Drop-off-reinitiation triggered by EF-G-driven mistranslocation and its alleviation by EF-P

Abstract In ribosomal translation, peptidyl transfer occurs between P-site peptidyl-tRNA and A-site aminoacyl-tRNA, followed by translocation of the resulting P-site deacylated-tRNA and A-site peptidyl-tRNA to E and P site, respectively, mediated by EF-G. Here, we report that mistranslocation of P-site peptidyl-tRNA and A-site aminoacyl-tRNA toward E and A site occurs when high concentration of EF-G triggers the migration of two tRNAs prior to completion of peptidyl transfer. Consecutive incorporation of less reactive amino acids, such as Pro and d-Ala, makes peptidyl transfer inefficient and thus induces the mistranslocation event. Consequently, the E-site peptidyl-tRNA drops off from ribosome to give a truncated peptide lacking the C-terminal region. The P-site aminoacyl-tRNA allows for reinitiation of translation upon accommodation of a new aminoacyl-tRNA at A site, leading to synthesis of a truncated peptide lacking the N-terminal region, which we call the ‘reinitiated peptide’. We also revealed that such a drop-off-reinitiation event can be alleviated by EF-P that promotes peptidyl transfer of Pro. Moreover, this event takes place both in vitro and in cell, showing that reinitiated peptides during protein synthesis could be accumulated in this pathway in cells.


INTRODUCTION
In ribosomal translation, nascent polypeptide chains are elongated by repeating the following three fundamental steps: (i) peptidyl transfer from the P-site peptidyl-tRNA onto the A-site aminoacyl-tRNA catalyzed by the peptidyl transferase center of ribosome, (ii) translocation of the resulting P-site deacylated-tRNA and A-site peptidyl-tRNA to the E site and P site, respectively, mediated by EF-G and (iii) accommodation of a new aminoacyl-tRNA onto the empty A site mediated by EF-Tu (1). Although the P-site peptidyl-tRNA is indispensable for the peptidyl transfer re-action, the peptidyl-tRNA can be lost under certain conditions by drop-off from the ribosome, leading to truncation of peptides (2)(3)(4)(5). For instance, elongation of inefficient substrates such as consecutive L-proline (Pro) residues (6)(7)(8)(9) or some kinds of nonproteinogenic amino acids, e.g. ␤-, D-and N-methyl-amino acids, often suffers from peptidyl-tRNA drop-off (5). Moreover, since peptidyl-tRNA is stabilized at the P-site by interaction with the ribosomal exit tunnel, ones with short nascent peptides are more prone to drop-off due to insufficient interaction with the tunnel. Thus, drop-off is promoted by some macrolide antibiotics that hinder the interaction between the nascent peptide and the exit tunnel (10)(11)(12)(13)(14).
The dropped peptidyl-tRNA is eventually hydrolyzed by peptidyl-tRNA hydrolase (PTH) at the ester bond between peptide and 2 /3 -hydroxyl group of tRNA (15), ending up with generation of a truncated peptide lacking the Cterminal region, which we refer to as 'drop-off peptide (DoP)'. On the other hand, the fate of the ribosome that has lost the peptidyl-tRNA is unclear due to the lack of comprehensive study of such a complex. Given the aminoacyl-tRNA as well as the mRNA still remain in ribosome, it is possible that the translation reinitiates by the migration of the aminoacyl-tRNA from A site to P site along with the mRNA, followed by accommodation of a new aminoacyl-tRNA onto the empty A site. In that case, a truncated peptide lacking the N-terminal region, which we call 'reinitiated peptide (RiP)', could be generated utilizing the remaining aminoacyl-tRNA as an N-terminal building block. Indeed, two related papers published by our group indicated such a possibility that not only peptidyl-tRNA drop-off but also RiP synthesis could occur (5,16). However, the mechanism how drop-off-reinitiation occurs to generate such RiPs remains elusive.
As for the translocation of the remaining aminoacyl-tRNA from A site to P site, EF-G would be the possible candidate responsible for such an event, similar to the case with the canonical translocation (17). Based on the fact that EF-G directly interacts with only A-site tRNA (18), we assumed that EF-G moves the A-site aminoacyl-tRNA toward the P site, and thereby the P-site peptidyl-tRNA is pushed out ester (Asn-DBE), L-glutamate-3,5-dinitrobenzyl ester (Glu-DBE), L-leucine-3,5-dinitrobenzyl ester (Leu-DBE), L-isoleucine-3,5-dinitrobenzyl ester (Ile-DBE), L-phenylalanine-cyanomethyl ester (Phe-CME), Lmethionine-3,5-dinitrobenzyl ester (Met-DBE), Nbiotinylated L-phenylalanine-cyanomethyl ester ( Bio F-CME) and D-alanine-3,5-dinitrobenzyl ester (D-Ala-DBE)] were synthesized by previously reported methods (19,20). Aminoacylation was carried out at 0˚C in a reaction mixture containing 50 mM HEPES-KOH (pH 7.5), 600 mM MgCl 2 , 20% DMSO, 25 M dFx or eFx, 25 M tRNA and 5 mM activated amino acids. eFx was used for aminoacylation of Phe-CME and Bio F-CME and dFx for that of other amino acids. Reaction time was 2 h for all the activated amino acids. The aminoacyl-tRNAs were recovered by ethanol precipitation, and then pellet was washed twice with 70% ethanol containing 0.1 M sodium acetate (pH 5.2), once with 70% ethanol and dissolved in 1 mM sodium acetate (pH 5.2).

