Two-step binding kinetics of tRNAGly by the glyQS T-box riboswitch and its regulation by T-box structural elements

T-box riboswitches are cis-regulatory RNA elements that regulate mRNAs encoding for aminoacyl tRNA synthetases or proteins involved in amino acid biosynthesis and transport. Rather than using small molecules as their ligands, as do most riboswitches, T-box riboswitches uniquely bind tRNA and sense their aminoacylated state. Whereas the anticodon and elbow regions of the tRNA interact with Stem I, located in the 5’ portion of the T-box, sensing of the aminoacylation state involves direct binding of the NCCA sequence at the tRNA 3’ end to the anti-terminator sequence located in the 3’ portion of the T-box. However, the kinetic trajectory that describes how each of these interactions are established temporally during tRNA binding remains unclear. Using singlemolecule fluorescence resonance energy transfer (smFRET), we demonstrate that tRNA binds to the riboswitch in a two-step process, first with anticodon recognition followed by NCCA binding, with the second step accompanied by an inward motion of the 3’ portion of the T-box riboswitch relative to Stem I. By using site-specific mutants, we further show that the T-loop region of the T-box significantly contributes to the first binding step, and that the K-turn region of the T-box influences both binding steps, but with a more dramatic effect on the second binding step. Our results set up a kinetic framework describing tRNA binding by T-box riboswitches and highlight the important roles of several T-box structural elements in regulating each binding step. SIGNIFICANCE Bacteria commonly use riboswitches, cis-regulatory RNA elements, to regulate the transcription or translation of the mRNAs upon sensing signals. Unlike small molecule binding riboswitches, T-box riboswitches bind tRNA and sense their aminoacylated state. T-box modular structural elements that recognize different parts of a tRNA have been identified, however, how each of these interactions is established temporally during tRNA binding remains unclear. Our study reveals that tRNA binds to the riboswitch in a two-step mechanism, with anticodon recognition first, followed by binding to the NCCA sequence at the 3’ end of the tRNA with concomitant conformational changes in the T-box. Our results also highlight the importance of the modular structural elements of the T-box in each of the binding steps.


