1. Biochemistry and Chemical Biology
  2. Structural Biology and Molecular Biophysics
Download icon

Dynamics of ribosomes and release factors during translation termination in E. coli

Research Article
  • Cited 15
  • Views 2,589
  • Annotations
Cite this article as: eLife 2018;7:e34252 doi: 10.7554/eLife.34252

Abstract

Release factors RF1 and RF2 promote hydrolysis of peptidyl-tRNA during translation termination. The GTPase RF3 promotes recycling of RF1 and RF2. Using single molecule FRET and biochemical assays, we show that ribosome termination complexes that carry two factors, RF1–RF3 or RF2–RF3, are dynamic and fluctuate between non-rotated and rotated states, whereas each factor alone has its distinct signature on ribosome dynamics and conformation. Dissociation of RF1 depends on peptide release and the presence of RF3, whereas RF2 can dissociate spontaneously. RF3 binds in the GTP-bound state and can rapidly dissociate without GTP hydrolysis from termination complex carrying RF1. In the absence of RF1, RF3 is stalled on ribosomes if GTP hydrolysis is blocked. Our data suggest how the assembly of the ribosome–RF1–RF3–GTP complex, peptide release, and ribosome fluctuations promote termination of protein synthesis and recycling of the release factors.

https://doi.org/10.7554/eLife.34252.001

eLife digest

Inside cells, molecular machines called ribosomes make proteins using messenger RNA as a template. However, the template contains more than just the information needed to create the protein. A ‘stop codon’ in the mRNA marks where the ribosome should stop. When this is reached a group of proteins called release factors removes the newly made protein from the ribosome.

Bacteria typically have three types of release factors. RF1 and RF2 recognize the stop codon, and RF3 helps to release RF1 or RF2 from the ribosome so that it can be recycled to produce another protein. It was not fully understood how the release factors interact with the ribosome and how this terminates protein synthesis.

Adio et al. used TIRF microscopy to study individual ribosomes from the commonly studied bacteria species Escherichia coli. This technique allows researchers to monitor movements of the ribosome and record how release factors bind to it. The results of the experiments performed by Adio et al. show that although RF1 and RF2 are very similar to each other, they interact with the ribosome in different ways. In addition, only RF1 relies upon RF3 to release it from the ribosome; RF2 can release itself. RF3 releases RF1 by forcing the ribosome to change shape. RF3 then uses energy produced by the breakdown of a molecule called GTP to help release itself from the ribosome.

Most importantly, the findings presented by Adio et al. highlight that the movements of ribosomes and release factors during termination are only loosely coupled rather than occur in a set order. Other molecular machines are likely to work in a similar way. The results could also help us to understand the molecular basis of several human diseases, such as Duchenne muscular dystrophy and cystic fibrosis, that result from ribosomes not recognizing stop codons in the mRNA.

https://doi.org/10.7554/eLife.34252.002

Introduction

Termination of protein synthesis occurs when a translating ribosome encounters one of the three universally conserved stop codons UAA, UAG or UGA. In bacteria, the release of the nascent peptide is promoted by release factors RF1 and RF2 which recognize the stop codons in the A site and hydrolyze the ester bond in the peptidyl-tRNA bound to the P site, allowing the nascent peptide to leave the ribosome through the polypeptide exit tunnel (Dunkle and Cate, 2010; Nakamura et al., 1996). RF1 and RF2 bind to the ribosome in the space between the small and large ribosomal subunits. RF1 and RF2 differ in their stop codon specificity: RF1 utilizes a conserved PET motif to recognize UAG and UAA codons, whereas RF2 uses an SPF motif to recognize UGA and UAA codons. Both RF1 and RF2 have a universally conserved GGQ motif which promotes the catalysis of peptidyl-tRNA hydrolysis (Seit-Nebi et al., 2001); mutations of the GGQ motif to GAQ or GGA inhibit peptide release (Frolova et al., 1999; Mora et al., 2003; Shaw and Green, 2007; Zavialov et al., 2002). After peptide release, RF1 and RF2 dissociate from the post-termination complex to allow for the next steps of translation. The dissociation is accelerated by RF3, a translational GTPase that binds and hydrolyses GTP in the course of termination (Freistroffer et al., 1997; Koutmou et al., 2014; Zavialov et al., 2002). In addition to canonical termination, RF2 takes part in non-canonical termination events such as post-peptidyl transfer quality control (Zaher and Green, 2009) and ribosome rescue on truncated mRNAs (Kurita et al., 2014).

There are two different models concerning the sequence of events during termination, including the timing of peptide release, the order of RF1, RF2 and RF3 binding and dissociation, and the role of nucleotide exchange in RF3 and GTP hydrolysis. The first model of translation termination was proposed by Ehrenberg and colleagues (Zavialov et al., 2001; Zavialov et al., 2002). Based on nitrocellulose filtration experiments, the authors reported that free RF3 has a much higher affinity for GDP (Kd = 5.5 nM) than for GTP (Kd = 2.5 µM) or GDPNP (Kd = 8.5 µM) (Zavialov et al., 2001), which would imply that at cellular GTP/GDP concentrations RF3 is expected to be predominantly in the GDP form. The exchange of GDP for GTP occurs only when RF3–GDP binds to the ribosome in complex with RF1 or RF2 (Zavialov et al., 2001). In the absence of the nucleotide, RF3-dependent RF1/2 recycling is slow, which has been interpreted as an indication for a high-affinity complex of apo-RF3 to the ribosome–RF1/2 complex (Zavialov et al., 2001). Furthermore, because RF3-dependent turnover GTPase activity was stimulated by peptidyl-tRNA hydrolysis, the authors suggested that RF3 binds to the ribosome–RF1 complex only after the peptide is released. Based on these results, Ehrenberg et al. suggested the following sequential mechanism of termination: RF1/RF2 bind to the ribosome and hydrolyze peptidyl-tRNA, allowing RF3–GDP to enter the ribosome occupied by RF1 or RF2 to form an unstable encounter complex. Dissociation of GDP leads to a stable high-affinity complex with RF3 in the nucleotide-free state. The subsequent binding of GTP by RF3 promotes RF1/RF2 dissociation. In the final step, RF3 hydrolyses GTP and as a result dissociates from the ribosome (Zavialov et al., 2001; Zavialov et al., 2002).

An alternative model was proposed based on the kinetic and thermodynamic analysis of GTP/GDP binding to RF3 by ensemble kinetics and equilibrium methods. The results of those experiments indicated that the affinity of RF3 to GDP and GTP is on the same order of magnitude (5 nM and 20 nM, respectively [Koutmou et al., 2014; Peske et al., 2014]). As the cellular GTP concentration is at least 10 times higher than the GDP concentration (Bennett et al., 2009), these affinities imply that nucleotide exchange in RF3 can occur spontaneously, off the ribosome, and thus RF3 could enter the ribosome in either the GTP- or GDP-bound form. Consistent with previous findings (Zavialov et al., 2001; Zavialov et al., 2002), ribosome–RF1 or ribosome–RF2 complexes accelerate nucleotide exchange in RF3 (Koutmou et al., 2014; Peske et al., 2014); however, this effect is independent of peptide release, because also a catalytically inactive RF2 mutant activates nucleotide exchange in RF3 (Peske et al., 2014; Zavialov et al., 2002). Binding of GTP to RF3 in complex with the ribosome and RF2 is rapid (130 s−1) (Peske et al., 2014), and thus the lifetime of the apo-RF3 state would be too short to assume a tentative physiological role. Peptide release results in the stabilization of the RF3–GTP–ribosome complex, thereby promoting the dissociation of RF1/2, followed by GTP hydrolysis and dissociation of RF3–GDP from the ribosome (Peske et al., 2014).

Efficient translation termination not only requires the coordinated action of the release factors, but also entails conformational dynamics of the factors and the ribosome. The key conformational motions of the ribosome during termination and in general in all phases of translation include the rotation of ribosomal subunits relative to each other, the swiveling motion of the body and head domains of the small ribosomal subunit, the movement of the ribosomal protein L1 toward or away from the E-site tRNA, and the movement of tRNAs between classic and hybrid conformation. These motions are loosely coupled and gated by ligands of the ribosome such as translation factors and tRNAs (Adio et al., 2015; Chen et al., 2011; Cornish et al., 2008; Horan and Noller, 2007; Sharma et al., 2016; Shi and Joseph, 2016; Valle et al., 2003; Wasserman et al., 2016). Crystal structures show that termination complexes with RF1 or RF2 are predominantly in the non-rotated (N) state. The P-site tRNA in the complexes is in the classical state and the L1 stalk in an open conformation (Jin et al., 2010; Korostelev et al., 2008; Laurberg et al., 2008; Weixlbaumer et al., 2008). A single molecule fluorescence resonance energy transfer (smFRET) study showed that binding of RF1 to termination complexes stabilizes the open conformation of the L1 stalk, whereas in the absence of RF1 termination complexes make reversible transitions between the open and closed state (Sternberg et al., 2009); the rotation of the ribosomal subunits was not investigated directly in that study. The high sequence similarity between RF1 and RF2 suggests that the two factors interact with the ribosome in the same manner and promote peptide release by a similar mechanism (Freistroffer et al., 1997; Zavialov et al., 2001). However, structures of RF1 or RF2 bound to termination complexes reveal differences regarding the interaction with the L11 region of the 50S subunit (Korostelev et al., 2008; Laurberg et al., 2008; Petry et al., 2005; Pierson et al., 2016; Rawat et al., 2006; Rawat et al., 2003; Weixlbaumer et al., 2008). Thus, it is not clear whether RF1 and RF2 follow the same mechanism and whether they respond in the same way to the recruitment of RF3 to termination complexes.

In the absence of RF1/RF2, binding of RF3 with a non-hydrolyzable GTP analog to the ribosomes where the nascent peptide has been released induces formation of the rotated (R) state of the ribosome, with the tRNA in the P/E hybrid state and the closed conformation of the L1 stalk (Gao et al., 2007; Jin et al., 2011; Zhou et al., 2012). A very similar effect of RF3–GDPNP was found by smFRET (Sternberg et al., 2009). However, it is much less clear what happens when RF1/RF2 and RF3 bind to the ribosome together. Modeling of the atomic structures of RF1 and RF2 into the cryo-EM structure of RF3-bound post-release complex suggests that the RF3-induced ribosome rearrangements break the interactions of RF1/RF2 with both the decoding center and the L11 region of the ribosome, leading to the release of RF1/RF2 (Gao et al., 2007). In this model, stable binding of RF1 and RF3 is mutually exclusive. On the other hand, a cryo-EM structure of ribosomes in complex with a deacylated tRNA in the P site, RF1, and RF3 in the apo form, that is, in the absence of added nucleotide, suggest that both factors can bind simultaneously to the ribosome (Pallesen et al., 2013). smFRET measurements carried out with post-release complexes in the presence of excess RF1 showed that the addition of RF3–GTP induced short-lived transitions from the L1-open to the L1-closed state which were not observed in the absence of RF3. This suggests that the two factors can bind to the ribosome simultaneously (Sternberg et al., 2009). No structural studies are available on the interaction of RF3 with RF2-bound complexes. The interaction of RF3 with the ribosomes prior to peptide release has not been studied.

Here, we use TIRF microscopy to monitor smFRET signals reporting on subunit rotation to follow changes in ribosome conformation in response to RF1, RF2 and RF3 and the binding of each individual release factor to the ribosome during termination. Our results demonstrate how the recruitment of release factors change the ribosome conformation in termination complexes, how the dissociation of the factors is achieved, show differences in the function of RF1 and RF2, and explain the importance of GTP binding and hydrolysis by RF3.

Results

RF1 and RF2 have distinct effects on ribosome dynamics

To monitor the rotation of the ribosomal subunits during termination, we utilized ribosomes with fluorescent labels attached to the small subunit protein S6 and the large subunit protein L9, S6-Cy5 and L9-Cy3, respectively. This FRET pair has been extensively characterized in both smFRET and ensemble kinetics experiments and reports on the formation of the non-rotated (N) or the rotated (R) state of the ribosome (Cornish et al., 2008; Ermolenko et al., 2013; Sharma et al., 2016). We prepared termination complexes on an mRNA which is translated up to the stop codon UAA recognized by both RF1 and RF2. The complexes contain a peptidyl-tRNA in the P site and have a stop codon in the A site; those complexes are denoted as pre-hydrolysis complexes (PreHC). In the absence of termination factors, PreHC is found predominantly in a state with the FRET efficiency of 0.73 ± 0.01 (denoted as 0.7 FRET state in the following) (Figure 1A, Figure 1—figure supplement 1; Supplementary file 1). Previous work has shown that this state corresponds to the N state of the ribosome (Cornish et al., 2008; Qin et al., 2014; Sharma et al., 2016). A small fraction of complexes shows a FRET state with an efficiency of 0.52 ± 0.02 (0.5 FRET state), which corresponds to the R state of the ribosome. While peptidyl-tRNA generally favors the N state, the ability of ribosomes with peptidyl-tRNA in the P site to adopt the R state at room temperature has been demonstrated previously by smFRET and cryo-EM (Cornish et al., 2008; Fischer et al., 2010; Ling and Ermolenko, 2015). The distribution of FRET efficiencies and thus the ratio between N and R conformations of PreHC is independent of the tRNA in the P site and of the presence of a single N-terminal amino acid (fMet) or a dipeptide (fMetPhe, fMetVal or fMetLys) at the P-site tRNA (Figure 1—figure supplement 1; Supplementary file 1). This finding has prompted us to use the PreHC with fMet-tRNAfMet in the P site and a stop codon in the A site as a minimal model system, following previous publications which used this approach to study termination (Casy et al., 2018; Jin et al., 2010; Koutmou et al., 2014; Kuhlenkoetter et al., 2011; Pallesen et al., 2013; Pierson et al., 2016; Shi and Joseph, 2016; Sternberg et al., 2009).

Figure 1 with 3 supplements see all
Subunit rotation of termination complexes in the presence of release factors.

Histograms and Gaussian fits of normalized FRET distributions of S6/L9-labeled termination complexes in the presence of saturating RF1/RF2 concentrations. (A,D) PreHC and PostHC in the absence of RFs. In (D) PostHC* was generated by addition of puromycin to PreHC. (B,E) Same as A,D in the presence of RF1(GAQ) (1 µM) and RF1 (1 µM), respectively. PostHC is formed by the action of RF1. (C,F) Same as B,E in the presence of RF2(GAQ) (1 µM) and RF2 (1 µM), respectively. Cartoons show the complex composition. Grey triangles and brown circles represent the formyl group and the amino acid of fMet, respectively; stars indicate the positions of the Cy3 (green) and Cy5 (red) labels. The red shade of histogram in (D) indicates frequent reversible transitions between N and R states. The grey shade of all other histograms indicates that transitions were observed in less than 20% of traces. FRET values (Supplementary file 1) are calculated from three independent data sets. See also Figure 1—figure supplement 1, Figure 1—figure supplement 2 Figure 1—figure supplement 3 and Supplementary file 1.

https://doi.org/10.7554/eLife.34252.003

To probe the effect of RF1/RF2 binding on subunit rotation, we mixed PreHC with RF1 or RF2 at saturating concentrations of the factors (Zavialov et al., 2002). Peptide release was avoided by using RF1(GAQ) or RF2(GAQ) mutants which are catalytically deficient (Frolova et al., 1999; Zavialov et al., 2002) (Figure 1—figure supplement 2A). Binding of RF1(GAQ) to PreHC stabilizes the N state (Figure 1B). The fraction of the PreHC in the R state, albeit small, is somewhat higher with RF2(GAQ) than with RF1(GAQ) (Figure 1B,C).

The hydrolysis of the ester bond between the tRNA and the nascent peptide in PreHC leads to the formation of post-hydrolysis complex (PostHC). To prepare PostHC without the use of termination factors, we released nascent peptides with the help of puromycin, an analog of the A-site aminoacyl-tRNA that reacts with the peptidyl-tRNA in the P site to form peptidyl-puromycin (which then dissociates from the ribosome) and a deacylated tRNA in the P site. These complexes are denoted as PostHC*. The FRET histogram of PostHC* in the absence of the factors indicate the presence of two states, the 0.5 FRET (R) state and the 0.7 FRET (N) state (Figure 1D, Supplementary file 1). FRET time courses of individual ribosomes show reversible transitions between the N and R states (Figure 1—figure supplement 1B). The exact distribution of states depends on the P-site tRNA (Figure 1—figure supplement 1; Supplementary file 1) (Cornish et al., 2008) with tRNAfMet behaving similarly to tRNAVal, thus underscoring the suitability of the minimal model system.

To test the effect of RF1 and RF2 on subunit rotation of PostHC, we added saturating concentrations of the wild type RF1 or RF2 to PreHC to allow peptide release. RF1 halts PostHC in the N state (Figure 1E, Figure 1—figure supplement 2B,D), in agreement with the previous smFRET study where RF1 binding stabilizes the L1 stalk in the open state (Sternberg et al., 2009). Binding of RF2 to PostHC shifts the equilibrium toward the N state, but not to the same extent as RF1 (Figure 1F, Figure 1—figure supplement 2C,E). Complexes with RF2 make occasional N to R transitions, in particular with RF2(GAQ) bound to PostHC (Figure 1—figure supplement 2C). These initial observations suggest that although both factors favor the N state, RF1 appears more efficient than RF2.

To further probe the potential differences between RF1 and RF2, we monitored subunit rotation in response to factor binding in real time. We injected catalytic amounts of release factors to PreHC and PostHC and recorded the time courses of FRET signal changes (Figure 1—figure supplement 3). RF1(GAQ) binding to PreHC does not change the FRET efficiency appreciably, as the complex is predominantly in the N state with or without the factor (Figure 1—figure supplement 3A). Also the binding of wild-type RF1 to PreHC with subsequent peptide release does not change the FRET efficiency (Figure 1—figure supplement 3B), supporting the notion that stabilization of the N state by RF1 is independent of peptide release (Figure 1B,E). PostHC without factor fluctuates between the N and R state; binding of RF1 to PostHC halts fluctuating ribosomes in the N state and prevents excursions to the R state (Figure 1—figure supplement 3C).

With RF2 the picture is somewhat different. PreHC–RF2(GAQ) is predominantly in the N state (Figure 1—figure supplement 3D). However, binding of wild type RF2 and peptide release shift the distribution toward the R state (Figure 1—figure supplement 3E). The resulting PostHC fluctuates between N and R states as shown by synchronization of FRET traces to the first N to R transition. PostHC obtained by puromycin treatment also shows reversible N to R transitions which remain undisturbed by the addition of RF2 (Figure 1—figure supplement 3F). Although the binding of the factors is not directly monitored in these experiments, the differences in the rotation pattern suggest that RF1 and RF2 have distinct effects on ribosome dynamics. Such differences may result from a shorter residence time of RF2 compared to RF1 on the ribosome, which we tested in the following experiments.

