Allosteric collaboration between elongation factor G and the ribosomal L1 stalk directs tRNA movements during translation

Single-cysteine variants of ribosomal proteins L1 and L9 were fluorescently labeled with Cy5- and Cy3-maleimides, respectively, and reconstituted into large, 50S ribosomal subunits purified from an L1/L9 double-deletion strain of Escherichia coli (Fig. 1). Functional testing of dual-labeled 50S subunits using a standard primer extension inhibition assay (16, 17) demonstrated ≈90% activity through the first round of translation elongation and ≈70% activity in a second round of translation elongation (Methods, supporting information (SI) Methods and Fig. S1).

Dual-labeled 50S subunits were used to enzymatically prepare a ribosomal initiation complex (INI) containing fMet-tRNA^fMet at the P site (9, 13, 18). Delivery of Phe-tRNA^Phe, in complex with elongation factor Tu (EF-Tu) and GTP in the presence of EF-G to INI generates a POST complex, POST_fM/F (where the subscript denotes the presence of deacylated tRNA^fMet in the E site and fMet-Phe-tRNA^Phe in the P site) (9, 13, 18).

By using POST_fM/F and INI, two PRE complexes were formed. Delivery of EF-Tu(GTP)Lys-tRNA^Lys to POST_fM/F generates PRE_F/K (where the F/K subscript now denotes deacylated tRNA^Phe at the P site and fMet-Phe-Lys-tRNA^Lys at the A site). Likewise, delivery of EF-Tu(GTP)Phe-tRNA^Phe to INI generates PRE_fM/F, carrying deacylated tRNA^fMet at the P site and fMet-Phe-tRNA^Phe at the A site (9, 13, 18). Two corresponding PRE-complex analogs were formed by reacting POST_fM/F with puromycin to generate PMN_F/−, containing a deacylated tRNA^Phe at the P site and a vacant A site, and reacting INI with puromycin to generate PMN_fM/−, carrying a deacylated tRNA^fMet at the P site and a vacant A site (9, 13, 18) .

Analysis of steady-state smFRET vs. time trajectories reveals the presence of three trajectory subpopulations within each of the PRE/PMN complexes (Fig. 2). For all PRE/PMN complexes, the major subpopulation exhibits fluctuations between two well-defined FRET states centered at 0.56 ± 0.01 and 0.34 ± 0.01 FRET (Figs. 2 and 3). Based on close agreement with the smFRET_L1-L9 values predicted from cryo-EM reconstructions of the open and closed L1 stalk [≈0.67 FRET and ≈0.35 FRET, assuming R₀ ≈ 55 Å (19, 20)], the 0.56 and 0.34 FRET states were assigned to the open and closed L1 stalk conformations, respectively. Thus, we will refer to this trajectory subpopulation as SP_fluct. The remaining two subpopulations exhibit either stable 0.56 FRET (SP_open) or stable 0.34 FRET (SP_closed) before fluorophore photobleaching (Figs. 2 and 3). SP_open is attributed to: (i) contaminating amounts of POST_fM/F or INI that failed to react with EF-Tu(GTP)aminoacyl-tRNA or puromycin and (ii) PRE/PMN complexes that exhibited photobleaching directly out of the open conformation before undergoing a open → closed L1 stalk transition. SP_closed is attributed to PRE/PMN complexes that exhibited photobleaching directly out of the closed conformation before undergoing a closed → open L1 stalk transition. We note here that PRE/PMN trajectories that occupy SP_open and SP_closed do not correspond to static subpopulations of PRE/PMN complexes that are somehow distinct from the fluctuating subpopulation of PRE/PMN complexes. Rather, the occupancies of SP_fluct, SP_open, and SP_closed are simply determined by a competition between open ⇄ closed L1 stalk transitions and photobleaching from the open and closed states (Methods and refs. 18 and 21; see also Table S1).

Fluctuations within SP_fluct for all PRE/PMN complexes occur within one frame (Fig. 2), suggesting that open ⇄ closed L1 stalk transitions occur without sampling any intermediate state(s), at least not within our time resolution (0.10 sec frame⁻¹, see SI Methods for information regarding the time resolution of the smFRET data). Consistent with this, transition density plots reveal the existence of two major L1 stalk transitions, open → closed and closed → open, with no evidence of any significantly populated intermediate state(s) (Fig. S2).

