Anatomy of footprint extension in ribosome profiling reveals translational landscape mediated by S1 protein in bacteria


 Ribosome profiling — RNase footprinting of ribosome-bound mRNA — has been a unique and powerful method, applied to widespread organisms to survey ribosome traversal along mRNAs. In contrast to eukaryotes, bacterial ribosome footprints show a broad range of sizes, reflecting the differential states of ribosomes. However, the origin remains unclear. Here, we show that rotated state of ribosomes and intramolecular RNA duplexes each extend bacterial ribosome footprints at the 5′ end. Combining elongation inhibitors, cryo-electron microscopy, and ribosome profiling, we demonstrated that the rotated state of ribosomes results in long footprints. Along the subunit rotation, ribosomal protein S1 — a 30S-subunit RNA-binding protein — sterically protects mRNA at the 5′ end of the ribosome from RNase digestion and facilitates elongation of the ribosome. Moreover, we found that ribosomes stalled on ycbZ mRNA generate prolonged footprints because of their unique RNA secondary structure proximal to ribosomes. Through the studies of footprint extension, our results revealed S1-mediated stabilization of translation elongation and provide ribosome profiling approach to probe the conformational diversity of ribosomes in bacteria.


Introduction
Ribosomes traverse along mRNAs to decode codons and synthesize amino acid chains (Voorhees and Ramakrishnan, 2013). This translation elongation is accompanied with tRNA translocation in the inter subunit cavity driven by global conformational rearrangements of the 70S ribosome, such as a rotation of two subunits (Voorhees and Ramakrishnan, 2013). In the non-rotated state, ribosomes possess peptidyl-tRNA at the P site and the codon at the vacant A site undergoes the initial selection of the appropriate aminoacyl-tRNA with the aid of EF-Tu. After the peptidyl transfer reaction, the small subunit (30S) rotates relative to the large subunit (50S) in a counterclockwise manner (i.e., rotated state) with the guide of EF-G and then dynamic positional rearrangement of tRNAs is induced: tRNAs in ribosomes occupy hybrid E/P and P/A positions, in which P-and A-site tRNA heads lean toward the E and P sites of the 50S subunit, respectively.
Finally, the translocation of ribosomes is prompted by GTP hydrolysis of EF-G, and then the ribosome returns to a non-rotated state for the next round of the elongation cycle. This elongation process is even more complicated in cells; various factors (e.g., tRNA availabilities, RNA secondary structures, and nascent peptide-ribosome exit tunnel interaction) halt the elongation cycle in the middle and stall ribosomes on mRNAs (Nakatogawa and Ito, 2002;Lu and Deutsch, 2008;Yanagitani et al., 2011;Dana and Tuller, 2012;Brandman et al., 2012;Charneski and Hurst, 2013;Pop et al., 2014;Hussmann et al., 2015;Weinberg et al., 2016;Zhang et al., 2017;Dao Duc and Song, 2018;Ibrahim et al., 2018;Diament et al., 2018;Mohammad et al., 2019;Sharma et al., 2019).
Ribosome profiling has been proven to be a powerful and sensitive tool to measure ribosome movement along mRNA in vivo (Ingolia et al., 2009;McGlincy and Fujita et al. p4 Ingolia, 2017;Iwasaki and Ingolia, 2017). Harnessing the protection of RNA fragments by ribosomes from RNase digestion and sequencing of the "ribosome footprints" allows us to survey ribosome occupancy across the transcriptome. Since RNase accessibility determines the resultant size of the RNA fragment, it is reasonable to suppose that the conformational rearrangements in different ribosomal states affect the footprint lengths.
While eukaryotic ribosomes generate two distinct footprint sizes depending upon A-site tRNA occupancy (Lareau et al., 2014;Matsuo et al., 2017;Kurihara et al., 2018;Wu et al., 2019), E. coli ribosomes have been reported to have a broad footprint size range (Mohammad et al., 2016;Mohammad et al., 2019). However, the rationales of the origin of widely distributed footprint size and its relevance to ribosome states remain unknown.
Here, we show that 5′-extended, long footprints highly correlated to the rotated state of ribosomes in bacteria. Testing a series of translation inhibitors in ribosome profiling, we found that ribosomes trapped in the rotated state result in 35-40 nt footprints, while non-rotated ribosomes generate 23-28 nt footprints. Ribosomal protein S1, which composes the small ribosomal subunit and is located close to the mRNA exit channel, contributed to footprint extension toward the 5′ end. The interaction between S1 and mRNA stabilized elongation step and facilitated protein synthesis from long coding sequences (CDS). In addition, we found that a subset of codon positions still retained long footprints regardless of drug treatment. A ribosome pause site on ycbZ mRNA coincided with highly structured RNA docked to ribosomes, rendering footprints resistant to RNase treatment. This work provides an insight into stabilization mechanism of rotating ribosomes during elongation reactions and a framework to link ribosome footprint size and ribosome states in bacteria. Fujita et al. p5