Binding of RF1 and RF2 to the ribosome

To measure how long the factors remain bound to the ribosome, we prepared Cy5-labeled RF1 and RF2, as well as the respective RF1/2(GAQ) mutants and ribosomes containing Cy3-labeled protein L11 (Adio et al., 2015; Chen et al., 2011; Geggier et al., 2010; Holmberg and Noller, 1999; Stöffler et al., 1980) (Figure 2—figure supplement 1A). The biochemical activity of labeled release factors was indistinguishable from that of the unlabeled counterparts (Figure 2—figure supplement 1B,C) and the factors were fully methylated (Figure 2—figure supplement 2). L11 constitutes part of the factor binding site (Pallesen et al., 2013; Petry et al., 2005; Rawat et al., 2006; Rawat et al., 2003). Recruitment of the factors to the ribosome is expected to result in high FRET efficiency. Binding of RF1 or RF1(GAQ) to either PreHC or PostHC results in a single FRET population centered at 0.72 ± 0.02 (0.7 FRET) (Figure 2A–C, Supplementary file 1). RF1 and RF1(GAQ) are stably bound to the ribosome, in agreement with previous biochemical reports on dissociation rates of RF1 and RF1(GAQ) (0.005–0.1 s−1; [Koutmou et al., 2014; Shi and Joseph, 2016]). The estimated upper limit of the dissociation rate in our experiments is 0.2 s−1 (Supplementary file 1), defined by the photobleaching rate of the FRET dye pair with kphotobleaching = 0.07–0.19 s−1 at the given imaging conditions (Materials and methods). Binding of RF2 or RF2(GAQ) to PreHC or PostHC leads to single FRET populations with efficiencies between 0.6 and 0.7 (Figure 2D–F, Supplementary file 1). However, the residence time of RF2 is much shorter compared to RF1 or RF1(GAQ), with the koff values in the range from 0.8 ± 0.1 s−1 to 1.3 ± 0.2 s−1 (Figure 2D–F, Supplementary file 1). Peptide hydrolysis has no visible effect on RF1 and only a minor effect on RF2 dissociation (in the absence of RF3).

Figure 2 with 3 supplements see all
Residence times of RF1 and RF2 on PreHC and PostHC.

(A–C) smFRET upon addition of RF1-Cy5 or RF1(GAQ)-Cy5 to PreHC or PostHC labeled at protein L11 with Cy3. FRET values (mean ± sd) center at 0.72 ± 0.02 (A), 0.71 ± 0.01 (B), and 0.71 ± 0.01 (C). (D–F) smFRET upon addition of RF2-Cy5 or RF2(GAQ)-Cy5 to PreHC or PostHC labeled at protein L11 with Cy3. FRET values (mean ± sd) center at 0.65 ± 0.03 (D), 0.56 ± 0.05 (E), and 0.70 ± 0.04 (F). Experiments were carried out with catalytic amounts of labeled release factors (10 nM). Individual traces were combined in contour plots. FRET histograms are plotted to the right of the contour plots. In (B,E) the peptide was released using puromycin. In the (C,F) peptide was released by RF1 and RF2, respectively. FRET signals were synchronized to the beginning of the FRET signal. koff is the rate of RF1 or RF2 dissociation. Mean FRET values and rate constants with sd were calculated from three independent data sets. See also Figure 2—figure supplement 1, Figure 2—figure supplement 2, Figure 2—figure supplement 3 and Supplementary file 1.

https://doi.org/10.7554/eLife.34252.007

The difference in the dissociation rates of RF1 and RF2 implies that RF1 needs an auxiliary factor, RF3, to help it to dissociate from the ribosome, whereas RF2 may be able to recycle independent of RF3. This notion is consistent with previous reports (Petropoulos et al., 2014; Zavialov et al., 2002) and is further supported by our peptide hydrolysis turnover assay (Figure 2—figure supplement 3A). With catalytic amounts of RF1 in the absence of RF3, that is, when RF1 turnover depends on its intrinsic dissociation rate from the ribosome, termination is essentially blocked, whereas in the presence of RF3 RF1-mediated peptide release is very efficient. In contrast, even catalytic amounts of RF2 are sufficient to complete peptide release from PreHC, although RF3 accelerates the reaction by a factor of 10. Thus, RF3 is essential for RF1, but not for RF2 recycling. Addition of RF3 to PreHC–RF2(GAQ) or PostHC–RF2 complexes makes the complexes more dynamic (Figure 2—figure supplement 3B–E). Our results demonstrate that during canonical termination RF1 and RF2 interact with termination complexes in somewhat different ways, as they have different residence times on the ribosome and respond differently to the presence of RF3.

Interaction with RF3–GTP

Next, we studied how RF3 affects ribosome dynamics and promotes the dissociation of RF1/RF2. To investigate the effect of RF3 on subunit rotation in the absence of RF1 or RF2, we added saturating concentrations of RF3 to S6/L9-labeled PreHC (Figure 3A, Figure 2—figure supplement 1). Binding of RF3 to PreHC, which in the absence of the factor is in the N state, strongly shifts the equilibrium toward the R state (Figure 3A), that is, RF3 has the opposite effect on subunit rotation than RF1 or RF2. The traces are now highly dynamic and show reversible N to R transitions (Supplementary file 1). This finding seems unexpected as ribosomes with peptidyl-tRNA in the P site favor the N state. However, previous cryo-EM and smFRET studies have indicated that those complexes can in fact adopt the R state (Cornish et al., 2008; Fischer et al., 2010; Ling and Ermolenko, 2015). Thus, RF3 appears to bias spontaneous fluctuations of peptidyl-tRNA in the PreHC, rather than induce a previously disallowed conformation.

Figure 3 with 1 supplement see all
Interaction of RF3–GTP with termination complexes.

(A,D) Subunit rotation of S6/L9-labeled PreHC and PostHC* in the presence of excess RF3 (1 µM) with GTP (1 mM). (B,E) Distribution of L1-open and L1-closed states in PreHC and PostHC* labeled at tRNAfMet and protein L1 in the presence of excess RF3 (1 µM) with GTP (1 mM). Smooth red and gray lines show distributions of states without RF3. Color code is the same as in histograms (red, frequent transitions between the states; gray, transitions in less than 20% of traces). (C,F) Contour plots representing the residence time of RF3-Cy5 (10 nM) on PreHC and PostHC* labeled at L11 by Cy3. FRET time courses were synchronized to the beginning of the first FRET event. FRET values (mean ±sd) 0.62 ± 0.02 (C) and 0.64 ± 0.04 (F) are from three independent data sets and plotted to the right of the contour plots. koff is the rate of RF3 dissociation. See also Figure 3—figure supplement 1 and Supplementary file 1.

https://doi.org/10.7554/eLife.34252.011

To further characterize the conformational changes of the ribosome induced by RF3, we probed the position of the P-site tRNA relative to protein L1. We used a FRET pair with the donor label at the tRNA (fMet-tRNAfMet-Cy3) and the acceptor label on ribosomal protein L1 (L1-Cy5). The two labels are close together and give a high FRET signal when ribosomes are in the L1-closed state and move apart to give a low FRET signal when ribosomes are in the L1-open state (Fei et al., 2009; Fei et al., 2008; Munro et al., 2010a; Munro et al., 2010b; Munro et al., 2010c; Sternberg et al., 2009). FRET histograms of PreHC in the absence of RF3 are dominated by a low-FRET population (0.32 ± 0.01) and do not show transitions to other states (Figure 3—figure supplement 1A, Supplementary file 1). This indicates that ribosomes are predominantly in the L1-open state with fMet-tRNAfMet in the classic P/P state, in agreement with previous studies (Cornish et al., 2009; Fei et al., 2008; Sternberg et al., 2009). RF3 induces dynamic transitions from the low FRET state to a high FRET state (0.74 ± 0.02), which suggests that ribosomes transiently sample the L1-closed state with fMet-tRNAfMet in a hybrid-like P/E state (Figure 3B). This state is short lived (kclosed→open = 6.0 ± 0.8 s−1; Supplementary file 1). The transition rate is faster than subunit rotation (kR→N = 2.2 ± 0.4 s−1) suggesting that the two processes are not tightly coupled, consistent with the previous smFRET work (Munro et al., 2010a; Wasserman et al., 2016) and cryo-EM reconstructions (Fischer et al., 2010).

We then monitored the dissociation of RF3 from PreHC using FRET between RF3–Cy5 and L11–Cy3 (Figure 3C). Labeling of RF3 did not change its catalytic properties (Figure 2—figure supplement 1A,D). The dissociation rate of RF3 from PreHC is koff = 5.9 ± 1.1 s−1 (Figure 3C; Supplementary file 1). Thus, RF3-GTP can bind to PreHC and alter its conformation as shown by the rotation of subunits and movement of the peptidyl-tRNA into a P/E-like state, but the residence time of the factor on the ribosome is short.

To test whether the interaction of RF3 with termination complexes depends on peptide release, we then studied the effect of RF3 on subunit rotation of PostHC prepared by puromycin treatment (PostHC*) (Figure 3D–F). S6/L9-labeled PostHC* fluctuates between 0.5 and 0.7 FRET states (Figure 1—figure supplement 1). RF3 binding shifts the distribution toward the 0.5 FRET state, indicating that the R state is stabilized (Figure 3D). The L1–tRNA FRET pair shows an enrichment of the high FRET state corresponding to the P/E state of the tRNA (Figure 3E). While subunits are stabilized in the R state and do not fluctuate toward N state, the L1-tRNA label shows reversible transitions between P/E and P/P conformations (Supplementary file 1). This suggests that also in PostHC subunit rotation and the formation of a hybrid-like state are not tightly coupled. Dissociation of RF3 from PostHC is as rapid as from PreHC, koff = 5.4 ± 1.3 s−1 (Figure 3F, Supplementary file 1).

Our results show that RF3 facilitates the formation of the R state with the tRNA in a P/E-like orientation on both PreHC and PostHC. RF3 dissociation is not directly coupled to subunit rotation, as the rate of R to N transitions is lower than that of RF3 dissociation (Figure 3, Supplementary file 1). The residence time of RF3 on the ribosome is nearly identical on Pre- and PostHC which indicates that the presence of RF3 on the ribosome is not regulated by peptide release. We also note that the observed RF3 dissociation rates are much higher than the rate of GTP hydrolysis by RF3 (Peske et al., 2014; Shi and Joseph, 2016; Zavialov et al., 2001). This implies that rapid RF3 dissociation is independent of GTP hydrolysis.

Interplay between RF1 and RF3

Next, we studied the interplay between RF1, RF3 and ribosomes during termination. We compared three different termination conditions including PreHC, PostHC* prepared by puromycin treatment, and PreHC which was converted to PostHC in situ upon the interaction with RF1. For each condition, we monitored (i) subunit rotation in the presence of saturating concentrations of both RF1 and RF3; (ii) RF1-Cy5 binding to the ribosome at saturating concentrations of unlabeled RF3; and (iii) RF3-Cy5 binding to the ribosome at saturating concentrations of unlabeled RF1 (Figure 4).

Interplay between RF1 and RF3–GTP.

(A,E,I) Subunit rotation of S6/L9-labeled Pre- and PostHC measured at saturating RF1 and RF3–GTP concentrations (1 µM each). Grey line represents FRET distribution in the absence of RF3. (B,F,J) Contour plots representing the residence time of RF1-Cy5/RF1(GAQ)-Cy5 ribosomes labeled at protein L11 by Cy3 in the presence of excess RF3 (1 µM). Time courses were synchronized to the beginning of the FRET signal. FRET values (mean ±sd) are 0.67 ± 0.02 (B), 0.50 ± 0.03 and 0.76 ± 0.02 (F), and 0.53 ± 0.04 (I). (C,G,K) Contour plots representing the residence time of RF3-Cy3 on ribosomes labeled at protein L11 by Cy3 in the presence of excess RF1 or RF1(GAQ) (1 µM). FRET values (mean ± sd) are 0.51 ± 0.03 (C), 0.51 ± 0.03 (G), and 0.51 ± 0.03 (K). (D,H,L) Comparison of the rates of RF1 and RF3 dissociation and subunit rotation. (A–D) Interactions with PreHC. (E–H) Interactions with PostHC* obtained by puromycin treatment. (I–L) Interactions with PostHC which is formed in situ using RF1. All values are mean ± sd from three independent data sets. See also Supplementary file 1.

https://doi.org/10.7554/eLife.34252.013

To follow the interactions of RF1 and RF3 with PreHC (Figure 4A,B,C), we again used the RF1(GAQ) mutant, which ensures that peptidyl-tRNA in PreHC is not hydrolyzed. While RF1(GAQ) alone stabilizes the N state (grey line in Figure 4A; Figure 1B) and RF3 alone induces transitions from the N to the R state (Figure 3A), in the presence of saturating amounts of RF1(GAQ) and RF3 together the N and R states are almost equally populated (Figure 4A). Ribosomes show rapid reversible N to R transitions indicating that RF3 can promote subunit rotation even when RF1 is present. RF1(GAQ) binding to PreHC–RF3 results in a single FRET population centered at a FRET efficiency of 0.67 ± 0.02, similar to the 0.7 FRET when RF1 binds to the ribosome in the absence of RF3. The RF1(GAQ) dissociation rate is low, <0.3 s−1 (Figure 4B,D; Supplementary file 1), in agreement with previous reports (0.14 ± 0.02 s−1, [Koutmou et al., 2014]). RF3 binds to PreHC–RF1 (Figure 4C), but the FRET efficiency for the RF3-L11 pair is reduced compared to the complex in the absence of RF1 (0.51 ± 0.03 and 0.62 ± 0.02 in the presence and absence of RF1, respectively; Supplementary file 1). Thus, the orientation of RF3 on the PreHC is shifted by RF1, whereas the position of RF1 appears unchanged, at least with respect to L11. The rate of RF3 dissociation in the presence of RF1(GAQ) is 1.3 ± 0.1 s−1, which is higher than the dissociation rate of RF1(GAQ), but about fivefold slower than that of RF3 in the absence of RF1 (Figure 3C; Figure 4D; Supplementary file 1), indicating that RF1 stabilizes the binding of RF3 to PreHC. Dwell time distributions for N or R state in the presence of RF1(GAQ) and RF3 are biphasic, suggesting the presence of two populations of each complex. The majority of ribosomes display rapid transitions (>70%, kN→R = 5.9 ± 0.6 s−1, kR→N = 2.9 ± 0.4 s−1; Figure 4D; Supplementary file 1) that are faster than RF1 or RF3 dissociation, indicating that subunits can rotate while both factors are bound to the ribosome. Low rotation rates (kN→R = 1.30 ± 0.07 s−1, kR→N = 0.80 ± 0.05 s−1; <30% of ribosomes) are also observed with RF3 alone and thus may represent subunit rotation after RF1 dissociation (Supplementary file 1). The observed shift of PreHC–RF1 from the predominantly N to a fluctuating ensemble of N and R states upon RF3 addition, together with the altered RF3 position and the decreased RF3 dissociation rate when the two factors are bound suggest that the complex undergoes conformational adjustments when RF1 and RF3–GTP are bound simultaneously.

Next, we monitored subunit rotation in PostHC. For better comparison with the results obtained with PreHC and RF1(GAQ), we first prepared PostHC* by puromycin treatment of PreHC and studied the interactions with RF1(GAQ) and RF3–GTP (Figure 4E,F,G,H). In the presence of RF1 and RF3, the majority of complexes undergo rapid N to R transitions and the equilibrium is shifted toward the R state (Figure 4E). The mean FRET efficiency for RF1(GAQ) binding to PostHC*–RF3 changes to 0.67 ± 0.02 compared to 0.50 ± 0.03 for RF1(GAQ) binding to PreHC–RF3 (Figure 4B,F) or 0.71 ± 0.01 for binding to complexes in the absence of RF3 (Figure 2A–C). The decrease in FRET efficiency suggests that peptide release allows a rearrangement of the complex which alters the position of RF1 relative to L11. The FRET efficiency for RF3 binding to either PreHC–RF1(GAQ) or PostHC*–RF1(GAQ) is 0.51 ± 0.03 (Figure 4C,G), as compared to 0.62 and 0.64, respectively, for binding to PreHC or PostHC in the absence of RF1 (Figure 3C,F). This suggests that the position of RF3 on PreHC and PostHC is affected by the presence of RF1, but not by peptide release (Figures 3F and 4G). The dissociation rates are 1.3 ± 0.2 s−1 and 1.3 ± 0.1 s−1 for RF1(GAQ) and RF3, respectively (Figure 4F–H; Supplementary file 1). A small fraction (8%) of complexes that release RF1(GAQ) slowly (koff = 0.12 ± 0.07 s−1) is likely due to incomplete peptide hydrolysis by puromycin. The rotation rates (kN→R = 4.20 ± 0.08 s−1, kR→N = 2.50 ± 0.03 s−1) are somewhat higher than RF1 and RF3 dissociation rates, but the most prominent effect of peptide release is the acceleration of RF1 dissociation from <0.3 s−1 to 1.3 ± 0.2 s−1 (Figure 4D,H; Supplementary file 1).

Similar effects are observed when instead of puromycin we used wild-type RF1 to convert PreHC to PostHC (Figure 4I–L): at saturating concentrations of RF1 and RF3 the R state of PostHC is enriched and the complexes show reversible N to R transitions (Figure 4I; Supplementary file 1). RF1 and RF3 are bound to PostHC in the 0.5 FRET state (Figure 4J,K; Supplementary file 1). The dissociation rates are 1.2 ± 0.4 s−1 for RF1 (>70% of ribosomes; Figure 4J; Supplementary file 1) and 1.3 ± 0.2 s−1for RF3 (Figure 4K; Supplementary file 1). Thus, RF1 stabilizes the binding of RF3 on PreHC or PostHC, whereas RF3 destabilizes RF1 binding, but only after peptide release. Peptide release also allows an adjustment in the positions of both factors relative to L11. Thus, peptide release is a major determinant for RF1, but not RF3, dissociation.

Because the kinetics of subunit rotation is faster than RF3 and RF1 dissociation, it remains unclear from which state, N or R, the factors dissociate. To test whether R state formation is required for RF3 dissociation, we used the antimicrobial peptide apidaecin 137 (Api) as a tool to trap RF1 on termination complexes. Api binds into the exit tunnel of PostHC and prevents RF1/RF2 dissociation (Florin et al., 2017). When we monitor subunit rotation in the presence of saturating concentrations of RF1, RF3 and Api, the PostHC–RF1–RF3–Api complex is stalled in the N state (Figure 5A). In the absence of RF1 Api does not alter the relative fraction of N and R states induced by RF3 (Figure 5B). In the PostHC–RF1–Api–RF3 complex, RF1 is stably bound in the 0.7 FRET state (Figure 5C). RF3 is bound in 0.5 FRET state and dissociates with the rate of 1.2 ± 0.1 s−1 (Figure 5D). These data suggest that RF3 can dissociate independent of subunit rotation from termination complexes that are exclusively in the N state as well as from termination complexes that show mixed N and R populations.