Preparation of EF-P for in vitro translation reactions
Escherichia coli efp gene was cloned into a modified pET28a(+) vector that has PreScission protease recognition site instead of thrombin site (Supplementary Table S1I). E. coli epmA and epmB genes were cloned into pETDuet-1 vector. These vectors were co-introduced into E. coli Rosetta2 (DE3) pLysS. The cells were cultured in LB medium with 0.5 mM IPTG for 2 h at 37˚C and lysed by sonication. The cell lysate was applied to HiTrap TALON crude column (Cytiva) to purify the histidine-tagged EF-P. The column was washed with buffer A (20 mM Tris-HCl (pH 8.0), 200 mM NaCl, 2 mM imidazole and 1 mM 2-mercaptoethanol) and then the histidine-tagged EF-P was eluted by buffer A containing 500 mM imidazole. Turbo3C protease was added to the eluate for cleaving the histidine-tag and dialyzed against the buffer A at 4˚C overnight. The sample was applied on the HiTrap TALON crude column, and the flowthrough and wash fractions were recovered as EF-P without histidine-tag. Then, the protein was concentrated by Amicon Ultra 10k centrifugal filter (Merck Millipore).

Expression of peptides in the reconstituted cell-free translation system
Peptides were expressed by utilizing modified FIT (Flexible in vitro translation system), in which unnecessary amino acids, aminoacyl-tRNA synthetases (aaRSs) and EF-P were withdrawn from the following mixture (See also supplementary Table S2 for the details of the translation conditions). The complete FIT system contains 2 M template mRNA, 0.5 mM each 20 proteinogenic L-amino acids (Ala, Cys, Asp, Glu, Phe, Gly, His, Ile, Lys, Leu, Met, Asn, Pro, Gln, Arg, Ser, Thr, Val, Trp, Tyr), 2 mM ATP, 2 mM GTP, 1 mM CTP, 1 mM UTP, 20 mM creatine phosphate, 50 mM HEPES-KOH (pH 7.6), 100 mM potassium acetate, 12.8 mM magnesium acetate, 2 mM spermidine, 1 mM DTT, 1 mM tris(2-carboxyethyl)phosphine (TCEP), 0.1 mM 10-HCO-H4folate, 1.5 mg/ml E. coli tRNAs, 1.  CAU without Met, MetRS and 10-HCO-H4folate or with 100 M Bio F-tRNA fMet CAU , 0.5 mM Met and MetRS without 10-HCO-H4folate. In expression of representative peptides with various kinds of nascent peptide sequences prior to the three consecutive prolines, peptides were expressed with 12.5 M Bio F-tRNA fMet CAU without 10-HCO-H4folate. The reaction mixture was incubated at 37˚C for 20 min. In the identification of dropped peptidyl-tRNA or its nascent chain, the translation reaction mixture was incubated with 10 M kanamycin (Nacalai tesque) for 5 min at 37˚C, followed by incubation with 1 M PTH for 10 min at 37˚C, if necessary. In separation of the dropped peptidyl-tRNA and the peptidyl-tRNA trapped in the ribosome, the reaction mixture was mixed with equivalent volume of MeOH, centrifuged for 3 min at 13 000 rpm, and the supernatant and the precipitation was separated. The precipitation was re-resolved into 1× FIT buffer (20 mM  In the identification by MALDI-TOF MS, the translation products were purified by anti-FLAG antibody if necessary. After addition of equal amount of 2× TBS buffer (100 mM Tris-HCl (pH 7.6) and 300 mM NaCl), the reaction mixture was incubated with anti-FLAG antibody agarose gel (Sigma) for 1 h at room temperature, followed by wash with 1× TBS buffer (50 mM Tris-HCl (pH 7.6) and 150 mM NaCl) twice and eluted with 0.1% TFA solution. Reaction mixture and elution were desalted with SPE C-tip (Nikkyo Technos) and eluted with 1.2 l of 80% acetonitrile and 0.5% acetic acid solution containing 50% saturated (R)-cyano-4-hydroxycinnamic acid (Bruker Dalton-ics). MALDI-TOF MS analysis was performed using ul-trafleXtreme (Bruker Daltonics) in reflector positive mode. Peptide calibration standard II (Bruker Daltonics) was used for the external mass calibration.
In the identification and quantification by LC-ESI MS, proteins were precipitated by incubation of the reaction solution with equal volume of methanol on ice for 5 min, centrifuged at 13 000 rpm for 3 min, and centrifuged with 4× volume of 1% TFA at 13 000 rpm for 3 min. The resulting solution was injected to the LC-ESI MS (Xevo™ G2-XS, Waters) assembled with Acquity UPLC BEH C18 column (1.7 m, 300Å, 2.1 × 150 mm, Waters) and eluted by gradient of water/acetonitrile containing 0.1% formic acid.

Profiling of nascent peptide sequence against peptidyl-tRNA drop-off and RiP generation
Nascent chain-dependent Pro/Gly incorporation efficiency was analyzed by an mRNA display-based profiling system. The cDNA of Bio F-peptide-mRNA/cDNA conjugates were eluted by 1× PCR mix by incubation at 95˚C for 5 min as 'Pull-down' fraction. After addition of Taq DNA polymerase into 1× PCR mixture, the amount of cDNA in the 'Initial' fraction and 'Pull-down' fraction was quantified by qPCR and amplified by PCR in optimum thermal cycles.