Remarkably, the drugs could be sub-classified into two groups based on the footprint size distribution. The first group of inhibitors, including chloramphenicol (a standard for bacterial ribosome profiling), linezolid, and blasticidin, showed that the footprints peaked at ~28 nt, as observed in a drug-free condition (Fig. 1a). This size has been observed in earlier works (Mohammad et al., 2016;Mohammad et al., 2019). In contrast, the second group of inhibitors -capreomycin, viomycin, and spectinomycin -extended ribosome footprints toward 35-43 nt (Fig. 1b).
Indeed, the sub-classification corresponded to the ribosome conformations trapped by the drugs. The first group of inhibitors (chloramphenicol, linezolid, and blasticidin) binds to the peptidyl transferase center (PTC) of the large subunit (50S), blocks peptide bond formation, and thus stabilizes ribosomes in a non-rotated state (Wilson et al., 2008;Bulkley et al., 2010;Dunkle et al., 2010;Svidritskiy et al., 2013).
The second group of inhibitors, capreomycin and viomycin are known to bind to the 16S rRNA decoding center located on the 30S-subunit A site, thereby inhibiting the translocation of mRNA and tRNAs (Ermolenko et al., 2007;Cornish et al., 2008;Stanley et al., 2010;Dunkle et al., 2011;Wang et al., 2012). Spectinomycin binds to the head domain of the 30S small subunit (Borovinskaya et al., 2007;Carter et al., 2000) to block head swiveling after subunit rotation. The dose-dependent effect of the second group of drugs further supported that the prolonged footprints were due to compound treatment . Although footprint extension was observed for all the second group of inhibitors, the weaker effects of spectinomycin might originate from the mechanistic difference in drugs to halt ribosomes in the rotated state.
To further characterize the longer footprints, we categorized the reads obtained into two bins based on the read length (short; 23-28 nt and long; 35-40 nt). Meta-gene analysis confirmed the enrichment of long footprints along the CDS in capreomycin treatment compared with chloramphenicol treatment (Fig.1c, d). Moreover, the 3-nt periodicity of footprints (Fig. 1c,d) provided further evidence that the longer footprints emerged from elongating ribosomes, as observed in regular short footprints. The high reproducibility of footprints on each CDS showed that capreomycin treatment did not bias ribosome footprints toward specific mRNAs ( Supplementary Fig. 1d). Similarly, the drug did not alter ribosome occupancy on each codon species ( Supplementary Fig. 1e).
Thus, capreomycin extended footprints from elongating ribosomes throughout codons across the transcriptome. Fujita et al. p7 Since the earlier structural basis of capreomycin action has been restricted in Thermus thermophilus 70S (Stanley et al., 2010), we examined whether the drugs trap ribosome conformations in E. coli by cryo-electron microscopy single particle analysis (cryo-EM) i,j,and Supplementary Table 1). The three-dimensional classification based on the global conformation of the 70S ribosome showed that the majority (52%) of the E. coli ribosomes treated with capreomycin were populated in the rotated state with two tRNAs at hybrid positions (P/E-tRNA and A/P-tRNA) ( Fig. 1f and Supplementary Fig. 2b). Notably, the distinct mass of cryo-EM density attributed for capreomycin was clearly observed at the 16S rRNA decoding center (Fig. 1h). The sable binding of capreomycin flips A1492 and A1493 bases, which maintain fidelity of decoding. This observation is consistent with earlier observations in Thermus thermophilus (Stanley et al., 2010). Predictably, in the chloramphenicol-treated sample, the major population of particles converged to non-rotated ribosomes with chloramphenicol bindings (Fig. 1e, g, and Supplementary Fig. 2a). We also observed unpredicted rotated state of ribosomes with P/E tRNA ( Supplementary Fig. 2a, i, j), accompanied with A/P tRNA, which was barely observed as fragmented cryo-EM densities in the lower contour level. This is feasible that chloramphenicol binds to the PTC after tRNAs' leaning and subunit rotation, although we suspect this situation is not frequent. Taken together, the ribosome conformations confined by the drugs led to the differential footprint size in ribosome profiling. Fujita et al. p8 The structure-footprint size correlation allowed us to address ribosome conformations at a sub-codon resolution. Here we revisited to the reported translation pause sites. To investigate the naïve ribosome conformation, ribosome profiling without drug treatment was analysed. The tnaC pause site at the stop codon is an example: two L-Trp molecules bind to the pocket formed between the ribosome exit tunnel and the nascent chain, position PTC in an incompatible conformation, and then trap the ribosome in a nonrotated state (Gong et al., 2001;Bischoff et al., 2014). As expected in the structural study, we observed the prominent accumulation of short footprints at the corresponding pause site (Fig. 1i).

