The ABCF proteins in Escherichia coli individually cope with ‘hard-to-translate’ nascent peptide sequences

Abstract

Organisms possess a wide variety of proteins with diverse amino acid sequences, and their synthesis relies on the ribosome. Empirical observations have led to the misconception that ribosomes are robust protein factories, but in reality, they have several weaknesses. For instance, ribosomes stall during the translation of the proline-rich sequences, but the elongation factor EF-P assists in synthesizing proteins containing the poly-proline sequences. Thus, living organisms have evolved to expand the translation capability of ribosomes through the acquisition of translation elongation factors. In this study, we have revealed that Escherichia coli ATP-Binding Cassette family-F (ABCF) proteins, YheS, YbiT, EttA and Uup, individually cope with various problematic nascent peptide sequences within the exit tunnel. The correspondence between noncanonical translations and ABCFs was YheS for the translational arrest by nascent SecM, YbiT for poly-basic sequence-dependent stalling and poly-acidic sequence-dependent intrinsic ribosome destabilization (IRD), EttA for IRD at the early stage of elongation, and Uup for poly-proline-dependent stalling. Our results suggest that ATP hydrolysis-coupled structural rearrangement and the interdomain linker sequence are pivotal for handling ‘hard-to-translate’ nascent peptides. Our study highlights a new aspect of ABCF proteins to reduce the potential risks that are encoded within the nascent peptide sequences.

Graphical Abstract

Introduction

Numerous proteins in organisms show unique amino acid sequences that contribute to a wide range of cellular functions. This diversity in proteins is an essential driving force underlying the evolution of life. Therefore, the ribosome, responsible for protein synthesis, must synthesize an immense repertoire of amino acid sequences with high versatility. The ribosome itself has been considered very robust because of the variety of proteins expressed and functioning in the cells. However, recent advances have illuminated that the translation system is robust, but the ribosome itself is not universally capable of synthesizing all sequences with equal efficiency.

Ribosome-mediated translation elongation is believed to be controlled by mRNA sequences such as secondary structure and codon usage. In addition, decades of studies have demonstrated that the nascent polypeptide chain being synthesized by the ribosome itself also regulates translation. The nascent peptide interacts with the ribosomal exit tunnel and exerts various influences on ribosomal function (1–3) Particularly, in consecutive proline sequences, translation elongation is significantly hampered due to the inhibition of transpeptidation (4–6). However, organisms have evolved a solution to this challenge by obtaining a translation elongation factor EF-P (eukaryote: eIF5A), which rescues translation stalling at proline-rich motifs, enabling the synthesis of proteins containing such sequences (7–9). EF-P and eIF5A act on the peptidyl-tRNA within the P-site of the stalled ribosome from the vacant E-site, promoting transpeptidation and alleviating translation stalling (10,11).

In addition to the poly-proline sequence, translation of poly-basic (positive charge) or poly-acidic (negative charge) sequences is also challenging for the ribosome. When a polypeptide chain passing through the ribosome tunnel contains an abundance of positively charged amino acids, translation elongation can be hindered due to interactions with the negatively charged tunnel structure (12,13). Conversely, the translation of poly-acidic sequences can lead to difficulties in maintaining the intact ribosome complex conformation, resulting in stochastic premature termination termed intrinsic ribosome destabilization (IRD) (14–17). Thus, the growing nascent peptide can impact ribosomal activity through the interplay of ribosome structure and chemical properties.

On the other hand, several of the translation-inhibitory nascent peptides are harnessed to the survival strategies of the organisms. The nascent peptide of Escherichia coli SecM interacts with the ribosome tunnel, distorting the position of SecM peptidyl-tRNA itself (18,19). This inhibits the peptidyl-transfer reaction, resulting in the translation ‘arrest’. This phenomenon promotes the expression of downstream SecA translocase, and the expressed SecA releases the ribosome arrested by SecM in its translocase activity-dependent manner (20). Other arrest peptides that function as ‘monitoring substrates’ to sense membrane transport activity as well as SecM, are widely found among bacteria (21–23). Furthermore, such regulatory nascent peptides, are also found in eukaryotes (24,25), underscoring their efficacy in gene regulation.

While excepting for the regulatory nascent peptides, noncanonical translations such as translation stalling or IRD, are in general potential risks of protein synthesis. While EF-P acts on the translation of poly-proline sequences, it has not been shown whether other translation factors play similar roles in other ‘hard-to-translate’ sequences. In recent years, a subfamily of the ATP Binding Cassette subfamily-F, known as ABCF proteins that are conserved from bacteria to eukaryotes, has been reported to rescue the ribosome trapped by antibiotics (26–31). Such antibiotics-resistant (ARE)-ABCF proteins are constituted from the two nucleotide-binding domains (NBDs), the interdomain linker sequence, and C-terminal extension (CTE) (31). Once bound to the vacant E-site of the ribosome, similar to EF-P, the interdomain linker of ABCF proteins is inserted into the catalytic core of the ribosome. This rearranges the architecture of the peptidyl transferase center (PTC) and the tRNA within the P-site, effectively preventing the binding of drugs. Furthermore, it has been reported that EttA, one of the ABCFs in E. coli, interacts with and stabilizes P-site tRNA, thereby promoting early-stage translation (32,33). These imply that ABCF proteins have the potential to rearrange the catalytic core of translation, thereby rescuing ribosomes in various aberrant states. Although a variety of ABCF proteins have been identified across different organisms, ranging from E. coli to humans (31), their physiological functions remain elusive.

Antibiotics and ‘hard-to-translate’ nascent peptide sequences, while differing in detailed mechanisms, both ultimately inhibit translation by causing structural problems within the ribosome. Based on this similarity, we hypothesized that ABCF proteins might exert inhibitory effects on various noncanonical translations induced by nascent peptides. From a series of biochemical and genetic analyses using the E. coli translation system, we found that each of the four ABCF proteins in E. coli specifically alleviates different types of problematic nascent peptides, from potential risky sequences to the programmed arrest sequence, SecM. A class of ABCF proteins would increase the robustness of the translation system and contribute to the foundation of protein evolution.

Materials and methods

E. coli strains, plasmids and primers

Escherichia coli strains, plasmids, and oligonucleotides used in this study are listed in Supplementary Tables S1, S2, and S3, respectively. The KEIO collection library (34) was obtained from the National BioResource Project (NBRP). Plasmids were constructed using standard cloning procedures and Gibson assembly. Detailed schemes are summarized in Supplementary Table S2, and sequences of constructed plasmids are available in the Mendeley repository. Phage P1-mediated transduction was used to introduce the antibiotic-selection marker insertion mutation in ettA (JW4354), uup (JW0932), yheS (JW3315), ybiT (JW0804) or efp (JW4107) into BW25113. Subsequently, the FRT-Km^R-FRT cassette was removed from the chromosome of the transductants using pCP20, as described (35).