Nucleic Acids Research, 2022, Vol. 50, No. 5 2739 Amplified cDNAs were purified by phenol/chloroform extraction and ethanol precipitation. Adopter sequences for next generation sequencing were added to the purified cD-NAs by PCR and purified by NucleoSpin Gel and PCR Clean-up (Macherey-Nagel). The purified cDNAs were sequenced by Miseq and Miseq reagent kit v3 (150 cycles) (Illumina). The ratio of a nascent peptide chain n in the 'Initial' or 'Pull-down' fraction, P n(Ini) and P n(PD−/+) , enrichment of n in the presence or absence of EF-P through the selection process, E n-/+, and incorporation enhancement by EF-P, W n were calculated as following equations;  Table S1I). cDNAs of YhhM (purchased from Integrated DNA Technologies) were amplified by adopter PCR in order to add 15 nt homologous regions and purified by NucleoSpin Gel and PCR Clean-up (Macherey-Nagel). pBAD-HisA ColE1 ori plasmid (Thermo Fisher Invitrogen) were amplified with the same homologous region as those added to the cDNA of proteins by inversion PCR. PCR reaction contained 50% KODone premix (TOY-OBO), 0.3 M forward and reverse primers, and 1.00 ng/l linear or circular template DNA. PCR reaction was performed at 94˚C for 120 s, followed by repeat of 94˚C for 10 s, 56˚C for 30 s and 68˚C for 60 s/1 kb. DNAs were purified by 1w/v% agarose gel electrophoresis. The resulting cDNA and linear plasmids were assembled by In-Fusion HD cloning kit (Takara bio). Transformation was carried out with 50 l E. coli DH5␣ competent cells with 2.5 l of In-Fusion reaction mixture, incubated on ice for 10 min, heat-shocked for 45 s at 42˚C, incubated on ice for 2 min, and spread on LB plates containing 100 g/ml ampicillin at 37˚C for overnight. Single colonies were picked and cultivated in 4 ml LB broth with 100 g/ml ampicillin at 37˚C for overnight, and then plasmid was extracted by fast gene plasmid mini kit (Nippon Gene) and sequenced (Fasmaq). pBAD-P BAD -YhhM-fh plasmid was prepared in the same procedure shown above. pBAD-efp-p15A ori plasmid was prepared as follows. First, pBAD-efp-ColE1 ori plasmid was prepared by In-Fusion assembly (Takara bio) of an inversion PCR product of pBAD-HisA ColE1 ori plasmid (Thermo Fisher Invitrogen) and PCR product of efp gene from pET28a(+)efp plasmid. Transformation was carried out with 50 l E. coli DH5␣ competent cells with 2.5 l of In-Fusion reaction mixture, following the same procedure as preparation of pBAD-P BAD -YhhM-His 6 plasmid using ampicillin. Then a DNA fragment containing araC-P BAD promotor-efp was amplified by PCR of the pBAD-efp-ColE1 ori plasmid and a DNA fragment containing CmR-p15A ori was amplified by inversion PCR from pLysS plasmid, which were assembled by In-Fusion reaction. Transformation was carried out with 50 l E. coli DH5␣ competent cells with 2.5 l In-Fusion reaction mixture, following the same procedure as preparation of pBAD-P BAD -YhhM-His 6 plasmid using 30 g/ml chloramphenicol in LB. The recombinant plasmid was prepared and sequenced following the procedure shown above.

Complementation of E. coli cells lacking chromosomal efp gene by pBAD-efp-p15A ori plasmid
Expression and post-translational modification of EF-P was confirmed by using E. coli JW4107 strain lacking its chromosomal efp gene (E. coli efp) (National BioResource Project-E. coli). E. coli efp was transformed with pBAD-efp-p15A ori plasmid. The resulting E. coli efp/pBAD-efp-p15A ori (E. coli efp/+efp) was cultivated in 0.5 L LB with 50 g/ml kanamycin and 30 g/ml chloramphenicol for 4.5 h at 37˚C and expression of efp gene was induced by 0.02% L-arabinose for 4 h at 37˚C. The E. coli pellet was recovered by centrifugation at 8000 rpm for 10 min, flash-frozen and stored at −80˚C. The pellet was resuspended with lysis buffer (20 mM Tris-HCl (pH 7.4), 200 mM NaCl, 1 mM DTT and 0.1 mM PMSF), sonicated and centrifuged at 15 000 g for 10 min. The supernatant was filtered by Minisart NML GF and Minisart NML hydrophile (Sartorius) and applied to HiTrap TALON crude (Cytiva). Then, EF-P was eluted with imidazole gradient and digested by 100 ng/l Glu-C proteinase (Sigma) in 50 mM Tris-HCl (pH 8.0). After methanol precipitation and trifluoroacetic acid precipitation, the solution was analyzed by LC-ESI MS (Xevo TM G2-XS, Waters) assembled with Acquity UPLC BEH C18 column (1.7 m, 300Å, 2.1 × 150 mm, Waters) and eluted by gradient of water/acetonitrile containing 0.1% formic acid.

Expression and purification of YhhM-fh in E. coli efp and E. coli efp/+efp
After transformation of E. coli efp strain and E. coli efp/+efp stain by pBAD-P BAD -YhhM-fh plasmid, the resulting single colony was pre-cultivated in 2 ml of LB with 100 g/ml ampicillin and 50 g/ml of kanamycin for the E. coli efp stain and with additional 30 g/ml chloramphenicol for E. coli efp/+efp stain at 37˚C for overnight. The preculture broth was added to 100 ml of LB broth containing the same antibiotics, cultivated at 37˚C for 4 h, and induced with f.c. 0.02% L-arabinose at 37˚C for 4 h. E. coli pellet was recovered by centrifugation at 8000 rpm for 10 min, flash-frozen and stored at −80˚C. The pellet was resuspended with lysis buffer (20 mM Tris-HCl (pH 7.4), 200 mM NaCl, 1 mM DTT and 0.1 mM PMSF), sonicated and centrifuged at 15 000 g for 10 min. The supernatant was applied to pre-equilibrated 1 ml slurry of His Super Flow (Cytiva) and eluted by imidazole gradient. The fractions containing YhhM-fh were analyzed by glycine-SDS-PAGE. The band corresponding to YhhM-fh was eluted from the gel and purified by PD10 (Cytiva) in order to remove DTT. Then, YhhM-fh was further purified by anti-FLAG M2 agarose gel (Sigma) and concentrated by Amicon Ultra 10k MW cut-off (Merck Millipore).