Conformational annotations of characterized ribosome pause sites by ribosome profiling
The second example of ribosome pauses focuses on secM, decoding Pro166 at the A site (Nakatogawa and Ito, 2001;Nakatogawa and Ito, 2002;Muto et al., 2006;Zhang et al., 2015). This nascent chain upstream of the pause site interacts with the ribosome exit tunnel and eventually induces two different consequences in the structure to be stalled (Zhang et al., 2015): inactivation of the peptidyl transfer reaction and long duration of peptidyl-tRNA at the A/P hybrid position. As a result, ribosomes are paused in both non-rotated and rotated states (Zhang et al., 2015). Concordant with the structural analysis, ribosomes on the secM pause site accumulated both long and short footprints ( Fig. 1j). Taken together, the characterization of ribosome pause sites further supported the categorization of rotated and non-rotated states by footprint lengths.

Long footprints are extended at the 5′ end
We were intrigued by the molecular basis of the footprint extension. To investigate the relative direction of extension on footprints, we calculated ribosome occupancies around Fujita et al. p9 stop codons which enrich the footprints in general, assigning the 5′ and 3′ ends of the ribosome footprints. The ribosome under termination or recycling was denoted by a region between the peaks in 5′ and 3′ assigning that locates upstream and downstream of stop codon, respectively (Fig. 2a). Whereas the 3′ ends of footprints were at the same position in short and long footprints, the 5′ ends were shifted along the read extension ( Fig. 2a). Thus, the major difference in footprint length originated from variable 5′ ends.

Ribosomal protein S1 protects the 5′ end of ribosome footprints in rotated state of ribosomes
Given the variable 5′ ends, we reasoned that factor(s) close to the 5′ ends of ribosomes may affect RNase accessibility. An apparent candidate is ribosomal protein S1 (S1), which lies adjacent to the mRNA exit channel in the 30S subunit (Byrgazov et al., 2015;Demo et al., 2017;Beckert et al., 2018). S1 is an RNA binding protein and consists of six repeats of OB-fold domains. The first two OB-fold repeats directly interact with 30S (Byrgazov et al., 2015), and the subsequent four domains can bind to mRNA (Boni et al., 1991;Sengupta et al., 2001;Qu et al., 2012;Duval et al., 2013), covering a 10-nt-long region (Qu et al., 2012) To test the contribution of S1 to footprint extension, we attenuated the RNA binding ability of the protein. Although S1 could not be depleted from cells because of its essential role in general translation, the partial deletion of the RNA binding region was feasible (S1 ∆56 strain) (Duval et al., 2013) (Supplementary Fig. 3a, b). Ribosome profiling from Fujita et al. p10 the strain with truncated S1 showed reduced long footprints with capreomycin ( Fig. 2c).
On the other hand, the short footprints generated by chloramphenicol treatment were not affected by mutated S1 (Fig. 2b). Thus, these data indicated that S1 prolongs the footprints in a ribosome conformation-specific manner. S1 has been proposed to function in translation initiation in a form of 30S; S1 directly associates with the U-rich region of mRNA on the upstream of the start codon, recruits the 30S subunit to mRNA, and facilitates the recognition of the start codon (Duval et al., 2013). In contrast, S1's effect on ribosome footprint extension suggested that S1 also associates with mRNA on the elongating 70S ribosome. To examine this scenario, we monitored the position of S1 along transcripts by UV-C crosslinking immunoprecipitation sequencing with an infrared labeled linker (irCLIP-seq) (Zarnegar et al., 2016) (Supplementary Fig. 3c). As expected by the role of S1 in translation initiation, we observed the pile-up of the reads upstream of initiation codons ( Supplementary Fig. 3d). The peak position (~20 nt upstream of the start codon) was consistent with the location of S1 on the 30S ribosome (Boni et al., 1991). In addition to the 5′ UTR, the irCLIP-seq reads were also found in the CDS ( Supplementary Fig. 3d), providing evidence of the interaction of S1 with mRNA in the elongating 70S.
Indeed, the enrichments of irCLIP-seq reads by capreomycin treatment around the stop codon corresponded closely to the footprint extension by the drug (Fig. 2d). Thus, the data suggested that, in the rotated state, S1 on the 70S ribosome has a higher chance of associating with the mRNA region close to the mRNA exit channel. It is most likely that the S1 interaction sterically withstands RNase treatment and thus extends the footprints at the 5′ ends.

