), phycoerythrin-conjugated anti-CD69 (H1.2F3), phycoerythrinconjugated
anti-CD122 (TM-b1), fluorescein isothiocyanate–conjugated
anti-CD62L (Mel-14), fluorescein isothiocyanate–conjugated antibody to
Armenian and Syrian hamster IgG (G192-1), and anti-CD16/32 (2.4G2).
Allophycocyanin-conjugated anti-CD45.1 (A20), phycoerythrin- and allophycocyanin–Alexa
Fluor 750–conjugated anti-CD45.2 (104) and fluorescein
isothiocyanate–conjugated anti-CD103 (2E7) were from eBiosciences. H-2K b
gB(498–505)–phycoerythrin tetramer was generated at the Department of
Microbiology and Immunology of The University of Melbourne. Dead cells
were excluded by propidium iodide staining. Sphero calibration particles (BD
Pharmingen) were added to samples to allow calculation of cell numbers.
For analysis of the homeostatic turnover of memory cells, 1 mg sterile BrdU
(5-bromodeoxyuridine) was injected intraperitoneally on 7 consecutive days.
For short-pulse experiments, 1.25 mg BrdU was injected and mice were
analyzed 1 h later. Uptake was detected with a BrdU Flow kit according
to the manufacturer’s instructions (BD Pharmingen). A FACSCalibur or
ARTICLES
FACSCanto II, plus Cell QuestPro software (BD Biosciences) and FlowJo
software (TreeStar), were used for analysis.
Transplantation experiments. Latently infected dorsal root ganglia were
transplanted under the kidney capsules of syngeneic recipients as described 22 .
For transplantation of previously HSV-infected skin tissue, donor mice were
killed by carbon dioxide administration 2–4 weeks after HSV infection of the
left flank skin. After flank skin was clipped and then treated with Veet
depilation cream (Reckitt Benckiser), mice were perfused through the left
ventricle with 10 ml Hank’s balanced-salt solution. Donor skin tissue (previously
infected skin with an area of 1–1.5 cm 1 cm) was removed in aseptic
conditions for storage in ice-cold buffer. Naive recipient mice were anesthetized
and the skin of their flanks and backs was clipped, was treated with Veet cream
and was disinfected with 70% (vol/vol) ethanol. Graft beds were prepared by
carefully snipping away of the skin layer with sterile curved scissors. Finally,
graft tissue was placed onto the graft bed and was secured with Jelonet gauze
dressing (Smith & Nephew) as well as Micropore and Transpore tape (both
from 3M Health Care). For the next 4 d, recipient mice were given analgesics in
their drinking water (paracetamol; 1.3 mg/ml). After 8–10 d, bandages were
removed and engraftment of transplanted skin was monitored daily.
Immunohistochemistry. Skin tissues were fixed for 2 h on ice in 4% (vol/vol)
paraformaldehyde and 10% (wt/vol) sucrose in PBS and were frozen in Tissue
Tek (Sakura Finetek). Sections were cut on a cryomicrotome (CM3050S; Leica),
were air-dried and were fixed in ice-cold acetone. Epidermal sheets were
prepared by incubation of full-thickness skin with Dispase II (2.5 mg/ml;
Roche), followed by manual separation of the dermal and epidermal layers and
fixation in 4% (vol/vol) paraformaldehyde and 10% (wt/vol) sucrose in PBS.
Keratin expression was visualized by incubation with polyclonal rabbitanti–mouse
keratin 1 (AF 109; Covance) followed by Alexa Fluor 594–
conjugated donkey anti-rabbit (A21207; Molecular Probes). Slides were
mounted with Vectashield containing DAPI (4,6-diamidino-2-phenylindole;
Vector Laboratories). Images were acquired with a fluorescence microscope
(DMI 4000B) and digital camera (DFC 490) with IM50 software (all from
Leica) and the Adobe Photoshop program.
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank S. Tevethia (Pennsylvania State University) for WSN/NA/gB;
P. Doherty (The University of Melbourne) for WSN/NA/OVA; J. Yewdell (U.S.
National Institutes of Health) for vaccinia-NP.; D. Masopust for discussions; and
H. Kosaka for suggesting that fixed drug eruptions are caused by resident
T cells. Supported by the Australian National Health and Medical Research
Council, the Howard Hughes Medical Institute and the German Research
Foundation (GE1666/1-1 to T.G.).
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Welsh, R.M., Selin, L.K. & Szomolanyi-Tsuda, E. Immunological memory to viral
infections. Annu. Rev. Immunol. 22, 711–743 (2004).
2. Bachmann, M.F., Kundig, T.M., Hengartner, H. & Zinkernagel, R.M. Protection against
immunopathological consequences of a viral infection by activated but not resting
cytotoxic T cells: T cell memory without ‘‘memory T cells’’? Proc. Natl. Acad. Sci. USA
94, 640–645 (1997).
3. Bachmann, M.F., Wolint, P., Schwarz, K. & Oxenius, A. Recall proliferation potential of
memory CD8 + T cells and antiviral protection. J. Immunol. 175, 4677–4685
(2005).
4. Sallusto, F., Geginat, J. & Lanzavecchia, A. Central memory and effector memory T cell
subsets: function, generation, and maintenance. Annu. Rev. Immunol. 22, 745–763
(2004).
5. Sallusto, F., Lenig, D., Forster, R., Lipp, M. & Lanzavecchia, A. Two subsets of memory
T lymphocytes with distinct homing potentials and effector functions. Nature 401,
708–712 (1999).
6. Wherry, E.J. et al. Lineage relationship and protective immunity of memory CD8 T cell
subsets. Nat. Immunol. 4, 225–234 (2003).
7. Masopust, D., Vezys, V., Marzo, A.L. & Lefrancois, L. Preferential localization of effector
memory cells in nonlymphoid tissue. Science 291, 2413–2417 (2001).
8. Reinhardt, R.L., Khoruts, A., Merica, R., Zell, T. & Jenkins, M.K. Visualizing the
generation of memory CD4 T cells in the whole body. Nature 410, 101–105 (2001).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 529

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
9. Kaech, S.M., Wherry, E.J. & Ahmed, R. Effector and memory T-cell differentiation:
implications for vaccine development. Nat. Rev. Immunol. 2, 251–262
(2002).
10. Clark, R.A. et al. The vast majority of CLA + T cells are resident in normal skin.
J. Immunol. 176, 4431–4439 (2006).
11. Hikono, H. et al. T-cell memory and recall responses to respiratory virus infections.
Immunol. Rev. 211, 119–132 (2006).
12. Harris, N.L., Watt, V., Ronchese, F. & Le Gros, G. Differential T cell function and fate in
lymph node and nonlymphoid tissues. J. Exp. Med. 195, 317–326 (2002).
13. Klonowski, K.D. et al. Dynamics of blood-borne CD8 memory T cell migration in vivo.
Immunity 20, 551–562 (2004).
14. Masopust, D. et al. Activated primary and memory CD8 T cells migrate to nonlymphoid
tissues regardless of site of activation or tissue of origin. J. Immunol. 172, 4875–4882
(2004).
15. Hawke, S., Stevenson, P.G., Freeman, S. & Bangham, C.R. Long-term persistence of
activated cytotoxic T lymphocytes after viral infection of the central nervous system.
J. Exp. Med. 187, 1575–1582 (1998).
16. Hogan, R.J. et al. Activated antigen-specific CD8 + T cells persist in the lungs
following recovery from respiratory virus infections. J. Immunol. 166, 1813–1822
(2001).
17. Simmons, A. & Nash, A.A. Zosteriform spread of herpes simplex virus as a model of
recrudescence and its use to investigate the role of immune cells in prevention of
recurrent disease. J. Virol. 52, 816–821 (1984).
18. van Lint, A. et al. Herpes simplex virus-specific CD8 + T cells can clear established lytic
infections from skin and nerves and can partially limit the early spread of virus after
cutaneous inoculation. J. Immunol. 172, 392–397 (2004).
19. Stanberry, L.R. Pathogenesis of herpes simplex virus infection and animal models for
its study. Curr.Top.Microbiol.Immunol.179, 15–30 (1992).
20. Simmons, A. & Tscharke, D.C. Anti-CD8 impairs clearance of herpes simplex virus from
the nervous system: implications for the fate of virally infected neurons. J. Exp. Med.
175, 1337–1344 (1992).
21. Khanna, K.M., Bonneau, R.H., Kinchington, P.R. & Hendricks, R.L. Herpes simplex
virus-specific memory CD8 + T cells are selectively activated and retained in latently
infected sensory ganglia. Immunity 18, 593–603 (2003).
22. Wakim, L.M., Waithman, J., van Rooijen, N., Heath, W.R. & Carbone, F.R. Dendritic
cell-induced memory T cell activation in nonlymphoid tissues. Science 319, 198–202
(2008).
23. Liu, T., Khanna, K.M., Chen, X., Fink, D.J. & Hendricks, R.L. CD8 + Tcellscanblock
herpes simplex virus type 1 (HSV-1) reactivation from latency in sensory neurons.
J. Exp. Med. 191, 1459–1466 (2000).
24. Mueller, S.N., Heath, W., McLain, J.D., Carbone, F.R. & Jones, C.M. Characterization of
two TCR transgenic mouse lines specific for herpes simplex virus. Immunol. Cell Biol.
80, 156–163 (2002).
25. Stock, A.T., Jones, C.M., Heath, W.R. & Carbone, F.R. CTL response compensation for
the loss of an immunodominant class I-restricted HSV-1 determinant. Immunol. Cell
Biol. 84, 543–550 (2006).
26. Cepek, K.L. et al. Adhesion between epithelial cells and T lymphocytes mediated by
E-cadherin and the aEb7 integrin. Nature 372, 190–193 (1994).
27. Pauls, K. et al. Role of integrin a E(CD103)b 7 for tissue-specific epidermal localization
of CD8 + T lymphocytes. J. Invest. Dermatol. 117, 569–575 (2001).
28. Hemler, M.E. VLA proteins in the integrin family: structures, functions, and their role on
leukocytes. Annu. Rev. Immunol. 8, 365–400 (1990).
29. Ray, S.J. et al. The collagen binding a1b1 integrin VLA-1 regulates CD8 T cellmediated
immune protection against heterologous influenza infection. Immunity 20,
167–179 (2004).
30. Ely, K.H., Cookenham, T., Roberts, A.D. & Woodland, D.L. Memory T cell populations in
the lung airways are maintained by continual recruitment. J. Immunol. 176, 537–543
(2006).
31. Hogan, R.J. et al. Long-term maintenance of virus-specific effector memory CD8 + Tcells
in the lung airways depends on proliferation. J. Immunol. 169, 4976–4981 (2002).
32. Zammit, D.J., Turner, D.L., Klonowski, K.D., Lefrancois, L. & Cauley, L.S. Residual
antigen presentation after influenza virus infection affects CD8 T cell activation and
migration. Immunity 24, 439–449 (2006).
33. Nash, A.A. et al. Different roles for L3T4 + and Lyt2 + T cell subsets in the control of an
acute herpes simplex virus infection of the skin and nervous system. J. Gen. Virol. 68,
825–833 (1987).
34. Jennings, S.R., Bonneau, R.H., Smith, P.M., Wolcott, R.M. & Chervenak, R. CD4positive
T lymphocytes are required for the generation of the primary but not the
secondary CD8-positive cytolytic T lymphocyte response to herpes simplex virus in
C57BL/6 mice. Cell. Immunol. 133, 234–252 (1991).
35. Smith, C.M. et al. Cognate CD4 + T cell licensing of dendritic cells in CD8 + Tcell
immunity. Nat. Immunol. 5, 1143–1148 (2004).
36. Mintern, J.D., Davey, G.M., Belz, G.T., Carbone, F.R. & Heath, W.R. Cutting edge:
precursor frequency affects the helper dependence of cytotoxic T cells. J. Immunol.
168, 977–980 (2002).
37. Knickelbein, J.E. et al. Noncytotoxic lytic granule-mediated CD8 + T cell inhibition of
HSV-1 reactivation from neuronal latency. Science 322, 268–271 (2008).
38. Liu, T., Khanna, K.M., Carriere, B.N. & Hendricks, R.L. Gamma interferon can prevent
herpes simplex virus type 1 reactivation from latency in sensory neurons. J. Virol. 75,
11178–11184 (2001).
39. van Lint, A.L. et al. Latent infection with herpes simplex virus is associated with
ongoing CD8 + T-cell stimulation by parenchymal cells within sensory ganglia. J. Virol.
79, 14843–14851 (2005).
40. Mintern, J.D., Guillonneau, C., Carbone, F.R., Doherty, P.C. & Turner, S.J. Cutting edge:
tissue-resident memory CTL down-regulate cytolytic molecule expression following
virus clearance. J. Immunol. 179, 7220–7224 (2007).
41. Campbell, D.J. & Butcher, E.C. Rapid acquisition of tissue-specific homing phenotypes
by CD4 + T cells activated in cutaneous or mucosal lymphoid tissues. J. Exp. Med. 195,
135–141 (2002).
42. Mora, J.R. & von Andrian, U.H. T-cell homing specificity and plasticity: new concepts
and future challenges. Trends Immunol. 27, 235–243 (2006).
43. Zhu, J. et al. Virus-specific CD8 + T cells accumulate near sensory nerve endings in
genital skin during subclinical HSV-2 reactivation. J. Exp. Med. 204, 595–603
(2007).
44. Mizukawa, Y. et al. Direct evidence for interferon-g production by effector-memory-type
intraepidermal T cells residing at an effector site of immunopathology in fixed drug
eruption. Am. J. Pathol. 161, 1337–1347 (2002).
45. Shiohara, T. & Moriya, N. Epidermal T cells: their functional role and disease relevance
for dermatologists. J. Invest. Dermatol. 109, 271–275 (1997).
46. Boyman, O. et al. Spontaneous development of psoriasis in a new animal model shows
an essential role for resident T cells and tumor necrosis factor-a. J. Exp. Med. 199,
731–736 (2004).
47. Stock, A.T. et al. Optimization of TCR transgenic T cells for in vivo tracking of immune
responses. Immunol. Cell Biol. 85, 394–396 (2007).
530 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
T cell antigen receptor signaling and immunological
synapse stability require myosin IIA
Tal Ilani 1 , Gaia Vasiliver-Shamis 2 , Santosh Vardhana 2 , Anthony Bretscher 1,3 & Michael L Dustin 2,3
Immunological synapses are initiated by signaling in discrete T cell antigen receptor microclusters and are important for the
differentiation and effector functions of T cells. Synapse formation involves the orchestrated movement of microclusters toward
the center of the contact area with the antigen-presenting cell. Microcluster movement is associated with centripetal actin flow,
but the function of motor proteins is unknown. Here we show that myosin IIA was necessary for complete assembly and
movement of T cell antigen receptor microclusters. In the absence of myosin IIA or its ATPase activity, T cell signaling was
interrupted ‘downstream’ of the kinase Lck and the synapse was destabilized. Thus, T cell antigen receptor signaling and the
subsequent formation of immunological synapses are active processes dependent on myosin IIA.
The specific and long-lasting interface between a T cell and an antigenpresenting
cell (APC), called the ‘immunological synapse’, is critical
for the afferent and efferent limbs of the adaptive immune response 1,2 .
The supramolecular organization of the immunological synapse was
described more than a decade ago 3–5 , yet the mechanisms leading to
its formation and persistence are unknown. No function for motor
proteins in signaling and synapse formation by cells of the immune
system has been established 6,7 .
The first step in synapse formation is engagement of the T cell
antigen receptor (TCR) with the appropriate major histocompatibility
complex–antigenic peptide complex, which leads to actin-dependent
microcluster formation and recruitment of signaling components
to form a signalosome within seconds 8–10 . The TCR signalosome
includes tyrosine-phosphorylated forms of the kinase Lck (A001394),
the kinase Zap70 (A002396) and the membrane adaptor Lat
(A001392) and excludes the transmembrane phosphatase CD45
(refs. 8,9,11–13). The contact area expands by integrin-mediated
spreading as TCR microclusters continue to form at the outer
edge 11,13 . Over a period of minutes, the microclusters move to the
center of the contact area, where they fuse into larger clusters and
become part of the nonmotile central supramolecular activation
cluster (cSMAC) 13 . As there is less tyrosine phosphorylation in the
cSMAC, it has been suggested to be the site of inactivation of old
clusters, whereas new microclusters form at the periphery 9,13,14 .The
formation and movement of new TCR microcluster–based signalosomes
toward the cSMAC sustains signaling 13 .
The driving force for protein rearrangement in the immunological
synapse is unknown, although actomyosin-driven contraction has
been proposed to drive TCR movement 15 . An alternative has been
Received 11 December 2008; accepted 5 March 2009; published online 6 April 2009; doi:10.1038/ni.1723
ARTICLES
proposed on the basis of size-dependent segregation of proteins
coupled with receptor-ligand interaction kinetics and membrane
dynamics 16 . T cell synapses have been shown to have a centripetal
version of retrograde actin flow 2,17 , a process that propels growth
cones of neurons and other motile cells 18 . A close examination of the
centripetal movement of TCR microclusters shows that it is F-actin
dependent and that they move at about half of the speed of the
underlying actin cytoskeleton (140 nm/s versus 320 nm/s, respectively)
and can change course to move around barriers 2,17 . It has been
proposed that intermittent coupling between the retrograde actin
flow and the microclusters may drive centripetal movement, but the
function of ‘motors’ in this process is not known.
Members of the nonmuscle myosin II subfamily are critical to many
cellular functions, including cell polarization, migration, adhesion and
cytokinesis 19 . Members of the myosin II family are composed of a
heavy-chain dimer; each heavy chain is associated with two myosin
light chains (MLCs). Nonmuscle myosin II is activated by phosphorylation
of the MLCs to induce assembly into bipolar filaments and
contraction after interaction with actin filaments 19,20 . Three genes
encode mammalian nonmuscle myosin II heavy chains, producing the
following three isoforms: MyH9 (A004003), MyH10 and MyH14
(refs. 21,22). Of those three isoforms, only MyH9 is dominant in
T cells 6,23 . MyH9 pairs with regulatory MLCs to form a complex we
refer to here by its common name, myosin IIA. T cell crawling and the
movement of beads attached to the surface of T cells have been
shown to require myosin IIA–mediated contractility 6,24 .Inbothof
those studies, the immunological synapse seemed to form in
the absence of myosin IIA activity or in cells depleted of myosin
IIA by small interfering RNA (siRNA). Myosin IIA was recruited
1 Department of Molecular Biology and Genetics, Weill Institute for Cell and Molecular Biology, Cornell University, Ithaca, New York, USA. 2 Molecular Pathogenesis
Program, Helen L. and Martin S. Kimmel Center for Biology and Medicine of the Skirball Institute of Biomolecular Medicine, New York University School of Medicine,
New York, New York, USA. 3 These authors contributed equally to this work. Correspondence should be addressed to A.B. (apb5@cornell.edu) or M.L.D.
(michael.dustin@med.nyu.edu).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 531