Dissociation of RF3 from RF1-bound PostHC in the presence of Api.

(A,B) Subunit rotation of S6/L9-labeled Pre- and PostHC with or without RF1 (1 µM), with saturating RF3 concentrations (1 µM) and Api (1 µM). Red lines in (A) and (B) represent FRET distribution in the absence of Api. (C) Contour plot representing the residence time of RF1-Cy5 on PostHC-Cy3 in the presence of saturating RF3 concentration (1 µM) and Api (1 µM). FRET values (mean ± sd) center at 0.71 ± 0.01. (D) Contour plot representing the residence time of RF3-Cy5 on PostHC-Cy3 in the presence of saturating RF1 concentration (1 µM) and Api (1 µM). FRET values (mean ± sd) center at 0.55 ± 0.04. All values are mean ± sd from three independent data sets. See also Supplementary file 1.

https://doi.org/10.7554/eLife.34252.014

The role of GTP binding and hydrolysis

By analogy with other GTPases, GTP hydrolysis by RF3 is expected to regulate the dissociation of RF3 from the ribosome. In contrast to all other GTPases, RF3 was suggested to bind to the PostHC-RF1 complex in the GDP-bound form; the ribosome-induced rapid release of GDP should stabilize RF3 binding, while subsequent GTP binding induces a conformational change of the ribosome and the release of RF1 (Sternberg et al., 2009; Zavialov et al., 2002). We first tested these models using a biochemical turnover peptidyl-tRNA hydrolysis assay and compared the effect of different nucleotides on factor recycling (Figure 6A,B). When both RF1 and RF3 are sub-stoichiometric to PreHC, such that 10 cycles of RF1 and RF3 turnover are required to convert all PreHC to PostHC, peptide release is only observed in the presence of GTP (Figure 6A). In excess of RF3, when only RF1 has to turnover, efficient peptide release is observed with wild type RF3 in the presence of GTP, GTPγS or GDPNP (Figure 6B). Also RF3(H92A)–GTP, a RF3 mutant deficient in GTP hydrolysis, induces efficient recycling of RF1, contrary to previous reports (Gao et al., 2007), but consistent with a recent kinetic study (Shi and Joseph, 2016). Apo-RF3 has no activity, again consistent with previous reports (Shi and Joseph, 2016; Zavialov et al., 2002). The low activity in the presence of GDP is most likely due to a minor contamination with GTP. Thus, GTP hydrolysis is not required for RF1 recycling but is necessary to ensure recycling of RF3, while the apo and GDPforms of RF3 appear inactive.

Figure 6 with 2 supplements see all
The role of GTP hydrolysis for RF1 and RF3 recycling.

(A,B) Effect of different nucleotides on peptidyl-tRNA hydrolysis (GTP, black circles; GTPγS, green circles; GDPNP, blue circles; GDP, red circles; no nucleotide, grey circles) or in the presence of RF3(H92A) and GTP (purple circles). Control experiments are in the absence of RF3 (blue crosses). Error bars represent the range of two technical replicates. (A) Peptide hydrolysis was performed by incubating PreHC (100 nM) with RF3 (10 nM) and the respective nucleotides (1 mM); reactions were started with the addition of RF1 (10 nM). (B) Same as in (A), but at 1 µM RF3. (C,D) FRET distribution reporting on subunit rotation of S6/L9-labeled PreHC in the presence of saturating amounts of RF3–GDPNP (C) or RF1 with RF3–GDPNP (D) (1 µM RF each). Red lines represent the distribution of FRET states with RF3–GTP. (E,F) Contour plots representing the residence time of RF3-Cy5 (10 nM) on PostHC labeled at L11 by Cy3 in the presence of GDPNP (1 mM) without RF1 (E) or (F) in the presence of saturating RF1 concentration (1 µM). FRET values (mean ± sd) center at 0.71 ± 0.01 and 0.40 ± 0.01 (E) and 0.58 ± 0.03 (F). All values are mean ± sd from three independent data sets. See also Figure 6—figure supplement 1, Figure 6—figure supplement 2 and Supplementary file 1.

https://doi.org/10.7554/eLife.34252.015

Next, we sought to understand how different nucleotides affect the interaction of RF3 with termination complexes. RF3–GTP promotes R state formation, which can be used as readout for the ribosome interaction with RF3 in complex with different nucleotides (Figure 6—figure supplement 1). PreHC in the presence of excess RF3–GDP or RF3 in the apo form are predominantly in the N state and do not show transitions to the R state (Figure 6—figure supplement 1A,B); the ratio of N and R states is identical to that in the PreHC in the absence of RF3 (Figure 1A). Also RF1-bound PostHC in the presence of excess RF3–GDP or apo-RF3 are predominantly in the N state and the distribution of states is very similar to that in RF1-bound termination complexes (Figure 6—figure supplement 1C,D and Figure 1E, respectively). Together, these experiments suggest that RF3-GDP and apo-RF3 are not able to induce the R state in termination complexes. By analogy, smFRET experiments monitoring the position of the L1 stalk show that addition of RF3-GDP or apo-RF3 does not change the ribosome conformation (Sternberg et al., 2009). For a more direct observation of RF3-GDP or apo-RF3 binding to the ribosome, we made an attempt to follow FRET between RF3-Cy5 and termination complexes labeled at protein L11 with Cy3. However, we did not find any FRET events indicative of RF3 binding in the presence of GDP or with apo-RF3 (data not shown). These observations suggest that although RF3-GDP or apo-RF3 must bind to PostHC–RF1 in some way, because this interaction accelerates nucleotide exchange in RF3 (Koutmou et al., 2014; Peske et al., 2014; Shi and Joseph, 2016; Zavialov et al., 2001) the interaction must be transient and does not engage the factor at its binding site at L11 unless GTP is bound.

We then asked whether GTP hydrolysis by RF3 is required to induce subunit rotation. We replaced GTP with a non-hydrolysable analog, GDPNP, which is extensively used in structural studies. RF3–GDPNP can bind to the PreHC or PostHC obtained by addition of RF1 and induces formation of the R state, albeit not to the same extent as RF3–GTP and with fewer transitions between N and R states (Figure 6C,D). The same tendencies are observed with RF3(H92A)–GTP or RF3–GTPγS (Figure 6—figure supplement 1E,F,G). The exact fraction of the R state and dynamic ribosomes depends on the choice of nucleotide, which may indicate that the ability of RF3–GDPNP or RF3–GTPγS to form a stable complex with the ribosome is reduced compared to RF3–GTP.

We then tested whether GTP hydrolysis is required for RF3 dissociation from the ribosome. The dissociation rate of RF3–GDPNP from PostHC* in the absence of RF1 is koff = 0.34 ± 0.04 s−1, much lower than with GTP (Figure 6E and Supplementary file 1). In contrast, in the presence of saturating RF1 concentrations dissociation of RF3–GDPNP from PostHC-RF1 is as rapid as with GTP (Figure 6F and Supplementary file 1), indicating that GTP hydrolysis is not essential when RF1 is present. Experiments with RF3–GTPγS gave very similar results (Figure 6—figure supplement 2). At saturating RF3 concentrations, dissociation of RF1 from PostHC is independent of GTP hydrolysis (Figure 6—figure supplement 2), consistent with the biochemical data (Figure 6B). In the simplest model, these findings can be interpreted as an indication for the role of GTP hydrolysis in RF3 dissociation from termination complexes in the absence of RF1. They also explain why RF1 turnover is impaired at sub-stoichiometric RF3 concentrations when GTP hydrolysis is blocked (Figure 6A): those RF3 molecules that bind to ribosomes lacking RF1 remain stalled if GTP is not hydrolyzed, thereby depleting the pool of RF3 which has to turnover to stimulate RF1 dissociation. Thus, the only reaction where GTP hydrolysis or an authentic GTP conformation appears to play an essential role is the dissociation of RF3 from termination complexes in the absence of RF1.

Discussion

Our experiments show how release factors navigate through the landscape of possible ribosome conformations during translation termination (Figure 7A). Release factors not only change the ratio between the N and R states, but also alter the fraction of the ribosomes that make transient fluctuations between the states. Each factor alone has its distinct signature on ribosome conformation and dynamics. Binding of RF1 to either PreHC or PostHC favors the static N state; protein L1 adopts an open conformation, which correlates with a classical state of the P-site tRNA. The N state of the ribosome–RF1 complex has been also captured by structural studies (James et al., 2016; Korostelev et al., 2008; Laurberg et al., 2008; Petry et al., 2005; Weixlbaumer et al., 2008). Surprisingly, we find that PreHC–RF2 is more dynamic, and has a higher fraction of the R states than the complex with RF1. Furthermore, RF2 can dissociate equally well from the PreHC and PostHC and is less dependent on the action of RF3 than RF1 (Figure 2—figure supplement 3A, Figure 7B). With its high dissociation rate, RF2 action may depend on the ratio between the rate of peptide release and factor dissociation, for example, if the rate of peptidyl-tRNA hydrolysis is about 10 s−1 (Indrisiunaite et al., 2015; Kuhlenkoetter et al., 2011) and the rate of RF2 dissociation is ~1 s−1 (this paper), the factor can achieve efficient peptide release before dissociating. Thus, RF1 and RF2 – albeit fulfilling a similar function during canonical termination – differ in their ability to affect ribosome dynamics.

The mechanism of translation termination.

(A) Ribosome dynamics in the presence of RF1, RF2, and RF3. Ribosome fluctuations are color-coded from static (gray) to dynamic (red) and correlated to the fraction of N and R state in the respective complex. (B) Summary of the dissociation rate constants of RF1, RF2 and RF3 from different type of complexes. Bars representing the dissociation of RF1 are colored teal, RF2 gray, RF3 magenta. (C) The landscape of ribosome conformations with RF1 and RF3. The ribosome states (N and R, PreHC and PostHC) are indicated. Red arrows indicate rapid reaction, blue arrows static or slowly exchanging states, with the preferential direction indicated by color gradient; single-headed arrows indicate irreversible steps of peptidyl-tRNA hydrolysis. See also Supplementary file 1.

https://doi.org/10.7554/eLife.34252.018

Binding of RF3–GTP to termination complexes shifts the conformational distribution toward the R state (Figure 7A). The PreHC–RF3 complex is dynamic, whereas the PostHC–RF3 is stabilized in the R state, consistent with the previous smFRET work (Sternberg et al., 2009) and structural studies (Gao et al., 2007; Jin et al., 2011; Zhou et al., 2012). After peptide release, RF1 and RF3 or RF2 and RF3 together shift the distribution of ribosome conformations towards the middle of the dynamic spectrum (Figure 7A). The rates of ribosome fluctuations are in the range of 2–6 s−1, somewhat faster than in the absence of the factors, 0.5–2.6 s−1 (Supplementary file 1).

One open question is what drives the dissociation of RF1 and RF3 from the ribosome (Figure 7B). Dissociation of RF1 from the static N state is very slow. RF3 accelerates the dissociation, which correlates with increased ribosome dynamics and frequent transitions from N to R state. However, dynamic transitions alone are not sufficient to induce RF1 dissociation from the ribosome, because peptide release is crucial to allow RF1 to dissociate rapidly. Peptide release leads to a change in the orientation of RF1 with respect to L11. On the other hand, peptide release alone is not sufficient, as the dissociation rate of RF1 from the PostHC is slow in the absence of RF3. Thus, RF1 dissociation is promoted by the concerted action of RF3, which stimulates subunit rotation and may directly displace RF1 from its original binding site, and by peptide release, which allows a conformational adjustment of RF1.

RF3 dissociation is independent of peptide release or the ribosome dynamics, but is affected by the presence of RF1 or RF2, which stabilize RF3 binding to the ribosome and change conformation/position of RF3 relative to L11. In the presence of RF1, RF3 efficiently dissociates from the N state even in the absence of GTP hydrolysis (this paper and [Shi and Joseph, 2016]). The order of RF1 and RF3 dissociation appears random, because the rates of factor release are quite similar and the exact sequence depends on experimental conditions (this paper; [Koutmou et al., 2014; Shi and Joseph, 2016]). In those cases where RF1 happens to dissociate before RF3 has left the ribosome, GTP hydrolysis completes RF3 recycling. In summary, subunit rotation, peptide release, conformational changes of the factors, and GTP hydrolysis together drive dissociation of RF1 and RF3. However, kinetically these movements are not directly coupled, that is the dissociation rates of the factors and the rates of subunit rotation are independent of each other but are individually defined by the dynamic properties of the complex. Thus, translation termination is a stochastic process that utilizes loosely coupled motions of its players to complete protein synthesis and release the newly synthesized nascent chain toward its cellular destination.

Our results lead to the following model of translation termination for RF1 (Figure 7C). Among all possible reaction routes, two appear most likely, either via RF1 binding to PreHC, followed by peptide release and RF3–GTP recruitment, or through simultaneous binding of RF1 and RF3–GTP to PreHC followed by peptide release. The resulting complex PostHC–RF1–RF3–GTP can make rapid transitions between the N and R states. RF1 and RF3 change their relative positions and can now both rapidly dissociate from the ribosome. The order of events is not deterministic: multiple ribosome conformations, ribosome dynamics and the lack of strong coupling between the rates of subunit rotation and the dissociation of RF1 and RF3 seem characteristic features of RF1-dependent termination.

This work provides an unexpected view on the role of nucleotide exchange and GTP hydrolysis by RF3. Although RF3-GDP or apo-RF3 can bind to the ribosome carrying RF1/RF2 (Peske et al., 2014; Zavialov et al., 2001), this interaction does not result in the recruitment of the factor to its binding site at the vicinity of L11. In vitro in the absence of GTP, apo-RF3 can form a relatively stable complex with PostHC–RF1 (Pallesen et al., 2013; Shi and Joseph, 2016), but this binding does not alter the dynamics of subunit rotation and does not accelerate RF1 dissociation (this paper and [Sternberg et al., 2009]). Rather, the GTP-bound form of RF3 is required to stimulate ribosome dynamics and RF1 dissociation from PostHC. Given the moderate difference in the affinities of RF3 for GTP and GDP, at cellular concentrations a large fraction of RF3 is in the GTP form. Furthermore, given the high GTP association rate, apo-RF3 will be immediately converted into the functionally active GTP form (Peske et al., 2014); thus, the apo-RF3–ribosome complex can only be a transient intermediate. The present experiments, most of which are performed in the presence of a GTP regeneration system, which does not allow for accumulation of the GDP- or apo-form of RF3, show efficient factor binding, peptide release and factor recycling. We thus have no indication for an active role of nucleotide exchange or for an essential role of the GDP- or the apo-form of RF3 in termination at cellular conditions and we consider the respective models unlikely.

Unexpectedly, our data suggest that GTP hydrolysis or an authentic GTP-bound form of RF3 are required to release RF3 that is arrested on the ribosome in the absence of RF1. At the first glance, the low dissociation rate of RF3–GDPNP from the ribosome appears to contradict the results of the experiments with RF3–GTP, which show that factor dissociation is not coupled to GTP hydrolysis (Figure 3C,F). We hypothesize that upon binding to the ribosome, RF3 can either form an initial binding complex from which the factor can dissociate rapidly, or enter an engaged complex, from which RF3 can only dissociate after GTP hydrolysis (Figure 6E). In principle, this should result in biphasic dissociation time courses of RF3-GTP with a second slow phase corresponding to the rate of GTP hydrolysis, which we did not observe. However, in the presence of GTP the fraction of RF3 molecules that enter the engaged state may be too small to capture. As RF3–GDPNP appears to have a lower affinity to the ribosome than RF3–GTP, the transient initial RF3–ribosome complex might be too short-lived to be detected and only the stable engaged complexes are captured. Alternatively, GDPNP, as well as GTPγS or the RF3(H92A) mutant may induce a conformation that hinders RF3 from dissociation but is hardly populated in the presence of GTP; in this case, the effects are purely conformational and not due to GTP hydrolysis as such.

Available structures of ribosome-bound RF3 suggest that RF3 is arrested on ribosomes in the R state (Gao et al., 2007; Jin et al., 2011; Zhou et al., 2012). This could explain why PostHC, with its higher propensity to be in the R state than the PreHC, is more efficient in stimulating GTP hydrolysis by RF3 (Zavialov et al., 2002). In this respect, RF3 appears to be an unusual GTPase that differs from other translational GTPases, such as EF-G, EF-Tu and IF2, where GTP hydrolysis is coupled to key steps on the reaction pathway of the factors and is required on all ribosome complexes. Rather, the internal clock of the RF3 GTPase (Peske et al., 2014) acts as a rescue mechanism to release RF3 recruited to complexes that do not contain RF1. This scenario is realistic at the concentrations of factors in the cell where RF3 is much more abundant than RF1 (Schmidt et al., 2016).

The smFRET data presented here for a simple model system present a starting point to study dynamics of more natural termination complexes containing long peptide nascent chains. While model termination systems are fully functional in all steps of termination and the rate of GTP hydrolysis by RF3 is similar with the fM-Stop and fMFTI-Stop termination contexts (Zavialov and Ehrenberg, 2003), the length of the nascent peptide and the nature of the P-site tRNA may attenuate the ribosome dynamics. While currently such complexes are biochemically too heterogeneous to study, further development of smFRET techniques toward multicolor detection and better time resolution may provide a tool to decipher the dynamics of these heterogeneous assemblies.