S1 stabilizes elongation ribosomes and facilitates protein synthesis from long CDS.
Given that S1 interacts with mRNA in CDS ( Supplementary Fig. 3d), we reasoned that S1 may have a role in translation elongation in addition to translation initiation. Indeed, calculating footprint distribution on each CDS [i.e. polarity score (Schuller et al., 2017)], we found that S1 ∆56 strain biased the footprint toward 5′ direction, especially in long CDS (300 codons or longer) (Fig. 3a). These data suggested that elongation in S1 ∆56 strain was delayed or aborted.
To test this possibility, we set out in vitro translation with S1-depleted ribosomes. To avoid the effect of S1 depletion on translation initiation, we designed the firefly luciferase reporter fused to sodB 5′ UTR, which allows S1-independent initiation (Duval et al., 2013) (Fig. 3b, top). The extension of CDS with single or tandem GFP sequences in reporter reduced protein synthesis driven by S1-depleted ribosomes (Fig. 3b, bottom), as a readout of elongation defects. The supplementation of recombinant S1 protein ( Supplementary Fig. 3e) rescued the translation reactions (Fig. 3b, bottom). We note that S1 depletion per se did not strongly affect the translation in our condition, probably because S1 was dropped off ribosomes during preparation and thus at substoichiometric in in vitro translation system (PURE system) ( Supplementary Fig. 3e) as reported in another ribosome protein (Chadani et al., 2017). Taken together, these data suggested that S1 facilitates translation elongation, associating with mRNAs during rotation of ribosomes.

Footprint extension beyond ribosome rotation
Although we found that the inter-subunit rotation of ribosomes contributes to differential footprint sizes, a fraction of extended footprints could be attributed to other rationales.
For example, duplex formation between the Shine-Dalgarno (SD) sequence in mRNA and the anti-SD sequence in 16S rRNA, which is docked in the chamber on 30S (Kaminishi et al., 2007) was reported to block trimming by RNase (Mohammad et al., 2016;Mohammad et al., 2019). Our ribosome profiling also showed that the SD-like Grich sequence (e.g., UGG) resulted in longer footprint accumulation ~20-25 nt downstream ( Supplementary Fig. 4a), where the anti-SD sequence of 16S rRNA is located. Such extension was minor overall but was still found in a fraction of ribosome footprints in the chloramphenicol-treated ribosome profiling library. Indeed, the regression analysis of two Gaussian distribution formulas modeled two populations, which consisted of a major population peak at ~25 nt and a minor population peak at ~35 nt ( Supplementary Fig. 4b).
To identify a factor altering the footprint length beyond the inter-subunit rotation, we surveyed a specific codon position biased toward long footprints across the transcriptome irrespective of drug treatment. Considering footprint sizes, we calculated the ribosome pause score, which is the ratio of reads to average reads per codon at the given mRNA (Schuller et al., 2017), at the A-site codon under drug-free conditions (Fig.   4a). Here, we defined 64 codons in which long footprints were over-represented compared with short footprints (Fig. 4a). Many of these still possessed long footprints even after chloramphenicol treatment (Fig. 4b), which should convert footprints to short footprints. Consistent with the footprint extension by the SD-like sequence, a fraction of the long-footprint-biased pause sites possessed a low free energy (∆G) between the anti-SD sequence on 16S rRNA and the sequence upstream of the pause site (Supplementary Fujita et al. p13 Fig. 4c, e.g., lower than red dot line). However, a subset of the codons (such as ycbZ pause site, see below for the details) still could not be explained by the SD-like sequence ( Supplementary Fig. 4c, e.g., higher than top black dot line), implying other molecular bases for footprint extension.