Rates for L1 stalk closing and opening (k_close and k_open) were extracted by using dwell time analyses of SP_fluct for all PRE/PMN complexes (Methods, Fig. S2, and Table 1). In addition, rates for the formation and disruption of the L1 stalk-tRNA contact (k_L1•tRNA and k_L1○tRNA) have been previously reported for PRE complexes analogous to PRE_F/K and PMN_F/− (ref. 13 and Table 1) and are measured and reported here for PRE_fM/F and PMN_fM/− (Fig. S3 and Table 1). Table 1 demonstrates the close agreement between k_close and k_L1•tRNA and between k_open and k_L1○tRNA for all PRE/PMN complexes. Significantly, k_close and k_open exhibit a dependence on the presence of an A-site peptidyl-tRNA that very closely mirrors the dependence observed for k_L1•tRNA and k_L1○tRNA (compare changes in k_close to those in k_L1•tRNA and changes in k_open to those in k_L1○tRNA for PRE_F/K vs. PMN_F/− and PRE_fM/F vs. PMN_fM/−). The slight discrepancy between k_close and k_L1•tRNA in PRE_fM/F vs. PMN_fM/− most likely originates from the presence of the Cy3 fluorophore on tRNA^fMet in the k_L1•tRNA measurement (note that Cy3 on tRNA^fMet is at a different position than on tRNA^Phe) or, less likely, suggests that coupling between closing of the L1 stalk and movement of tRNA into the hybrid configuration might depend on the identity of P-site tRNA. Likewise, Table 1 demonstrates that k_close and k_open exhibit a dependence on the identity of the P-site tRNA that very closely mirrors the dependence observed for k_L1•tRNA and k_L1○tRNA (compare changes in k_close to those in k_L1•tRNA and changes in k_open to those in k_L1○tRNA for PRE_fM/F vs. PRE_F/K and PMN_fM/− vs. PMN_F/−). Here, our observations are consistent with the well-documented propensity of tRNA^fMet to occupy the classical configuration (11, 22, 23). Collectively, our data provide strong support for the tight coupling of open → closed L1 stalk and classical → hybrid tRNA transitions on the one hand and closed → open L1 stalk and hybrid → classical tRNA transitions on the other.

We have previously shown that addition of 1 μM EF-G in the presence of GDPNP (a nonhydrolyzable GTP analog) to a PMN complex analogous to PMN_F/− significantly inhibits L1•tRNA → L1○tRNA transitions, thereby shifting the L1○tRNA ⇄ L1•tRNA equilibrium strongly toward L1•tRNA (13). In a completely analogous manner, addition of 1 μM EF-G(GDPNP) to PMN_F/− strongly inhibits closed → open L1 stalk transitions such that the open ⇄ closed L1 stalk equilibrium shifts to favor the closed L1 stalk conformation, and the rate of photobleaching from the closed L1 stalk conformation effectively outcompetes closed → open transitions; the overall effect is a decrease in the occupancy of SP_fluct and a corresponding increase in the occupancy of SP_closed (Figs. 2 and 4A and Fig. S4). Because EF-G(GDPNP)-bound PMN_F/− occupies SP_closed, k_open, and k_close for EF-G(GDPNP)-bound PMN_F/− cannot be calculated (Table 1). Nevertheless, it is clear that without directly contacting either the P-site tRNA or the L1 stalk, binding of EF-G(GDPNP) to PMN_F/− suppresses both L1•tRNA → L1○tRNA and closed → open L1 stalk transitions; this observation strongly suggests that during translocation, EF-G-ribosome interactions allosterically regulate tRNA as well as L1 stalk dynamics.