Purification of ABCF proteins

BL21 (DE3) harboring the plasmids encoding the N-terminally His₆-tagged YheS or C-terminally His₆-tagged EttA, Uup, or YbiT were grown overnight at 37°C in LB medium supplemented with a selection antibiotic (100 μg/ml ampicillin or 20 μg/ml chloramphenicol). The following day, cultures were inoculated into fresh LB medium containing a selection antibiotic and grown at 37°C until A₆₀₀ reached 0.6. Subsequently, the appropriate inducer (0.05% arabinose or 1 mM IPTG) was added to express the ABCF proteins, and cultures were further cultivated for 2 h. Then cells were harvested by centrifugation (5000 × g, 15 min at 4°C).

Cell pellets were washed, and resuspended in the disruption buffer optimized for each ABCF protein as follows:

EttA or YbiT:

20 mM Tris–HCl pH 7.5, 150 mM NaCl, 10% glycerol, 7 mM β-mercaptoethanol (ME), 10 mM imidazole, EDTA-free cOmplete™ mini protease inhibitor cocktail (Roche)

Uup :

25 mM Tris–HCl pH 7.5, 2 mM MgCl₂, 60 mM KCl, 7 mM β-ME, 10 mM imidazole, EDTA-free cOmplete™ mini protease inhibitor cocktail

YheS :

20 mM Tris–HCl pH 7.5, 500 mM NaCl, 10% glycerol, 7 mM β-ME,

10 mM imidazole, EDTA-free cOmplete™ mini protease inhibitor cocktail

The cells were disrupted by sonication, and debris was removed by ultracentrifugation (100 000 × g, 40 min at 4°C). The recovered lysate was loaded onto a column filled with Ni-NTA Superflow (QIAGEN). After washing the column with a 10-times column volume of wash buffer {20 mM Tris–HCl pH 7.5, 500 mM NaCl, 10% glycerol, 7 mM β-ME, 20 mM imidazole, in case of the Δhelix mutants, 20 mM Mg(OAc)₂ was also included}, ABCF protein was eluted by wash buffer containing 50, 100 or 300 mM imidazole step by step. The fraction containing the ABCF protein was concentrated using Amicon Ultra 30 kDa (Merck) and PD-10 column (Cytiva) with solvent exchanged to storage buffer (20 mM Tris–HCl pH 7.5, 150 mM NaCl, 10% glycerol, 7 mM β-ME). For Uup and its derivatives, wash buffer and storage buffer were modified as follows.

Wash buffer for Uup:

25 mM Tris–HCl pH 7.5, 2 mM MgCl₂, 60 mM KCl, 7 mM β-ME, 20 mM imidazole

Storage buffer for Uup:

25 mM Tris–HCl pH 7.5, 2 mM MgCl₂, 60 mM KCl, 20% glycerol, 7 mM β-ME

Purified proteins were stored at −80°C.

Circular dichroism measurements

The far-UV spectra (260–190 nm) were obtained using JASCO J-820 Spectropolarimeter (JASCO Corporation). The ABCF proteins and their derivatives were dissolved at 0.05–0.1 mg/ml in 10 mM potassium phosphate (pH 7.5). All measurements were performed at 20°C using quartz cuvettes of 1 mm path length. Each spectrum represents an average of eight measurements, subtracting the blank values calculated from buffer measurements conducted under identical conditions.

In vitro translation and product analysis

The coupled transcription-translation reaction was performed using PUREfrex v1.0 (GeneFrontier) in the presence of ³⁵S-methionine {EasyTag L-[35S]-Methionine (PerkinElmer) or ARS0110A(American Radiolabeled Chemicals, Inc.)} or Cy5-Met-tRNA^fMet at 37°C, as described previously (14). DNA templates were prepared by PCR, as summarized in Supplementary Table S4. The reaction mixture was treated with 1 μM of purified Pth for 10 min at 37°C, or 200 μg/ml of puromycin for 10 min at 37°C where indicated. The ABCF proteins purified as described above were added at the final 1 μM. The reaction was stopped by the addition of TCA, washed by ice-cooled acetone, dissolved in SDS sample buffer (62.5 mM Tris–HCl, pH 6.8, 2% SDS, 10% glycerol, 50 mM DTT) that had been treated with RNAsecure (Ambion). Finally, the sample was divided into two portions, one of which was incubated with 50 μg/ml of RNase A (Promega) at 37°C for 30 min, and separated by a WIDE Range SDS-PAGE system (Nakalai Tesque). Images were visualized and analyzed by Amersham™ Typhoon™ scanner RGB system (GE Healthcare) using a 635 nm excitation laser and LPR emission filter or FLA7000 image analyzer (GE Healthcare).

β-Galactosidase assay

Escherichia coli cells harboring both the lacZ reporter plasmid and ABCF-expressing plasmid were grown overnight at 37°C in LB medium supplemented with 100 μg/ml ampicillin and 20 μg/ml chloramphenicol. On the next day, they were inoculated into fresh LB medium containing 2 × 10⁻³ or 2 × 10⁻⁴% arabinose, 100 μg/ml ampicillin and 20 μg/ml chloramphenicol were grown at 37°C. Expression of each ABCF protein was induced by the addition of the following concentration of IPTG as follows:

EttA and YbiT: at 10 μM, pre-included

Uup: 100 μM, pre-included

YheS: 100 μM, added when A₆₆₀ reached 0.2

After further inoculation (A₆₆₀ = ∼0.6), 20 μl portions were subjected to a β-galactosidase assay as described previously (14,15).

To evaluate the ‘downstream translation frequency’, the Miller unit (m.u.) of the reporter carrying a hard-to-translate sequence was normalized by that of control reporter pCY2518 (GFP-LacZ, for GFP-20D, 20E, 3P, 6P or 10P) or pBAD30-lacZ (for 5D, or 5E-LacZ), as follows:

$$\begin{eqnarray*} &&{\rm{Downstream\, translation }}\left( \% \right){\rm{ }} = \\ &&{\rm{m}}.{\rm{u}}.{\rm{ }}\left[ {{\rm{reporter }}\left( {{\hbox{hard-to-translate}}\,{\rm{sequenc}}{{{\rm{e}}}^ + }} \right)} \right] \\ && /{\rm{m}}.{\rm{u}}.\left[ {{\rm{reporter}}\left( {{\rm{no}}\,{\hbox{hard-to-translate}}\,{\rm{sequence}}} \right)} \right]\end{eqnarray*}$$

Quantification of the signals from gel images

The ratio of the translation-completed chain (full-length: FL) against RNase-sensitive polypeptidyl-tRNAs (pep-tRNAs), which signifies the frequency of the noncanonical translation events including pausing and premature termination, was calculated by the following formula.

$$\begin{eqnarray*} &&{\rm{peptidyl - tRNA }}\left( \% \right){\rm{ }} =\\ && {\rm{ }}\left( {{\rm{pep - tRNA}}} \right){\rm{ }}/{\rm{ }}\left[ {\left( {{\rm{FL}}} \right) + \left( {{\rm{pep - tRNA}}} \right)} \right]\end{eqnarray*}$$

The synthesis ratio of the ‘Full-length’ product was calculated as the ratio of the full-length product (FL) against the polypeptide part of the peptidyl-tRNAs (truncate) by the following formula. Subsequently, these values were normalized to the control experiment.