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
Control
Blebb
0 s 6 s 12 s 18 s 24 s 72 s
to the synapse 6 , but its activation and function in signaling and
synapse formation were not firmly established.
Here we show that the actin-based ‘molecular motor’ myosin IIA is
an essential participant in the formation and persistence of immunological
synapses and TCR signaling. Myosin IIA was rapidly activated
after TCR engagement, and its activity was essential for the centripetal
movement of TCR microclusters. Additionally, both immunological
synapse stability and signaling ‘downstream’ of the TCR required
intact myosin IIA.
RESULTS
TCR microcluster movement requires myosin IIA
As translocation of TCR microclusters is an essential part of the
formation of the immunological synapse, we first assessed whether
myosin IIA was required for this motion. TCR microclusters can be
tracked with a supported planar bilayer system and total internal
reflection fluorescence (TIRF) microscopy 11,13 .WeusedTIRFmicroscopy
to image the motion of TCR microclusters in Jurkat T cells
on supported planar bilayers containing laterally mobile Alexa Fluor
568–labeled antibody to TCR (anti-TCR; OKT3) and intercellular
adhesion molecule 1 (ICAM-1) 17 . In agreement with published
studies 17 , TCR microclusters in Jurkat T cells moved centripetally
with an average velocity (± s.e.m.) of 0.15 ± 0.05 mm/s (P o 0.0001;
Fig. 1a and Supplementary Movie 1 online) to generate the cSMAC.
The average displacement of a microcluster from its point of formation
to the cSMAC was 2.6 ± 0.8 mm (P o 0.0001), and the
meandering index, calculated as displacement divided by track length,
was 0.83 ± 0.09 (P o 0.0001); both are consistent with published
values 17 . To assess the involvement of myosin IIA activity in microcluster
translocation, we first treated the Jurkat T cells with blebbistatin,
a well established specific inhibitor of myosin II ATPase
activity 25 . Jurkat T cells pretreated for 10 min with blebbistatin
(50 mM) formed microclusters but showed less directed microcluster
movement, with an average speed of 0.06 ± 0.02 mm/s, a displacement
of 0.25 ± 0.13 mm and a meandering index of 0.17 ± 0.09 (P o 0.0001
for all measurements; Fig. 1a and Supplementary Fig. 1 and Supplementary
Movie 2 online). We detected equivalent blockade of
microcluster centripetal motion with ML-7 an inhibitor of myosin
light-chain kinase (MLCK; Supplementary Movie 3 online). We
noted similar effects on the inhibition of microcluster movement
when we inhibited myosin IIA activity in primary human CD4 + T cells
by treatment with blebbistatin (Supplementary Movies 4–6 online).
b
TCR Myosin
120
100
80
60
40
20
0
Thus, the microclusters continued to move at 40% the speed of
control Jurkat T cells but with over fourfold more meandering and
only 10% of the displacement of control Jurkat T cells (treated with
DMSO). In mature synapses with formed cSMACs, blebbistatin
treatment did not disrupt the cSMAC, but the peripheral TCR
microclusters ceased directed movement shortly after addition of the
drug (Supplementary Movie 7 online). These results suggest that
myosin IIA activity is required for the centripetal movement of TCR
microclusters but not for microcluster formation.
To further assess the involvement of myosin IIA in the translocation
of TCR microclusters, we targeted MyH9 by siRNA. Jurkat
T cells did not recover sufficiently from nucleofection of control
siRNA to form mature synapses (data not shown). As myosin II is
required for cytokinesis 19 , siRNA vectors that require growth and
selection would also not be usable. Therefore, we ‘knocked down’
MyH9 in primary activated human CD4 + T cells, which recover well
from nucleofection. The best knockdown efficiency achieved in the
primary T cells was 35%, as assessed by immunoblot analysis (data
not shown). However, immunofluorescence analysis showed that
this was due to nearly complete knockdown of MyH9 in one third of
cells (data not shown). We analyzed microcluster tracking on planar
bilayers of all cells in several microscopic fields while indexing the
x-y coordinates of the fields, then fixed the cells and stained for
intracellular MyH9, which allowed us to identify the cells in which
MyH9 was ‘knocked down’ in the previously tracked and indexed
cells. Primary T cells depleted of MyH9 failed to form the typical
condensed cSMAC and instead had small, scattered TCR microclusters
(Fig. 1b). TCR microclusters in cells treated with control
siRNA had an average centripetal velocity of 0.12 ± 0.034 mm/s, with
an average displacement of 2.2 ± 0.53 mm and a meandering index of
0.85 ± 0.07 (P o 0.0001 for all measurements). TCR microclusters
in MyH9-deficient cells had a speed of 0.062 ± 22 mm/s, a displacement
of 0.26 ± 0.11 mm and a meandering index of 0.25 ± 0.11
(P o 0.0001 for all measurements; Supplementary Fig. 1). Notably,
there was significantly less TCR accumulation at the cSMAC
(P o 0.0001) but only slightly, nonsignificantly less total TCR
in the entire contact area in cells depleted of MyH9 (Fig. 1c).
These results obtained by siRNA knockdown of MyH9 expression
reproduced the results obtained by inhibition of myosin II activity
with blebbistatin and ML7. Thus, myosin IIA activity is required for
the translocation of TCR microclusters to form a cSMAC but not for
the formation of TCR microclusters.
c
TCR (% of control)
Control
siRNA
Total cSMAC
Figure 1 Effect of the inhibition or depletion of myosin IIA on the centripetal motion of TCR microclusters. (a) TIRF microscopy of microclusters in control
Jurkat T cells (top) or cells pretreated with blebbistatin (Blebb; bottom) added to a planar lipid bilayer containing Alexa Fluor 568–labeled anti-TCR and
ICAM-1 and imaged during the initial minutes of synapse formation. Yellow plus symbols indicate initial microcluster localization; red circles indicate
microcluster location; red lines indicate tracks of individual clusters. (b) Microscopy of primary human CD4 + cells unaffected by siRNA treatment (left;
n ¼ 1 cell per image) or treated with MYH9-specific siRNA (arrows; n ¼ 1 cell per image), then added for 20 min to a planar lipid bilayer containing Alexa
Fluor 568–labeled anti-TCR (red) and ICAM-1, and then fixed and stained for myosin IIA (green). Scale bars (a,b), 5 mm. (c) Total TCRs and TCRs at the
center of the contact area (cSMAC) in control and myosin IIA–depleted (siRNA) cells (n ¼ 30 cells per condition). Data are representative of seven (a) or
three (b) experiments with 26 samples per condition or three experiments (c error bars, s.e.m.).
532 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a
b
Stim (min):
p-MLC
Coomassie
MLC
0
0.5 1 3 5 10 30
F-actin Myosin HC Merge
Myosin IIA is activated during T cell stimulation
Our initial results indicated that myosin IIA participates in the
formation of the immunological synapse. Activation of myosin IIA
through phosphorylation of MLCs during the formation of the
immunological synapse has not been evaluated so far. We therefore
examined the phosphorylation status of MLCs in Jurkat T cells
stimulated either with soluble OKT3, which activates the TCR only,
or with ‘superantigen’ presented by Raji B cells as APCs, which engages
the TCR and multiple adhesion and costimulatory molecules. MLCs
were not detectably phosphorylated in resting Jurkat T cells, but within
30 s of stimulation by soluble OKT3, they became phosphorylated and
this phosphorylation was sustained for at least 30 min (Fig. 2a). In
resting Jurkat T cells, myosin IIA was uniformly distributed in the
cytoplasm, whereas after stimulation with soluble OKT3, myosin IIA
and its phosphorylated MLCs rapidly became enriched at the area of
TCR clusters at the plasma membrane (Fig. 2b,c).
In synapses formed between Jurkat T cells and ‘superantigen’loaded
Raji B cells, we detected the typical accumulation of TCR,
F-actin and ezrin at the contact site as described before 26,27 .Inatwocell
system, either the T cell or B cell could contribute to this protein
redistribution, yet the results obtained with immune synapses between
Jurkat and Raji B cells were identical to results obtained with Jurkat
T cells stimulated with soluble OKT3 and were indicative of a
seemingly normal immune synapse (Figs. 2 and 3). The synapse
was also highly enriched for myosin IIA, with a distribution very
similar to that of the TCR (Fig. 3a). We obtained similar results with
primary human CD4 + T cells (Supplementary Fig. 2 online). The
recruitment of activated myosin IIA to the immunological synapse is
consistent with the observed function of myosin IIA in the movement
of TCR microclusters and cSMAC formation.
Figure 3 Effect of the inhibition of myosin IIA
activity on the formation of the immunological
synapse. (a,b) Microscopy of Jurkat T cells
pretreated for 10 min with DMSO (a) or50mM
blebbistatin (b), incubated for 5 min with SEE
superantigen–loaded B cells (prestained with
CMTPX; red), then fixed and stained for TCR,
ezrin, F-actin or myosin II heavy chain (green).
Numbers above images indicate percent cells
with similar protein distribution (n ¼ 30 cells per
condition). Scale bars, 5 mm. (c) Conjugate
formation by cells treated as described in a,b
(n ¼ 50 cells per condition). Data are representative
of three experiments (error bars (c), s.d.).
c
TCR Myosin HC Merge
TCR p-MLC Merge
Myosin HC p-MLC Merge
DMSO
Blebb
Figure 2 Phosphorylation and redistribution
of myosin IIA during activation of T cells.
(a) Abundance of phosphorylated MLC (p-MLC)
and total MLC (MLC) in total T cell lysates at
various times during stimulation (Stim) by soluble
anti-TCR (OKT3). (b) Microscopy of resting Jurkat
T cells fixed and stained for F-actin (green) and
myosin IIA heavy chain (HC; red). (c) Microscopy
of Jurkat T cells stimulated for 1 min with OKT3,
then fixed and stained for TCR, myosin heavy
chain and phosphorylated MLC. Cells showing
colocalization: 83%, TCR and myosin IIA heavy
chain; 92%, TCR and phosphorylated MLC; 90%,
myosin IIA heavy chain and phosphorylated MLC.
Scale bars (b,c), 5 mm. Data are representative of
three experiments (n ¼ 30 cells per condition).
Immunological synapse stability requires myosin IIA
To understand the consequences of myosin IIA activity for the
immunological synapse, we determined the effect of pretreating
Jurkat T cell with 50 mM blebbistatin on synapse formation with
‘superantigen’-loaded Raji B cells. Unexpectedly, we found that
inhibition of myosin II activity did not inhibit the concentration of
TCR, ezrin, F-actin or myosin IIA itself at the contact site between the
two cells (Fig. 3b). As we could not use siRNA in the Jurkat model, we
used both ML7 and an additional inhibitor, Y27632, which inhibits
Rho-associated kinase (ROCK). Both ROCK and MLCK phosphorylated
and activated MLCs, and both ML7 and Y27632 inhibited
phosphorylation of MLCs during T cell stimulation with soluble
anti-TCR (Supplementary Fig. 3 online). Conjugates between ‘superantigen’-loaded
Raji B cells and Jurkat T cells pretreated with either of
those drugs had apparently normal accumulation of myosin IIA
(Supplementary Fig. 4 online). We obtained similar results with
primary human CD4 + cells pretreated with blebbistatin and incubated
for 5 min with Raji B cells (Supplementary Fig. 2). These data
confirm and extend earlier indications that the first attachment step of
immunological synapse formation does not require myosin IIA
activity 6 . These results were also consistent with the ability of
blebbistatin-treated or myosin IIA–depleted cells to form immature
immunological synapses on planar bilayers containing TCR microclusters
and ICAM-1 (Fig. 1).
Although the conjugates formed after inhibition of myosin IIA
seemed normal, we found that fewer total conjugates formed with
T cells pretreated with 50 mM blebbistatin than with control cells
treated with DMSO (Fig. 3c). Conjugate formation was not further
decreased by pretreatment with 100 mM blebbistatin (data not shown),
which suggested that the residual conjugate formation was not simply
a TCR (89%) Ezrin (86%) F-actin (93%) Myosin HC (83%) c
b
TCR (92%) Ezrin (79%) F-actin (90%) Myosin HC (81%)
T cells in
conjugates (%)
ARTICLES
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 533
50
40
30
20
10
0
DMSO
Blebb

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
DMSO
(pretreat)
Blebb
(pretreat)
Blebb
(1 min)
Blebb
(12 min)
Before blebb After blebb
T
T
T
T
T
T
B
B
T
B
B B
T
T
T
B
B
an effect of partial inhibition of myosin IIA activity. To explore the
basis for the lower conjugate number, we examined the effect of
blebbistatin addition before and after conjugate formation. We first
immobilized ‘superantigen’-loaded Raji B cells in dishes with coverslip
inserts and then added Jurkat T cells. We monitored conjugate
formation and stability by differential interference contrast (DIC)
microscopy, adding blebbistatin at various times relative to conjugate
formation (Fig. 4a). The addition of blebbistatin resulted in lower
stability of formed conjugates so that only about 20% remained 2 min
after drug addition (Fig. 4b). Jurkat T cells treated with blebbistatin
formed unstable synapses that only lasted for 102 ± 14 s (Fig. 4c),
whereas control T cells formed stable synapses that persisted for over
20 min (data not shown). The addition of blebbistatin at various times
after conjugate formation resulted in instability and detachment within
1–2 min of drug addition, with an average time of 109 s (Fig. 4c).
As blebbistatin resulted in the same instability regardless of the time of
addition after synapse formation, we concluded that myosin IIA activity
is needed to maintain the stability of both early and mature
synapses. We obtained similar results by inhibiting myosin IIA activation
with 10 mM ML7(Supplementary Movie 8 online) and with
primary human CD4 + cells (Supplementary Movies 9,10 online).
B
b
Synapses (%)
100
80
60
40
20
0
170
150
130
110
90
70
50
30
Synapse duration (s) c
DMSO
Blebb
0 1 2 3 4 5 10 12
Blebb addition (min)
Figure 5 Effect of the inhibition of myosin IIA activity on intracellular Ca 2+
concentrations. (a) Microscopy of Jurkat T cells incubated with Fluo-LoJo,
then mixed with SEE superantigen–loaded B cells and allowed to form
immunological synapses, followed by the addition of DMSO or blebbistatin.
Scale bars, 5 mm. (b) Change in intensity over time of Fluo-LoJo in the cells
in a treated with DMSO (control) or blebbistatin at time 0 (n ¼ 3cells
per condition), presented relative to the average sustained signal in
‘superantigen’-activated cells, set as 100%. Numbers in parentheses (key)
identify different samples. (c) Emission ratios of Jurkat T cells incubated
with Fura-2AM, then added to a planar lipid bilayer containing anti-TCR and
ICAM-1 for 15 min before imaging, assessed every 15 s by fluorescence
microscopy and presented as the ratio of absorbance at 340 nm to
absorbance at 340 nm (340/380). Gray shaded area indicates the addition
of DMSO or blebbistatin 15 min after the addition of cells to bilayer (except
for pretreated cells, which did not receive additional blebbistatin). The low
and high calcium ratios corresponding to cells in EGTA with Mg 2+ and
without Ca 2+ or ionomycin were also determined (data not shown). Data are
representative of three experiments (a,b) or two independent experiments (c;
n ¼ 17 cells per condition).
Figure 4 Effect of the inhibition of myosin IIA activity on the stability of the
immunological synapse. (a) DIC imaging of conjugate formation and stability
of Jurkat T cells pretreated (pretreat) with DMSO or blebbistatin and added
to SEE superantigen–loaded B cells immobilized in dishes with coverslip
inserts, followed by the formation of immunological synapses (top two rows),
or of Jurkat T cells without pretreatment added to the B cells described
above, followed by the addition of blebbistatin 1 min or 12 min after synapse
formation (bottom two rows), assessed before (left) and 1–2 min after (right)
treatment with blebbistatin. T, T cell; B, B cell; white arrows, immunological
synapses; black arrows, loss of synapse. (b) Synapses in the cells in a
present 2 min after the addition of blebbistatin. (c) Duration of synapses in
the cells in a after the addition of blebbistatin at various times after synapse
formation (horizontal axis). Data are representative of five experiments (a),
or five experiments with 35 cells per condition (b,c; error bars, s.d.).
We next assessed whether synapse breakdown resulted from
perturbation of the typical accumulation of the adhesion proteins
LFA-1 and ICAM-1 at the peripheral SMAC 5 after inhibition of
myosin IIA activity with blebbistatin. Pretreatment with blebbistatin
led to a more peripheral distribution of these interactions, consistent
with impaired transport toward the center. However, we did not detect
a difference in the intensity of these interactions relative to that of
control cells treated with DMSO (Supplementary Fig. 5 online).
Thus, instability of the immunological synapse after the inhibition of
myosin IIA activity was not due to initial failure of LFA-1 activation.
Ca 2+ signaling requires myosin IIA activity
One of the earliest and most easily monitored signaling events after
T cell activation is a rapid increase in cytoplasmic Ca 2+ (ref. 28).
A published study has shown that treatment with butanedione
monoxide, a less specific inhibitor of myosin II than blebbistatin, in
activated primary CD4 + T cells leads to a less sustained increase in
Ca 2+ after stimulation and a partial blockade of the movement of
membrane proteins to the synapse 24 . To explore if synapse instability
correlated with loss of Ca 2+ signaling, we preloaded Jurkat T cells with
the fluorescent, cytoplasmic, Ca 2+ -sensitive indicator dye Fluo-LoJo
and assessed the effect of blebbistatin on cytoplasmic Ca 2+ in response
to ‘superantigen’-loaded Raji B cells. Although control Jurkat T cells
maintained higher cytoplasmic Ca 2+ concentrations (Fig. 5a,b), the
addition of blebbistatin (50 mM) to an existing immunological synapse
led to a rapid decrease in Ca 2+ concentrations within 1 min (Fig. 5a,b
a b
Time (s) DMSO Blebb
0
9
18
27
33
B
B
B
B
B
B
B
B
B
B
Intensity (%)
c
Fura-2 (340/380)
110
100
90
80
70
60
50
0 15 30 45
Time (s)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1 2 3 4 5 6
Time (min)
DMSO (1)
DMSO (2)
DMSO (3)
Blebb (1)
Blebb (2)
Blebb (3)
DMSO
Blebb
Blebb
(pretreat)
534 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a
0 min
1 min
5 min
b 700
1 min c
600
5 min
TCR cluster (nm)
500
400
Control
300
200
100
0
Blebb: None Before
stim
Blebb pretreat Blebb after TCR
After
stim
α-TCR:
Blebb:
p-Src
p-Lat
p-Zap70
Actin
and Supplementary Fig. 6b online). We obtained similar results with
ML-7 (Supplementary Movie 11 and Supplementary Fig. 7 online).
We detected a similar decrease in Ca 2+ concentrations in primary
human CD4 + cells after inhibition of myosin IIA (Supplementary
Movies 12–14 online). For a more quantitative measurement of
cytoplasmic Ca 2+ changes, we loaded Jurkat T cells with the ratiometric
Ca 2+ -indicator dye Fura-2AM and calculated emission ratios.
The addition of 50 mM blebbistatin to cells with preformed synapses
diminished cytoplasmic Ca 2+ concentrations to baseline within less
than 2 min, whereas control cells maintained high Ca 2+ concentrations
(Fig. 5c). Pretreatment with 50 mM blebbistatin blocked the
TCR-inducedincreaseinCa 2+ altogether (Fig. 5c). To rule out the
possibility that emission-intensity changes resulted from autofluorescence
of blebbistatin, we preloaded T cells with Fluo-LoJo and added
blebbistatin without any TCR stimulation. Blebbistatin fluorescence
was negligible in our assays (Supplementary Fig. 6). Moreover, we
found that the addition of 50 mM blebbistatin to the cells, followed by
illumination, had no toxic effect (data not shown). Notably, in all
these experiments, the decrease in cytoplasmic Ca 2+ concentration
preceded the detachment of the immunological synapse, which
indicated that myosin IIA activity is necessary for sustained Ca 2+
signaling ‘downstream’ of the TCR in the immunological synapse
independently of any effects on adhesion.
The serial-triggering model holds that one major histocompatibility
complex–bound antigenic peptide engages a large number of TCRs in
successive rounds, contacting about 50–200 receptors per antigenic
peptide 29 . This model is compatible with the demonstration that ten
complexes of peptide and major histocompatibility complex in the
T cell–APC interface can sustain signaling long enough to generate
interleukin 2 (ref. 30). If myosin IIA is needed only to promote an
active process of serial triggering, then increasing the number of TCRs
triggered in parallel might overcome the requirement for myosin IIA
activity. To test that possibility, we explored whether more activating
anti-TCR could overcome the effect of blebbistatin on Ca 2+ signaling.
We preloaded Jurkat T cells with Fluo-LoJo and then stimulated the
cells with increasing concentrations of anti-TCR. Once the cytoplasmic
Ca 2+ concentrations had risen, we added 50 mM blebbistatinand
monitored Ca 2+ concentrations for an additional 1 min. The decrease
in cytoplasmic Ca 2+ concentration was independent of the concentration
of activating antibody, with a similar decrease in cells stimulated
–
–
+
–
+
+
ARTICLES
Figure 6 Effect of the inhibition of myosin IIA activity on TCR microclusters.
(a) Microscopy of Jurkat T cells stimulated with Alexa Fluor 488–conjugated
anti-CD3 without additional treatment (Control), after pretreatment for
10 min with 50 mM blebbistatin (Blebb pretreat) or treatment with 50 mM
blebbistatin after TCR stimulation (Blebb after TCR), then imaged
immediately (0 min) or at 1 and 5 min after stimulation. Scale bars, 5 mm.
Data are representative of five experiments with two cells per condition per
time point. (b) Quantitative analysis of the results in a. Dataare
representative of five experiments (error bars, s.e.m.; n ¼ 35 clusters per
bar). (c) Immunoblot analysis of Src phosphorylated at tyrosine 416 (p-Src),
Zap70 phosphorylated at tyrosine 319 (p-Zap70) and Lat phosphorylated at
tyrosine 191 (p-Lat), in Jurkat T cells left untreated (a-TCR ) or treated for
2 min with anti-CD3 (a-TCR +; OKT3) with (+) or without ( ) blebbistatin
pretreatment. Bottom, actin protein abundance (loading control). Results
are representative of three independent experiments.
with between 10 mg/ml and 500 mg/ml of antibody (Supplementary
Fig. 8 online). This result challenges the possibility of insufficient
TCR engagement as a mechanism to account for the decrease in
Ca 2+ signaling.
TCR signaling requires myosin IIA activity
Our results suggested that myosin IIA might be important for the
function of the TCR signalosome. The simplest way to activate the
formation of TCR signalosomes is based on the addition of soluble
anti-CD3e to Jurkat T cells 31 , shown above to activate MLC phosphorylation.
We incubated Jurkat T cells with fluorescence-tagged
anti-CD3e and monitored TCR distribution and biochemical indicators
of TCR signalosome assembly (phosphorylation of Lck, Zap70
and Lat). Control Jurkat T cells initially showed a uniform surface
fluorescence that aggregated into microclusters with a diameter of
280 ± 70 nm by 1 min, followed by coalescence into larger clusters of
456 ± 88 nm after 5 min of stimulation (Fig. 6a,b). When we
pretreated Jurkat T cells for 5 min with 50 mM blebbistatin and
then stimulated the cells for 1 min with labeled anti-TCR, the TCR
clusters were slightly smaller, with a diameter of 217 ± 63 nm.
However, progression in cluster size in the blebbistatin-treated cells
was minimal, reaching a diameter of 247 ± 66 nm after 5 min of
stimulation (Fig. 6a,b). We next explored the effect on microclusters
when we added blebbistatin at 1 or 5 min after stimulation. In both
cases, 5 min after the addition of blebbistatin, the cluster was smaller,
with a diameter of 217 ± 64 nm or 258 ± 59 nm for 1 min or 5 min,
respectively (Fig. 6a,b). These results collectively show that TCR
microclusters about 217 nm in diameter can form in the absence of
myosin IIA activity, yet their coalescence into larger clusters and their
maintenance in larger clusters require myosin IIA activity. We
obtained similar results with primary human CD4 + cells (Supplementary
Fig. 2). When we treated Jurkat T cells for 2 min with anti-
CD3e and then analyzed phosphorylated signalosome components by
direct immunoblot of lysates, we found that phosphorylation of Src
kinases, which probably included phosphorylated Lck, was similar
with or without blebbistatin pretreatment (Fig. 6c). In contrast,
phosphorylation of Zap70 and Lat was substantially lower after
blebbistatin pretreatment (Fig. 6c). We obtained similar results with
primary CD4 + T cells (Supplementary Fig. 2). We also assessed
whether Jurkat T cells pretreated with blebbistatin increased their
Ca 2+ in response to stimulation with soluble anti-CD3e. T cells
preloaded with Fluo-LoJo and stimulated with soluble anti-TCR
underwent a robust Ca 2+ response, whereas cells pretreated with
blebbistatin failed to increase Ca 2+ concentrations in response
to stimulation (Fig. 5c and Supplementary Fig. 8). These results
indicate quantitative defects in TCR microcluster size and defective
signalosome function in a synapse-free assay.
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 535