Addition of 1 μM EF-G(GDPNP) to PMN_fM/− has a dramatically different effect than that observed for PMN_F/−. Rather than shifting the trajectory subpopulation occupancy toward SP_closed, binding of EF-G(GDPNP) to PMN_fM/− leads to preferential occupancy of SP_fluct (Fig. 2 and Fig. S4). However, contour plots of the time evolution of population FRET reveal that, like PMN_F/–, PMN_fM/− preferentially occupies the closed conformation of the L1 stalk in the presence of EF-G(GDPNP) (Fig. 4). To investigate the kinetic basis for the preferential occupancy of the closed L1 stalk conformation, we determined k_close and k_open for PMN_fM/− in the absence and presence of 1 μM EF-G(GDPNP) (Table 1 and Table S2). The data in Table 1 demonstrate that EF-G(GDPNP) primarily increases k_close by ≈8-fold and has only a relatively minor effect on k_open. Thus, rather than suppressing closed → open L1 stalk transitions, as was observed for PMN_F/−, EF-G(GDPNP) increases k_close by destabilizing the open conformation of the L1 stalk in PMN_fM/−, resulting in an overall shift of the open ⇄ closed equilibrium analogous to what is observed for EF-G(GDPNP) binding to PMN_F/− (i.e., the equilibrium shifts to favor the closed L1 stalk conformation). Here, the increased k_close and unchanged k_open yield a decrease in the occupancy of SP_open and a corresponding increase in the occupancy of SP_fluct and, to a lesser extent, SP_closed (Figs. 2 and 4). These results reveal that, although the overall effect of EF-G binding to PRE complexes is to shift the open ⇄ closed L1 stalk equilibrium toward the closed L1 stalk conformation, distinct kinetic mechanisms that depend on the identity of the P-site tRNA are used to accomplish this. Fully consistent with these results, smFRET_L1-tRNA experiments on a PMN complex analogous to PMN_fM/− reveal that in the presence of 1 μM EF-G(GDPNP), the majority of smFRET_L1-tRNA trajectories fluctuate between L1○tRNA and L1•tRNA, with a preference for L1•tRNA that is primarily driven by an ≈4-fold increase in k_L1•tRNA (Fig. S3 and Table 1).

In addition to characterization of L1 stalk dynamics within PRE/PMN complexes, the smFRET_L1-L9 signal allows investigation of L1 stalk dynamics within POST complexes (Fig. 5). The majority of POST_fM/F trajectories occupy SP_open, indicating a strong preference for the open L1 stalk conformation (Figs. 2 and 5A). Within SP_fluct, k_close = 0.10 ± 0.13 sec⁻¹ and k_open = 0.99 ± 0.16 sec⁻¹, also yielding a preference for occupying the open L1 stalk conformation. Because of heterogeneity in the tRNA occupancy of the E site, we generated a homogenous POST_−/F complex by quantitatively dissociating tRNA^fMet from POST_fM/F (refs. 24 and 25 and Fig. S5). Fig. S6 shows that quantitative dissociation of the E-site tRNA decreases the occupancy of SP_fluct in favor of SP_open, strengthening the preference of the POST complex for the open L1 stalk conformation.

To test the generality of these results, we repeated these experiments on POST_F/K. In contrast to POST_fM/F, we find that only a minority of POST_F/K trajectories occupy SP_open. Instead, the majority of POST_F/K trajectories occupy SP_fluct, with k_close = 0.31 ± 0.09 sec⁻¹ and k_open = 0.76 ± 0.22 sec⁻¹, again generating a preference for the open L1 stalk conformation (Figs. 2 and 5B). Because heterogeneity in the E-site tRNA occupancy of POST_fM/F and POST_F/K is similar (Fig. S5), our data suggest that L1 stalk dynamics in POST complexes may depend on the identity of the E-site tRNA. Quantitative dissociation of deacylated tRNA^Phe from the E site of POST_F/K (Fig. S5) reveals that the majority of POST_−/K trajectories occupy SP_open, completely analogous to our observations on POST_−/F (Fig. S6). Thus, although the L1 stalk within POST complexes exhibits an overall preference for the open conformation, the kinetics underlying this preference depend on the presence and identity of the E-site tRNA. This observation implies that each tRNA species might make slightly different and unique binding interactions with the ribosomal E site.

