ResQ, a release factor-dependent ribosome rescue factor in the Gram-positive bacterium Bacillus subtilis

Rescue of the ribosomes from dead-end translation complexes, such as those on truncated (non-stop) mRNA, is essential for the cell. Whereas bacteria use trans-translation for ribosome rescue, some Gram-negative species possess alternative and release factor (RF)-dependent rescue factors, which enable an RF to catalyze stop codon-independent polypeptide release. We now discover that the Gram-positive Bacillus subtilis has an evolutionarily distinct ribosome rescue factor named ResQ. Genetic analysis shows that B. subtilis requires the function of either trans-translation or ResQ for growth, even in the absence of proteotoxic stresses. Biochemical and cryo-EM characterization demonstrates that ResQ binds to non-stop stalled ribosomes, recruits homologous RF2, but not RF1, and induces its transition into an open active conformation. Although ResQ is distinct from E. coli ArfA, they use convergent strategies in terms of mode of action and expression regulation, indicating that many bacteria may have evolved as yet unidentified ribosome rescue systems.


Introduction 32
Faithful translation requires accurate initiation, elongation and termination. In 33 translation termination, the stop codon situated in the A-site of the ribosome recruits a 34 release factor (RF), which then hydrolyzes the peptidyl-tRNA ester bond to release the 35 polypeptide product from the ribosome. In bacteria, RF1 recognizes UAA and UAG 36 while RF2 recognizes UAA and UGA through their PxT and SPF codon recognition 37 motifs, respectively 1 . These RFs contain the hydrolysis active site motif, GGQ, for 38 catalysis. Polypeptide release is then followed by dissociation of the ribosome from 39 mRNA into the small and large subunits by a processes mediated by ribosome recycling 40 factor (RRF) and elongation factor G (EF-G) 2 . 41 However, the termination-recycling event can be perturbed when mRNA has 42 aberrant features, one of which is the absence of an in-frame stop codon. The mRNA 43 lacking a stop codon, called a non-stop mRNA, causes stalling of the ribosome at the 3' 44 end because recruitment of RFs to the ribosome requires the interaction of a stop codon 45 recognition motif of the RF with a cognate stop codon. Since the role of termination is 46 not only to define the end of the protein, but also to recycle the ribosome for the next 47 dCas9 was uninduced in the absence of xylose (Fig. 1b, left panels). Growth of the 143 resQ + cells was not affected by 1% xylose, which induced dCas9 to silence ssrA or 144 smpB (lanes 1, 4, 7 and 10, right panels). By contrast, growth of the ΔresQ mutant was 145 severely impaired in the presence of xylose, which led to CRISPRi-mediated repression 146 of ssrA or smpB (lanes 2, 5, 8 and 11, right panels). Expression of resQ + from an 147 ectopic locus restored the growth defect associated with the trans-translation deficiency, 148 substantiating that ResQ is responsible factor for the synthetic growth phenotype (lanes 149 3, 6, 9 and 12). These results indicate that ResQ is required for optimal growth in the 150 absence of sufficient activity of trans-translation in B. subtilis. 151 152

ResQ is expressed from a non-stop mRNA due to internal transcription 153 termination and regulated negatively by trans-translation 154
The resQ gene contains 71 sense codons followed by a stop codon. We note that it 155 contains a typical rho-independent transcription terminator sequence within the coding 156 region, raising an intriguing possibility that ResQ is translated naturally from a non-stop 157 mRNA lacking the 3' region including the stop codon (Fig. 2a). If this were the case, 158 rescue mechanisms, trans-translation and ResQ-dependent peptidyl-tRNA hydrolysis 192 (see below). The trans-translation-dependent down-regulation indicates that ResQ is the 193 secondary ribosome-rescue factor that is only produced upon dysfunction of 194 trans-translation, the primary ribosome rescue mechanism. Thus, the internal 195 transcription terminator in ResQ provides the means for this bacterium to accomplish 196 the compensatory and vectorial regulation for the maintenance of ribosome rescue 197 capability. In this context, ResQ bears a striking similarity to the E. coli alternative 198 rescue factor, ArfA, which is also synthesized from a non-stop mRNA 17,18 . 199 200

ResQ recruits RF2 to hydrolyze non-stop peptidyl-tRNAs 201