Sequence identification of YhhM-fh digested with Lys-N by LC-ESI MS/MS
0.67 g of YhhM-fh expressed in E. coli efp and 1.3 g of YhhM-fh expressed in E. coli efp/+efp was digested by 0.027 g/l Lys-N (Thermo Fisher Invitrogen) with 0.05 M NH 4 HCO 3 (pH 8.2) in 30 l reaction mixture at 37˚C for 16 h. The reaction mixtures were desalted by SPE C-tip (Nikkyo Technos) and eluted with 30 l of 80% acetonitrile with 0.5% acetic acid solution. The eluates were analyzed by LC-ESI MS E and LC-ESI MS/MS (Xevo™ G2-XS, Waters) assembled with Acquity UPLC BEH C18 column (1.7 m, 300Å, 2.1 × 150 mm, Waters) and eluted by gradient of water/acetonitrile containing 0.1% formic acid. First, the parent ions were screened by MS E (Data-Independent Acquisition, DIA) method. m/z values corresponding to parent peptide fragments were screened by UniFi software (Waters). In order to sequence the N-terminal fragments, parent ions were fragmented by MS/MS mode. In MS/MS analysis, parent ions having a specific m/z value was selected by the quadrupole and fragmented by collision-induced dissociation using ramping of collision energy from 10 to 40 V. The daughter ions were identified by MassLynx software (Waters) after deconvolution of raw MS/MS spectra.

Peptidyl-tRNA drop-off and reinitiated peptide synthesis upon incorporation of consecutive Pro residues
To observe the EF-G-driven mistranslocation event, we conducted in vitro translation of mRNAs into model peptides containing consecutive Pro residues with titration of EF-G concentration. mRNA2 encodes a model full-length peptide FLP2 that have three consecutive Pro, whereas mRNA1 encodes a negative control peptide FLP1 that has three Gly in place of Pro ( Figure 1A). Since peptidyl transfer between consecutive Pro is extremely inefficient due to the poor abilities in peptidyl donor and acceptor (9,21,22), we expected to observe drop-off-reinitiation in translation of mRNA2 but not in mRNA1. Peptides were radioisotope-labelled by incorporating [ 14 C]-aspartic acid (Asp, D) in the C-terminal FLAG sequence (Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys) for quantification of product bands in tricine SDS-PAGE by autoradiography. In translation of mRNA1, only the full-length product FLP1 was observed [ Figure 1B (Figure 2A, B: mRNA1 and mRNA2). The drop-off peptide DoP2 was not detected because PTH is not added in this analysis and thus DoP2 should not be released from the tRNA. These results indicate that the mistranslocation event occurs with fMet-Lys-Lys-Lys-Pro-Pro-tRNA at P site and Pro-tRNA at A site; the former dropped off and the latter migrated to P site, then translation is reinitiated from the Pro-tRNA to generate the RiP2.
In order to analyze the effect of the number of consecutive Pro residues on mistranslocation, translation of mR-NAS1 and mRNAS2, expressing the full-length FLPS1 and FLPS2 containing one and two consecutive Pro residues, respectively, was conducted, and the products were compared to those of mRNA1 and mRNA2 ( Figures 1A, B, 2A, Supplementary Figure S1A). An intense peak of RiPS2 along with FLPS2 were observed for mRNAS2, similar to the products of RiP2 and FLP2 expressed from the mRNA2 template. On the other hand, the mRNAS1 template produced only the full-length peptide FLPS1. This indicates that two consecutive incorporations of Pro is less efficient than its single incorporation, promoting the RiP synthesis. In the case of mRNAS2 translation, the expression level of the full-length FLPS2 dramatically decreased at higher concentrations of EF-G [ Figure 1B: mRNAS2, EF-P (−)], indicating that EF-G indeed triggers peptidyl-tRNA drop-off and thereby inhibits the usual level of FLPS2 synthesis. The expression level of RiPS2 increased at such high concentrations of EF-G, indicating that EF-G mediates the migration of A-site aminoacyl-tRNA to P site for the reinitiation. Consequently, the ratio of the RiPs increased in an EF-G concentration-dependent manner. A similar tendency was also observed in the translation of mRNA2, in which the ratio of RiP2 increases at higher concentrations of EF-G [ Figure 1B: mRNA2, EF-P (−)]. These results show that both inefficient peptidyl transfer (i.e. consecutive Pro) and frequent translocation (i.e. high EF-G concentration) lead to the drop-off-reinitiation event. Since not only EF-G but also RF2, RF3 and RRF have been previously suggested to induce peptidyl-tRNA drop-off (23)(24)(25), we also titrated them in the translation of mRNA2; however, no significant enhancement in the RiP2 synthesis was observed with higher concentration of those factors ( Figure 1C). Thus, we concluded that RF2, RF3 and RRF were not involved in the drop-off-reinitiation event.
Next, we analyzed the effect of EF-P on the drop-offreinitiation. It has been previously reported that EF-P recognizes a specific D-arm motif found in the tRNA Pro isoacceptors and accelerates peptidyl transfer in the elongation of Pro (7,8,26). EF-P is also involved in alleviation of translocation defects arising from miscoding and frameshifting (27,28). In translation of mRNA2, EF-P greatly enhanced the expression level of FLP2 and reduced that of RiP2 ( Figure 1B Since EF-P requires ␤-lysinylation at Lys34 for its full activity, we also evaluated the effect of using an unmodified EF-P in translation of mRNA2 into FLP2 and RiP2 (Supplementary Figure S2). Consequently, even the unmodified EF-P could suppress RiP synthesis and enhance the ratio   Figure 2D. of FLP, although its effect was weaker than the use of ␤lysinylated EF-P. This result indicates that the EF-P in the E site blocks the movement of the P-site peptidyl-tRNA to the E site, thereby suppressing peptidyl-tRNA drop-off. Both the acceleration of peptidyl transfer and the blocking of the movement of the P-site tRNA should contribute to the suppression of drop-off-reinitiation event.
