Insights into Genome Recoding from the Mechanism of a Classic +1-Frameshifting tRNA 1 Insights into Genome Recoding 1 from the Mechanism of a Classic +1-Frameshifting tRNA 2 3 4 Howard Gamper1,5, Haixing Li2,5, Isao Masuda1, D. Miklos Robkis3, Thomas Christian1, 5 Adam B. Conn4, Gregor Blaha4, E. James Petersson3, Ruben L. Gonzalez, Jr2,#, 6 and Ya-Ming Hou1,#,* 7 8 9 1Department of Biochemistry and Molecular Biology, Thomas Jefferson University, 10 Philadelphia, PA 19107, USA 11 2Department of Chemistry, Columbia University, New York, NY 10027, USA 12 3Department of Chemistry, University of Pennsylvania, Philadelphia, PA 19104, USA 13 4Department of Biochemistry, University of California, Riverside, CA 92521, USA 14 5These authors contributed equally to this work. 15 #Corresponding authors: 16 rlg2118@columbia.edu (T) 212-854-1096; (F) 212-932-1289; ORCID: 0000-0002-1344-5581 17 ya-ming.hou@jefferson.edu (T) 215-503-4480; (F) 215-503-4954; 18 ORCID: 0000-0001-6546-2597 19 20 *Lead contact: Ya-Ming Hou (ya-ming.hou@jefferson.edu) 21 22 Running Title: Mechanism of SufB2-induced +1 frameshifting 23 24 25 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 2 ABSTRACT 26 While genome recoding using quadruplet codons to incorporate non-proteinogenic amino 27 acids is attractive for biotechnology and bioengineering purposes, the mechanism through which 28 such codons are translated is poorly understood. Here we investigate translation of quadruplet 29 codons by a +1-frameshifting tRNA, SufB2, that contains an extra nucleotide in its anticodon loop. 30 Natural post-transcriptional modification of SufB2 in cells prevents it from frameshifting using a 31 quadruplet-pairing mechanism such that it preferentially employs a triplet-slippage mechanism. 32 We show that SufB2 uses triplet anticodon-codon pairing in the 0-frame to initially decode the 33 quadruplet codon, but subsequently shifts to the +1-frame during tRNA-mRNA translocation. 34 SufB2 frameshifting involves perturbation of an essential ribosome conformational change that 35 facilitates tRNA-mRNA movements at a late stage of the translocation reaction. Our results 36 provide a molecular mechanism for SufB2-induced +1 frameshifting and suggest that engineering 37 of a specific ribosome conformational change can improve the efficiency of genome recoding. 38 39 Key words: SufB2 frameshift suppressor tRNA, +1 ribosomal frameshifting, quadruplet codon, 40 genome expansion, m1G37 methylation 41 42 43 44 45 46 47 48 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 3 INTRODUCTION 49 The ability to recode the genome and expand the chemical repertoire of proteins to include 50 non-proteinogenic amino acids promises novel tools for probing protein structure and function. 51 While most recoding employs stop codons as sites for incorporating non-proteinogenic amino 52 acids, only two stop codons can be simultaneously recoded due to the cellular need to reserve 53 the third stop codon for termination of protein synthesis. The use of quadruplet codons as 54 additional sites for incorporating non-proteinogenic amino acids has thus emerged as an attractive 55 alternative1,2. Recoding at a quadruplet codon requires a +1-frameshifting tRNA that is 56 aminoacylated with the non-proteinogenic amino acid of interest. The primary challenge faced by 57 this technology has been the low efficiency with which the full-length protein carrying the non-58 proteinogenic amino acid can be synthesized. One reason for this is the poor recoding efficiency 59 of the +1-frameshifting aminoacyl (aa)-tRNA, and the second is the failure of the +1-frameshifting 60 aa-tRNA to compete with canonical aa-tRNAs that read the first three nucleotides of the 61 quadruplet codon at the ribosomal aa-tRNA binding (A) site during the aa-tRNA selection step of 62 the translation elongation cycle. While directed evolution by synthetic biologists has yielded +1-63 frameshifting tRNAs, efficient recoding requires cell lines that have been engineered to deplete 64 potential competitor tRNAs3-8. These problems emphasize the need to better understand the 65 mechanism through which quadruplet codons are translated by +1-frameshifting tRNAs. 66 In bacteria, +1-frameshifting tRNAs that suppress single-nucleotide insertion mutations that 67 shift the translational reading frame to the +1-frame have been isolated from genetic studies9,10. 68 These +1-frameshifting tRNAs typically contain an extra nucleotide in the anticodon loop – a 69 property that has led to the proposal of two competing models for their mechanism of action. In 70 the quadruplet-pairing model, the inserted nucleotide joins the triplet anticodon in pairing with the 71 quadruplet codon in the A site and this quadruplet anticodon-codon pair is translocated to the 72 ribosomal peptidyl-tRNA binding (P) site11. In the triplet-slippage model, the expanded anticodon 73 loop forms an in-frame (0-frame) triplet anticodon-codon pair in the A site and subsequently shifts 74 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 4 to the +1-frame at some point later in the elongation cycle12,13, possibly during translocation of the 75 +1-frameshifting tRNA from the A to P sites14 or within the P site15. The triplet-slippage model is 76 supported by structural studies of ribosomal complexes in which the expanded anticodon-stem-77 loops (ASLs) of +1-frameshifting tRNAs have been found to use triplet anticodon-codon pairing 78 in the 0-frame at the A site16-18 and in the +1-frame at the P site19. Nonetheless, these structures 79 do not eliminate the possibility that two competing triplet pairing schemes (0-frame and +1-frame) 80 can co-exist when a quadruplet codon motif occupies the A site15, that some amount of +1 81 frameshifting can occur via the quadruplet-pairing model, and that the quadruplet-pairing model 82 may even dominate for particular +1-frameshifting tRNAs, codon sequences, and/or reaction 83 conditions10. We also do not know how each model determines the efficiency of +1 frameshifting 84 or whether any competition between the two models is driven by the kinetics of frameshifting or 85 the thermodynamics of base pairing. In addition, virtually all natural tRNAs contain a purine at 86 nucleotide position 37 on the 3'-side of the anticodon (http://trna.bioinf.uni-leipzig.de/), which is 87 invariably post-transcriptionally modified and is important for maintaining the translational reading 88 frame in the P site15. While most +1-frameshifting tRNAs sequenced to date also contain a purine 89 nucleotide at position 378, we do not know whether it is post-transcriptionally modified or how the 90 modification affects +1 frameshifting. Perhaps most importantly, while the structural studies 91 described above provide snapshots of the initial and final states of +1 frameshifting, they do not 92 reveal where, when, or how the shift occurs, thereby precluding an understanding of the structural 93 basis and mechanism of +1 frameshifting. These open questions have limited our ability to 94 increase the efficiency of genome recoding at quadruplet codons. 95 To address these questions, we have investigated the mechanism of +1 frameshifting by 96 SufB2 (Figure 1a), a +1-frameshifting tRNA that was isolated from Salmonella typhimurium as a 97 suppressor of a single C insertion into a proline (Pro) CCC codon20. The observed high +1-98 frameshifting efficiency of SufB2 at the CCC-C motif, nearly 80-fold above background20, 99 demonstrates its ability to successfully compete with the naturally occurring ProL and ProM 100 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 5 isoacceptor tRNAs that read the CCC codon. Using the ensemble ‘codon-walk’ methodology21 101 and single-molecule fluorescence resonance energy transfer (smFRET), we have compared the 102 +1 frameshifting activity of SufB2 relative to its closest counterpart, ProL, at a CCC-C motif, and 103 determined the position and timing of the shift. Our results show that SufB2 is naturally N1-104 methylated at G37 in cells, generating an m1G37 that blocks quadruplet pairing and forces SufB2 105 to use 0-frame triplet anticodon-codon pairing to decode the quadruplet codon at the A site. 106 Additionally, we find that SufB2, and likely all +1-frameshifting tRNAs, shifts to the +1-frame during 107 the subsequent translocation reaction in which the translational GTPase elongation factor (EF)-G 108 catalyzes the movement of SufB2 from the A to P sites (i.e., a triplet-slippage mechanism). More 109 specifically, we show that this frameshift occurs in the later steps of translocation, during which 110 EF-G catalyzes a series of conformational rearrangements of the ribosomal pre-translocation 111 (PRE) complex that enable the tRNA ASLs and their associated codons to move to their 112 respective post-translocation positions within the ribosomal small (30S in bacteria) subunit22-28. 113 Thus, efforts to increase the recoding efficiency of +1-frameshifting tRNAs should focus on 114 enforcing a triplet anticodon-codon pairing in the 0-frame at the A site and directed evolution to 115 optimize conformational rearrangements of the ribosomal PRE complex during the late stages of 116 translocation. 117 118 RESULTS 119 Native-state SufB2 is N1-methylated at G37 and is readily aminoacylated with Pro 120 SufB2 contains an extra G37a nucleotide inserted between G37 and U38 of ProL20 (Figure 121 1a). Whether the extra G37a is methylated and how it affects methylation of G37 is unknown. We 122 thus determined the methylation status of the G37-G37a motif using RNase T1 cleavage inhibition 123 assays and primer extension inhibition assays. We first generated a plasmid-encoded SufB2 by 124 inserting G37a into an existing Tac-inducible plasmid encoding Escherichia coli ProL29, which has 125 an identical sequence to S. typhimurium ProL. We then expressed and purified the plasmid-126 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 6 encoded SufB2 and ProL from an E. coli ProL knock-out (ProL-KO) strain30 containing all the 127 endogenous enzymes necessary for processing SufB2 and ProL to their S. typhimurium native 128 states such that they possess the full complement of naturally occurring post-transcriptional 129 modifications (termed the native-state tRNAs). In addition, we prepared in vitro transcripts of 130 SufB2 and ProL lacking all post-transcriptional modifications (termed the G37-state tRNAs), or 131 enzymatically methylated with purified E. coli TrmD30,31 such that they possess only the N1-132 methylation at G37 and no other post-transcriptional modifications (termed the m1G37-state 133 tRNAs). In the case of SufB2, RNase T1 cleavage inhibition assays demonstrated cleavage at 134 G37 and G37a of the G37-state tRNA, but inhibition of cleavage at either position upon treatment 135 with TrmD (Figure 1b), indicating that both nucleotides are N1-methylated in the m1G37-state 136 tRNA. 137 Primer extension inhibition assays, which were previously validated by mass spectrometry 138 analysis30, showed inhibition of extension at G37 and G37a in m1G37- and native-state SufB2 139 (Figure 1c), confirming that both nucleotides are N1-methylated in these species. Notably, N1 140 methylation shifted almost entirely to G37 in native-state SufB2, indicating that m1G37 is the 141 dominant methylation product in cells. As a control, no inhibition of extension at G37 or G37a was 142 observed for G37-state SufB2. Complementary kinetics experiments showed that the yield and 143 rate of N1-methylation of G37-state SufB2 were similar to those of G37-state ProL (Figure 1d). 144 Likewise, kinetics experiments revealed that the yield and rate of aminoacylation of native-state 145 SufB2 with Pro were similar to those of native-state ProL (Figure 1e). In contrast, aminoacylation 146 of G37-state SufB2 was inhibited (Figure 1f). These results demonstrate that the native-state 147 SufB2 synthesized in cells is quantitatively N1-methylated to generate m1G37 and is readily 148 aminoacylated with Pro. 149 150 SufB2 promotes +1 frameshifting using triplet-slippage and possibly other mechanisms 151 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 7 We next determined the mechanism(s) through which SufB2 promotes +1 frameshifting in a 152 cellular context. We created a pair of isogenic E. coli strains expressing SufB2 or ProL from the 153 chromosome in a trmD-knockdown (trmD-KD) background30. This background strain was 154 designed to evaluate the effect of m1G37 on +1 frameshifting and it was generated by deleting 155 chromosomal trmD and controlling cellular levels of m1G37 using arabinose-induced expression 156 of the human counterpart trm5, which is competent to stoichiometrically N1-methylate intracellular 157 tRNA substrates30. The isogenic pair of the SufB2 and ProL strains were measured for +1 158 frameshifting in a cell-based lacZ reporter assay in which a CCC-C motif was inserted into the 2nd 159 codon position of lacZ such that a +1-frameshifting event at the motif was necessary to synthesize 160 full-length b-galactosidase (b-Gal)29. The efficiency of +1 frameshifting was calculated as the ratio 161 of b-Gal expressed in cells containing the CCC-C insertion relative to cells containing an in-frame 162 CCC insertion. 163 In the m1G37-abundant (m1G37+) condition, SufB2 displayed a high +1-frameshifting 164 efficiency (8.2%, Figure 2a) relative to ProL (1.4%). In the m1G37-deficient (m1G37–) condition, 165 SufB2 exhibited an even higher efficiency (20.8%) and, consistent with our previous work29, ProL 166 also displayed an increased efficiency (7.0%) relative to background (1.4%). Because N1-167 methylation in the m1G37+ condition was stoichiometric (Figure 1c), thereby preventing 168 quadruplet-pairing, we attribute the 8.2% efficiency of SufB2 in this condition as arising exclusively 169 from triplet-slippage. In the m1G37– condition, we observed an increase in +1-frameshifting 170 efficiency of SufB2 to 20.8%. While multiple mechanisms may exist for the increased +1 171 frameshifting, the exploration of both triplet-slippage and quadruplet-pairing is one possibility. 172 To confirm our results, we performed similar studies with the isogenic SufB2 and ProL strains 173 on the endogenous E. coli lolB gene, encoding the outer membrane lipoprotein. The lolB gene 174 naturally contains a CCC-C motif at the 2nd codon position such that +1 frameshifting at this motif 175 would decrease protein synthesis due to premature termination. As a reference, we used E. coli 176 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 8 cysS, encoding cysteinyl-tRNA synthetase (CysRS)30, which has no CCC-C motif in the first 16 177 codons and would be less sensitive to +1 frameshifting at CCC-C motifs during protein synthesis. 178 The ratio of protein synthesis of lolB to cysS for the control sample ProL in the m1G37 condition, 179 measured from Western blots (Methods), was normalized to 1.00, denoting that lolB and cysS 180 were maximally translated in the 0-frame without +1 frameshifting (i.e., a relative +1 frameshifting 181 efficiency of 0.00) (Figures 2b, 2c). In the m1G37+ condition, SufB2 displayed a ratio of LolB to 182 CysRS of 0.62, indicating an increase in the relative +1 frameshifting efficiency to 0.38, and in the 183 m1G37– condition, it displayed a ratio of 0.17, indicating an increase in the relative +1 184 frameshifting efficiency to 0.83 (Figures 2b, 2c). Similarly, ProL in the m1G37– condition displayed 185 a ratio of LolB to CysRS of 0.47, indicating an increase in the +1-frameshifting efficiency to 0.53. 