Materials and methods

Key resources table
Reagent type (species)
or resource
DesignationSource or referenceIdentifiersAdditional information
Strain, strain
background (E. coli)
JW3947-1Keio collectionCGSC#: 12041E. coli rplA knockout strain
Sequence-based
reagent
Start-stop mRNAIBA (Göttingen)N/ARNA oligonucleotide:
5’-GGCAAGGAGGUAAAUAAU
GUAAACGAUU-3’
Sequence-based
reagent
mMetStopIBA (Göttingen)N/ARNA oligonucleotide:
5′-Biotin-CAACCUAAAACUUACACA
CCCGGCAAGGAGGUAAAUAAU
GUAAACGAUU-3′
Sequence-based
reagent
mMetPheStopIBA (Göttingen)N/ARNA oligonucleotide:
5‘-Biotin-CAACCUAAAACUUACACACCC
GGCAAGGAGGUAAAUAAUGUUU
UAAACGAUU-3 ‘
Sequence-based
reagent
mMetLysStopIBA (Göttingen)N/ARNA oligonucleotide:
5‘-Biotin-CAACCUAAAACUU
ACACACCCGGCAAGGAGGUA
AAUAAUGAAGUAAACGAUU-3 ‘
Sequence-based
reagent
mMetValStopIBA (Göttingen)N/ARNA oligonucleotide:
5‘-Biotin-CAACCUAAAACUUAC
ACACCCGGCAAGGAGGUAAAU
AAUGGUUUAAACGAUU-3 ‘
Peptide,
recombinant protein
RF2(GAQ)
(recombinant protein)
PMID: 12419223
Peptide,
recombinant protein
RF1(GAQ)
(recombinant protein)
PMID: 12419223
Peptide,
recombinant protein
RF1(S167C) (recombinant protein)PMID: 19597483Single-cysteine RF1
Peptide,
recombinant protein
RF2(C273)
(recombinant protein)
This paperSingle-cysteine RF2
Peptide,
recombinant protein
RF3(L233C) (recombinant protein)This paperSingle-cysteine RF3
Peptide,
recombinant protein
L1(T202C)
(recombinant protein)
PMID: 18471980Single-cysteine L1
Peptide,
recombinant protein
Apidaecin137 (API) (peptide)NovoPro
Biosciences Inc.
N/A
Chemical
compound, drug
Cy3-maleimideGE HealthcarePA23031
Chemical
compound, drug
Cy5-maleimideGE HealthcarePA25031
Software,
algorithm
MatlabMathWorks
Software,
algorithm
Prism GraphPadGraphPad Software,
La Jolla California
USA, www.graphpad.com
Software,
algorithm
Matlab code vbFREThttp://vbfret.sourceforge.net/Described in
Bronson et al. (2009)

Buffers

All smFRET experiments were performed in imaging buffer (50 mM Tris-HCl pH 7.5, 70 mM NH4Cl, 30 mM KCl, 15 mM MgCl2, 1 mM spermidine, 8 mM putrescine, 2.5 mM protocatechuic acid, 50 nM protocatechuate-3,4-dioxygenase (from Pseudomonas), 1 mM Trolox (6-hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid), and 1 mM methylviologen). Peptide hydrolysis experiments were performed in TAKM7 buffer (50 mM Tris-HCl pH 7.5, 70 mM NH4Cl, 30 mM KCl, 7 mM MgCl2).

Labeled ribosomes, release factors and tRNA

Request a detailed protocol

The preparation and functional characterization of ribosomes labeled with Cy3 at protein L11 and double-labeled at S6-Cy5 and L9-Cy3 was carried out as described (Adio et al., 2015; Sharma et al., 2016). E. coli strain lacking L1 were obtained from the Keio collection (CGSC#: 12041) and ΔL1 ribosomes purified according to the protocol used for native ribosomes (Rodnina and Wintermeyer, 1995). A single cysteine was introduced at position T202 of L1 and the protein purified as described in Fei et al. (2008). L1(T202C) was fluorescence labeled with Cy5-maleimide (GE Healthcare) and purified using a 5 ml HiTrap SP HP cation exchange chromatography column (GE Healthcare). ΔL1 ribosomes were reconstituted by incubation with a 5-fold molar excess of L1-Cy5 for 30 min at 37°C. Excess protein was removed by centrifugation through a 30% sucrose cushion in 50 mM Tris-HCl pH 7.5, 70 mM NH4Cl, 30 mM KCl, 15 mM MgCl2, pH 7.5.

The RF2 construct was cloned from the E. coli K12 strain and contains the natural T246A replacement (Wilson et al., 2000). Catalytically impaired RF1(G234A) (RF1(GAQ)) and RF2(G251A) (RF2(GAQ)), and the respective single-cysteine variant RF1(S167C) (Sternberg et al., 2009; Wilson et al., 2000), RF2(C273) and RF3(L233C) were generated by Quickchange mutagenesis according to the standard protocol. Native cysteines were replaced by serine or alanine based on the sequence conservation analysis performed using the Consurf database. RF1 and RF2 were purified and in vitro methylated as described (Kuhlenkoetter et al., 2011). RF3 was purified by affinity chromatography on a Ni-IDA column (Macherey-Nagel) followed by ion exchange chromatography on a HiTrapQ column (Peske et al., 2014). Prior to labeling, methylated RF1 and RF2 were incubated for 30 min with a 10-fold molar excess of TCEP (Sigma) at room temperature (RT). Cy5 maleimide (GE Healthcare) was dissolved in DMSO and added to the proteins (5- to 10-fold molar excess). Labeling was performed for 2 hr at RT and quenched by addition of a 10-fold molar excess of 2-mercaptoethanol over dye. Excess dye was removed by gel filtration on a PD-10 column (GE Healthcare). tRNAfMet was labeled at position s4U8 with Cy3-maleimide (Fei et al., 2010) and aminoacylated and purified as described (Milon et al., 2007).

mRNA

Request a detailed protocol

All mRNAs used in the smFRET experiments are labeled with biotin at the 5´end and were purchased from IBA (Göttingen, Germany). The following sequences were used: mMetStop

5′-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGUAAACGAUU-3′ mMetPheStop

5‘-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGUUUUAAACGAUU-3‘ mMetLysStop

5‘-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGAAGUAAACGAUU-3‘ mMetValStop

5‘-Biotin-CAACCUAAAACUUACACACCCGGCAAGGAGGUAAAUAAUGGUUUAAACGAUU-3 ‘

For the peptide hydrolysis experiments, ribosome complexes were assembled on the synthetic model mRNA, 5’-GGCAAGGAGGUAAAUAAUGUAAACGAUU-3’ (IBA) with a start codon followed by a stop codon.

Sample preparation for smFRET TIRF experiments

Request a detailed protocol

Initiation complex formation was carried out by incubating ribosomes (100 nM) with a three-fold excess of IF1, 2 and 3, fMet-tRNAfMet, mRNA and 1 mM GTP in TAKM7 for 30 min at 37°C. To form initiation complexes with fMet-tRNAfMet-Cy3, equal amounts of ribosomes and tRNA were used. In case of the mRNA coding for fMetStop, the initiation complex was used as PreHC. To generate PreHC on other mRNAs, an equal volume of ternary complex was added containing EF-Tu (1 µM) incubated with GTP (1 mM), phosphoenolpyruvate (3 mM) and pyruvate kinase (0.1 mg/ml) in TAKM7 for 15 min at 37°C, followed by addition of Phe-tRNAPhe, Lys-tRNALys or Val-tRNAVal (500 nM). Addition of EF-G (100 nM) and GTP (1 mM) induced tRNA translocation to form PreHC that contains peptidyl tRNA in the P site and displays the UAA stop codon in the A site.

TIRF experiments

Request a detailed protocol

Complexes were diluted to 1 nM with smFRET buffer (50 mM Tris-HCl, 70 mM NH4Cl, 30 mM KCl, 15 mM MgCl2, 1 mM spermidine and 8 mM putrescine). Biotin/PEG functionalized cover slips were incubated for 5 min at room temperature with the same buffer containing additionally BSA (10 mg/ml) and neutravidin (1 µM) (Thermo Scientific). Excess neutravidin was removed by washing the cover slip with buffer containing BSA (1 mg/ml). Ribosome complexes were applied to the surface and immobilized through the mRNA-biotin:neutravidin interaction. Images were recorded at a rate of 30 frames/s after exchanging the buffer with imaging buffer at room temperature (22°C) (Adio et al., 2015).

To monitor subunit rotation of L9/S6-labeled ribosomes in the presence of release factors at steady-state conditions, imaging buffer was supplemented with RF1, RF2 and/or RF3 (1 µM each). In experiments with RF1(GAQ) or RF2(GAQ), the observation time was limited to <10 min in order to minimize peptide hydrolysis due to residual factor activity. In experiments monitoring subunit rotation by RF3 in the GTP form or in complex with non-hydrolysable GTP analogs, imaging buffer was additionally supplemented with the energy recycling system (1 mM GTP or 1 mM GDPNP or 1 mM GTPγS, 3 mM phosphoenolpyruvate and 0.1 mg/ml pyruvate kinase). FRET signals reporting on the time course of subunit rotation during termination were obtained by injecting RF1 or RF2 (100 nM) in imaging buffer to immobilized PreHC or PostHC.

To measure FRET signals reporting on the residence time of labeled release factors on PreHC or PostHC labeled at protein L11 with Cy3, the complexes were immobilized on the cover slip. Movies were recorded upon addition of Cy5-labeled RF1, RF2 or RF3 to a final concentration of 10 nM in imaging buffer. To study the residence time of Cy5-labeled RF3 or to study the residence time of Cy5-labeled RF1 or RF1(GAQ) on ribosomes in the presence of unlabeled RF3, imaging buffer was supplemented with unlabeled RF3 (1 µM), GTP (1 mM), phosphoenolpyruvate (3 mM) and pyruvate kinase (0.1 mg/ml). To study the residence time of Cy5-labeled RF3 in the presence of RF1, imaging buffer was supplemented with unlabeled RF1 (1 µM), GTP (1 mM), phosphoenolpyruvate (3 mM) and pyruvate kinase (0.1 mg/ml).

To monitor FRET signals reporting on the conformation of the P-site tRNA PreHC or PostHC labeled on protein L1(C202-Cy5) and on fMet-tRNAfMet(thioU8-Cy3) or tRNAfMet(U8-Cy3) were immobilized on the coverslip. Movies were recorded upon addition of imaging buffer or imaging buffer containing RF3 (1 µM). In experiments with RF3 imaging, buffer was additionally supplemented with the energy recycling system (1 mM GTP or 1 mM GDPNP or 1 mM GTPγS, 3 mM phosphoenolpyruvate and 0.1 mg/ml pyruvate kinase) (Sternberg et al., 2009).

Data analysis

Request a detailed protocol

Fluorescence time courses for donor (Cy3) and acceptor (Cy5) were extracted as described (Adio et al., 2015; Fei et al., 2008; Roy et al., 2008). A semi-automated algorithm (Matlab) was used to select anti-correlated fluorescence traces (correlation coefficient <0.1) exhibiting characteristic single fluorophore fluorescence intensities (Adio et al., 2015). Time traces for further analysis were selected from the dataset by choosing only those traces that contained single photobleaching steps for Cy3 and Cy5 (as recommended in [Fei et al., 2008]). The bleed-through of the Cy3 signal into the Cy5 channel was corrected using an experimentally determined coefficient (~0.13 in our experimental system [Adio et al., 2015]). All trajectories were smoothed over three data points. FRET efficiency was defined as the ratio of the measured emission intensities, Cy5/(Cy3 +Cy5) (Roy et al., 2008). FRET-histograms were fitted to Gaussian distributions using Matlab code (Adio et al., 2015). Mean FRET values (mean ±sd) and population distribution (p=area under the curve ± sd) were calculated from three independent datasets and are summarized in Supplementary file 1.

The vbFRET software package (http://vbfret.sourceforge.net/) (Bronson et al., 2009) was used for hidden Markov model (HMM) analysis of the FRET data. Time trajectories with only one transition per trace and with the FRET changes of less than 0.1 were excluded from further kinetic analysis (Fei et al., 2008; Sternberg et al., 2009). Individual time-resolved FRET traces were compiled into FRET probability density plots (contour plots) (Blanchard et al., 2004; Munro et al., 2007). For the experiments measuring subunit rotation of PostHC upon binding of RF1 in real time, FRET traces were synchronized at the transition to the stable N state. For the experiments measuring subunit rotation of PreHC upon binding of RF2 in real time, FRET traces were synchronized to the first N to R transition. In experiments measuring the residence time of labeled release factors, FRET traces are synchronized to the beginning of the FRET event reporting on the binding of the factor to the ribosome. One-dimensional histograms at the right side of the contour plots summarize FRET values of the first 10–30 time frames (0.3–1.0 s) of the FRET signals. The photobleaching rates of the S6/L9-FRET pair were estimated as described (Adio et al., 2015) from the non-fluctuating 0.7 FRET trajectories obtained with PreHC, PreHC-RF1(GAQ) and PostHC-RF1, as well as from the non-fluctuating 0.5 FRET trajectories of PostHC*-RF3(GTP); the photobleaching rates were in the range of 0.07–0.19 s−1, comparable to 0.05–0.3 s−1 in (Sternberg et al., 2009). Dwell times of individual FRET states in traces with multiple FRET states were calculated from idealized traces (Bronson et al., 2009). Dwell time histograms were fitted to either one- or two-exponential function. Rates (k) were calculated by taking the inverse of dwell times. Rate constants ± standard deviations were determined from three independent datasets as described in Fei et al. (2011); Sternberg et al. (2009); Wasserman et al. (2016) and summarized in Supplementary file 1.

Peptide hydrolysis assay

Request a detailed protocol

PreHC was prepared as described (Peske et al., 2014) and purified through sucrose cushion centrifugation. After centrifugation, ribosome pellets were resuspended in TAKM7, frozen in liquid nitrogen and stored at −80°C. The extent of initiation was better than 95% as determined by nitrocellulose filtration and radioactive counting. PreHC (100 nM) was incubated with RF3 at the indicated concentration and nucleotide (1 mM) for 15 min at 37°C. Pyruvate kinase (0.1 mg/ml) and phosphoenol pyruvate (3 mM) were added in all experiments performed in the presence of GTP. Time courses were started by addition of RF1 or RF2 (10 nM). Samples were quenched with a solution containing TCA (10%) and ethanol (50%). After centrifugation (30 min, 16,000 g), the amount of released f[3H]Met in the supernatant was quantified by radioactive counting.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
    Mutants of Escherichia coli lacking ribosomal protein L11
    1. G Stöffler
    2. E Cundliffe
    3. M Stöffler-Meilicke
    4. ER Dabbs
    (1980)
    The Journal of Biological Chemistry 255:10517–10522.
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66

Decision letter

  1. Rachel Green
    Reviewing Editor; Johns Hopkins School of Medicine, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Dynamics of ribosomes and release factors during translation termination in E. coli" for consideration by eLife. Your article has been evaluated by a Senior Editor and three reviewers, one of whom is a member of our Board of Reviewing Editors.

We have received comments from three experts in the field and these reviewers have discussed their independent views to reach a decision. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

While all three reviewers found your data on the function of bacterial termination factors RFs1-3 interesting, and all appreciated the major effort to develop the story, the consensus view was that: (1) there were substantive experimental issues that still need to be addressed and (2) even if these concerns are addressed, the manuscript will lack a cohesive narrative that would be of broad interest to the readers of eLife.

The reviews are included but the critical points that the reviewers focused on were:

a) All three reviewers were surprised to see arguments suggesting that the ribosome can assume a rotated conformation independent of tRNAs moving into a hybrid state of binding (though this surprising observation was never very clearly articulated and rationalized to the reader). One possibility would be to form complexes with more authentic peptidyl-tRNAs (rather than fMet-tRNA) since it is very difficult to imagine that these peptidyl-tRNAs could migrate to the E site. Alternatively, the rotated state being observed with an intermediate FRET value (~0.5) may not be the true rotated state but instead a different intermediate or binding state (controls need to be performed to confirm the authenticity of this FRET state as rotated).

b) The differences between RF2 and RF1 behavior were very interesting but given the reduced affinity of RF2 relative to RF1, there are a number of experimental concerns. Is RF2 fully methylated? Is the site of modification (a different site than used for RF1) problematic (data not provided)? Are complexes fully saturated or would incomplete binding explain some of the heterogeneity (i.e. it is compositional)? This latter possibility seems critical to establish since RF2 does seem to bind more weakly. The biphasic nature of many of the dwell time histograms are suggestive of such compositional heterogeneity.

c) Finally, the data are not presented with error bars or statistical analyses of significance.

Reviewer #1:

This manuscript by Adio et al. uses smFRET, ensemble FRET, and ensemble kinetic assays to probe conformational changes on the ribosome associated with peptide release and release factor binding (RF1, RF2, and RF3). Several new findings are described that refute the older model of RF3 function, particularly the effect of nucleotide binding and hydrolysis, though that particular point has been addressed by the Rodnina lab previously. The data for RF2 appear to contradict previous structural work though the experiments here were performed on tRNAs lacking a true peptide in the exit tunnel. Finally, the fact that ribosomal subunit rotation and factor binding and dissociation are not kinetically coupled in a straightforward manner makes it difficult to propose a clear mechanistic model.

Conclusions from this work:

1) While RF1 stabilizes the non-rotated state, RF2 enriches the rotated state of the ribosome and dissociates quickly (a new and controversial finding). RF2 dissociates quickly from either pre or post termination complexes even without RF3.

2) RF3-GTP stabilizes the rotated state (previously known) and dissociates quickly from pre or post termination complexes.

3) Binding of RF3 to ribosomes bound with RF1 changes the conformation of both factors and stabilizes RF3 binding. RF1 and RF3 dissociation appears random; if RF3 dissociates first (from the non-rotated state), it does not require GTP hydrolysis, but if RF1 dissociates first, RF3 requires GTP hydrolysis to leave from the rotated state. (This finding is new and important).

4) Rotation, peptide release, GTP hydrolysis are not kinetically coupled in a straightforward manner.

Concerns:

In the first paragraph of the subsection “RF1 and RF2 have distinct effects on ribosome dynamics”, the authors state that the tRNA in the PreHC does not affect FRET dynamics (and for these experiments they use fMet-tRNA). However, the figures seem to depict a dipeptidyl-tRNA. This should be clarified in the text, figures, and figure captions. More importantly, it raises the question as to why they observe the rotated (R) state in the PreHC upon either RF2 or RF3 binding. Is this allowed rotation specific to the use of this minimal substrate fMet-tRNA? What would happen with a longer peptidyl-tRNA that extends into the exit tunnel? This is an important question because this finding suggests (contrary to the literature) that the hybrid tRNA states and rotated state of the ribosome can be separated.

In Figure 3 and Figure 6G, release of RF3 is much faster than GTP hydrolysis, but one of the main conclusions is that RF3 requires GTP hydrolysis to be released from the rotated state (Figure 6D, H).

Reviewer #2:

This paper is an attempt to use single-molecule FRET to dissect the events during termination of bacterial translation. Termination in bacteria involves class I release factors RF1 and RF2 that recognize the stop codon in the A site of the ribosome and trigger peptide release. A second GTPase class II release factor RF3 binds to the ribosome and accelerates dissociation of release factors from the ribosome, setting the stage for recycling (the final stage of translation).

Although the actions of RF1 and RF2 have become clearer over the years, the action of RF3 remains less so. In particular, there have been competing models. A model proposed by Ehrenberg's lab suggested that the ribosome functions as a GTP exchange factor for RF3 that is closely related to its mechanism.

This paper expands on recent work by both the Rodnina and Green labs to show a number of things. Firstly, it suggests that RF1 and RF2 do not function in precisely the same way. RF2 is less stably bound and does not require the action of RF3 to dissociate. Moreover, while RF1 strongly favors the non-rotated state without RF3, the same is not true for RF2. Building on previous papers in the Rodnina/Wintermeyer and Green labs, the paper provides strong evidence that RF3 in fact binds the ribosome in the GTP form, and that GTP hydrolysis is not a prerequisite for RF1 or RF2 release. Moreover, the apo and GDP forms are not capable of catalyzing release of RF1.