Since not only Pro but also some kinds of nonproteinogenic amino acids, such as D-amino acids, cause inefficient peptidyl transfer reaction, we also evaluated the frequency of reinitiated peptide synthesis in two-consecutive incorporation of D-Ala into FLPS2a using mRNAS2 ( Figure  3A). D-Ala was pre-charged onto an artificial suppressor tRNA, tRNA GluE2 CGG , using flexizyme and introduced at CCG codons by means of genetic code reprogramming in a reconstituted translation system that lacks Pro. As a result, two types of reinitiated peptides, RiPS1a and RiPS2, were detected by MALDI-TOF MS ( Figure 3B). RiPS1a and RiPS2 could not be separated on a tricine-SDS-PAGE gel due to the only single amino acid difference between them. Therefore, the sum of the expression level of these two peptides were estimated and compared to that of the full-length peptide, FLPS2a ( Figure 3C, D). Consequently, the ratio of the RiPs significantly increased at higher EF-G concentrations. This tendency was consistent with that of consecutive Pro incorporation.

Detection of drop-off peptides
In order to detect the drop-off peptide, translation of mRNA2 was conducted in the presence of [ 14 C]-Lys so that all of FLP2, DoP2 and RiP2 were radio-labeled, and then PTH (15) was added after 20-min translation to remove tRNA moiety of the dropped peptidyl-tRNA (DoP2-tRNA). We observed a band corresponding to DoP2 at high PTH concentrations (Supplementary Figure S1B, DoP2), while the intensity of a band corresponding to DoP2-tRNA decreased by elevation of the PTH concentrations (Supplementary Figure S1B: DoP2-tRNA). However, the band of DoP2-tRNA did not completely disappear at 1 M or above. We hypothesized that this is because unhydrolyzed [ 14 C]-Lys-tRNA Lys or peptidyl-tRNA protected from PTH by stalled ribosome overlapped on this band (29). In order to confirm this, not only PTH but also RNase A were added after 20-min translation, by which the PTHsensitive band completely disappeared ( Supplementary Figure S1C). This is probably because RNase A could decompose both dropped and ribosome-protected peptidyl-tRNAs. The band at the bottom that was observed in the presence of RNase A should be a peptidyl-adenosine (DoP2-adenosine) derived from the DoP2-tRNA (Supplementary Figure S1C: 50 g/ml RNase A). Then, the DoP2 was also analyzed by LC-ESI MS ( Figure 2D). Addition of RNase A gave two peaks bearing the same m/z values ( Figure 2D, 2, m/z = 335.844, relative peak intensities were 671 and 3244, indicated by black arrow-heads), both of which were identified as DoP2-adenosine molecules by LC-ESI MS/MS analysis ( Figure 2E). In the presence of PTH as well as RNase A, the intensity of DoP2-adenosine decreased by 10-fold, and instead a peak corresponding to DoP2 appeared ( Figure 2D, 2 and 3: Relative peak intensity of DoP2-adenosine decreased from 671 to 76 at RT = 5.74 min, and from 3242 to 325 at RT = 6.15 min. The intensity of DoP2 was 2485.). The expression of DoP2-adenosine was significantly suppressed in the presence of EF-P (Figure 2D: 2 and 5, relative peak intensity change from 671 to 300 at RT = 5.74 min, and from 3244 to 1268 at RT = 6.15 min).

Effects of length and sequence of nascent peptides on mistranslocation
As explained above, the length of nascent peptide of Psite peptidyl-tRNA should be one of critical determinants of drop-off frequency. Therefore, we examined different lengths of nascent peptides with EF-G titration ( Figure  4A: FLPL1-4). Peptides were [ 14 C]-Asp-labelled in the Cterminal FLAG, and analyzed by tricine-SDS-PAGE and autoradiography ( Figure 4B, C). In the absence of EF-P, the expression of FLPL1-4 yielded the corresponding full-length peptides along with RiP2 ( Figure 4B: FLPL1-4, grey bars; RiP2, red bars). At the high EF-G concentration (3 M), shorter peptides, FLPL3 and 4, exhibited low FLPL ratios [FLPL3/(FLPL3 + RiP2) and FLPL4/(FLPL4 + RiP2), respectively], indicating more frequent peptidyl-tRNA drop-off due to their insufficient stabilization via interaction with the exit tunnel. On the other hand, longer peptides, FLPL1 and 2, exhibited significantly higher FLPL ratios. (Figure 4C: ratio of FLPL1, 2, 3 and 4 were 75, 78, 44 and 43% at 3 M EF-G). The identities of FLPL1-4 and RiP2 were also confirmed by MALDI-TOF MS using non-radio-labeled peptides ( Figure 4D).
It has been also reported that certain nascent peptide sequences interact with the exit tunnel so strongly as to induce ribosomal stalling and slow down the peptidyl transfer (30). There is also an argument for a possibility that purinerich Shine-Dalgarno (SD)-like mRNA ORF sequences interact with the pyrimidine-rich anti-SD sequence of 16S rRNA to cause ribosome stalling (31,32). Therefore, we next analyzed the effect of such nascent peptide and corresponding mRNA sequences. mRNA2 is translated into FLP2 containing positively-charged Lys-Lys-Lys residues prior to the Pro-Pro-Pro, which is encoded by a purine-rich SD-like sequence (AAGAAGAAG in Figure 5A). On the other hand, mRNA6 is translated into FLP5 bearing hydrophobic residues (Leu-Ile-Phe) encoded by non-SD-like pyrimidine-rich codons (CUCAUCUUC) followed by three consecutive Pro. Translation of mRNA6 into FLP5 yielded only small amount of RiP2 [ Figure 5C . We also confirmed that the ratio of reinitiated peptide increased at higher EF-G concentra-  Figure 5F, mRNA6/FLP5, EF-P(−)] and decreased in the presence of EF-P [ Figure 5D, E, F: mRNA3/FLP3, mRNA2/FLP2, mRNA5/FLP4 and mRNA6/FLP5, EF-P(+)]. These results indicate that polar amino acids encoded by such SD-like mRNA sequences, mRNA2/FLP2 and mRNA5/FLP4, induce the mistranslocation event. However, it was yet unclear which is more or the most critical factor, mRNA context or peptide sequence. Therefore, we changed the relationship between codons and amino acids by means of genetic code reprograming (33) (Figure 5G), by which FLP3 bearing hydrophobic Leu-Leu-Leu was expressed from both the non-SD-like mRNA3 and SDlike mRNA4 ( Figure 5C: lane 1 and 2, Figure 5E, H) and FLP5 with hydrophobic Leu-Ile-Phe was expressed from both the non-SD-like mRNA6 and SD-like mRNA7 (Figure 5C: lanes 7 and 8, Figure 5E, H). Ribosomal syntheses of FLP3 or FLP5 bearing the hydrophobic residues yielded the full-length peptide regardless of the mRNA context. Likewise, regardless of the context of mRNA template, ribosomal syntheses of FLP2 and FLP4 bearing the polar Lys-Lys-Lys and Asn-Lys-Glu residues, respectively, resulted in a noticeable amount of truncated RiP2 along with the full-length peptides ( Figure 5C: lanes 3, 4, 5 and 6, Figure 5E, H). These results showed that induction of the drop-off-reinitiation event could be dictated by the peptide sequence rather than the mRNA context ( Figure 5B). We propose a possibility that the electrostatic interaction between polar/charged nascent peptides and the exit tunnel may hold the peptidyl-tRNA at an inactive position in the peptidyl transferase center (PTC) and cause inefficient peptidyl transfer reaction.

