|Home | About | Journals | Submit | Contact Us | Français|
Cyclic diguanylate (c-di-GMP) is a bacterial second messenger important for physiologic adaptation and virulence. Class-I c-di-GMP riboswitches are phylogenetically widespread and thought to mediate pleiotropic genetic responses to the second messenger. Previous studies suggest that the RNA aptamer domain switches from an extended free state to a compact, c-di-GMP-bound conformation in which two helical stacks dock side-by-side. Single molecule fluorescence resonance energy transfer (smFRET) experiments reveal that the free RNA exists in four distinct populations that differ in dynamics in the extended and docked conformations. In the presence of c-di-GMP and Mg2+, a stably docked population (>30 min) becomes predominant. smFRET mutant analysis demonstrates that tertiary interactions distal to the c-di-GMP binding site strongly modulate the RNA population structure, even in the absence of c-di-GMP. These allosteric interactions accelerate ligand recognition by pre-organizing the RNA, favoring rapid c-di-GMP binding.
Riboswitches are gene-regulatory mRNA domains that recognize small molecule metabolites and second messengers.1–8 With the exception of the catalytic glmS ribozyme-riboswitch,9, 10 these regulators function by adopting ligand occupancy-dependent conformations that modulate transcription, translation or alternative splicing. The substructure of a riboswitch that suffices for specific ligand binding in vitro is known as its “aptamer” domain. Biophysical characterization of the aptamer domains of different riboswitch classes shows that their global structural response to ligand binding is idiosyncratic.11 The aptamer domain of the flavin-mononucleotide (FMN) riboswitch from Bacillus subtilis is largely pre-organized in the absence of FMN at physiologic Mg2+ concentrations, while that of the class I S-adenosylmethionine (SAM) riboswitch from the same organism is only partly ordered absent its cognate metabolite, and adopts its most compact form upon SAM binding.11 Moreover, the degree of compaction induced by ligand binding or by Mg2+ varies even between aptamer domains of the same riboswitch class [e.g. the thiamine-pyrophosphate (TPP) riboswitches from Escherichia coli and Arabidopsis thaliana]12. Although it has been suggested that this variation may reflect the distinct regulatory requirements of each genetic locus,12 the functional significance of ligand binding-induced global folding of riboswitch aptamer domains remains to be established.
Cyclic diguanylate (c-di-GMP) is a bacterial second messenger involved in the regulation of a variety of complex physiological adaptations including motility, virulence, biofilm formation, and cell cycle progression.13–16 Two structurally distinct classes of c-di-GMP riboswitches have been described.17, 18 Of these, the class-I riboswitches (c-di-GMP-I) are the most widespread. Over 500 different c-di-GMP-I riboswitches have been identified in a wide range of bacteria, and are often found in multiple copies in bacterial genomes. As many as 30 different genetic loci in Geobacter uraniumreducens appear to be under c-di-GMP-I riboswitch regulation.19 Crystallographic structure determinations of the aptamer domain of the c-di-GMP-I riboswitch associated with the tfoX gene of Vibrio cholerae bound to the second messenger revealed that the RNA consists of three helices, paired regions P1a, P1b, and P2, joined in a three-way junction by joining regions J1a/b, J1b/2, and J2/1a (Fig. 1a,b).20, 21 c-di-GMP binds at the junction, participating in an interaction network between P1a, P1b, and the three joining regions. The conserved A47 from J1b/2 intercalates between the two guanine bases of the second messenger (Fig. 1a,b, purple). The bound c-di-GMP and A47 mediate continuous coaxial stacking between P1b and P1a. P2 docks side-by-side with P1b. This arrangement appears to be stabilized by two sets of phylogenetically conserved tertiary interactions distal to the junction, in addition to the c-di-GMP-bound junction itself. First, the GNRA tetraloop (GT) that caps P1b docks against a tetraloop receptor (TR) in P2. Second, an interhelical Watson-Crick pair is formed between C44 of P1b and G83 of P2 (Fig. 1a,b, green and orange).20, 21
Small-angle X-ray scattering (SAXS) experiments revealed a dramatic compaction of the class-I riboswitch aptamer domain induced by c-di-GMP binding in the presence of physiologic [Mg2+].12, 20 Low-resolution molecular envelopes calculated from the SAXS data suggest that, in absence of c-di-GMP, the RNA adopts an extended conformation in which P1b and P2 are splayed apart, and neither the GT/TR interaction nor the C44•G83 base pairing takes place. Nuclease protection and in-line probing experiments are consistent with disruption of both sets of tertiary interactions in the absence of c-di-GMP, and also suggest that P1a becomes disordered under these conditions.18, 20 Previous studies of large catalytic RNAs have shown that tertiary interactions promote RNA folding within compact intermediates resulting from an early divalent-cation induced collapse, in which the helices interact but are not yet stably docked.22 For instance, in the case of the Azoarcus group I ribozyme, a GT/TR interaction has been shown cooperatively to promote tertiary structure throughout the RNA, increasing the speed and accuracy of its folding.23 Unlike these catalytic RNAs, which require only divalent cations to achieve their native state, riboswitch aptamer domains have evolved to recognize small molecules concomitant with folding.