Ribosomes are stalled on ycbZ
Then, we characterized the codon sites accompanied by drug-independent long footprints.
Because we considered the over-accumulation of long footprints in the survey of codon positions, the sites naturally had higher footprints overall and thus are candidates for ribosome pause sites. To test the stalling of ribosomes on those sites, we performed in vitro integrated nascent chain profiling (iNP) (Chadani et al., 2016) using a reconstituted cell-free translation system (Shimizu et al., 2001) with 35 S methionine. In this experiment, the accumulation of intermediate peptidyl-tRNA, which migrates slowly in SDS-PAGE at neutral pH because of tRNA (~20 kDa) conjugation, was used as a readout of ribosome pause. Along with well-characterized secM, 6 out of 9 selected mRNAs with longer footprint pause sites -ispH, napD, nuoA, sseA, ycbZ, ykgF, hybA, hybB, and lpxC ( Fig.   4a, b) -led to high molecular weight products (Fig. 4c, d). The disappearance of the signals after RNase treatments validated the presence of a tRNA moiety on the peptide ( Fig. 4c).
Among the pause sites with long footprints, we focused the site found in ycbZ just before the stop codon ( a unit -on the exact same site (Pro585 at the P site and Asn586 at the A site) further supported that the elongation reaction on the ycbZ transcript is slow enough for trailing ribosomes to catch up to the leading ribosome ( Fig. 5b and Supplementary Fig. 5b). To uncover the minimum region critical for the pause, we designed a reporter for in vitro translation. Given that the amino acid/RNA contexts around the pause sites have been reported to impede translation elongation (Nakatogawa and Ito, 2002;Lu and Deutsch, 2008;Yanagitani et al., 2011;Dana and Tuller, 2012;Brandman et al., 2012;Charneski and Hurst, 2013;Weinberg et al., 2016;Dao Duc and Song, 2018), we placed the 30 amino acids upstream from the pause site, which ranges from the A site to the end of the nascent chain held within the ribosome exit tunnel, into a reporter (Fig. 5c, top and Supplementary Fig. 5c). The translated products were detected by Western blotting with a HA-tag inserted in front of the pause site. Indeed, the inserted segment of ycbZ was sufficient to drive RNase-sensitive peptidyl-tRNA from the reporter (Fig. 5c, bottom).
The substitution of the 30 amino acid sequence into the unrelated tag sequence (streptavidin binding protein or SBP) reduced the pause ( Supplementary Fig. 5d).
We further narrowed down the critical amino acid for ribosome pause. Taking the amino acid sequence conservation among species (e.g., Vibrio, Yersinia, Klebsiella, and Salmonella) ( Supplementary Fig. 5c) and residues immediately upstream of the pause site into account, we performed alanine-scanning mutagenesis. We found that Asn586, which is assumed to correspond to the A site of the paused ribosome, is crucial for pausing translation ( Supplementary Fig. 4d).
To gain evidence of ribosome pause at the expected site, we investigated the tRNAs held in the stalled ribosome. The ribosome-nascent chain complex (RNC) on the pause site was affinity-purified via His-tag on the nascent peptide, followed by a sucrose Fujita et al. p15 density gradient (Supplementary Fig. 5e-g). Sequencing of the tRNAs within the complex revealed that tRNA Asn GUU and tRNA Pro CGG, whose anti-codons correspond to the A and P sites of codons of the pause site, were enriched in the RNC (Fig. 5d).
Direct visualization of the stalled ribosome by single particle cryo-EM further supported the assignment of the paused codon. The structural features of the cryo-EM density corresponding to P-site tRNA in the non-rotated state 70S was compatible with peptidyl-tRNA Pro CGG (Fig. 5e, f), which was consistent with ribosome profiling, tRNAseq, and biochemical analysis (Fig. 4). Thus, we concluded that the ribosomes translating ycbZ mRNA were stalled on Pro585-Asn586 at their P-A sites (Fig. 5a, b).

mRNA secondary structure alters RNase accessibility
During the cryo-EM analysis of the stalled ribosomes on ycbZ mRNA, we could not find any rational structural feature to explain footprint extension in ribosome profiling; the ribosome was non-rotated state with P-site tRNA. Thus, we reasoned that, rather than ribosomes, ycbZ mRNA has an element that alters RNase accessibility and that, analogous to the SD/anti-SD duplex, the secondary structure of mRNA may be such a block. High base pairing of ycbZ mRNAs upstream of the pause site was observed in DMS-seq (Burkhardt et al., 2017) (Fig. 5g). In addition, RNA secondary structure prediction by CentroidFold (Sato et al., 2009) indicated base pairing in footprint extended region ( Supplementary Fig. 6a). Similar duplex formation was also found in other sites with drug-independent long footprints ( Supplementary Fig. 6a, b). Taken together, our results showed the diverse origins of footprint extension, primarily ribosome subunit rotation and secondarily mRNA secondary structure (Fig. 6).