$$\begin{eqnarray*} {\rm{Synthesis ratio}}\left( \% \right) &=& \left[ {{\rm{FL}}\left( {{\rm{RNas}}{{{\rm{e}}}^ + }} \right)} \right]/\left[ {{\rm{FL}}\left( {{\rm{RNas}}{{{\rm{e}}}^ + }} \right)} \right] \\ &&+ \left[ {{\rm{truncate}}\left( {{\rm{RNas}}{{{\rm{e}}}^ + }} \right)} \right]\end{eqnarray*}$$

The proportion of peptidyl-tRNA that is sensitive to puromycin was calculated with the following formula.

$$\begin{eqnarray*} &&{\rm{Pep - tRNA }}( {{\rm{P}}{{{\rm{m}}}^{\rm{S}}}} ) = \\ && 100-[ {\rm{pep - tRNA}}( {{\rm{Puromycin - treated}}} )/\\ && \qquad\quad {\rm{pep - tRNA}}( {{\rm{control}}} ) ] \times 100\end{eqnarray*}$$

The radioactivity (³⁵S-methionine) or Cy5-fluorescence proportion of ‘FL’ and ‘pep-tRNA’ among the samples was quantified by the Multi Gauge software (Fujifilm).

Proteomic analysis

Samples for proteomic analysis were prepared as described previously (36). In brief, cells were grown in LB medium until the OD₆₆₀ reached ∼0.5. After the cells were harvested, they were washed and resuspended with PBS buffer (137 mM NaCl, 8.1 mM Na₂HPO₄, 2.68 mM KCl, 1.47 mM KH₂PO₄, pH 7.4). The suspension was mixed with an equal volume of 10% of TCA. After standing on ice for at least 10 min, the samples were centrifuged, and the supernatant was removed by aspiration. Precipitates were washed twice with acetone, by vigorous mixing. Proteins were dissolved in PTS solution (12 mM sodium deoxycholate, 12 mM sodium lauryl sulfate, 100 mM Tris–HCl, pH 9.0) and 50 μg of total protein at a concentration of 1 μg/μl was processed to reduction by dithiothreitol (DTT), alkylation with iodoacetamide, and limited digestion by Trypsin/Lys-C Mix (Promega, USA). Then samples were extracted by ethyl acetate, evaporated, dissolved in MS buffer-A (0.1% TFA and 2% acetonitrile), desalted by StageTip composed of an SDB-XC Empore disk (3M, USA), again evaporated, and resolved in MS buffer-A. Then samples were subjected to the LC–MS/MS measurement.

The LC–MS/MS measurements (SWATH-MS acquisition) (37) were conducted with an Eksigent nanoLC 415 and TripleTOF 6600 mass spectrometer (AB Sciex, USA). The trap column used for nanoLC was a 5.0 mm × 0.3 mm ODS column (L-column2, CERI, Japan) and the separation column was a 12.5 cm × 75 μm capillary column packed with 3 μm C18-silica particles (Nikkyo Technos, Japan). The SWATH acquisition was performed three times for each sample. Data analysis of the SWATH acquisition was performed using the DIA-NN software with default settings (38). The library for SWATH acquisition was obtained from the SWATH atlas {http://www.swathatlas.org, the original data are in Midha et al. (39)}. Only the proteins detected in all three measurements for both samples were used for the fold change calculation. The obtained protein intensities were averaged by using an in-house R script. The P-value was determined by Welch's t-test and corrected by the Benjamini-Hochberg method for multiple comparisons, using the ‘p.adjust’ function in R (for Mac, version 4.3.1).

Dataset for evolutionary analyses

We downloaded the full database dump (part I; https://stringdb-downloads.org/download/items_schema.v12.0.sql.gz) from STRING v12.0 (40). We used 2002 STRING ‘core’ bacterial species spanning ≥ 42 phyla.

Phylogenetic analysis of ABCF family proteins

We extracted 7146 proteins annotated as COG0488 (‘ATPase components of ABC transporters with duplicated ATPase domains’). We then inferred a phylogenetic tree for a subset of the sequences annotated with common subfamily names, where ‘preferred_name’ begins with or ‘protein_annotation’ contains either of ‘ettA’, ‘uup’, ‘ybiT’, ‘yheS’, ‘ydiF’, ‘yfmM’, ‘yfmR’, ‘ykpA’ or ‘yjjK’ in a case-insensitive manner. The resulting 948 sequences were aligned with the E-INS-i algorithm of MAFFT v7.520 (41) (‘–genafpair –maxiterate 1000’). The multiple sequence alignment without trimming was used for phylogenetic tree estimation by IQ-Tree v2.2.5 (42). The model selection was performed using ModelFinder Plus ('-m MFP') from general substitution models supported by RAxML (‘–msub nuclear –mset raxml’). The LG + F + I + R10 model was selected. The ultrafast bootstrapping was performed for 1000 times (‘-B 1000’).

Classification of ABCF family proteins

We visually inspected the multiple sequence alignment and the phylogenetic tree of the 948 ABCF proteins and followed three steps: we (i) (re-)classified all proteins as one of EttA, Uup, YheS or YbiT, (ii) classified them as EttA, Uup or YheS + YbiT based on the tree topology and finally (iii) distinguished YheS from YbiT based on the presence of C-terminal domain.

To classify all 7146 ABCF family protein sequences, we placed 6198 sequences into the tree above using the ‘add’ functionality of MAFFT ('–add') with the E-INS-i algorithm and feeding the output to the phylogenetic placement tool EPA-ng v0.3.8 (43) with the LG + F + I + R10 model. The resulting jplace file was converted to the Newick format using guppy in pplacer v1.1.alpha19 (44), adopting the most likely placement for each sequence. In rare cases where sequences were placed into branches splitting the subfamilies, they were regarded as ‘unclassified’ and excluded in the subsequent analysis.

We defined C-terminal domains as ≥80 amino-acid regions following ABC transporter 2 based on the domain architecture of E. coli YheS (P63389 of UniProt release 2303_5). ABC transporters 1 and 2 are denoted as nucleotide-binding domains or NBDs in the main text.

Phylogenetic profiling

We excluded fragmented sequences that had only <400 informative columns shared by 90% or more sequences in the alignment. To investigate the association of ABCF gene families with genomic contents of potential ‘hard-to-translate’ sequences, we extracted whole protein sequences from STRING and counted the following cases: (i) for the proline cluster, ‘XPPX’ motifs defined by Peil et al. (45) (both ‘strong’ motifs ‘PP[PWDNG] or [PDA]PP’ and ‘strong’ + ‘weak’ motifs ‘PP[PWDNGHRYLF] or [PDANCRWLTYI]PP’ were investigated), (ii) for the N-terminal poly-acidic cluster, Asp or Glu appearing three times or more in six residues following heading Met and (iii) for the gene-wide poly-acidic cluster, Asp or Glu appearing three times or more in a window of size six. We also specifically investigated NOG370323 [SecM], COG0653 [Preprotein translocase subunit SecA (ATPase, RNA helicase)] and COG0231 [Translation elongation factor P (EF-P)/translation initiation factor 5A (eIF-5A)] using the STRING database dump.