186 187 SufB2 can insert non-proteinogenic amino acids at CCC-C motifs 188 We next asked whether SufB2 can deliver non-proteinogenic amino acids to the ribosome by 189 inducing +1 frameshifting at a CCC-C motif (Figure 2d). We inserted a CCC-C motif at the 5th 190 codon position of the E. coli folA gene, encoding dihydrofolate reductase (DHFR). A SufB2-191 induced +1 frameshifting event at the insertion would result in full-length DHFR, whereas the 192 absence of +1 frameshifting would result in a C-terminal truncated DHFR fragment (DC). SufB2 193 was aminoacylated with non-proteinogenic amino acids using a Flexizyme32 and subsequently 194 tested in [35S]-Met-dependent in vitro translation reactions using the E. coli PURExpress system. 195 The resulting protein products were separated by sodium dodecyl sulfate (SDS)-polyacrylamide 196 gel electrophoresis and quantified by phosphorimaging. Control experiments with no SufB2 or 197 with a non-acylated SufB2 showed no full-length DHFR, demonstrating that synthesis of full-198 length DHFR depended upon SufB2 delivery of an amino acid as a result of +1 frameshifting at 199 the CCC-C motif. We showed that SufB2 was able to deliver Pro, Arg, Val, and the Pro analogs 200 cis-hydroxypro, trans-hydroxypro, azetidine, and thiapro (Supplementary Figure 1) to the 201 ribosome in response to the CCC-C motif, and that the efficiency of delivery by G37-state SufB2 202 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 9 was generally higher than that by native-state SufB2. Notably, the PURExpress system contains 203 all canonical tRNAs, including ProL and ProM, indicating the ability of SufB2 to successfully 204 compete with these tRNAs. 205 206 SufB2 uses triplet pairing in the 0-frame at the A site 207 To determine at which step in the elongation cycle SufB2 undergoes +1 frameshifting in 208 response to a CCC-C motif, we used an E. coli in vitro translation system composed of purified 209 components and supplemented with requisite tRNAs and translation factors to perform a series 210 of ensemble rapid kinetic studies. We began with a GTPase assay that reports on the yield and 211 rate with which the translational GTPase EF-Tu hydrolyzes GTP upon delivery of a ternary 212 complex (TC), composed of EF-Tu, [g-32P]-GTP, and prolyl-SufB2 (SufB2-TC) or ProL (ProL-TC), 213 to the A site of a ribosomal 70S initiation complex (70S IC) carrying an initiator fMet-tRNAfMet in 214 the P site and a programmed CCC-C motif at the A site. The results of these experiments showed 215 that the yield and rate of GTP hydrolysis (kGTP,obs) upon delivery of SufB2-TC were quantitatively 216 similar to those of ProL-TC for both the native- and G37-state tRNAs (Figure 3a). 217 We next performed a dipeptide formation assay that reports on the synthesis of a peptide 218 bond between the [35S]-fMet moiety of a P-site [35S]-fMet-tRNAfMet in a 70S IC and the Pro moiety 219 of a SufB2- or ProL-TC delivered to the A site. This assay revealed that the rate of [35S]-fMet-Pro 220 (fMP) formation (kfMP,obs) for SufB2-TC was within 2-fold of that for ProL-TC for both the native- 221 and G37-state tRNAs (Figure 3b, Table S2). 222 To test whether native-state SufB2-TC can effectively compete with ProL-TC for delivery to 223 the A site and peptide-bond formation, we varied the dipeptide formation assay such that an 224 equimolar mixture of each TC was used in the reaction (Figure 3c). Since aminoacylation of both 225 tRNAs with Pro would create dipeptides of the same identity (i.e., fMP), we used a Flexizyme to 226 aminoacylate them with different amino acids and generate distinct dipeptides. Control 227 experiments showed that ProL charged with Pro or Arg (Figure 3c, Bars 1 and 2) and SufB2 228 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 10 charged with Pro or Arg (Bars 3 and 4) generated the same amount of fMP and fMR, indicating 229 that the amino-acid identity did not affect the level of dipeptide formation. We found that the 230 amount of dipeptide formed by SufB2-TC and ProL-TC in these competition assays was similar, 231 although the amount formed by SufB2-TC was slightly less (45% vs. 55%), in both the native- 232 (Bars 5-8) and G37-state tRNAs (Supplementary Figure 2a). These competition experiments 233 provide direct evidence that SufB2-TC effectively competes with ProL-TC for delivery to the A site 234 and peptide-bond formation. 235 Collectively, the results of our GTPase-, dipeptide formation-, and competition assays indicate 236 that SufB2-TC is delivered to the A site and participates in peptide-bond formation in the same 237 way as ProL-TC, suggesting that SufB2 uses triplet pairing in the 0-frame at the A site that 238 successfully competes with triplet pairing by ProL. To support this interpretation, we measured 239 kfMP,obs in our dipeptide formation assay, using G37-state SufB2-TC and a series of mRNA variants 240 in which single nucleotides in the CCC-C motif were substituted. We showed that kfMP,obs did not 241 decrease upon substitution of the 4th nucleotide of the CCC-C motif, but that it decreased 242 substantially upon substitution of any of the first three nucleotides of the motif (Figure 3d, 243 Supplementary Figure 2b). Thus, triplet pairing of SufB2 to the first three Cs of the CCC-C motif 244 is necessary and sufficient for rapid delivery of the tRNA to the A site and its participation in 245 peptide-bond formation. 246 247 The A-site activity of SufB2 depends on the sequence of the anticodon loop 248 We next asked how delivery of SufB2-TC to the A site and peptide-bond formation depend on 249 the sequence of the SufB2 anticodon loop. Starting from G37-state SufB2, we created two 250 variants containing a G-to-C substitution in nucleotide 37 (G37C) or 34 (G34C) within the 251 anticodon loop and adapted our dipeptide formation assay to measure the fMP yield and kfMP,obs 252 generated by each variant at the CCC-C motif at the A site. We showed that the G37C variant 253 resulted in a fMP yield of 32% and a kfMP,obs of 0.14 ± 0.01 s–1, most likely by triplet pairing of 254 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 11 nucleotides 34-36 of the anticodon loop with the 0-frame of the CCC-C motif (Figure 4a). In 255 contrast, the G34C variant resulted in a fMP yield of 30% and a kfMP,obs of 0.28 ± 0.04 s–1, most 256 likely by triplet pairing of nucleotides 35-37 of the anticodon loop with the 0-frame of the CCC-C 257 motif (Figure 4b). Our interpretation that nucleotides 35-37 of the anticodon loop of the G34C 258 variant most likely triplet pair with the 0-frame of the CCC-C motif is consistent with the 259 observations that the fMP yield and kfMP,obs of the G34C variant are similar and 2-fold higher, 260 respectively, than those of the G37C variant. If nucleotides 34-36 of the anticodon loop of the 261 G34C variant were to form a triplet pair with the CCC-C motif, we would have expected it to pair 262 in the +2-frame, which would have most likely reduced the fMP yield and kfMP,obs of the G34C 263 variant relative to the G37C variant. These results suggest that G37-state SufB2 exhibits some 264 plasticity as to whether it can undergo triplet pairing with anticodon loop nucleotides 34-36 or 35-265 37, consistent with a previous study33. 266 267 SufB2 shifts to the +1-frame during translocation 268 Although SufB2 uses triplet pairing in the 0-frame when it is delivered to the A site, it is a 269 highly efficient +1-frameshifting tRNA (Figure 2). We therefore asked whether +1 frameshifting 270 occurs during or after translocation of SufB2 into the P site. We addressed this question by 271 adapting our previously developed tripeptide formation assays29. We rapidly delivered EF-G and 272 an equimolar mixture of G37-state SufB2-, tRNAVal-, and tRNAArg-TCs to 70S ICs assembled on 273 an mRNA in which the 2nd codon was a CCC-C motif and the 3rd codon was either a GUU codon 274 encoding Val in the +1 frame or a CGU codon encoding Arg in the 0-frame. As soon as 275 translocation of the PRE complex and the associated movement of SufB2 from the P to A sites 276 formed a ribosomal post-translocation (POST) complex with an empty A site in these experiments, 277 tRNAVal- and tRNAArg-TC would compete for the codon at the A site to promote formation of an 278 fMPV tripeptide or an fMPR tripeptide. Thus, the fMPV yield and kfMPV,obs report on the sub-279 population of SufB2 that shifted to the +1-frame, whereas the fMPR yield and kfMPR,obs report on 280 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 12 the sub-population that remained in the 0-frame29,34. The results showed that the yield of fMPV 281 was much higher than that of fMPR (90% vs. 10%, Figure 5a), demonstrating the high efficiency 282 with which G37-state SufB2 induces +1 frameshifting. Notably, relative to the +1 frameshifting of 283 ProL we have previously reported29, kfMPV,obs of SufB2 (0.09 s–1) was comparable to the rate of +1 284 frameshifting of ProL during translocation (0.1 s–1) rather than that of +1 frameshifting after 285 translocation into the P site (~10–3 s–1)29, indicating that SufB2 underwent +1 frameshifting during 286 translocation. Our observation that the fMPV yield plateaus at 90% at long reaction times 287 suggests that the sub-populations of SufB2 that will shift to the +1-frame and remain in the 0-288 frame are likely established in the A site, even before EF-G binds to the PRE complex. Given that 289 SufB2 exhibits triplet pairing in the 0-frame at the A site (Figures 3a-c, Supplementary Table 2, 290 and Supplementary Figure 2a) and shifts into the +1-frame during translocation (Figure 5a), the 291 two sub-populations of SufB2 in the A site seem to differ primarily in their propensity to undergo 292 +1 frameshifting during translocation. The sub-population that encompasses 90% of the total 293 would exhibit a high propensity of undergoing +1 frameshifting during translocation, whereas the 294 sub-population that encompasses 10% of the total would exhibit a low propensity of undergoing 295 +1 frameshifting during translocation, preferring instead to remain in the 0-frame. 296 We next determined whether the 10% sub-population of G37-state SufB2 that remained in the 297 0-frame during translocation could undergo +1 frameshifting after arrival at the P site. We varied 298 our tripeptide formation assay so as to deliver the TCs in two steps separated by a defined time 299 interval (Figure 5b). In the first step, G37-state SufB2-TC and EF-G were delivered to the 70S IC 300 to form a POST complex, which was then allowed the opportunity to shift to the +1-frame over a 301 systematically increasing time interval. In the second step, an equimolar mixture of tRNAArg- and 302 tRNAVal-TCs was delivered to the POST complex. The results showed that fMPV was rapidly 303 formed at a high yield and exhibited a kfMP+V,obs (where the “+” denotes the time interval between 304 the delivery of translation components) that did not increase as a function of time. In contrast, 305 fMPR was formed at a low yield and exhibited a kfMP+R,obs that did not decrease as a function of 306 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 13 time. Together, these results indicate that the sub-population of P site-bound SufB2 in the 0-frame 307 does not undergo +1 frameshifting. This interpretation is supported by the observation that EF-P, 308 an elongation factor which we showed suppresses +1 frameshifting within the P site29, had no 309 effect on the yield of fMPV yield (Supplementary Figure 2c and Supplementary Table 3). 310 Having shown that +1 frameshifting of SufB2 occurs only during translocation, we evaluated 311 the effect of m1G37 on the frequency of this event. We began by delivering G37-, m1G37-, or 312 native-state SufB2-TCs together with EF-G to 70S ICs to form the corresponding POST 313 complexes and then delivered an equimolar mixture of tRNAArg- and tRNAVal-TCs to each POST 314 complex to determine the relative formation of fMPV and fMPR. The results showed that m1G37- 315 and native-state SufB2 displayed a reduced fMPV yield and a concomitantly increased fMPR yield 316 relative to G37-state SufB2 (Figures 5c, Supplementary Figures 2d-f), consistent with the notion 317 that the presence of m1G37 compromises +1 frameshifting. 318 We then used the same tripeptide formation assay to determine how +1 frameshifting during 319 translocation of G37-state SufB2 depends on the identity of the 4th nucleotide of the CCC-C motif. 320 A series of POST complexes were generated by delivering G37-state SufB2-TCs and EF-G to 321 70S ICs programmed with a CCC-N motif at the 2nd codon position. Each POST complex was 322 then rapidly mixed with tRNAVal-TC to monitor the yield of fMPV and kfMP+V,obs (Figure 5d). The 323 results showed a high fMPV yield and high kfMP+V,obs at the CCC-[C/U] motifs, but a low yield and 324 low kfMP+V,obs at the CCC-[A/G] motifs. This indicates that high-efficiency of SufB2-induced +1 325 frameshifting during translocation requires the presence of a [C/U] at the 4th nucleotide of the 326 CCC-C motif. Because SufB2 in these experiments was in the G37-state, it is possible that a sub-327 population underwent +1 frameshifting via quadruplet-pairing with the [C/U] at the 4th nucleotide 328 of the CCC-[C/U] motif during translocation. It is also possible that a sub-population underwent 329 +1 frameshifting via triplet-slippage, which could potentially be inhibited by the presence of [G/A] 330 at the 4th nucleotide of the motif. To verify that the POST complex formed with the CCC-A 331 sequence was largely in the 0-frame, we rapidly mixed the complex with an equimolar mixture of 332 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 14 tRNASer-TC, cognate to the next A-site codon in the 0-frame (AGU), and tRNAVal-TC, cognate to 333 the next A-site codon in the +1-frame (GUU) (Figure 5e). The results showed a high yield and 334 high kfMP+S,obs, supporting the notion that the POST complex formed with the CCC-A motif was 335 largely in the 0-frame. Thus, the 4th nucleotide of the CCC-C motif plays a role in determining +1 336 frameshifting during translocation of SufB2 from the A site to the P site. 337 338 The +1-frameshifting efficiency of SufB2 depends on sequences of the anticodon loop and 339 the CCC-C motif 340 To determine whether the +1-frameshifting efficiency of SufB2 during translocation is influenced 341 by sequences of the anticodon loop and the CCC-C motif, we performed tripeptide formation 342 assays and monitored the yield of fMPV. In these experiments, we varied the sequence of the 343 SufB2 anticodon loop and/or the CCC-C motif at the 2nd codon position of the mRNA. To explore 344 the possibilities of both triplet-slippage and quadruplet-pairing, we used variants of G37-state 345 SufB2. We showed that variants with the potential to undergo quadruplet-pairing with the CCC-C 346 motif resulted in fMPV yields of 87% and 62% (Figures 4c, d). The different yields suggest that 347 G37-state SufB2 variants can induce triplet-slippage and/or engage in quadruplet-pairing with 348 different efficiencies during translocation. Analogous experiments showed that SufB2 variants 349 that were restricted to triplet-pairing resulted in reduced fMPV yields (26% and 20%, respectively) 350 upon pairing with a CCC-C motif (Figures 4e, f). Collectively, these results suggest that there is 351 considerable plasticity in the mechanisms that SufB2 uses to induce +1 frameshifting during 352 translocation and in the efficiencies of these mechanisms. 