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a CD3 Src (pY416) Merge
b
DMSO
Blebb
DMSO
Blebb
DMSO
Blebb
CD3
CD3
Zap70 (pY319) Merge
Lat (pY191) Merge
Src (pY416)
intensity (%)
Zap70 (pY319)
intensity (%)
Lat (pY191)
intensity (%)
120
100
80
TCR signalosome function can also be evaluated in a synapse-based
system with supported planar bilayers presenting OKT3 (ref. 17).
T cells interacting with a planar bilayer containing OKT3 and ICAM-1
for 5 min had a central condensed TCR cluster surrounded by
peripheral microclusters containing TCRs, as well as phosphorylated
Zap70 and Lat (Fig. 7), similar to published studies 9 .Whenweadded
Jurkat T cells pretreated with 50 mM blebbistatin to the bilayers,
followed by staining with a specific antibody to each phosphorylated
protein, we found that phosphorylated Src was localized together with
TCR microclusters, but the abundance of phosphorylated Zap70 and
Lat, as measured by fluorescence intensity, was significantly lower, by
80% each (P o 0.0001; Fig. 7). We also extended this analysis to
primary CD4 + T cells treated with control and MyH9-specific siRNA
during activation; this resulted in nearly complete knockdown of
myosin IIA in one third of the cells. We found that knockdown
of myosin IIA diminished phosphorylation of Src by only 25%
(P o 0.0001) but lowered phosphorylation of Zap70 at tyrosine 319
by 80% (P o 0.0001) and phosphorylation of Lat by 70%
(P o 0.0001). These data demonstrate, by both pharmacological and
reverse-genetic approaches, that myosin IIA is required for amplification
of TCR signaling between the Lck and Zap70 activation steps.
DISCUSSION
Here we have provided evidence that myosin IIA is central to synapse
assembly and signaling, being necessary for TCR signaling, the
centripetal motion and fusion of microclusters during the formation
60
40
20 0
120
100
80
60
40
20 0
120
100
80
60
40
20 0
DMSO
Blebb
DMSO
Blebb
DMSO
Blebb
Control
siRNA
Myosin IIA
siRNA
Control
siRNA
Myosin IIA
siRNA
Control
siRNA
Myosin IIA
siRNA
CD3
Src (pY416) Merge
CD3 Zap70 (pY319) Merge
CD3 LatT (pY191) Merge
Src (pY416)
intensity (%)
Zap70 (pY319)
intensity (%)
Lat (pY191)
intensity (%)
120
100
80
of the immunological synapse, and synapse persistence. Published
work has shown that the F-actin cytoskeleton is required for all of
these processes 17,32 and that TCR engagement induces actin polymerization
through recruitment of the adaptor protein Nck and Wiskott-
Aldrich syndrome protein to the TCR microclusters 33 . Our study has
shown that after T cell engagement, myosin IIA was activated by
phosphorylation of MLCs and its activity was necessary for proper
assembly of the signalosome. Inhibition of myosin IIA activity with
the highly specific myosin II inhibitor blebbistatin or depletion of
myosin IIA expression with specific siRNA resulted in a complete
halting of directed motion of microclusters, prevented the formation
of the cSMAC and prevented amplification of TCR signals after Lck
activation. Whether myosin IIA activity was inhibited pharmacologically,
in which case myosin IIA was still recruited to the synapse, or if
its expression was diminished by siRNA, in which case it was
profoundly depleted from the synapse, the formation of initial small
TCR microclusters remained intact. However, these clusters did not
increase in size, did not fully signal and did not undergo directed
translocation. Thus, we have defined distinct F-actin-dependent and
actomyosin-dependent phases of T cell activation and immunological
synapse formation.
The potential involvement of myosin II in the formation of
immunological synapses has been reported before. One study showed
that movement of ICAM-1-coated beads on T cells after activation by
a B cell is inhibited by butanedione monoxime with concurrent
decrease in Ca 2+ signaling, although the B cell–T cell conjugates
60
40
20
0
Control Myosin IIA
siRNA siRNA
120
100
80
60
40
20
0
Control Myosin IIA
siRNA siRNA
120
100
80
60
40
20
0
Control Myosin IIA
siRNA siRNA
Figure 7 Effect of the inhibition or depletion of myosin IIA on signaling in T cells. (a) Localization of TCR microclusters and phosphorylated Src, Zap70
and Lat (left) in Jurkat T cells (with or without blebbistatin pretreatment; left margin) added for 25 min to a planar lipid bilayer containing Alexa Fluor
568–labeled anti-TCR and ICAM-1, then fixed and stained with antibody to Src phosphorylated at tyrosine 416 (Src(pY416)), Zap70 phosphorylated at
tyrosine 319 (Zap70(pY319)) or Lat phosphorylated at tyrosine 191 (Lat(pY191)). Right, quantification of protein phosphorylation. Data are representative of
three experiments (n ¼ 15 cells per bar; error bars, s.d.). (b) Localization of TCR microclusters and phosphorylated Src, Zap70 and Lat (left) in primary
human CD4 + cells treated with control or MYH9-specific siRNA and added for 25 min to a planar lipid bilayer containing Alexa Fluor 568–labeled anti-TCR
and ICAM-1, then fixed and stained as described in a. Myosin IIA–depleted cells were identified by the lack of central TCR clustering (as in Fig. 6b). Right,
quantification of protein phosphorylation. Data are representative of two experiments (n ¼ 15 cells per bar; error bars, s.d.).
536 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
remain stable 24 . The authors hypothesized that myosin II–mediated
transport is delivering components to the immunological synapse that
are needed for sustained signaling. Another study showed that myosin
IIA is necessary for T cell motility and uropod maintenance and
postulated that inhibition of myosin IIA filament formation is
required for the T cell ‘stop signal’ after antigen encounter 6 .These
authors also reported that immunological synapse formation seemed
unaffected by pretreatment with blebbistatin. That result is in agreement
with our finding that immunological synapses formed with
blebbistatin-treated T cells were initially similar to synapses with
control cells. The T cell blasts used in the earlier study 6 have high
constitutive LFA-1 activity, so myosin IIA–dependent signaling is not
required for conjugate formation. We focused on two systems here,
Jurkat T cells and primary human T cells, in which basal LFA-1
activity is low and ‘inside-out’ signaling through the TCR is required
for conjugate formation 34 . In retrospect, evidence of spreading and
contraction in the immunological synapse formation process was
visible in earlier studies 5,9 and has been explicitly described for
B cell synapse formation without indicating involvement of myosin
II (ref. 35). Contractile oscillations have been reported at the outer
edge of the immunological synapse formed by T cells 32 .Contractile
oscillations require myosin IIA in fibroblasts. Our results suggest that
this is also probably true in lymphocytes 36 .
Myosin II–based cortical movement has been documented in
several other situations. Myosin II is necessary for cortical tension
and functions in the contractile ring during cytokinesis 37,38 .Several
studies have suggested that an imbalance in cortical tension contributes
to cytokinesis, with cortical loosening at the cell poles and
enhanced tension at the cell equator, leading to equatorial movement,
assembly and contraction of the contractile ring 39 . In a related
mechanism, anterior-posterior polarity in the one-cell nematode
embryo is established by myosin II–mediated cortical contraction,
which moves granules and fate determinants toward the future
anterior pole 40 . It is possible that a related myosin II–dependent
cortical tension may move TCR microclusters toward the center of the
immunological synapse. Such cortical tension seemed to be required
for TCR signalosome function even in the absence of a synapse, on the
basis of results obtained with soluble OKT3. The reported particle-size
requirements for T cell stimulation may arise from the need for
myosin IIA–mediated tension across an interface or crosslinked
protein network 41,42 . Myosin IIA–mediated cortical tension may be
required for the rearrangement of cytoskeletally associated ‘protein
islands’ into functional signalosomes 43 .
The activation of myosin II by phosphorylation of its MLCs can
be mediated by several different kinases, including the calciumcalmodulin–dependent
MLCK 44 , ROCK and protein kinase C 45 .
Shortly after stimulation of T cells, Vav1, a Rho guanine-exchange
factor, is recruited to TCR microclusters through interaction with the
adaptor protein SLP-76, which is then followed by the recruitment of
the GTP-binding protein Cdc42 and ROCK 46,47 . T cell stimulation also
results in more cytoplasmic Ca 2+ , which is known to activate MLCK 44 .
We have shown that treatment with either the ROCK inhibitor Y27632
or the MLCK inhibitor ML-7 inhibited phosphorylation of MLCs after
T cell stimulation. Thus, both kinases take part in activating myosin II
even when the TCR is triggered by OKT3. As myosin IIA activity was
necessary to maintain higher Ca 2+ concentrations, a plausible model is
that Rho-GTP–activated ROCK initially phosphorylates MLCs, then
Ca 2+ concentrations increase, which maintains light-chain phosphorylation
through persistent activation of MLCK. Thus, one crucial function
of myosin IIA activity is to maintain signaling that then feeds back
to maintain higher Ca 2+ concentrations and active myosin IIA.
ARTICLES
This is the first report to our knowledge to link myosin II activity to
signaling through an immunoreceptor. In examining the ‘downstream’
signaling pathway, we found that phosphorylation of the Src family
kinases was unimpaired by either inhibition or depletion of myosin
IIA, whereas ‘downstream’ signaling, including phosphorylation of
Zap70 and Lat and an increase in cytosolic Ca 2+ , were much more
dependent on myosin IIA activity. The truncation in signaling ‘downstream’
of Lck was not due to defects in adhesion, as inhibition of
myosin IIA activity in Jurkat T cells stimulated with soluble OKT3 also
resulted in less phosphorylation of Zap70 and Lat and a decrease in
intracellular Ca 2+ concentrations to baseline. Our data support a twostep
model in which initial conjugate formation involving the formation
of TCR microclusters, recruitment of myosin IIA and activation
of Lck are all independent of myosin IIA activity, whereas amplification
of signaling and microcluster movement are dependent on
myosin IIA activity. Our work here and published work indicates
that there is careful ‘tuning’ of myosin IIA activity during T cell
activation, with negative regulation through inhibition of the formation
of thick filaments 6 and positive regulation through phosphorylation
of MLCs, which leads to maintenance of the cortical tension
needed for TCR signaling and synapse stabilization.
METHODS
Cells and antibodies. Jurkat T cells and Raji B cells were from American Type
Culture Collection. Human peripheral blood lymphocytes were isolated from
citrate-anticoagulated whole blood by dextran sedimentation (Blood Centers
of America–hemerica), followed by density separation over Ficoll-Hypaque
(Sigma). The resulting mononuclear cells were washed in PBS and were further
purified by nylon wool and plastic adherence as described48 . Human peripheral
CD4 + blasts were prepared as described49 . Anti-ezrin and anti–myosin II heavy
chain have been described50 . Affinity-purified polyclonal antibodies to MLC
phosphorylated at serine 19 (3671), Src phosphorylated at tyrosine 416 (used to
measure phosphorylated Lck, the most abundant Src member in T cells; 2101),
Zap 70 phosphorylated at tyrosine 319 (2701), and Lat phosphorylated at tyrosine
191 (3584) were from Cell Signaling. OKT3 mouse anti–human CD3 was
purified from an OKT3 hybridoma cell line (14-0037; eBioscience). Rhodaminephalloidin,
Alexa Fluor 568–phalloidin, Alexa Fluor 488–conjugated donkey
anti-mouse (21202) and Alexa Fluor 568–conjugated goat anti-rabbit (11011)
and goat anti-mouse (11004) were from Invitrogen. Horseradish peroxidase–
conjugated goat anti-rabbit (W4011) was from Promega. All procedures were
approved by the Health and Safety Committee of Cornell University.
Immunofluorescence. Cells were plated on poly-L-lysine-coated glass slides,
were fixed for 30 min at 25 1C with 3.7% (wt/vol) formaldehyde, were made
permeable by treatment for 2 min with 0.1% (vol/vol) Triton X-100 in PBS and
then were rinsed three times in PBS. Cells were then incubated for 1 h with 5%
(vol/vol) BSA in PBS, then were incubated for 1 h with primary antibody in 5%
(vol/vol) BSA in PBS, washed in PBS and incubated for 1 h with the
appropriate secondary antibody (or phalloidin) in 5% (vol/vol) BSA in PBS.
After additional washes, 5 ml Vectashield (Vector Labs) was added to cells and
slides were covered with coverslips. Cells were viewed with a Nikon Eclipse
TE2000-U (100 objective; numerical aperture, 1.4) with the Perkin Elmer
UltraVIEW LCI spinning-disk confocal imaging system and a Hamamatsu
12-bit C4742-95 digital charge-coupled device camera.
Immunoblot analysis. Jurkat and primary T cells were lysed and were resolved
by SDS-PAGE, followed by transfer to polyvinylidene difluoride membranes
(Immobilon-P; Millipore) with a semidry transfer system (Integrated Separation
Systems). After 1 h of blocking in 5% (wt/vol) milk in Tris-buffered saline
with Tween, membranes were incubated for 1 h with primary antibody, then
were washed and were incubated for 1 h with the appropriate horseradish
peroxidase–conjugated secondary antibody. Blots were developed by the
enhanced chemiluminescence system (Amersham).
Cell stimulation and conjugate formation. Jurkat and primary human T cells
were activated for various times with the antibody OKT3 (10 mg/ml). For
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 537