To dissect the interplay of cations, second-messenger ligand and tertiary interactions in riboswitch folding, we have now analyzed the c-di-GMP-I aptamer domain using smFRET.24–29 These studies confirm that this RNA samples extended and compact conformations. However, smFRET analysis, which can uncover structural dynamics of individual molecules that would otherwise be hidden in ensemble-averaged experiments, reveal that the aptamer domain is kinetically partitioned into four distinct populations: two that, in the experimental timeframe, remain statically docked or undocked (compact or extended, respectively), one that fluctuates between docked and undocked but is preferentially in the docked state, and another that fluctuates but is preferentially undocked. The population structure shifts in response to Mg2+ and c-di-GMP concentration, such that at saturating second messenger and physiologic Mg2+ concentration, the majority of the molecules are statically docked. smFRET analysis of site-directed mutants that disrupt the GT/TR or C44•G83 tertiary interactions indicates that these are required not only for binding of c-di-GMP, but also profoundly impact the population structure of the RNA in the absence of ligand. Thus, we find that these tertiary interactions, which are distant from the c-di-GMP binding site, serve to pre-organize the aptamer domain. In vivo, this would allow the nascent riboswitch transcript to fold and recognize its ligand rapidly and thus respond effectively to varying intracellular levels of the second messenger.
We incorporated a Cy3 (donor) and a Cy5 (acceptor) fluorophores near the distal tips of P2 and P1b, respectively, of an RNA construct based on the Vibrio cholerae tfoX c-di-GMP riboswitch (Fig. 1a–c).29 With this labeling scheme, the extended conformation is expected to result in a low FRET ratio, while the compact conformation is expected to result in a high FRET ratio.20 Characteristic smFRET time trajectories in standard conditions (which contain 2.5 mM Mg2+, Methods) are shown in Figure 1d. The aptamer domain RNA exhibits FRET ratios of 0.2 and 0.8, which may correspond to the extended and docked conformations, respectively, deduced from the SAXS reconstructions.20 States with intermediate FRET efficiencies were not observed with a time resolution of 33 ms (Methods). In the absence of c-di-GMP, most molecules (49 ± 4% of 282, Supplemental Table 1) remain in a low FRET state over the time of the experiment (a few minutes). We refer to this population as static undocked. A smaller population (27 ± 5%) resides primarily in the high FRET state with brief excursions into the low FRET state. We refer to this population as dynamic docked. Two other minor populations are also observed: one in which molecules exhibit high FRET for the duration of the experiment (7 ± 2%), and another in which molecules display primarily low FRET with brief excursions into a high FRET state (17 ± 8%). We refer to these populations as static docked and dynamic undocked, respectively. We used dwell time analysis to determine the rate constants of docking and undocking for the two dynamic populations (Fig. 2 and Supplemental Fig. 1).29 In the absence of c-di-GMP, kdock and kundock for the dynamic docked population are 6.6 ± 0.2 s−1 and 1.0 ± 0.1 s−1, respectively, while kdock and kundock for the dynamic undocked population are 1.6 ± 0.1 s−1 and 6.7 ± 0.1 s−1, respectively. Thus, the dynamic docked molecules spend most of the time in the docked conformation and the dynamic undocked molecules spend most of the time in the undocked conformation. The distinct kinetic properties of these two populations become readily apparent by scatter analysis (Fig. 2b).