353 354 An smFRET signal that reports on ribosome dynamics during individual elongation cycles 355 To address the mechanism of SufB2-induced +1 frameshifting during translocation, we used 356 a previously developed smFRET signal to determine whether and how SufB2 alters the rates with 357 which the ribosome undergoes a series of conformational changes that drive and regulate the 358 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 15 elongation cycle35 (Figures 6a-c). This signal is generated using a ribosomal large, or 50S, subunit 359 that has been Cy3- and Cy5-labeled at ribosomal proteins bL9 and uL1, respectively, to report on 360 ‘opening’ and ‘closing’ of the L1 stalk of the 50S subunit. Accordingly, individual FRET efficiency 361 (EFRET) vs. time trajectories recorded using this signal exhibit transitions between two FRET states 362 corresponding to the ‘open’ (EFRET = ~0.55) and ‘closed’ (EFRET = ~0.31) conformations of the L1 363 stalk (Figure 6d). 364 Previously, we have shown that open→closed and closed→open L1 stalk transitions correlate 365 with a complex series of conformational changes that take place during an elongation cycle35-37. 366 The L1 stalk initially occupies the open conformation as an aa-tRNA is delivered to the A site of 367 a 70S IC or POST complex and peptide-bond formation generates a PRE complex that is in a 368 global conformation we refer to as global state (GS) 1. The PRE complex then undergoes a large-369 scale structural rearrangement that includes an open→closed transition of the L1 stalk so as to 370 occupy a second global conformation we refer to as GS2 (i.e., the 0.55→0.31 EFRET transition 371 denoted by the rate k70S IC→GS2 in Figures 6d and e, corresponding to the multi-step 70S IC→GS2 372 transition in Figure 6a). Subsequently, in the absence of EF-G, the L1 stalk goes through 373 successive closed→open and open→closed transitions as the PRE complex undergoes multiple 374 GS2→GS1 and GS1→GS2 transitions that establish a GS1⇄GS2 equilibrium (i.e., the 0.55⇄0.31 375 EFRET transitions denoted by the rates kGS1→GS2 and kGS2→GS1 and the equilibrium constant Keq = 376 (kGS1→GS2)/(kGS2→GS1) in Figure 6d, corresponding to the GS1⇄GS2 transitions in Figure 6a). In the 377 presence of EF-G, however, a single closed→open L1 stalk transition reports on conformational 378 changes of the PRE complex as it undergoes EF-G binding and completes translocation (i.e., the 379 0.31→0.55 EFRET transition denoted by the rate kGS2→POST in Figures 6d and e, corresponding to 380 the multi-step GS2→POST transition that takes place in the presence of EF-G and bridges across 381 Figures 6a and b). Using this approach, we have successfully monitored the conformational 382 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 16 dynamics of ribosomal complexes during individual elongation cycles36,38-41, including in a study 383 of –1 frameshifting41. 384 385 SufB2 interferes with elongation complex dynamics during late steps in translocation 386 We began by asking whether SufB2 alters the dynamics of elongation complexes during the 387 earlier steps of the elongation cycle. We stopped-flow delivered SufB2- or ProL-TC to 70S ICs 388 and recorded pre-steady-state movies during delivery, and steady-state movies 1 min post-389 delivery (Figures 6a, d, and f, Supplementary Figures 3, 4a, and 4b). The results showed that k70S 390 IC→GS2, as well as kGS1→GS2, kGS2→GS1, and Keq at 1 min post-delivery, for SufB2-TC were each less 391 than 2-fold different than the corresponding value for ProL-TC (Supplementary Table 4). The 392 close correspondence of these rates indicates that SufB2-TC is delivered to the A site, 393 participates in peptide-bond formation, undergoes GS2 formation, and exhibits GS1→GS2 and 394 GS2→GS1 transitions within the GS1⇄GS2 equilibrium in a manner that is similar to ProL-TC, 395 consistent with the results of ensemble kinetic assays (Figures 3a-c, Supplementary Table 2, and 396 Supplementary Figure 2a) and thereby strengthening our interpretation that SufB2 uses triplet 397 pairing in the 0-frame at the A site during the early stages of the elongation cycle that precede 398 EF-G binding and EF-G-catalyzed translocation. Although we could not confidently detect the 399 presence of two sub-populations of A site-bound SufB2 in the smFRET data that might differ in 400 their propensity of undergoing +1 frameshifting, as suggested by the results presented in Figure 401 5a, it is possible that the distance between our smFRET probes and/or the time spent in one of 402 the observed FRET states are not sensitive enough to detect the structural and/or energetic 403 differences between these sub-populations of A site-bound SufB2. The development of different 404 smFRET signals and/or the use of variants of SufB2 and/or the CCC-C motif with different 405 propensities of undergoing +1 frameshifting may allow future smFRET investigations to identify 406 and characterize such sub-populations. 407 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 17 We then investigated whether SufB2 alters the dynamics of elongation complexes during the 408 later steps of the elongation cycle. We stopped-flow delivered SufB2- or ProL-TC and EF-G to 409 70S ICs and recorded pre-steady-state movies during delivery, and steady-state movies 1, 3, 10, 410 and 20 min post-delivery (Figures 6b, e, and g, Supplementary Figures 4c, 4d, and 5). The results 411 showed that k70S IC→GS2 for SufB2 and ProL-TC were within error of each other (Supplementary 412 Table 5), again suggesting that SufB2-TC is delivered to the A site, participates in peptide-bond 413 formation, and undergoes GS2 formation in a manner that is similar to ProL-TC. Notably, the k70S 414 IC→GS2s obtained in the presence of EF-G were within error of the ones obtained in the absence of 415 EF-G, consistent with reports that EF-G has little to no effect on the rate with which PRE 416 complexes undergo GS1→GS2 transitions37,42. 417 Once it transitions into GS2, however, the SufB2 PRE complex can bind EF-G37,42 and we find 418 that it becomes arrested in an EF-G-bound GS2-like conformation for up to several minutes, 419 during which it slowly undergoes a GS2→POST transition (Figure 6g, Supplementary Figure 5). 420 While the limited number of time points did not allow rigorous determination of kGS2→POST for the 421 SufB2 PRE complex, visual inspection (Figure 6g) and quantitative analysis (Supplementary 422 Tables 5 and 6) showed that the GS2→POST reaction was complete between 3 and 10 min post-423 delivery (i.e., kGS2→POST = ~0.0017–0.0060 s–1). Remarkably, this range of kGS2→POST is up to 2-3 424 orders of magnitude lower than kGS2→POST measured for the ProL PRE complex (Supplementary 425 Table 5). It is also up to 2-3 orders of magnitude lower than kGS2→POST for a different PRE complex 426 measured using a different smFRET signal under the same conditions43 and the rate of 427 translocation measured using ensemble rapid kinetic approaches under similar conditions44,45. 428 This observation suggests that SufB2 adopts a conformation within the EF-G-bound PRE complex 429 that significantly impedes conformational rearrangements of the complex that are known to take 430 place during late steps in translocation. These rearrangements include the severing of interactions 431 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 18 between the decoding center of the 30S subunit and the anticodon-codon duplex in the A site22-432 25; forward and reverse swiveling of the ‘head’ domain of the 30S subunit27,28 associated with 433 opening and closing, respectively, of the ‘E-site gate’ of the 30S subunit26; reverse relative rotation 434 of the ribosomal subunits46,47; and opening of the L1 stalk35,37,48. Collectively, these dynamics 435 facilitate movement of the tRNA ASLs and their associated codons from the P and A sites to the 436 E and P sites of the 30S subunit. 437 We next explored whether SufB2 alters the dynamics of elongation complexes after it is 438 translocated into the P site. We prepared PRE-like complexes carrying deacylated SufB2 or ProL 439 in the P site and a vacant A site (denoted PRE–A complexes) and recorded steady-state movies 440 for the resulting GS1⇄GS2 equilibria (Figures 6c and h, Supplementary Figure 6). The results 441 showed that kGS1→GS2 and kGS2→GS1 for the SufB2 PRE–A complex were 45% lower and 36% higher, 442 respectively, than for the ProL PRE–A complex, driving a 2.5-fold shift towards GS1 in the 443 GS1⇄GS2 equilibrium (Supplementary Table 7), suggesting that SufB2 adopts a conformation at 444 the P site that is different from that of ProL. Consistent with this interpretation, a recent structural 445 study has shown that the conformation of P site-bound SufA6, a +1-frameshifting tRNA with an 446 extra nucleotide in the anticodon loop, is significantly distorted relative to a canonical tRNA49. 447 448 DISCUSSION 449 Here we leverage the high efficiency of recoding by SufB2 to identify the steps of the 450 elongation cycle during which it induces +1 frameshifting at a quadruplet codon, thus answering 451 the key questions of where, when, and how +1 frameshifting occurs. We are not aware of any 452 other studies of +1 frameshifting that have addressed these questions as precisely. In addition to 453 elucidating the determinants of reading-frame maintenance and the mechanisms of SufB2-454 induced +1 frameshifting, our findings reveal new principles that can be used to engineer genome 455 recoding with higher efficiencies. 456 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 19 Integrating our results with the available structural, biophysical, and biochemical data on the 457 mechanism of translation elongation results in the structure-based model for SufB2-induced +1 458 frameshifting that we present in Figure 7. In this model, POST complexes to which SufB2 or ProL 459 are delivered exhibit virtually indistinguishable conformational dynamics in the early steps of the 460 elongation cycle, up to and including the initial GS1→GS2 transition. However, POST complexes 461 to which SufB2 is delivered exhibit a kGS2→POST that is more than an order-of-magnitude slower 462 than those to which ProL is delivered. Notably, kGS2→POST comprises a series of conformational 463 rearrangements of the EF-G-bound PRE complex that facilitate translocation of the tRNA ASLs 464 and associated codons within the 30S subunit. These rearrangements encompass the severing 465 of decoding center interactions with the anticodon-codon duplex in the A site22-25; forward and 466 reverse head swiveling27,28,50 and associated opening and closing, respectively, of the E-site 467 gate26; reverse relative rotation of the subunits46,47; and opening of the L1 stalk35,37,48 (steps PRE-468 G2 to PRE-G4, denoted with red arrows, in Figure 7). Given the importance of these 469 rearrangements in translocation of the tRNA ASLs and their associated codons within the 30S 470 subunit, we propose that SufB2-mediated perturbation of these rearrangements underlies +1 471 frameshifting. More specifically, because SufB2 does not seem to impede the reverse relative 472 rotation of the subunits or opening of the L1 stalk during the GS2→GS1 transitions within the 473 GS1⇄GS2 equilibrium in the absence of EF-G (compare kGS2→GS1 for SufB2-TC vs. ProL-TC in 474 Supplementary Table 4), it most likely interferes with the severing of decoding center interactions 475 with the anticodon-codon duplex in the A site and/or forward and/or reverse head swiveling and 476 associated opening and/or closing, respectively, of the E-site gate. The latter rearrangement is 477 particularly important for movement of the tRNA ASLs and their associated codons within the 30S 478 subunit26-28,50, suggesting that SufB2-mediated perturbation of head swiveling may make the most 479 important contribution to +1 frameshifting. Consistent with this, a recent structural study showed 480 that upon forward head swiveling, the ASLs of the P- and A-site tRNAs can disengage from their 481 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 20 associated codons and occupy positions similar to a partial +1 frameshift, even in the presence 482 of a non-frameshift suppressor tRNA in the A site and the absence of EF-G51. 483 While previous structural studies have demonstrated that +1 frameshifting tRNAs bind to the 484 A site in the 0-frame16,17,49 and to the P site in the +1-frame19, these studies lacked EF-G and the 485 observed structures were obtained by directly binding a deacylated +1 frameshifting tRNA to the 486 P site. Specifically, a +1 frameshifting peptidyl-tRNA was not translocated from the A to P sites, 487 as would be the case during an authentic translocation event. In contrast, our elucidation of the 488 +1-frameshifting mechanism was executed in the presence of EF-G and is based on extensive 489 comparison of the kinetics with which SufB2 and ProL undergo individual reactions of the 490 elongation cycle (i.e., aa-tRNA selection, peptide-bond formation, and translocation) and the 491 associated conformational rearrangements of the elongation complex. Additionally, all of our in 492 vitro biochemical assays, and most of our ensemble rapid kinetics assays were performed under 493 the conditions in which the A site is always occupied by an aa- or peptidyl-tRNA, leaving no 494 chance of a vacant A site. Therefore, the +1 frameshifting mechanism we present here is distinct 495 from that presented by Farabaugh and co-workers13, in which the ribosome is stalled due to a 496 vacant A site, thus giving the +1-frameshifting-inducing tRNA at the P site an opportunity to 497 rearrange into the +1-frame. The fact that all well-characterized +1-frameshifting tRNAs contain 498 an extra nucleotide in the anticodon loop, despite differences in their primary sequences, the 499 amino acids they carry, and whether the extra nucleotide is inserted at the 3'- or 5'-sides of the 500 anticodon, suggests that the results we report here for SufB2 are likely applicable to other +1-501 frameshifting tRNAs with an expanded anticodon loop. 502 While an expanded anticodon loop is a strong feature associated with +1 frameshifting, it is 503 not associated with –1 frameshifting, which instead is typically induced by structural barriers in 504 the mRNA that stall a translating ribosome from moving forward, thus providing the ribosome with 505 an opportunity to shift backwards in the –1 direction10,52. Given the unique role of the expanded 506 anticodon loop in +1 frameshifting, here we have identified the determinants that drive the 507 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 21 ribosome to shift in the +1 direction. We show that SufB2 exclusively uses the triplet-slippage 508 mechanism of +1 frameshifting in the m1G37+ condition, but that it explores other mechanisms 509 (e.g., quadruplet-pairing) in the m1G37– condition during translocation from the A site to the P 510 site. Under conditions that only permit the triplet-slippage mechanism (e.g., in the presence of 511 m1G37), SufB2 exhibits a relatively low +1-frameshifting efficiency of ~30%, whereas under 512 conditions that permit quadruplet-pairing during translocation (e.g., in the absence of m1G37), it 513 exhibits a relatively high +1-frameshifting efficiency of ~90% (Figures 4c-f, 5a). This feature is 514 observed in various sequence contexts. One advantage of a quadruplet-pairing mechanism 515 during translocation is that it would enhance the thermodynamic stability of anticodon-codon 516 pairing during the large EF-G-catalyzed conformational rearrangements that PRE complexes 517 undergo during translocation to form POST complexes. Nonetheless, SufB2 is naturally 518 methylated with m1G37 (Figure 1c), indicating that it makes exclusive use of the triplet-slippage 519 mechanism in vivo. This mechanism is likely also exclusively used in vivo by all other +1-520 frameshifting tRNAs that have evolved from canonical tRNAs to retain a purine at position 37, 521 which is almost universally post-transcriptionally modified to block quadruplet-pairing 522 mechanisms. 