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
stimulation with B cells, Raji B cells were labeled with the fluorescent dye
CellTracker Red (CMTPX; Molecular Probes) and were loaded with the
‘superantigen’ staphylococcus enterotoxin E (SEE; 2 mg/ml; Toxin Technology).
Equal numbers of T cells and B cells were incubated together.
Conjugate stability and DIC microscopy. Raji B cells were loaded with SEE
superantigen and then were immobilized in dishes containing coverslip inserts
(MatTek) and viewed on an Axiovert 100 TV microscope (Carl Zeiss) equipped
with a charge-coupled device (C4742-95-12ERG; Hamamatsu) with a DIC
prism and Openlab 4.0 software (Improvision). After initial B cell imaging,
Jurkat T cells or primary human T cells were added to plates and cells were
allowed to form conjugates. Blebbistatin (50 mM), ML7 (10 mM) or DMSO was
added at various times and conjugates were continuously imaged. Movies were
analyzed with ImageJ software.
Ca 2+ assays. For nonratiometric assays, Jurkat T cells and primary human
T cells were loaded with 1 mM Fluo-LoJo (TefLabs), then were added to
SEE-loaded Raji B cells and allowed to form synapses or were stimulated with
OKT3. Blebbistatin, ML7 or DMSO was added at various times and fluorescence
intensity was measured with the spinning-disk confocal imaging system.
For ratiometric analysis, Jurkat T cells were loaded with 2.5 mM Fura-2AM
(acetoxymethyl ester; Molecular Probes) as described 13 .
Bilayer assembly and TIRF microscopy. Glass-supported bilayers of dioleoyl
phosphatidylcholine incorporating 0.01% (wt/vol) 1,2-dioleoyl-sn-glycero-3phosphatidylcholine,
1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(cap
biotinyl) were prepared in flow chambers (Bioptechs) as described 5 . Bilayers
were loaded with monobiotinylated-564-OKT3. Cells were allowed to settle and
form contacts with the bilayer before imaging. An Olympus inverted IX-70
microscope equipped with Hamamatsu 12-bit C9100 1.1B charge-coupled
device and a TIRF objective from Olympus were used for all bilayer imaging.
Microclusters were analyzed with Volocity 4.2 (Improvision).
Transfection with siRNA. CD4 + human T cell blasts (3 10 6 )atday4after
isolation were transfected by electroporation with Amaxa nucleofector technology
according to the manufacturer’s instructions. Two siRNA duplexes specific
for human MYH9 and negative control siRNA were used (Dharmacon). Cells
were cultured for 48 h and were analyzed by immunoblot or immunofluorescence.
Suppression of the target protein was verified by immunoblot.
Statistical analysis. Prism software was used for nonparametric t-tests.
Accession codes UCSD-Nature Signaling Gateway (http://www.signalinggateway.org):
A001394, A002396, A001392 and A004003.
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank D. Garbett for help with data analysis with Volocity and for
comments, and D.W. Pruyne for help in setting up DIC microscopy. Supported
by the European Molecular Biology Organization (T.I.) and the US National
Institutes of Health (GM36652 to A.B.; and AI44931 and Nanomedicine
Development Center EY16586 to M.L.D.).
AUTHOR CONTRIBUTIONS
The laboratories of A.B. and M.L.D. did independent work on the involvement of
myosin IIA in the formation of immunological synapses and continued the work
collaboratively focusing on studies in the human system initiated by T.I. and A.B.;
T.I. conceived and did the experiments in Figures 2–4, 5a,b and 6; T.I., G.V.-S.
and S.V. collaborated on Figures 1, 5c and 7; and T.I. and A.B. wrote the first
draft of the manuscript, which M.L.D. extensively edited and revised.
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Davis, M.M. The ab T cell repertoire comes into focus. Immunity 27, 179–180 (2007).
2. DeMond, A.L., Mossman, K.D., Starr, T., Dustin, M.L. & Groves, J.T. T cell receptor
microcluster transport through molecular mazes reveals mechanism of translocation.
Biophys. J. 94, 3286–3292 (2008).
3. Monks, C.R., Freiberg, B.A., Kupfer, H., Sciaky, N. & Kupfer, A. Three-dimensional
segregation of supramolecular activation clusters in T cells. Nature 395, 82–86
(1998).
4. Dustin, M.L. et al. A novel adaptor protein orchestrates receptor patterning and
cytoskeletal polarity in T-cell contacts. Cell 94, 667–677 (1998).
5. Grakoui, A. et al. The immunological synapse: a molecular machine controlling T cell
activation. Science 285, 221–227 (1999).
6. Jacobelli, J., Chmura, S.A., Buxton, D.B., Davis, M.M. & Krummel, M.F. A single class
II myosin modulates T cell motility and stopping, but not synapse formation. Nat.
Immunol. 5, 531–538 (2004).
7. Combs, J. et al. Recruitment of dynein to the Jurkat immunological synapse. Proc. Natl.
Acad. Sci. USA 103, 14883–14888 (2006).
8. Bunnell, S.C. et al. T cell receptor ligation induces the formation of dynamically
regulated signaling assemblies. J. Cell Biol. 158, 1263–1275 (2002).
9. Campi, G., Varma, R. & Dustin, M.L. Actin and agonist MHC-peptide complexdependent
T cell receptor microclusters as scaffolds for signaling. J. Exp. Med. 202,
1031–1036 (2005).
10. Huse, M. et al. Spatial and temporal dynamics of T cell receptor signaling with a
photoactivatable agonist. Immunity 27, 76–88 (2007).
11. Yokosuka, T. et al. Newly generated T cell receptor microclusters initiate and sustain T
cell activation by recruitment of Zap70 and SLP-76. Nat. Immunol. 6, 1253–1262
(2005).
12. Douglass, A.D. & Vale, R.D. Single-molecule microscopy reveals plasma membrane
microdomains created by protein-protein networks that exclude or trap signaling
molecules in T cells. Cell 121, 937–950 (2005).
13. Varma, R., Campi, G., Yokosuka, T., Saito, T. & Dustin, M.L. T cell receptor-proximal
signals are sustained in peripheral microclusters and terminated in the central
supramolecular activation cluster. Immunity 25, 117–127 (2006).
14. Lee, K.H. et al. The immunological synapse balances T cell receptor signaling and
degradation. Science 302, 1218–1222 (2003).
15. Dustin, M.L. & Cooper, J.A. The immunological synapse and the actin cytoskeleton:
molecular hardware for T cell signaling. Nat. Immunol. 1, 23–29 (2000).
16. Chakraborty, A.K. How and why does the immunological synapse form? Physical
chemistry meets cell biology. Sci. STKE 2002, PE10 (2002).
17. Kaizuka, Y., Douglass, A.D., Varma, R., Dustin, M.L. & Vale, R.D. Mechanisms for
segregating T cell receptor and adhesion molecules during immunological synapse
formation in Jurkat T cells. Proc. Natl. Acad. Sci. USA 104, 20296–20301 (2007).
18. Lin, C.H., Espreafico, E.M., Mooseker, M.S. & Forscher, P. Myosin drives retrograde Factin
flow in neuronal growth cones. Neuron 16, 769–782 (1996).
19. Conti, M.A. & Adelstein, R.S. Nonmuscle myosin II moves in new directions. J. Cell Sci.
121, 11–18 (2008).
20. Tan, J.L., Ravid, S. & Spudich, J.A. Control of nonmuscle myosins by phosphorylation.
Annu. Rev. Biochem. 61, 721–759 (1992).
21. Simons, M. et al. Human nonmuscle myosin heavy chains are encoded by two genes
located on different chromosomes. Circ. Res. 69, 530–539 (1991).
22. Golomb, E. et al. Identification and characterization of nonmuscle myosin II–C,
a new member of the myosin II family. J. Biol. Chem. 279, 2800–2808 (2004).
23. Maupin, P., Phillips, C.L., Adelstein, R.S. & Pollard, T.D. Differential localization of
myosin-II isozymes in human cultured cells and blood cells. J. Cell Sci. 107,
3077–3090 (1994).
24. Wulfing, C. & Davis, M.M. A receptor/cytoskeletal movement triggered by costimulation
during T cell activation. Science 282, 2266–2269 (1998).
25. Straight, A.F. et al. Dissecting temporal and spatial control of cytokinesis with a myosin
II Inhibitor. Science 299, 1743–1747 (2003).
26. Blanchard, N. et al. Strong and durable TCR clustering at the T/dendritic cell immune
synapse is not required for NFAT activation and IFN-g production in human CD4 +
Tcells.J. Immunol. 173, 3062–3072 (2004).
27. Ilani, T., Khanna, C., Zhou, M., Veenstra, T.D. & Bretscher, A. Immune synapse
formation requires ZAP-70 recruitment by ezrin and CD43 removal by moesin.
J. Cell Biol. 179, 733–746 (2007).
28. Weiss, A., Imboden, J., Shoback, D. & Stobo, J. Role of T3 surface molecules in human
T-cell activation: T3-dependent activation results in an increase in cytoplasmic free
calcium. Proc. Natl. Acad. Sci. USA 81, 4169–4173 (1984).
29. Valitutti, S., Muller, S., Cella, M., Padovan, E. & Lanzavecchia, A. Serial triggering of
many T-cell receptors by a few peptide-MHC complexes. Nature 375, 148–151
(1995).
30. Krogsgaard, M. et al. Agonist/endogenous peptide-MHC heterodimers drive T cell
activation and sensitivity. Nature 434, 238–243 (2005).
31. Janeway, C.A. Jr. & Bottomly, K. Responses of T cells to ligands for the T-cell receptor.
Semin. Immunol. 8, 108–115 (1996).
32. Sims, T.N. et al. Opposing effects of PKCy andWASponsymmetrybreakingand
relocation of the immunological synapse. Cell 129, 773–785 (2007).
33. Barda-Saad, M. et al. Dynamic molecular interactions linking the T cell antigen
receptor to the actin cytoskeleton. Nat. Immunol. 6, 80–89 (2005).
34. Dustin, M.L. & Springer, T.A. T-cell receptor cross-linking transiently stimulates
adhesiveness through LFA-1. Nature 341, 619–624 (1989).
35. Fleire, S.J. et al. B cell ligand discrimination through a spreading and contraction
response. Science 312, 738–741 (2006).
36. Giannone, G. et al. Lamellipodial actin mechanically links myosin activity with
adhesion-site formation. Cell 128, 561–575 (2007).
37. Pasternak, C., Spudich, J.A. & Elson, E.L. Capping of surface receptors and concomitant
cortical tension are generated by conventional myosin. Nature 341, 549–551
(1989).
538 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
38. Mabuchi, I. & Okuno, M. The effect of myosin antibody on the division of starfish
blastomeres. J. Cell Biol. 74, 251–263 (1977).
39. Matsumura, F. Regulation of myosin II during cytokinesis in higher eukaryotes. Trends
Cell Biol. 15, 371–377 (2005).
40. Munro, E., Nance, J. & Priess, J.R. Cortical flows powered by asymmetrical contraction
transport PAR proteins to establish and maintain anterior-posterior polarity in the early
C. elegans embryo. Dev. Cell 7, 413–424 (2004).
41. Mescher, M.F. Surface contact requirements for activation of cytotoxic T lymphocytes.
J. Immunol. 149, 2402–2405 (1992).
42. Galbraith, C.G., Yamada, K.M. & Sheetz, M.P. The relationship between force and focal
complex development. J. Cell Biol. 159, 695–705 (2002).
43. Lillemeier, B.F., Pfeiffer, J.R., Surviladze, Z., Wilson, B.S. & Davis, M.M. Plasma
membrane-associated proteins are clustered into islands attached to the cytoskeleton.
Proc. Natl. Acad. Sci. USA 103, 18992–18997 (2006).
44. Gallagher, P.J., Herring, B.P., Griffin, S.A. & Stull, J.T. Molecular characterization of a
mammalian smooth muscle myosin light chain kinase. J. Biol. Chem. 266,
23936–23944 (1991).
ARTICLES
45. Ludowyke, R.I., Peleg, I., Beaven, M.A. & Adelstein, R.S. Antigen-induced secretion of
histamine and the phosphorylation of myosin by protein kinase C in rat basophilic
leukemia cells. J. Biol. Chem. 264, 12492–12501 (1989).
46. Koretzky, G.A., Abtahian, F. & Silverman, M.A. SLP76 and SLP65: complex regulation
of signalling in lymphocytes and beyond. Nat. Rev. Immunol. 6, 67–78 (2006).
47. Zeng, R. et al. SLP-76 coordinates Nck-dependent Wiskott-Aldrich syndrome protein
recruitment with Vav-1/Cdc42-dependent Wiskott-Aldrich syndrome protein activation
at the T cell-APC contact site. J. Immunol. 171, 1360–1368 (2003).
48. Dustin, M.L. & Springer, T.A. Lymphocyte function-associated antigen-1 (LFA-1)
interaction with intercellular adhesion molecule-1 (ICAM-1) is one of at least three
mechanisms for lymphocyte adhesion to cultured endothelial cells. J. Cell Biol. 107,
321–331 (1988).
49. Vasiliver-Shamis, G. et al. HIV-1 envelope gp120 induces a stop signal and virological
synapse formation in non-infected CD4 + Tcells.J. Virol. 82, 9445–9457 (2008).
50. Bretscher, A. Rapid phosphorylation and reorganization of ezrin and spectrin accompany
morphological changes induced in A-431 cells by epidermal growth factor. J. Cell
Biol. 108, 921–930 (1989).
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 539

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
HoxC4 binds to the promoter of the cytidine
deaminase AID gene to induce AID expression, classswitch
DNA recombination and somatic hypermutation
Seok-Rae Park 1,2 , Hong Zan 1,2 , Zsuzsanna Pal 1 , Jinsong Zhang 1 , Ahmed Al-Qahtani 1 , Egest J Pone 1 ,
Zhenming Xu 1 , Thach Mai 1 & Paolo Casali 1
The cytidine deaminase AID (encoded by Aicda in mice and AICDA in humans) is critical for immunoglobulin class-switch
recombination (CSR) and somatic hypermutation (SHM). Here we show that AID expression was induced by the HoxC4
homeodomain transcription factor, which bound to a highly conserved HoxC4-Oct site in the Aicda or AICDA promoter. This site
functioned in synergy with a conserved binding site for the transcription factors Sp1, Sp3 and NF-jB. HoxC4 was ‘preferentially’
expressed in germinal center B cells and was upregulated by engagement of CD40 by CD154, as well as by lipopolysaccharide
and interleukin 4. HoxC4 deficiency resulted in impaired CSR and SHM because of lower AID expression and not some other
putative HoxC4-dependent activity. Enforced expression of AID in Hoxc4 –/– B cells fully restored CSR. Thus, HoxC4 directly
activates the Aicda promoter, thereby inducing AID expression, CSR and SHM.
Class-switch recombination (CSR) and somatic hypermutation
(SHM) are critical for the maturation of antibody responses to foreign
and self antigens. CSR recombines switch-region DNA located
upstream of exons of the constant heavy-chain (C H) region, thereby
altering immunoglobulin CH regions and endowing antibodies with
new biological effector functions. SHM introduces mainly point
mutations in immunoglobulin variable regions, thereby providing
the structural substrate for the selection by antigen of antibody
mutants with higher affinity. Despite advances in the identification
of some factors involved in CSR and SHM, details of the mechanisms
of these processes remain elusive. Both CSR and SHM require
activation-induced cytidine deaminase (AID), which is expressed by
activated B cells, mainly in germinal centers of peripheral lymphoid
organs 1,2 . AID initiates CSR and SHM by deaminating deoxycytosine
residues to yield deoxyuracil-deoxyguanine mispairings in
DNA 3–8 . These mispairings trigger DNA-repair processes, which
entails the introduction of mutations in variable-(diversity)-joining
(V(D)J) regions or DNA breaks, including double-stranded DNA
breaks, which lead to nonclassical nonhomologous end joining
and CSR 3,5,9–14 .
The mechanisms governing the transcriptional regulation of the
gene encoding AID (AICDA in humans and Aicda in mice) remain to
be elucidated. A conserved region in the first intron of Aicda containing
two E-boxes, the consensus sequence for binding of the transcription
factor E2A (A000804), has been suggested to contribute to the
regulation of Aicda transcription through recruitment of the E2A
Received 17 December 2008; accepted 9 March 2009; published online 12 April 2009; doi:10.1038/ni.1725
helix-loop-helix transcription factor E47 and the inhibitor of DNA
binding helix-loop-helix protein Id3, respectively 15 . The B cell lineage–
specific transcription factor Pax5 is thought to act with E2A proteins
in controlling Aicda transcription 16 . However, this was not confirmed
by another study, which has instead suggested involvement of the Sp1
family of ubiquitous zinc-finger transcription factors. These regulate
various promoters by binding to deoxyguanine-deoxycytosine,
deoxyguanine-deoxyadenine or deoxyguanine-deoxythymine boxes in
activating the Aicda promoter 17 .
Hox proteins are highly conserved helix-loop-helix homeodomaincontaining
transcription factors that regulate cellular differentiation
and organogenesis 18,19 . Genes encoding these proteins, which are
clustered chromosomally, are expressed in a temporally and spatially
regulated way 20,21 . Among the human genes, those encoding the
HoxC proteins, particularly HOXC4, are expressed mainly in lymphoid
cells 22 . HOXC4 expression increases through sequential stages
of B cell development 22–25 , from uncommitted hematopoietic progenitors
in the bone marrow to mature B cells in the periphery,
particularly when they are activated and proliferating. Malignant
B cells, including those of mantle cell lymphoma, Burkitt’s lymphoma
and B cell chronic lymphocytic leukemia, have aberrant AID expression
26,27 and abundant HoxC4 expression 22,28 . HoxC4 induces the
human 3¢ E a enhancer elements, particularly DNAse I–hypersensitive
sites 1 and 2 (called ‘hs1,2’ here), in a B cell development stage–
specific way 25 . HoxC4 binds to a HoxC4-Oct motif with the sequence
5¢-ATTTGCAT-3¢ in hs1,2 (refs. 24,25), which is conserved in humans,
1 Institute for Immunology, School of Medicine and School of Biological Sciences, University of California, Irvine, California, USA. 2 These authors contributed equally to
this work. Correspondence should be addressed to P.C. (pcasali@uci.edu).
540 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a
Aicda expression (fold) Hoxc4 expression (fold)
50
40
30
20
10
0
660
600
480
360
240
120
0
Bone
marrow
Thymus
Spleen
Peyer’s
patch
Liver
Heart
b
HoxC4
AID
PCNA
β-actin
PNA hi B220 +
PNA lo B220 +
mice, rats and rabbits, and acts in synergy with homeodomain
transcription factors Oct1 and Oct2 (A001704; collectively called
‘Oct’ here) and the Oca-B coactivator (A001696) to induce hs1,2 in
B cells 24,25 . HOXC4 expression is induced by stimuli that induce the
differentiation of germinal center B cells and AICDA expression 24,25 ,
such as CD154 (A000536) and interleukin 4 (IL-4; A001262), which
suggests involvement of HoxC4 in CSR and SHM.
In this study, we test the hypothesis that HoxC4 regulates AID
expression in human and mouse B cells. We show that HoxC4 bound
to a HoxC4-Oct motif in the AICDA and Aicda promoters that is
conserved in humans, chimps, mice, rats, dogs and cows. Binding of
HoxC4 to this cis element activated the AICDA and Aicda promoters
and induced AID expression, thereby inducing CSR and SHM. In this
function, HoxC4 acted in synergy with an equally conserved upstream
binding site for the transcription factors Sp1 and Sp3 (collectively
called ‘Sp’ here) and NF-kB intheAICDA and Aicda promoters.
RESULTS
HoxC4 and AID are induced in germinal center B cells
HoxC4 is upregulated in human immunoglobulin D–negative (IgD – )
CD38 + germinal center B cells 24,25 , which express AID and undergo
CSR and SHM. Stimulation of human IgD + CD38 – naive B cells with
an agonistic monoclonal antibody (mAb) to CD40 and human IL-4
upregulates HoxC4 and induced AID expression 24,25 .Herewefurther
analyzed the expression of Hoxc4 and Aicda in the bone marrow,
thymuses, spleens, Peyer’s patches, livers and hearts of wild-type
C57BL/6 mice. Real-time quantitative RT-PCR showed that like
Aicda, Hoxc4 was expressed mainly in the spleen and Peyer’s patches,
which contain a large proportion of hypermutating and switching
B cells but not in nonlymphoid organs such as the liver or heart
(Fig. 1a). To further address the correlation between the expression of
HoxC4 and AID, we isolated PNA hi B220 + (germinal center) B cells
and PNA lo B220 + (non–germinal center) B cells from the spleens of
8- to 10-week-old C57BL/6 mice 14 d after immunizing the mice with
NP 16-CGG (16 molecules of 4-hydroxy-3-nitrophenyl acetyl coupled
to 1 molecule of chicken g-globulin) and measured HoxC4 and AID,
as well as PCNA, a multifunctional protein critical for DNA replication
and repair that has high expression in actively dividing cells.
HoxC4 was specifically expressed in PNA hi B220 + germinal center
B cells, which also had high expression of AID and PCNA (Fig. 1b).
Stimulation of mouse spleen B cells with bacterial lipopolysaccharide
(LPS) and IL-4 or CD154 and IL-4, which induce the differentiation of
germinal center B cells and Aicda expression, upregulated Hoxc4
expression by 10- to 15-fold (Fig. 1c) and induced CSR to IgG1
(data not shown), which indicates that HoxC4 is involved in inducing
AID expression.
c
Aicda expression (fold) Hoxc4 expression (fold)
16
14
12
10
8
6
4
2
0
12
10
8
6
4
2
0
+ nil
+ LPS + IL-4
+ CD154 + IL-4
1 2
Experiment
3
ARTICLES
Figure 1 Hoxc4 expression correlates with Aicda expression. (a) Real-time
quantitative RT-PCR analysis of Hoxc4 and Aicda transcripts in bone
marrow, thymus, spleen, Peyer’s patches, liver and heart of C57BL/6 mice,
normalized to expression of Gapdh (encoding glyceraldehyde phosphate
dehydrogenase) and presented relative to mRNA in bone marrow, set as 1.
Data are from three independent experiments (mean and s.e.m. of triplicate
samples). (b) Immunoblot analysis of HoxC4, AID, PCNA and b-actin in
PNA hi B220 + (germinal center) B cells and PNA lo B220 + (non–germinal
center) B cells from spleens of mice immunized with NP 16-CGG. Data are
representative of two independent experiments. (c) Real-time quantitative
RT-PCR analysis of the expression of Hoxc4 and Aicda mRNA in C57BL/6
splenic B cells stimulated for 3 d with LPS plus IL-4 or with CD154 plus
IL-4, normalized to CD79b transcripts and presented relative to expression
in unstimulated B cells (+ nil), set as 1. Data are from three independent
experiments (mean and s.e.m.; n ¼ 3 mice).
HoxC4 deficiency impairs the antibody response to NP-CGG
We used Hoxc4 –/– mice to address the function of HoxC4 in CSR and
SHM. Two laboratories independently generated two lines of HoxC4deficient
mice; homozygous mutants of both lines had esophageal
defects and abnormal cervical and thoracic vertebral development and
suffered high postnatal mortality rates 29,30 . In these mice, the expression
of Hoxc5 and Hoxc6, which are in the same gene cluster as Hoxc4,
was lower, probably because of the effect of the neighboring neomycinresistance
selection cassette inserted in the Hoxc4 locus 29,30 .Toobviate
that effect, another laboratory generated a third Hoxc4 –/– strain in
which the neomycin-resistance cassette was deleted by Cre recombinase
through two flanking loxP sites in the germline (Supplementary Fig. 1
online), thereby leaving the expression of Hoxc5 and Hoxc6 unaltered
(A.M. Boulet and M.R. Capecchi, unpublished data). Those mice have
since been lost, but frozen Hoxc4 +/– sperm on the C57BL/6 background
was preserved. Using that Hoxc4 +/– sperm, we rederived Hoxc4 +/– mice
by in vitro fertilization and bred them to obtain Hoxc4 –/– mice. These
Hoxc4 –/– mice were born at the expected mendelian ratios, did not
suffer the high postnatal mortality rate of the earlier HoxC4-deficient
mouse lines 29,30 and developed to adulthood.
In unimmunized Hoxc4 –/– mice, serum IgM titers were normal.
However, the average serum IgG1 concentration was less than
0.6 mg/ml, compared with 1.2 mg/ml in their Hoxc4 +/+ littermates
(data not shown), which suggested impairment of CSR. We immunized
four pairs of 8- to 10-week-old littermate Hoxc4 –/– and Hoxc4 +/+
mice with NP 16-CGG and analyzed blood from these mice for titers of
IgM and IgG1; IgM and IgG1 bound to NP30 (30 molecules of NP
coupled to BSA); and IgM and IgG1 bound with high affinity to NP 3
(3 molecules of NP coupled to BSA; Fig. 2a). Titers of total IgM and
NP30-binding IgM were not significantly different in Hoxc4 –/– and
Hoxc4 +/+ mice. Hoxc4 –/– mice, however, had slightly less NP 3-binding
IgM and significantly lower titers of total IgG1 and NP 30-binding
IgG1, as well as of high-affinity NP3-binding IgG1. The defective
antibody response to NP-CGG did not reflect altered development of
plasma cells or memory B cells, as the proportions of B220 lo CD138 +
cells and CD38 hi B cells among NP-binding IgG1 B cells in NP16-
CGG-immunized Hoxc4 –/– mice were similar to those of their
Hoxc4 +/+ littermates (Fig. 2b,c). Instead, it reflected the lower overall
IgG1 and, together with the slightly lower NP 3-binding IgM activity, a
lower binding affinity for NP.
HoxC4 deficiency does not alter germinal center formation
The defective antibody response to NP-CGG in Hoxc4 –/– mice was not
due to obvious defects in lymphoid differentiation. In these mice, the
size of the spleen and the number and size of Peyer’s patches were
similar to those of Hoxc4 +/+ mice (data not shown). Moreover, the
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 541