Evolutionary residue conservation of YheS

We downloaded the latest version (commit ID: 16be890) of ConSurf (46) from https://github.com/Rostlab/ConSurf. Minimum patches were applied to let the program run on our environment. The input alignment was compiled by recomputing MAFFT E-INS-i with 1478 YheS sequences. The reference structure of YheS was downloaded from AlphaFold Protein Structure Database (47). We specified the LG model for the substitution model.

Statistical analyses

Statistical analyses were conducted by using R (https://www.r-project.org) and Python (https://www.python.org).

Description of the structural data

Published or predicted structural data were obtained from PDBJ (https://pdbj.org/, 3J5S) (32), or AlphaFold Protein Structure Database (https://alphafold.ebi.ac.uk, P43672, P63389, P0A9U3) (47), respectively. The three-dimensional representation was generated by using PyMOL software (2.5.0).

Results

E. coli ABCF protein YheS can release the ribosome arrested by the SecM nascent peptide

Within the E. coli K-12 strain, there are four distinct ABCF proteins, each sharing <50% sequence similarity with one another (Figure 1A). These proteins commonly consist of two nucleotide-binding domains (NBDs) and unique interdomain linker sequences. The C-terminal extension (CTE), found in several ARE-ABCFs, is also present in YheS and Uup.

Figure 1.

E. coli YheS releases the ribosome arrested by SecM nascent peptide. (A) Four endogenous ABCF proteins, that possess characteristic interdomain linkers, individually cope with the ribosome in the aberrant state(s). The interdomain linker of each ABCF was described from published data (32) (PDB: 3J5S) or AlphaFold Protein Structure database (47). (B) The secM mRNA was translated by PUREfrex in the presence of Cy5-Met-tRNA and indicated ABCF proteins (final 1 μM). Samples were separated by neutral pH SDS-PAGE with optional RNase A (RN) pretreatment. The peptidyl-tRNA and tRNA-released truncated peptide are schematically indicated. ‘FL’ stands for the full-length product. (C) Proportion of the SecM peptidyl-tRNA quantified from gel images represented in (B) (#). (D) Deletion of the α-helix abolishes the function of YheS. The proportion of the SecM peptidyl-tRNA synthesized in the presence of YheS or its derivative lacking the linker helix (Supplementary Figure S1C). ***P-value < 0.005, n.s.: no significant difference (Welch's t-test). (#) (E) The secM mRNA was translated by PUREfrex in the presence of ³⁵S-methionine for 30 min. After the addition of cold methionine (final 2 mg/ml) and purified YheS (final 1 μM), a small portion of PUREfrex mixture was withdrawn at the indicated time and mixed with an excessive amount of TCA solution. Subsequently, the ratio of the peptidyl-tRNA at the indicated time was calculated from gel images (Supplementary Figure S2D) and plotted. (#) (F) Schematic of the domain structure of YheS and the introduced mutations. NBD stands for nucleotide-binding domain. (G) Arrest-releasing activity of YheS mutants. The secM mRNA was translated by PUREfrex in the presence of Cy5-Met-tRNA. After 10 min from the addition of YheS or its derivatives, a portion was withdrawn and mixed with the final 200 μg/ml puromycin (Pm) and further incubated for 10 min. (H) Proportion of the SecM peptidyl-tRNA quantified from gel images represented in (G). (#) (I) Proportion of the puromycin-reactive SecM peptidyl-tRNA quantified from gel images represented in (G). (#) (#) The mean values ± SE estimated from three independent technical replicates are shown.

Various kinds of nascent peptide sequences trigger noncanonical translations, such as ribosome stalling or premature termination. Notably, these aberrant phenomena commonly lead to the accumulation of peptidyl-tRNA molecules, detectable through gel electrophoresis analysis (14,48). In this study, we mainly assessed the impact of each ABCF on noncanonical translations by evaluating the accumulation of peptidyl-tRNA. Moreover, to eliminate potential indirect effects of the ABCF proteins in E. coli, we conducted experiments using the reconstituted in vitro translation system (PURE system; PUREfrex v1.0) (49). Quality evaluations of purified ABCF proteins were summarized in Supplementary Figure S1.

Escherichia coli ABCF proteins promoted the synthesis of various kinds of hard-to-translate sequences, which we will delve into shortly. However, what surprised us the most among the sequences we tested was the ability of YheS to release the SecM-induced translation arrest (Figure 1). In the case of secM mRNA, most of the translation product accumulates as peptidyl-tRNA, indicating the strong arrest activity of the SecM nascent peptide (Figure 1B, lanes 1 and 2, Figure 1C). Doublet bands represent stalling before and after the transpeptidation to the 166th Pro-tRNA within the A-site, respectively (Supplementary Figure S2A) (50). Conversely, the presence of YheS specifically enhanced the production of the full-length polypeptide (FL), indicating that ribosomes reach the stop codon beyond the arrest peptide motif (Figure 1B, lanes 7 and 8). The YheS-induced disappearance of shorter peptide products (20 kDa), which would be accommodated within the ‘following’ ribosomes, also suggests the resolution of the leading ribosome stalled at the Pro^166th codon.

Previous studies have shown that the interdomain linker of ARE-ABCFs plays a critical role in rescuing the ribosomes trapped by antibiotics (26–31). Therefore, we investigated whether the interdomain linker of YheS is similarly essential for its function. Strikingly, the deletion of the α-helix within the interdomain linker completely inactivated YheS (Figure 1D, Supplementary Figure S1C). This result indicates that the interdomain linker of YheS plays a crucial role in releasing the SecM-arrested ribosome. In addition, the arrest-releasing activity of YheS was independent of the signal sequence of SecM (Supplementary Figure S2B and S2C). These findings indicate that YheS employs a distinct mechanism from the well-established secretion-dependent process to release SecM-induced translation arrest (18,20,51). We also discovered that YheS was capable of releasing the translation arrest even after the SecM nascent peptide had already arrested the ribosome (Figure 1E and Supplementary Figure S2D). In contrast, YheS had minimal impact on other ribosome-arresting peptides, such as E. coli tnaC (52), E. coli speFL (53), apdP in Sinorhizobium medicae (22), or an artificial arrest peptide (54) (Supplementary Figure S2E and S2F). Based on these results, we conclude that YheS efficiently rescues the ribosomes arrested by SecM, while showing limited effects on other ribosome-arresting peptides.

Structural features required for the function of YheS

Previous studies on ARE-ABCFs have shown that several structural features were critical for their functions. For instance, the ATPase activity of NBD, which would trigger the structural rearrangements like other ABC family proteins (55–57), is indispensable for the function of ABCFs (26,31,33,58). Moreover, the C-terminal extension, also present in YheS (and Uup), is essential for the function of ARE-ABCFs (27,58). To investigate their importance, we introduced mutations at E175 and E456 residues, which constitute the catalytic core of the ATPase activity of the two ABC cassettes, or deleted the C-terminal extension of YheS (Figure 1F).