To assess the dynamics of the L1 stalk within a homogeneous POST complex containing a fully occupied E site, we artificially delivered 1 μM deacylated tRNA^fMet to POST_−/F to generate POST_fM*/F and deacylated tRNA^Phe to POST_−/K to generate POST_F*/K. In stark contrast to our results for POST_fM/F and POST_F/K, containing authentically translocated E-site tRNAs, we find that the majority of POST_fM*/F and POST_F*/K trajectories preferentially occupy the closed L1 stalk conformation (compare Fig. 5A with Fig. S6D and Fig. 5B with Fig. S6E). Preliminary subpopulation and kinetic analysis of POST_fM*/F and POST_F*/K (Fig. S6) indicate that, similar to our results for POST complexes containing authentically translocated E-site tRNAs, L1 stalk dynamics in POST complexes containing artificially delivered E-site tRNAs may also depend on the identity of the E-site tRNA. It should be stated, however, that possible compositional heterogeneity arising from the incomplete binding of deacylated tRNA to the ribosomal E site and/or from reverse translocation of POST_fM*/F and POST_F*/K (these experiments were conducted in the absence of EF-G) (26, 27) precludes detailed subpopulation and kinetic analysis. Regardless, it is clear that although authentically translocated E-site tRNAs exhibit a strong preference for the open conformation of the L1 stalk, artificially delivered E-site tRNAs instead generate a preference for the closed L1 stalk conformation. Presumably, this closed L1 stalk conformation is identical to the “half-closed” conformation that has been observed by Cornish et al. in similarly prepared POST complexes (i.e., containing an artificially delivered E-site tRNA) (15). Thus, it seems that our smFRET_L1-L9 signal cannot distinguish between the fully closed L1 stalk conformation observed in PRE/PMN complexes and the half-closed L1 stalk conformation observed in POST complexes containing an artificially delivered E-site tRNA; this is perhaps not surprising given the smaller dynamic range of the smFRET_L1-L9 signal in this work relative to the L1-L33 smFRET signal in Cornish et al. (15) Thus, whether or not the half-closed L1 stalk conformation observed by Cornish et al. in POST complexes containing artificially delivered E-site tRNAs is sampled in POST complexes containing authentically translocated E-site tRNAs remains to be determined. Regardless, our results demonstrate that L1 stalk dynamics within POST complexes are sensitive to the mechanism through which the deacylated tRNA enters the E site.

Previously, we have proposed that PRE/PMN complexes spontaneously and reversibly fluctuate between two major conformational states: global state 1 (GS1), encompassing classically bound tRNAs, an open L1 stalk and a nonratcheted ribosome and global state 2 (GS2), encompassing hybrid-bound tRNAs, a closed L1 stalk and a ratcheted ribosome (13). Consistent with this model, our smFRET_L1-L9 results demonstrate that the L1 stalk within PRE/PMN complexes exists in an open ⇄ closed dynamic equilibrium that exhibits kinetics closely matching those of the classical ⇄ hybrid tRNA (9) and the L1○tRNA ⇄ L1•tRNA (13) dynamic equilibria. Most notably, all three equilibria have matching kinetic responses toward the occupancy of the A site by a peptidyl-tRNA and the identity of the P-site tRNA. These kinetic data demonstrate the close coupling between tRNA and L1 stalk dynamics within PRE/PMN complexes. In addition to our smFRET_tRNA-tRNA (9), smFRET_L1-tRNA (13) and smFRET_L1-L9 studies, further support for the GS1 ⇄ GS2 model is provided by two recent smFRET studies from Ha, Noller, and coworkers demonstrating the close correlation between the equilibrium constants governing the nonratcheted ⇄ ratcheted ribosome and open ⇄ closed L1 stalk equilibria (14, 15). In complete agreement with the smFRET results and the GS1 ⇄ GS2 model, two recent cryo-EM studies applied particle classification methods to reveal the existence of two major PRE complex conformations within a single pretranslocation sample, corresponding to GS1 and GS2, respectively (7, 8). Despite the excellent agreement between the GS1 ⇄ GS2 model and the available smFRET and cryo-EM data, however, it remains possible that short-lived and/or rarely sampled intermediates within GS1 → GS2 and/or GS2 → GS1 transitions have thus far eluded detection by smFRET experiments or cryo-EM reconstructions. Future smFRET experiments recorded at higher-time resolution or using ribosome-targeting small-molecule translocation inhibitors or mutagenized ribosomes may prove useful tools for uncovering such short-lived and/or rarely sampled intermediates. Nevertheless, the GS1 ⇄ GS2 model represents a simple dynamic model that is consistent with the available data and provides a convenient framework for describing the global dynamics of the translating ribosome.