523 The key insight from this work suggests an entirely novel pathway to increase the efficiency 524 of genome recoding at quadruplet codons. While initial success in genome recoding has been 525 achieved by engineering the anticodon-codon interactions of a +1-frameshifting-inducing tRNA at 526 the A site6,53, or by engineering a new bacterial genome with a minimal set of codons for all amino 527 acids54, we suggest that efforts to engineer the ‘neck’ structural element of the 30S subunit that 528 regulates head swiveling would be as, or even more, effective. This can be achieved by screening 529 for 30S subunit variants that exhibit high +1-frameshifting efficiencies mediated by +1-530 frameshifting tRNAs at quadruplet codons while preserving 0-frame translation by canonical 531 tRNAs at triplet codons. Specifically, head swiveling is driven by the synergistic action of two 532 hinges within the 16S ribosomal RNA elements that comprise the 30S subunit neck55. Hinge 1 is 533 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 22 composed of two G-U wobble base pairs that are separated by a bulged G within helix 28 (h28), 534 while hinge 2 is composed of a GACU linker between h34 and h35/36 within a three-helical 535 junction with h38. Co-engineering these two hinges by directed evolution should identify such 30S 536 subunit variants. To complement the directed evolution approach, we suggest that our recently 537 developed time-resolved cryogenic electron microscopy (TR cryo-EM) method56,57 can be used 538 to obtain structures of SufB2 and ProL in EF-G-bound PRE complexes captured in intermediate 539 states of translocation. Such cryo-EM structures would help further define how the two hinges 540 that control head swiveling are differentially modulated during translocation of SufB2 vs. ProL to 541 provide a structure-based roadmap for engineering them. In addition, detailed comparison of such 542 structures would offer the opportunity to identify ribosomal structural elements beyond the two 543 hinges that play a role in +1 frameshifting and can thus serve as additional targets for engineering. 544 Furthermore, antibiotics that bind to the 30S subunit and act as translocation modulators can be 545 exploited to further increase the +1-frameshifting efficiency at a quadruplet codon with either 546 wildtype or highly efficient 30S subunit variants. Implemented in combination and integrated into 547 a recently described in vivo ‘designer organelle’ strategy58, these approaches should provide a 548 novel and powerful platform for increasing the efficiency of genome recoding at quadruplet codons 549 with minimal off-target effects. 550 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 23 METHODS 551 Construction of E. coli strains. E. coli strains that expressed a plasmid-borne ProL or SufB2 for 552 isolation of native-state tRNAs were made in a ProL-KO strain, which was constructed by inserting 553 the Kan-resistance (Kan-R) gene, amplified by PCR primers from pKD4, into the ProL locus of E. 554 coli BL21(DE3) using the l-Red recombination method59, followed by removal of the Kan-R gene 555 using FLP recombination30. The pKK223-SufB2 plasmid was made by site-directed mutagenesis 556 to introduce G37a into the pKK223-ProL plasmid29. E. coli strains that expressed ProL or SufB2 557 from the chromosome as an isogenic pair for reporter assays were made using the l-Red 558 technique30. To construct the E. coli SufB2 strain, the SufB2 gene was PCR-amplified from 559 pKK223-SufB2, and the 5' end of the amplified gene was joined with Kan-R (from pKD4) by PCR 560 using reverse-2 primer, while the 3' end was homologous to the ProL 3' flanking region. The PCR-561 amplified SufB2-Kan product was used to replace ProL in l-Red expressing cells. An isogenic 562 counterpart strain expressing ProL-Kan was also made. These ProL-Kan and SufB2-Kan loci 563 were independently transferred to the trmD-KO strain29 by P1 transduction, followed by pCP20-564 dependent FLP recombination, generating the isogenic pair of ProL and SufB2 strains in the trmD-565 KO background. These strains were transformed with pKK223-3-lacZ reporter plasmid that has 566 the CCC-C motif at the 2nd codon position of the lacZ gene, and the b-Gal activity was measured29. 567 All primer sequences used in this work are shown in Supplementary Table 1. 568 569 Preparation of translation components for ensemble biochemical experiments. The mRNA 570 used for most in vitro translation reactions is shown below, including the Shine-Dalgarno 571 sequence, the AUG start codon, and the CCC-C motif: 572 5'-GGGAAGGAGGUAAAAAUGCCCCGUUCUAAG(CAC)7. 573 Variants of this mRNA had a base substitution in the CCC-C motif. All mRNAs were transcribed 574 from double-stranded DNA templates with T7 RNA polymerase and purified by gel 575 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 24 electrophoresis. E. coli strains over-expressing native-state tRNAfMet, tRNAArg (anticodon ICG, 576 where I = inosine), and tRNAVal (anticodon U*AC, where U* = cmo5U) were grown to saturation 577 and were used to isolate total tRNA. The over-expressed tRNA species in each total tRNA sample 578 was aminoacylated by the cognate aminoacyl-tRNA synthetase and used directly in the TC 579 formation reaction and subsequent TC delivery to 70S ICs or POST complexes. E. coli tRNASer 580 (anticodon ACU) was prepared by in vitro transcription. Aminoacyl-tRNAs with the cognate 581 proteinogenic amino acid were prepared using the respective aminoacyl-tRNA synthetase and 582 those with a non-proteinogenic amino acid were prepared using the dFx Flexizyme and the 3,5-583 dinitobenzyl ester (DBE) of the respective amino acid (Supplementary Figure 1). Aminoacylation 584 and formylation of tRNAfMet were performed in a one-step reaction in which formyl transferase and 585 the methyl donor 10-formyltetrahydrofolate were added to the aminoacylation reaction29. 586 Aminoacyl-tRNAs were stored in 25 mM sodium acetate (NaOAc) (pH 5) at –70 °C, as were six-587 His-tagged E. coli initiation and elongation factors and tight-coupled 70S ribosomes isolated from 588 E. coli MRE600 cells. Recombinant His-tagged E. coli EF-P bearing a b-lysyl-K34 was expressed 589 and purified from cells co-expressing efp, yjeA, and yjeK and stored at –20 °C29. 590 591 Preparation of translation components for smFRET experiments. 30S subunits and 50S 592 subunits lacking ribosomal proteins bL9 and uL1 were purified from a previously described bL9-593 uL1 double deletion E. coli strain35,60 using previously described protocols35,37,60. A previously 594 described single-cysteine variant of bL9 carrying a Gln-to-Cys substitution mutation at residue 595 position 18 (bL9(Q18C))35 and a previously described single-cysteine variant of uL1 carrying a 596 Thr-to-Cys substitution mutation at residue position 202 (uL1(T202C))35,37 were purified, labeled 597 with Cy3- and Cy5-maleimide, respectively, to generate bL9(Cy3) and uL1(Cy5), and 598 reconstituted into the 50S subunits lacking bL9 and uL1 following previously described 599 protocols35. The reconstituted bL9(Cy3)- and uL1(Cy5)-labeled 50S subunits were then re-purified 600 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 25 using sucrose density gradient ultracentrifugation35,43. 50S subunits lacking bL9(Cy3) and/or 601 uL1(Cy5) or harboring unlabeled bL9 and/or uL1 do not generate bL9(Cy3)-uL1(Cy5) smFRET 602 signals and therefore do not affect data collection or analysis. Previously, we have shown that 603 70S ICs formed with these bL9(Cy3)- and uL1(Cy5)-containing 50S subunits can undergo 604 peptide-bond formation and two rounds of translocation elongation with similar efficiency as 70S 605 ICs formed with wild-type 50S subunits35. 606 The sequence of the mRNA used for assembling ribosomal complexes for smFRET studies 607 is shown below, including the Shine-Dalgarno sequence, the AUG start codon, and the CCC-C 608 motif: 609 5'-GCAACCUAAAACUCACACAGGGCCCUAAGGACAUAAAAAUGCCCCGUU 610 AUCCUCCUGCUGCACUCGCUGCACAAAUCGCUCAACGGCAAUUAAGGA. 611 The mRNA was synthesized by in vitro transcription using T7 RNA polymerase, and then 612 hybridized to a previously described 3’-biotinylated DNA oligonucleotide (Supplementary Table 613 1) that was complementary to the 5' end of the mRNA and was chemically synthesized by 614 Integrated DNA Technologies60. Hybridized mRNA:DNA-biotin complexes were stored in 10 mM 615 Tris-OAc (pH = 7.5 at 37 ºC), 1 mM EDTA, and 10 mM KCl at –80 ºC until they were used in 616 ribosomal complex assembly. Aminoacylation and formylation of tRNAfMet (purchased from MP 617 Biomedicals) was achieved simultaneously using E. coli methionyl-tRNA synthetase and E. coli 618 formylmethionyl-tRNA formyltransferase60. Expression and purification of IF1, IF2, IF3, EF-Tu, 619 EF-Ts, and EF-G were following previously published procedures60. 620 621 Preparation and purification of SufB2 and ProL. Native-state SufB2 was isolated from a 622 derivative of E. coli JM109 lacking the endogenous ProL, but expressing SufB2 from the pKK223-623 3 plasmid (Supplementary Table 1), while native-state ProL was purified from total tRNA isolated 624 from E. coli JM109 cells over-expressing ProL from the pKK223-3 plasmid. The ProL-KO strain 625 lacking the endogenous ProL was described previously30. Each native-state tRNA was isolated 626 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 26 by a biotinylated capture probe attached to streptavidin-derivatized Sepharose beads29. G37-state 627 SufB2 and ProL were also prepared by in vitro transcription. Each primary transcript contained a 628 ribozyme domain on the 5'-side of the tRNA sequence, which self-cleaved to release the tRNA. 629 m1G37-state SufB2 and ProL were prepared by TrmD-catalyzed and S-adenosyl methionine 630 (AdoMet)-dependent methylation of each G37-state tRNA. Due to the lability of the aminoacyl 631 linkage to Pro, stocks of SufB2 and ProL aminoacylated with Pro were either used immediately 632 or stored no longer than 2-3 weeks at –70 °C in 25 mM NaOAc (pH 5.0). 633 634 Primer extension inhibition assays. Primer extension inhibition analyses of native-, G37-, and 635 m1G37-state SufB2 and ProL were performed as described30. A DNA primer complementary to 636 the sequence of C41 to A57 of SufB2 and ProL was chemically synthesized, 32P-labeled at the 637 5'-end by T4 polynucleotide kinase, annealed to each tRNA, and was extended by Superscript III 638 reverse transcriptase (Invitrogen) at 200 units/µL with 6 µM each dNTP in 50 mM Tris-HCl (pH 639 8.3), 3 mM MgCl2, 75 mM KCl, and 1 mM DTT at 55 °C for 30 min, and terminated by heating at 640 70 °C for 15 min. Extension was quenched with 10 mM EDTA and products of extension were 641 separated by 12% denaturing polyacrylamide gel electrophoresis (PAGE/7M urea) and analyzed 642 by phosphorimaging. In these assays, the length of the read-through cDNA is 54-55 nucleotides, 643 as in the case of the G37-state SufB2 and ProL, whereas the length of the primer-extension 644 inhibited cDNA products is 21-22 nucleotides, as in the case of the m1G37-state and native-state. 645 646 RNase T1 cleavage inhibition assays. RNase T1 cleaves on the 3'-side of G, but not m1G. 647 Cleavage of tRNAs was performed as previously described29. Each tRNA (1 µg) was 3'-end 648 labeled using Bacillus stearothermophilus CCA-adding enzyme (10 nM) with [α-32P]ATP at 60 °C 649 in 100 mM glycine (pH 9.0) and 10 mM MgCl2. The labeled tRNA was digested by RNase T1 650 (Roche, cat # 109193) at a final concentration of 0.02 units/µL for 20 min at 50 °C in 20 mM 651 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 27 sodium citrate (pH 5.5) and 1 mM ethylene diamine tetraacetic acid (EDTA). The RNA fragments 652 generated from cleavage were separated by 12% PAGE/7M urea along with an RNA ladder 653 generated by alkali hydrolysis of the tRNA of interest. Cleavage was analyzed by 654 phosphorimaging. 655 656 Methylation assays. Pre-steady-state assays under single-turnover conditions61 were performed 657 on a rapid quench-flow apparatus (Kintek RQF-3). The tRNA substrate was heated to 85 °C for 658 2.5 min followed by addition of 10 mM MgCl2, and slowly cooled to 37 °C in 15 min. N1-methylation 659 of G37 in the pre-annealed tRNA (final concentration 1 µM) was initiated with the addition of E. 660 coli TrmD (10 µM) and [3H]-AdoMet (Perkin Elmer, 4200 DPM/pmol) at a final concentration of 15 661 µM in a buffer containing 100 mM Tris-HCl (pH 8.0), 24 mM NH4Cl, 6 mM MgCl2, 4 mM DTT, 0.1 662 mM EDTA, and 0.024 mg/mL BSA in a reaction of 30 µL. The buffer used was optimized for TrmD 663 in order to evaluate its in vitro activity61. Reaction aliquots of 5 µL were removed at various time 664 points and precipitated in 5% (w/v) trichloroacetic acid (TCA) on filter pads for 10 min twice. Filter 665 pads were washed with 95% ethanol twice, with ether once, air dried, and measured for 666 radioactivity in an LS6000 scintillation counter (Beckman). Counts were converted to pmoles 667 using the specific activity of the [3H]-AdoMet after correcting for the signal quenching by filter 668 pads. In these assays, a negative control was always included, in which no enzyme was added 669 to the reaction61, and signal from the negative control was subtracted from signal of each sample 670 for determining the level of methylation. 