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
IgM (mgeq/ml) IgG1 (mgeq/ml)
1.6
1.2
0.8
0.4
0.0
P = 0.019 P = 0.022 P = 0.026
NP30-binding IgG1
EC50 (× 103 )
1.6
1.2
0.8
0.4
0.0
Hoxc4 +/+ Hoxc4 –/–
Hoxc4 +/+ Hoxc4 –/–
NS
80
NS
60
40
20
0
NP30-binding IgM
EC50 (× 103 )
b c
Hoxc4 +/+
Hoxc4 –/–
B220-PE
80
60
40
20
0
NP3-binding IgG1
EC50 (× 103 )
NP3-binding IgM
EC50 (× 103 )
Hoxc4 +/+ Hoxc4 –/–
80
60
40
20
0
P = 0.048
80
60
40
20
0
10 4
10 3
10 2
10 1.8
1.9
1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
33.3
34.3
10 0
10
CD138-FITC IgG1-APC CD38-PECY7
1 10 2 10 3 10 4 10 0 10 1 10 2 10 3 10 4
10 0 10 1 10 2 10 3 10 4
Hoxc4 +/+
Hoxc4 –/–
NP-PE
number of B cells and T cells, the proportion of CD4 + T cells, and the
viability of B cells and T cells in the spleen and Peyer’s patches, as
analyzed by staining with 7-amino-actinomycin D, were also similar to
those of Hoxc4 +/+ mice (Fig. 3a–d). After stimulation with LPS and
IL-4, Hoxc4 –/– B lymphocytes were similar to Hoxc4 +/+ cells in terms
of cell cycle, as analyzed by staining with propidium iodide, and cell
division rate, as measured by incorporation of the vital dye CFSE
(Fig. 3e,f). In Hoxc4 –/– mice, the number and architecture of germinal
centers in the spleen, the proportion of proliferating B cells, as shown
by in vivo incorporation of the thymidine analog BrdU, as well as the
proportion of PNA hi germinal center B cells in both spleen and Peyer’s
patches, were similar to those of Hoxc4 +/+ mice (Fig. 3g–i), which
suggested that the defective antibody response to NP 16-CGG in
Hoxc4 –/– mice reflected an intrinsic impairment of the CSR and
SHM ‘machinery’.
HoxC4 deficiency impairs CSR and SHM
To assess the effect of HoxC4 deficiency on CSR, we stimulated
spleen B cells with LPS (to induce switching to IgG2b and IgG3),
with LPS or CD154 plus IL-4 (to induce switching to IgG1 and
IgE), with LPS or CD154 plus IFN-g (A001238; to induce
switching to IgG2a), or with LPS or CD154 in the presence of
transforming growth factor-b1 (TGF-b1; A002271), IL-5, IL-4 and
dextran-conjugated mAb to d-chain (to induce switching to IgA).
After 4 d of stimulation with LPS alone or LPS plus cytokines, the
proportion of Hoxc4 –/– B cells positive for surface IgG1, IgG2a,
IgG2b, IgG3, IgA and IgE was 54%, 49%, 46%, 36%, 43% and 57%
lower, respectively, than that of the corresponding Hoxc4 +/+
B cells (Supplementary Fig. 2 online); in cultures stimulated
with CD154 and cytokines, the proportion of Hoxc4 –/– B cells
positive for surface IgG1, IgG2a and IgE was 49%, 69% and 79%
lower, respectively. Consistent with those findings, after 7 d,
secretion of IgG1, IgG2a, IgG2b, IgG3, IgA and IgE by Hoxc4 –/–
B cells stimulated with LPS alone or LPS plus cytokines was as
much as 48%, 55%, 37%, 35%, 40% and 66% lower, respectively
Cells
Figure 2 Impaired antibody response in Hoxc4 –/– mice. (a) Titers of
circulating IgM and IgG1 (left), NP 30-binding IgM and IgG1 (middle) and
high-affinity NP 3-binding IgM and IgG1 (right) in Hoxc4 +/+ and Hoxc4 –/–
littermates (n ¼ 4 pairs of mice; each symbol represents an individual
mouse) immunized with NP16-CGG and ‘boosted’ 21 d later, measured 7 d
after the boost injection and presented as milligram equivalents (mgeq; left)
or the number of dilutions needed to reach 50% of saturation binding
(EC50; middle and right). NS, not significant (P values, paired t-test).
(b,c) Development of plasma cells and memory B cells in Hoxc4 +/+ and
Hoxc4 –/– littermates 14 d after immunization with NP 16-CGG. (b) Surface
expression of B220 and CD138 on spleen cells. Numbers in outlined areas
indicate percent B220 lo CD138 + (plasma) cells as percentage of total B220 +
cells. (c) Flow cytometry of spleen cells stained with FITC-labeled PNA,
phycoerythrin (PE)-labeled NP, allophycocyanin (APC)-labeled mAb to
mouse IgG1 and phycoerythrin-indotricarbocyanine (PECy7)-labeled mAb to
mouse CD38. Insets (left), NP-binding surface IgG1 + B cells; right, numbers
above bracketed lines indicate percent CD38 + cells among those gated
NP-binding IgG1 + B cells. Data are representative of four (a) or three
(b,c) independent experiments.
(Fig. 4a). Impaired CSR in Hoxc4 –/– B cells was not due to altered
proliferation, as after 2, 3 or 4 d of culture with LPS plus IL-4 or
with CD154 plus IL-4, Hoxc4 –/– B cells completed the same
number of divisions as their Hoxc4 +/+ counterparts did (Figs. 3f
and 4b).Itwasalsonotduetoalteredplasma cell differentiation,
as after 4 d of culture with LPS alone, LPS plus IL-4, or CD154
plus IL-4, the number of CD138 + B220 lo plasma cells that emerged
from Hoxc4 –/– B cells was similar to that of their Hoxc4 +/+
counterparts (Fig. 4c). Consistent with those findings, expression
of the transcription factors Blimp-1 and IRF4, which are
critical for plasma cell differentiation, was similar in Hoxc4 –/–
and Hoxc4 +/+ B cells, as determined by real-time quantitative
RT-PCR, at 5 d after stimulation with LPS and IL-4 (data not
shown). Further, lower CSR in Hoxc4 –/– Bcellswasnotdueto
impairment of germline transcription of the intervening heavychainregionandconstantheavy-chainregion(I
H-C H), which is
necessary for CSR. Real-time quantitative RT-PCR showed that the
abundance of germline transcripts of I m-C m, I g3-C g3, I g1-C g1,
I g2b-C g2b, I g2a-C g2a, I e-C e and I a-C a in Hoxc4 –/– B cells stimulated
for 3 d with LPS alone or LPS plus cytokines was similar to
that in their Hoxc4 +/+ B cell counterparts (Fig. 5 and Supplementary
Fig. 3 online), whereas post-recombination I m-C H transcripts,
which are generated by CSR, were significantly less
abundant in Hoxc4 –/– B cells, by as much as 89.2%. Thus,
HoxC4 deficiency impairs CSR to all isotypes without affecting
germline I H-C H transcription.
To asses the effect of HoxC4 deficiency on SHM, we analyzed the
IgH J H4–intronic enhancer (iE m) sequence downstream of rearranged
V J558DJ H4 DNA in PNA hi B220 + germinal center B cells from
Peyer’s patches of 12-week-old unimmunized littermate Hoxc4 –/– and
Hoxc4 +/+ mice. In these mice, the proportion of Peyer’s patch
PNA hi B220 + germinal center B cells was similar (Fig. 3h). Analysis
of 324 and 319 JH4-iEm intronic DNA sequences (720 base pairs (bp))
from Hoxc4 –/– and Hoxc4 +/+ mice showed that Hoxc4 –/– mice had
59% fewer mutations (P o 0.00001; Fig. 6a). The lower mutation
frequency was associated with similarly fewer mutations of
deoxyguanine-deoxycytosine and deoxyadenine-deoxythymine (Supplementary
Fig. 4 online) and was not due to impaired transcription
of the rearranged V J558DJ H genes, as shown by specific real-time
quantitative RT-PCR of Peyer’s patch B cells from these Hoxc4 –/– mice
and their Hoxc4 +/+ littermates (Fig. 6b). Thus, HoxC4 deficiency
substantially impairs SHM without altering the spectrum of the
residual mutations or VHDJH transcription.
542 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
a
Hoxc4 +/+
Hoxc4
Cells
–/–
Spleen
Peyer’s
patch
B220-PE
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
Hoxc4 +/+
Hoxc4 –/–
51.9 2.7 53.9 3.5
5.8
59.1
39.6
6.4
5.8
55.9
37.6
5.0
2.4 32.1 3.4 35.6
10 4
10 3
10 2
10 1
10 0
10 1
10 0
10 4
10 3
10 2
CD3-FITC CD3-FITC CD3-FITC B220-PE
HoxC4 deficiency impairs AID expression
CSR and SHM require transcription of the Igh locus and AID
expression. The much lower CSR and SHM, together with the normal
abundance of mature VHDJH-Cm and germline IH-CH transcripts, in
Hoxc4 –/– B cells prompted us to hypothesize modulation of AID
expression by HoxC4. We stimulated spleen Hoxc4 +/+ and Hoxc4 –/–
B cells for 0, 12, 24, 48 and 72 h with LPS plus IL-4 or with CD154
plus IL-4. Expression of Aicda mRNA could be detected by real-time
quantitative RT-PCR as early as 24 h and peaked within 48–72 h of
stimulation in Hoxc4 +/+ B cells. In Hoxc4 –/– B cells, Aicda expression
was more than 70% lower after 72 h of stimulation (Fig. 7a). The
lower abundance of Aicda transcripts was associated with much less
AID protein (Fig. 7b), which further suggests that HoxC4 regulates
Aicda expression.
Binding of transcription factors to the Aicda promoter
To address the possibility that HoxC4 modulates Aicda expression
by binding to a cis-regulatory element of this gene, we analyzed the
sequence upstream of the putative transcription-initiation site of
Aicda (Supplementary Fig. 5 online). We identified eight motifs
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
Hoxc4 +/+
Hoxc4 –/–
Hoxc4 +/+
Hoxc4 –/–
b c d
CD4-PerCP
Hoxc4 +/+
Hoxc4 –/–
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
7-AAD
B220-PE
10 4
10 3
10 2
10 1
10 0
e f
g
Hoxc4 +/+
Hoxc4 –/–
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
Cells
Day 1
Day 2 Day 2 Day 3
250
200
150
100
50
G0-G1 G0-G1 = 90.8%
S S = 3.1%
G2-M G2-M = 4.9%
G0-G1 = 41.4%
G0-G1
S
S = 25.2%
G2-M
G2-M = 16.1%
250
200
150
100
50
765 4321
0 76543210
0
250
200 G0-G1
G0-G1 = 90.4%
G0-G1 = 43.3%
G0-G1
150 S S = 3.1%
S = 28.5%
S
100 G2-M G2-M = 4.3%
G2-M = 14.3%
50
G2-M
0
0 200 400 600 800 1,000 0 200 400 600 800 1,000
0
250
200
150
100
50
0
76543
2 1 0
76543210
DNA CFSE
Hoxc4 +/+
Hoxc4 –/–
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
h i
Spleen
Peyer’s
patch
Figure 3 HoxC4 deficiency does not affect B cell, T cell or CD4 + T cell
numbers, death of B cells and T cells in spleens and Peyer’s patches or
B cell cycle or division or alter germinal center formation and in vivo B cell
proliferation. (a–d) Flow cytometry of spleen and Peyer’s patch cells stained
with phycoerythrin-labeled mAb to B220 and FITC-labeled mAb to CD3 (a),
FITC-labeled mAb to CD4 and peridinine chlorophyll protein complex
(PerCP)-labeled mAb to CD4 (b), 7-amino-actinomycin D (7-AAD) and
FITC-labeled mAb to CD3 (c), or 7-amino-actinomycin and phycoerythrinlabeled
mAb to B220 (d). Data are representative of three experiments.
(e) Cell cycle analysis of Hoxc4 +/+ and Hoxc4 –/– splenic B cells stimulated
10 4
10 3
10 2
10 1
10 0
conserved in humans, chimps, mice, rats, dogs and cows (Supplementary
Fig. 5). The first six motifs did not fulfill the minimal
criteria for known transcription factor–binding sites by weightmatrix
search with the Match program (threshold score, 0.75). The
final two were a HoxC4- or Oct-binding site with a sequence of
5¢-ATTTGAAT-3¢ (residues –29 to –22 in humans and mice; scores,
1.0 for HoxC4 and 0.93 for Oct), which was nearly identical to the
conserved HoxC4-Oct motif that is critical in inducing the human
IGH 3¢ E a enhancer elements 24,25 , and an upstream Sp–NF-kB–
binding site with a sequence of 5¢-GGGGAGGAGCC-3¢ (residues
–57 to –47 in humans and mice 17 ; scores, 0.93 for Sp1 and Sp3, and
0.96 for NF-kB). This sequence (5¢-GGGGAGGAGCC-3¢) hasbeen
suggested to be a Pax5-binding site 16 , but it did not satisfy, in our
analysis, the minimum requirement for such a binding site (a score
of 0.61). Both the putative HoxC4-Oct–binding site and the
Sp–NF-kB–binding site were identical in all six species analyzed.
To determine the function of the region from position –349 to
position –1 in the Aicda promoter (tentatively defined as the Aicda
promoter on the basis of its high conservation) in the regulation of
Aicda transcription, we constructed pGL3–luciferase reporter vectors
7-AAD
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
B220-PE
Hoxc4 +/+
PNA-FITC BrdU-APC
ARTICLES
with LPS plus IL-4 and collected after 1 and 2 d for propidium iodide staining and flow cytometry for measurement of DNA content and cells in the G0-G1,
S and G2-M phases. Data are representative of three experiments. (f) Cell division by Hoxc4 +/+ and Hoxc4 –/– splenic B cells labeled with CFSE, cultured with
LPS plus IL-4, and collected 2 and 3 d later for flow cytometry. Progressive left shift of fluorescence indicates cell division; numbers in plots indicate cell
generations. Data are from one representative of three experiments. (g) Staining of germinal centers in spleen sections prepared 10 d after immunization of
Hoxc4 +/+ and Hoxc4 –/– mice with NP16-CGG. Scale bars, 200 mm. Results are representative of three experiments. (h) Flow cytometry of cells obtained from
Peyer’s patches from NP16-CGG–immunized Hoxc4 +/+ and Hoxc4 –/– mice and stained with phycoerythrin-labeled mAb to B220 and FITC-labeled PNA. Data
are representative of three experiments (n ¼ 3 pairs of mice). (i) In vivo proliferation of B cells from spleens and Peyer’s patches of 10-week-old Hoxc4 +/+
and Hoxc4 –/– mice (n ¼ 3 pairs) immunized with NP16-CGG, then, 10 d later, injected with BrdU (1 mg) twice within 16 h and analyzed 4 h after the final
injection; cells were stained with phycoerythrin-labeled mAb to B220 or that mAb plus FITC-labeled PNA; incorporated BrdU was detected by flow cytometry
with allophycocyanin-labeled mAb to BrdU. Data are representative of three experiments. Numbers in quadrants (a–d,h,i) indicate percent cells in each.
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 543
10 4
10 3
10 2
10 1
10 0
Hoxc4 +/+
10 4
10 3
10 2
10 1
10 0
Hoxc4 –/–
10 4
10 3
10 2
10 1
10 0
Hoxc4 –/–
10 3
10 2
10 1
10 0
10 4