In the presence of saturating c-di-GMP (1 μM, standard conditions, 2.5 mM Mg2+, Methods), the static docked population becomes predominant (61 ± 4% of 250 molecules, Supplemental Table 1), suggesting that this population is ligand-bound. Both the static undocked and dynamic docked populations decrease significantly in the presence of ligand (to 29 ± 6% and 2 ± 2%, respectively). The docking and undocking rate constants for the dynamic populations in the presence of 100 nM c-di-GMP were similar to those in the absence of c-di-GMP (Fig. 2 and Supplemental Fig. 1), suggesting that these dynamic populations do not have c-di-GMP bound. Upon inspection of >100 time trajectories, we found that the static docked population can form from any of the other three populations (Fig. 3a), providing further evidence that the static docked population is ligand-bound and indicating that any of these populations are able to bind c-di-GMP and form the stable docked conformation. Experiments with lower laser power and longer exposure showed the static docked conformation has a lifetime longer than 30 min (Fig. 3b), consistent with previous results.21
To determine the second messenger binding affinity, we measured the fraction of c-di-GMP bound-riboswitch (static docked population) as a function of c-di-GMP concentration under standard conditions. In the absence of c-di-GMP, 7 ± 2% of riboswitches fold into the static docked conformation. This fraction increases with c-di-GMP concentration and saturates at 68 ± 9%, with the remaining molecules persisting in the undocked conformation. A fit to the Langmuir equation yields a KD = 90 ± 20 nM (Fig. 4a,b), comparable with KD’s from isothermal titration calorimetry (ITC) experiments performed with single-chain aptamer domain constructs (Supplemental Table 2). This demonstrates the surface immobilization of the two-chain RNA construct (Fig. 1a) does not adversely affect folding or ligand binding. However, the affinity determined by smFRET and ITC is several orders of magnitude weaker than that based on electrophoretic gel mobility-shift analyses,21 suggesting that the aptamer domain behaves differently in the polyacrylamide gel matrix.
To test the selectivity of the c-di-GMP-I riboswitch, we performed similar experiments using c-di-AMP, a structural analog of c-di-GMP and recently discovered putative bacterial second messenger.20, 21, 30 In bulk experiments, c-di-AMP does not bind the c-di-GMP-I riboswitch.20, 21 The aptamer domain displays similar behavior in smFRET experiments in the presence of 1 μM c-di-AMP as in the absence of c-di-GMP, with comparable population distributions (Supplemental Table 1). The docking and undocking rate constants for the dynamic molecules are also similar (Fig. 2 and Supplemental Fig. 1), supporting our interpretation that the dynamic molecules are not bound to the second messenger.
Overall, these data show the c-di-GMP-I riboswitch aptamer domain adopts docked and undocked conformations similar to those deduced from SAXS analysis.20 The RNA is kinetically partitioned into four distinct populations that do not readily interconvert in the timeframe of our experiments (several minutes) under steady-state conditions: static undocked and docked as well as dynamic undocked and docked. The RNA binds c-di-GMP tightly and selectively to fold into the stable docked conformation, and this results in population redistribution as the molecules in the three other populations are competent for second messenger binding.
Previous SAXS analysis of the V. cholerae c-di-GMP-I riboswitch aptamer domain indicates that the RNA undergoes global compaction as the concentration of Mg2+ is raised from 2.5 mM to 10 mM.20 This is reminiscent of the behavior of other RNAs with complex structure, such as the group I intron and RNase P, which undergo Mg2+ ion-induced folding.31–33 Those large ribozymes attain their native conformations upon Mg2+ ion-induced folding, as judged by their full catalytic activity.34, 35 In contrast, Kratky analysis of the SAXS data on the c-di-GMP-I aptamer domain implies that, unlike the group I intron and RNase P, the riboswitch remains locally disordered even at high Mg2+ concentration until c-di-GMP is bound.20 Divalent cations can facilitate RNA folding non-specifically, as part of a diffusely condensed ionic atmosphere, or by making specific, direct interactions with the RNA.36 Our smFRET experiments indicate that both modes of action are operative in the c-di-GMP-I riboswitch.
