Xenogen IVIS-200 reducing background autofluorescence technical ...

Methods for Reducing the Effects of Background
Autofluorescence Using IVIS Imaging Technology
T.L. Troy and B.W. Rice
Xenogen Corporation
December 7, 2004
Introduction
The ability to track fluorescent proteins, dyes, quantum dots, and other fluorescent
reporters in vivo is a rapidly growing technology known as molecular imaging. The IVIS
Imaging Systems developed at Xenogen are optimized for imaging both bioluminescent
and fluorescent reporters in vivo. Fluorescence equipment is standard on the IVIS
Imaging System 200 Series while the IVIS Imaging System 100 or 50 Series is available
with the XFO-12 Fluorescence Option. Table 1 summarizes the standard four fluorescent
filter sets raging from the green to the near infrared (NIR).
One of the challenges associated with in vivo fluorescence imaging is to minimize the
effects of background fluorescence, or autofluorescence. Autofluorescence is a
fluorescent signal originating from substances other than the fluorophore of interest. The
level of autofluorescence ultimately determines the limit of detection for fluorescent
probes. Sources of autofluorescence include both an instrument component due to the
optics and filters of the imaging system and a biological component from the animal
tissue itself. Although tissue autofluorescence is typically much higher than instrument
autofluorescence, particularly in the visible wavelength range, it is desirable to minimize
all sources of autofluorescence in order to increase detection sensitivity. This technical

ulletin describes techniques developed at Xenogen to reduce the effects of
autofluorescence using IVIS Imaging technology and the Living Image software.
Units for Fluorescent Images
Before discussing methods for subtracting autofluorescent background, it is useful to
understand how fluorescent images are displayed and quantified in the Living Image
software.
IVIS Imaging Systems are calibrated relative to a National Institutes of
Standards and Technology traceable radiance standard so that images measured in
relative light units (CCD camera counts) can be converted to physical units of surface
radiance (photons/sec/cm 2 /steradian).
Fluorescent images have an additional
complication, however, because the measured surface radiance depends on the
illumination intensity. IVIS Imaging Systems use a top illumination scheme (light shines
down on the subject), which means that the illumination intensity can vary somewhat
across the field-of-view and will change as the field-of-view changes. To reduce the
effects of these illumination variations, fluorescent images can be normalized by dividing
the fluorescent image by a reference illumination image. The reference illumination
image is an image of a white plate taken with the excitation filter in place and an open (or
neutral density) emission filter; this image is saved on disk prior to shipping the
instrument. The resulting “normalized” fluorescent image is unitless and is called a
fluorescent efficiency image.
A fluorescent efficiency image is independent of the
illumination intensity. The value of each pixel in an efficiency image represents the
fractional ratio of fluorescent emitted photons per incident excitation photon. Typical
values for fluorescent efficiency images are very low, in the range of 10 -7 – 10 -3 [1].

Instrumentation Background Subtraction
Through the use of fused silica optics, low-autofluorescence materials, and high quality
filters, IVIS
Imaging Systems are designed to minimize autofluorescence and
background caused by instrumentation. However, because of the highly sensitive CCD
camera utilized in these instruments, a residual background may be detected during
fluorescent imaging. The background is often higher near the edges and corners of the
CCD, resulting in a ring-shaped pattern as shown in Figure 1b. The Living Image
software enables you to subtract this background using the following procedure. First
take a fluorescent image of the subject (Figure 1a) and evaluate whether the background
is significant. If it is beneficial to subtract the background, remove the subject from the
imaging chamber and take an image of the empty chamber using the same imaging
parameters as the original measurement (for example, time, binning, f/stop, FOV, lamp
level) (Figure 1b).
If a background image is available, the Living Image software provides the “Sub Fluor
Bkg” check box (Figure 1c, upper right). If this check box is selected, the background
(Figure 1b) is subtracted from the original image (Figure 1a), resulting in the final
background-subtracted image shown in Figure 1c. This procedure reduced the ring that
was observed in Figure 1a. Note also that the autofluorescence from the anesthesia nose
cone was also reduced since the manifold was inside the IVIS imaging chamber during
the instrument background measurement.