Binding of EF-G(GDPNP) to PMN complexes strongly shifts the open ⇄ closed L1 stalk equilibrium toward the closed conformation by regulating k_open and/or k_closed. Given that EF-G(GDPNP) binds near the A site of the PRE complex, ≈170 Å away from the hinge region of the L1 stalk (28–31), our data demonstrate that EF-G(GDPNP) regulates L1 stalk dynamics allosterically, through its interactions with the ribosome upon binding to the PMN complex. An attractive hypothesis, consistent with the close coupling of ratcheting, L1 stalk and tRNA dynamics stipulated by the GS1 ⇄ GS2 model, is that EF-G(GDPNP) establishes interactions with the ribosome that directly stabilize the ratcheted conformation of the ribosome, indirectly leading to stabilization of the hybrid P-site tRNA configuration and the closed L1 stalk conformation. Indeed, the discovery that vacant Saccharomyces cerevisiae ribosomes (i.e., not containing tRNA substrates) predominantly exist in a ratcheted conformation with a closed L1 stalk (32) suggests the possibility that the coupling between intersubunit ratcheting and L1 stalk closure might be independent of the presence of a P-site tRNA and may instead be encoded within the architecture of the ribosome itself. Evidence for a similar possibility in prokaryotic ribosomes comes from the correlation between the equilibrium constants governing the nonratcheted ⇄ ratcheted ribosome and open ⇄ closed L1 stalk equilibria in vacant E. coli ribosomes (14, 15); the correlation between the forward and reverse rates of these two processes, however, remains to be investigated. Regardless, in this framework, ratcheting and L1 stalk closure would function allosterically to promote and stabilize the hybrid tRNA configuration during translocation. Future experiments exploring the role of ribosomal structural elements in regulating ratcheting, L1 stalk and tRNA dynamics should allow testing of these hypotheses.

Using smFRET_L1-tRNA and smFRET_L1-L9 signals as reporters for the GS1 ⇄ GS2 equilibrium, we find that EF-G(GDPNP) can shift the GS1⇄ GS2 equilibrium toward GS2 through at least two distinct kinetic mechanisms, the choice of which is regulated by the identity of the P-site tRNA. When tRNA^Phe occupies the P site, EF-G(GDPNP) almost completely suppresses GS2 → GS1 transitions whereas when tRNA^fMet occupies the P site, EF-G(GDPNP) has an almost negligible effect on GS2 → GS1 transitions, instead increasing the rate of GS1 → GS2 transitions by ≈4- to 8-fold. This latter result strongly suggests that EF-G can bind directly to PRE complexes in the GS1 state and actively promote the GS1 → GS2 transition, although the extent to which this is observed strongly depends on the identity of the P-site tRNA.

The conformational dynamics of the L1 stalk observed within POST complexes, which are likely uncoupled from intersubunit ratcheting dynamics because POST complexes are not expected to ratchet (11, 14, 30), are altogether distinct from those observed within PRE/PMN complexes. We find that L1 stalk dynamics within POST complexes are sensitive to the presence and identity of the E-site tRNA as well as to the mechanism through which the tRNA enters the E site. In the presence of a vacant E site, the L1 stalk is almost uniformly found in the stable open conformation. The presence of an authentically translocated E-site tRNA, however, triggers open ⇄ closed L1 stalk fluctuations where k_open and k_close depend on the identity of the E-site tRNA. Despite the differing kinetics, POST complexes containing either authentically translocated tRNA^fMet or tRNA^Phe both favor the open L1 stalk conformation; this stands in contrast with the uniformly half-closed L1 stalk conformation observed within POST complexes containing an artificially delivered E-site tRNA (15) and in X-ray crystal structures of POST-like ribosomes carrying what are likely artificially delivered E-site tRNAs (33–35). Consistent with these observations, we find that artificial delivery of a deacylated tRNA into the E site of a POST complex triggers a strong preference for the closed L1 stalk conformation.