Finally, the paper shows that the temporal order of dissociation of RF1 and RF3 is random. Rather the complex favors a form that facilitates dissociation of both.

This is a nice study that is a clear advance on previous papers and helps clarify some of the confusion in this area. The difference between RF1 and RF2 is surprising.

The implication of their control (with just tRNAs) is that the rotated state with RF1,2/3 is similar that reached with tRNAs during translocation. However, a P/E tRNA state is not possible for the preHC complex, because it requires a deacylated tRNA. So what is the R state that RF2 reaches in the preHC? This is never actually addressed.

A larger question they might have asked is why not all bacteria have RF3, and whether this is related to a possible role of RF3 in quality control shown by the Green lab.

In summary, the paper deals with details of the mechanism of translational termination in bacteria that although important, will be of interest to only a handful of people even in the ribosome field. My own feeling is that it is not clear that it belongs in a general interest journal like eLife.

Reviewer #3:

The manuscript by Adio et al. describes an smFRET study of translation termination. Specifically, Adio et al. attempt to investigate whether RF2-meidated termination follows the same mechanism as RF1-mediated termination, the dynamics of the RF1-RF3-bound and RF2-RF3-bound termination complexes that are intermediates in the termination pathway, and the role of guanine nucleotide in RF3 function. The first two of these are open questions in the field and the third remains controversial, with two different models represented in the current literature. Given this, the answers to these questions would undoubtedly be of importance to the field and, in principle, the manuscript by Adio, et al. would be perfectly appropriate for publication in eLife.

However, as described in greater detail below, important controls are missing and, in several key instances, it is not clear that the data support the conclusions. In some ways, it almost seems as if the work presented here is a work in progress that is not yet finished and ready for publication. Thus, the authors would need to address these concerns before the appropriateness of this work for publication in eLife could be properly assessed.

1) The authors find that, unlike RF1, RF2 has a relatively low affinity for pre- and post-hydrolysis termination complexes and binds only transiently to these complexes. As the authors point out, this is a very surprising result. As such, it raises many important concerns that could be easily addressed by controls:

1a) The authors do not demonstrate whether their RF2 (or RF1) constructs are methylated at the Q of the GGQ motif in domain 3. Given that the Ehrenberg group has shown that the affinity of RF2 for termination complexes and the catalytic activity of RF2 on termination complexes are both dependent on this post-translational modification (Pavlov, et al. (1998) J Molec Biol and Dincbas-Renqvist, et al. (2000) EMBO J), it is important that the authors demonstrate that the surprising results they have obtained with RF2 are not due to the lack of this post-translational modification.

1b) The authors claim that their fluorophore-labeled RF2 (and RF1) construct are as active as their unlabeled counterparts, but the data are not shown. Given the results that the authors have obtained with their fluorophore-labeled RF2 construct, it seems to me that controls demonstrating that both the affinity and the catalytic activity of the authors' fluorophore-labeled RF2 construct are unchanged relative to unlabeled RF2 must be shown. It is also important to specify whether this comparison is being made to the unlabeled, single-cysteine mutant RF2 construct or to the unlabeled, fully wildtype RF2 construct. All of this is made more important by the fact that the authors have mutated and fluorophore-labeled a position on RF2, A237, that is very different from the position that they have mutated and labeled on RF1, has not been previously characterized, and is located in domain 3, where it could easily affect the affinity and/or catalytic activity of RF2 in a manner similar to that which is observed for the methylation of the Q in the GGQ motif in domain 3.

2) In several instances, the authors seem to interpret their data under the assumption that the termination complexes are saturated with unlabeled components, without having convincingly argued or demonstrated that the complexes are saturated:

2a) In the experiments in which unlabeled RF3 is added to termination complexes that are interacting with fluorophore-labeled RF2 shown in Figure 2—figure supplement Figure 2A, C. What is the affinity of the unlabeled RF3 for these complexes? Is the concentration of unlabeled RF3 that the authors use for these experiments high enough such that the complexes are saturated? How dependent are the interpretation of these data and the conclusions that are drawn on the complexes being saturated with unlabeled RF3? It seems like the extremely low, 10 nM concentrations of fluoropore-labeled RF2 and the possibility that, at any one time, the termination complexes are only partially occupied with RF3 would generate compositional heterogeneity that would make the data hard to interpret. The authors should address these questions through controls (e.g., titrations of fluorophore-labeled RF2 and/or unlabeled RF3) and/or revisions to the manuscript.

2b) Similar considerations apply to the interpretation of the experiments in which the authors characterize the dynamics of intersubunit rotation in the presence of unlabeled RF1 and RF3 or unlabeled RF2 and RF3. Particularly careful attention should be paid to the unlabeled RF2 and RF3 experiments, since the authors have discovered that RF2 binds to termination complexes with a very low affinity such that the termination complexes may not be saturated with RF2 at the RF2 concentrations that are used for these experiments. Such a scenario would again result in compositional heterogeneity that would make interpretation of the unlabeled RF2 and RF3 experiments difficult and, in addition, would challenge the appropriateness of comparing these results with the results of the RF1 and RF3 experiments in which the termination complexes are more likely to be saturated with RF1 (or at least have lower compositional heterogeneity due to the higher affinity of RF1 for termination complexes).

3) The authors need to be much more cautious regarding their assignment of the 0.5 FRET state that is observed in various intersubunit FRET experiments recorded in the presence of RF2. The authors have assigned this FRET state as corresponding to the rotated state of the ribosome and have made no distinction between this rotated state of the ribosome and the rotated state of the ribosome that is observed in other contexts (e.g., in the absence of any factors).

Nonetheless, as the authors and many others have pointed out, structural/steric considerations associated with intersubunit rotation make such an assignment very surprising. Given that it is based on a single FRET measurement on a single construct, how confident can the authors really be about this assignment? Is it possible that local conformational changes involving S6 and/or L9, but not associated with global rotation of the ribosome (or at least a full, global rotation of the ribosome) could lead to a decrease in the distance between the fluorophores so as to generate this decrease in FRET? What about photophysical considerations, could binding of RF2 have directly or indirectly altered the photophysical properties of one and/or the other fluorophore in a manner that is independent of intersubunit rotation? How do the authors' observation that RF2 has a low affinity for, and binds only transiently to, termination complexes play into this? Is it possible that sampling of the rotated state of the ribosome only happens under conditions in which RF2 has dissociated from the termination complex due to the low affinity (this relates to the concerns regarding whether the complexes are actually saturated with RF2)? Unless the authors can present arguments or controls to eliminate these alternative interpretations or, better yet, provide additional, independent data that RF2-bound termination complexes can occupy the rotated state, I don't think the assignment of this FRET state is supported by the data that has been presented here.

4) The fits to many of the dwell time histograms are biphasic, which indicates the presence of kinetic heterogeneity in the corresponding smFRET experiments. In each case, the authors should analyze the individual trajectories to determine and report whether a particular experiment exhibits static or dynamic heterogeneity and what the most likely origin of that heterogeneity is. The authors should be particularly attentive to static heterogeneity, which may be indicative of compositional heterogeneity arising from termination complexes that may not be saturated by a particular factor.

5) With the exception of Figure 6A, Figure 6B, and Figure 2—figure supplement Figure 2E, the data that are plotted and graphed do not have error bars. Additionally, the amplitudes and rate constants presented in Supplementary file 1 do not have standard deviations. Thus, it is not clear that the majority of the experiments were repeated and, if they were repeated, it is not clear why the authors have not performed and reported the statistical analyses necessary for assessing the reproducibility of the results and the validity of the interpretations. The authors should repeat the experiments and/or perform and report the statistical analyses of the data.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for submitting your article "Dynamics of ribosomes and release factors during translation termination in E. coli" for consideration by eLife. Your article has been evaluated by James Manley (Senior Editor) and three reviewers, one of whom is a member of our Board of Reviewing Editors. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We have received comments from three reviewers (two of them new since the previous version). As you will read in the detailed comments, all three reviewers appreciated the substantial amount of work contained in the manuscript and the many interesting insights derived from the data. However, the reviewers remain concerned about the physiological relevance of the short peptidyl-tRNA ribosome complexes being studied. The authors offer cryoEM evidence that these tRNAs do readily sample the rotated state, especially at higher temperatures, and indeed, reviewer 2 suggests that there is additional literature supporting this point that should be clearly cited. Despite these arguments, the authors should acknowledge that these ribosome complexes carrying short peptidyl-tRNAs may not fully reflect the behavior of longer, more physiologically relevant, peptidyl-tRNAs where the classic ribosome configurations are more stabilized. As such, the authors should acknowledge that the detailed examinations here provide a starting point for defining the complex molecular events of termination rather than a definitive description.

In addition to these general concerns, reviewer 2 had numerous concerns about the statistical analysis throughout the study (in particular, whether the FRET histograms of the data subsets actually look like the FRET histograms of the total population). Reviewers 2 and 3 felt that conclusions were generally overstated given the limitations of the data (specifics are detailed in the reviews). Despite these limitations, all three reviewers felt that this manuscript contained important insights for the ribosome field, first on the relative differences in behavior between RF1 and RF2 on the ribosome (which is undoubtedly relevant to their different in vivo roles), and second, on the dynamics of RF1 and RF2 on the ribosome, and how they are impacted by the GTPase RF3. Finally, all three reviewers felt that the critical insights of the manuscript were often lost in the dense writing style, the overstatement of conclusions, and the lack of clarity in relating the work to some previous studies. At this stage, eLife will consider the manuscript for publication if the reviewer issues can be thoughtfully and completely addressed, both at the level of de-compressing the manuscript, clarifying the critical statistics, and more cautiously stating the conclusions.

Reviewer #1:

This revision of the manuscript by Adio et al. addresses most points raised by reviewers in the previous submission. All three reviewers were concerned about the peptidyl-tRNA migrating to the rotated state and the implications of this for the termination mechanism – the authors argue that this has been observed previously with relatively short peptidyl-tRNAs (they note in particular in cryoEM structures by Fischer et al.) and therefore rationalizes the data here showing rotation into the R state in complexes containing peptidyl-tRNAs. There were broad concerns related to the modification state of the RFs (whether fully methylated), the impact of the fluorescent labels on their behavior, the saturation of factors in various experiments, and the statistics of the analysis. The authors have systematically addressed each of these concerns to my satisfaction.

This paper represents a tour de force analysis with mountains of data (smFRET, ensemble FRET, and ensemble kinetic assays) on the conformational changes of various factors, tRNAs and the ribosome itself associated with peptide release and release factor binding (RF1, RF2, and RF3). These studies lead to several new findings that are important for defining the roles of these critical factors in translation termination, and indeed in defining the roles of such factors more broadly in biology. The most important findings are:

1) RF1 and RF2 behave quite distinctly on the preHC and postHC complexes – RF1 stabilizes the non-rotated state, RF2 enriches the rotated state and dissociates quickly (even without RF3). These differences are interesting in light of the auxiliary roles played by RF2 in quality control mechanisms in bacteria (both post-peptidyl QC as characterized by Zaher et al. and ArfA-mediated rescue). While the data don't tell us why these factors behave differently, they provide a biophysical basis for thinking about their distinct in vivo functions.

2) RF3-GTP stabilizes the rotated state and dissociates quickly from pre or post termination complexes.

3) Binding of RF3 to ribosomes bound with RF1 changes the conformation of both factors and stabilizes RF3 binding. RF1 and RF3 dissociation appears random; if RF3 dissociates first (from the non-rotated state), it does not require GTP hydrolysis, but if RF1 dissociates first, RF3 requires GTP hydrolysis to leave from the rotated state. It might be useful for the authors to compare the rates that they observe for RF1 departure as promoted by RF3 to previous studies determined by fluorescence (Koutmou et al.). Again, the differences here relative to RF2 (which does not depend on RF3 function) are interesting.

4) Rotation, peptide release, GTP hydrolysis are not coupled in a straightforward manner. These data are extremely dense, and include differences in behavior related to the type of GTP analog used (as previously reported in biochemical and structural studies). This section may have been the most difficult to sort through and I wonder whether the unnaturally short substrates (short peptidyl-tRNAs) and the lack of active release during the experiment (postHC complexes prepared by puromycin release) limited the impact/accuracy of the conclusions.

Overall, I feel the manuscript contains a substantial amount of important data on the dynamics and function of termination factors on the ribosome during translation termination. These data fit nicely with earlier studies by the same group detailing the critical role of the RF3-GTP cycle during these same steps (and extended here). The challenge for the manuscript remains that it is extremely dense and the main points are often lost in the detailed discussions of complex experiments. As just one example, the FRET distribution plots are layered with color (blue, pink and red, which all look very similar), to give the dimension of dynamics – which is useful and important, but nevertheless overwhelming. I broadly support publication of this work in eLife but would ask that the authors take one more pass to increase the accessibility of their main conclusions. Perhaps the problem is this: there are two stories here (1) the details of the functional cycle of RF1 and RF3 on the ribosome and (2) the distinctions in behavior between RF1 and RF2 on the ribosome. Yes, these are related stories, but presented together, the reader struggles to figure out whether to pay attention to commonalities or differences.

Reviewer #2:

The manuscript by Adio et al. describes an integration of single-molecule imaging and ensemble kinetic studies to explore the process of translation termination on the bacterial ribosome mediated by either release factor 1 or 2 (REF1, RF2) in concert with RF-3. The authors present a multitude of experiments describing the impacts of RF1 or RF2 binding on the conformational dynamics of various ribosome complexes, the rates of peptide release and the effects of RF3 on these various processes. Included in these investigations are direct measurements of factor binding interactions with the ribosome via FRET measurements as well as GTP hydrolysis studies to address an open question in the field about the role of GTP hydrolysis in the release mechanism and to refute reports that RF3-GDP is the physiological substrate for the ribosome in termination. There is no doubt that the synthesis of all of these data required tremendous effort both technically and intellectually. Although I did not see the original manuscript, it would appear that the addition of a second structural perspective on the classical-hybrid equilibrium (shown in Figure 3B, E) increase one's confidence in the interpretation that acylated tRNAs can indeed achieve hybrid-like configurations (see more about this below).

Although respectful of the amount of work that went into the present manuscript, my overarching conclusion is that it is exceedingly complicated. The salient physiologically relevant conclusions from the study are hard to grasp. The integration of ensemble and single-molecule experiments is of course extremely helpful at times as it provides confidence and grounding, but the number of experimental systems examined and the speculative conclusions made are dizzying, making it hard to keep track of key considerations. For instance, quantitative analyses are only provided for the subset of molecules that exhibit dynamics: it is not immediately clear how the proportion of non-dynamic/static molecules in each experiment affects the interpretations that are made; the existence of "static" and "dynamic" classes gives rise to a general concern about contributions of biochemical heterogeneities to the analyses presented. Do histograms of the small subset of dynamic molecules mirror the ensemble? The analyses presented are particular concerning given the authors use of/interpretation of these rate information, which is sometimes based on just 20-40 molecules from the hundreds that are measured (Figure 1A, B, E; Figure 5A; Figure 1—figure supplement 1A, C, E; Figure 1—figure supplement 2B, D; Figure 6—figure supplement 1A, C, D). Error bars on the individual measurements seem to be lacking throughout. The underlying basis of the static and dynamic populations is not clearly explained and should be clarified. As written, the manuscript seems to imply that this is expected from the biochemical system. But this is not clear to me where this notion comes from. The easier explanation is that this arises from rapid fluorophore photobleaching prior to evident conformational changes – in this context, I was unable to find the photobleaching rates for the distinct systems examined in the manuscript but it appears to be rapid (i.e. 0.5-1 per second) and thus a limiting feature to the experimental setup. Each of these concerns seem more or less consistent with the reviewer comments provided during initial review of the manuscript. My sense is that these considerations are likely to render the manuscript challenging to distill for the general reader.

One of the points raised by the initial reviewers is that significant complexities in the interpretation of the data presented arise from the use of ribosome complexes bearing short peptide mimics (fMET-Phe, fMET, NAc-Phe), which allows the ribosome to fluctuate between classical and hybrid states in the absence (and presence) of RFs. Although the authors choose to reference their own cryo-EM work indicating that the small subunit spontaneously and reversibly rotates at elevated temperatures, multiple studies prior and subsequent to their chosen reference have indicated hybrid-like states can be achieved with acylated-tRNA in the P site and that such hybrid-like configurations are sensitive to the nature and length of the nascent peptide (see Cornish et al., 2008; Cornish et al., 2009; Munro et al., EMBO 2008; Alejo et al., PNAS 2017). Referring to such states as "Hybrid" is inappropriate as structural data indicate that this includes A76 interactions with the C2394 region.

The average reader is likely to be confused by this potentially non-physiological aspect of the termination studies presented. As the initial reviewers suggested, had the study been performed with longer nascent chains, as is expected for the physiological substrate of release factors during translation termination, this complexity could likely have been entirely avoided. The fact that it is difficult to prepare such complexes (as the authors indicate in their response to reviewers) is fully appreciated, but a focused study on the propensity of the chosen complexes to achieve hybrid configurations and the nuanced impacts of such dynamics on release factor binding could be seen as an appropriate prerequisite for the present studies and their interpretation of the results presented.

For example, one of the key takeaways from the present study from the perspective of the general reader is that RF1 and RF2 perform differently in termination based on their measured off rates. Yet, such differences may simply arise from small distinctions in the binding energies of RF1 and RF2 to the ribosomes, which reveal themselves as potentially meaningful in the context of the non-physiological substrates examined. Such binding idiosyncrasies may be further exacerbated by the GGQ mutation that is used to prevent peptide release (as seems to be evident when comparing Figure 1—figure supplement 2C and E). What is the argument that these distinctions are physiologically important? Focused studies on the binding differences of RF1 and RF2 (concrete measurements of affinity, on and off rates, beyond the information that is presented) and clarification as to why such differences are important in the context of RF3 functions seems relevant to discuss but is presently lacking.

More readily addressable issues of concern include the following:

1) The use of language that could be considered inaccurate, misleading or too interpretive.

For instance, in the Introduction it states, "An alternative model was proposed, when it turned out that the affinity of RF3 to GDP and GDP[…]" Aren't these simply different measurements? As written it implies that the prior studies were just wrong and the studies by Koutmou and Peske are just right. How are the authors so sure that this is the case? Was the prior work retracted due to a technical error?

It is written in the Introduction, "thus the lifetime of the apo-RF3 complex would be too short to assume a tentative physiological role". The work cited seems to make a thermodynamic rather than a kinetic argument and the statement as written seems to create a false context.

