Edwards et al., Curr Opin Struct Biol 2007

Metabolite recognition by riboswitches Edwards, Klein and Ferré-D’Amaré 275
structure of the metabolite-binding domain of the first
class to be discovered (the SAM-I riboswitch) reveals
an architecture that is distinctly different from the
inverted-h fold of the purine and TPP riboswitches
(Figure 1e,f) [15 ]. The SAM-I riboswitch contains two
stacks (P1-P4 and P2a-P3) that, rather than packing sideby-side,
cross at an angle of 708. A pseudoknot coupled
to a kink-turn [16] atop P2 appears to stabilize this
fold. The SAM-binding site is located at the interface
of the minor grooves of P1 and P3, and has features
reminiscent of the metabolite-binding sites of both the
TPP and the purine riboswitches. Like TPP, SAM
bridges two helical stacks. Like the purine riboswitches,
SAM makes van der Waals contact with the 3 0 strand of
the P1 switch helix.
In many Gram-positive bacteria, the glmS ribozyme-riboswitch
is part of the 5 0 -UTR of the mRNA that encodes
glucosamine-6-phosphate (GlcN6P) synthetase [17]. This
riboswitch has a self-cleavage activity that becomes activated
when it binds GlcN6P. The structure of the glmS
ribozyme [18 ,19 ] consists of three parallel helical stacks
(Figure 1g,h). A doubly pseudoknotted core (P1-P2-P2.1-
P2.2) is buttressed by a peripheral RNA domain (P4-P4.1).
The solvent-exposed GlcN6P-binding pocket is composed
of two highly distorted major grooves and abuts the site of
self-cleavage, reflecting the coenzyme function of GlcN6P
(see the review by Scott in this issue). Among currently
characterized riboswitches, the glmS ribozyme is unique
because it adopts its active structure in the absence of its
metabolite ligand [18 ,20 ]. As other ribozymes, such as
the natural hairpin ribozyme [21,22] and the in vitro
selected Diels–Alderase [23], also assemble rigid active
sites, this disparity might reflect the different constraints
under which ribozymes and RNAs that function by alternative
folding evolved.
Ligand recognition by riboswitches
The purine riboswitch recognizes its ligand almost exclusively
through hydrogen-bonding interactions that satisfy
nearly all possible acceptors and donors of the purine
(Figure 2a). Purine riboswitch structures have been
solved bound to 2,6-diaminopurine [24 ] and 2,4,6-triaminopyrimidine
[25], in addition to the biological activators
hypoxanthine [6], guanine [7] and adenine [7]. The
purine ligand is primarily recognized by residue 74 of the
riboswitch, a pyrimidine, through Watson–Crick pairing
[6,7,26]. In addition, U51 and the 2 0 -OH of U22 hydrogen
bond to the N3/N9 edge (corresponding to the sugar edge
of nucleotides) and the N7, respectively, of the purine.
Reliance on Watson–Crick (as opposed to Hoogsteen)
pairing for recognition enables the same RNA scaffold to
regulate either adenine or guanine metabolism by having
U74 or C74, respectively. Gilbert et al.[24 ] noted that the
purine ligand makes poor stacking interactions and proposed
that this enhances the discriminatory role of Watson–Crick
pairing with residue 74.
The two helical stacks of the thi-box riboswitch separately
recognize the aminopyrimidine and pyrophosphate moieties
of TPP. J3/2 of the ‘pyrimidine sensor helix’ (the P1-
P2-P3 stack) adopts a canonical T-loop fold [9 ,10 ,11 ].
Binding of the aminopyrimidine ring of TPP to G40 of this
T-loop (Figure 2b) mimics a tertiary interaction between
the D- and T-loops in the classic L-shaped fold of tRNA.
The aminopyrimidine of TPP and G40 replace G18 and
C55, respectively (purines and pyrimidines are reversed
between the riboswitch and the tRNA). Mimicry of an all-
RNA structure by an exogenous small molecule is reminiscent
of ATP binding by an in vitro selected aptamer
RNA, whereby the ATP completes a GNRA tetraloop
[27,28]. Rather than directly binding to the negatively
charged pyrophosphate of TPP, the ‘pyrophosphate sensor
helix’ (the P4-P5 stack) of the riboswitch coordinates the
pyrophosphate mostly through two solvated divalent
metals ions (the exception is G78) [9 ,10 ,11 ,29]. Thus,
the riboswitch effectively binds a positively charged TPP–
cation complex. Structures of the TPP riboswitch bound to
three metabolite analogs suggest that the pyrimidine sensor
helix is largely preformed, whereas the pyrophosphate
sensor helix becomes organized concomitant with binding
of the TPP–cation complex [11 ].
The SAM-I riboswitch sandwiches its ligand between two
parallel helices [15 ]. P1 recognizes the ribose–sulfur
backbone of SAM primarily through van der Waals contacts
(Figure 2c). By contrast, the P3 helix binds the
adenine ring and the amino acid by making several
hydrogen bonds and stacking interactions. Interestingly,
the RNA does not directly recognize the methionine
e-methyl group. Rather, the positively charged sulfur atom
makes a favorable electrostatic interaction with the partial
negative charge on O2 of U7. This might explain the
enhanced binding of SAM compared to S-adenosylhomocysteine
and other non-positively charged analogs [12,30].
Recognition of GlcN6P, a simple phosphorylated hexosamine
sugar, by the glmS ribozyme-riboswitch presents a
challenge for RNA that is distinct from those posed by
purines, TPP and SAM, all of which have nucleotide-like
substructures. Crystal structures of the glmS ribozymeriboswitch
were solved bound to glucose-6-phosphate
(Glc6P), an isosteric competitive inhibitor (antagonist)
of GlcN6P [18 ], and to the authentic activator GlcN6P
([19 ]; DJ Klein and AR Ferré-D’Amaré, unpublished),
revealing that GlcN6P and Glc6P are equivalently positioned
in the glmS ribozyme-riboswitch active site. The
sugar hydroxyl groups hydrogen bond to G1, C2, A50 and
G65 (Figure 2d). The amine of GlcN6P hydrogen bonds
to a water molecule, U51 and the 5 0 oxygen of G1; the last
is the leaving group of the transesterification reaction
catalyzed by the ribozyme. As in the TPP riboswitch,
the phosphate of the metabolite interacts with the RNA
through two solvated divalent metal ions ([19 ]; DJ Klein
and AR Ferré-D’Amaré, unpublished). TPP, SAM and
www.sciencedirect.com Current Opinion in Structural Biology 2007, 17:273–279