INTRODUCTION
Riboswitches are cis-regulatory RNA elements that recognize and respond to defined external signals to affect transcription or translation of downstream messenger RNAs (mRNAs) (1)(2)(3).
Riboswitches generally consist of two domains: a sensory or aptamer domain and a regulatory domain or expression platform. The expression platform can adopt different conformations in response to ligand binding to the aptamer, and in this way controlling gene expression outcome (1). The aptamer of each riboswitch class contains conserved sequence motifs and unique secondary or tertiary structural elements that help distinguish and bind specific ligands (4).
Bacterial T-box riboswitches represent a unique class of riboswitches that do not bind small molecule ligands, instead they recognize and bind tRNA molecules and sense directly their aminoacylation state (5). T-box riboswitches serve as excellent paradigms to understand RNA-RNA interactions and RNA-based regulation.
T-box riboswitches are found in Gram-positive bacteria and are usually located in the region upstream of mRNA sequences encoding aminoacyl tRNA synthetases and proteins involved in amino acid biosynthesis and transport and hence participate directly in amino acid homeostasis (5). In general, the aptamer domain of all T-box riboswitches contains a long stem, Stem I, responsible for specific tRNA binding (6). The expression platform can adopt either a terminator or anti-terminator conformation, depending on whether the bound tRNA is charged or uncharged (5,7). In most T-box riboswitches, binding of a charged tRNA to the T-box leads to rho independent transcription termination whereas an uncharged tRNA stabilizes the antiterminator conformation and leads to transcription read-through (5,7). Whereas Stem I and the anti-terminator domain are highly conserved among T-box riboswitches, the region connecting them can vary. The Bacillus subtilis glyQS T-box riboswitch, involved in glycine regulation, represents one of the simplest T-box riboswitches (8). Only a short linker and a small stem, Stem III, connect Stem I and the anti-terminator domain (Fig. 1).
Recognition of tRNA by a T-box riboswitch involves three main structural elements of the tRNA: the anticodon region, the "elbow" region formed by the conserved T-and D-loops, and the 3' NCCA sequence (Fig. 1B). The anticodon and elbow regions of the tRNA interact with Stem I directly. Stem I contains several phylogenetically conserved structural motifs (6), including a Kturn motif, a specifier loop, a distal bulge, and an apical loop (6) (Fig. 1). Bioinformatics and structural analyses have collectively revealed the interactions between Stem I and the tRNA (9)(10)(11). Specifically, the co-crystal structures of Stem I/tRNA complexes show that Stem I flexes to follow closely the tRNA anticodon stem and interacts directly with the anticodon loop and the elbow through its proximal and distal ends, respectively (11). The distal bulge and the apical loop fold into a compact structural module of interdigitated T-loops (12,13), which interact directly with conserved unstacked nucleobases at the tRNA elbow (9,11). In addition, the structures revealed that Stem I turns sharply around two hinge regions using a conserved dinucleotide bulge and the K-turn motif (11,14). Sensing of the aminoacylation state involves direct binding of the tRNA 3' end to a highly conserved bulge in the T-box, the t-box sequence (15) (Fig. 1). A free NCAA end can base pair with the t-box sequence, enabling the anti-terminator conformation, whereas a charged NCCA end prevents the formation of the t-box/NCCA interactions, leading to the more stable terminator conformation (5,7). Importantly, discrimination between the charged and uncharged tRNA does not require any additional proteins, such as EF-Tu (16), and is driven solely by RNA/RNA interactions.
Although there are no atomic-level structural details on the interactions between tRNA and the anti-terminator region, Small Angle X-ray Scattering (SAXS)-derived models of the entire B.
subtilis glyQS T-box riboswitch in complex with tRNA are available (17,18). The two models are distinct, one presenting a more compact structure where all the previously observed interactions between Stem I and tRNA are preserved and the 3' NCAA sequence of the tRNA helps stabilize a coaxial stem formed by Stem III and the anti-terminator region (17), while the second model shows a more extended and relaxed structure where the interactions with the anticodon are preserved but the contacts with the tRNA elbow are not present (18). In addition, there is a dearth of information on the kinetics of the binding process. Whereas it is clear that tRNA recognition involves several specific interactions, their binding temporal sequence remains elusive. In addition, it is unclear whether sensing of the 3' end of the tRNA involves any additional conformational changes in the T-box. Here, by introducing donor-acceptor fluorophore pairs at several locations in the tRNA and the T-box riboswitch, and using single-molecule fluorescence resonance energy transfer (smFRET), we demonstrate the temporal order of events in the trajectory of tRNA binding. Our results demonstrate that tRNA binds to the riboswitch in two steps, with its anticodon being recognized first, followed by NCCA binding accompanied with an inward motion of the 3' region of the T-box riboswitch, including Stem III and the anti-terminator stem, relative to Stem I. In addition, by introducing mutations at different locations of the T-box, we further show that the two-step binding kinetics is regulated by structural elements in the T-box riboswitch.