671 672 Aminoacylation assays. Each SufB2 or ProL tRNA was aminoacylated with Pro by a 673 recombinant E. coli ProRS expressed from the plasmid pET22 and purified from E. coli BL21 674 (DE3)62. Each tRNA was heat-denatured at 80 ºC for 3 min, and re-annealed at 37 ºC for 15 min. 675 Aminoacylation under pre-steady state conditions was performed at 37 ºC with 10 µM tRNA, 1 676 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 28 µM ProRS, and 15 µM [3H]-Pro (Perkin Elmer, 7.5 Ci/mmol) in a buffer containing 20 mM KCl, 10 677 mM MgCl2, 4 mM dithiothreitol (DTT), 0.2 mg/mL bovine serum albumin (BSA), 2 mM ATP (pH 678 8.0), and 50 mM Tris-HCl (pH 7.5) in a reaction of 30 µL. Reaction aliquots of 5 µL were removed 679 at different time intervals and precipitated with 5% (w/v) TCA on filter pads for 10 min twice. Filter 680 pads were washed with 95% ethanol twice, with ether once, air dried, and measured for 681 radioactivity in an LS6000 scintillation counter (Beckman). Counts were converted to pmoles 682 using the specific activity of the [3H]-Pro after correcting for signal quenching by filter pads. 683 684 Cell-based +1-frameshifting reporter assays. Isogenic E. coli strains expressing chromosomal 685 copies of SufB2 or ProL were created in a previously developed trmD-knockdown (trmD-KD) 686 background, in which the chromosomal trmD is deleted but cell viability is maintained through the 687 arabinose-induced expression of a plasmid-borne trm5, the human counterpart of trmD29,30 that is 688 competent for m1G37 synthesis to support bacterial growth (Supplementary Table 1). Due to the 689 essentiality of trmD for cell growth, a simple knock-out cannot be made. We chose human Trm5 690 as the maintenance protein in the trmD-KD background, because this enzyme is rapidly degraded 691 in E. coli once its expression is turned off to allow immediate arrest of m1G37 synthesis. In the 692 isogenic SufB2 and ProL strains, the level of m1G37 is determined by the concentration of the 693 added arabinose in a cellular context that expresses ProM as the only competing tRNAPro species. 694 In the m1G37+ condition, where arabinose was added to 0.2% in the medium, tRNA substrates 695 of N1-methylation were confirmed to be 100% methylated by mass spectrometry, whereas in the 696 m1G37– condition, where arabinose was not added to the medium, tRNA substrates of N1-697 methylation were confirmed to be 20% methylated by mass spectrometry30. Each strain was 698 transformed with the pKK223-3 plasmid expressing an mRNA with a CCC-C motif at the 2nd codon 699 position of the reporter lacZ gene. To simplify the interpretation, the natural AUG codon at the 5th 700 position of lacZ was removed. A +1 frameshift at the CCC-C motif would enable expression of 701 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 29 lacZ. The activity of b-Gal was directly measured from lysates of cells grown in the presence or 702 absence of 0.2% arabinose to induce or not induce, respectively, the plasmid-borne human trm5. 703 In these assays, decoding of the CCC-C codon motif would be mediated by SufB2 and ProM in 704 the SufB2 strain, and would be mediated by ProL and ProM in the ProL strain. Due to the presence 705 of ProM in both strains, there would be no vacancy at the CCC-C codon motif. 706 707 Cell-based +1 frameshifting lolB assays. To quantify the +1-frameshifting efficiency at the 708 CCC-C motif at the 2nd codon position of the natural lolB gene, the ratio of protein synthesis of 709 lolB to cysS was measured by Western blots. Overnight cultures of the isogenic strains expressing 710 SufB2 or ProL were separately inoculated into fresh LB media in the presence or absence of 0.2% 711 arabinose and were grown for 4 h to produce the m1G37+ and m1G37– conditions, respectively. 712 Cultures were diluted 10- to 16-fold into fresh media to an optical density (OD) of ~0.1 and grown 713 for another 3 h. Cells were harvested and 15 µg of total protein from cell lysates was separated 714 on 12% SDS-PAGE and probed with rabbit polyclonal primary antibodies against LolB (at a 715 10,000 dilution) and against CysRS (at a 20,000 dilution), followed by goat polyclonal anti-rabbit 716 IgG secondary antibody (Sigma-Aldrich, #A0545). The ratio of protein synthesis of lolB to cysS 717 was quantified using Super Signal West Pico Chemiluminescent substrate (Thermo Fischer) in a 718 Chemi-Doc XR imager (Bio-Rad) and analyzed by Image Lab software (Bio-Rad, SOFT-LIT-170-719 9690-ILSPC-V-6-1). To measure the +1-frameshifting efficiency, we measured the ratio of protein 720 synthesis of lolB to cysS for each tRNA in each condition, and we normalized the observed ratio 721 in the control sample (i.e., ProL in the m1G37+ condition) to 1.0, indicating that protein synthesis 722 of these two genes was in the 0-frame and no +1 frameshifting. A decrease of this ratio was 723 interpreted as a proxy of +1 frameshifting at the CCC-C motif at the 2nd codon position of lolB. 724 From the observed ratio of each sample in each condition, we calculated the +1 frameshifting 725 efficiency relative to the control sample. 726 727 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 30 Cell-free PURExpress in vitro translation assays. The folA gene, provided as part of the E. 728 coli PURExpress (New England BioLabs) in vitro translation system, was modified by site-directed 729 mutagenesis to introduce a CCC-C motif into the 5th codon position. If SufB2 induced +1 730 frameshifting at this motif, a full-length DHFR would be made, whereas if SufB2 failed to do so, a 731 C-terminal truncated fragment (DC) would be made due to premature termination of protein 732 synthesis. Because SufB2 has no orthogonal tRNA synthetase for aminoacylation with a non-733 proteinogenic amino acid, we used the Flexizyme ribozyme technology32 for this purpose. 734 Coupled in vitro transcription-translation of the modified E. coli folA gene containing the CCC-C 735 motif at the 5th codon position was conducted in the presence of [35S]-Met using the PURExpress 736 system. SDS-PAGE analysis was used to detect [35S]-Met-labeled polypeptides, which included 737 the full-length DHFR, the DC fragment, and a DN fragment that likely resulted from initiation of 738 translation at a cryptic site downstream from the CCC-C motif (Figure 2d). The fraction of the full-739 length folA gene product, the DC fragment, and the DN fragment was calculated from the amount 740 of each in the sum of all three products. We attribute the overall low recoding efficiency (0.5 – 741 5.0%) as arising from a combination of the rapid hydrolysis of the prolyl linkage, which is the least 742 stable among aminoacyl linkages63, and the lack of SufB2 re-acylation in the PURExpress system. 743 In these assays, each tRNA was tested in the G37-state and each was normalized by the 744 flexizyme aminoacylation efficiency, which was ~30% for Pro and Pro analogues. The 745 PURExpress contained all natural E. coli tRNAs, such that the CCC-C codon motif would not have 746 a chance of vacancy even when a specific CCC-reading tRNA was absent. 747 748 Rapid kinetic GTPase assays. Ensemble GTPase assays were performed using the codon-walk 749 approach, in which an E. coli in vitro translation system composed of purified components is 750 supplemented with the requisite tRNAs and translation factors to interrogate individual steps of 751 the elongation cycle. Programmed with a previously validated synthetic AUG-CCC-CGU-U mRNA 752 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 31 template29,34, a 70S IC was assembled that positioned the AUG start codon and an initiator fMet-753 tRNAfMet at the P site and the CCC-C motif at the A site. Reactions to monitor the EF-Tu-754 dependent hydrolysis of GTP during delivery and accommodation of a TC to the A site were 755 conducted at 20 °C in a buffer containing 50 mM Tris-HCl (pH 7.5), 70 mM NH4Cl, 30 mM KCl, 7 756 mM MgCl2, 1 mM DTT, and 0.5 mM spermidine29. Each TC was formed by incubating EF-Tu with 757 8 nM [g-32P]-GTP (6000 Ci/mmole) for 15 min at 37 ºC, after which aminoacylated SufB2 or ProL 758 was added and the incubation continued for 15 min at 4 ºC. Unbound [g-32P]-GTP was removed 759 from the TC solution by gel filtration through a spin cartridge (CentriSpin-20; Princeton 760 Separations). Equal volumes of each purified TC and a solution of 70S ICs were rapidly mixed in 761 the RQF-3 Kintek chemical quench apparatus29. Final concentrations in these reactions were 0.5 762 µM for the 70S IC; 0.8 µM for mRNA; 0.65 µM each for IFs 1, 2, and 3; 0.65 µM for fMet-tRNAMet; 763 1.8 µM for EF-Tu; 0.4 µM for aminoacylated SufB2 or ProL; and 0.5 mM for cold GTP. The yield 764 of GTP hydrolysis and kGTP,obs upon rapid mixing of each TC with excess 70S ICs were measured 765 by removing aliquots of the reaction at defined time points, quenching the aliquots with 40% formic 766 acid, separating [g-32P] from [g-32P]-GTP using thin layer chromatography (TLC), and quantifying 767 the amount of each as a function of time using phosphorimaging29. We adjusted reaction 768 conditions such that the kGTP,obs increased linearly as a function of 70S IC concentration. 769 770 Rapid kinetic di- and tripeptide formation assays. Di- and tripeptide formation assays were 771 performed using the codon-walk approach described above in 50 mM Tris-HCl (pH 7.5), 70 mM 772 NH4Cl, 30 mM KCl, 3.5 mM MgCl2, 1 mM DTT, 0.5 mM spermidine, at 20 °C unless otherwise 773 indicated29. 70S ICs were formed by incubating 70S ribosomes, mRNA, [35S]-fMet-tRNAfMet, and 774 IFs 1, 2, and 3, and GTP, for 25 min at 37 °C in the reaction buffer. Separately, TCs were formed 775 in the reaction buffer by incubating EF-Tu and GTP for 15 min at 37 °C followed by adding the 776 requisite aa-tRNAs and incubating in an ice bath for 15 min. In dipeptide formation assays, 70S 777 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 32 ICs templated with the specified variants of an AUG-NNN-NGU-U mRNA were mixed with SufB2-778 TC or ProL-TC. fMP formation was monitored in an RQF-3 Kintek chemical quench apparatus. In 779 tripeptide formation assays, 70S ICs templated with the specified variants of the AUG-NCC-NGU-780 U mRNA were mixed, either in one step or in two steps, with equimolar mixtures of SufB2-, tRNAVal 781 (anticodon U*AC, where U* = cmo5U)-, and tRNAArg (anticodon ICG, where I = inosine)-TCs and 782 EF-G. Formation of fMPV and fMPR were monitored in an RQF-3 Kintek chemical quench 783 apparatus. Tripeptide formation assays with one-step delivery of TCs were initiated by rapidly 784 mixing the 70S IC with two or more of the TCs in the RQF-3 Kintek chemical quench apparatus. 785 Final concentrations in these reactions were 0.37 µM for the 70S IC; 0.5 µM for mRNA; 0.5 µM 786 each for IFs 1, 2, and 3; 0.25 µM for [35S]-fMet-tRNAfMet; 2.0 µM for EF-G; 0.75 µM for EF-Tu for 787 each aa-tRNA; 0.5 µM each for the aa-tRNAs; and 1 mM for GTP. For tripeptide formation assays 788 with one-step delivery of G37-state SufB2-, tRNAVal-, and tRNAArg-TCs to the 70S ICs, the yield 789 of fMPV and kfMPV,obs report on the activity of ribosomes that shifted to the +1-frame, whereas the 790 yield of fMPR and kfMPR,obs report on the activity of ribosomes that remained in the 0-frame29,34. 791 We chose G37-state SufB2 to maximize its +1-frameshifting efficiency but native-state tRNAVal 792 and tRNAArg to prevent them from undergoing unwanted frameshifting (note that, for simplicity, 793 we have not denoted the aminoacyl or dipeptidyl moieties of the tRNAs). Tripeptide formation 794 assays with two-step delivery of TCs29 were performed in a manner similar to those with one-step 795 delivery of TCs, except that the 70S ICs were incubated with a SufB2- or ProL-TC and 2.0 µM 796 EF-G for 0.5-10 min, as specified, followed by manual addition of an equimolar mixture of tRNAArg- 797 and tRNAVal-TCs. Reactions were conducted at 20 °C unless otherwise specified, and were 798 quenched by adding concentrated KOH to 0.5 M. After a brief incubation at 37 °C, aliquots of 0.65 799 µL were spotted onto a cellulose-backed plastic TLC sheet and electrophoresed at 1000 V in 800 PYRAC buffer (62 mM pyridine, 3.48 M acetic acid, pH 2.7) until the marker dye bromophenol 801 blue reached the water-oil interface at the anode29. The position of the origin was adjusted to 802 maximize separation of the expected oligopeptide products. The separation of unreacted [35S]-803 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 33 fMet and each of the [35S]-fMet-peptide products was visualized by phosphorimaging and 804 quantified using ImageQuant (GE Healthcare) and kinetic plots were fitted using Kaleidagraph 805 (Synergy software). 806 807 Assembly and purification of 70S ICs, TCs, POST, and PRE–A complexes for use in smFRET 808 experiments. 70S ICs were assembled in a manner analogous to those for the ensemble rapid 809 kinetic studies described above, except that the mRNA containing an AUG-CCC-CGU-U coding 810 sequence was 5'-biotinylated and the 50S subunits were labeled with bL9(Cy3) and uL1(Cy5). 811 More specifically, 70S ICs were assembled in three steps. First, 15 pmol of 30S subunits, 27 pmol 812 of IF1, 27 pmol of IF2, 27 pmol of IF3, 18 nmol of GTP, and 25 pmol of biotin-mRNA in 7 µL of 813 Tris-Polymix Buffer (50 mM Tris-(hydroxymethyl)-aminomethane acetate (Tris-OAc) (pH25°C = 814 7.0), 100 mM KCl, 5 mM NH4OAc, 0.5 mM Ca(OAc)2, 0.1 mM EDTA, 10 mM 2-mercaptoethanol 815 (BME), 5 mM putrescine dihydrochloride, and 1 mM spermidine (free base)) at 5 mM Mg(OAc)2 816 were incubated for 10 min at 37 ºC. Then 20 pmol of fMet-tRNAfMet in 2 µL of 10 mM KOAc (pH = 817 5) was added to the reaction, followed by an additional incubation of 10 min at 37 ºC. Finally, 10 818 pmol of bL9(Cy3)- and uL1(Cy5)-labeled 50S subunits in 1 µL of Reconstitution Buffer (20 mM 819 Tris-HCl (pH25°C = 7.8), 8 mM Mg(OAc)2, 150 mM NH4Cl, 0.2 mM EDTA, and 5 mM BME) was 820 added to the reaction to give a final volume of 10 µL, followed by a final incubation of 10 min at 821 37 ºC. The reaction was then adjusted to 100 µL with Tris-Polymix Buffer at 20 mM Mg(OAc)2, 822 loaded onto a 10-40% (w/v) sucrose gradient prepared in Tris-Polymix Buffer at 20 mM Mg(OAc)2, 823 and purified by sucrose density gradient ultracentrifugation to remove any free mRNA, IFs, and 824 fMet-tRNAfMet. Purified 70S ICs were aliquoted, flash frozen in liquid nitrogen, and stored at –80 825 ºC until use in smFRET experiments. 826 TCs were prepared in two steps. First, 300 pmol of EF-Tu and 200 pmol of EF-Ts in 8 µL of 827 Tris-Polymix Buffer at 5 mM Mg(OAc)2 supplemented with GTP Charging Components (1 mM 828 GTP, 3 mM phosphoenolpyruvate, and 2 units/mL pyruvate kinase) were incubated for 1 min at 829 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 34 37 ºC. Then, 30 pmol of aa-tRNA in 2 µL of 25 mM NaOAc (pH = 5) was added to the reaction, 830 followed by an additional incubation of 1 min at 37 ºC. This results in a TC solution with a final 831 volume of 10 µL that was then stored on ice until used for smFRET experiments. 