Comprehensive analysis of nascent peptide sequences that induce drop-off-reinitiation
To further explore nascent peptide sequences that induce drop-off-reinitiation, we applied a means of an mRNA display-based profiling (34). We constructed an mRNA library bearing 1-3 NNN random codons preceding three consecutive CCG Pro codons followed by a linker sequence ( Figure 6A: a mixture of mRNA libraries 1, 2 and 3), where the N-terminal N-biotinyl-phenylalanine ( Bio F) was installed by the initiation reprogramming. The peptide library expressed from this mRNA library was designed to contain 8,420 different peptide sequences to which 3 end of the mRNA library was ligated via a puromycin linker for the formation of peptide-mRNA fusion ( Figure 6B). We then translated the mRNA sequences with 0.03, 0.26 or 10 M of EF-G in the absence or presence of EF-P, producing the respective mRNA-peptide fusions followed by reversetranscription. Since only the full-length peptide-mRNA fusions should have the N-terminal Bio F, they were selectively recovered by streptavidin beads via the N-terminal biotin. Thereby, the deep sequencing of such cDNAs would reveal the populations of fully expressed peptide sequences.
By processing the sequencing data, proportion of a particular peptide sequence 'n' in the initial library was estimated as P n(Ini) , and that in the pulled-down library ob-  (mRNA2-7). The canonical genetic code assigns hydrophobic amino acids (sky-blue) to pyrimidine-rich codons (sky-blue), and polar amino acids (orange) to purine-rich codons (orange). Reprogrammed genetic codes (see also G) assign hydrophobic amino acids (sky-blue) to purine-rich codons (orange), and polar amino acids (orange) to pyrimidine-rich codons (sky-blue). The lysine residues that appear in the FLAG sequences are assigned to the codons   , green), followed by 1-3 residues of any kinds of the 20 proteinogenic amino acids (X 2 , X 3 and X 4 , blue), three consecutive prolines (P, red), and a linker peptide. The initiator AUG codon was reprogrammed to Bio F, whereas the elongator AUG codon was assigned to Met. See also Supplementary Figure S3 for the details. UAG codon was made vacant to conjugate an mRNA and the corresponding peptide via Puromycin (Pu). (B) Schematic depiction of the comprehensive analysis of the effect of nascent peptide sequences on the three consecutive Pro incorporation. Puromycin-ligated mRNA libraries were translated with 0.03, 0.26 or 10 M EF-G in the absence or presence of 5 M EF-P, and then conjugated with the corresponding peptides via puromycin and reverse-transcribed. The resulting Bio F-peptide-mRNA-cDNA conjugates were selectively pulled down by streptavidin magnetic beads. Truncated peptides caused by drop-off-reinitiation were not recovered due to the lack of N-terminal Bio F. The pull-downed cDNA was amplified by PCR and sequenced. Proportion of a particular sequence 'n' in the initial library was estimated as P n (Ini) , and that in the pulled-down library obtained in the absence or presence of EF-P was estimated as P n(PD−) or P n(PD+) , respectively. Then, the enrichment of a sequence n (E n− and E n+ ) were calculated by normalizing the P n(PD−) and P n(PD+) by P n(Ini) . The suppression of drop-off by EF-P was defined as W n = E n+ − E n− . RT: reverse transcription, SAV: streptavidin. (C-E) Histograms of E n− , E n+ , and W n sorted by the amino acid property of position X 2 , X 3 and X 4 with 0.26 M (C), 0.03 M (D), or 10 M (E) EF-G. Histograms were separately drawn with red if X i is proline (P), with orange if X i is polar amino acid (D, E, R, K, H, Q, N), with grey if X i is small amino acid (S, T, G, C, A), or with sky-blue if X i is hydrophobic amino acid (V, I, L, M, F, Y, W). tained in the absence or presence of EF-P was estimated as P n(PD−) or P n(PD+) , respectively. Then, enrichment of the peptide sequence n after pull-down compared to the initial library was evaluated as E n− and E n+ , which were calculated by log 2 [P n(PD−) / P n(Ini) ] and log 2 [P n(PD+) / P n(Ini) ], respectively. These values reflect the efficiency of full-length peptide synthesis for the particular sequence n. The effect of EF-P on suppression of drop-off-reinitiation was evaluated as W n = E n+ − E n− . Then, E n− , E n+ and W n of all peptide sequences were summarized and visualized in heat maps (Supplementary Figure S4A-C). Those values were also plotted in histograms ( Figure 6C-E, Supplementary Figure S5A, left), in which all peptides were categorized into four groups in terms of the property of amino acids at X 2 , X 3 and X 4 positions [polar residues (DERKHQN), small residues (STGCA), hydrophobic residues (VILM-FYW) and Pro (P)].