In the Results its states that "the exact distribution of states depends on P-site tRNA […]" referencing Fei et al., 2011, when prior literature on the classical – to – hybrid equilibrium (N to R transition) were prior to report this finding. The authors should include references to Cornish et al., 2008; Cornish et al., 2009 and Munro et al., 2010 and Munro et al., 2010, which are already included in the bibliography.

The discussion in the subsection “RF1 and RF2 have distinct

effects on ribosome dynamics” regarding Figure 1—figure supplement 3 seem to suggest that there is information about factor binding in the data presented but this figure is confusing. The authors want to interpret differences between RF1 and RF2 'contour' plots that are post-synchronized as differences in binding and unbinding, but direct information about binding and unbinding is not present in these data. These types of plots may indicate differences in dynamics within the bound complexes and this needs to be clarified. While I have no doubt that there may be binding and unbinding information underlying the differences observed, the language used in this section is too strong. The authors should also try to put sufficient information in the figure legends to enable one to discern the relevant information about the experiment but looking at the figure legend alone. From the legend of Figure 1—figure supplement 3 it is impossible to know what ribosome complex was being examined (where does the FRET come from) without going back to the main text.

In the results it states that "the P/E hybrid state is short lived and decays rapidly (koff = 6 per second)" As this is an experiment done in the presence of RF3, the rate constant koff seems to imply unbinding of RF3, whereas I think the authors mean the classical to hybrid transition.

It also states that "suggesting the two processes are not tightly coupled, consistent with previous notions based on cryo-EM […] and smFRET work" While subtle, as stated this sentence implies that the cryoEM work came first and the smFRET work came later. But the opposite is true.

In reference to Figures 4C, G and K it is stated that "The FRET efficiency for the RF3-L11 pair is closer to 0.5, changed from 0.7, which is observed in the absence of RF1 (Figure 3C). Thus, the orientation of RF3 in the complex differs from that formed in the absence of RF1" While such conclusions may indeed be correct, statistical analyses on one dimensional histograms comprised of biological repeats are needed to make this conclusion without a second structural perspective to support this interpretation. The difference is not obvious and could be the result of day-to-day variabilities in microscope performance and/or fluorophore behaviors.

In reference to Figure 4 it is stated that "Thus, RF1 and RF3 can reside simultaneously on the Post HC in an arrangement where the position of RF1 and RF3 relative to L11 is shifted compared to the complex with only a single factor". Here, it seems too strong to make such statements without direct evidence of RF1 and RF3 binding.

Reviewer #3:

The manuscript by Adio et al. explores the mechanism of release factor recycling in bacteria using single-molecule FRET approach. The authors took advantage of different labeling strategies to look at ribosome, tRNA and release factor conformations at different stages of the termination process. These were used to draw two main conclusions: 1) Unlike RF1, RF2 can dissociate efficiently from the ribosome in the absence of RF3 (although RF3 helps); 2) The order of binding and dissociation events appear to be stochastic and unlike other translational GTPases, GTP hydrolysis by RF3 serves as a rescue mechanism for the factor.

Overall the paper was well written and the experiments appear to have been well executed. I was not however convinced that the paper offered significant new insights into the mechanism of translation termination. The notion that RF1 and RF2 are slightly different is not new (for example ArfA rescue of ribosomes only works in the presence of RF2 and not RF1). Furthermore the authors played down the effect of RF3 on RF2 (the fold change in termination rates under substoichiometric conditions is similar to that observed for RF1). Whether these differences also apply to other peptidyl tRNAs was not explored. As for the second main point of the paper, the Ruben group has also looked at the dynamics of the tRNAs and ribosome during recycling (albeit not to the same extent in this paper) and similar observations were made.

As for the presentation of the data, I felt that the results were over interpreted and the main conclusion was to a certain extent based on conjecture. For instance, the idea that the process of recycling is stochastic is based on the observation that many of the rates being similar (RF1, RF3 dissociation for example), but these experiments were conducted in the absence of peptide release (either using post hydrolysis complex or pre-hydrolysis one with GAQ RF1). It's unclear whether peptide release plays a role in the conformational changes. This is relevant, as under normal conditions release factors must bind in the presence of a peptidyl-tRNA and promote peptide release before it is recycled. At the end, I was left wondering how the different assays recapitulate what happens under normal conditions.

Another major point is the authors' assertion that GTP hydrolysis by RF3 merely serves to rescue the factor from nonproductive binding events. The data is based on the observation that in the presence of excess RF3, GTP hydrolysis in not required for RF1 turnover. The data in Figure 6A then suggests that RF3 is as likely to bind in a nonproductive fashion as to a productive one.

Given the overall contribution of the paper to the field together with the less than ideal interpretation of the data, my overall enthusiasm of the paper was tempered.

https://doi.org/10.7554/eLife.34252.022

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

Reviewer #1:

[…] Concerns:

In the first paragraph of the subsection “RF1 and RF2 have distinct effects on ribosome dynamics”, the authors state that the tRNA in the PreHC does not affect FRET dynamics (and for these experiments they use fMet-tRNA). However, the figures seem to depict a dipeptidyl-tRNA. This should be clarified in the text, figures, and figure captions.

We changed the figures to depict the formyl group by a different symbol than that representing an amino acid. The respective text has been changed/added (Figure 1 legend).

More importantly, it raises the question as to why they observe the rotated (R) state in the PreHC upon either RF2 or RF3 binding. Is this allowed rotation specific to the use of this minimal substrate fMet-tRNA? What would happen with a longer peptidyl-tRNA that extends into the exit tunnel? This is an important question because this finding suggests (contrary to the literature) that the hybrid tRNA states and rotated state of the ribosome can be separated.

Previous cryo-EM reconstructions have demonstrated that peptidyl-tRNA can assume a hybrid/rotated state (Fischer et al., 2010). There is a notion in the field that a peptidyl tRNA (and fMet-tRNA as a model peptidyl-tRNA) cannot move to the hybrid state. The cryo-EM work by Fischer et al. shows that this pertains to complexes prepared at low temperatures (4°C), but is not true for higher temperatures. At 18°C, the distribution of the subunit rotation states is bimodal, with the discernable peaks of non-rotated (about 60%) and rotated (40%) states, whereas at 37°C almost any rotation state is possible. Given that the smFRET data presented in this paper are obtained at 22°C, it is not surprising that fMet-tRNA can move into the rotated/hybrid state, which consistent with the cryoEM results (Figure 4 in Fischer et al., 2010).

To further validate the orientation of fMet-tRNAfMet in the PreHC in the presence of RF2 or RF3 we performed additional experiments. We monitored the movement of fMet-tRNAfMet into the hybrid state using FRET between the donor label at the tRNA (fMet-tRNAfMet(Cy3)) and the acceptor label on the ribosomal protein L1 (L1(Cy5)). This FRET pair has been extensively validated in previous studies as a reporter to probe whether the P-site tRNA is in the classical P/P state or in the hybrid P/E state (Munro et al., 2010a, Munro et al., 2010b, Munro et al., 2010c, Fei et al., 2008, Sternberg et al., 2009, Fei et al., 2009). We find that in the presence of RF2 the fraction of tRNAs in the P/E state is 20% which corresponds to the fraction ribosomes in the rotated state under the same conditions (Figure 3—figure supplement 1B). We also provide further controls as to the stability of fMet-tRNAfMet in the presence of RF2(GAQ) mutant (Figure 1—figure supplement 2E). In the presence of RF3 the fraction of fMet-tRNAfMet adopting the P/E state is almost 40% and the traces are extremely dynamic (now shown as Figure 3B, E). These findings further support our conclusion that RF3 changes the conformational dynamics of the PreHC. We also provide additional text explaining that peptidyl-tRNA can adopt a hybrid/rotated state (subsection “Interaction with RF3-GTP”).

Concerning the experiments with PreHC carrying a longer peptidyl chain, those are well beyond the scope of this paper for several reasons. First, fMet-tRNAfMet has proven to be a good analog of peptidyltRNA with respect to studies of translation termination and several groups working on termination use it as a model system (Sternberg et al., 2009, Shi et al., 2016, Koutmou et al., 2014); the peptide length appears to have a small (if at all) effect the rate of peptidyl-tRNA hydrolysis (Indrisiunaite et al., 2015), which is not relevant for understanding the fundamental mechanism of termination. Second, the whole set of new experiments with a longer peptidyl-tRNA presents a significant technical challenge, because it requires preparation of homogeneous FRET-labeled ribosome complexes and validation of their biochemical and photophysical performance. Preparing such fully homogeneous complexes is challenging, because the efficiency of each translation step might be less than 100%; such technical problems cannot be solved within a reasonable time scale. Given that the hybrid/rotated state of peptidyl-tRNA has been reported for a peptidyl-tRNA (fMetVal-tRNAVal)by Fischer et al. (2010), and our data show that also fMet-tRNAfMet can adopt a hybrid/rotated state, we feel that repeating the experiments with a longer peptidyl moiety would bring too little additional information to justify the enormous effort necessary to perform such additional experiments.

In Figure 3 and Figure 6G, release of RF3 is much faster than GTP hydrolysis, but one of the main conclusions is that RF3 requires GTP hydrolysis to be released from the rotated state (Figure 6D, H).

In the absence of RF1, RF3 can rapidly dissociate from some ribosomes, but these experiments do not show whether RF3 dissociates from the N or R state of the ribosome (Figure 3). In contrast, ensemble kinetics of Figure 6D shows that the back-rotation from the R state is very slow and the rate of backrotation corresponds to the rate of GTP hydrolysis (~0.5 s-1, Peske et al., 2014). Therefore, we suggest that the rapid dissociation monitored by smFRET reflects RF3–GTP dissociation from the N state, whereas RF3 release from the R state appears too slow to be captured by the smFRET technique. This is plausible, as the expected rate of GTP hydrolysis at smFRET conditions (22°C) is slower than 0.5 s-1, which is measured at 37°; the rates <0.2 s-1 are usually not resolved in our smFRET experiments. There is no contradiction between the data sets, as smFRET and ensemble measurements show different parts of the mechanism. In the presence of RF1, dissociation of RF3 is entirely independent of GTP hydrolysis. We changed the discussion in the subsection “The role of GTP binding and hydrolysis” to make this point clearer.

The confusion comes primarily from the RF3 dissociation experiments carried out with non-hydrolysable analogs, which do not behave as authentic GTP in our experiments. This is evident from experiments showing that RF3–GDPNP and RF3–GTPγS do not induce dynamic fluctuations of the ribosome–RF3 complex (Figure 6—figure supplement 1). If dynamic fluctuations are necessary for the rapid dissociation of RF3, and this is induced only by an authentic GTP-like conformation of RF3, then non-hydrolysable analogs are not suitable to study this question. These considerations motivated us to remove Figure 6G, H from the main text to the Figure 6—figure supplement 4. The text is modified to describe the potential problems of non-hydrolysable analogs.

Reviewer #2:

[…] The implication of their control (with just tRNAs) is that the rotated state with RF1,2/3 is similar that reached with tRNAs during translocation. However, a P/E tRNA state is not possible for the preHC complex, because it requires a deacylated tRNA. So what is the R state that RF2 reaches in the preHC? This is never actually addressed.

The notion that the ribosomes with peptidyl-tRNA cannot adopt rotated/hybrid state is based on structural work (cryo-EM or X-ray) that typically used low temperatures for complex preparation, whereas our experiments are performed at 22°C. The results of cryo-EM studies (Fischer et al., 2010), who have extensively studied the distribution of the ribosome rotation states at different temperatures, show that peptidyl-tRNA (fMetVal-tRNAVal in their experiments) favors the nonrotated/classical state only at low temperatures (4°C). In contrast, at 18°C the distribution of subunit rotation states is bimodal, with the discernable peaks of non-rotated (about 60%) and rotated (40%) states, whereas at 37°C almost any rotation state is possible (Figure 4 in Fischer et al., 2010). Given that the smFRET data presented in this manuscript are obtained at 22°C, it is thus not surprising that fMettRNAfMet can move into the rotated/hybrid state, which is entirely consistent with the bimodal distribution revealed by the cryo-EM. We agree that this was not sufficiently explained in the manuscript and added the pertinent discussion in the revised manuscript.

Moreover, to address the question whether fMet-tRNAfMet in PreHC can sample the P/E hybrid state we used FRET labels attached to the ribosomal protein L1 (L1(Cy5)) and fMet-tRNAfMet (fMet-tRNAfMet(Cy3)). This FRET pair has been used in previous studies as a reporter to probe whether the P-site tRNA is in the classic P/P state or in the hybrid P/E state (Munro et al., 2010a, Munro et al., 2010b, Munro et al., 2010c, Fei et al., 2008, Sternberg et al., 2009, Fei et al., 2009). Our data clearly show that the presence of RF2 or RF3 not only promote subunit rotation but also the hybrid state formation of the tRNA (Figure 3C, F and Supplementary file 1).

A larger question they might have asked is why not all bacteria have RF3, and whether this is related to a possible role of RF3 in quality control shown by the Green lab.

This is an interesting question which is entirely beyond the scope of this paper. In E. coli, deletion of RF3 induces a higher expression of RF2 (Baggett et al., 2017). It is likely that bacteria that do not express RF3 compensate by having more RF2, or release RF1 with the help of other, yet uncharacterized factors (e.g. HflX).

In summary, the paper deals with details of the mechanism of translational termination in bacteria that although important, will be of interest to only a handful of people even in the ribosome field. My own feeling is that it is not clear that it belongs in a general interest journal like eLife.

On this point we politely disagree with the reviewer. We think that it is important to get the mechanism of translation termination right and make essential corrections to the inaccurate models that made their way to the textbooks. On a broader scale, this work shows that simple mechanistic models are not suitable to describe the dynamics of complex machineries, which may be an important lesson not only for the ribosome field but also for others dealing with macromolecular ensembles.

Reviewer #3:

[…] 1) The authors find that, unlike RF1, RF2 has a relatively low affinity for pre- and post-hydrolysis termination complexes and binds only transiently to these complexes. As the authors point out, this is a very surprising result. As such, it raises many important concerns that could be easily addressed by controls:

1a) The authors do not demonstrate whether their RF2 (or RF1) constructs are methylated at the Q of the GGQ motif in domain 3. Given that the Ehrenberg group has shown that the affinity of RF2 for termination complexes and the catalytic activity of RF2 on termination complexes are both dependent on this post-translational modification (Pavlov, et al. (1998) J Molec Biol and Dincbas-Renqvist, et al. (2000) EMBO J), it is important that the authors demonstrate that the surprising results they have obtained with RF2 are not due to the lack of this post-translational modification.

The factors are fully methylated. This is now shown in Figure 2—figure supplement 2 and stated in the respective text.

1b) The authors claim that their fluorophore-labeled RF2 (and RF1) construct are as active as their unlabeled counterparts, but the data are not shown. Given the results that the authors have obtained with their fluorophore-labeled RF2 construct, it seems to me that controls demonstrating that both the affinity and the catalytic activity of the authors' fluorophore-labeled RF2 construct are unchanged relative to unlabeled RF2 must be shown.

The activity is now documented in Figure 2—figure supplement 1.

It is also important to specify whether this comparison is being made to the unlabeled, single-cysteine mutant RF2 construct or to the unlabeled, fully wildtype RF2 construct.

The comparison is to the fully wild type unlabeled RF1 or RF2 (Figure 2—figure supplement 1).

All of this is made more important by the fact that the authors have mutated and fluorophore-labeled a position on RF2, A237, that is very different from the position that they have mutated and labeled on RF1, has not been previously characterized, and is located in domain 3, where it could easily affect the affinity and/or catalytic activity of RF2 in a manner similar to that which is observed for the methylation of the Q in the GGQ motif in domain 3.

The labeled position is C273, we apologize for the misprint. The activity of the factor is not changed (Figure 2—figure supplement 1).

2) In several instances, the authors seem to interpret their data under the assumption that the termination complexes are saturated with unlabeled components, without having convincingly argued or demonstrated that the complexes are saturated:

2a) In the experiments in which unlabeled RF3 is added to termination complexes that are interacting with fluorophore-labeled RF2 shown in Figure 2—figure supplement 2A, C. What is the affinity of the unlabeled RF3 for these complexes? Is the concentration of unlabeled RF3 that the authors use for these experiments high enough such that the complexes are saturated? How dependent are the interpretation of these data and the conclusions that are drawn on the complexes being saturated with unlabeled RF3? It seems like the extremely low, 10 nM concentrations of fluoropore-labeled RF2 and the possibility that, at any one time, the termination complexes are only partially occupied with RF3 would generate compositional heterogeneity that would make the data hard to interpret. The authors should address these questions through controls (e.g., titrations of fluorophore-labeled RF2 and/or unlabeled RF3) and/or revisions to the manuscript.

The concentration of RF3 is 10-fold over the saturation concentration determined from the RF3-dependence of RF1 turnover rate (see Figure 2—figure supplement 1D and Zavialov et al., 2002). Despite this very large excess, it is difficult to entirely exclude a minor fraction of ribosomes that do not have RF3 at the moment when RF2 arrives. We would have easily recognized such complexes, because they are mostly static and favor the N state (Supplementary file 1). If some of the complexes have been taken for analysis, we would slightly underestimate the effect of RF3, which would make the observed trend even stronger.

2b) Similar considerations apply to the interpretation of the experiments in which the authors characterize the dynamics of intersubunit rotation in the presence of unlabeled RF1 and RF3 or unlabeled RF2 and RF3. Particularly careful attention should be paid to the unlabeled RF2 and RF3 experiments, since the authors have discovered that RF2 binds to termination complexes with a very low affinity such that the termination complexes may not be saturated with RF2 at the RF2 concentrations that are used for these experiments.

Despite the difference in the koff values for RF1 and RF2, the complexes are saturated with factors at the high concentrations used in the inter subunit experiments (1 µM) (Figure 2—figure supplement 1B, C and Zavialov et al., 2002).

Such a scenario would again result in compositional heterogeneity that would make interpretation of the unlabeled RF2 and RF3 experiments difficult and, in addition, would challenge the appropriateness of comparing these results with the results of the RF1 and RF3 experiments in which the termination complexes are more likely to be saturated with RF1 (or at least have lower compositional heterogeneity due to the higher affinity of RF1 for termination complexes).

The ribosomes that do not have either RF1 or RF2 would show the rotation distribution of the complexes with RF3 alone. However, the distribution between the N and R states is clearly different with RF3 alone (73% R state) and RF3 with RF1 (43%) or RF3 and RF2 (56%). If some of the complexes lack RF1 or RF2, the effect would be underestimated, which does not change the trend, but makes it even stronger. We do not interpret small, statistically insignificant differences in rotation distribution between RF1+RF3 and RF2+RF3 experiments.

3) The authors need to be much more cautious regarding their assignment of the 0.5 FRET state that is observed in various intersubunit FRET experiments recorded in the presence of RF2. The authors have assigned this FRET state as corresponding to the rotated state of the ribosome and have made no distinction between this rotated state of the ribosome and the rotated state of the ribosome that is observed in other contexts (e.g., in the absence of any factors).