832 To prepare PRE–A complexes, we first needed to assemble POST complexes. POST 833 complexes were assembled by first preparing a 10-µL solution of 70S IC and a 10-µL solution of 834 TC as described above. Separately, a solution of GTP-bound EF-G was prepared by incubating 835 120 pmol EF-G in 5 µL of Tris-Polymix Buffer at 5 mM Mg(OAc)2 supplemented with GTP 836 Charging Components for 2 min at room temperature. Then 10 µL of the 70S IC, 10 µL of the TC, 837 and 2.5 µL the GTP-bound EF-G solution were mixed, and incubated for 5 min at room 838 temperature and for additional 5 min on ice. The resulting POST complex was diluted by adjusting 839 the reaction volume to 100 µL with Tris-Polymix Buffer at 20 mM Mg(OAc)2 and purified via 840 sucrose density gradient ultracentrifugation as described above for the 70S ICs. Purified POST 841 complexes were aliquoted, flash frozen in liquid nitrogen, and stored at –80 ºC until use in 842 smFRET experiments. PRE–A complexes were then generated by mixing 3 µL of POST complex, 843 2 µL of a 10 mM puromycin solution (prepared in Nanopure water and filtered using a 0.22 µm 844 filter), and 15 µL of Tris-Polymix Buffer at 15 mM Mg(OAc)2 and incubating the mixture for 10 min 845 at room temperature. PRE–A complexes were used for smFRET experiments immediately upon 846 preparation. 847 848 smFRET imaging using total internal reflection fluorescence (TIRF) microscopy. 70S ICs or 849 PRE–A complexes were tethered to the PEG/biotin-PEG-passivated and streptavidin-derivatized 850 surface of a quartz microfluidic flowcell via a biotin-streptavidin-biotin bridge between the biotin-851 mRNA and the biotin-PEG37,43. Untethered 70S ICs or PRE–A complexes were removed from the 852 flowcell, and the flowcell was prepared for smFRET imaging experiments, by flushing it with Tris-853 Polymix Buffer at 15 mM Mg(OAc)2 supplemented with an Oxygen-Scavenging System (2.5 mM 854 protocatechuic acid (pH = 9) (Sigma Aldrich) and 250 nM protocatechuate-3,4-dioxygenase (pH 855 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 35 = 7.8) (Sigma Aldrich))64 and a Triplet-State-Quencher Cocktail (1 mM 1,3,5,7-cyclooctatetraene 856 (Aldrich) and 1 mM 3-nitrobenzyl alcohol (Fluka))65. 857 Tethered 70S ICs or PRE–A complexes were imaged at single-molecule resolution using a 858 laboratory-built, wide-field, prism-based total internal reflection fluorescence (TIRF) microscope 859 with a 532-nm, diode-pumped, solid-state laser (Laser Quantum) excitation source delivering a 860 power of 16-25 mW as measured at the prism to ensure the same power density on the imaging 861 plane. The Cy3 and Cy5 fluorescence emissions were simultaneously collected by a 1.2 862 numerical aperture, 60´, water-immersion objective (Nikon) and separated based on wavelength 863 using a two-channel, simultaneous-imaging system (Dual ViewTM, Optical Insights LLC). The Cy3 864 and Cy5 fluorescence intensities were recorded using a 1024 ´ 1024 pixel, back-illuminated 865 electron-multiplying charge-coupled-device (EMCCD) camera (Andor iXon Ultra 888) operating 866 with 2 ´ 2 pixel binning at an acquisition time of 0.1 seconds per frame controlled by software 867 μManager 1.4. This microscope allows direct visualization of thousands of individual 70S ICs or 868 PRE-A complexes in a field-of-view of 115 × 230 µm2. Each movie was composed of 600 frames 869 in order to ensure that the majority of the fluorophores in the field-of-view were photobleached 870 within the observation period. For stopped-flow experiments using tethered 70S ICs, we delivered 871 0.25 µM of G37-state SufB2- or ProL-TC in the absence of EF-G or, when specified, in the 872 presence of a 2 µM saturating concentration of EF-G. Stopped-flow experiments proceeded by 873 recording an initial pre-steady-state movie of a field-of-view that captured conformational changes 874 taking place during delivery followed by recording of one or more steady-state movies of different 875 fields-of-view that captured conformational changes taking place the specified number of minutes 876 post-delivery. 877 878 Analysis of smFRET experiments. For each TIRF microscopy movie, we identified 879 fluorophores, aligned Cy3 and Cy5 imaging channels, and generated fluorescence intensity vs. 880 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 36 time trajectories for each pair of Cy3 and Cy5 fluorophores using custom-written software 881 (manuscript in preparation; Jason Hon, Colin Kinz-Thompson, Ruben L. Gonzalez) as described 882 previously66. For each time point, Cy5 fluorescence intensity values were corrected for Cy3 883 bleedthrough by subtracting 5% of the Cy3 fluorescence intensity value in the corresponding Cy3 884 fluorescence intensity vs. time trajectory. EFRET vs. time trajectories were generated by using the 885 Cy3 fluorescence intensity (ICy3) and the bleedthrough-corrected Cy5 fluorescence intensity (ICy5) 886 from each aligned pair of Cy3 and Cy5 fluorophores to calculate the EFRET value at each time point 887 using EFRET = (ICy5 / (ICy5 + ICy3)). 888 For both pre-steady-state and steady-state movies (Figures 6d-6h and Supplementary 889 Figures 3, 5, and 6, Supplementary Tables 4-7), an EFRET vs. time trajectory was selected for 890 further analysis if all of the transitions in the fluorescence intensity vs. time trajectory were anti-891 correlated for the corresponding, aligned pair of Cy3 and Cy5 fluorophores, and the Cy3 892 fluorescence intensity vs. time trajectory underwent single-step Cy3 photobleaching, 893 demonstrating it arose from a single ribosomal complex. In the case of pre-steady-state movies 894 (Figures 6d-6g, Supplementary Figures 3 and 5 and Tables 4-6), EFRET vs. time trajectories had 895 to meet two additional criteria in order to be selected for further analysis: (i) EFRET vs. time 896 trajectories had to be stably sampling EFRET = 0.55 prior to TC delivery, thereby confirming that 897 the corresponding ribosomal complex was a 70S IC carrying an fMet-tRNAfMet at the P site and 898 (ii) EFRET vs. time trajectories had to exhibit at least one 0.55→0.31 transition after delivery of TCs, 899 thereby confirming that the corresponding 70S IC had accommodated a Pro-SufB2 or Pro-ProL 900 into the A site, that the A site-bound Pro-SufB2 or Pro-ProL had participated as the acceptor in 901 peptide-bond formation, and that the resulting PRE complex was capable of undergoing 902 GS1→GS2 transitions. We note that the second criterion might result in the exclusion of EFRET vs. 903 time trajectories in which Cy3 or Cy5 simply photobleached prior to undergoing a 0.55→0.31 904 transition, and could therefore result in a slight overestimation of k70S IC→GS2 and/or kGS1→GS2 (see 905 below for a detailed description of how k70S IC→GS2, kGS1→GS2, and other kinetic and thermodynamic 906 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 37 parameters were estimated). Nonetheless, the number of such EFRET vs. time trajectories should 907 be exceedingly small. This is because the rates with which the fluorophore that photobleached 908 the fastest, Cy5, entered into the photobleached state (Æ) from the GS1, GS2, EF-G-bound GS2-909 like, and POST states were kGS1→Æ = 0.04 ± 0.02 s–1, kGS2→Æ = 0.07 ± 0.01 s–1, kGS2(G)→Æ = 0.07 ± 910 0.01 s–1 (where the subscript “(G)” denotes experiments performed in the presence of EF-G), and 911 kPOST→Æ 0.05 ± 0.02 s–1, respectively (see below for a detailed description of how kGS1→Æ, kGS2→Æ, 912 kGS2(G)→Æ, and kPOST→Æ were estimated). These rates are, on average, about 11-fold lower than 913 those of k70S IC → GS2 and kGS1 → GS2 (0.3–0.6 s–1 and 0.58–0.82 s–1 (Supplementary Table 4)). 914 Consequently, we do not expect the measurements of k70S IC→GS2 and kGS1→GS2 to be limited by 915 Cy3 or Cy5 photobleaching. Additionally, even if k70S IC→GS2 and kGS1→GS2 were slightly 916 overestimated, they would be expected to be equally overestimated for SufB2- and ProL 917 ribosomal complexes given that the rate of photobleaching would be expected to be very similar 918 for SufB2- and ProL ribosomal complexes. Furthermore, because we are primarily concerned with 919 the relative values of k70S IC→GS2 and kGS1→GS2 for SufB2- vs. ProL ribosomal complexes, rather 920 than with the absolute values of k70S IC→GS2 and kGS1→GS2 for the SufB2- and ProL ribosomal 921 complexes, such slight overestimations do not affect the conclusions of the work presented here. 922 To calculate k70S IC→GS2 and the corresponding error from the pre-steady-state experiments, 923 we analyzed the 70S IC survival probabilities (Supplementary Figure 4, Tables 4 and 5)37,67. 924 Briefly, for each trajectory, we extracted the time interval during which we were waiting for the 925 70S IC to undergo a transition to GS2 and used these ‘waiting times’ to construct a 70S IC survival 926 probability distribution, as shown in Supplementary Figure 4. All 70S IC survival probability 927 distributions were best described by a single exponential decay function of the type 928 𝑌 = 𝐴e("#/𝜏!"# %&) , (1) 929 where Y is survival probability, A is the initial population of 70S IC, t is time, and τ70S IC is the 930 time constant with which 70S IC transitions to a PRE complex in the GS2 state. k70S IC→GS2 was 931 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 38 then calculated using the equation k70S IC→GS2 = 1 / τ70S IC. Errors were calculated as the standard 932 deviation of technical triplicates. 933 Six sets of kinetic and/or thermodynamic parameters were calculated from hidden Markov 934 model (HMM) analyses of the recorded movies. These parameters are defined here as: (i) 935 kGS1→GS2, kGS2→GS1, and Keq from the pre-steady-state and steady-state movies recorded for the 936 delivery of SufB2- and ProL-TCs in the absence of EF-G (Figures 6d, 6f, and Supplementary 937 Figure 3 and Table 4); (ii) kGS2→POST from the pre-steady-state movie recorded for the delivery of 938 ProL-TC in the presence of EF-G (Figures 6e, 6g, and Supplementary Figure 5 and Table 5); (iii) 939 the fractional population of the POST complex from the pre-steady-state and steady-state movies 940 recorded for the delivery of SufB2- and ProL-TCs in the presence of EF-G (Figures 6e, 6g, and 941 Supplementary Figure 5 and Table 5); (iv) kGS1→GS2, kGS2→GS1, and Keq from a sub-population of 942 PRE complexes that lacked an A site-bound, deacylated SufB2 in the steady-state movies 943 recorded for the longer time points (i.e., 3, 10, and 20 min) after the delivery of SufB2-TC in the 944 presence of EF-G (Figures 6g, Supplementary Table 6); (v) kGS1→GS2, kGS2→GS1, and Keq from the 945 steady-state movies recorded for the SufB2- and ProL PRE–A complexes (Figures 6h and 946 Supplementary Figure 6 and Table 7); and (vi) kGS1→Æ, kGS2→Æ, kGS2(G)→Æ, and kPOST→Æ from the 947 movies described in (i)-(v) (Figures 6d-6h, Supplementary Figures 3, 5, and 6, and reported two 948 paragraphs above). To calculate these parameters, we extended the variational Bayes approach 949 we introduced in the vbFRET algorithm68 to estimate a ‘consensus’ (i.e., ‘global’) HMM of the 950 EFRET vs. time trajectories. In this approach, we use Bayesian inference to estimate a single, 951 consensus HMM that is most consistent with all the EFRET vs. time trajectories in a movie, rather 952 than to estimate a separate HMM for each trajectory in the movie. To estimate such a consensus 953 HMM, we assume each trajectory is independent and identically distributed, thereby enabling us 954 to perform the inference using the likelihood function 955 ℒ = ∏ ℒ& & ∈ )*+,-.)/*0-1 , (2) 956 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 39 where ℒ& is the variational approximation of the likelihood function for a single trajectory. 957 Subsequently, the single, consensus HMM that is most consistent with all of the trajectories is 958 estimated using the expectation-maximization algorithm that we have previously described68. 959 Viterbi paths (Supplementary Figures 3, 5, and 6), representing the most probable hidden-state 960 trajectory, were then calculated from the HMM using the Viterbi algorithm69. Based on extensive 961 smFRET studies of translation elongation using the bL9(Cy3)-uL1(Cy5) smFRET signal35,36,38, we 962 selected a consensus HMM composed of three states for further analysis of the data. For 963 calculation of the kinetic and/or thermodynamic parameters in (i), (iv), and (v), the three states 964 corresponded to GS1, GS2, and Æ and for calculation of the kinetic and/or thermodynamic 965 parameters in (ii) and (iii), the three states corresponded to EF-G-bound GS2-like, POST, and Æ. 966 The transition matrix of the consensus HMM was then used to calculate kGS1→GS2 and kGS2→GS1 in 967 (i), (iv), and (v); kGS2→POST in (ii); kGS1→Æ, kGS2→Æ, kGS2(G)→Æ, and kPOST→Æ in (vi); and the errors 968 corresponding to each of these parameters. This transition matrix consists of a 3 x 3 matrix in 969 which the off-diagonal elements correspond to the number of times a transition takes place 970 between each pair of the GS1, GS2, and Æ states (in (i), (iv), (v), and (vi)) or each pair of the EF-971 G-bound GS2-like, POST, and Æ states (in (ii) and (vi)) and the on-diagonal elements correspond 972 to the number of times a transition does not take place out of the GS1, GS2, and Æ states (in (i), 973 (iv), (v), and (vi)) or out of the EF-G-bound GS2-like, POST, and Æ states (in (ii) and (vi)). Each 974 element of this matrix parameterizes a Dirichlet distribution, from which we calculated the mean 975 and the square root of the variance for four transition probabilities pGS1→GS2, pGS2→GS1, pGS1→Æ, and 976 pGS2→Æ (in (i), (iv), (v), and (vi)) or for three transition probabilities pGS2→POST, pGS2(G)→Æ, and pPOST→977 Æ (in (ii) and (vi)). These transition probabilities were then used to calculate the corresponding 978 four rate constants, kGS1→GS2, kGS2→GS1, kGS1→Æ, and kGS2→Æ (in (i), (iv), (v), and (vi)) or three rate 979 constants, kGS2→POST, kGS2(G)→Æ, and kPOST→Æ (in (ii) and (vi)) using the equation 980 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 40 𝑘 = − ln(1 − 𝑝) 𝑡 , (3) 981 where t is the time interval between data points (t = 0.1 s). We propagated the error for the 982 transition probabilities into the error for the rate constants using the equation 983 𝜎2 = 𝜎3 (1 − 𝑝) × 𝑡 , (4) 984 where 𝜎3 is the standard deviation of the variance of p and 𝜎2 is the standard deviation of the 985 variance of k. Keq in (i), (iv), and (v) was determined using the equation Keq = kGS1→GS2 / kGS2→GS1. 986 The fractional populations of the POST complex in (iii) and the corresponding errors were 987 calculated by marginalizing, which in this case simply amounts to calculating the mean and the 988 standard error of the mean, for the conditional probabilities of each EFRET data point given each 989 hidden state. Because the data points preceding the initial 70S IC→GS2 transition in the pre-990 steady-state movies do not contribute to the kinetic and/or thermodynamic parameters in (i)-(vi), 991 these data points were not included in the analyses that were used to determine these 992 thermodynamic parameters. 