The average E n values of peptides bearing Pro at X 2 , X 3 and X 4 were always less than zero regardless of EF-G concentration and presence of EF-P, indicating that Pro at these positions strongly promote drop-off-reinitiation ( Figure 6C Regarding other types of amino acids than Pro, no significant tendency in average E n values was observed at positions X 2 and X 3 , whereas at position X 4 polar residues showed significantly lower distribution in E n− , indicating that drop-offreinitiation is promoted by polar residues at position X 4 in the absence of EF-P ( Figure 6C Since E n+ at position X 4 did not show any polarity-dependent difference ( Figure 6C-E), drop-offreinitiation can be suppressed by EF-P, which resulted in higher W n values of polar residues.
In order to confirm the reliability of this profiling system, 6 representative sequences with the highest E n− and 8 sequences with the lowest E n− were individually translated in vitro at 10 M EF-G in the absence or presence of EF-P, and quantified by LC-ESI MS (Supplementary Figure  S5B, indicated by black arrows, S5E). The sequences with high E n− could be efficiently translated into full-length peptides along with only a trace amount of reinitiated peptide RiPX even in the absence of EF-P (Supplementary Figure  S5F, blue sequences, S5B). On the other hand, the sequences with low E n− could not be translated into full-length at all in the absence of EF-P (Supplementary Figure S5F, orange sequences), but EF-P enabled expression of full-length peptides (Supplementary Figure S5G). We also performed sequence profiling by replacing the triple Pro of the mRNA library with a triple Gly, in which E n− and E n+ were less biased than those of the triple-Pro library, and EF-P had no effect as expected (Supplementary Figure S5A, right, S5D, S4D-F).

In-cell reinitiated peptide synthesis caused by drop-offreinitiation
Our in vitro studies described above undoubtedly have shown that the occurrence of drop-off-reinitiation led to an accumulation of the reinitiated peptide during the translation, so we wondered if this could be observed even in a cellular system. For this study, we chose five Escherichia coli proteins, YjgZ, PrpR, YhhM, RutD and YdcO, which have two or three consecutive Pro residues at the N-terminal region. We first set up in vitro translation experiments where the respective peptides corresponding to the N-terminal region (14-19 amino acid residues) of these proteins and observed their expression in the absence and presence of EF-P and PTH (Supplementary Figure S6A, FLPN1-5). The respective peptides were radio-labelled with [ 35 S]-methionine (Met or M) at the N-terminus and analyzed by tricine-SDS-PAGE and autoradiography. The bands whose intensities increased in the absence of EF-P and PTH were derived from the dropped peptidyl-tRNA (Supplementary Figure S6B, green arrows). In order to confirm the identities of both DoP and RiP, non-labelled peptides were also expressed in the absence and presence of EF-P (both in the presence of PTH) and subject to LC-ESI MS analysis (Supplementary Figure S6C-G). We observed DoPs originating from peptidyl-tRNA drop-off and RiPs originating from reinitiation at various positions. Notably, expression of FLPN1 (YjgZ), FLPN2 (PrpR) and FLPN4 (RutD) resulted in mainly full-length peptides and relatively small amounts of truncated peptides, whereas FLPN3 (YhhM) and FLPN5 (YdcO) showed severe peptidyl-tRNA drop-off and reinitiation, leading to accumulation of more DoPs and RiPs. Interestingly, in the expression of FLPN5 (MRLFSIPPPTLLAGFLAVL-FLAG), MRLF-SIP was the dominant DoP, indicating that peptidyl-tRNA drop-off mainly occurred at the first Pro residue; However, not PPTLLAGFLAVL-FLAG but PTLLAGFLAVL-FLAG and TLLAGFLAVL-FLAG were the dominant RiPs, indicating that the translation reinitiated from the third Pro or the next Thr by skipping one or two Pro residues. This is probably because the drop-off event occurred successively at the Pro-Pro-Pro sequence.
In this cell-based assay, an overexpressed protein, YhhMfh, was analyzed because the use of a specific tag sequence was necessary for the efficient isolation and detection of the fragments. However, the result has strongly supported a possibility where the drop-off-reinitiation could occur in the cellular system. The evolutionary importance of EF-P is also supported by this result, in turn suggesting why EF-P and its homologs are widely conserved in the three domains of life. Thus, the drop-off-reinitiation can frequently occur without EF-P, resulting in an accumulation of aberrant proteins that lack their N-terminal region.

DISCUSSION
Although the mechanism and processing pathway of peptidyl-tRNA drop-off event have been previously studied (2)(3)(4), the fate of the ribosome complex that lost the peptidyl-tRNA has not been well understood. Here, we reported for the first time that such a ribosome complex is subject to EF-G-dependent translocation and then reinitiation occurs to yield an RiP ( Figure 8B). We referred this event to as drop-off-reinitiation and confirmed that it occurs both in vitro and in cells. The frequency of drop-offreinitiation is regulated by the following two factors: (i) translocation frequency and (ii) peptidyl transfer efficiency ( Figure 8C).