Nonetheless, as the authors and many others have pointed out, structural/steric considerations associated with intersubunit rotation make such an assignment very surprising. Given that it is based on a single FRET measurement on a single construct, how confident can the authors really be about this assignment? Is it possible that local conformational changes involving S6 and/or L9, but not associated with global rotation of the ribosome (or at least a full, global rotation of the ribosome) could lead to a decrease in the distance between the fluorophores so as to generate this decrease in FRET? What about photophysical considerations, could binding of RF2 have directly or indirectly altered the photophysical properties of one and/or the other fluorophore in a manner that is independent of intersubunit rotation? How do the authors' observation that RF2 has a low affinity for, and binds only transiently to, termination complexes play into this? Is it possible that sampling of the rotated state of the ribosome only happens under conditions in which RF2 has dissociated from the termination complex due to the low affinity (this relates to the concerns regarding whether the complexes are actually saturated with RF2)? Unless the authors can present arguments or controls to eliminate these alternative interpretations or, better yet, provide additional, independent data that RF2-bound termination complexes can occupy the rotated state, I don't think the assignment of this FRET state is supported by the data that has been presented here.

The FRET pair used here to monitor subunit rotation is one of the best characterized FRET pairs (Sharma et al., 2016, Belardinelli et al., 2016, Quin et al., 2014, Ermolenko et al., 2007, Ermolenko et al., 2013, Cornish et al., 2008). It has been used to study subunit rotation on free ribosomes, diverse ribosome complexes with a number of different tRNAs and with and without elongation factors. Noller et al. and our group have shown that the S6-L9 FRET pair does not affect the ribosome function (Sharma et al., 2016, Belardinelli et al., 2016, Ermolenko et al., 2007, Ermolenko et al., 2013, Cornish et al., 2008, Hickerson et al., 2005, Majumdar et al., 2005). Clegg and Noller have tested all potential photophysical effects (Hickerson et al., 2005, Majumdar et al., 2005) and we have further characterized the behavior of the fluorophores upon peptide bond formation and translocation (Sharma et al., 2016, Belardinelli et al., 2016). We find the same two states – previously characterized as N and R – without the factors or with EF-G, EF-Tu, RF1, RF2, or RF3 regardless of the affinity of the factors. The FRET efficiencies are identical to those measured in other ribosome complexes and there is no peak broadening in FRET histograms. Termination factors cannot affect the reporter directly, because they bind from a different side of the ribosome, and there is no indication for the existence of some yet unobserved indirect effects. The available structures do not reveal anything unusual in the S6-L9 region upon termination factor binding; furthermore they demonstrate that the N state favors by RF1 or RF2 is the same N state as in the absence of the factors and likewise the R state formed with RF3 is similar to that formed during translocation. Thus, there is absolutely nothing to indicate that the two states we observe here are grossly different from the well-characterized N and R states observed so far with a wide variety of ribosome complexes.

Although there is no indication for the existing “alternative” rotation state, we made an additional experiment to show that the formation of the rotated state is accompanied by the tRNA movement into the hybrid state using another well-established FRET pair with the labels on the ribosomal protein L1 (L1(Cy5)) and fMet-tRNAfMet (fMet-tRNAfMet(Cy3)). These experiments clearly show that ribosome rotation is accompanied by the hybrid state formation (Figure 3B, E); thus, two of the hallmarks of the R/hybrid H/classic states are satisfied. The law of parsimony does not leave us any other choice but to conclude that the N and R states formed in the presence of termination factors are the same as sampled by all other ribosome complexes studies so far. We assume that the attempt to re-interpret the R state observed in these experiments as an alternative conformation stems from the deeply-routed believe that a peptidyl-tRNA cannot move into the R/hybrid state. This is however true for low temperatures only, whereas at room temperature the complexes can adopt both N/classic and R/hybrid state (Fischer et al., 2010, Figure 4). (see also replies to reviewers 1 and 2).

4) The fits to many of the dwell time histograms are biphasic, which indicates the presence of kinetic heterogeneity in the corresponding smFRET experiments. In each case, the authors should analyze the individual trajectories to determine and report whether a particular experiment exhibits static or dynamic heterogeneity and what the most likely origin of that heterogeneity is. The authors should be particularly attentive to static heterogeneity, which may be indicative of compositional heterogeneity arising from termination complexes that may not be saturated by a particular factor.

We analyze each trace individually and we separate traces into static (no transitions) and dynamic traces (Supplementary file 1). We report (as indicated in Supplementary file 1) transition rates of dynamic traces only. There is no kinetic heterogeneity in the static traces. The decay rate of static traces has little biologic meaning as it is dominated by the photobleaching rate of the FRET dyes. The compositional heterogeneity is addressed in our replies above. The two rates obtained from two-exponential fitting are difficult to interpret, and in most cases we focus on the major (70-100%) component. In some cases, the heterogeneity has a biological meaning, which is discussed in the text.

5) With the exception of Figure 6A, Figure 6B, and Figure 2—figure supplement 2E, the data that are plotted and graphed do not have error bars. Additionally, the amplitudes and rate constants presented in Supplementary file 1 do not have standard deviations. Thus, it is not clear that the majority of the experiments were repeated and, if they were repeated, it is not clear why the authors have not performed and reported the statistical analyses necessary for assessing the reproducibility of the results and the validity of the interpretations. The authors should repeat the experiments and/or perform and report the statistical analyses of the data.

All statistics are now presented in Supplementary file 1.

[Editors' note: the author responses to the re-review follow.]

Reviewer #1:

[…] This paper represents a tour de force analysis with mountains of data (smFRET, ensemble FRET, and ensemble kinetic assays) on the conformational changes of various factors, tRNAs and the ribosome itself associated with peptide release and release factor binding (RF1, RF2, and RF3). These studies lead to several new findings that are important for defining the roles of these critical factors in translation termination, and indeed in defining the roles of such factors more broadly in biology. The most important findings are:

1) RF1 and RF2 behave quite distinctly on the preHC and postHC complexes – RF1 stabilizes the non-rotated state, RF2 enriches the rotated state and dissociates quickly (even without RF3). These differences are interesting in light of the auxiliary roles played by RF2 in quality control mechanisms in bacteria (both post-peptidyl QC as characterized by Zaher et al. and ArfA-mediated rescue). While the data don't tell us why these factors behave differently, they provide a biophysical basis for thinking about their distinct in vivo functions.

A notion on the non-canonical roles of RF2 has been included in the Introduction, first paragraph.

2) RF3-GTP stabilizes the rotated state and dissociates quickly from pre or post termination complexes.

3) Binding of RF3 to ribosomes bound with RF1 changes the conformation of both factors and stabilizes RF3 binding. RF1 and RF3 dissociation appears random; if RF3 dissociates first (from the non-rotated state), it does not require GTP hydrolysis, but if RF1 dissociates first, RF3 requires GTP hydrolysis to leave from the rotated state. It might be useful for the authors to compare the rates that they observe for RF1 departure as promoted by RF3 to previous studies determined by fluorescence (Koutmou et al.). Again, the differences here relative to RF2 (which does not depend on RF3 function) are interesting.

The comparison with the rates reported in Koutmou et al. is included in the first paragraph of the subsection “Binding of RF1 and RF2 to the ribosome” and in the second paragraph of the subsection “Interplay between RF1 and RF3”. As stated in the text, we find that the data are in very good agreement given the different methods used to determine the dissociation rates.

4) Rotation, peptide release, GTP hydrolysis are not coupled in a straightforward manner. These data are extremely dense, and include differences in behavior related to the type of GTP analog used (as previously reported in biochemical and structural studies). This section may have been the most difficult to sort through and I wonder whether the unnaturally short substrates (short peptidyl-tRNAs) and the lack of active release during the experiment (postHC complexes prepared by puromycin release) limited the impact/accuracy of the conclusions.

We agree that the section on the role of GTP hydrolysis is complex. Also reviewer 2 commented that a combination of smFRET and ensemble kinetics is difficult to follow. To answer this criticism, we have restructured the text and removed the ensemble kinetics data. The arguments are now based on the smFRET data alone, which, we think, is easier to read (all ensemble and smFRET data give the same results anyway). We do not think that the short peptidyl substrates play a role here, as the key experiments on RF3 dissociation are carried out with PostHC where the peptide is released anyway. Furthermore, there is no indication that the puromycin treatment produces complexes that are different from those obtained with RF1, see the extended comparison between these complexes in Figure 4. The real surprise to us is why GTP hydrolysis is not needed for the recycling of the PostHC–RF1–RF3 complex, which is a naturally obtained Post complex with a deacylated tRNA in the P site.

Generally, we are surprised about the fundamental criticism concerning the short peptidyl-tRNAs as substrates, as this is a standard used by many in the termination field, e.g. Koutmou et al., 2014; Sternberg et al., 2009; Florin et al., 2017; Shi and Joseph, 2016; and a recent paper on rotation dynamics during termination by the Cornish lab (Casey et al., 2018). Ehrenberg and co-workers provided the direct comparison of the GTPase activity of RF3 on RF2-bound termination complexes with a tetrapeptide (MFTI) or fMet bound to P-site tRNA (Zavialov et al., 2003). They find that GTP hydrolysis rates are similar with MFTI and fMet which suggests that there is no significant difference in the interaction of termination factors with ribosomes carrying a short or a slightly longer peptide. Making the complexes with an authentic peptidyl-tRNA is not feasible, because the stop codons are read-through at the in vitro translation conditions in the absence of release factors and in the presence of large excesses of ternary complexes even at high accuracy conditions.

Overall, I feel the manuscript contains a substantial amount of important data on the dynamics and function of termination factors on the ribosome during translation termination. These data fit nicely with earlier studies by the same group detailing the critical role of the RF3-GTP cycle during these same steps (and extended here). The challenge for the manuscript remains that it is extremely dense and the main points are often lost in the detailed discussions of complex experiments. As just one example, the FRET distribution plots are layered with color (blue, pink and red, which all look very similar), to give the dimension of dynamics – which is useful and important, but nevertheless overwhelming. I broadly support publication of this work in eLife but would ask that the authors take one more pass to increase the accessibility of their main conclusions. Perhaps the problem is this: there are two stories here (1) the details of the functional cycle of RF1 and RF3 on the ribosome and (2) the distinctions in behavior between RF1 and RF2 on the ribosome. Yes, these are related stories, but presented together, the reader struggles to figure out whether to pay attention to commonalities or differences.

In order to make the representation of smFRET data less complex, we simplified the color scheme and now use only two colors (red and gray) to distinguish static from dynamic complexes. Furthermore, we removed the percentage of dynamic traces (% Dyn) from the histograms to simplify the graphs. We realized that as presented, this value apparently leads to confusion (see reply 1 to comments of reviewer 2). These values are still given in Supplementary file 1 where it is perhaps easier to understand. To make the writing style less dense, we extended the description of the experiments. Concerning the RF2 story, we are reluctant to separate it from the RF1/RF3 story, as they are intricately related. We hope that the general decompression of the paper helps to understand the results.

Reviewer #2:

[…] Although respectful of the amount of work that went into the present manuscript, my overarching conclusion is that it is exceedingly complicated. The salient physiologically relevant conclusions from the study are hard to grasp. The integration of ensemble and single-molecule experiments is of course extremely helpful at times as it provides confidence and grounding, but the number of experimental systems examined and the speculative conclusions made are dizzying, making it hard to keep track of key considerations. For instance, quantitative analyses are only provided for the subset of molecules that exhibit dynamics: it is not immediately clear how the proportion of non-dynamic/static molecules in each experiment affects the interpretations that are made; the existence of "static" and "dynamic" classes gives rise to a general concern about contributions of biochemical heterogeneities to the analyses presented. Do histograms of the small subset of dynamic molecules mirror the ensemble? The analyses presented are particular concerning given the authors use of/interpretation of these rate information, which is sometimes based on just 20-40 molecules from the hundreds that are measured (Figure 1A, B, E; Figure 5A; Figure 1—figure supplement 1A, C, E; Figure 1—figure supplement 2B, D; Figure 6—figure supplement 1A, C, D).

The histograms displayed in the manuscript represent the entire population of molecules, rather than just a subset of dynamic traces. We realized that the “% Dyn” in the figures was probably misleading and removed it from the graphs. These values are still available in Supplementary file 1. The definition of static and dynamic traces is provided in the legend to Figure 1 and Supplementary file 1.

Error bars on the individual measurements seem to be lacking throughout.

Supplementary file 1 provides the detailed statistical analysis including the standard deviations of all measurements. It is impossible to provide statistical deviation in contour plots or time trajectories, but the information derived from these data (mean and sd of the FRET values, distribution of different FRET values in the population, and rate constants) is given either in the text, figure legends, or Supplementary file 1. As we now added the FRET distribution histograms (see below), we also included the mean ± sd for FRET values in figure legends (they are also available in Supplementary file 1). As indicated in figure legends and Materials and methods, standard deviations were calculated from three independent datasets. In figures and the calculations of statistics we followed the standards in the smFRET field (see Alejo et al., 2017, Ning et al., 2014; Wassermann et al., 2015; Elvekrog et al., 2013; Fei et al., 2011). We would be grateful to the reviewer for more specific comment as to which statistical parameter is still lacking.

The underlying basis of the static and dynamic populations is not clearly explained and should be clarified. As written, the manuscript seems to imply that this is expected from the biochemical system. But this is not clear to me where this notion comes from. The easier explanation is that this arises from rapid fluorophore photobleaching prior to evident conformational changes – in this context, I was unable to find the photobleaching rates for the distinct systems examined in the manuscript but it appears to be rapid (i.e. 0.5-1 per second) and thus a limiting feature to the experimental setup.

The existence of static and dynamic populations of different ribosome complexes has been consistently observed by many groups in the smFRET field, including the papers by Sternberg et al., 2009 and Casy et al., 2018 for the termination complexes and the groups of Gonzalez, Blanchard, Goldman/Cooperman, Cornish and us for other complexes during translation. The basis for the existence of static and dynamic populations is not known and the clarification of this point is beyond the scope of the manuscript. It is not a biochemical phenomenon related to the homogeneity of the PreHC or PostHC (which are tested and are uniform in all experiments), but it depends on the presence/absence of the peptide and on factor binding, in particular of RF3. Throughout the revised version, we simplified the description of ribosome fluctuations and removed the% Dyn which caused obvious confusion (see above). The stimulating effect of RF3–GTP on ribosome dynamics is shown in Figures 3 and 4 and described in the Results section.

The photobleaching rate is in the range of 0.07-0.19 s-1, very similar to photobleaching rates published by other groups (for example kphotobleaching=0.05 ± 0.01 s-1-0.29 ± 0.03 s-1 in Sternberg et al., 2009). It is certainly much less than the 0.5-1 s-1 suggested by the reviewer, because if it were so high, we would not be able to measure dissociation rates as low as 0.2 s-1. The details of photobleaching estimations are described in Materials and methods (subsection “Data analysis”).

Each of these concerns seem more or less consistent with the reviewer comments provided during initial review of the manuscript. My sense is that these considerations are likely to render the manuscript challenging to distill for the general reader.

We made an utmost effort to satisfy the reviewers’ comments in the initial review. We included additional experiments including activity titrations of labeled termination factors with respect to unlabeled factors (Figure 2—figure supplement 1), provided mass spec analysis of labeled termination factors (Figure 2—figure supplement 2), and performed additional FRET measurements using dyes bound to tRNAfMet and protein L1 (Figure 3B, E and Figure 3—figure supplement 1). The detailed statistics is provided in Supplementary file 1. We also rearranged the text and added citations to support our finding that termination complex can enter R/hybrid-like states in the presence of termination factors. Disregarding all this work of the 1st revision does not seem fair, particularly for the issues of statistics and photobleaching (see answers above).

One of the points raised by the initial reviewers is that significant complexities in the interpretation of the data presented arise from the use of ribosome complexes bearing short peptide mimics (fMET-Phe, fMET, NAc-Phe), which allows the ribosome to fluctuate between classical and hybrid states in the absence (and presence) of RFs. Although the authors choose to reference their own cryo-EM work indicating that the small subunit spontaneously and reversibly rotates at elevated temperatures, multiple studies prior and subsequent to their chosen reference have indicated hybrid-like states can be achieved with acylated-tRNA in the P site and that such hybrid-like configurations are sensitive to the nature and length of the nascent peptide (see Cornish et al., 2008; Cornish et al., 2009; Munro et al., EMBO 2008; Alejo et al., PNAS 2017). Referring to such states as "Hybrid" is inappropriate as structural data indicate that this includes A76 interactions with the C2394 region.

We thank the reviewer for the support of the notion that short peptidyl-tRNAs can adopt an R state. One reason for citing the cryo-EM work is that it directly shows the distribution of the rotated states at elevated temperatures. We included the references suggested by the reviewer and used a more appropriate term “hybrid-like” instead of “hybrid”. We note, however, that we included those citations that report on subunit rotation with peptidyl-tRNA, rather than with deacylated tRNA (Munro et al., 2010) which is not disputed and is not relevant in the given context (subsection “RF1 and RF2 have distinct effects on ribosome dynamics”).

The average reader is likely to be confused by this potentially non-physiological aspect of the termination studies presented. As the initial reviewers suggested, had the study been performed with longer nascent chains, as is expected for the physiological substrate of release factors during translation termination, this complexity could likely have been entirely avoided. The fact that it is difficult to prepare such complexes (as the authors indicate in their response to reviewers) is fully appreciated, but a focused study on the propensity of the chosen complexes to achieve hybrid configurations and the nuanced impacts of such dynamics on release factor binding could be seen as an appropriate prerequisite for the present studies and their interpretation of the results presented.

The use of short model substrates is a standard approach in the field, as essentially every group working on termination uses this experimental system (Koutmou et al., 2014; Sternberg et al., 2009; Shi and Joseph, 2016; Casey et al., 2018). In that sense, we provide the data that can be compared with the bulk of previous work. Importantly, the current model of termination is based on very similar short peptidyl-tRNAs, so the ability to compare to the previous data is essential. There is no indication so far that these complexes are in any sense non-physiological. Pre-hydrolysis complexes with short peptidyl-tRNAs can bind termination factors, hydrolyze peptidyl-tRNA and proceed to recycling and thus the major features should represent termination complexes regardless of the peptidyl length. The potential effect of longer side chains for the dynamics of the PreHC is now indicated in the last paragraph of the Discussion.