Discussion
Since development of ribosome profiling (Ingolia et al., 2009), the method has been used to address three major issues in cells: translation efficiency (over-or under-representation of footprint counts vs. RNA-Seq on CDS), coding region assignment, and elongation speed (Ingolia, 2014;Brar and Weissman, 2015;Fujita et al., 2019). In addition, our work indicated that by focusing on footprint length, ribosome profiling allows us to discuss the state of ribosome rotation during elongation in cells.
In eukaryotes, ribosome footprints show two distinct populations with peaking at sizes of ~22 nt and ~29 nt (Lareau et al., 2014;Matsuo et al., 2017;Kurihara et al., 2018;Wu et al., 2019). Indeed, a recent study reported that the footprint sizes represent differential RNase accessibility at the 3′ end originating from the accommodation of tRNA at the A site: the tRNA extends the footprint toward the 3′ end of footprints (Wu et al., 2019). In contrast, the bacterial ribosome footprint may not be reflected by A-site tRNA since the extended direction is the 5′ side rather than the 3′ end. Thus, the direction and molecular basis of footprint extension are essentially disparate in prokaryotes and eukaryotes. Notably, ribosomal protein S1 is found only in prokaryotes; thus, 5′ end variability could not occur in eukaryotes.
As shown in this study, the footprint length is directly associated with the rotation of ribosomal subunits in E. coli. In eukaryotic ribosome footprints, this is a challenge since the pre-peptidyl transfer state (non-rotated ribosome with tRNA at the A site) and pre-translocation state (rotated ribosome with leaned tRNA at the A/P site) can both form long footprints and thus could not be distinguished. However, the cocktail of cycloheximide and tigecycline, which inhibits tRNA translocation from the P site to the tRNA accommodation (Jenner et al., 2013), respectively, allows minimization of the fraction of pre-peptidyl transfer (non-rotated) states and eventually increases the pretranslocation state (rotated) ribosome fraction in long footprints (Wu et al., 2019).
During the study of ribosome pauses with long footprints, we found that translation paused on the CDS of ycbZ. The current cryo-EM structure of the paused ribosomes could not determine the origin of the slowdown of the elongation cycle at this site, possibly because of the averaged view of multiple states of the isolated ribosomes.
Moreover, the clear drop-off of A-site Asn-tRNA, which was assessed by tRNA-seq, also hampered our understanding. Although the YcbZ remains uncharacterized, its role in the suppression of stop codon readthrough was suggested (Gagarinova et al., 2016). Future studies regarding the molecular function of the YcbZ protein and the role of pause during its synthesis are warranted.
In conclusion, the states of ribosomes during the elongation cycle revealed by footprint length provide a useful framework and resource to identify the translational landscape in bacteria.

E. coli strains and DNA constructs
The E. coli strains and plasmids used in this study are listed in Supplementary Table 2 and Supplementary Table 3. The ASKA library clones (Kitagawa et al., 2005) and Keio collection strains (Baba et al., 2006) were kind gifts from the National Institute of Genetics (NIG), Japan. P1 bacteriophage-mediated transduction was conducted to introduce the ycbZ deletion mutation in the Keio collection strain JW0938-KC (Baba et al., 2006) into MG1655, and then the FRT-Km R -FRT cassette was removed from the chromosome of the ycbZ deletion strain using pCP20 (Datsenko and Wanner, 2000).
DNA fragments encoding 85 amino acids of ftsQ, the HRV 3C cleavage site, and the HAtag were inserted into the pCA24N-ycbZ motif, to fuse in-frame with the ycbZ motif, by In-Fusion HD (TaKaRa). Tandem stop codons were further introduced downstream of the stop codon of the ycbZ motif.

and inserted into pCI-neo-Fluc-EGFP by
In-Fusion HD (TaKaRa). The inserted EGFP was placed upstream of Fluc to fuse inframe. For pCI-neo-2xEGFP-Fluc-EGFP, two tandem EGFPs fragments were inserted.

Ribosome profiling
Cell sampling and library preparation were performed as previously described (McGlincy and Ingolia, 2017) with some modifications. A saturated culture of E. coli MG1655 cells, which were grown in LB media overnight at 37°C, was diluted at a 1:100 ratio into fresh LB media and grown at 37°C to an OD600 of 0.4-0.5. Cells were collected by filtration After depleting the reads originating from noncoding RNAs, the remaining reads were mapped to the E. coli genome sequence (NC_000913.2). Empirically, we defined an A-site position, which is essentially the same as the 3′ assignment as described previously (Woolstenhulme et al., 2015). Polarity scores were calculated as previously described (Schuller et al., 2017).

tRNA-seq
tRNAs from in vitro translation reactions and RNCs (prepared as described below) were purified using a mirVanaTM miRNA Isolation Kit (Thermo Fisher Scientific). After the deacylation reaction, RNAs were dephosphorylated, linker-ligated, circularized, and

S1 irCLIP-seq
The irCLIP linker was prepared as previously described ( Reads were analysed as described in the "Ribosome profiling" section. Reads at a given position were normalized by average reads per nucleotide at each gene.