We characterized ResQ biochemically by examining whether it could induce 202 polypeptide release, as expected for an alternative ribosome rescue factor. To do this, 203 we purified ResQ in the form of ResQ62-His 6 , which lacks the C-terminal 9 amino acid 204 residues encoded by the resQ gene but not by the resQ mRNA (see above); a 205 hexahistidine (His 6 ) tag was attached to the C-terminus to aid purification. We 206 Bs hybrid PURE system 29 , a modified version of the PURE coupled 213 transcription-translation system 30 , in which the original E. coli ribosomes were replaced 214 with B. subtilis ribosomes. We omitted RFs in the Bs hybrid PURE system unless 215 otherwise stated. 216 We used DNA fragments encoding GFP but without an in-frame stop codon 217 (GFP-ns) to direct in vitro transcription and translation with Bs hybrid PURE system 218 and separated the translation products by neutral pH SDS-PAGE, which preserved the 219 peptidyl-tRNA ester bond. The major product, migrating between the 42 kDa and the 55 220 kDa markers, represents the peptidyl-tRNA (GFP-tRNA; Fig. 3a, lane 1), as treatment 221 of the sample with RNase A before electrophoresis down-shifted this band to the 222 had been mutated to GAQ, no longer stimulated hydrolysis of GFP-tRNA even in the 239 presence of ResQ (Fig. 3b, lanes 5 and 6). RF2 possesses the conserved SPF motif as an 240 essential element for the stop codon recognition in termination 31 . We addressed whether 241 this motif is required for the ribosome rescue function of RF2 by mutating it to SPT, 242 which abolishes the termination activity 31 . In the in vitro peptidyl-tRNA hydrolysis 243 assay, the RF2(SPT) was as active as the wild-type RF2 in the ResQ-dependent 244 cleavage of GFP-tRNA (Fig. 3b, lanes 9- (Fig. 3c, lanes 1 and 2), it did not work with the E. coli RFs (lanes 3 and 4). 277 Interestingly, ResQ was functional when the substrate was translated by the E. coli 278 ribosome, provided that the B. subtilis RFs were available (lanes 9-10). These results 279 suggest that ResQ interaction with RF2 is species-specific, but its interaction with the 280 ribosome is rather promiscuous. 281 By contrast, we could show that E. coli ArfA requires both the RF2 as well as the 282 ribosomes to be derived from E. coli. Specifically, ArfA did not work if combined with 283 the homologous RF2 when the substrate was translated using Bs PURE system (Fig. 3c non-stop mRNAs ( Fig. 3c and Supplementary Fig. 3), we formed ResQ-RF2-non-stop 296 70S ribosome (ns70S) complexes by incubating B. subtilis ResQ and RF2 with E. coli 297 ns70S complexes used previously for ArfA 38 . By substituting the wildtype B. subtilis 298 RF2 with a catalytically inactive GGP mutant 39,40 , peptidyl-tRNA hydrolysis and 299 therefore recycling of the ns70S complex was prevented (Supplementary Fig. 3). 300 Cryo-EM analysis of the ResQ62His-RF2-GGP-ns70S complex (herein referred as 301 ResQ-RF2-ns70S) and extensive in silico sorting of this dataset yielded a major 302 Table 1). The cryo-EM density for ResQ was well-resolved with local resolution 307 ranging between 3.0-3.6 Å (Fig. 4b), enabling residues 2-55 of ResQ to be modelled de 308 novo (Fig. 4c,d). ResQ contains an N-terminal α-helix α1 (residues 4-17) followed by a 309 short α-helical turn (α2, residues 21-25), as well as a β-strand (β1, residues 35-38) and a 310 short C-terminal α-helix (α3, 40-47) followed by a positively charged region (residues 311 48-55) (Fig. 4d). 312 313

Interaction of ResQ with the non-stop 70S ribosome 314
The binding site of ResQ is located predominantly on the 30S subunit in the 315 vicinity of the decoding center, where it spans from the top of helix 44 (h44) of the 316 16S rRNA past the ribosomal protein uS12 and reaches into the mRNA channel 317 formed by the head and body of the 30S (Fig. 4a). The overall binding site of ResQ 318 interactions with the major groove of h44 and the minor groove of H71 of the 23S 322 rRNA (Fig. 4e). In addition, Arg25 within helix α2 of ResQ, which is conserved in 323 all ResQ sequences ( Supplementary Fig. 2), stacks upon U1915 and flips C1914 out 324 of H69, where it stacks upon His133 of RF2 (Fig. 4f). This contrasts with the 325 canonical conformation of C1914 within H69 that is observed during translation 326 termination (Fig. 4g) as well as ArfA-mediated ribosome rescue. Indeed, these 327 N-terminal α-helices have no counterpart in ArfA, instead the N-terminus of ArfA is 328 unstructured and folds back to interact with uS12 36-38,41-43 (Supplementary Fig. 5a-c). 