993 994 QUANTIFICATION AND STATISTICAL ANALYSES 995 All ensemble biochemical experiments and cell-based reporter assays were repeated at least 996 three times and the mean values and standard deviations for each experiment or assay are 997 reported. Technical replicates of all smFRET experiments were repeated at least three times and 998 trajectories from all of the technical replicates for each experiment were combined prior to 999 generating the surface contour plot of the time evolution of population FRET and modeling with 1000 the HMM. Mean values and errors for the transition rates and fractional populations determined 1001 from modeling with an HMM are reported (for details see “Analysis of smFRET experiments” in 1002 Methods). Mean values and standard deviations for the k70S IC→GS2s were determined from 1003 technical triplicates of the survival plots analysis for each experiment and are reported. 1004 1005 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 41 DATA AND CODE AVAILABILITY 1006 1007 Data Availability 1008 With the exception of the smFRET data, all other data supporting the findings of this study are 1009 presented within this article. Due to the lack of a public repository for smFRET data, the smFRET 1010 data supporting the findings of this study are available from the corresponding authors upon 1011 request. Source data are provided with this paper. 1012 Code Availability 1013 The code used to analyze the TIRF movies in this study is described in a manuscript in preparation 1014 (Jason Hon, Colin Kinz-Thompson, Ruben L. Gonzalez), where R.L.G. is the corresponding 1015 author. Therefore, the code is available from R.L.G, upon request. 1016 1017 1018 1019 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 42 REFERENCES 1020 1. Wang, K., Schmied, W.H. & Chin, J.W. Reprogramming the genetic code: from triplet to 1021 quadruplet codes. Angew Chem Int Ed Engl 51, 2288-97 (2012). 1022 2. Chen, Y. et al. Controlling the Replication of a Genomically Recoded HIV-1 with a 1023 Functional Quadruplet Codon in Mammalian Cells. ACS Synth Biol 7, 1612-1617 (2018). 1024 3. Lee, B.S., Kim, S., Ko, B.J. & Yoo, T.H. An efficient system for incorporation of unnatural 1025 amino acids in response to the four-base codon AGGA in Escherichia coli. Biochim 1026 Biophys Acta 1861, 3016-3023 (2017). 1027 4. Chatterjee, A., Lajoie, M.J., Xiao, H., Church, G.M. & Schultz, P.G. A bacterial strain with 1028 a unique quadruplet codon specifying non-native amino acids. Chembiochem 15, 1782-6 1029 (2014). 1030 5. Niu, W., Schultz, P.G. & Guo, J. An expanded genetic code in mammalian cells with a 1031 functional quadruplet codon. ACS Chem Biol 8, 1640-5 (2013). 1032 6. Wang, N., Shang, X., Cerny, R., Niu, W. & Guo, J. Systematic Evolution and Study of 1033 UAGN Decoding tRNAs in a Genomically Recoded Bacteria. Sci Rep 6, 21898 (2016). 1034 7. Neumann, H., Wang, K., Davis, L., Garcia-Alai, M. & Chin, J.W. Encoding multiple 1035 unnatural amino acids via evolution of a quadruplet-decoding ribosome. Nature 464, 1036 441-4 (2010). 1037 8. Wang, K. et al. Optimized orthogonal translation of unnatural amino acids enables 1038 spontaneous protein double-labelling and FRET. Nat Chem 6, 393-403 (2014). 1039 9. Atkins, J.F., Loughran, G., Bhatt, P.R., Firth, A.E. & Baranov, P.V. Ribosomal 1040 frameshifting and transcriptional slippage: From genetic steganography and 1041 cryptography to adventitious use. Nucleic Acids Res 44, 7007-78 (2016). 1042 10. Atkins, J.F. & Bjork, G.R. A gripping tale of ribosomal frameshifting: extragenic 1043 suppressors of frameshift mutations spotlight P-site realignment. Microbiol Mol Biol Rev 1044 73, 178-210 (2009). 1045 11. Roth, J.R. Frameshift suppression. Cell 24, 601-2 (1981). 1046 12. Bossi, L. & Roth, J.R. Four-base codons ACCA, ACCU and ACCC are recognized by 1047 frameshift suppressor sufJ. Cell 25, 489-96 (1981). 1048 13. Qian, Q. et al. A new model for phenotypic suppression of frameshift mutations by 1049 mutant tRNAs. Mol Cell 1, 471-82 (1998). 1050 14. Weiss, R.B., Dunn, D.M., Shuh, M., Atkins, J.F. & Gesteland, R.F. E. coli ribosomes re-1051 phase on retroviral frameshift signals at rates ranging from 2 to 50 percent. New Biol 1, 1052 159-69 (1989). 1053 15. Jager, G., Nilsson, K. & Bjork, G.R. The phenotype of many independently isolated +1 1054 frameshift suppressor mutants supports a pivotal role of the P-site in reading frame 1055 maintenance. PLoS One 8, e60246 (2013). 1056 16. Fagan, C.E., Maehigashi, T., Dunkle, J.A., Miles, S.J. & Dunham, C.M. Structural 1057 insights into translational recoding by frameshift suppressor tRNASufJ. RNA 20, 1944-54 1058 (2014). 1059 17. Maehigashi, T., Dunkle, J.A., Miles, S.J. & Dunham, C.M. Structural insights into +1 1060 frameshifting promoted by expanded or modification-deficient anticodon stem loops. 1061 Proc Natl Acad Sci U S A 111, 12740-5 (2014). 1062 18. Dunham, C.M. et al. Structures of tRNAs with an expanded anticodon loop in the 1063 decoding center of the 30S ribosomal subunit. RNA 13, 817-23 (2007). 1064 19. Hong, S. et al. Mechanism of tRNA-mediated +1 ribosomal frameshifting. Proc Natl Acad 1065 Sci U S A 115, 11226-11231 (2018). 1066 20. Sroga, G.E., Nemoto, F., Kuchino, Y. & Bjork, G.R. Insertion (sufB) in the anticodon loop 1067 or base substitution (sufC) in the anticodon stem of tRNA(Pro)2 from Salmonella 1068 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 43 typhimurium induces suppression of frameshift mutations. Nucleic Acids Res 20, 3463-9 1069 (1992). 1070 21. Caliskan, N., Katunin, V.I., Belardinelli, R., Peske, F. & Rodnina, M.V. Programmed -1 1071 frameshifting by kinetic partitioning during impeded translocation. Cell 157, 1619-31 1072 (2014). 1073 22. Taylor, D.J. et al. Structures of modified eEF2 80S ribosome complexes reveal the role 1074 of GTP hydrolysis in translocation. EMBO J 26, 2421-31 (2007). 1075 23. Khade, P.K. & Joseph, S. Messenger RNA interactions in the decoding center control 1076 the rate of translocation. Nat Struct Mol Biol 18, 1300-2 (2011). 1077 24. Liu, G. et al. EF-G catalyzes tRNA translocation by disrupting interactions between 1078 decoding center and codon-anticodon duplex. Nat Struct Mol Biol 21, 817-24 (2014). 1079 25. Abeyrathne, P.D., Koh, C.S., Grant, T., Grigorieff, N. & Korostelev, A.A. Ensemble cryo-1080 EM uncovers inchworm-like translocation of a viral IRES through the ribosome. Elife 5, 1081 doi: 10.7554/eLife.14874 (2016). 1082 26. Schuwirth, B.S. et al. Structures of the bacterial ribosome at 3.5 A resolution. Science 1083 310, 827-34 (2005). 1084 27. Pulk, A. & Cate, J.H. Control of ribosomal subunit rotation by elongation factor G. 1085 Science 340, 1235970 (2013). 1086 28. Ratje, A.H. et al. Head swivel on the ribosome facilitates translocation by means of intra-1087 subunit tRNA hybrid sites. Nature 468, 713-6 (2010). 1088 29. Gamper, H.B., Masuda, I., Frenkel-Morgenstern, M. & Hou, Y.M. Maintenance of protein 1089 synthesis reading frame by EF-P and m(1)G37-tRNA. Nat Commun 6, 7226 (2015). 1090 30. Masuda, I. et al. tRNA Methylation Is a Global Determinant of Bacterial Multi-drug 1091 Resistance. Cell Syst 8, 302-314 e8 (2019). 1092 31. Christian, T. & Hou, Y.M. Distinct determinants of tRNA recognition by the TrmD and 1093 Trm5 methyl transferases. J Mol Biol 373, 623-32 (2007). 1094 32. Murakami, H., Ohta, A., Ashigai, H. & Suga, H. A highly flexible tRNA acylation method 1095 for non-natural polypeptide synthesis. Nat Methods 3, 357-9 (2006). 1096 33. Walker, S.E. & Fredrick, K. Recognition and positioning of mRNA in the ribosome by 1097 tRNAs with expanded anticodons. J Mol Biol 360, 599-609 (2006). 1098 34. Gamper, H.B., Masuda, I., Frenkel-Morgenstern, M. & Hou, Y.M. The UGG Isoacceptor 1099 of tRNAPro Is Naturally Prone to Frameshifts. Int J Mol Sci 16, 14866-83 (2015). 1100 35. Fei, J. et al. Allosteric collaboration between elongation factor G and the ribosomal L1 1101 stalk directs tRNA movements during translation. Proc Natl Acad Sci U S A 106, 15702-1102 7 (2009). 1103 36. Ning, W., Fei, J. & Gonzalez, R.L., Jr. The ribosome uses cooperative conformational 1104 changes to maximize and regulate the efficiency of translation. Proc Natl Acad Sci U S A 1105 111, 12073-8 (2014). 1106 37. Fei, J., Kosuri, P., MacDougall, D.D. & Gonzalez, R.L., Jr. Coupling of ribosomal L1 stalk 1107 and tRNA dynamics during translation elongation. Mol Cell 30, 348-59 (2008). 1108 38. Fei, J., Richard, A.C., Bronson, J.E. & Gonzalez, R.L., Jr. Transfer RNA-mediated 1109 regulation of ribosome dynamics during protein synthesis. Nat Struct Mol Biol 18, 1043-1110 51 (2011). 1111 39. Boel, G. et al. The ABC-F protein EttA gates ribosome entry into the translation 1112 elongation cycle. Nat Struct Mol Biol 21, 143-51 (2014). 1113 40. Chen, B. et al. EttA regulates translation by binding the ribosomal E site and restricting 1114 ribosome-tRNA dynamics. Nat Struct Mol Biol 21, 152-9 (2014). 1115 41. Kim, H.K. et al. A frameshifting stimulatory stem loop destabilizes the hybrid state and 1116 impedes ribosomal translocation. Proc Natl Acad Sci U S A 111, 5538-43 (2014). 1117 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 44 42. Munro, J.B., Wasserman, M.R., Altman, R.B., Wang, L. & Blanchard, S.C. Correlated 1118 conformational events in EF-G and the ribosome regulate translocation. Nat Struct Mol 1119 Biol 17, 1470-7 (2010). 1120 43. Blanchard, S.C., Kim, H.D., Gonzalez, R.L., Jr., Puglisi, J.D. & Chu, S. tRNA dynamics 1121 on the ribosome during translation. Proc Natl Acad Sci U S A 101, 12893-8 (2004). 1122 44. Studer, S.M., Feinberg, J.S. & Joseph, S. Rapid kinetic analysis of EF-G-dependent 1123 mRNA translocation in the ribosome. J Mol Biol 327, 369-81 (2003). 1124 45. Wintermeyer, W. & Rodnina, M.V. Translational elongation factor G: a GTP-driven motor 1125 of the ribosome. Essays Biochem 35, 117-29 (2000). 1126 46. Ermolenko, D.N. et al. Observation of intersubunit movement of the ribosome in solution 1127 using FRET. J Mol Biol 370, 530-40 (2007). 1128 47. Ermolenko, D.N. & Noller, H.F. mRNA translocation occurs during the second step of 1129 ribosomal intersubunit rotation. Nat Struct Mol Biol 18, 457-62 (2011). 1130 48. Cornish, P.V. et al. Following movement of the L1 stalk between three functional states 1131 in single ribosomes. Proc Natl Acad Sci U S A 106, 2571-6 (2009). 1132 49. Nguyen, H.A., Hoffer, E.D. & Dunham, C.M. Importance of a tRNA anticodon loop 1133 modification and a conserved, noncanonical anticodon stem pairing in tRNACGGProfor 1134 decoding. J Biol Chem 294, 5281-5291 (2019). 1135 50. Guo, Z. & Noller, H.F. Rotation of the head of the 30S ribosomal subunit during mRNA 1136 translocation. Proc Natl Acad Sci U S A 109, 20391-4 (2012). 1137 51. Zhou, J., Lancaster, L., Donohue, J.P. & Noller, H.F. Spontaneous ribosomal 1138 translocation of mRNA and tRNAs into a chimeric hybrid state. Proc Natl Acad Sci U S A 1139 116, 7813-7818 (2019). 1140 52. Korniy, N., Samatova, E., Anokhina, M.M., Peske, F. & Rodnina, M.V. Mechanisms and 1141 biomedical implications of -1 programmed ribosome frameshifting on viral and bacterial 1142 mRNAs. FEBS Lett 593, 1468-1482 (2019). 1143 53. Lajoie, M.J. et al. Genomically recoded organisms expand biological functions. Science 1144 342, 357-60 (2013). 1145 54. Wang, K., de la Torre, D., Robertson, W.E. & Chin, J.W. Programmed chromosome 1146 fission and fusion enable precise large-scale genome rearrangement and assembly. 1147 Science 365, 922-926 (2019). 1148 55. Mohan, S., Donohue, J.P. & Noller, H.F. Molecular mechanics of 30S subunit head 1149 rotation. Proc Natl Acad Sci U S A 111, 13325-30 (2014). 1150 56. Kaledhonkar, S. et al. Late steps in bacterial translation initiation visualized using time-1151 resolved cryo-EM. Nature 570, 400-404 (2019). 1152 57. Chen, B. et al. Structural dynamics of ribosome subunit association studied by mixing-1153 spraying time-resolved cryogenic electron microscopy. Structure 23, 1097-105 (2015). 1154 58. Reinkemeier, C.D., Girona, G.E. & Lemke, E.A. Designer membraneless organelles 1155 enable codon reassignment of selected mRNAs in eukaryotes. Science 363(2019). 1156 59. Datsenko, K.A. & Wanner, B.L. One-step inactivation of chromosomal genes in 1157 Escherichia coli K-12 using PCR products. Proc Natl Acad Sci U S A 97, 6640-5 (2000). 1158 60. Fei, J. et al. A highly purified, fluorescently labeled in vitro translation system for single-1159 molecule studies of protein synthesis. Methods Enzymol 472, 221-59 (2010). 1160 61. Christian, T., Lahoud, G., Liu, C. & Hou, Y.M. Control of catalytic cycle by a pair of 1161 analogous tRNA modification enzymes. J Mol Biol 400, 204-17 (2010). 1162 62. Zhang, C.M., Perona, J.J., Ryu, K., Francklyn, C. & Hou, Y.M. Distinct kinetic 1163 mechanisms of the two classes of Aminoacyl-tRNA synthetases. J Mol Biol 361, 300-11 1164 (2006). 1165 63. Peacock, J.R. et al. Amino acid-dependent stability of the acyl linkage in aminoacyl-1166 tRNA. RNA 20, 758-64 (2014). 1167 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 45 64. Aitken, C.E., Marshall, R.A. & Puglisi, J.D. An oxygen scavenging system for 1168 improvement of dye stability in single-molecule fluorescence experiments. Biophys J 94, 1169 1826-35 (2008). 1170 65. Gonzalez, R.L., Jr., Chu, S. & Puglisi, J.D. Thiostrepton inhibition of tRNA delivery to the 1171 ribosome. RNA 13, 2091-7 (2007). 1172 66. Desai, B.J. & Gonzalez, R.L., Jr. Multiplexed, bioorthogonal labeling of multicomponent, 1173 biomolecular complexes using genomically encoded, non-canonical amino acids. 1174 bioRxiv doi: 10.1101/730465(2019). 1175 67. MacDougall, D.D. & Gonzalez, R.L., Jr. Translation initiation factor 3 regulates switching 1176 between different modes of ribosomal subunit joining. J Mol Biol 427, 1801-18 (2015). 1177 68. Bronson, J.E., Fei, J., Hofman, J.M., Gonzalez, R.L., Jr. & Wiggins, C.H. Learning rates 1178 and states from biophysical time series: a Bayesian approach to model selection and 1179 single-molecule FRET data. Biophys J 97, 3196-205 (2009). 1180 69. Viterbi, A.J. Error bounds for convolutional codes and an asymptotically optimum 1181 decoding algorithm. IEEE Trans. Inform. Theory 13, 260-269 (1967). 1182 70. Agirrezabala, X. et al. Visualization of the hybrid state of tRNA binding promoted by 1183 spontaneous ratcheting of the ribosome. Mol Cell 32, 190-7 (2008). 1184 1185 1186 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 46 ACKNOWLEDGEMENTS 1187 We thank Dr. Hajime Tokuda for rabbit polyclonal anti-LolB antibodies, Dr. Colin Kinz-1188 Thompson and Korak Kumar Ray for help with smFRET data analysis. R.L.G. and H.L. thank the 1189 Columbia University Precision Biomolecular Characterization Facility for access to and support of 1190 instrumentation. This work was supported by NIH grants GM134931 to Y-M.H. and GM119386 to 1191 R.L.G., a Charles H. Revson Foundation Postdoctoral Fellowship in Biomedical Science 19-24 to 1192 H.L., a Japanese JSPS overseas postdoctoral fellowship to I.M., and NSF grant CHE-1708759 to 1193 E.J.P. 1194 1195 AUTHOR CONTRIBUTIONS 1196 H.G. conceived of and performed ensemble rapid kinetic assays, R.L.G. and H.L. conceived 1197 of and designed smFRET assays, H.L. performed smFRET assays, I.M. performed cell-based 1198 reporter assays, D.M.R. and E.J.P. generated aminoacyl-DBE derivatives, T.C. performed G37 1199 methylation and aminoacylation assays, and A.B.C. and G.B. provided E. coli 70S ribosomes. 1200 Y.M.H. and R.L.G. wrote the manuscript. 1201 1202 COMPETING FINANICAL INTERESTS 1203 The authors declare no competing interests. 