EF-G is the critical regulator that determines translocation frequency (Figure 8C left). In the canonical translo-cation, EF-G mediates migration of P/E-site deacylated-tRNA and A/P-site peptidyl-tRNA in the hybrid state to E/E and P/P site, respectively, after completion of peptidyl transfer ( Figure 8A) (17). In contrast, in the mistranslocation, P/P-site peptidyl-tRNA and A/A-site aminoacyl-tRNA migrate toward E/E and P/P site, respectively, before peptidyl transfer completes ( Figure 8B). It has been previously reported that N-acetyl-Phe-tRNA in the P site can translocate toward E site, albeit less efficiently than the canonical translocation (41). Presumably, the nascent peptide on the peptidyl-tRNA perturbs the interaction between the tRNA and 23S rRNA in the E site and thereby makes the translocation less efficient (42,43). It has been also reported that a replacement of the peptidyl group of A-site tRNA with an aminoacyl group inhibits the translocation to P site (44). Moreover, since EF-G should bind preferably to the rotated ribosome/tRNAs in the hybrid state compared to those in the non-rotated classical state, the mistranslocation would not efficiently take place at a low concentration of EF-G. Therefore, an elevation of the EF-G concentration should be required to induce the mistranslocation with a higher frequency, yielding more RiPs. The concentration of EF-G in E. coli cell is estimated to be approximately twofold higher than that of 70S ribosome (45), which corresponds to 2.4 M in the case of our in vitro translation system whose concentration of ribosome is 1.2 M. In this EF-G concentration range, peptidyl-tRNA drop-off is induced but suppressed by EF-P in vitro ( Figure  1B). A similar trend was observed in the in-cell experiment, in which the YhhM-fh lacking its N-terminus was detected but largely suppressed by EF-P.
Susceptibility of P-site peptidyl-tRNA to drop-off is another regulator of the drop-off-reinitiation frequency (Figure 8C left). Shorter nascent peptides of peptidyl-tRNA increased the drop-off susceptibility due to the insufficient interaction with the exit tunnel, and resulted in more RiP synthesis ( Figure 4). In E. coli cells, there are 2115 proteins and putative proteins containing two or more consecutive Pro, 6% of which are located within N-terminal 24 amino acid residues (46). When the nascent peptide on peptidyl-Pro-tRNA is shorter than the depth of the exit tunnel, RiP synthesis possibly tends to occur via drop-off-reinitiation. If that is the case, the population of RiPs should not be ignorable, although more comprehensive in-cell studies should be performed to verify this.
As for the regulators of peptidyl transfer efficiency (Figure 8C right), we evaluated the effect of consecutive Pro residues in model peptides and proteins. Increasing the number of consecutive Pro, RiPs derived from mistranslocation were more frequently observed. The inefficient peptidyl transfer observed for the consecutive Pro elongation can be alleviated by adding EF-P to the translation system, by which drop-off-reinitiation was also suppressed. EF-P recognizes the P-site peptidyl-Pro-tRNA Pro and stabilizes its CCA end for shaping suitable geometry for peptidyl transfer (47). In addition to consecutive Pro, certain nascent peptide sequences that interact with the ribosomal exit tunnel also make peptidyl transfer inefficient. We conducted a comprehensive analysis of 8,420 nascent peptide sequences preceding the three consecutive Pro, showing that polar amino acids adjacent to the Pro stretch significantly  Ribosome catalyzes peptidyl transfer by which a new peptide bond is formed between a nascent peptide on P-site peptidyl-tRNA (blue) and an amino acid on A-site aminoacyl-tRNA (grey). In the canonical translocation mediated by EF-G (purple), P-site deacylated-tRNA and A-site peptidyl-tRNA migrate to E site and P site, respectively. In accommodation, an aminoacyl-tRNA corresponding to the next A-site codon is delivered by EF-Tu. By repeating the three reactions, a full-length peptide is synthesized. (B) Schematic illustration of P-site peptidyl-tRNA drop-off and generation of reinitiated peptide caused by mistranslocation and drop-off-reinitiation. Consecutive incorporation of prolines (P, red) makes peptidyl transfer reaction inefficient. Before completion of the inefficient peptidyl transfer, drop-off of the peptidyl-tRNA and migration of A-site aminoacyl-tRNA to P site occur, which we refer to as mistranslocation. Then, translation is reinitiated by accommodation of a new aminoacyl-tRNA on to A site to give a reinitiated peptide, RiP. (C) Candidates of translocation frequency regulators (1, left) and peptidyl transfer frequency regulators (2, right). Those indicated by red letters are confirmed translocation frequency regulators (short nascent peptide and EF-G) and peptidyl transfer frequency regulators (consecutive Pro incorporation, EF-P and sequence of nascent peptide).
promoted drop-off-reinitiation, which was also suppressed by EF-P ( Figure 6, Supplementary Figure S4A-C). Our result looks consistent with previous related studies where Pro and polar amino acids, such as Arg, Asp, His and Lys, were preferred at position X of the stalling motif, XPP or XPPP, although the contribution of polar residues was not clearly discussed in the papers (48)(49)(50). These studies also showed that the sequence dependent ribosome stalling was alleviated by EF-P. Not only consecutive Pro motifs but also other types of arrest peptides and ribosome destabilizing peptides have also been reported. For example, expression of TnaC stalls with the P-site peptidyl-Pro-tRNA (51), whereas that of SecM stalls with the A-site Pro-tRNA (52). MgtL peptide (MEPDPTPLPRRRLKLFR) and peptides bearing 10 consecutive acidic residues also destabilize ribosome upon Mg 2+ starvation, leading to a dissociation of the ribosomal subunits and eventually translation abortion (53). Although the mechanisms for the ribosomal stalling and destabilization are not fully understood, both Pro and polar/acidic residues in such peptides may force the peptidyl-tRNA inactive for peptidyl transfer reaction by electrostatic repulsion with the tunnel. Therefore, these peptide sequences also act as potent drop-off-reinitiation inducers ( Figure 8C right).
Drop-off-reinitiation induced by mistranslocation is distinct from other ribosome rescue pathways mediated by specific quality control factors, such as tmRNA/SmpB (54,55), ArfA (56) and ArfB (57). These rescue pathways do not generate RiPs from the stalled ribosome. Since both A and P sites are occupied by peptidyl-and aminoacyl-tRNAs, respectively, in the case of drop-off-reinitiation, these rescue factors cannot enter the A site where they work. Although we could detect RiPs in E. coli, it is still unclear whether such truncated peptides/proteins are rapidly and selectively degraded or can circumvent such degradation pathway to survive. If the N-end rule pathway applies to the truncated proteins (58), the lack of fMet and the alternate N-terminal amino acids (typically Pro) would play a role in protein stability. Elucidation of these issues should be important for future works.