329 The C-terminal region of ResQ extends from the decoding center into the 330 mRNA channel and would be incompatible with the presence of a full-length mRNA 331 ( Fig. 4h), but compatible with a truncated non-stop mRNA (Fig. 4i). ResQ exhibits 332 a modest overlap with the second (+2) and third (+3) nucleotide of the A-site codon, 333 but extensive steric clashes would be expected for the subsequent positions (+4 334 onwards) (Fig. 4i), similar to that observed previously for ArfA 36-38,41-43 335 ( Supplementary Fig. 5d-i). Thus, ResQ may also recycle ribosomes stalled on 336 non-stop mRNAs with 1-3 nucleotides extending into the A-site, as shown 337 experimentally for ArfA 44-46 . The positively charged C-terminus of ResQ can form 338 multiple hydrogen bond interactions with 16S rRNA nucleotides that comprise the 339 mRNA channel (Fig. 4j). While the interaction network is generally distinct from 340 that observed for ArfA, we note that the mode of contact between the side chains of 341 Lys49 and His50 of ResQ with U534 of the 16S rRNA appears to be shared by 342 ArfA 36-38,41-43 (Supplementary Fig. 5j-l). (β1) to augment the β-sheet of the superdomain d2/d4 of RF2 (Fig. 5a). The overall 350 which is involved in the specificity of recognition of the first and second positions of 357 UGA/UAA stop codons 31,39,47 (Fig. 5c). Importantly, the structure illustrates that 358 ResQ, like ArfA, does not interact with the SPF motif and therefore does not directly 359 mimic the presence of a stop codon ( Fig. 5c and Supplementary Fig. 6g-i), which is 360 consistent with our observation that mutations in the SPF motif that impair RF2 361 termination activity, do not affect ResQ-RF2-mediated ribosome recycling (Fig. 3b). RF-dependent ribosome rescue pathway, which had previously been known to occur 380 only in Gram-negative bacteria (Fig. 6a). In this pathway in B. subtilis, ResQ plays a 381 critical role in the hydrolytic release of the incomplete polypeptide from the non-stop 382 stalled ribosomes. It does so by recruiting RF2 in a stop-codon-independent manner to 383 the otherwise dead-end translation complex, as shown by our biochemical experiments 384 using purified components. Our cryo-EM structure also reveals that ResQ recognizes 385 the empty mRNA channel of a non-stop ribosome complex to recruit and stabilize the 386 active (open) conformation of RF2 on the ribosome (Fig. 6b), in a similar but distinct 387 manner to ArfA 36-38,41-43 (Fig. 6c). 388 In vivo, the resQ deletion mutation exhibits a synthetic lethal phenotype when 389 combined with CRISPRi-mediated knock-down or deletion of either SsrA or SmpB. propose that ResQ is an RF-dependent ribosome rescue factor in Gram-positive bacteria, 416 such as Bacillus (Fig. 6a). 417 In the process of canonical translation termination, stop codon recognition by the 418 SPF motif of RF2 is a prerequisite for the process in which RF2 is accommodated into 419 the A-site of the ribosome and adopting a catalytically active (open) conformation 420 where the GGQ motif is directed into the PTC 39,47,50,51 . However, our structural data 421 show that ResQ does not directly mimic the stop codon in the A-site (Fig. 5C), 422 consistent with our observations that the stop codon recognition motif SPF is not 423 required for ribosome rescue ( Fig. 3b; see ref. 16 Fig.  447 5j-l). Secondly, both ArfA and ResQ contain short ß-strands that augment the ß-sheet in 448 domain 2/4 of RF2, which we presume is important to recruit RF2 to the ribosome 449 ( Supplementary Fig. 5a-c). 450 Here we show that ResQ works with RF2, but not RF1, reminiscent of the partner 451 selectivity described previously for ArfA 16 . Analysis of the contacts between ResQ and 452 RF2 within the ResQ-RF2-ns70S complex, and comparison with the sequence 453 alignments between B. subtilis RF2 and RF1, indicated that there are two main regions 454 in the RFs that are likely to be responsible for the selectivity of ResQ (Supplementary 455 Fig. 8). These encompass residues within the ß-sheet of domain 2/4 of RF2 that are in 456 however, is the lack of sequence conservation between the switch loops of B. subtilis 460 RF1 and RF2. In fact, the switch loop of RF1 is one residue longer than in RF2 461 ( Supplementary Fig. 8b,d). We also note that the RF-dependent rescue factors show 462 low interspecies compatibility (Fig. 3c). The incompatibility of ResQ to