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a P = 0.034 P = 0.0049 P = 0.010 P = 0.020 P = 0.049 P = 0.0068 b
IgG1 (ngeq/ml)
1,400
1,200
1,000
800
600
400
200
0
Hoxc4 +/+
Hoxc4 –/–
IgG2b (ngeq/ml)
Hoxc4 +/+
Hoxc4 –/–
IgG3 (ngeq/ml)
Hoxc4 +/+
Hoxc4 –/–
consisting of the 349-bp Aicda promoter and/or the 490-bp flanking 5¢
region (Fig. 7c) upstream of the gene encoding the luciferase reporter,
combined with nothing more or with conserved region1 (cr1) and/or
cr2, which are in the first intron in the Aicda 15,16 .Weusedthese
vectors to transfect human Ramos B cells, which spontaneously
express AICDA and undergo SHM, and mouse CH12F3-2A B cells,
which express Aicda and undergo CSR after stimulation with LPS,
IL-4 and TGF-b1. We cultured these B cells and measured luciferase
activity after 16 h (Ramos) or 24 h (CH12F3-2A). The construct that
contained both the Aicda promoter and its flanking 5¢ region had
11- to 15-fold more activity than did the empty vector in Ramos and
CH12F3-2A B cells (Fig. 7d). Neither cr1 nor cr2 showed substantial
enhancer activity in either Ramos or CH12F3-2A B cells. In addition,
whereas the construct that contained only the Aicda promoter
promoted transcription as efficiently as that containing both the
Aicda promoter and its flanking 5¢ region, the construct that included
only the flanking 5¢ region had only ‘background’ activity similar to
that of empty vector in both Ramos B cells and stimulated CH12F3-
2A B cells. These experiments show that the 349-bp Aicda promoter
region had full promoter activity, whereas neither cr1 nor cr2
enhanced Aicda promoter activity in human or mouse B cells.
To address the specificity of the two evolutionarily conserved
5¢-ATTTGAAT-3¢ and 5¢-GGGGAGGAGCC-3¢ motifs, we did
electrophoretic mobility-shift assay (EMSA) with wild-type and
mutated oligonucleotide probes encompassing residues –65 to –14
of the human AICDA promoter sequence and containing both the
HoxC4-Oct–binding site and the Sp–NF-kB–binding site (Sp-Hox),
or an oligonucleotide probe encompassing residues –65 to –44 of the
AICDA promoter sequence containing only the Sp–NF-kB–binding
site (Sp; Supplementary Fig. 6a online). Incubation of nuclear
extracts of human 4B6 B cells, which spontaneously express AICDA
and undergo CSR, or Ramos B cells, with the Sp-Hox probe gave
rise to four main protein–DNA complexes: A and A¢, and B and
B¢ (Supplementary Fig. 6b,c), specific for the binding of HoxC4-Oct
IgG2a (ngeq/ml)
300
250
200
150
100
50
0
c
Hoxc4 +/+
Hoxc4 –/–
Hoxc4 +/+
Hoxc4 –/–
B220-PE
IgA (ngeq/ml)
120
100
80
60
40
20
0
Hoxc4 +/+
Hoxc4 –/–
IgE (ngeq/ml)
and Sp–NF-kB, respectively. The mutated oligonucleotides Sp-Hox mut
(in which the HoxC4-Oct–binding site was disrupted), Sp mut -Hox (in
which the Sp–NF-kB–binding site was disrupted) and Sp mut -Hox mut
(in which both sites were disrupted) did not efficiently achieve
competition for the formation of complexes B and B¢, AandA¢, or
all four complexes, respectively. We confirmed those results by EMSA
with the mutated oligonucleotides as radiolabeled probes; Sp-Hox mut
gave rise only to complexes B and B¢, Sp mut -Hox yielded only A and
A¢, and Sp mut -Hox mut gave rise to none of the four complexes.
Incubation of nuclear extracts of 4B6 or Ramos B cells with the Sp
mRNA (normalized)
50
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0
Hoxc4 +/+
Hoxc4 –/–
Hoxc4 +/+
Hoxc4 –/–
Hoxc4 +/+
Hoxc4 –/–
NS NS NS NS NS NS NS
Hoxc4 +/+
Hoxc4 –/–
10 4
10 3
10 2
10 1
10 0
10 4
10 4
10 3
10 3
10 2
10 2
10 1
10 1
10 0
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 4
10 3
10 3
10 2
10 2
10 1
10 1
10 0
10 0
654321 0
20
16
12
8
4
CFSE
0
0 1 2 3 4
Cell division
5 6
654321 0
LPS + IL-4
12
Hoxc4
10
8
6
4
2
CFSE
0
0 1 2 3 4
Cell division
5 6
CD154 + IL-4
+/+
Hoxc4 –/–
10
CD138-FITC
4
10 3
10 2
10 1
10 0
10 4
10 4
10 3
10 3
10 2
10 2
10 1
10 1
10 0
10 0
10 4
10 3
10 2
10 1
10 0
10 4
10 3
10 2
10 1
10 0
LPS LPS + IL-4 CD154 + IL-4
Figure 4 Impaired CSR in Hoxc4 –/– B cells.
(a) Immunoglobulin titers in supernatants of
cultures of Hoxc4 +/+ and Hoxc4 –/– splenic
B cells stimulated for 7 d with LPS alone
(IgG2b and IgG3), or with LPS plus IL-4
(IgG1 and IgE), IFN-g (IgG2a) or TGF-b1,
IL-4, IL-5 and dextran-conjugated mAb to
d-chain (IgA). P values, paired t-test. Data are
from three experiments with four pairs of
Hoxc4 +/+ and Hoxc4 –/– mice. (b) Proliferation
of Hoxc4 +/+ and Hoxc4 –/– B cells labeled with
CFSE and stimulated with LPS plus IL-4 (top) or with CD154 plus IL-4 (bottom) to induce switching to IgG1. P ¼ 0.0051, for LPS plus IL-4, or
P ¼ 0.0069, for CD154 plus IL-4 (overall; paired t-test). Numbers above plots (left) indicate cell divisions. Data are from two independent experiments.
(c) Plasma cell–differentiation of Hoxc4 +/+ and Hoxc4 –/– B cells stimulated for 4 d with LPS alone, LPS plus IL-4, or CD154 plus IL-4, then stained with
phycoerythrin-labeled mAb to mouse B220 and FITC-labeled mAb to mouse CD138 and analyzed by flow cytometry. Numbers in outlined areas indicate
percent B220loCD138 + (plasma) cells among total cells. Data are from one representative of three independent experiments.
40
30
20
10
0
IgG1-APC
IgG1-APC
Hoxc4 +/+
Hoxc4 –/–
I µ -C µ
Iγ3-Cγ3 Iγ1-Cγ1 Iγ2b-Cγ2b P = 0.00026
Iγ2a-Cγ2a P = 0.000046
Iε-Cε P = 0.00034
Iα-Cα P = 0.0022
I µ -Cγ3 P = 0.0019
I µ -Cγ1 P = 0.0028
I µ -Cγ2b I µ -Cγ2a I µ -Cε I µ -Cα Germline I H -C H transcripts Post-recombination I µ -C H transcripts
Figure 5 HoxC4 deficiency does not alter germline IH-CH transcripts but
results in lower expression of post-recombination Im-CH transcripts. Realtime
quantitative RT-PCR analysis of germline IH-CH transcripts (left) and
post-recombination I m-C H transcripts (right) in Hoxc4 +/+ and Hoxc4 /
splenic B cells cultured for 3 d with LPS alone (I g2b-C g2b, I g3-C g3, I m-C g2b
and I m-C g3) or with LPS plus IL-4 (I m-C m,I g1-C g1, I E-C E,I mC g1 and I m-C E),
IFN-g (I g2a-C g2a and I m-C g2a), or TGF-b1, IL-4, IL-5 and dextran-conjugated
mAb to m-chain (I a-C a and I m-C a); expression is normalized to Cd79b
expression and is presented relative to the expression in Hoxc4 +/+ B cells,
set as 1. P values, paired t-test. Data are representative of three experiments
(mean and s.e.m. of triplicate samples.) with one representative of three
pairs of Hoxc4 +/+ and Hoxc4 –/– mice (other pairs, Supplementary Fig. 3).
544 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
IgG1 (%)
IgG1 (%)

© 2009 Nature America, Inc. All rights reserved.
a b
Hoxc4 +/+
Frequency
of mutations
Hoxc4 –/–
Frequency
of mutations
Pair 1
>10
8
7
6
5
4
3
2
1
0
9 >10
7
8
6
4
3
2
1
0
>10
8
5
4
1
0
3.53 × 10
578
4 ≥10
3
2
4 68
>10
3
2
6
2
9
1
1
1
–3
2.70 × 10 –3
2.56 × 10 –3
Pair 2 Pair 3
105
107 107
108 106 110
0
1.35 × 10 –3
1.15 × 10 –3
probe gave rise to three main protein-DNA complexes, C, C¢ and C¢¢,
specific for the binding of Sp–NF-kB. We further confirmed the
binding specificity of HoxC4, Oct1 and Oct2 to the 5¢-ATTTGAAT-
3¢ site, and the binding specificity of Sp1, Sp3 and NF-kB tothe5¢-
GGGGAGGAGCC-3¢ site, by supershift or inhibition of the formation
of the respective protein-DNA complexes by using a specific mAb to
HoxC4 and specific antibody to Oct1 (anti-Oct1), anti-Oct2, anti-Sp1,
anti-Sp3 or antibody to the p52 subunit of NF-kB. No supershift or
inhibition of protein-DNA complex involving the Sp-Hox or the Sp
probe was achieved with a Pax5-specific antibody. These experiments
show that HoxC4, Oct1 and Oct2 bind specifically to the conserved
5¢-ATTTGAAT-3¢ site in the AICDA promoter, whereas Sp1, Sp3 and
NF-kB, but not Pax5, bind specifically to the conserved 5¢-GGGGAG
GAGCC-3¢ site in the AICDA promoter.
Cis elements critical for Aicda and AICDA promoter activation
To determine the contribution of the conserved cis elements, including
HoxC4-Oct– and Sp–NF-kB–binding sites to the activity of the Aicda
0
1.12 × 10 –3
0
V J558 DJ H -C µ transcripts (fold)
1.6
Hoxc4
1.4
1.2
1.0
0.8
0.6
0.4
0.2
NS
0
Pair 1
+/+
Hoxc4 –/–
NS NS
Pair 2 Pair 3
Figure 6 Somatic mutation in the immunoglobulin heavy-chain intronic J H4-iEm DNA of Peyer’s patch
PNA hi B220 + (germinal center) B cells. (a) Proportion of sequences carrying various numbers (along
margins) of mutations over the 720-bp J H4-iE m DNA from Peyer’s patch PNA hi B220 + (germinal center)
B cells from 12-week-old Hoxc4 –/– and Hoxc4 +/+ littermates (n ¼ 3 pairs of mice); below, frequency of
somatic mutation; center, number of sequences analyzed (spectrum of mutation, Supplementary
Fig. 4). Data are from three experiments. (b) Real-time quantitative RT-PCR analysis of V J558DJ H-C m
transcripts in Peyer’s patch B cells from the mice in a, normalized to CD79b expression and presented
relative to the expression in Hoxc4 +/+ B cells. Data are from three experiments (mean and s.e.m. of
triplicate samples) with three pairs of Hoxc4 +/+ and Hoxc4 –/– mice.
Figure 7 HoxC4 deficiency impairs AID expression, which depends on the
conserved HoxC4-Oct–binding site in the Aicda promoter. (a) Quantitative
RT-PCR analysis of Aicda expression in RNA (2 mg) from Hoxc4 +/+ and
Hoxc4 –/– splenic B cells cultured for 0, 12, 24, 48 and 72 h in the
presence of LPS plus IL-4 (left) or CD154 plus IL-4 (right), normalized
to Cd79b expression and presented relative to expression in Hoxc4 +/+
B cells at time 0. P values, paired t-test. Data are representative of three
independent experiments (mean and s.e.m.). (b) Immunoblot analysis of AID
and b-actin in the cells in a. Data are representative of three independent
experiments. (c) Promoter regions of human AICDA and mouse Aicda (AIDpro),
the flanking 5¢ region (5¢R), cr1 and cr2 in the first intron, and exons
1–5 (coding regions, light blue). Numbers above indicate the number of
nucleotides in each region. (d) Aicda promoter activity in mouse CH12F3-2A
B cells (CSR inducible) and human Ramos B cells (spontaneous SHM)
transfected with pGL3-Basic luciferase (luc) reporter vectors containing
various combinations of elements of the Aicda locus (left) and then treated
with LPS, IL-4 and TGF-b1 to induce CSR (CH12F3-2A) or given no further
treatment (nil; Ramos), assessed as luciferase activity measured after or
24 h (CH12F3-2A) or 16 h (Ramos) culture and presented relative to the
luciferase activity of cells transfected with empty vector (pGL3). Data are
representative of three independent experiments (mean and s.e.m.).
promoter, we constructed luciferase reporter
vectors containing the mouse Aicda promoter
sequence (residues –349 to –1) with mutation
or deletion of the HoxC4-Oct–binding site
and/or the Sp–NF-kB–binding site. In addition
to the conserved HoxC4-Oct–binding
site with the sequence 5¢-ATTTGAAT-3¢, we
identified a putative HoxC4-binding site with
a sequence of 5¢-ATTT-3¢ in the mouse and
rat promoter (residues –155 to –158 of
mouse Aicda) but not in the human, chimp
or dog promoter (Supplementary Fig. 5).
Deletion of this site did not alter the Aicda
promoter activity (Fig. 8a). In contrast, deletion
of the Hoxc4-Oct motif resulted in
promoter activity that was 71%, 64% and
88% lower in mouse CH12F3-2A B cells
induced with LPS, IL-4 and TGF-b, in
human 4B6 cells, and in Ramos B cells,
respectively. To determine the relative contribution
of the binding of HoxC4 and Oct to
the promoter activity of the HoxC4-Oct
motif as a whole, we mutated 5¢-ATTT-
GAAT-3¢ to 5¢-CTTTGAAT-3¢, thereby disrupting the binding of
HoxC4 but retaining the binding of Oct 25 ,orto5¢-ATTTGCCG-3¢,
thereby abrogating the binding of Oct1-Oct2 but not of HoxC4
(mutated residues underlined) 25 . Mutation of the HoxC4 motif in
the HoxC4-Oct site resulted in transcription that was 47%, 36% and
55% lower, whereas mutation of the Oct site resulted in transcription
that was 28%, 21% and 55% lower, in CH12F3-2A, 4B6 and Ramos B
cells, respectively. Thus, both the HoxC4 and Oct motifs of the
HoxC4-Oct–binding site contribute to the Aicda promoter activity,
as further confirmed by the up-to-88% loss of Aicda promoter activity
a
b
c
16
Hoxc4 +/+
Hoxc4 +/+
Hoxc4 –/–
Hoxc4 –/–
P = 0.00067
12
10
P = 0.0017
P = 0.0046
12
P = 0.019
8
8
4
6
4
2
P = 0.0047
0
0
0 12 24 48 72
Time (h) Time (h)
Hoxc4
AID
β-actin
Time (h) 0 12 24 48 72 0 12 24 48 72 0 12 24 48 72 0 12 24 48 72
LPS + IL-4
CD154 + IL-4
+/+
Hoxc4 –/–
0 12 24 48 72
Aicda expression
Human AICDA
ARTICLES
5′R AID-pro 84
5,748 1471,380 270 294 115 470 2,104
1 cr1 cr2
2 3 4 5
5′R AID-pro 84
Mouse Aicda
1 cr1
–839 –349 –1
5,369
cr2
147 1,423 270
2 3
530 115 379
4
1,759
5
d
CH12F3-2A
(LPS + IL-4 + TGF-β1)
Ramos
(nil)
pGL3–
luc
pGL3–5′R–AID-pro–cr1–cr2
5′R AID-pro
luc cr1 cr2
pGL3–5′R–AID-pro–cr1
luc cr1
pGL3–5′R–AID-pro–cr2
luc cr2
pGL3–5′R–AID-pro
luc
pGL3–5′R
luc
pGL3–AID-pro
luc
0 2 4 6 8 10 12 0 5 10 15 20
Luciferase activity Luciferase activity
(fold)
(fold)
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 545

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
a
b
pGL3–AID-pro–enh
–349
HoxC4 Sp–NF-κB Hoxc4 Oct
–349 –158
Empty vector
–155 –57 –45 –29 –22 –1
Human
Mouse
WT
Mut0
Mut1
Mut2
Mut3
Mut4
Mut5
Mut6
Mut7
Aicda
promoter
when the entire Hoxc4-Oct–binding site was deleted. Further, mutation
of the conserved Sp–NF-kB–binding site to 5¢-AAAAAGGAAA-3¢
resulted in promoter activity that was 73%, 80% and 63% lower in
CH12F3-2A, 4B6 and Ramos B cells, respectively. In agreement with
that result, deletion of this site resulted in promoter activity that
was 85%, 68% and 82% lower in CH12F3-2A, 4B6 and Ramos
B cells, respectively. Finally, deletion of the HoxC4-Oct–binding site
combined with mutation or deletion of the Sp–NF-kB–binding site
abrogated Aicda promoter activity. These experiments show that the
conserved HoxC4-Oct–binding site is important to Aicda promoter
activity and, together with the conserved Sp–NF-kB–binding site, is
indispensable for full transcriptional activation of Aicda.
To confirm the relevance of our EMSA and luciferase reporter
experiments, we precipitated chromatin from human 4B6 and
–1
luc
4B6
Ramos
2E2 (nil)
2E2 (anti-huCD40 + IL-4)
CH12F3-2A (nil)
CH12F3-2A (LPS + IL-4 + TGF-β1)
Spleen (nil)
Spleen (LPS + IL-4)
Spleen (CD154 + IL-4)
SV40 enhancer
CH12F3-2A 4B6
(LPS + IL-4 + TGF-β1)
(nil)
0
100
200
300
400
500
Luciferase activity
(relative)
0
100
200
c
400
500
Luciferase activity
(relative)
Ramos
(nil)
Luciferase activity
(relative)
Input
Mo IgG
Rab IgG
Oct1
Oct2
HOXC4 or Hoxc4 AICDA or Aicda GAPDH or Gapdh AICDA or Aicda promoter
Figure 8 The conserved HoxC4-Oct– and Sp–NF-kB–binding sites are essential for full Aicda promoter activity, and HoxC4, Oct1, Oct2, Oca-B, Pax5, Sp1,
Sp3 and NF-kB (p52) are recruited to the Aicda promoter in B cells expressing AICDA or Aicda and undergoing CSR or SHM. (a) Luciferase activity of
mouse CH12F3-2A B cells (CSR inducible), human 4B6 B cells (spontaneous CSR) and human Ramos B cells (spontaneous SHM) transfected with pGL3-
Enhancer luciferase reporter constructs containing wild-type Aicda promoter (WT) or mutated Aicda promoter (Mut1–Mut7) in which the HoxC4-Oct– and/or
Sp–NF-kB–binding sites were deleted or disrupted by site-directed mutagenesis (left; lower case indicates mutated nucleotides), and then treated with LPS,
IL-4 and TGF-b1 to induce CSR (CH12F3-2A) or given no further treatment (nil; 4B6 and Ramos), assessed after 24 h (CH12F3-2A) or 16 h (Ramos and
4B6) of culture and presented relative to the luciferase activity of cells transfected with empty vector. Data are representative of three independent
experiments (mean and s.e.m.). (b) Semiquantitative RT-PCR of the expression of the genes encoding HoxC4 (left), AID (middle) and GAPDH (loading
control; right) in human 4B6 and Ramos B cells; human 2E2 B cells left unstimulated (nil) or stimulated with mAb to human CD40 (anti-huDC40) plus
IL-4; mouse CH12F2-2A B cells left unstimulated or stimulated with LPS, IL-4 and TGF-b1; and C57BL/6 wild-type splenic B cells left unstimulated or
stimulated with LPS plus IL-4, or CD154 plus IL-4. Wedges (top) indicate serial twofold dilution of template cDNA. Results are representative of three
independent experiments. (c) PCR of crosslinked chromatin precipitated from the cells in b with preimmune control mouse (Mo) or rabbit (Rab) IgG, mouse
mAb to HoxC4, or rabbit anti-Oct1, anti-Oct2, anti-Oca-B, anti-Pax5, anti-Sp1, anti-Sp3 or anti–NF-kB p52, amplified with primers for the AICDA or Aicda
promoter. Results are representative of three independent experiments.
Ramos B cells with anti-HoxC4, anti-Oct1, anti-Oct2, anti-Oca-B,
anti-Sp1, anti-Sp3 or anti–NF-kB p52. In DNA precipitated by
each antibody, we readily identified the AICDA promoter sequence
(Fig. 8b,c). The specificity of those findings was further proven by
our ability to readily detect AICDA or Aicda promoter DNA in
chromatin-immunoprecipitation assays involving human 2E2 B cells,
which can be induced to express AID and undergo CSR after treatment
with mAb to CD40 and cytokines (such as IL-4), mouse CH12F3-2A B
cells, which can be induced to express AID and undergo CSR after
treatment with LPS, IL-4 and TGF-b1, as well as spleen B cells from
wild-type C57BL/6 mice activated by LPS plus IL-4 or by CD154 plus
IL-4. Induction of CSR in 2E2 or CH12F3-2A B cells or primary mouse
spleen B cells by these stimuli resulted in substantial Hoxc4 expression
and recruitment of HoxC4, Oct1, Oct2, Oca-B, Sp1, Sp3 and NF-kB to
546 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY
300
0
40
80
Oca-B
120
IgG
HoxC4
Pax5
Sp1
Sp3
NF-κB