Derivatives of YheS were purified (Supplementary Figure S1), and their activity was assessed by using the PURE system (Figure 1G). The deletion of the C-terminal extension abolished the activity of YheS, as well as ARE-ABCF. In contrast, the ATPase mutants exhibited different effects from each other. The E171Q mutant retained a slight degree of arrest-releasing activity, whereas the E456Q mutant completely lost its activity (Figure 1H). Intriguingly, the addition of the EQ₂ mutant led to a more significant accumulation of peptidyl-tRNAs and nearly completely blocked the action of puromycin (Figure 1H and I). These observations suggest that ATP hydrolysis within the two NBDs does not function symmetrically. Moreover, YheS in the initial binding state, prior to ATP hydrolysis, may not necessarily alleviate the SecM-induced translation arrest; instead, it might engage in interactions that strengthen the translation arrest. This would reflect the YheS-dependent stabilization of the SecM peptidyl-tRNA, or the distortion of peptidyl-tRNA by insertion of YheS’s interdomain linker, as observed in the studies on the ARE-ABCFs (26–31).

Structural and functional characteristics of the interdomain linker

Comparative analysis among the YheS homologs revealed that the amino acid sequence of the interdomain linker is well conserved (Figure 2A), suggesting its importance in releasing the translation arrest. We then attempted to identify critical features within the interdomain linker of YheS via systematic alanine substitution. For this purpose, we employed the GFP-SecM-LacZ reporter presented in Supplementary Figure S3A, which expresses LacZ exclusively when YheS releases the translation arrest. All mutations tested in Figure 1 abolished the YheS-dependent expression of LacZ, confirming the validity of the reporter (Supplementary Figure S3A). Note that the expression of YheS did not influence the translation of the GFP-SecM-LacZ reporter carrying the arrest-attenuating mutation (Supplementary Figure S3B), and endogenous YheS had poorly influenced SecM-induced translation arrest under the conditions we tested. (Supplementary Figure S3C). Then, we introduced alanine substitution mutation into residues from Y227 to I326 residues one by one and evaluated their arrest-releasing activity. As depicted in Figure 2B, the substitution of various residues impacted the function of YheS, with major effects attributed to three regions of the structure predicted by AlphaFold2 (47). Firstly, while most mutations introduced into the α-helix had no significant effect, residues F234, R241, and Y249 strongly influenced the function of YheS (Figure 2B, panel 3). These residues all face inside the linker, suggesting their importance in maintaining the linker structure. The second region is approximately eight residues from the tip of the α-helix and appears important for interaction with tRNA. Mutations in this region had particularly strong effects, with residue F266 at the tip of the α-helix (Figure 2B, panel 4). This residue might have a function similar to the F237 of ribosome protection protein VmlR (27). Thirdly, mutations of residues F298, F300, and F302, located at the base of the linker, also showed significant effects (Figure 2B, panel 5). Considering that mutating all three residues almost hampered the function of YheS (Supplementary Figure S3D), these phenylalanine residues would be important for fixing the linker to the 2nd NBD, thus contributing to the proper maintenance of the entire structure. Notably, residues with strong influence, as mentioned above, were highly conserved among YheS homologs (Figure 2B, panel 2) (46). Thus, we speculate that, in addition to the tRNA interaction domain, the maintenance of its unique linker structure is also crucial for the function of YheS.

Figure 2.

Structural features of the interdomain linker in YheS. (A) The three-dimensional structure of YheS predicted by AlphaFold2 (P63389) (47) was colored according to the ConSurf conservation score (46). Higher scores mean higher conservation. The C-terminal tail (from A532 to N636) was omitted in the right panel for clarity. (B) panel 1: Systematic alanine/serine scanning of the interdomain linker of YheS, spanning residues Y227 to I326. The arrest-releasing activity of each YheS mutant was evaluated as shown in Supplementary Figure S3A, and their activity was normalized relative to that of wild type YheS (set to 1). The mean values ± SE estimated from three independent biological replicates are shown. Panel 2: ConSurf conservation score of the interdomain linker. Panels 3–5: expanded view of F234, R241 and Y249 residues within the linker helix (panel 3), the tip of the linker helix including R265, F266 and K272 (panel 4), and F298, F300 and F302 residues located at the base of the interdomain linker (panel 5).

EttA and YbiT alleviate intrinsic ribosome destabilization (IRD) triggered by poly-acidic sequences.

Given the arrest-releasing activity of YheS, we aimed to investigate whether other ABCFs also play a role in coping with nascent peptide-dependent noncanonical translations. Our previous research has demonstrated that the translation of poly-acidic sequences triggers intrinsic ribosome destabilization (IRD), leading to Pth-mediated premature termination (14–17). For instance, over 50% of the synthesized product from a model GFP mRNA containing the 20D poly-acidic sequence accumulated as Pth-sensitive peptidyl-tRNA (Figure 3A, lanes 1–4, Figure 3B). Pth is incapable of hydrolyzing the peptidyl-tRNA within the intact 70S complex (59,60). Therefore, the sensitivity of peptidyl-tRNA to Pth serves as a distinguishing factor between the two similar yet distinct noncanonical translations, ribosome stalling and IRD.

Figure 3.

YbiT and EttA repress IRD-dependent premature termination. (A) Lanes 1–4: the GFP-20D mRNA was translated by PUREfrex in the presence of Cy5-Met-tRNA. Samples were treated with puromycin (Pm) or peptidyl-tRNA hydrolase (Pth) as indicated, and separated by neutral pH SDS-PAGE with optional RNase A (RN) pretreatment. The peptidyl-tRNA and tRNA-released truncated peptide are schematically indicated. ‘FL’ stands for the full-length product. Lanes 5–14: The GFP-20D mRNA was translated by PUREfrex in the presence of indicated ABCF proteins (final 1 μM). (B) Panel 1: the ratio of the GFP-20D peptidyl-tRNA, calculated as described in the methods. Panel 2: The ratio of the full-length product (FL), calculated as described in the methods. (#) (C) The 3E-lacZα mRNA was translated by PUREfrex in the presence of ³⁵S-methionine and indicated ABCF proteins (final 1 μM). Samples were treated with Pth as indicated and separated by neutral pH SDS-PAGE with optional RNase A (RN) pretreatment. ‘fMet’ stands for fMet-tRNA. (D) The ratio of the 3E peptidyl-tRNA in the presence of indicated ABCF was quantified from the gel images represented in Figure 3C. (#) (E) The ratio of the GFP-20D (left panel) or 3E (right panel) peptidyl-tRNA in the presence of EttA, EttAΔhelix, YbiT or YbiTΔhelix. ***P-value < 0.005, n.s. : no significant difference (Welch's t-test). (#) (F) The GFP-20D mRNA was translated by PUREfrex in the presence of YbiT or its ATPase-deficient EQ₂ mutant. The time intervals during which YbiT or YbiT-EQ₂ was present during the 30 min reaction are indicated on the gel image. (G) The E. coli yagN mRNA carrying the ^110thDPDPEPE^116th IRD motif was translated and analyzed as (A). (H) The ratio of the YagN peptidyl-tRNA in the presence of indicated ABCF was quantified from the gel images represented in (G). (#) (I) The E. coli mgtL-lacZ mRNA carrying the N-terminal ^2ndEPDP^5th IRD motif was translated and analyzed as (C). (J) The ratio of the MgtL peptidyl-tRNA in the presence of indicated ABCF was quantified from the gel images represented in (I). (#) (#) The mean values ± SE estimated from three independent technical replicates are shown.