work with 463 E. coli RF2 is not surprising given the relatively low sequence conservation observed 464 within the switch region (Supplementary Fig. 8d). Moreover, sequence differences are 465 also observed with α-helix 7 between B. subtilis and E. coli RF2 that could contribute to 466 the interactions with ResQ (Supplementary Fig. 8c). Collectively, these differences Because ResQ is not effectively produced in trans-translation proficient cells, but 499 induced strikingly upon dysfunction of trans-translation, it is likely to represent a 500 secondary, back-up rescue system that compensates for defects in trans-translation. This 501 scenario reinforces the notion that the proteolytic function characteristically associated 502 with the SmpB-SsrA system is not essential for growth, as reported previously 13,14 . 503 Indeed, the GFP-ns non-stop product accumulates in the ssrA-deleted cell (Fig. 4a), 504 indicating that the products of the ResQ system are not necessarily toxic. Thus, the 505 growth-essential roles of trans-translation and ResQ are in their ribosome recycling 506 functions, rather than in the tagging-proteolysis that the ResQ system lacks. Cell 507 viability would require a sufficient pool of the uncompromised ribosomes, which 508 maintains the translation capacity of the cell. Whereas the two pathways share the 509 essential function required for growth-supporting ribosome rescue, the proteolytic 510 functions of the trans-translation and that of the RqcH tail-adding system could become 511 more important under more severe stress conditions 49 . 512 The regulatory scheme of ResQ expression is strikingly similar to that 513 elucidated for ArfA regulation in E. coli 17,18 ; the arfA mRNA is also subject to 514 RNase III-dependent cleavage and/or transcription termination such that ArfA only 515 accumulates when trans-translation is defective. It is noteworthy that the two 516 evolutionarily unrelated rescue factors employ a similar scheme of regulation. 517 Convergent acquisition of such regulatory mechanisms may be more common for 518 factors that have evolved recently and which have functions related to the firmly 519 established and relatively rigid constituent of the cell, such as the ribosome and 520 translation factors. 521 In summary, our study reveals that bacteria, both Gram-negative and 522 Gram-positive, have RF-dependent mechanisms of ribosome rescue that allow for the 523 stop-codon independent liberation of the polypeptide from the ribosome on the non-stop 524 mRNA (Fig. 6a). However, the crucial adapter proteins, ArfA, ArfT and ResQ, are 525 unrelated in amino acid sequence (Supplementary Fig. 2). The modes of their To isolate B. subtilis mutants whose viability depends on trans-translation, we 576 used the BKE library (a collection of single-gene knockout mutants covering the 3,968 577 non-essential genes, which had been disrupted by replacement with the erythromycin 578 resistance marker) as the source of gene knockouts. We pooled the BKE strains and 579 prepared a genomic DNA mixture using Wizard genome DNA purification kit 580 (Promega). We used this DNA preparation to transform an smpB-deleted strain of B. 581 subtilis that harbored a rescue plasmid carrying smpB + and lacZ + (pNAB1286), which 582 was constructed from pLOSS* with a temperature sensitive (Ts) replication system. 583 plasmid from bacteria that did not need the rescue plasmid, followed by further growth 585 at 37°C overnight and plating on LB agar containing 40 µg/mL X-Gal 586 (5-bromo-4-chloro-3-indolyl β-D-galactopyranoside), 1 mM IPTG, 12.5 µg/mL 587 lincomycin and 1 µg/mL erythromycin. We picked up blue (lacZ + ) colonies to obtain 588 transformants that retained the smpB + -lacZ + rescue plasmid even after the 589 high-temperature incubation and prepared chromosomal DNA from them. Genes that 590 had been disrupted by the erythromycin-resistance marker were determined by PCR 591 amplification and DNA sequencing of the mutant-specific barcode sequence, using 592 appropriate primers 27 . 593

CRISPR interference 595