Tissue Autofluorescence
Autofluorescence from animal tissue is usually much higher than instrument background,
particularly in the visible wavelength range.
In vivo autofluorescence can appear
throughout the internal organs of the animal, but is brightest near the surface where the
excitation light is the strongest. Tissue autofluorescence is caused by the endogenous
chromophores in animal tissue, including elastin, collagen, tryptophan, NADH,
porphyrins, and flavins. These chromophores have absorption bands centered from the
ultraviolet to blue regions of the spectrum and emission bands throughout the visible to
near-infrared spectrum [2]. In addition to tissue autofluorescence, it has been observed
that chlorophyll from the alfalfa in standard mouse food also fluoresces in the near
infrared part of the spectrum [1].
Figure 2 shows autofluorescent images of control mice (no exogenous fluorophore
present) fed a regular diet that contains alfalfa (Figure 2a) or an alfalfa-free diet (Figure
2b). The images were taken using the four filter sets listed in Table 1. The biggest
distinction between the two sets of data is the large autofluorescent signal concentrated in
the intestinal area of the animal that was observed using the Cy5.5 and ICG filter sets.
This signal is due to the chlorophyll in the regular diet. After the diet was changed to the
alfalfa-free version, the signal in the intestinal area decreased to levels similar to the rest
of the body. Table 2 lists the values of the average tissue autofluorescent efficiency
measured for each filter set. These data show a general decrease in signal with increasing
wavelength and show that the alfalfa-free diet not only lowers the intestinal signal for

Cy5.5, but also slightly lowers the overall autofluorescence for the GFP and DsRed filter
sets.
Figure 3 shows the autofluorescent excitation and emission spectra from the surface
tissue of a living nude mouse for each filter passband. The excitation spectrum for each
filter set was measured using the center wavelength of the corresponding emission filter;
likewise, the emission spectrum was measured using the center wavelength of the
excitation filter for each filter passband. All of the data were then normalized to the
autofluorescent efficiency of GFP listed in Table 2. These data, measured on the lower
dorsal region of a mouse fed a regular diet (Figure 3a) or an alfalfa-free diet (Figure 3b),
also demonstrate a general decrease in autofluorescence towards the NIR region of the
spectrum. The chlorophyll emission peak, centered around 670 nm, is clearly evident in
the regular diet spectrum.
Autofluorescent Subtraction Using Background Filters
The high levels of tissue autofluorescence shown in Figure 2 limit the sensitivity of
detection of exogenous fluorophores, particularly in the visible wavelength range from
400 to 700 nm. Even in the NIR, there is still a low level of autofluorescence. Therefore
it is desirable to be able to subtract out the tissue autofluorescence from a fluorescent
measurement. The IVIS Imaging Systems implement a subtraction method based on the
use of blue-shifted background filters. The objective of these blue-shifted excitation
filters or “background” filters (see Table 1), is to excite the tissue autofluorescence
without exciting the fluorophore. The background filter image can then be subtracted

from the primary excitation filter image using an appropriate scale factor, thus reducing
autofluorescence. The assumption here is that the tissue excitation spectrum is much
broader than the excitation spectrum of the fluorophore of interest and that the spatial
distribution of autofluorescence does not vary much with small shifts in the excitation
wavelength.
Figure 4 shows a typical example of filter passbands and fluorescent
spectra. The fluorescent dye spectra (for PKH26 in this case) is shown by the dashed
curves. The excitation and emission filter passbands are set up to match the excitation
and emission spectra for the dye. The blue-shifted background filter excites only a small
part of the PHK26 curve, yet strongly excites the tissue autofluorescence (solid curve).
The background filter tissue autofluorescent correction procedure is as follows. First,
acquire two images, one using the background filter and one using the primary excitation
filter (use the same emission filter for both measurements).
The image sequence
acquisition can be automated using the Sequential Imaging option in the Living Image
software. Next, determine the scale factor by taking the ratio of the autofluorescent
signal measured using the background filter to the autofluorescent signal measured using
the excitation filter in a region on the animal with no fluorophore present. The scale
factor accounts for different levels of tissue autofluorescence due to different excitation
wavelengths and filter transmission characteristics. Subtract the background image from
the primary image using the Image Math panel in the Living Image software (Figure 5).
This panel shows the individual click numbers of the images that were loaded in the
sequence. The images on the right correspond to the highlighted image numbers on the
left. Choose the subtraction mode, A-Bk, from the Result = drop-down list and enter the

value for the scale factor, k, in the variable control box. To display the corrected image,
click Display Results for Measuring. The corrected image results in a significant
reduction in autofluorescence while leaving most of the fluorophore signal.
This procedure is illustrated in Figure 6 for all four filter sets using a control mouse (12-
week old female nude (Nu/nu) mouse on an alfalfa-free diet). The scale factor, k, for each
filter set is shown in the figure and was determined from the average tissue
autofluorescence over the entire animal. Table 3 lists the average tissue autofluorescence
and the root mean square (RMS) error measured before and after filter background
subtraction. Because the background filter subtraction technique essentially reduces the
autofluorescent signal to zero, the RMS error is used to quantify the tissue
autofluorescent value.
Table 3 also includes the improvement factor that could be
expected for this example. The improvement was calculated by taking the ratio of the
average autofluorescence before correction to the RMS error after correction.
Figure 7 shows an example of this technique using a fluorescent maker. In this example,
1x10 6 HeLa-luc/PKH26 cells were subcutaneously implanted into the left flank of a 6-8
week old female Nu/nu mouse. Figure 4 shows the spectrum for HeLa-luc/PKH26 cells,
the autofluorescent excitation spectrum of mouse tissue, and passbands for the
background filter (DsRed Bkg), the primary excitation filter (DsRed), and the emission
filter (DsRed). Figure 7 shows the IVIS images using the primary excitation filter, the
background excitation filer, as well as the autofluorescent-corrected image.
The
corrected image was obtained using a background scale factor of 1.4, determined by