© 2009 Nature America, Inc. All rights reserved.
a b
TAC-Aicda TAC
B220-PE
Aicda expression (relative) Aicda expression (relative)
10 0
10 1
10 2
10 3
10 4
Hoxc4
72.3
+/+
Hoxc4 –/–
22.0 76.9 9.7
5.6 0.05 12.3 0.07
59.1 33.2 58.4 34.6
10
7.5 0.2 0.2
0
10 1
10 2
10 3
10 4
6.7
10 0
10 1
10 2
IgG1-FITC
10 3
10 4
10 0
10 1
10 2
10 3
10 4
Real-time
P = 0.00041
1.2
1.0
0.8
0.6
0.4
0.2
0
TAC
TAC-Aicda
TAC
TAC-Aicda
Germline Iµ-Cµ expression
(relative)
Germline Iγ1-Cγ1 expression
(relative)
Real-time
NS
1.2
1.0
0.8
0.6
0.4
0.2
0
TAC
TAC-Aicda
TAC
TAC-Aicda
Semiquantitative
P = 0.000047
1.2
1.0
0.8
0.6
0.4
0.2
0
Circle Iγ1-Cµ expression
(relative)
Post-rec Iµ-Cγ1 expression
(relative)
TAC
TAC-Aicda
Semiquantitative Real-time Real-time
P = 0.0066
1.2
1.2
NS
P = 0.00021
1.2
1.0
1.0
1.0
0.8
0.8
0.8
0.6
0.6
0.6
0.4
0.4
0.4
0.2
0.2
0.2
0
0
0
TAC
TAC-Aicda
the AICDA and Aicda promoters. Because it has been suggested that
Pax5 is (indirectly) recruited to the Aicda promoter 16 ,weprecipitated
chromatin in these human and mouse B cells with anti-Pax5; in these
immunoprecipitated DNA complexes, we readily detected AICDA and
Aicda promoter sequences, respectively. These findings show that
HoxC4, Oct1, Oct2, Oca-B, Sp1, Sp3 and NF-kB are recruited to the
AICDA and Aicda promoters in human and mouse B cells, respectively,
that express AID and undergo CSR or SHM.
Enforced AID expression restores CSR in Hoxc4 –/– B cells
We then sought to demonstrate that the defective CSR in HoxC4deficient
B cells was actually due to impairment of AID expression and
not some other HoxC4-dependent activity. For this, we enforced
expression of AID in Hoxc4 –/– B cells to restore CSR. We used a
control retroviral vector containing human IL2RA, which encodes the
human IL-2 receptor (CD25; TAC), and an AID-expression retroviral
construct containing human IL2RA and Aicda (TAC-Aicda) 15 (Supplementary
Fig. 7 online). We transduced LPS-activated spleen
Hoxc4 +/+ and Hoxc4 –/– B cells with TAC and TAC-Aicda and stimulated
them with LPS and IL-4 for 72 h and 96 h before analyzing CSR.
Consistent with the results we obtained with untransduced Hoxc4 +/+
and Hoxc4 –/– B cells, CSR was much lower in Hoxc4 –/– B cells
transduced with the TAC control retrovirus than in their Hoxc4 +/+
counterparts (Fig. 9a,b). In Hoxc4 +/+ B cells, enforced expression of
AID increased CSR to IgG1 by about 50%. Transduction of Hoxc4 –/–
B cells with TAC-Aicda retrovirus increased Aicda expression. It did
not modulate the expression of germline I m-C m and I g1-C g1 transcripts
but did increase CSR to IgG1 to an extent similar to that of Hoxc4 +/+
B cells transduced with TAC-Aicda retrovirus, as shown by the greater
proportion of B cells positive for surface IgG1 and more circle I g1-C m
and post-recombination I m-C g1 transcripts. These experiments show
that the defective AID expression and CSR of Hoxc4 –/– B cells are
restored by enforced AID expression, which further indicates that
HoxC4 modulates CSR by regulating AID expression.
DISCUSSION
In B cells, AID expression is tightly regulated 31–34 , possibly in an
activation-dependent way in conditions in which CSR and SHM
unfold. The specificity and amount of AID expression are probably
controlled by a complex combination of various tissue-specific transcription
factors, both activators and repressors. Here we have
provided evidence that by binding to the highly conserved 5¢-ATTT
GAAT-3¢ motif in the Aicda promoter, HoxC4 activates this promoter,
thereby modulating AID expression, CSR and SHM. In this function,
HoxC4 acts in synergy with Oct1-Oct2, which also bind to the
ARTICLES
Figure 9 Enforced expression of AID restores CSR in Hoxc4 –/– Bcells.
(a) Flow cytometry of the surface expression of B220 and IgG1 in Hoxc4 +/+
and Hoxc4 –/– B cells activated with LPS and transduced with TAC or
TAC-Aicda retrovirus, then cultured for 3 or 4 d with LPS plus IL-4.
Numbers in quadrants indicate percent cells in each. Data are from one
representative of five independent experiments. (b) Real-time or
semiquantitative RT-PCR analysis (above graphs) of Aicda expression,
germline Im-Cm and Ig1-Cg1 transcripts, circle Ig1-Cm transcripts and postrecombination
(Post-rec) I m-C g1 transcripts in the cells in a, normalized to
CD79b transcripts and presented as the ratio of the expression in Hoxc4 –/–
cells to the expression in Hoxc4 +/+ cells. P values, t-test. Data are
representative of three independent experiments (mean and s.e.m.).
5¢-ATTTGAAT-3¢ motif, by a mechanism similar to that reported
for HoxC4-Oct–mediated activation of the human 3¢ E a hs1,2
enhancer element 25 . In agreement with that, Hoxc4 –/– mice or
B cells show defective CSR and SHM despite normal expression of
mature V HDJ H and germline I H-C H transcripts 31 . The full restoration
of CSR by enforced expression of AID in Hoxc4 –/– B cells indicates
that induction of AID is the main pathway through which HoxC4
regulates CSR and, probably, SHM. Mutation of the 5¢-ATTTGAAT-3¢
site to 5¢-GCTTGAATT-3¢, which does not disrupt the binding
of Oct and introduces a putative Sp-binding site, does not seem to
alter Aicda or AICDA promoter activity in mouse B lymphoma
M12 and CH33 cells or human embryonic kidney 293 cells, respectively
16 . In our experiments, mutation of 5¢-ATTTGAAT-3¢ to disrupt
HoxC4-binding activity while preserving the Oct-binding activity
or to abrogate the binding of Oct1-Oct2 but not HoxC4 binding 25
resulted in less Aicda transcription, thereby emphasizing the function
of these homeodomain transcription factors in AID expression
and adding another dimension to the function of Oct1-Oct2 in
B cell differentiation.
As we have shown in human B cells, HoxC4 serves an important
and complex function in the regulation of the IgH locus 24,25,35 .The
putative dampening effect of HoxC4 on the baseline activity of the
I g and I e promoters would be effectively lifted and overridden by
the strong activation of these promoters by CD40 signaling and
CD154-induced HoxC4-mediated activation of the hs1,2 enhancer
element 24,25,35 . The idea that HoxC4 is involved in the induction of
AID expression is further strengthened by the demonstration that
like AID, HoxC4 is expressed mainly in germinal center B cells of
both humans 24 and mice (as reported here), and that engagement
of CD40 by CD154 and treatment with cytokines, which induce the
expression of AICDA and Aicda, also induces HOXC4 and Hoxc4 in
human B cells 24 and mouse B cells (data presented here), respectively.
Furthermore, NF-kB-binding sites in both the human and
mouse HOXC4 and Hoxc4 promoters underpin the upregulation of
HoxC4 by CD40 signaling (unpublished data).
Consistent with the idea that the defect in CSR and SHM
manifested by Hoxc4 –/– B cells was due to a failure to induce
Aicda expression, both Aicda transcripts and AID protein, as
induced by LPS and IL-4 or CD154 and IL-4, were much less
abundant in Hoxc4 –/– B cells. The tissue and differentiation-stage
specificity of HoxC4 expression would account to a great extent for
the precise regulation of AID expression. In Hoxc4 –/– mice, the
lower AID expression was reflected in vivo in the impairment of the
maturation of the T cell–dependent antibody response. In these
mice, although Aicda expression was much lower, germinal center
formation seemed normal. That phenotype is reminiscent of that of
Aicda +/– mice 36 , which also have normal germinal centers in the
presence of much lower Aicda expression 37,38 , and contrasts with
that of Aicda / mice, in which activated B cells accumulate and
form giant germinal centers 36 .
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 547

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
Pax5 has been suggested to serve a function in Aicda expression by
binding to the 5¢-GGGGAGGAGCC-3¢ site in the Aicda promoter 16 .
This cis element, however, does not fulfill the requirements of a
consensus Pax5-binding motif and did not bind Pax5 in our experiments
and those by others 17 . Our finding of a lack of specificity of the
conserved 5¢-GGGGAGGAGCC-3¢ cis element for Pax5 suggests
that recruitment of Pax5 to the Aicda promoter, as showed by
chromatin-immunoprecipitation assay, occurred indirectly through
other DNA-binding transcription factors, perhaps Sp1 or Sp3, by a
mechanism similar to the interaction between the estrogen receptor
and Sp1 on certain estrogen-responsive promoters 39 .
The B cell development–related transcription factor E47 has been
suggested to contribute to the enhancement of Aicda promoter activity
by binding to E-boxes in cr2 of the first intron of Aicda 21,22 .E47induced
enhancement of Aicda promoter activity would be modulated
by the E2A inhibitor Id3 (ref. 15). The presence of cr2 in human
Ramos B cells and mouse CH12F3-2A B cells (as reported here) and in
mouse BaF/3 pro–B cells and M12 B cells 16 did not enhance transcription
of the luciferase reporter driven by the Aicda promoter. This
might reflect the dispensability of E2A transcription factors in the
expression of AICDA and Aicda, CSR and, possibly, SHM, as shown by
analysis of E2A-deficient B cells 40,41 or the muted baseline activity of
cr2 that resulted from ‘preferential’ binding of Id2 and/or Id3 to this
region. These experiments, however, cannot rule out the possibility
that like the IgH and immunoglobulin-k intronic enhancers, which
contain many E-box sites and show a high enhancer activity in
germinal center B cells 42 , cr2 and E proteins act together with
HoxC4 to synergistically induce AID expression.
The HoxC4-mediated activation of the Aicda promoter was
further enhanced by the upstream conserved 5¢-GGGGAG
GAGCC-3¢ site, which, as we have also shown, recruited Sp1, Sp3
and NF-kB. Sp1 and Sp3 bind directly to DNA through their zincfinger
motifs and enhance gene transcription. These proteins are
ubiquitously expressed and are directly involved not only in the
regulation of basal transcription and expression of ‘housekeeping’
genes but also in the expression of developmentally controlled
genes. In our experiments, the Sp–NF-kB–binding site was able to
partially mediate Aicda promoter activity in the absence of the
HoxC4-Oct–binding site and possibly accounted for the residual
AID expression, CSR and SHM in Hoxc4 –/– B cells and mice.
We have shown that Oca-B was also recruited to the AICDA and
Aicda promoters in B cells undergoing CSR or SHM, probably
through interaction with Oct1 and/or Oct2 (ref. 43). By clamping
together the POU H and POU S subdomains of Oct1 and Oct2, Oca-B
would increase the affinity of these homeodomain proteins for DNA,
thereby potentiating HoxC4- and Oct1-Oct2-mediated activation of
the Aicda promoter. Oca-B is important in HoxC4- and Oct1-Oct2mediated
activation of the human 3¢ Ea enhancer hs1,2 element 25 .In
agreement with that, mice that lack Oca-B have impaired CSR 43 .Our
results here have identified another function for Oct1-Oct2 and Oca-B
activity: regulating AID expression. Overall, our findings offer fundamental
insights into the mechanisms of activation of the AICDA and
Aicda promoters and induction of AID expression, CSR and SHM.
The possibility that the induction of HoxC4 by stimuli other than
CD40 signaling, LPS and cytokines, such as hormones (unpublished
data), modulates AID expression and, therefore, antibody diversification
in health and disease should be addressed.
METHODS
Hoxc4 –/– mice. In the Hoxc4 –/– mice used here (A.M. Boulet and
M.R. Capecchi), Hoxc4 was disrupted through insertion of a loxP siteinexon
2ofHoxc4 at the sequence encoding the amino-terminal end (between the
third and the fourth codons) of the homeobox, which introduces a stop codon
at the insertion site and yields a nonfunctional truncated protein (lacking 95%
of the homeodomain; unpublished observations). We obtained Hoxc4 +/– frozen
sperm and rederived Hoxc4 +/– mice by in vitro fertilization through the services
of the transgenic mouse facility of the University of California at Irvine. These
Hoxc4 +/– mice were on a C57BL/6 background after backcrossing of the
129Sv/Ev founder strain with C57BL/6 mice. Hoxc4 –/– mice and their wildtype
littermates were bred in pathogen-free conditions. The Institutional
Animal Care and Use Committee of University of California at Irvine approved
all animal experiments.
B cells and T cells. The number of B cells (B220 + ) and T cells (CD3 + ), the
proportion CD4 + T cells and dead B cells and T cells, and the proportion of
PNA hi B cells, plasma cells (B220 lo CD138 + ) and NP-binding CD38 hi IgG1 +
memory B cells was assessed by flow cytometry with a FACSCalibur (BD
Biosciences). Single-cell suspensions were prepared from spleens or Peyer’s
patches of Hoxc4 –/– and Hoxc4 +/+ mice and were stained with phycoerythrinlabeled
mAb to mouse B220 (RA3-6B2; BD Biosciences), fluorescein isothiocyanate
(FTIC)-labeled mAb to mouse CD3 (17A2; BioLegend), peridinine
chlorophyll protein complex–labeled mAb to mouse CD4 (GK1.5; BioLegend),
7-amino-actinomycin D (BD Biosciences), FITC-labeled peanut agglutinin
(PNA), FITC-labeled mAb to mouse CD138 (281-2) or allophycocyaninlabeled
anti–mouse IgG1 (X56; BD Biosciences), phycoerythrin-labeled NP
(Biosearch Technologies) and phycoerythrin-indotricarbocyanine–labeled mAb
to mouse CD38 (90; eBiosciences). Single-cell suspensions of B220 + cells were
prepared from spleens or Peyer’s patches with the EasySep Mouse B Cell
Enrichment kit (StemCell Technologies). For the preparation of PNA hi (germinal
center) B cells, spleen or Peyer’s patch B cells were stained with
phycoerythrin-labeled mAb to mouse CD45R and FITC-labeled PNA. Labeled
lymphocytes were then sorted with a MoFlo cell sorter (Dako), which yielded
PNA hi B220 + cells that were 95% pure.
B cell lines. Ramos B cells were monitored for spontaneous SHM 44 .Monoclonal
4B6 and 2E2 B cell lines were derived from the CSR- and SHM-inducible
human monoclonal IgM + IgD + CL-01 B cell line 45–50 by sequential subcloning
and selection for spontaneous and inducible CSR, respectively: 4B6 B cells are
IgM + IgD + with an ‘early’ germinal center phenotype and undergo spontaneous
CSR to IgG, IgE and IgA 24 ; 2E2 B cells are IgM + IgD + and undergo CSR to IgG,
IgE and IgA after stimulation with an agonistic mAb to human CD40 and
appropriate cytokines 51 . The mouse B lymphoma cell line CH12F3-2A was
obtained from T. Honjo. CH12F3-2A cells have surface expression of IgM and
switch to IgA after stimulation with CD154 or LPS in the presence of IL-4 and
TGF-b1 (ref. 52). All these monoclonal B cells were cultured in FBS-RPMI
(RPMI-1640 medium (Invitrogen) supplemented with 10% (vol/vol) heatinactivated
FBS (Hyclone), 2 mM L-glutamine and 1 antibiotic-antimycotic
mixture (penicillin (100 units/ml), streptomycin (100 mg/ml) and amphotericin
B fungizone (0.25 mg/ml); Invitrogen)).
B cell cycle and proliferation. Cell cycle was analyzed by staining with
propidium iodide 50 . Proliferation was analyzed with the CellTrace CFSE Cell
Proliferation kit (Molecular Probes). Cells were washed in serum-free Hank’s
balanced-salt solution (Invitrogen) and were resuspended at a density of 1
10 6 cells per ml. After the addition of an equal volume of 2.4 mM CFSE
(carboxyfluorescein succinimidyl ester), cells were incubated for 12 min at
37 1C and then were washed in FBS-RPMI. Cells were then diluted and were
cultured in the presence or absence of LPS (20 mg/ml) from Escherichia coli
(serotype 055:B5; Sigma-Aldrich) and recombinant mouse IL-4 (5 ng/ml; R&D
Systems), then were collected at various times after activation and analyzed by
flow cytometry. For in vivo B cell proliferation, mice were immunized
with NP 16-CGG. After 10 d, they were injected intraperitoneally twice within
16 h with 1 mg BrdU (5-bromodeoxyuridine) and were killed 4 h after the last
injection. Cells from the spleen or Peyer’s patches were stained with
phycoerythrin-labeled mAb to mouse B220 (BD Biosciences) or that mAb
together with FITC-labeled PNA. Incorporated BrdU was stained with
allophycocyanin-labeled mAb to BrdU with the APC BrdU Flow kit (BD
Biosciences) and were analyzed by flow cytometry.
548 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Analysis of in vitro CSR. Enriched spleen B cells were cultured at a density of
1 10 6 cell/ml in FBS-RPMI with 0.05 mM b-mercaptoethanol. Cells were
stimulated with LPS (20 mg/ml) or CD154-expressing membrane fragments of
baculovirus-infected Sf21 insect cells (called ‘CD154’ here) 53 and the following
reagents: nothing more for CSR to IgG3 and IgG2b; recombinant mouse IL-4
(5 ng/ml) for CSR to IgG1 and IgE; IFN-g (50 ng/ml; PeproTech) for CSR to
IgG2a; or TGF-b1 (1 ng/ml; R&D Systems), recombinant mouse IL-5 (5 ng/ml;
R&D Systems) and dextran-conjugated mAb to d-chain (3 ng/ml; provided by
C.M. Snapper) for CSR to IgA. Cells were collected on day 4 for analysis of
surface immunoglobulins after being stained with FITC-labeled rat mAb to
mouse IgG1 (A85-1), mouse IgG2a (R19-15), mouse IgG2b (R12-3), mouse
IgG3 (R40-82) or mouse IgA (C10-3), or phycoerythrin-labeled rat mAb to
mouse CD45R (B220; RA3-6B2; all from BD Biosciences). Cells were fixed with
1% (vol/vol) paraformaldehyde in PBS and were analyzed by flow cytometry.
Specific enzyme-linked immunosorbent assays involving 96-well plates coated
with polyclonal goat antibody F(ab¢) 2 to the appropriate mouse isotype
(Southern Biotechnology Associates) were used to measure IgG1, IgG2a, IgG2b,
IgG3, IgA and IgE in culture supernatants of in vitro–stimulated spleen
Hoxc4 +/+ and Hoxc4 –/– B cells. Supernatants were serially diluted twofold from
1:5 to 1:640 and then were added to the plates (100 ml per well), which were
incubated for 1 h at 25 1C. After plates were washed, biotin-labeled isotypespecific
mAbs were added, followed by visualization with horseradish
peroxidase–streptavidin (Supplementary Methods online). Concentrations of
immunoglobulin isotypes were determined by interpolation with a calibrated
standard curve for each isotype. Assays were done in triplicate.
Quantitative real-time RT-PCR and semiquantitative RT-PCR. RNA was
extracted with the RNeasy Mini Kit according to the manufacturer’s protocol
(Qiagen). Residual DNA was removed by treatment with DNase I (Invitrogen).
First-strand cDNA was synthesized from total RNA (2 mg for each experiment)
with the SuperScript Preamplification system and oligo(dT) primer (Invitrogen).
The expression of germline IH-CH, post-recombination Im-CH, mature
VJ558DJH-Cm, Hoxc4 and Aicda transcripts was quantified by real-time quantitative
RT-PCR 54 with the appropriate primers (Supplementary Table 1 online;
Operon). In some cases, circle I g1-C m transcripts, HOXC4, Hoxc4, AICDA and
Aicda transcripts were analyzed by specific semiquantitative RT-PCR with serial
twofold dilutions so there was a nearly linear relationship between the amount
of cDNA used and the intensity of the PCR product. A DNA Engine Opticon 2
Real-Time PCR Detection system (Bio-Rad Laboratories) was used for realtime
quantitative RT-PCR to measure the incorporation of SYBR Green
(DyNAmo HS kit; New England Biolabs) according to the following protocol:
50 1C for 2 min and 95 1C for 10 min; then 40 cycles of 95 1C for 10 s, 60 1Cfor
20 s, 72 1C for 30 s and 80 1C for 1 s; data acquisition at 80 1C; and then 72 1C
for 10 min. Melting-curve analysis was done from 72 1C to951C and samples
were incubated for another 5 min at 72 1C. The change in cycling threshold
(DDC T) method was used for data analysis.
Analysis of somatic mutations in intronic JH4-iEl DNA. Peyer’s patch B cells
were obtained from unimmunized 12-week-old littermate Hoxc4 –/– and
Hoxc4 +/+ C57BL/6 mice and were used for analysis of somatic mutations in
intronic DNA downstream of rearranged VJ558DJH4 genes. Platinum Pfx DNA
polymerase (Invitrogen) was used for amplification of genomic DNA. The
intronic IgH region downstream of rearranged V J558DJ H was amplified by
nested PCR 55 involving two V H J558 framework region 3–specific forward
primers and two reverse primers specific for sequences downstream of JH4
(Supplementary Table 1), which yielded DNA of approximately 900 bp if
aJ H4 rearrangement occurred. PCR conditions were 35 cycles of 94 1C for45s,
58 1C for 45 s and 68 1C for 1 min. PCR products were cloned into the
pCR-Blunt II-TOPO vector (Invitrogen) and sequenced. Only sequences from
rearrangements involving JH4-iEm were analyzed. Sequences were analyzed with
MacVector 7.2.3 software.
Identification of putative transcription factor–binding sites. Putative transcription
factor–binding sites in Aicda promoter sequence were identified
by weight-matrix search with Match software (BIOBASE) which integrates
the TRANSFAC 6.0 database and uses its positional weight matrices
for analysis. Scores indicate the degree of fitness of the putative binding
ARTICLES
site with the consensus sequence according to the following parameters:
a score 1.0 is 100%; the cut-off score is 0.75.
Aicda promoter–luciferase reporter assays. Reporter constructs consisted of
the pGL3-Basic (Fig. 7d) or pGL3-Enhancer (Fig. 8d) firefly luciferase reporter
vector (Promega) and various sequences from the Aicda locus. Sequences
consisting of 839 bp of flanking 5¢ region plus the Aicda promoter, 490 bp of
flanking 5¢ region, or 349 bp of the Aicda promoter, as well as cr1 and cr2 of the
first intron, were amplified by PCR from C57BL/6 mouse genomic DNA and
were inserted upstream and/or downstream of the luciferase gene in pGL3
vector. Various mutant reporters were constructed with the QuikChange Site-
Directed Mutagenesis kit (Stratagene). Sequences of constructs were confirmed
by at least two sequencing reactions. The reporter construct and the constitutively
active Renilla reniformis luciferase–producing vector pRL-TK (Promega)
were transfected together into human Ramos and 4B6 B cells and mouse
CH12F3-2A B cells by electroporation24,25,35 . Firefly and renilla luciferase
activity was measured with the Dual-Luciferase Reporter Assay system according
to the manufacturer’s instructions (Promega).
Chromatin immunoprecipitation. B cells (5 10 7 )weretreatedfor10minat
25 1C with 1% (vol/vol) formaldehyde for crosslinkage of chromatin. After cells
were washed with cold PBS containing protease inhibitors (Roche), chromatin
was separated with nuclear lysis buffer (10 mM Tris-HCl, 1 mM EDTA, 0.5 M
NaCl, 1% (vol/vol) Triton X-100, 0.5% (wt/vol) sodium deoxycholate and 0.5%
(wt/vol) sarcosyl, pH 8.0) and was resuspended in IP-1 buffer (20 mM Tris-
HCl, pH 8.0, 200 mM NaCl, 2 mM EDTA, 0.1% (wt/vol) sodium deoxycholate,
0.1% (wt/vol) SDS and protease inhibitors). Chromatin was sonicated to yield
DNA fragments approximately 200–1,000 bp in size, was precleared with
agarose beads bearing protein G (Santa Cruz Biotechnology) and then was
incubated overnight at 4 1C with mAb to HoxC4 (1E9; Novus Biologicals) or
rabbit polyclonal anti-Oct1 (C-21), anti-Oct2 (C-20), Oca-B (C-20), anti-Pax5
(N-19), anti-Sp1 (H-225), anti-Sp3 (D-20) or anti–NF-kB p52 (K-27; all from
Santa Cruz Biotechnology). After that incubation, immune complexes were
isolated with agarose beads bearing protein G, were eluted with elution buffer
(50 mM Tris-HCl, 0.5% (vol/vol SDS, 200 mM NaCl and 100 mg/ml of
proteinase K, pH 8.0) and then were incubated overnight at 65 1C forreversal
of the formaldehyde crosslinks. DNA was extracted with phenol-chloroform
and precipitated with ethanol and then was resuspended in 10 mM Tris-HCl,
pH 8.0, and 1 mM EDTA. DNA sequences were confirmed by PCR with the
appropriate primers (Supplementary Table 1).
Retroviral transduction of B cells. The TAC and TAC-Aicda retroviral
constructs 15 were obtained from C. Murre. For the generation of retrovirus,
pCSretTAC-based constructs were transfected into the HEK293T packaging cell
line with the ProFection Mammalian Transfection system (Promega). The
retroviral constructs were used to transduce mouse spleen B cells as reported 15 .
Additional methods. Information on immunoblot analysis, immunization
with NP16-CGG and titration of NP-binding IgM and IgG1, histology, and
EMSA and EMSA supershift-blocking assays is available in the Supplementary
Methods online.
Statistical analyses. Differences in the frequency and spectrum of mutations in
Hoxc4 –/– and Hoxc4 +/+ mice were analyzed with the chi-squared test. Differences
in immunoglobulin titers, CSR and mRNA expression were analyzed with
paired t-tests.
Accession codes. UCSD-Nature Signaling Gateway (http://www.signalinggateway.org):
A000804, A001704, A001696, A000536, A001262, A001238 and
A002271.
Note: Supplementary information is available on the Nature Immunology website.
ACKNOWLEDGMENTS
We thank M.R. Capecchi and A.M. Boulet (University of Utah) for Hoxc4 +/–
frozen mouse sperm; T. Fielder for technical efforts; L. Khidr, L. Phan,
B. Gupta, J. Feng and Y. Zhong for handling Hoxc4 +/– mice; C. Murre
(University of California, San Diego) for the Aicda retroviral construct,
T. Honjo (Kyoto University) for CH12F3-2A cells; C.M. Snapper (Uniformed
Services University of the Health Sciences) for dextran-conjugated mAb to
NATURE IMMUNOLOGY VOLUME 10 NUMBER 5 MAY 2009 549