Consequently, GFP-20D mRNAs were subjected to translation using the PUREfrex system, which includes purified E. coli ABCF proteins. Notably, the accumulation of GFP-20D peptidyl-tRNA decreased significantly upon the introduction of YbiT, one of the ABCFs, into the translation mixture (Figure 3A, lanes 13 and 14, Figure 3B). This reduction in peptidyl-tRNA accumulation correlated with an increase in the production of full-length polypeptide, indicating that YbiT enhances the ability of ribosomes to successfully navigate through the translation of the 20D sequence (Figure 3B, panel 2). Similar results were observed when translating other poly-acidic 20E sequences (Supplementary Figure S4A and S4B) or the 56-DE repeats found in the yeast YOR054C ORF (Supplementary Figure S4C and S4D).

It should be noted that poly-acidic sequences like the 20D repeats examined in Figure 3A are present in the eukaryotic proteome but not in the prokaryotic E. coli proteome. However, N-terminal DE/P-enriched sequences in E. coli, characterized by the presence of three or more DE/P residues within a 6 amino acid window, have been shown to induce frequent IRD (15). To assess the influence of ABCFs on the translation of N-terminal IRD sequences, we employed the 3E-lacZα mRNA, which encodes three consecutive glutamate residues in the N-terminal region. Once again, the inclusion of YbiT prevented IRD in the early stage of translation (Figure 3C, lanes 13–15, Figure 3D). Additionally, EttA, one of the ABCFs, exhibited an inhibitory effect on the N-terminal IRD as well (Figure 3C, lanes 4–6). Identical results were obtained even when the 3E motif was substituted with the 3D motif (Supplementary Figure S4E and S4F). The activity of EttA and YbiT depends on the interdomain linker and the ATPase activity, as was the case for YheS (Figure 3E, F, Supplementary Figure S4G–I). Intriguingly, the activity of EttA appears to be context-dependent, possibly influenced by the length of the nascent peptide within the ribosome tunnel (Supplementary Figure S4J). Previous reports have shown that EttA promotes the early stage of translation elongation (32,33), which is consistent with the findings here. Collectively, these findings highlight that YbiT and EttA alleviate the poly-acidic sequence-induced IRD in a similar but distinct manner from each other. However, it should be noted that both EttA and YbiT had minimal influence on IRD during the translation of yagN (DPDPEPE) or mgtL (EPDPTP), which have a proline-intermitted acidic amino acid sequence (Figure 3G-J) (14).

ABCF-mediated alleviation of IRD in living E. coli cells

We conducted further investigation into whether the effects of YbiT or EttA on IRD could be observed in living E. coli cells. Initially, we assessed the impact of ABCFs on IRD in the middle of the ORF. The GFP-20D-LacZ or GFP-20E-LacZ reporter was expressed in E. coli cells lacking ettA, ybiT, or both (Figure 4A). Additionally, we tested the ΔbL31 strain (ΔrpmE ΔykgM), known to exhibit more frequent occurrences of IRD (14), as a positive control. Deletion of ybiT or bL31, but not ettA, resulted in decreased translation efficiency of both GFP-20D-LacZ and GFP-20E-LacZ mRNA, consistent with the in vitro results (Figure 4A). This phenotype was rescued by the trans-expressed YbiT, but not by the YbiTΔhelix mutant (Figure 4B).

Figure 4.

ABCFs-mediated alleviation of IRD in living E. coli cells. (A) The GFP-20D-LacZ (panel 1) or GFP-20E-LacZ reporter (panel 2) was expressed in E. coli BW25113 cells (wild type) and its derivatives as indicated. The expression level of GFP-20D-LacZ or GFP-20E-LacZ was normalized to that of GFP-LacZ control reporter and plotted as the downstream translation frequency, respectively (as detailed in the materials and methods). (#) (B) Downstream translation frequency of the GFP-20D-LacZ (panel 1) or GFP-20E-LacZ reporter (panel 2) in E. coli ΔybiT cells expressing wild type YbiT or YbiTΔhelix mutant. (#) (C) The 5D-LacZ (panel 1) or 5E-LacZ reporter (panel 2) was expressed in E. coli BW25113 cells (wild type) and its derivatives as indicated. The expression level of 5D-LacZ or 5E-LacZ was normalized to that of the LacZ reporter carrying no DE-repeats and plotted, respectively. (#) (D) Downstream translation frequency of the 5D-LacZ in E. coli ΔettAΔybiT cells expressing wild type EttA, YbiT or their derivatives. (#) (E) E. coli pth^ts mutant (G101D) (61) carrying the indicated mutations were serially spotted onto the LB agar plate, and incubated for 1 overnight at 30 or 39°C, respectively. Representative of three independent experiments is shown. (F) Downstream translation frequency of the MgtL-LacZ reporter, which carries the N-terminal ^2ndEPDP^5th IRD motif, in E. coli BW25113 cell and its derivatives. *P-value < 0.05, n.s.: no significant difference (Welch's t-test). (#) (G) Expression of MgtA, whose expression negatively correlates with the frequency of IRD during the translation of mgtL, was estimated as the expression level of translationally fused LacZ reporter. ***P-value < 0.005, *P-value < 0.05, n.s.: no significant difference (Welch's t-test). (#) (#) The mean values ± SE estimated from three independent biological replicates are shown.

EttA exclusively alleviated the N-terminal IRD. Therefore, we prepared other reporters, 5D-LacZ or 5E-LacZ, in which five consecutive acidic amino acids were translated immediately after translation initiation (Figure 4C). The translation efficiency of these mRNA decreased significantly in the absence of EttA and bL31 (Figure 4C). In contrast to the IRD in the middle of the ORF, the absence of YbiT had a moderate effect on N-terminal IRD (5D: P = 0.124; 5E: P = 0.064). However, trans-expressed YbiT promoted the translation of the N-terminal DE-repeats as well as EttA (Figure 4D). This indicates that endogenous YbiT preferentially alleviates IRD in the middle of ORF; however, it also has the potential to prevent N-terminal IRD, as observed in in vitro experiments.

In addition to the reporter assays, the deletion of ettA or ybiT enhanced the temperature-sensitive phenotype of pth G101D mutant (Figure 4E) (61). Pth resolves the IRD-derived abortive peptidyl-tRNA (14,17), and its shortage exhibits a synthetic growth defect with the ΔbL31 mutation (Figure 4E). From these results, we concluded that endogenous EttA and YbiT have a physiological role in preventing IRD in living E. coli cells. However, the deletion of these ABCFs had minimal influences on the translation of mgtL and the mgtL-regulated MgtA expression (Figure 4F and G) (14). This result was reproduced in the comparative quantitative proteomics analyses (Supplementary Figure S5A–C). ABCFs-mediated alleviation could disturb the Mg²⁺-sensing function of mgtL, so this regulatory IRD sequence might have evolved not to be rescued by these factors.