To analyze the role of Mg2+ in c-di-GMP riboswitch folding, we measured the fraction of each RNA population as a function of [Mg2+]. In the absence of both, Mg2+ and c-di-GMP, almost all of the molecules (96 ± 6% of 75 molecules) reside in the static undocked population (Supplemental Table 1). In the absence of Mg2+ ions, addition of saturating c-di-GMP does not alter the static undocked population. In very high [Mg2+] (50 mM), but in the absence of c-di-GMP, the majority of molecules reside in the dynamic docked and undocked populations, as a consequence of a substantial decrease in the static undocked population. Fitting the fraction of dynamic molecules (including both dynamic docked and undocked) as a function of Mg2+ ion concentration to the Langmuir equation yields KMg = 1.2 ± 0.2 mM (Supplemental Fig. 2). Even under these elevated [Mg2+], only a small fraction of the molecules (10 ± 5%) are statically docked (Supplemental Table 1). These results indicate that both c-di-GMP and Mg2+ are necessary to drive a majority of the molecules into the static docked state, and further supports our assignment of this population to the ligand bound, highly structured conformation characterized crystallographically and by SAXS.20
To determine the interplay of Mg2+ ion and second messenger binding, we measured the fraction of c-di-GMP bound-riboswitch (static docked population) as a function of [Mg2+]. In the presence of saturating c-di-GMP, the static docked population increases concomitant with increasing Mg2+ concentration (Fig. 4c). Fitting the data to the Langmuir equation yields a KMg = 0.5 ± 0.1 mM (Fig. 4d circles), a value near the physiological range of concentration for the cation. In the absence of c-di-GMP, there is no increase in the static docked population (Fig. 4d triangles), indicating that Mg2+ and c-di-GMP binding are strongly cooperative. Experiments in the presence of higher monovalent ion concentrations but no Mg2+ (50 mM Tris-HCl, pH 8.0, 150 mM KCl, and 30 mM NaCl) show that the aptamer domain cannot fold into the docked conformation and remains entirely in the static undocked conformation even in the presence of saturating c-di-GMP (Supplemental Table 1). These results suggest that the role of Mg2+ ion is not solely electrostatic screening required to fold the riboswitch into an ligand binding-competent conformation, but that it specifically mediates binding to the second messenger. Indeed, crystal structures reveal a hydrated Mg2+ ion at the second messenger-binding pocket37 where it bridges a phosphate of c-di-GMP with those of residues G19 and G20 of the RNA, and our smFRET titration may be reporting on this tightly bound cation.
To characterize the role of the GT/TR and C44•G83 tertiary interactions (Fig. 1a)20, 21 in the formation and stability of the docked structure, we introduced point mutations to prevent their formation. First, we examined the C44•G83 base pair by mutating G83 to C. smFRET analyses of this G83C mutant show that both in the absence or presence of 1 μM c-di-GMP, most molecules (95 ± 9% and 80 ± 6 %, respectively) reside in the static undocked population (Fig. 5, Supplemental Table 1). The G83C mutant RNA is therefore unable to attain either the dynamic or static docked states, even with c-di-GMP present. These results are consistent with the results of bulk ITC experiments that show that the G83C mutant RNA is severely disrupted in c-di-GMP binding, and exhibits a 1200-fold increase in the apparent KD for c-di-GMP even at elevated Mg2+ concentration (10 mM) relative to wild-type RNA (Supplemental Table 2). Mutation of G83 to U also severely disrupts affinity for c-di-GMP with a 250-fold increase in KD at 10 mM Mg2+ (Supplemental Table 2).
We also mutated C44 to A, which should prevent its interhelical base pairing with G83. smFRET experiments show that in the absence of c-di-GMP (Fig. 5), the majority of the molecules (93 ± 3%) reside in the static undocked population. However, in the presence of 1 μM c-di-GMP this mutant exhibits a 25 ± 7% population in the static docked conformation (Fig. 5, Supplemental Table 1), indicating that the C44A mutation does not destabilize the docked state as much as the G83C mutation. This result is also consistent with the binding affinity measured by bulk ITC, which shows an 80-fold increase in KD at 10 mM Mg2+ (Supplemental Table 2) relative to wild-type. A possible explanation for this result is that the C44A mutant is capable of forming a non-canonical base pair with G83. Overall, these data show that the tertiary C44•G83 base pair is essential for the aptamer domain to attain either the dynamic or static docked states.