DMS-seq data analysis
DMS-seq data from E. coli K-12 MG1655 (GSE77617) (Burkhardt et al., 2017) were processed as described in the "Ribosome profiling" section. The score at a given position was the average of reads in a 33-nucleotide window. Data obtained at 37°C were compared to those at 95°C.

Cryo-EM single particle analysis
Sample preparation E. coli cells were collected and lysed as described in the "Ribosome profiling" section. magnification of x23,500, which is corresponding an objective pixel size at the specimen level of 1.47 Å. It should be noted that the data collection of specimens with antibiotics were performed on super-resolution mode. Therefore, pixel size became two times finer, which is corresponding to 0.735 Å shown in Table S1. The total exposure of electrons at the specimen level was around 50e-Å-2 and dose fractionated into 40 frames (Table S1).
Images were recoded as movie micrographs. Image processing were performed using RELION3 (Zivanov et al., 2018). Details of image processing are described as follows.
1) The dataset of the chloramphenicol-bound E. coli 70S ribosome Obtained movie micrographs in super-resolution mode were motion corrected using the program implemented in RELION with applying 2x binning by fourier cropping. CTF parameters were estimated by CTFFIND4.1 program (Rohou and Grigorieff, 2015).
Obtained 165K particles were initially subjected to 2D classification to discard 30S and junk particles ( Supplementary Fig. 2a). 157K remained particles were autorefined to reconstruct a consensus map. Subsequently, based on the alignment of consensus reconstruction, 3D classification was performed ( Supplementary Fig. 2a). As a result of 3D classification, the data set was classified into three subgroups: 70S in the rotated state (35%), dissociated 50S (5%), and 70S in the non-rotated state (60%). Rotated (55K particles) and nonrotated (94K particles) 70S subgroups were further auto-refined.
In addition, CTF-refinement, Bayesian polishing, and another round of auto-refinement were performed. Eventually, 70S ribosomes with chloramphenicol bindings in the rotated or nonrotated states both at resolution of 3.0 Å were obtained ( Supplementary Fig. 2a, c).
Local resolution distribution on obtained cryo-EM maps were calculated using the program implemented in RELION ( Supplementary Fig. 2e, f).
2) The dataset of the capreomycin-bound E. coli 70S ribosome Movie micrographs recorded on super-resolution mode were motion-corrected using the program implemented in RELION by 2x binning as described above. CTF parameters were estimated by CTFFIND4.1 program (Rohou and Grigorieff, 2015). Automated particle picking was performed with Gautomatch program. Obtained 158K particles were 2D classified to discard 30S and junk particles ( Supplementary Fig. 2b). Remained 154K particles converged to a consensus reconstruction by auto-refinement. Using this alignment information, the dataset was subject to 3D classification ( Supplementary Fig.   2b). The dataset was sorted into three subgroups: dissociated 50S (3%), 70S in the nonrotated state (41%), and 70S in the rotated state (52%). A subgroup of the rotated 70S (81K particles) was further auto-refined. Subsequently, CTF-refinement, Bayesian polishing, and auto-refinement was performed. Final reconstructed structure was at the resolution of 3.0 Å ( Supplementary Fig. 2b, c). Local resolution was estimated using the program implemented in RELION ( Supplementary Fig. 2g).
3) The dataset of the ycbZ-stalled E. coli 70S ribosome Movie micrographs were performed using the program implemented in RELION. CTF parameters were estimated by CTFFIND4.1 program (Rohou and Grigorieff, 2015).
Particle picking was performed with Gautomatch program. Obtained 77K particles was subject to 2D classification to discard 30S and junk particles. Remained 63K particles were auto-refined to a consensus reconstruction. The dataset subject to 3D classification.

Fujita et al. p31
Particles of 70S in the non-rotated state represented the major population ( Supplementary   Fig. 6c). This data set was further auto-refined and subsequently CTF refinement, Bayesian polishing, and another auto-refinement were performed. The obtained final reconstruction was at the resolution of 3.3 Å (Supplementary Fig. 2c, 6c). Local resolution distribution was calculated using the program implemented in RELION ( Supplementary Fig. 2h).