Previously we have reported a stable (i.e., nonfluctuating) high FRET signal between the L1 stalk and an authentically translocated E-site tRNA within a POST complex (13). In contrast, the data we present here demonstrates that the L1 stalk within an analogous POST complex, under identical sample conditions as our previous study, undergoes open ⇄ closed fluctuations. To reconcile these two observations, we propose a model in which the authentically translocated E-site tRNA is reconfigured within the E site such that the direct interaction between the tRNA and the L1 stalk is maintained during the open ⇄ closed fluctuations of the L1 stalk. This model is strongly supported by the observation that the E-site tRNA occupies one configuration in X-ray crystal structures of POST-like complexes bearing a closed, or half-closed, L1 stalk conformation (33–35) but occupies a notably different E-site tRNA configuration in cryo-EM reconstructions of POST complexes bearing an open L1 stalk conformation (30). As originally suggested by the authors of the cryo-EM study (30), reconfiguration of the E-site tRNA such that a direct contact with the opening L1 stalk is maintained may be mechanistically important for E-site tRNA release. That said, we find that k_open is ≈10-fold faster than the rate of passive tRNA dissociation from the E site in POST complexes, indicating that the L1 stalk/E-site tRNA can undergo numerous fluctuations before the E-site tRNA dissociates and strongly suggesting that opening of the L1 stalk is not rate limiting for E-site tRNA release. Finally, the observation that L1 stalk dynamics within a POST complex depend on the identity of the E-site tRNA may reflect a difference in the energetics of reconfiguring each tRNA species within the E site. The molecular basis for this difference likely originates from the slightly different interactions that each tRNA would be expected to make with structural elements of the ribosomal E site.

Collectively, our data demonstrate that differences in the interactions of tRNA^fMet and tRNA^Phe with the elongating ribosome can: (i) bias the kinetics of GS1 ⇄ GS2 transitions in PRE/PMN complexes; (ii) control the kinetic mechanism through which EF-G stabilizes the GS2 state during translocation; and (iii) regulate tRNA and L1 stalk dynamics within POST complexes. It remains to be investigated whether these differences are due to tRNA identity elements that uniquely distinguish initiator tRNA^fMet from all elongator tRNAs (36), thus suggesting that elongator tRNAs will generally exhibit kinetic behavior similar to tRNA^Phe, or whether similarly significant differences in kinetic behavior will be found even among elongator tRNAs. Future smFRET studies using an expanded set of elongator tRNAs and/or tRNA^fMet variants containing mutations to tRNA^fMet identity elements, should reveal which features of tRNA structure and tRNA-ribosome interactions are involved in regulating the kinetic behavior of PRE/PMN and POST complexes.

All experiments were performed in Tris-polymix buffer (50 mM Tris-OAc, 100 mM KCl, 5 mM NH₄OAc, 0.5 mM Ca(OAc)₂, 10 mM 2-mercaptoethanol, 5 mM putrescine and 1 mM spermidine) at 15 mM Mg(OAc)₂ and at pH_{25 °C} = 7.5 (9). smFRET trajectories were recorded by using a home-built total internal reflection fluorescence microscope (13, 18). Each smFRET trajectory was idealized as a hidden Markov model, by using the vbFRET software package (37). Dwell times spent in each state before transitioning were extracted from the idealized smFRET trajectories and the lifetime of each state was determined by exponential fitting of the corresponding one-dimensional population vs. time histogram (9, 13, 18). Transition rates were calculated by taking the inverse of the lifetimes and correcting for the rate of photobleaching from each state. Full methods and references can be found in the SIMethods.

We thank S. Das for managing the R.L.G. laboratory; J. Frank, H. Gao, and X. Agirrezabala for providing us with GS1 and GS2 structural models; and M. M. Elvekrog, D. D. MacDougall, and S. H. Sternberg for valuable discussions and comments on the manuscript. This work was supported by Burroughs Wellcome Fund Grant CABS 1004856 (to R.L.G.), National Science Foundation (NSF) Grant MCB 0644262 (to R.L.G.), National Institutes of Health (NIH)–National Institute of General Medical Sciences Grant 1RO1GM084288-01 (to R.L.G.), NIH Grants 1U54CA121852-01A1 and 5PN2EY016586-03 (to C.H.W.), and an NSF Research Experience for Undergraduates grant (to R.L.S.).