taking the ratio of the autofluorescent signals on the scruff of the animal. The numbers
shown in the figures are the peak radiance of the animal background within the region of
interest. In the corrected image, the RMS error is used to quantify the background. The
signal-to-background ratio of the original fluorescent image (DsRed filter) is 6.5. The
ratio increases to 150 in the corrected image, an improvement factor of 23.
This
improvement reduces the minimum number of cells necessary for detection from 1.5x10 5
to 6.7x10 3 .
Background Subtraction With Quantum Dots (QD)
The unique optical properties of QDs make them attractive for in vivo measurements.
These particles come in a wide selection of colors, can be excited over a broad
wavelength range, and have narrow emission bands at wavelengths that are directly
related to the particle size. QDs have high quantum yields, are highly photostable, and
due to their engineered surface chemistry, are biocompatible [4]. Figure 8a shows the
spectrum of a typical QD along with the autofluorescent excitation spectrum of mouse
tissue and the passbands for the background filter, primary excitation filter, and emission
filter (DsRed filter set). Because the excitation curve gradually increases towards shorter
wavelengths, the autofluorescent subtraction technique using a blue-shifted background
filter is not effective since the background filter not only excites tissue autofluorescence,
but is also more efficient in exciting the QD.
A better method to reduce tissue autofluorescence and improve detection sensitivity from
measurements of QD reporters is to use one excitation filter with several narrow band

emission filters as illustrated in Figure 8b. With this technique, one primary emission
filter captures the peak emission from the QD while the other secondary emission filters
capture autofluorescence of the tissue on either side of the peak. The same type of
subtraction technique is used, except the secondary emission filter images are averaged
together and then subtracted from the primary emission filter image. Although filter sets
similar to those in Figure 8b are not currently available as an “off-the-shelf” set, they are
available on a custom basis. The IVIS Imaging System 200 Series is particularly well
suited for this application because it is equipped with a 24-position emission filter wheel,
compare to six positions for the IVIS Imaging System 100 or 50 Series.
Conclusion
To fully exploit the power of in vivo fluorescent imaging, methods to reduce the effects
of autofluorescence and other sources of background are required. The broad band light
source and multi-position filter wheels of the IVIS Imaging Systems provide the
flexibility necessary for various background subtraction techniques. In addition, the
Image Math tool in the Living Image software enables convenient processing and
analysis of fluorescent images.
References
[1] Troy T, Jekic-McMullen D, Sambucett L, Rice B. (2004). Quantitative comparison of
the sensitivity of detection of fluorescent and bioluminescent reporters in animal models,
Mol Imaging 3: 9-23.
[2] Billinton N, Knight AW (2001). Seeing the wood through the trees: A review of
techniques for distinguishing green fluorescent protein from endogenous
autofluorescence. Anal Biochem. 291: 175-197.

[3] Tuchin V (2000). Tissue Optics. Bellingtham, WA: SPIE Press.
[4] Bruchez M, Moronne M, Gin P, Weiss S, Alivisatos A (1998). Semiconductor
nanocrystals as fluorescent biological labels. Science. 281: 2013-2016.

a)
b)
c)
Figure 1. Example of instrumentation background subtraction using an FVB female
control mouse with the DsRed filter set: a) uncorrected fluorescent image of subject; b)
image of instrument background; c) corrected image obtained by subtracting instrument
background (b) from the uncorrected fluorescent image of the subject (a).

a) Regular rodent food
Ventral
Dorsal
b) Alfalfa-free rodent food
Ventral
Dorsal
Figure 2. Fluorescent images of the ventral and dorsal sides of a control mouse (Nu/nu
female) illustrating the animal tissue autofluorescence for the GFP, DsRed, Cy5.5 and
ICG filter sets (displayed in terms of efficiency). The images were taken after the mouse
was fed for over a week with a) regular rodent food and b) alfalfa-free rodent food.