Uup and YbiT alleviate ribosome stalling induced by poly-proline or poly-basic sequences, respectively.

Previous studies on the nascent peptide have revealed that the translation of poly-basic or poly-proline amino acid sequence hampers the translation elongation, leading to ribosome stalling, respectively (1–9,12,13). In this section, we investigated the influence of ABCFs on ribosome stalling using consecutive twenty lysine (20K) or ten proline (10P) sequences. The addition of YbiT promoted the translation of the 20K or 20R sequence, resulting in a twofold increase in the synthesis of full-length products (Figure 5A, lanes 13 and 14, Figure 5B, panel 2, 20R in Supplementary Figure S6A and S6B). Uup, one of the ABCFs, weakened the poly-proline dependent ribosome stalling (Figure 5C, lanes 9 and 10, Figure 5D). However, EttA and YheS did not exhibit any discernible impact on the translation of GFP-20K, 20R, and 10P mRNA sequences. Loss of the α-helix within the interdomain linker or ATPase activity diminished the translation promotion activity, as well as other ABCFs (Supplementary Figure S6C–E). We also conducted a series of ABCF-chase experiments and found that the addition of ABCFs after the occurrence of the noncanonical translation altered the proportion of the ribosome stalling, but did not influence IRD-dependent premature termination (Supplementary Figure S6F). This implies that ABCFs can rescue ribosomes arrested by the nascent peptide but not those irreversibly split (17).

Figure 5.

Uup and YbiT alleviate poly-proline or poly-basic sequence-dependent ribosome stalling, respectively. (A) The translation profile of the GFP-20K mRNA, including a poly-basic amino acid sequence, was analyzed as Figure 3A. (B) Proportion of the GFP-20K peptidyl-tRNA (panel 1) and the full-length product (FL, panel 2). (#) (C) The translation profile of the GFP-10P mRNA, including a poly-proline sequence, was analyzed as Figure 3A. (D) Proportion of the GFP-10P peptidyl-tRNA (panel 1) and the full-length product (FL, panel 2). (#) (#) The mean values ± SE estimated from three independent technical replicates are shown.

The functional relationship between Uup and EF-P

The translation of the poly-proline sequence is effectively resolved by EF-P (4–9). The functional overlap between Uup and EF-P prompted us to compare their functionality on the poly-proline sequence. In the reconstituted translation system, EF-P enhanced the translation of GFP-10P mRNA more efficiently than Uup (Figure 6A and B). Similarly, the absence of EF-P caused a significant defect in the growth of E. coli cells, while the absence of Uup made almost no difference (Figure 6C). Consistent with these observations, the translation efficiency of the poly-proline sequence was significantly depressed in Δefp cells, and only slightly in Δuup cells (Supplementary Figure S7). These results suggest that EF-P acts as the major rescue factor for the proline-dependent translation stalling and endogenously expressed Uup is insufficient for this function at least under the conditions we tested. Consistent with this hypothesis, EF-P, rather than Uup, is widely conserved among diverse bacterial groups. On the other hand, bacterial species with genomes containing more abundant poly-proline sequences do not necessarily have more copies of Uup (Supplementary Figure S8).

Figure 6.

Functional comparison between Uup and EF-P. (A) The GFP-10P mRNA was translated by PUREfrex in the presence of an indicated concentration of EF-P (lanes 1–10) or Uup (lanes 11–20), respectively. (B) The ratio of GFP-10P peptidyl-tRNA in the presence of the indicated concentration of translation factors was quantified from the gel images represented in Figure 6A. (#) (C) E. coli BW25113 (wild type) and its derivatives were serially spotted onto the LB agar plate and incubated at 37°C for 1 overnight. Representative of three independent experiments is shown. (D) E. coli BW25113Δefp strain carrying the indicated plasmid was spotted onto LB agar plate containing chloramphenicol for the selection and 10 μM of IPTG to induce the expression of translation factors and incubated at 37°C for 1 overnight. Representative of three independent experiments is shown. (E) Downstream translation frequency of the GFP-3P-LacZ (panel 1), GFP-6P-LacZ (panel 2), or GFP-10P-LacZ reporter (panel 3) in E. coli BW25113Δefp cells expressing excessive Uup (in the presence of 100 μM IPTG). ***: P-value < 0.005, n.s.: no significant difference (Welch's t-test). (#) (#) The mean values ± SE estimated from three independent technical or biological replicates are shown.

However, it should be noted that the deletion of uup, but not other ABCFs, enhanced the Δefp-dependent growth defect (Figure 6C). Furthermore, trans-expressed Uup efficiently and exclusively complemented the growth defect of the Δefp strain (Figure 6D). These results indicate that Uup can compensate for the function of EF-P. Consistent with the assumption, trans-expressed Uup enhanced the translation of the GFP-LacZ reporters carrying the various lengths of proline repeats in the absence of EF-P (Figure 6E). These results indicate that Uup can alleviate the poly-proline-dependent ribosome stalling in living E. coli cells.

Discussion

In this study, we have successfully demonstrated that endogenous ABCF proteins in E. coli play a role in promoting the synthesis of ‘hard-to-translate’ sequences. A key observation is that each ABCF uniquely prevents or resolves distinct noncanonical translations. However, it should also be noted that our analyses were limited to the already-known problematic amino acid motifs and do not exclude the possibility that E. coli ABCFs have other clients or physiological roles.

It could be inferred that the structural characteristics, including the interdomain linker that likely interacts with tRNA, the nucleotide-binding domain, and the C-terminal extension in the case of YheS and Uup, collectively contribute to coping with noncanonical translations. From the analysis focusing on YheS, the ATP hydrolysis activity of the two NBDs, which is associated with subsequent structural rearrangement of other ABC proteins (55–57), is essential for the arrest-releasing activity of YheS. This implies that E. coli YheS rescues the SecM-arrested ribosome in an ATP-hydrolysis-coupled structural rearrangement, while several ARE-ABCFs are likely to dislodge the antibiotics without ATP hydrolysis (27–31). Structural analyses of ARE-ABCFs in Gram-positive bacteria have revealed that the differences in the interdomain linker determine what kinds of antibiotics they exclude from the ribosome. Intriguingly, the interdomain linker of ARE-ABCF induces a positional shift of the initiator tRNA, which is located within the P-site of the antibiotics-arrested ribosome (27–31). This repositioning could explain the arrest-releasing activity of YheS observed in this study. However, one of the notable differences from previous studies is the occupation of the exit tunnel by the nascent peptide, which could constrain tRNA dynamics. Due to these constraints on structural rearrangement, ATP hydrolysis would be required for YheS to release the translation arrest induced by SecM. Further analysis is needed to determine whether ATP hydrolysis is essential for releasing the translation arrest or solely required for the dissociation of YheS from the ribosome.