Next, we examined the role of the GT/TR interaction between P1b and P2 by mutating A33 to U (Fig. 1a). The crystal structures show that A33 flips out of the GAAA tetraloop to stack with A62 of the TR.20 Mutating A33 to U should prevent this stacking, and therefore, impair the GT/TR interaction. smFRET experiments using the A33U mutant in the absence of c-di-GMP show that the majority of molecules (68 ± 6%) reside in the static undocked population but 29 ± 8% of the molecules reside in the dynamic undocked population (Supplemental Table 1). This result shows that, unlike mutations that affect the C44•G83 base pair, mutational destabilization of the GT/TR does not prevent the riboswitch from transiently populating the docked conformation. However, the addition of 20 mM Mg2+ was not able to recover the dynamic docked population. In the presence of 1 μM c-di-GMP, the fraction of the molecules in the static undocked population decreases to 49 ± 5% while the static docked population increases to 37 ± 3%, indicating that this mutant can still bind c-di-GMP and form the static docked conformation. This result is in agreement with gel shift experiments, which show a 200–fold increase in apparent KD for this mutant.37 These mutational data indicate that the C44•G83 base pair between P1b and P2 is essential for the formation of the stable docked conformation, while the GT/TR tertiary interaction assists in stabilizing the aptamer in the docked conformation. Overall, our smFRET analyses of riboswitch mutants indicate that docking of P1b and P2 mediated by long-range tertiary interactions is required for formation of a c-di-GMP binding-competent aptamer domain conformation.
Single molecule methods have previously been employed to examine several riboswitches. The adenine and guanine riboswitches are closely related RNAs organized around a three-helix junction, where the purine base binds. In the crystal structures of their ligand-bound aptamer domains,38, 39 the loops that close the distal ends of two of the constituent helices of the riboswitch associate through a series of tertiary interactions. smFRET and force spectroscopy experiments imply that some of these long-range interactions can take place in the presence of Mg2+ ions alone, prior to purine binding.25, 28, 40–42 smFRET analyses of the class I and II SAM riboswitches suggest that these structurally unrelated RNAs can both transiently sample conformations similar to their respective ligand-bound states in the presence solely of Mg2, and that SAM binding occurs by conformational capture.26, 27 The Mg2+-induced, partially folded states of these riboswitches appear to be kinetic intermediates in pathways leading to their ligand-bound (native) states.27, 28, 40 This supports the inference that the partially folded states of these RNAs have structural similarity to their native fold.
Like the purine and SAM riboswitches, the c-di-GMP riboswitch can transiently adopt a global fold similar to that of its second messenger-bound form in the presence of Mg2+ alone. However, the results of our smFRET analysis reveal that this RNA folds in a complex landscape, populated by four classes of molecules that interconvert very slowly in the timeframe of the experiment. Such kinetic partitioning has been previously documented in bulk43 and at the single-molecule level44–47 for catalytic RNAs. The shift in the population structure of the RNA upon addition, first of physiologic concentrations of Mg2+, and then c-di-GMP, as well as the cooperativity exhibited by Mg2+ and c-di-GMP in formation of the stably docked population suggest that the high FRET state sampled by the dynamic undocked and dynamic docked populations has structural similarity to the native, second-messenger bound conformation. To provide support for this hypothesis, we generated RNAs with site-directed mutations targeting tertiary interactions that are present in the crystal structures of the ligand-bound aptamer domain. We first show calorimetrically that these mutant aptamer domains are impaired, to varying extents, in c-di-GMP binding. smFRET analysis reveals that the mutations perturb the folding landscape of the RNAs, such that decrease in the dynamic populations correlates directly with the deleterious impact of each mutation in ligand binding. This indicates that native-like tertiary interactions are indeed responsible for stabilizing the transiently folded states of the dynamic molecules, thereby demonstrating experimentally that the c-di-GMP riboswitch is pre-organized, rather than simply collapsed. Since the dynamic populations shift to the static docked, ligand-bound conformation in the presence of c-di-GMP, the pre-organization of the riboswitch enables rapid second messenger binding. Our findings parallel those of folding studies on large ribozymes, which have established the importance of forming native-like tertiary interactions early in the Mg2+ induced collapse of an RNA in order to achieve overall rapid folding.22