Normalized Intensity
Normalized Intensity
a) Regular rodent diet
1.0
0.8
0.6
Alfalafa Diet
GFP
DsRed
Cy5.5
ICG
0.4
0.2
0.0
400
500
600
700
Wavelength (nm)
800
900
b) Alfalfa-free diet
1.0
0.8
0.6
Alfalfa free diet
GFP
Cy5.5
DsRed
ICG
0.4
0.2
0.0
400
500
600
700
Wavelength (nm)
800
900
Figure 3. Autofluorescent excitation (-----) and emission ( _____ ) tissue spectra from the
lower dorsal side of a seven-week-old female Nu/nu mouse fed for a week with a) regular
and b) alfalfa-free rodent diet.

Figure 4. Spectral data describing the autofluorescent subtraction technique using a
background filter. The graph shows the excitation and emission spectrum of PKH26 and
the autofluorescent excitation spectrum of mouse tissue. Also included are the spectral
passbands for the blue-shifted background filter (DsRed Bkg), the primary excitation
filter (DsRed), and the emission filter used with this dye.

Figure 5. The Image Math Panel of the Living Image software is used to subtract out
tissue autofluorescence using background filters.

a) GFP filter set
- x
0.7 =
b) DsRed filter set
- x
0.5 =
c) Cy5.5 filter set
- x
1.2 =
d) ICG filter set
- x
2.6 =
Figure 6. Examples of autofluorescent background filter subtraction of a 12-week old
female Nu/nu control mouse fed a regular and an alfalfa free diet. The scale factor, k,
was determined from average autofluorescence measurements over the entire animal.
The GFP and DsRed images are displayed with the same color scale and the Cy5.5 and
ICG image are displayed with another color scale. Left column: Primary excitation filter
images; middle column: background filter images; right column: background-subtracted
images. All of the data displayed here are in terms of efficiency.

a) Primary Excitation filter b) Background filter c) Corrected image
(DsRed)
(DsRed Bkg)
Figure 7. Example of the autofluorescent subtraction technique using a background
excitation filter. The images show: a) primary excitation filter (DsRed), b) blue shifted
background excitation filter (DsRed Bkg), and c) corrected data. The corrected image
was obtained by subtracting the scaled background filter image (multiplied by 1.4) from
the primary filter image. The 6-week-old female Nu/nu mouse was injected
subcutaneously with 1x10 6 HeLa-luc/PKH26 cells in the left flank.

a) Two excitation filters
b) Several narrow band emission filters
Figure 8. Spectral properties for 605 QD showing the characteristic broad excitation and
narrow emission curve along with the autofluorescent excitation spectrum of mouse
tissue. The graphs illustrate different autofluorescent subtraction techniques using a) two
excitation filters and one emission filter, and b) one excitation filter with several narrow
band emission filters. Due to the broad excitation spectrum for QDs, narrow band
emission filters are more suitable for autofluorescent subtraction.

Table 1. Summary of the background, excitation and emission filters.
Background Excitation Emission
filter filter filter
Passband (nm) Passband (nm) Passband (nm)
GFP 410-440 445 - 490 515 -575
DsRed 460-490 500 - 550 575 - 650
Cy5.5 580-610 615 - 665 695 - 770
ICG 665-695 710 - 760 810 - 875
Table 2. Autofluorescent efficiency of mouse tissue for animals fed regular and alfalfafree
rodent food.
Regular rodent food Alfalfa free rodent food
Ventral Dorsal Ventral Dorsal
GFP 1.18E-04 1.06E-04 1.04E-04 1.01E-04
DsRed 5.46E-05 4.02E-05 4.27E-05 3.82E-05
Cy5.5 9.19E-05 2.16E-05 1.52E-05 1.07E-05
ICG 3.44E-06 2.45E-06 3.35E-06 3.06E-06
Table 3. Values of the autofluorescent efficiency (average and RMS error) from the
tissue of a 12-week old female Nu/nu mouse before and after background filter
correction. The improvement was determined by taking the ratio of the average
efficiency before and the RMS efficiency after correction.
Before Correction After Correction
Dorsal Ave efficiency RMS efficiency Ave efficiency RMS efficiency Improvement
GFP 8.9E-05 3.6E-05 -6.6E-08 1.6E-05 5.5
DsRed 3.4E-05 1.5E-05 3.1E-08 2.7E-06 12.9
Cy5.5 1.0E-05 3.4E-06 3.9E-09 1.6E-06 6.5
ICG 2.7E-06 6.2E-07 -6.6E-09 5.5E-07 4.9
Before Correction After Correction
Ventral Ave efficiency RMS efficiency Ave efficiency RMS efficiency Improvement
GFP 1.3E-04 4.2E-05 -2.7E-07 1.7E-05 7.7
DsRed 5.7E-05 1.8E-05 2.0E-07 6.0E-06 9.5
Cy5.5 1.6E-05 7.0E-06 3.7E-08 2.4E-06 6.6
ICG 3.1E-06 6.2E-07 -2.2E-09 7.7E-07 4.0