Previous structural studies have shown that E. coli EttA also interacts with tRNA, restricting the dynamics of the tRNA within the P-site (32,33). The interaction between EttA and tRNA might indeed impose constraints on the movement of the P-site tRNA, potentially leading to an inhibition of IRD. Our previous studies have shown that IRD is inhibited in situations where the dynamics of P-site tRNA are restricted by the preceding nascent chain (15,16). Additionally, molecular simulations have indicated the possibility of P-site tRNA adopting a different positioning due to the translation of the poly-acidic sequence (62). Investigating the function of EttA and YbiT is an intriguing subject not only from the perspective of how these proteins promote translation but also from the standpoint of understanding the fundamental mechanism underlying IRD. However, it is important to note that neither EttA nor YbiT completely suppresses IRD, and the possibility of these proteins inhibiting IRD through secondary effects apart from their primary targets cannot be ruled out.

We have unveiled the function of the ABCFs via biochemical experiments and in vivo reporter assays, however, the physiological significance of endogenous ABCF proteins in E. coli is not fully understood. Single-knockout strains for each ABCF showed almost no growth defects (Supplementary Figure S9B), minimal rearrangement of proteome landscape (Supplementary Figure S5A-C), or minimal alteration in the sensitivity to translation inhibitors such as erythromycin (Supplementary Figure S9C). However, the E. coli strain lacking all four ABCFs exhibited susceptibility to drugs related to translation elongation (Supplementary Figure S9C). In addition, synthetic growth defect was observed between pth^ts mutant and ettA or ybiT (Figure 4E), as well as between efp and uup (Figure 6C, the identical phenotype in Gram-positive bacteria was reported in a recently published paper and a preprint: 63,64). These demonstrate that each ABCF collectively contributes to maintaining the robustness of translation elongation. However, the effectiveness of Uup on the translation of the proline-repeats was moderate compared to that of EF-P, which almost exclusively alleviates the proline repeats (Figure 6). Combined with the fact that YbiT moderately promotes multiple ‘hard-to-translate’ sequences, ABCFs may be a broad and shallow rescue of ribosomes in a defective state while exhibiting some specificity. Due to these characteristics, their importance could significantly increase in stress environments where overall conditions deteriorate, while there may not be prominent phenotypes in situations where translation proceeds smoothly. Notably, previous reports have indicated the importance of EttA for long-term cultivation (33). Consistent with this, Boël and colleagues recently reported that EttA is upregulated by the osmotic stress and the limited nutrients to alleviate the N-terminal translation attenuation on the various endogenous genes (65). Consistent with this, we have recently shown that ribosome stalling, which could be induced by the limitation of amino acid, enhances IRD (17). The inherent risk of noncanonical translations caused by the nascent peptide is widespread throughout the proteome, meaning that organisms certainly need to alleviate this issue in order to survive in a variety of environments.

YheS resolves the translational arrest of SecM, which has a regulatory role in SecA expression (18,20,66,67). This might appear counterintuitive, especially considering the physiological role of SecM as a monitoring substrate for the protein secretion system (68). However, an alternative perspective is that YheS could act as a rescue system in cases where the secretion-dependent arrest-releasing mechanism could not work, such as due to the loss or absence of a signal sequence. Recent analyses by Chiba and colleagues suggested that regulatory nascent peptides like SecM are sporadically acquired during evolution (22). It is unclear whether the arrest motif or the arrest-releasing mechanism, such as the signal sequence in SecM, is acquired first during the evolution of arrest sequences. In the former case, the expression of an unreleasable arrest sequence is merely disadvantageous for survival. In this context, the ABCFs-mediated alleviation of translation stalling could function as a safety device during the evolution. YheS homologs are widely distributed across the bacterial species and are expected to have been acquired before SecM (Supplementary Figure S8). In addition, the amino acid sequence was less significantly conserved among the YheS homologs compared to the EttA or Uup homologs (Supplementary Figure S9A). These suggest that while YheS may originally serve a physiological function distinct from the arrest-releasing of SecM, YheS may have acquired the ability to release the translation arrest induced by SecM later in evolution. Consistent with this assumption, several mutations within the interdomain linker enhanced the arrest-releasing activity of YheS, suggesting that E. coli YheS is not entirely specialized for SecM (Figure 2). On the other hand, at the current stage where the release mechanism has been acquired, expression of YheS is generally repressed (69) and the deletion of the yheS gene had no influence on the SecM-induced translation arrest in the conditions we tested (Supplementary Figure S3C). These imply that the expression of yheS is regulated not to interrupt the SecM-controlled secretion homeostasis but to be expressed in certain stress conditions. Further study is required to fully understand the physiological significance of YheS and other ABCFs, including the gene expression regulation as demonstrated in ARE-ABCFs (30,70,71).

The limited impact of ABCFs on distinct noncanonical translations suggests that each ABCF may have specialized interactions with certain structural elements or sequences of translating ribosomes. A series of analyses on YheS highlights the significance of the interplay between the interdomain linker, potentially involved in tRNA interactions, and NBDs, responsible for initiating the entire structural rearrangement, in resolving the SecM-induced translational arrest. In the future, uncovering the ATP hydrolysis-coupled movements of the interdomain linker and understanding the details of the arrest-releasing activity of YheS may make it possible to redirect the targeting of YheS towards other than SecM through re-design of the linker sequence. Such artificial modification and evolution of ABCF might enable us to manipulate genetic information encoded within the amino acid sequence, rather than within DNA.

Data availability

All data generated or analyzed during this study are included in the manuscript and supporting information. Raw data files of plasmids used in this study are available in the Mendeley Data repository (DOI: 10.17632/x347y36n49.1). The mass spectrometry data have been deposited in the jPOST repository (JPST002944/PXD049710) (72).

Supplementary data

Supplementary Data are available at NAR Online.

Acknowledgements

We would like to express our gratitude to Hiraku Takada for valuable discussions. We also thank Shinobu Chiba and Takashi Kanamori for providing pCH2102 plasmid DNA and PUREfrex v1.0, purified Pth and EF-P, respectively. Additionally, we extend our thanks to the Bio-support Center at Tokyo Tech for DNA sequencing, the Cell Biology Center Research Core Facility at Tokyo Tech for the TripleTOF 6600 mass spectrometry measurements, and Naohiko Shimada and Atsushi Maruyama for the CD measurements.

Author contributions: Y.C., E.U., K.Y., T.N., T.I. and M. K. performed experiments; E.U. prepared the purified components for in vitro translation; S.Y. and W.I. performed the in silico analyses; Y.C. designed experiments and analyzed the results; Y.C., S.Y., E.U., K.Y., T.N., T.I., M.K., W.I. and H.T. conceived the study; Y.C. and H.T. supervised the entire project and wrote the manuscript.

Funding

MEXT Grants-in-Aid for Scientific Research [JP20H05925 to H.T., 23H02410 to Y.C.]; Ohsumi Frontier Science Foundation; Japan Foundation for Applied Enzymology; Takeda Science Foundation; Yamada Science Foundation (to Y.C.); JST CREST JPMJCR19S2 (to W.I.); JST SPRING JPMJSP2108 (to S.Y.). Funding for open access charge: MEXT Grants-in-Aid for Scientific Research [JP20H05925].

Conflict of interest statement. None declared.