The CRISPRi was performed as described previously 55 with some modification.  Table 7) 655 and allowed to continue at 37°C for 20 min. Samples were then mixed with the same 656 volume of 2xSDS-PAGE loading buffer. When indicated, they were further treated with 657 0.2 mg/mL RNaseA (Promega) at 37°C for 15 min before electrophoresis. Samples for 658 SDS-PAGE were heated at 65°C for 5 min and separated by 10% wide range gel 659 (Nacalai Tesque). Translation products were detected by Western blotting using ribosome-like features were then selected for 3D refinement using an E. coli 70S 688 ribosome as a reference structure (Supplementary Fig. 4a). 3D classification was then 689 performed, resulting in 317,095 particles containing P-site tRNA and RF2 690 ( Supplementary Fig. 4b) that were further selected for focus sorting on the RF2 691 ( Supplementary Fig. 4c). Focus sorting yielded a major population of 154,405 692 particles containing stoichiometric amounts of ResQ, P-site tRNA, RF2, which after a 693 final round of 3D refinement produced a final cryo-EM reconstruction with an average 694 resolution of 3.15 Å according to FSC 0.143 criterion (Supplementary Fig. 4d)  a, A schematic representation of the synthetic lethal screening to identify genes whose absence causes synthetic lethal phenotype with the deficiency of trans-translation. b, Xylose-inducible CRISPRi was targeted to ssrA (lines 1-3, 7-9) or smpB (lines 4-6, 7-9) in the B. subtilis strains indicated at the left by the genotypes of the resQ gene (ΔresQ/resQ + signifies the presence of resQ + in an ectopic locus). Cultures prepared in the absence of xylose were serially diluted (from 10 -2 to 10 -5 ) and spotted onto LB agar plates with or without 1% xylose, as indicated at the top, for incubation at 30°C (upper) or 37°C (lower) for 17 hours.  RF2 cleaves the GFP-tRNA non-stop translation product. In vitro translation using Bs hybrid PURE system was directed by the gfp-ns template. The reaction mixtures contained purified ResQ62-His 6 (lanes 7-12), purified B. subtilis RF1 (lanes 3, 4, 9, 10) and RF2 (lanes 5, 6, 11, 12), as indicated. Translation was allowed to proceed at 37°C for 20 min, and the products were divided into two parts, one of which was treated with RNaseA, as indicated. Samples were then analyzed by SDS-PAGE under neutral pH conditions, followed by anti-GFP immunoblotting. b, ResQ-dependent peptidyl-tRNA hydrolysis activity of RF2 requires its GGQ active site but not SPF stop codon recognition motif. In vitro translation using Bs hybrid PURE system was directed by the gfp-ns template in the presence of combinations of ResQ, wild type RF2 (lanes 3, 4, 9, 10), RF2(GAQ) (lanes 5, 6) and RF(FPT) (lanes 11, 12), as indicated. The translation products were analyzed by anti-GFP immunoblotting as described above. c, Interspecies compatibility of the RF-dependent rescue factor functions. In vitro translation of gfp-ns was carried out using Bs hybrid PURE system (lanes 1-8) or Ec PURE system (lanes 9-16) in the presence of combinations of purified ResQ (lanes 1-4, 9-12), E. coli ArfA (lanes 5-8, 13-16), RFs (RF1 plus RF2) purified from B. subtilis (lanes 1, 2, 5, 6, 9, 10, 13, 14) and RFs purified from E. coli (lanes 3, 4, 7, 8, 11,   12, 15, 16). The translation products were analyzed by anti-GFP immunoblotting as described above. Fig. 4 | Cryo-EM structure of ResQ-RF2-ns70S complex. a, Different views of the cryo-EM map of the ResQ-RF2-ns70S complex with isolated densities highlighting the 30S (yellow; hd, head and bd, body) and 50S (grey; CP, central protuberance) subunits, P-site tRNA (green), RF2 (orange) and ResQ (blue). b-c, Isolated electron density for ResQ (b) colored according to local resolution and (c) shown as mesh (grey) with fitted molecular model for ResQ. d, Model for ResQ with features highlighted corresponding to the schematic of ResQ protein, including α-helical and β-strand regions. e, The N-terminus of ResQ (blue) interacts both h44 of the 16S rRNA (yellow) and H69 and H71 of the 23S rRNA (grey). f, The conserved R25 of ResQ (blue) stacks upon U1915 and causes C1914 to flip out and stack upon H133 of RF2. g, Same view as (f) but showing the RF2 stop (lime) and the conformation of H69 for a canonical termination complex (PDB ID 4V5E 47 ). h, Transverse section of the 30S subunit (yellow) to reveal the mRNA channel showing a superimposition of full-length mRNA (FL-mRNA, cyan) with truncated non-stop mRNA (TR-mRNA, teal), P-site tRNA (green) and surface representations of ResQ (blue). i, Superimposition of FL-mRNA (cyan) with TR-mRNA (teal), P-site tRNA (green) and transparent surface representation of ResQ (blue). The first (+1), second (+2) and third (+3) nucleotides of the A-site codon of the FL-mRNA are indicated. j, Interaction of the C-terminus of ResQ (blue) with the 16S rRNA showing potential hydrogen bonds with yellow dashed lines.