© 2009 Nature America, Inc. All rights reserved.
ARTICLES
d-chain; Z. Yu for statistic analysis; A. Schaffer for discussions; and S. Sabet and
M. Kang for technical assistance. Supported by the US National Institutes of
Health (AI 045011, AI 079705 and AI 060573 to P.C.).
AUTHOR CONTRIBUTIONS
S.-R.P., H.Z., Z.P., J.Z., A.A.-Q., E.J.P., Z.X. and T.M. did experiments; H.Z.
designed experiments, analyzed data and prepared the manuscript; and P.C.
designed all experiments, analyzed the data, supervised the work and prepared
the manuscript.
Published online at http://www.nature.com/natureimmunology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Honjo, T., Kinoshita, K. & Muramatsu, M. Molecular mechanism of class switch
recombination: linkage with somatic hypermutation. Annu. Rev. Immunol. 20,
165–196 (2002).
2. Honjo, T., Muramatsu, M. & Fagarasan, S. Aid: How does it aid antibody diversity?
Immunity 20, 659–668 (2004).
3. Neuberger, M.S., Harris, R.S., Di Noia, J. & Petersen-Mahrt, S.K. Immunity through
DNA deamination. Trends Biochem. Sci. 28, 305–312 (2003).
4. Chaudhuri, J. & Alt, F.W. Class-switch recombination: interplay of transcription, DNA
deamination and DNA repair. Nat. Rev. Immunol. 4, 541–552 (2004).
5. Rada, C., Di Noia, J.M. & Neuberger, M.S. Mismatch recognition and uracil excision
provide complementary paths to both Ig switching and the A/T-focused phase of
somatic mutation. Mol. Cell 16, 163–171 (2004).
6. Maizels, N. Immunoglobulin gene diversification. Annu. Rev. Genet. 39, 23–46
(2005).
7. Di Noia, J.M. & Neuberger, M.S. Molecular mechanisms of antibody somatic hypermutation.
Annu. Rev. Biochem. 76, 1–22(2007).
8. Peled, J.U. et al. The biochemistry of somatic hypermutation. Annu. Rev. Immunol. 26,
481–511 (2008).
9. Diaz, M. & Casali, P. Somatic immunoglobulin hypermutation. Curr. Opin. Immunol.
14, 235–240 (2002).
10. Papavasiliou, F.N. & Schatz, D.G. Somatic hypermutation of immunoglobulin genes;
merging mechanisms for genetic diversity. Cell 109, s35–s44 (2002).
11. Wu, X. et al. Immunoglobulin somatic hypermutation: double-strand DNA breaks, AID
and error-prone DNA repair. J. Clin. Immunol. 23, 235–246 (2003).
12. Xu, Z. et al. DNA lesions and repair in immunoglobulin class switch recombination and
somatic hypermutation. Ann. NY Acad. Sci. 1050, 146–162 (2005).
13. Casali, P., Pal, Z., Xu, Z. & Zan, H. DNA repair in antibody somatic hypermutation.
Trends Immunol. 27, 313–321 (2006).
14. Yan, C.T. et al. IgH class switching and translocations use a robust non-classical endjoining
pathway. Nature 449, 478–482 (2007).
15. Sayegh, C.E., Quong, M.W., Agata, Y. & Murre, C. E-proteins directly regulate expression
of activation-induced deaminase in mature B cells. Nat. Immunol. 4, 586–593
(2003).
16. Gonda, H. et al. The balance between Pax5 and Id2 activities is the key to AID gene
expression. J. Exp. Med. 198, 1427–1437 (2003).
17. Yadav, A. et al. Identification of a ubiquitously active promoter of the murine
activation-induced cytidine deaminase (AICDA) gene. Mol. Immunol. 43, 529–541
(2006).
18. Pearson, J.C., Lemons, D. & McGinnis, W. Modulating Hox gene functions during
animal body patterning. Nat. Rev. Genet. 6, 839–904 (2005).
19. Zakany, J. & Duboule, D. The role of Hox genes during vertebrate limb development.
Curr. Opin. Genet. Dev. 17, 359–366 (2007).
20. Deschamps, J. & van Nes, J. Developmental regulation of the Hox genes during axial
morphogenesis in the mouse. Development 132, 2931–2942 (2005).
21. Lemons, D. & McGinnis, W. Genomic evolution of Hox gene clusters. Science 313,
1918–1922 (2006).
22. Bijl, J. et al. Expression of HOXC4, HOXC5, and HOXC6 in human lymphoid cell lines,
leukemias, and benign and malignant lymphoid tissue. Blood 87, 1737–1745
(1996).
23. Meazza, R. et al. Expression of HOXC4 homeoprotein in the nucleus of activated human
lymphocytes. Blood 85, 2084–2090 (1995).
24. Schaffer, A. et al. Selective inhibition of class switching to IgG and IgE by recruitment
of the HoxC4 and Oct-1 homeodomain proteins and Ku70/Ku86 to newly identified
ATTT cis-elements. J. Biol. Chem. 278, 23141–23150 (2003).
25. Kim, E.C., Edmonston, C.R., Wu, X., Schaffer, A. & Casali, P. The HoxC4 homeodomain
protein mediates activation of the immunoglobulin heavy chain 3¢ hs1,2 enhancer in
human B cells. Relevance to class switch DNA recombination. J. Biol. Chem. 279,
42258–42269 (2004).
26. Pasqualucci, L. et al. Expression of the AID protein in normal and neoplastic B cells.
Blood 104, 3318–3325 (2004).
27. Guikema, J.E. et al. Quantitative RT-PCR analysis of activation-induced cytidine
deaminase expression in tissue samples from mantle cell lymphoma and B-cell chronic
lymphocytic leukemia patients. Blood 105, 2997–2998 (2005).
28. Bijl, J.J. et al. HOXC4, HOXC5, and HOXC6 expression in non-Hodgkin’s lymphoma:
preferential expression of the HOXC5 gene in primary cutaneous anaplastic T-cell and
oro-gastrointestinal tract mucosa-associated B-cell lymphomas. Blood 90, 4116–4125
(1997).
29. Boulet, A.M. & Capecchi, M.R. Targeted disruption of hoxc-4 causes esophageal
defects and vertebral transformations. Dev. Biol. 177, 232–249 (1996).
30. Saegusa, H., Takahashi, N., Noguchi, S. & Suemori, H. Targeted disruption in the
mouse Hoxc-4 locus results in axial skeleton homeosis and malformation of the xiphoid
process. Dev. Biol. 174, 55–64 (1996).
31. Xu, Z. et al. Regulation of aicda expression and AID activity: relevance to somatic
hypermutation and class switch DNA recombination. Crit.Rev.Immunol.27, 367–397
(2007).
32. Crouch, E.E. et al. Regulation of AID expression in the immune response. J. Exp. Med.
204, 1145–1156 (2007).
33. Teng, G. et al. MicroRNA-155 is a negative regulator of activation-induced cytidine
deaminase. Immunity 28, 621–629 (2008).
34. Dedeoglu, F., Horwitz, B., Chaudhuri, J., Alt, F.W. & Geha, R.S. Induction of activationinduced
cytidine deaminase gene expression by IL-4 and CD40 ligation is dependent
on STAT6 and NFkB. Int. Immunol. 16, 395–404 (2004).
35. Schaffer, A., Cerutti, A., Shah, S., Zan, H. & Casali, P. The evolutionary conserved
sequence upstream of the human Ig Sg3 region is an inducible promoter: Synergistic
activation by CD40 ligand and IL-4 via cooperative NF-kB and STAT-6 binding sites.
J. Immunol. 162, 5327–5336 (1999).
36. Muramatsu, M. et al. Class switch recombination and hypermutation require activationinduced
cytidine deaminase (AID), a potential RNA editing enzyme. Cell 102,
553–563 (2000).
37. Jiang, C., Zhao, M.L. & Diaz, M. Activation-induced deaminase heterozygous MRL/lpr
mice are delayed in the production of high-affinity pathogenic antibodies and in the
development of lupus nephritis. Immunology 126, 102–113 (2009).
38. Takizawa, M. et al. AID expression levels determine the extent of cMyc oncogenic
translocations and the incidence of B cell tumor development. J. Exp. Med. 205,
1949–1957 (2008).
39. Saville, B. et al. Ligand-, cell-, and estrogen receptor subtype (a/b)-dependent
activation
at GC-rich (Sp1) promoter elements. J. Biol. Chem. 275, 5379–5387 (2000).
40. Kwon, K. et al. Instructive role of the transcription factor E2A in early B lymphopoiesis
and germinal center B cell development. Immunity 28, 751–762 (2008).
41. Schoetz, U., Cervelli, M., Wang, Y.-D., Fiedler, P. & Buerstedde, J.-M. E2A expression
stimulates Ig hypermutation. J. Immunol. 177, 395–400 (2006).
42. Odegard, V.H., Kim, S.T., Anderson, S.M., Shlomchik, M.J. & Schatz, D.G. Histone
modifications associated with somatic hypermutation. Immunity 23, 101–110 (2005).
43. Teitell, M.A. OCA-B regulation of B-cell development and function. Trends Immunol.
24, 546–553 (2003).
44. Sale, J.E. & Neuberger, M.S. TdT-accessibe breaks are scattered over the immunoglobulin
V domain in a constitutively hypermutating B cell line. Immunity 9, 859–869
(1998).
45. Cerutti, A. et al. CD40 ligand and appropriate cytokines induce switching to IgG, IgA,
and IgE and coordinated germinal center and plasmacytoid phenotypic differentiation
in a human monoclonal IgM + IgD + Bcellline.J. Immunol. 160, 2145–2157 (1998).
46. Zan, H. et al. Induction of Ig somatic hypermutation and class switching in a human
monoclonal IgM + IgD + cell line in vitro: Definition of the requirements and the
modalities of hypermutation. J. Immunol. 162, 3437–3447 (1999).
47. Zan, H., Cerutti, A., Schaffer, A., Dramitinos, P. & Casali, P. CD40 engagement triggers
switching to IgA1 and IgA2 in human B cells through induction of endogenous TGF-b.
Evidence for TGF-b but not IL-10-dependent direct Sm-Sa and sequential Sm-Sg,
Sg-Sa DNA recombination. J. Immunol. 161, 5217–5225 (1998).
48. Zan, H. et al. The translesion DNA polymerase z plays a major role in Ig and Bcl-6
somatic hypermutation. Immunity 14, 643–653 (2001).
49. Zan, H. et al. BCR engagement and T cell contact induce Bcl-6 hypermutation in
human B cells: association with initiation of transcription and identity with immunoglobulin
hypermutation. J. Immunol. 165, 830–839 (2000).
50. Zan, H., Wu, X., Komori, A., Holloman, W.K. & Casali, P. AID-dependent generation of
resected double-strand DNA breaks and recruitment of Rad52/Rad51 in somatic
hypermutation. Immunity 18, 727–738 (2003).
51. Zan, H. & Casali, P. AID- and Ung-dependent generation of staggered double-strand
DNA breaks in immunoglobulin class switch DNA recombination: A post-cleavage role
for AID. Mol. Immunol. 46, 45–61 (2008).
52. Muramatsu, M. et al. Specific expression of activation-induced cytidine deaminase
(AID), a novel member of the RNA-editing deaminase family in germinal center B cells.
J. Biol. Chem. 274, 18470–18476 (1999).
53. Rush, J.S., Liu, M., Odegard, V.H., Unniraman, S. & Schatz, D.G. Expression of
activation-induced cytidine deaminase is regulated by cell division, providing a
mechanistic basis for division-linked class switch recombination. Proc. Natl. Acad.
Sci. USA 102, 13242–13247 (2005).
54. Zan, H. et al. Lupus-prone MRL/faslpr/lpr mice display increased AID expression and
extensive DNA lesions, comprising deletions and insertions, in the immunoglobulin
locus: Concurrent upregulation of somatic hypermutation and class switch DNA
recombination. Autoimmunity 42, 89–103 (2009).
55. Zan, H. et al. The translesion DNA polymerase y plays a major role in immunoglobulin
gene somatic hypermutation. EMBO J. 24, 3757–3769 (2005).
550 VOLUME 10 NUMBER 5 MAY 2009 NATURE IMMUNOLOGY

© 2009 Nature America, Inc. All rights reserved.
Corrigendum: ADAR1 is essential for the maintenance of hematopoiesis and
suppression of interferon signaling
Jochen C Hartner, Carl R Walkley, Jun Lu & Stuart H Orkin
Nat. Immunol. 10, 109–115 (2009); published online 7 December 2008; corrected after print 9 April 2009
CORRIGENDA AND ERRAtA
In the version of this article initially published, the Cre-transgenic mouse is identified incorrectly as Tg(SV40-cre)1Jrg. The correct mouse strain
should be Tg(SCL6E5-Cre)1Jrg, and the citation describing this mouse (ref. 29) should be as follows: Gothert, J.R. et al. In vivo fate-tracing studies
using the Scl stem cell enhancer: embryonic hematopoietic stem cells significantly contribute to adult hematopoiesis. Blood 105, 2724–2732 (2005).
The error has been corrected in the HTML and PDF versions of the article.
Erratum: IL-4 inhibits TGF-β-induced Foxp3 + T cells and, together with
TGF-β, generates IL-9 + Foxp3 – effector T cells
Valérie Dardalhon, Amit Awasthi, Hyoung Kwon, George Galileos, Wenda Gao, Raymond A Sobel, Meike Mitsdoerffer, Terry B Strom,
Wassim Elyaman, I-Cheng Ho, Samia Khoury, Mohamed Oukka & Vijay K Kuchroo
Nat. Immunol. 9, 1347–1355 (2008); published online 9 November 2008; corrected after print 9 April 2009
In the version of this article initially published, graph axes are mislabeled in Figures 1e, 2d and 2e, and a gene symbol is misidentified in the legend
to Figure 2. The correct axis labels should be as follows: Figure 1e, middle right and far right vertical axes should end “(relative × 10 3 )”; Figure
2d,e, left vertical axes should read “Il9 mRNA (relative × 10 2 )”; Figure 2e, right vertical axis should read “Il10 mRNA (relative)”; and Figure 2e,
horizontal axes should read “WT” (in place of “Foxp3-GFP”) and “GATA3-KO” (in place of “STAT6-KO.Fox3-GFP”). The legend for Figure 2d
should state “relative to Hprt1 mRNA” and the legend to Figure 2e should have that phrase removed. The errors have been corrected in the HTML
and PDF versions of the article.
nature immunology volume 10 number 5 may 2009 551

© 2009 Nature America, Inc. All rights reserved.
Corrigendum: ADAR1 is essential for the maintenance of hematopoiesis and
suppression of interferon signaling
Jochen C Hartner, Carl R Walkley, Jun Lu & Stuart H Orkin
Nat. Immunol. 10, 109–115 (2009); published online 7 December 2008; corrected after print 9 April 2009
CORRIGENDA AND ERRAtA
In the version of this article initially published, the Cre-transgenic mouse is identified incorrectly as Tg(SV40-cre)1Jrg. The correct mouse strain
should be Tg(SCL6E5-Cre)1Jrg, and the citation describing this mouse (ref. 29) should be as follows: Gothert, J.R. et al. In vivo fate-tracing studies
using the Scl stem cell enhancer: embryonic hematopoietic stem cells significantly contribute to adult hematopoiesis. Blood 105, 2724–2732 (2005).
The error has been corrected in the HTML and PDF versions of the article.
Erratum: IL-4 inhibits TGF-β-induced Foxp3 + T cells and, together with
TGF-β, generates IL-9 + Foxp3 – effector T cells
Valérie Dardalhon, Amit Awasthi, Hyoung Kwon, George Galileos, Wenda Gao, Raymond A Sobel, Meike Mitsdoerffer, Terry B Strom,
Wassim Elyaman, I-Cheng Ho, Samia Khoury, Mohamed Oukka & Vijay K Kuchroo
Nat. Immunol. 9, 1347–1355 (2008); published online 9 November 2008; corrected after print 9 April 2009
In the version of this article initially published, graph axes are mislabeled in Figures 1e, 2d and 2e, and a gene symbol is misidentified in the legend
to Figure 2. The correct axis labels should be as follows: Figure 1e, middle right and far right vertical axes should end “(relative × 10 3 )”; Figure
2d,e, left vertical axes should read “Il9 mRNA (relative × 10 2 )”; Figure 2e, right vertical axis should read “Il10 mRNA (relative)”; and Figure 2e,
horizontal axes should read “WT” (in place of “Foxp3-GFP”) and “GATA3-KO” (in place of “STAT6-KO.Fox3-GFP”). The legend for Figure 2d
should state “relative to Hprt1 mRNA” and the legend to Figure 2e should have that phrase removed. The errors have been corrected in the HTML
and PDF versions of the article.
nature immunology volume 10 number 5 may 2009 551