The proposed folding pathway for the c-di-GMP riboswitch is shown in Fig. 6. In the absence of c-di-GMP, the riboswitch adopts an undocked conformation. In-line and nuclease probing experiments indicate that, in the absence of ligand, P1a is unpaired, thus allowing the formation of the downstream terminator stem-loop and preventing gene expression.18, 20, 21 A Mg2+-dependent population of riboswitches exhibits dynamic switching from this undocked conformation to a docked conformation with brief excursions into the undocked conformation at transcription relevant time scales (~150 ms). This unstable docked conformation is stabilized by the GT/TR and C44•G83 tertiary interactions. The cocrystal structures of the riboswitch20, 21 show that the nucleobase of A49 stacks underneath the C44•G83 base pair. This stacking may help propagate order from the interhelical base pair to the c-di-GMP binding site, in which A47 plays a central role (Fig. 1a). The Mg2+-dependent dynamic population offers a pre-organized structure that allows efficient cotranscriptional folding and ligand binding. Upon binding to the junction region, c-di-GMP completes the formation of a continuous helical stack between P1a and P1b, leading to stabilization of the docked conformation, including the P1a helix. Although we did not directly determine formation of the P1a helix in our experiments, in-line and nuclease probing experiments and crystal structures indicate that this helix is formed in the folded aptamer domain structure.18, 20, 21, 37 This helix is the molecular switch controlling gene expression:18 the anti-terminator stem would form, preventing formation of the terminator stem, allowing transcription of the downstream gene to proceed.
The biological significance of the large-scale, ligand-induced folding transition documented by SAXS for the c-di-GMP-I riboswitch has been unclear for two reasons. First, not every riboswitch class examined undergoes such a collapse concomitant with ligand binding.11, 12 Second, such large-scale reorganization of the aptamer domain couples the cost of the loss of conformational entropy to ligand recognition, thereby lowering the maximum achievable affinity. Our discovery that the c-di-GMP-I riboswitch aptamer domain transiently folds into a collapsed conformation that has a structure similar to that of the ligand-bound form suggests a role of the global folding transition in a process akin to kinetic proofreading: only molecules in which the P1b and P2 stems have folded correctly (i.e. in conformations compatible with making the allosteric GT/TR and C44•G83 tertiary interactions) will present a ligand binding site to the second messenger. It is noteworthy that of the various riboswitches whose Mg2+ and ligand induced collapse has been studied by SAXS,11, 20, 48, 49 the TPP riboswitch is the RNA that most closely mimics the behavior of the c-di-GMP-I riboswitch. Although their specific sequences are unrelated, the aptamer domains of the c-di-GMP and TPP riboswitches share a similar architecture comprised of a three-helix junction where the ligands bind, and tertiary interactions distal to the ligand binding site stabilizing the side-by-side packing of two helical stems. Biophysical experiments have shown the importance of the allosteric loop-loop interactions in ligand binding by the TPP riboswitch.50, 51 These distal interactions were found to form before the TPP binding site is fully organized.51 paralleling the results of our studies of the c-di-GMP-I riboswitch. Taken together, these studies suggest that large-scale pre-organization coupled to ligand binding may be a strategy to increase structural specificity, and hence accelerate ligand binding, by several riboswitch classes.
RNA purification and labeling, single molecule FRET and Isothermal Titration Calorimetry experiments were performed as previously described.11, 20, 29, 52 A detailed description of the methods is available as supporting information online.
We thank N. Kulshina for initial characterization of point mutants and N. Baird for calorimetry help. This work was supported in part by the NIH (R01 GM085116) and an NSF CAREER award (MCB-0747285) to D.R., and in part by the intramural program of the National Heart, Lung and Blood Institute, NIH (A.R.F.)