Binding of cognate tRNA by the glyQS T-box results in two distinct FRET states
To observe directly the binding of tRNA to the T-box, we placed the donor dye (Cy3) on the 3' end of a T-box fragment (T-box182), and the acceptor dye (Cy5) on the 5' end of the tRNA Gly , where the subscript "182" denotes the length of the T-box construct (SI Appendix, Figs. 1A and S1). T-box182 spans Stem I, the linker sequence, Stem III and the anti-terminator, but does not contain the terminator sequence, thereby preventing the transition to the terminator conformation.
A short RNA extension sequence was added to the 5' end of the T-box for surface immobilization To further confirm the assignment of the FRET states, we generated T-box149, where the antiterminator sequence is truncated (Figs. 1A and SI Appendix, S1). Based on the structure model from the SAXS data (17) we predicted that a Cy3 dye placed either at the end of Stem III (T-box149) or at the end of the anti-terminator stem (T-box182) are localized in close proximity in three dimensions, further confirmed by the distance measurement using smFRET (SI Appendix, Fig.   S4). Therefore, we expect that if tRNA Gly -Cy5 can reach the same fully bound state in T-box149 as in T-box182, a high FRET state centered at 0.7 would be observed. However, using T-box149-Cy3(3') in combination with tRNA Gly -Cy5, we again observed transient binding of tRNA Gly with a FRET value centered at ~0.4 with the same average lifetime as observed with the T-box182-Cy3 (3') and tRNA ∆NCCA -Cy5 combination (Fig. 2B-D and SI Appendix, Fig. S3B and C). Therefore these two complexes (T-box182 + tRNA ∆NCCA and T-box149 + tRNA Gly ) represent the same binding state of the tRNA, i.e. the state where binding of the anticodon to the specifier region has been established, but is unstable without the further interactions between the NCCA and the t-box region.
Collectively, our results suggest a two-step binding model where the establishment of the interaction with the anticodon precedes the interactions with the NCCA. Without the interaction between the NCCA and the t-box sequence the binding of tRNA Gly is not stable. From the binding kinetics of tRNA ∆NCCA , we estimated the association rate constant (k1) and the disassociation rate constant (k-1) for the first binding step to be (5.0±1.7)X10 5 s -1 ·M -1 and 0.28±0.04 s -1 , respectively (Figs. 5 and 6E and SI Appendix).

The transition from anticodon recognition to NCCA binding is rapid for uncharged tRNA
We classified smFRET traces for T-box182-Cy3(3') in complex with tRNA Gly -Cy5 into three types  (Type II) suggests that the NCCA/t-box interaction can break occasionally (Fig. 2B). We estimated the lifetime of the transiently sampled partially bound state to be 0.35±0.09 s (SI Appendix, Fig. S3A), ~10-fold shorter than the partially bound state without the NCCA end.
While the majority of the T-box molecules were already bound to tRNA Gly before starting data acquisition, we could detect that some molecules show real-time binding during imaging acquisition. We observed only a few traces briefly sampling the 0.4 FRET state from the zero FRET (unbound) state before reaching the 0.7 FRET state, while most traces directly sampled the 0.7 FRET state without a detectable 0.4 FRET, likely due to our imaging time resolution (100 ms per frame). We post-synchronized the FRET traces at the transition point from the zero FRET state to the first sampled 0.4 FRET state, and plotted in a time-evolved FRET histogram. From the time-evolved FRET histogram (Fig. 2F), we estimated roughly that the upper limit of the lifetime spent at the 0.4 FRET state is ~100 ms, very rapidly followed by establishment of NCCA/tbox interactions. In contrast, tRNA ∆NCCA could not pass the 0.4 FRET state. To capture better realtime binding, we performed a flow experiment, where tRNA Gly -Cy5 was flowed in to a chamber with immobilized T-box182-Cy3(3') during imaging acquisition. The corresponding postsynchronized time-evolved FRET histogram again shows a fast transition into the fully bound state (SI Appendix, Fig. S5). In addition, the association rate constant of tRNA Gly in the real-time flow experiment is (7.5±0.7)x10 5 s -1 ·M -1 , consistent with the k1 of tRNA ∆NCCA and confirming that the NCCA end of the tRNA does not participate in the first binding step.
From the real-time binding kinetics of tRNA Gly to T-box182, we estimated a transition rate constant from the partially bound state to the fully bound state (k2) of ~10 s -1 (Fig. 2F and SI Appendix, Fig.   S5). On the other hand, as transitions back to the partially bound state from the fully bound state were only observed in ~10% traces, we interpreted this to mean that the reverse transition rate constant (k-2) is very small, and the second binding step in the wild-type (WT) T-box with uncharged tRNA Gly is close to irreversible (Figs. 5 and 6E).