Another question is which conclusions may depend on the hypothetical (albeit not demonstrated) effect of longer nascent chains. The dissociation rates of each factor alone (Figures 2 and 3) are the same on Pre and PostHC and thus an effect of a shorter or longer peptide chain is rather unlikely. In the presence of RF1 or RF2 the ribosomes favor the N state; this preference is unlikely to change with longer peptides, as it is assumed that they limit ribosome dynamics to the N state. The extent of ribosome rotation in the presence of RF3 may be affected and we added a sentence in the text to acknowledge this (Discussion), but RF3 dissociation rate is again independent of the Pre- or PostHC state, which makes the effect of chain length unlikely. For the experiments with RF1 and RF3 bound to the ribosome together, the nuances of rotation are not critical, as we interpret the tendencies by comparing the complexes with RF1 or RF3 alone with the complex with two factors. As to the PostHC, which constitute 2/3 of complexes used in this work, the peptide length is not important as it is already released. Thus, we think that the potential effect of the longer peptides is not fundamental, although we introduced the requested note of caution and the explanation as to why the work with natural long peptidyl-tRNA is not feasible (Discussion).

For example, one of the key takeaways from the present study from the perspective of the general reader is that RF1 and RF2 perform differently in termination based on their measured off rates. Yet, such differences may simply arise from small distinctions in the binding energies of RF1 and RF2 to the ribosomes, which reveal themselves as potentially meaningful in the context of the non-physiological substrates examined. Such binding idiosyncrasies may be further exacerbated by the GGQ mutation that is used to prevent peptide release (as seems to be evident when comparing Figure 1—figure supplement 2C and E). What is the argument that these distinctions are physiologically important? Focused studies on the binding differences of RF1 and RF2 (concrete measurements of affinity, on and off rates, beyond the information that is presented) and clarification as to why such differences are important in the context of RF3 functions seems relevant to discuss but is presently lacking.

Our experiments show clear differences in the binding properties of RF1 and RF2. We show that the factors differ in the way they affect the rotational dynamics of the ribosome and respond to the presence of RF3 and have very different residence times on the ribosome. This clearly shows that RF1 and RF2 interact with the termination complexes in a different way, which is consistent with crystal structures (Introduction, fourth paragraph). To make this statement, we compare three different conditions:

i) With RF1(GAQ) or RF2(GAQ) bound to PreHC where peptide release is prevented by mutation of the GGQ motif.

ii) With RF1 or RF2 bound to PostHC where the peptide is released using puromycin.

iii) With wt RF1 or RF2 with PostHC that is formed by natural peptide release.

The differences between the RF1 and RF2 (i.e. different residence times and different subunit rotation patterns) is obvious at all conditions tested. The finding that the dissociation rates of RF1 and RF2 differ also on PostHC (where no peptide is present at all) shows that the difference between RF1 and RF2 is not dependent on peptide length. The observed difference in the subunit rotation measured at saturating conditions of the factors (referred to in Figure 1) cannot be simply due to affinity differences between RF1 and RF2 because the complex is saturated with the respective factor (Figure 2—figure supplement 1).

Regarding the biochemical properties of RF1(GAQ) and RF2(GAQ) mutants, it is known that the affinity to the ribosome is not affected by the mutation (Zavialov et al., 2002). We have a full biochemical and ensemble kinetics analysis that shows that their binding properties are identical to those of the respective wild type proteins, but we are reluctant to include yet another layer of data into the manuscript, as all referees agreed that the manuscript is heavily loaded already. The fact that the GAQ mutants behave similarly to the wild type proteins on PostHC (Figure 1 and Figure 1—figure supplement 2) and that binding of RF3 to PostHC with RF1 or RF1(GAQ) results in identical changes in FRET value and dissociation rate (Figure 4) shows that the mutation does not induce a peculiar behavior of RF1.

More readily addressable issues of concern include the following:

1) The use of language that could be considered inaccurate, misleading or too interpretive.

For instance, in the Introduction it states, "An alternative model was proposed, when it turned out that the affinity of RF3 to GDP and GDP […]" Aren't these simply different measurements? As written it implies that the prior studies were just wrong and the studies by Koutmou and Peske are just right. How are the authors so sure that this is the case? Was the prior work retracted due to a technical error?

The KD values presented by Zavialov et al., 2001 were obtained by nitrocellulose filtration quantifying the amount of radiolabeled nucleotide retained on the filter. NC filtration is a non-equilibrium method, which is sometimes unreliable and may underestimate the affinity for labile complexes. The papers that have used equilibrium methods (Koutmou et al., 2014 and Peske et al., 2014) provide more reliable estimates and explain the caveats of the nitrocellulose filtration; we do not think it is necessary to reiterate the published analysis. The issues with the NC filtration technique are also addressed in detail in Wilden et al., 2006 for EF-G. In that sense the choice of language was not interpretative, it summarized the existing publications. Nevertheless, we changed the sentence to avoid the statement disliked by the reviewer.

It is written in the Introduction, "thus the lifetime of the apo-RF3 complex would be too short to assume a tentative physiological role". The work cited seems to make a thermodynamic rather than a kinetic argument and the statement as written seems to create a false context.

The manuscript “Timing of GTP binding and hydrolysis by translation termination factor RF3” by Peske et al. studies both kinetic and thermodynamic aspects of nucleotide binding and exchange of RF3. Thus, there is no false context created, the argument is kinetic, as it is based on the reported rate constant of GTP binding to the nucleotide-free RF3.

In the Results its states that "the exact distribution of states depends on P-site tRNA" referencing Fei et al., 2011, when prior literature on the classical-to-hybrid equilibrium (N to R transition) were prior to report this finding. The authors should include references to Cornish et al., 2008; Cornish et al., 2009 and Munro et al., 2010 and Munro et al., 2010, which are already included in the bibliography.

We replaced Fei et al. with Cornish et al. 2008 which is the only paper addressing the extent of subunit rotation depending on the P site substrate.

The discussion in the subsection “RF1 and RF2 have distinct

effects on ribosome dynamics” regarding Figure 1—figure supplement 3 seem to suggest that there is information about factor binding in the data presented but this figure is confusing. The authors want to interpret differences between RF1 and RF2 'contour' plots that are post-synchronized as differences in binding and unbinding, but direct information about binding and unbinding is not present in these data. These types of plots may indicate differences in dynamics within the bound complexes and this needs to be clarified. While I have no doubt that there may be binding and unbinding information underlying the differences observed, the language used in this section is too strong. The authors should also try to put sufficient information in the figure legends to enable one to discern the relevant information about the experiment but looking at the figure legend alone. From the legend of Figure 1—figure supplement 3 it is impossible to know what ribosome complex was being examined (where does the FRET come from) without going back to the main text.

We present rotation data alongside RF1/RF2 binding data (Figure 1—figure supplement 3, Figure 2). The two types of experiments complement each other and relate the rotation data to the presence of RF1/RF2 on the ribosome. Changes in subunit rotation provide a readout for factor binding. We rephrased the main text to a more careful statement. We extended the legend of Figure 1—figure supplement 3 such that the experimental procedure becomes clear.

In the results it states that "the P/E hybrid state is short lived and decays rapidly (koff = 6 per second)…" As this is an experiment done in the presence of RF3, the rate constant koff seems to imply unbinding of RF3, whereas I think the authors mean the classical to hybrid transition.

It also states that "suggesting the two processes are not tightly coupled, consistent with previous notions based on cryo-EM[…] and smFRET work" While subtle, as stated this sentence implies that the cryoEM work came first and the smFRET work came later. But the opposite is true.

We changed the text und use kclosed→open to indicate transition rates between the P/P and P/E-like state. We also changed the order of cryo-EM and smFRET as pointed out by the reviewer.

In reference to Figures 4C, G and K it is stated that "The FRET efficiency for the RF3-L11 pair is closer to 0.5, changed from 0.7, which is observed in the absence of RF1 (Figure 3C). Thus, the orientation of RF3 in the complex differs from that formed in the absence of RF1" While such conclusions may indeed be correct, statistical analyses on one dimensional histograms comprised of biological repeats are needed to make this conclusion without a second structural perspective to support this interpretation. The difference is not obvious and could be the result of day-to-day variabilities in microscope performance and/or fluorophore behaviors.

We added one-dimensional histograms to illustrate FRET changes in response to factor binding together with mean FRET values ± sd in figure legends (Figure 2, 3, 4, 5, 6, Figure 1—figure supplement 3, Figure 6—figure supplement 2). Statistical analysis is also provided in Supplementary file 1. All histograms presented in this paper contain data from at least three individual datasets. Thus, day-to-day variabilities in microscope performance and/or fluorophore behaviors are already taken into consideration and integrated in the standard deviation of the measurements.

In reference to Figure 4 it is stated that "Thus, RF1 and RF3 can reside simultaneously on the Post HC in an arrangement where the position of RF1 and RF3 relative to L11 is shifted compared to the complex with only a single factor". Here, it seems too strong to make such statements without direct evidence of RF1 and RF3 binding.

We politely disagree. The biophysical properties (FRET efficiency with respect to L11 and the residence time) of RF1 and RF3 are very different when only one factor is bound compared to when they are present together. The change in the biophysical properties of RF1/RF3 can only result from the interaction with the respective other factor and therefore RF1 and RF3 must reside simultaneously on the ribosome. These changes can be used as a readout for factor binding, direct observation of both factors is not required to justify this statement.

Reviewer #3:

The manuscript by Adio et al. explores the mechanism of release factor recycling in bacteria using single-molecule FRET approach. The authors took advantage of different labeling strategies to look at ribosome, tRNA and release factor conformations at different stages of the termination process. These were used to draw two main conclusions: 1) Unlike RF1, RF2 can dissociate efficiently from the ribosome in the absence of RF3 (although RF3 helps); 2) The order of binding and dissociation events appear to be stochastic and unlike other translational GTPases, GTP hydrolysis by RF3 serves as a rescue mechanism for the factor.

Overall the paper was well written and the experiments appear to have been well executed. I was not however convinced that the paper offered significant new insights into the mechanism of translation termination. The notion that RF1 and RF2 are slightly different is not new (for example ArfA rescue of ribosomes only works in the presence of RF2 and not RF1).

The new mechanistic insights are the following:

1) RF1 and RF2 behave quite differently during canonical termination.

2) RF3–GTP can dissociate quickly from pre- or post-termination complexes without GTP hydrolysis.

3) Binding of RF3 to ribosome–RF1 complexes alters the positions of the two factors and stabilizes RF3 binding. The order of RF1 and RF3 dissociation appears random.

4) If RF1 is still present, dissociation of RF3 does not require GTP hydrolysis, but if RF1 dissociates first, RF3 dissociation is blocked if GTP hydrolysis cannot occur.

We suggest a model for translation termination which differs in all key points from current termination models and show how release factors affect each other’s function and ribosome dynamics.

Furthermore the authors played down the effect of RF3 on RF2 (the fold change in termination rates under substoichiometric conditions is similar to that observed for RF1). Whether these differences also apply to other peptidyl tRNAs was not explored.

We did not play down the effect of RF3 on RF2, as it is shown in Figure 2—figure supplement 3, where we report that even catalytic amounts of RF2 are sufficient to complete peptide release from PreHC. RF3 accelerates the reaction by a factor of 10. This effect is clearly much smaller than the effect of RF3 on RF1, which cannot recycle at all in the absence of RF3. Studying other tRNAs is well outside the scope of the paper, as other referees pointed out that it is already overloaded with experimental data.

As for the second main point of the paper, the Ruben group has also looked at the dynamics of the tRNAs and ribosome during recycling (albeit not to the same extent in this paper) and similar observations were made.

We refer to the work of the Gonzalez lab (Sternberg et al., 2009) on ribosome dynamics during termination. Indeed, some of the conclusions are in very good agreement with our results, such as the effect of RF3 on the L1 conformation of PostHC. However, Sternberg at al. did not compare the function of RF1 and RF2. Furthermore, they do not explicitly investigate the interaction of release factors with PreHC. As a result, they come to a different and less detailed mechanistic model. Our work complements the work presented by Sternberg et al. and provides new mechanistic insight into the importance of ribosome dynamics in general and in particular for the termination process.

As for the presentation of the data, I felt that the results were over interpreted and the main conclusion was to a certain extent based on conjecture. For instance, the idea that the process of recycling is stochastic is based on the observation that many of the rates being similar (RF1, RF3 dissociation for example), but these experiments were conducted in the absence of peptide release (either using post hydrolysis complex or pre-hydrolysis one with GAQ RF1). It's unclear whether peptide release plays a role in the conformational changes. This is relevant, as under normal conditions release factors must bind in the presence of a peptidyl-tRNA and promote peptide release before it is recycled. At the end, I was left wondering how the different assays recapitulate what happens under normal conditions.

The condition of active release is always included in our experiments.

To address the role of the nascent peptide for the dissociation of factors from termination complexes we compared three different conditions:

i) With RF1(GAQ) bound to PreHC where peptide release is prevented by mutation of the GGQ motif (Figure 1B-C, Figure 4A-D).

ii) With the same RF1(GAQ) bound to PostHC* where the peptide was pre-released using puromycin (Figure 1—figure supplement 2, Figure 4E-H).

iii) With wild type RF1 with PostHC that is formed by natural peptide release (Figure 1E-F, Figure 4I-L).

At all conditions, the rate of RF3 dissociation is the same, which shows that the interaction of RF3 with termination complexes in the presence of RF1 is independent of peptide release, subunit rotation and GTP hydrolysis. In contrast, RF1 dissociation is strongly dependent on peptide release. The rates of subunit rotation are faster than factor dissociation at all conditions. This highlights the stochastic aspect of RF3 interaction with termination complexes. Our model differs significantly from the linear termination models suggested so far and sheds light on how termination factors interact with each other and with the ribosome. We hope that there points became clearer in the revised version of the manuscript.

Another major point is the authors' assertion that GTP hydrolysis by RF3 merely serves to rescue the factor from nonproductive binding events. The data is based on the observation that in the presence of excess RF3, GTP hydrolysis in not required for RF1 turnover. The data in Figure 6A then suggests that RF3 is as likely to bind in a nonproductive fashion as to a productive one.

We changed Figure 6 and hope that this makes it clearer. Figure 6A and B show (biochemically) that RF1 recycling is blocked when GTP hydrolysis is abolished and RF3 is in sub-stoichiometric concentrations, but is efficient when RF3 is present in excess. We then show that RF3–GDPNP can bind to termination complexes with and without RF1 (Figure 6C, D). Finally, we show that the dissociation of RF3–GDPNP is slow in the absence and rapid in the presence of RF1. We simplified the text accordingly (subsection “The role of GTP binding and hydrolysis”) and added a more detailed discussion of these data in the Discussion. We never mention the term “non-productive binding” as we do not have an observable for that.

Given the overall contribution of the paper to the field together with the less than ideal interpretation of the data, my overall enthusiasm of the paper was tempered.

We hope that the text changes introduced upon revision have answered the main concerns of the reviewer and present our data and interpretation in a clearer way.

https://doi.org/10.7554/eLife.34252.023

Article and author information

Author details

  1. Sarah Adio

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Conceptualization, Formal analysis, Supervision, Validation, Investigation, Methodology, Writing—original draft, Writing—review and editing
    For correspondence
    Sarah.Adio@mpibpc.mpg.de
    Competing interests
    No competing interests declared
  2. Heena Sharma

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Resources, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Tamara Senyushkina

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Software, Formal analysis, Validation, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Prajwal Karki

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Resources, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6187-6506
  5. Cristina Maracci

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Conceptualization, Supervision, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3810-7180
  6. Ingo Wohlgemuth

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  7. Wolf Holtkamp

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Resources, Investigation, Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  8. Frank Peske

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Conceptualization, Supervision, Writing—review and editing
    Competing interests
    No competing interests declared
  9. Marina V Rodnina

    Department of Physical Biochemistry, Max Planck Institute for Biophysical Chemistry, Göttingen, Germany
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing
    For correspondence
    rodnina@mpibpc.mpg.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0105-3879

Funding

Deutsche Forschungsgemeinschaft (SFB 860)

  • Sarah Adio
  • Marina V Rodnina

Max-Planck-Institute for Biophysical Chemistry (Open-access funding)

  • Marina V Rodnina

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Marija Liutkute for preparation of RF2-Cy5 and Olaf Geintzer, Franziska Hummel, Sandra Kappler, Christina Kothe, Anna Pfeifer, Theresia Uhlendorf, Tanja Wiles, and Michael Zimmermann for expert technical assistance. We thank Dr. H Urlaub and the bioanalytical mass spectrometry facility for assistance with the analysis of RF1/RF2 methylation. The work was supported by the grants of the Deutsche Forschungsgemeinschaft (SFB860 for MVR and SA).

Reviewing Editor

  1. Rachel Green, Johns Hopkins School of Medicine, United States

Publication history

  1. Received: December 11, 2017
  2. Accepted: May 11, 2018
  3. Version of Record published: June 11, 2018 (version 1)

Copyright

© 2018, Adio et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,589
    Page views
  • 264
    Downloads
  • 15
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry and Chemical Biology
    2. Microbiology and Infectious Disease
    Qi Yang et al.
    Research Article Updated

    The Spike protein of SARS-CoV-2, its receptor-binding domain (RBD), and its primary receptor ACE2 are extensively glycosylated. The impact of this post-translational modification on viral entry is yet unestablished. We expressed different glycoforms of the Spike-protein and ACE2 in CRISPR-Cas9 glycoengineered cells, and developed corresponding SARS-CoV-2 pseudovirus. We observed that N- and O-glycans had only minor contribution to Spike-ACE2 binding. However, these carbohydrates played a major role in regulating viral entry. Blocking N-glycan biosynthesis at the oligomannose stage using both genetic approaches and the small molecule kifunensine dramatically reduced viral entry into ACE2 expressing HEK293T cells. Blocking O-glycan elaboration also partially blocked viral entry. Mechanistic studies suggest multiple roles for glycans during viral entry. Among them, inhibition of N-glycan biosynthesis enhanced Spike-protein proteolysis. This could reduce RBD presentation on virus, lowering binding to host ACE2 and decreasing viral entry. Overall, chemical inhibitors of glycosylation may be evaluated for COVID-19.

    1. Biochemistry and Chemical Biology
    Rajaa Boujemaa-Paterski et al.
    Research Article Updated

    Vinculin plays a fundamental role in integrin-mediated cell adhesion. Activated by talin, it interacts with diverse adhesome components, enabling mechanical coupling between the actin cytoskeleton and the extracellular matrix. Here we studied the interactions of activated full-length vinculin with actin and the way it regulates the organization and dynamics of the Arp2/3 complex-mediated branched actin network. Through a combination of surface patterning and light microscopy experiments we show that vinculin can bundle dendritic actin networks through rapid binding and filament crosslinking. We show that vinculin promotes stable but flexible actin bundles having a mixed-polarity organization, as confirmed by cryo-electron tomography. Adhesion-like synthetic design of vinculin activation by surface-bound talin revealed that clustered vinculin can initiate and immobilize bundles from mobile Arp2/3-branched networks. Our results provide a molecular basis for coordinate actin bundle formation at nascent adhesions.