Integrated nascent chain profiling (iNP)
iNP was performed as previously described (Chadani et al., 2016). The sample was run on Bolt 12% Bis-Tris Plus gel (Thermo Fisher Scientific) with Bolt MES SDS Running Buffer (Thermo Fisher Scientific).
Preparation of the S1-depleted 70S ribosomes S1 was biochemically depleted from ribosomes as previously described (Duval et al., 2013)  In vitro translation was conducted with PURE flex 1.0 (GeneFrontier), using S1-depleted or mock-treated ribosomes, at 37°C for 30 min. Luminescence was detected with a dual-luciferase reporter assay system (Promega) and GloMax (Promega).
(d) Read enrichment of S1 irCLIP with capreomycin (cap) over chloramphenicol (cm) treatment, around stop codon. The 5′ end of reads is shown. The enriched site, which is corresponded to the footprint extension at the stop codon, is indicated by black arrow. (a) The difference of polarity scores of footprints in S1 ∆56 strain compared S1 WT strain.
The mRNAs were sub-classed into four groups according to the CDS length.
(b) Schematic representation of reporters constructed (top). sodB 5′ UTR, which enables S1-independent translation initiation, is fused to firefly luciferase (Fluc). Single or tandem GFPs were inserted to extend CDS length. These reporters were translated by S1-    During bacterial ribosome profiling, short footprints (23-27 nt) are originated from nonrotated ribosomes, which could be trapped by chloramphenicol. Long footprints (35-40 nt) are generated by subunit rotation. Rotated ribosome (stabilized by capreomycin) leads to high probability of mRNA association by S1, which sterically inhibits a RNase attack.
A minor fraction of long footprints are originated by intra-molecular and inter-molecular (SD/andi-SD) base-paring, that are resistant to RNase digestion. (a) Image processing of 70S ribosomes with chloramphenicol addition. Briefly, 165K

Supplementary
particles were extracted from motion-corrected movie-micrographs followed by 2D classification to discard bad particles. Remained 157K particles were auto-refined to obtain the consensus reconstruction. Based on the alignment information of this Fujita et al. p38 reconstruction, 3D classification with local angular search was performed. Classification was converged into three structurally homogeneous subgroups; 50S (5%), a rotated 70S (35%), and a non-rotated 70S (60%). Rotated and non-rotated 70Ss were further processed through auto-refinement, CTF refinement, bayesian polishing, and autorefinement. Obtained cryo-EM structures reach 3.0 Å in the rotated 70S and 3.0 Å in the nonrotated 70S.
(b) Image processing of 70S ribosomes with capreomycin addition, performed in the same way as (A). Briefly, the extracted 158K particles were 2D classified to obtain good particles (154K particles). Remained particles were converged into the consensus reconstruction. Subsequently, 3D classification was performed and converged into three subgroups, 50S (3%), a rotated 70S (52%), and a non-rotated 70S (41%). Since the nonrotated ribosomes were tRNA-free, we focused on the rotated ribosomes. The rotated 70S with two tRNA binding were further processed as the same way described above. The obtained 70S ribosome in a rotated state is at the resolution of 3.0 Å.
(c and d) Fourier shell correlation curves of four cryo-EM structures (c) and models vs.  Fig. 3: Characterization of S1 irCLIP-seq.
(b) Western blot for S1 protein in the mutant strains.
(c) Immunopurified S1 protein ( (c) Conservation of the ycbZ pause motif (the amino acid position in E. coli is shown).
Mutated amino acids are indicated by arrowheads.
(d) Western blot for peptidyl-tRNA (probed by HA epitope) from the reporters run in natural pH SDS-PAGE.
(e) Schematic representation of the RNC purification procedure.
(f) Western blot for peptidyl-tRNA probed by HA-tag (left) and CBB staining of proteins in RNC used for cryo-EM analysis (right).
(g) Sucrose density gradient of immunopurified RNCs. The 70S fraction was used for downstream cryo-EM analysis.
Supplementary Fig. 6: Purification of RNCs on the ycbZ pause site.
Fujita et al. p41 (a) RNA secondary structure prediction of 60 nt from A-site at pause site with long footprint accumulation by CentroidFold (Sato et al., 2009). Footprint extended regions are shaded in red.
(b) RNA base-paring around pause sites (indicated by arrows) with long footprint accumulation, probed by DMS-seq (Burkhardt et al., 2017). Reads in a 33 nt window were summed and normalized. Red shades show footprint extended region upstream of A site.
(c) Image processing of 70S ribosomes stalled with ycbZ. 77K particles were extracted followed by 2D classification to obtain good particles (63K particles). The dataset was converged into the consensus reconstruction and 3D classification was performed. The dataset classified into six subgroups. The major population (64%) assigned to 70S ribosome in a non-rotated state was further processed. The obtained refined cryo-EM structure is at the resolution of 3.3 Å.  Duval et al., 2013