Establishment of the NCCA/t-box interaction is accompanied by conformational changes in the T-box riboswitch
We next investigated whether tRNA binding requires any conformational changes in the T-box itself. Using doubly labeled T-box182, with Cy3 at the 3' end and Cy5 at the 5' hybridization extension, we observed a high FRET state (centered at ~0.75) in the absence of tRNA (SI Appendix, Fig. S4). Based on the structural model (17), we estimated the distance between the 5' and 3' ends of the T-box182 to be ~36 Å (Fig. 1B). Our measured FRET value is slightly less than the predicted FRET value (~0.9), likely due to the engineered 5' extension sequence used to immobilize the T-box. No noticeable change was detected upon incubation with unlabeled tRNA Gly (SI Appendix, Fig. S4C), indicating that the 3' portion (Stem III plus the anti-terminator stem) does not move away from the 5' portion (Stem I). Given that the measured FRET efficiency of 0.75 is already located beyond the FRET sensitive region, it is unlikely that any inward motion of the 3' portion relative to the 5' could be detected. To overcome this limitation, we added extensions at both the 3' and 5' ends ( Fig. 3A and SI Appendix, Fig. S1). ITC experiments suggest that addition of a 5' and/or a 3' extension sequences to the T-BOX does not affect tRNA binding (SI Appendix, Fig. S5). With this intra-T-box FRET scheme, we observed a FRET shift from ~0.5 to ~0.65 when tRNA Gly was added (Fig. 3B), indicating that the 3' half of the T-box moved closer to the 5' half, potentially with the T-box becoming more compact due to the presence of the cognate tRNA Gly . Adding non-cognate tRNA Phe or tRNA ∆NCCA gave similar FRET values as the Tbox alone (Fig. 3B), suggesting that the conformational change is associated with binding of the NCCA, not with anticodon recognition.

The hinge of the intra-T-box conformational change is located near the K-turn region
Based on the known structures of Stem I in complex with tRNA Gly (11,14), it was hypothesized that the intra-T-box conformational change associated with the NCCA interaction is likely to involve the K-turn region. If this were the case, FRET between labels on the tRNA and near the K-turn region will be insensitive to the two binding states of the tRNA. Based on structural studies (11,14,17) (Fig. 1B), the 5' end of the T-box is in close proximity to the K-turn. We measured FRET between a Cy3 placed at the 5' end of the T-box (T-box182-Cy3(5')) and tRNA Gly -Cy5 (Fig.   4A). As predicted, using this FRET pair binding of both tRNA Gly and tRNA ∆NCCA generated a similar

A mutation in the T-loop region affects the first binding step but has minimal effect on the second binding step.
The interdigitated T-loops structure formed by the interactions between the distal bulge and the apical loop at the distal end of Stem I has been shown to be important for tRNA binding (9)(10)(11).
Specifically, C56 of T-box stacks on a nucleobase in the D-loop of tRNA, and a point mutation of C56 to U has been shown reduce the tRNA binding affinity by ~40 fold (11). We introduced the same mutation in the T-box182 backbone (T-boxC56U) (Fig. 6A and SI Appendix, S1). The smFRET trajectories for tRNA Gly , Fig. S7A). Comparison of tRNA Gly binding to T-boxC56U and T-box182 suggest that the C56U mutation does not affect the second binding step. To investigate whether the mutation at the T-loop region affects the first binding step, we analyzed the binding and dissociation of tRNA ∆NCCA -Cy5 to T-boxC56U-Cy3(3'). We found that the k1 of tRNA binding to T-boxC56U was roughly 16-fold slower compared to tRNA binding to T-box182, and the dissociation was roughly 2.5-fold faster compared to WT T-box ( Fig. 6E and SI Appendix, Fig.   S7), leading to a ~40 fold higher dissociation constant for the first binding step. Our results suggest that the T-loop region of the T-box is critical during the first binding step, potentially aiding in anticodon recognition, but does not contribute significantly to the second binding step.