1204 1205 CONTACT FOR REAGENT AND RESOURCE SHARING 1206 Further information and requests for resources and reagents should be directed to and will be 1207 fulfilled by the lead contacts Ruben L. Gonzalez, Jr. (rlg2118@columbia.edu) and Ya-Ming Hou 1208 (ya-ming.hou@jefferson.edu). 1209 1210 1211 1212 1213 1214 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 47 FIGURE LEGENDS 1215 1216 Figure 1. Methylation and aminoacylation of SufB2 and ProL. a Sequence and secondary 1217 structure of native-state SufB2, showing the N1-methylated G37 in red and the G37a insertion to 1218 ProL in blue. b RNase T1 cleavage inhibition assays of TrmD-methylated G37-state SufB2 1219 transcript confirm the presence of m1G37 and m1G37a. Cleavage products are marked by the 1220 nucleotide positions of Gs. L: the molecular ladder of tRNA fragments generated from alkali 1221 hydrolysis. c Primer extension inhibition assays identify m1G37 in native-state SufB2. Red and 1222 blue arrows indicate positions of primer extension inhibition products at the methylated G37 and 1223 G37a, respectively, which are offset by one nucleotide relative to ProL. The first primer extension 1224 inhibition product for SufB2 corresponds to m1G37a, the second corresponds to m1G37, while the 1225 primer extension inhibition product for ProL corresponds to m1G37. Due to the propensity of 1226 primer extension to make multiple stops on a long transcript of tRNA, the read-through primer 1227 extension product (54-55 nucleotides) had a reduced intensity relative to the primer extension 1228 inhibition products (21-22 nucleotides). Molecular size markers are provided by the primer alone 1229 (17 nucleotides) and the run-off products (54-55 nucleotides). d TrmD-catalyzed N1 methylation 1230 of G37-state SufB2 and ProL as a function of time. e, f ProRS-catalyzed aminoacylation. e 1231 Aminoacylation of native-state SufB2 and ProL. f Aminoacylation of G37-state SufB2 and ProL 1232 as a function of time. In panels b, c, gels were performed three times with similar results, while in 1233 panels d-f, the bars are SD of three independent (n = 3) experiments, and the data are presented 1234 as mean values ± SD. 1235 1236 Figure 2. SufB2-induced +1 frameshifting and genome recoding. a The +1-frameshifting 1237 efficiency in cell-based lacZ assay for SufB2 and ProL strains in m1G37+ and m1G37– conditions. 1238 The bars in the graph are SD of four, five, or six independent (n = 4, 5, or 6) biological repeats, 1239 and the data are mean values ± SD. b The difference in the ratio of protein synthesis of lolB to 1240 cysS for SufB2 and ProL strains in m1G37+ and m1G37– conditions relative to ProL in the m1G37+ 1241 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 48 condition. c Measurements underlying the bar plots in panel b. Each ratio was measured directly 1242 and the ratio of ProL in the m1G37+ condition was normalized to 1.0. The difference of each ratio 1243 relative to the normalized ratio represented the +1-frameshifting efficiency at the CCC-C motif at 1244 the 2nd codon of lolB. The bars in the graph are SD of three independent (n = 3) biological repeats, 1245 and the data are mean values ± SD. In a, b, decoding of the CCC-C motif was mediated by SufB2 1246 and ProM in the SufB2 strain, and by ProL and ProM in the ProL strain, where the presence of 1247 ProM ensured no vacancy at the CCC-C motif. The increased +1 frameshifting in the m1G37– 1248 condition vs. the m1G37+ condition indicates that SufB2 and ProL are each an active determinant 1249 in decoding the CCC-C motif. d SufB2-mediated insertion of non-proteinogenic amino acids at 1250 the CCC-C motif in the 5th codon position of folA using [35S]-Met-dependent in vitro translation. 1251 Reporters of folA are denoted by +/– CCC-C, where “+” and “–” indicate constructs with and 1252 without the CCC-C motif. SDS-PAGE analysis identifies full-length DHFR resulting from a +1-1253 frameshift event at the CCC-C motif by SufB2 pre-aminoacylated with the amino acid shown at 1254 the top of each lane, a DC fragment resulting from lack of the +1-frameshift event, and a DN 1255 fragment resulting from translation initiation at the AUG codon likely at position 17 or 21 1256 downstream from the CCC-C motif. Gel samples were derived from the same experiment, which 1257 was performed five times with similar results. Gels for each experiment were processed in parallel. 1258 Lane 1: full-length DHFR as the molecular marker; deacyl: deacylated tRNA. 1259 1260 Figure 3. SufB2 uses a triplet anticodon-codon pairing scheme at the A site. a GTP 1261 hydrolysis by EF-Tu as a function of time for delivery of G37- or native-state SufB2- or ProL-TC 1262 to the A site of a 70S IC. Although the concentration of TCs was limiting, which would limit the 1263 rate of binding of TCs to the 70S IC, the observed differences in the yield of GTPase activity 1264 indicated that binding was not the sole determinant, but that other factors, such as the identity 1265 and the methylation state of the tRNA, affected the GTPase activity. b Dipeptide fMP formation 1266 as a function of time for delivery of G37- or native-state SufB2- or ProL-TC to the A site of a 70S 1267 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 49 IC. Due to the limiting concentration of the 70S IC, which did not include the tRNA substrate, the 1268 yield of di- or tri-peptide formation assays was constant even with different tRNAs in TCs. c The 1269 yield of fMP and fMR in dipeptide formation assays in which equimolar mixtures of native-state 1270 SufB2-TC, carrying Pro and/or Arg, and/or native-state ProL-TC, carrying Pro and/or Arg, are 1271 delivered to 70S ICs. The mRNA in 70S ICs in (A-C) is AUG-CCC-CGU-U. d Dipeptide formation 1272 rate kfMP,obs for delivery of G37-state SufB2-TC to 70S ICs containing sequence variants of the 1273 CCC-C motif in the A site. In panels a, b, the bars in the graphs are SD of three independent (n 1274 = 3) experiments, in panel c, the bars in the graphs are SD of four independent (n = 4) experiments, 1275 and in panel d, the bars in the graphs are SD of three or four independent (n = 3 or 4) experiments. 1276 All data are presented as mean values ± SD. ∆t: a time interval, ND: not detected. 1277 1278 Figure 4. Plasticity of SufB2-induced +1 frameshifting. a fMP formation as a function of time 1279 upon delivery of the G37C variant of G37-state SufB2-TC to the A site of a 70S IC, allowing 1280 nucleotides 34-36 to pair with a CCC-C motif at the A site. b fMP formation as a function of time 1281 upon delivery of the G34C variant of G37-state SufB2-TC to the A site of a 70S IC, allowing 1282 nucleotides 35-37 to pair with a CCC-C motif. c-f Results of fMPV formation assays in which 1283 SufB2-TC is delivered to an A site programmed with a quadruplet codon at the 2nd position and 1284 sequences of the SufB2 anticodon loop and/or quadruplet codon are varied. Yields of fMPV 1285 formation represent +1 frameshifting during translocation of SufB2 from the A site to the P site. 1286 Possible +1-frame anticodon-codon pairing schemes of SufB2 during translocation: c G37-state 1287 SufB2 capable of frameshifting at a CCC-C motif via quadruplet pairing and/or triplet slippage, d 1288 G37C variant of G37-state SufB2 capable of frameshifting at a GCC-C motif via quadruplet pairing 1289 and/or triplet slippage, e m1G37-state SufB2 capable of frameshifting at a CCC-C motif via only 1290 triplet slippage, and f G37C variant of G37-state SufB2 capable of frameshifting at a CCC-C motif 1291 via only triplet slippage. In panels a, b, the bars in the graphs are SD of three (n = 3) independent 1292 experiments, and the data are presented as mean values ± SD. ∆t: a time interval. 1293 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 50 1294 Figure 5. SufB2 shifts to the +1-frame during translocation. a Relative fMPV and fMPR 1295 formation as a function of time upon rapid delivery of EF-G and an equimolar mixture of G37-state 1296 SufB2-, tRNAVal-, and tRNAArg-TCs to 70S ICs carrying a CCC-C motif in the A site. b Relative 1297 fMPV and fMPR formation as a function of time when a defined time interval is introduced between 1298 delivery of G37-state SufB2-TC and EF-G and delivery of an equimolar mixture of tRNAArg- and 1299 tRNAVal-TCs. c Relative fMPV and fMPR formation after reacting fMP-POST complexes with a 1300 mixture of tRNAVal- and tRNAArg-TCs based on the time courses in Supplementary Figures 2d-f. 1301 d fMPV formation as a function of time upon rapid delivery of tRNAVal-TC to an fMP-POST 1302 complex carrying a CCC-N motif in the A site. e Relative fMPV and fMPS formation as a function 1303 of time upon rapid delivery of an equimolar mixture of tRNAVal- and tRNASer-TCs to an fMP-POST 1304 complex carrying a CCC-A motif in the A site. In panels a-e, the bars are SD of three (n = 3) 1305 independent experiments and the data are presented as mean values ± SD. Arg: arginyl-tRNAArg; 1306 Val: valyl-tRNAVal. 1307 1308 Figure 6. SufB2 interferes with elongation complex dynamics during late steps of 1309 translocation. a-c Cartoon representation of elongation as a G37-state SufB2- or ProL-TC is 1310 delivered to the A site of a bL9(Cy3)- and uL1(Cy5)-labeled 70S IC; a in the absence, or b in the 1311 presence of EF-G, or c upon using puromycin (Pmn) to deacylate the P site-bound G37-state 1312 SufB2 or ProL and generate the corresponding PRE–A complex. The 30S and 50S subunits are 1313 tan and light blue, respectively; the L1 stalk is dark blue; Cy3 and Cy5 are bright green and red 1314 spheres, respectively; EF-Tu is pink; EF-G is purple; fMet-tRNAfMet is dark green; and SufB2 or 1315 ProL is dark red. d, e Hypothetical (top) and representative experimentally observed (bottom) 1316 EFRET vs. time trajectories recorded as ProL-TC is delivered to a 70S IC, d in the absence and e 1317 in the presence of EF-G as depicted in a, b. The waiting times associated with k70S IC→GS2, kGS1→GS2, 1318 kGS2→GS1, and kGS2→POST are indicated in each hypothetical trajectory. f, g, and h Surface contour 1319 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 51 plots of the time evolution of population FRET obtained by superimposing individual EFRET vs. time 1320 trajectories in the experiments in a, b, and c, respectively, for SufB2 (top) and ProL (bottom). N: 1321 the number of trajectories used to construct each contour plot. Surface contours are colored as 1322 denoted in the population color bars. For pre-steady-state experiments, the black dashed lines 1323 indicate the time at which the TC was delivered and the gray shaded areas denote the time 1324 required for the majority (54 - 68%) of the 70S ICs to transition to GS2. Note that the rate of 1325 deacylated SufB2 dissociation from the A site under our conditions is similar to that of EF-G-1326 catalyzed translocation, thereby resulting in the buildup of a PRE complex sub-population over 3-1327 20 min post-delivery that lacks an A site tRNA and is incapable of translocation. This sub-1328 population exhibits kGS1→GS2, kGS2→GS1, and Keq values similar to those observed in experiments 1329 recorded in the absence of EF-G (Supplementary Table 6). 1330 1331 Figure 7. Structure-based mechanistic model for SufB2-induced +1 frameshifting. A SufB2-1332 TC uses triplet anticodon-codon pairing in the 0-frame at a CCC-C motif, undergoes peptide-bond 1333 formation, and enables the resulting PRE complex to undergo a GS1→GS2 transition, all with 1334 rates similar to those of ProL-TC. During the GS1→GS2 transition, the 30S subunit rotates 1335 relative to the 50S subunit by 8º in the counter-clockwise (+) direction along the black curved 1336 arrow; the 30S subunit head swivels relative to the 30S subunit body by 5º in the clockwise (–) 1337 direction against the black curved arrow; the L1 stalk closes by ~60 Å; and the tRNAs are 1338 reconfigured from their P/P and A/A to their P/E and A/P configurations. EF-G then binds to the 1339 PRE complex to form PRE-G1 and subsequently catalyzes a series of conformational 1340 rearrangements of the complex (PRE-G1 to PRE-G4) that encompass further counter-clockwise 1341 and clockwise rotations of the subunits; severing of decoding center interactions with the 1342 anticodon-codon duplex in the A site; counter-clockwise and clockwise swiveling of the head and 1343 the associated opening and closing of the E-site gate; opening of the L1 stalk; and 1344 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 52 reconfigurations of the tRNAs as they move from the P and A sites to the E and P sites. It is during 1345 these steps, shown in red arrows within the gray shaded box, that SufB2 impedes forward and/or 1346 reverse swiveling of the head and the associated opening and/or closing of the E-site gate, 1347 facilitating +1 frameshifting. Next, EF-G and the deacylated tRNA dissociate from PRE-G4, 1348 leaving a POST complex ready to enter the next elongation cycle. The cartoons depicting PRE-1349 G1(GS1) and PRE-G1(GS2) were generated using Biological Assemblies 2 and 1, respectively, 1350 of PDB entry 4V9D. Due to the lack of an A-site tRNA or EF-G in 4V9D, cartoons of the A- and 1351 P-site tRNAs from previous structures1 were positioned into the two assemblies using the P-site 1352 tRNAs in 4V9D as guides and a cartoon of EF-G generated from 4V7D was manually positioned 1353 near the factor binding site of the ribosomes. The cartoons depicting PRE-G2, PRE-G3, and PRE-1354 G4 were generated from 4V7D, 4W29, and 4V5F, respectively, and colored as in Figure 6, with 1355 the head domain shown in orange. 1356 1357 1358 1359 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 53 Figure 1 1360 1361 1362 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 54 Figure 2 1363 1364 1365 1366 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 55 Figure 3 1367 1368 1369 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 56 Figure 4 1370 1371 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 57 Figure 5 1372 1373 1374 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 58 Figure 6 1375 1376 1377 P A 1.00 0.50 0.00 E F R E T 0 10 20 30 0 10 1 Time (s) Time (s) ProL-TC ProL-TC +EF-G Deliver TCsat 5 s Deliver TCs+EF-Gat 5 s N= k70S IC→GS kGS1→GS a kGS →GS1 A P. . Time e Time 0.55 ( 1 e ) 0. 1 ( 1 l se ) 0. 0.7 15 15 10 10 N= 5 1 ProL GS P E1 Stal P ST GS1 (EF-G) GS (EF-G) E GS GS1 E F R E T k70S IC→GS kGS →P ST 0 5 Time (s) 7.5.5 E F R E T N = N= N= N= 5 Suf N=1 N= 7 N= 1 ProL 0. 0.0 0. 0. 0. 0. 1.0 1. E F R E T Suf 0. 0.0 0. 0. 0. 0. 1.0 1. N=1 ProL 70S IC TC EF-G Pm N=1 Time (mi ) Time (mi ) .5 GS1 0. 5 0. 0 5 107.5 .5 0 5 Time (s) 10 Time (s) .5 0 5 Time (s) 7.5 a 0 1 0 1 10 N=17 0 0 N=55 Suf 0. 0. 0 5 Time (s) 10 0 5 Time (s) 7.5.5 0 5 Time (s) 7.5.5 0 5 Time (s) 7.5 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/ 59 Figure 7 1378 1379 1380 1381 60 Å POST PRE PRE-G1 PRE-G2 PRE-G3 PRE-G4 POST GS1 GS1 GS2 GS2 60 Å Intersubunit rotation 8° 8° 12° 2° 0° Head swive in - ° - ° 3° 21° 0° .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2021. ; https://doi.org/10.1101/2020.12.31.424971doi: bioRxiv preprint https://doi.org/10.1101/2020.12.31.424971 http://creativecommons.org/licenses/by-nc-nd/4.0/