A truncation of Stem III has a minor effect on tRNA binding
The functional role of Stem III is unclear. It has been speculated that Stem III might serve as a transcription stalling site to allow co-transcriptional folding and regulation of the T-box riboswitch (21,22). In addition, a SAXS data-derived model suggested coaxial stacking of Stem III and the anti-terminator stem, leading to a plausible role of Stem III in stabilizing the anti-terminator conformation in the presence of uncharged tRNA Gly (17). To investigate the latter hypothesis, we generated a T-box mutant (T-boxSIII-Δ4bp), in which four base pairs in Stem III are deleted to significantly shorten its length. smFRET studies using T-T-boxSIII-Δ4bp-Cy3(3') with tRNA ∆NCCA -Cy5 and tRNA Gly -Cy5 revealed insignificant difference in overall kinetics in the first and second step bindings ( Fig. 6 and SI Appendix, Fig. S8A). Noticeably, the T0.7 was around 50% shorter than that for the T-box182 (Fig. 6E), indicating that Stem III may contribute to the stabilization of the fully bound state, potentially through coaxial stacking with the anti-terminator stem, but the effect is minor.

A K-turn mutation affects both binding steps
As our smFRET data highlight the role of a region near the K-turn as the hinge of the tRNA binding-dependent conformational change, we investigated the role of the K-turn in regulating tRNA binding kinetics. We disrupted the K-turn (T-boxΔKT) by changing the 6 bulged nucleotides to 3 nucleotides (UCA) to replace the K-turn with a 3 base pair stem (Fig. 6A and SI Appendix,   S1). In contrast to binding of tRNA Gly to T-box182, binding to T-boxΔKT binding results in three FRET states centered on 0.2, 0.4, and 0.7. (Fig. 6B). While the exact boundary of each FRET state is difficult to determine accurately from the FRET histogram (Fig. 6C), a transition density plot (TDP) clearly revealed interconversion between the 0.2, 0.4, and 0.7 states (Fig. 6D) s -1 , respectively ( Fig. 6E and SI Appendix, Fig. S9C). The dramatically reduced k2_app (~100-fold slower than k2 of T-box182) and increased k-2_app in T-boxΔKT implies that inflexibility of the K-turn region largely inhibits the conformational change in the T-box required to form the NCCA/t-box interaction, and strongly destabilizes the fully bound state.

DISCUSSION
tRNA recognition by a T-box riboswitch is a bipartite process. Stem I is largely responsible for discriminating non-cognate tRNAs, while the t-box sequence in the expression platform senses the charged state of the tRNA. Here, we used smFRET to elucidate the binding kinetics of tRNA Gly by the glyQS T-box riboswitch. With three FRET pairs between different T-box riboswitches and tRNA ligands, our data collectively reveals a two-step binding model of uncharged tRNA Gly to the glyQS T-box riboswitch (Fig. 5). The first binding step involves recognition of the anticodon of the tRNA by the specifier sequence located in Stem I of the T-box riboswitch, leading to a partially bound state. In the second step, the 3' end of the T-box docks into the NCCA end of the tRNA through interactions with the t-box sequence, which leads to a fully bound state. Without the NCCA interaction, the binding of tRNA is unstable, with an average lifetime of ~4 s, whereas with interactions both with the anticodon and the NCCA end, the binding of tRNA is very stable, with an average lifetime > 24 s. The later lifetime measurement is limited by fluorophore photobleaching and is likely to be much longer.
While our manuscript was in preparation, Suddala et al. (24) reported a single-molecule study on tRNA binding to the glyQS T-box riboswitch and proposed a two-step binding model, very similar to our model. Although both studies propose highly consistent kinetic models, in the study by Suddala et al. (24), the FRET pair attached at the variable loop of the tRNA and the 3' or 5' ends of the glyQS T-box cannot distinguish between the partially bound state from the fully bound state; therefore the two binding states are distinguished by different dissociation rates of the tRNA from these states, aided by using a Stem I-only mutant that cannot interact with the NCCA end of the tRNA. Neither transitions between the partially bound and the fully bound states, nor the order of events during tRNA binding can be resolved in the study (24). In contrast, by employing a FRET pair located at the 5' end of the tRNA and the 3' end of the glyQS T-box, we observed directly two FRET states corresponding to the recognition of the anticodon (0.4 FRET) and the binding of the NCCA (0.7 FRET), therefore our study allows the discrimination between different states and generates a more complete kinetic framework describing the full trajectory of the tRNA binding process. Specifically, our data reveal that anticodon recognition precedes the NCCA end interactions, and that after anticodon recognition the commitment to further establishment of the NCCA/t-box interaction is high. The T-box/tRNA Gly complex transits rapidly from the partially bound state to the fully bound state, with a rate constant (k2) of 10 s -1 . Interestingly, our data also reveal that in the fully bound state the NCCA/t-box interaction is not highly-stable or ultra-stable.
Brief disruption of the NCCA/t-box interactions can occur, but transition back to the fully bound state is rapid, ~10 fold faster than dissociation of the tRNA from the partially bound state; therefore tRNA can remain bound during the breaking and reforming of the NCCA/t-box interaction.
However, overall, such transient breaking of NCCA/t-box interaction was only observed in ~10% of the total population, suggesting that the reverse transition rate constant (k-2) is very small, and the second binding step in the WT T-box with uncharged tRNA Gly is close to be irreversible.
The first binding step, interactions with Stem I, involves more than anticodon recognition. Crystal structures of Stem I in complex with tRNA (11,14) show the predicted interaction between the interdigitated T-loops in Stem I and the elbow region (D-and T-loops) of the tRNA. Our Tbox/tRNA FRET pairs and the intra-T-box FRET pair do not report directly on the formation of Tloops/elbow interactions, and therefore cannot resolve whether the T-loops/elbow interactions proceed before or after anticodon recognition. Nevertheless, the importance of this interaction is captured by smFRET measurements using a point mutation that impairs this interaction (11). The mutant shows a dramatic decrease in the association rate constant and a moderate increase in the dissociation rate constant, leading to an overall ~40 fold reduction on the binding affinity for the first step, but the second step is unaffected, suggesting that establishment of the T-loops/elbow interactions is an important part of the Stem I/tRNA recognition process, but plays a minimal role in NCAA recognition or binding. Consistent with the study by Suddala et al. (24), a FRET pair at the 5' and 3' ends of the T-box generates a high FRET value that is insensitive to any tRNA binding, suggesting that the T-box is largely pre-organized in a folded state before tRNA binding. However, by adding an additional extension sequence at the 3' end, we find that the 3' half of the T-box (including Stem III and the anti-terminator) moves inward relative to the 5' half (Stem I) of the T-box to accommodate the interaction with the NCCA end. The observation that such conformational change is only associated with the presence of tRNA Gly indicates that this intra-T-box conformational change is a concerted motion to aid in the establishment of the NCCA/tbox interaction. From the FRET pair placed on the tRNA and at the 5' end of Stem I, we confirm that the hinge is likely to be near the K-turn region. Furthermore, the crystal structures (11,14) show that Stem I flexes around the Kturn region and this flexing seems to be important to establish the interactions between the anticodon and specifier sequence. Our smFRET experiments using a mutant where the K-turn region is removed show that in the absence of the K-turn motif, both the first and second binding steps are significantly affected, emphasizing the importance of the flexing around the K-turn region in the overall recognition and binding process. Interestingly, Stem III does not appear to play a major role in the tRNA binding process in vitro, with a minor effect on the stability of the fully bound state. Stem III may contribute to the stabilization the anti-terminator conformation in vitro; however, potentially its major function is to create a pause site to coordinate with the cotranscriptional folding of the T-box (21,22) . fluorescence intensity vs. time trajectories were corrected for baseline and bleed-through in MATLAB as previously described (26) and fit with a Hidden Markov Model using vbFRET (27). Dwell times of each FRET state before transition to another FRET state were extracted from the idealized traces and fit with a single or double exponential decay with Origin 7.0 (OriginLab).

T-box and tRNA samples (SI
Full methods and references can be found in the SI Appendix.