Maintenance of the Shigella sonnei Virulence Plasmid Is Dependent on Its Repertoire and Amino Acid Sequence of Toxin-Antitoxin Systems

ABSTRACT Shigella sonnei is a major cause of bacillary dysentery and an increasing concern due to the spread of multidrug resistance. S. sonnei harbors pINV, an ∼210 kb plasmid that encodes a type III secretion system (T3SS), which is essential for virulence. During growth in the laboratory, avirulence arises spontaneously in S. sonnei at high frequency, hampering studies on and vaccine development against this important pathogen. Here, we investigated the molecular basis for the emergence of avirulence in S. sonnei and showed that avirulence mainly results from pINV loss, which is consistent with previous findings. Ancestral deletions have led to the loss from S. sonnei pINV of two toxin-antitoxin (TA) systems involved in plasmid maintenance, CcdAB and GmvAT, which are found on pINV in Shigella flexneri. We showed that the introduction of these TA systems into S. sonnei pINV reduced but did not eliminate pINV loss, while the single amino acid polymorphisms found in the S. sonnei VapBC TA system compared with S. flexneri VapBC also contributed to pINV loss. Avirulence also resulted from deletions of T3SS-associated genes in pINV through recombination between insertion sequences (ISs) on the plasmid. These events differed from those observed in S. flexneri due to the different distribution and repertoire of ISs. Our findings demonstrated that TA systems and ISs influenced plasmid dynamics and loss in S. sonnei and could be exploited for the design and evaluation of vaccines. IMPORTANCE Shigella sonnei is the major cause of shigellosis in high-income and industrializing countries and is an emerging, multidrug-resistant pathogen. A significant challenge when studying this bacterium is that it spontaneously becomes avirulent during growth in the laboratory through loss of its virulence plasmid (pINV). Here, we deciphered the mechanisms leading to avirulence in S. sonnei and how the limited repertoire and amino acid sequences of plasmid-encoded toxin-antitoxin (TA) systems make the maintenance of pINV in this bacterium less efficient compared with Shigella flexneri. Our findings highlighted how subtle differences in plasmids in closely related species have marked effects and could be exploited to reduce plasmid loss in S. sonnei. This should facilitate research on this bacterium and vaccine development.

this important pathogen and be exploited for the development and evaluation of vaccines.

RESULTS
The absence of gmvAT and ccdAB from pINV Ssonn contributes to plasmid loss. Spontaneous avirulence occurs more frequently in S. sonnei than S. flexneri due to pINV loss (14,15,17,26,27). pINV Sflex carries three functional plasmid maintenance TA systems, vapBC, gmvAT, and ccdAB, while gmvAT and ccdAB have been lost from pINV Ssonn through ancestral deletions (17). To determine whether the absence of ccdAB and gmvAT from pINV Ssonn is responsible for pINV loss, we analyzed a strain, ccdAB 1 / gmvAT 1 , with ccdAB and gmvAT and their promoters introduced into pINV Ssonn at sites corresponding to their positions in S. fexneri (17). The emergence of avirulent bacteria in ccdAB 1 /gmvAT 1 and wild-type (WT) S. sonnei was assessed in Congo red (CR) binding assays. Avirulent Shigella that lost T3SS expression do not bind CR when grown on CR-containing media and appear white (CR 2 ) while virulent T3SS-expressing bacteria bind CR and appear red (CR 1 ) (28)(29)(30). Therefore, we measured the emergence of avirulence as the emergence of CR 2 colonies and then characterized these colonies by multiplex PCR. To prevent pINV Ssonn loss before the start of experiments, we introduced a chloramphenicol resistance cassette (cat) into pINV in the WT strain and ccdAB 1 / gmvAT 1 downstream of vapC. Bacteria were initially grown in the presence of the antibiotic then allowed to replicate in the absence of chloramphenicol for ;25 generations by growth for 16 h at 37°C to allow for pINV loss during cell division. The emergence of avirulent colonies (CR 2 ) was then quantified by plating bacteria to media containing CR.
Consistent with previous work (17), the introduction of gmvAT and ccdAB into pINV Ssonn significantly reduced the emergence of avirulent CR 2 colonies, by almost two orders of magnitude compared with the WT strain (P = 0.0003 for S. sonnei 1/2 ccdAB/ gmvAT) (Fig. 1A) (17). To determine the molecular events leading to avirulence, we used multiplex PCR to examine CR 2 bacteria for the presence of (i) virF, which is present on pINV but not in the T3SS PAI and encodes a transcriptional regulator of T3SS genes, (ii) virB, which is in the PAI and encodes a transcriptional regulator, (iii) ori, the pINV origin of replication which was used as a proxy for the presence of pINV (26,31), and (iv) hns as a chromosomal control ( Fig. 1B and C). pINV loss was the dominant cause of avirulence in the WT strain, accounting for over 99% of CR 2 bacteria (Fig. 1B). The percentage of CR 2 colonies emerging from wild-type S. sonnei 53G (WT) and a strain with ccdAB and gmvAT introduced into pINV (ccdAB 1 /gmvAT 1 ) after growth at 37°C for ;25 generations. Solid line: mean (n = 9 biological replicates). ***, P # 0.001 by parametric Welch's t test. (B and C) The number of CR 2 colonies possessing pINV but lacking specified plasmid genes were assayed by multiplex PCR and shown as a percentage of all colonies. Loss of the entire plasmid is inferred by loss of the origin of replication (B). "Other" refers to CR 2 colonies that contain virB, virF and the origin of replication (C). The value of each category is indicated inside the corresponding bar portion. Eight independent CR 2 colonies were obtained from each of nine biological replicates (i.e., 72 colonies were analyzed per strain). **, P # 0.01; ****; P # 0.0001; n.s., not significant; one-sample t test (parametric comparison of mean test/WT strain to 1) and Wilcoxon signed-rank test (nonparametric comparison of the median of strain/WT to 1, per specific event). l.o.d., limit of detection.
Only 0.3% of avirulent bacteria emerging from wild-type S. sonnei retained pINV but had lost virB and/or virF (Fig. 1C). Introduction of ccdAB and gmvAT into pINV Ssonn significantly reduced pINV loss, which dropped from 2.5% to 0.02% of all colonies in the absence or presence of these TA systems, respectively (P , 0.0001) (Fig. 1B) and increased "other" events (P = 0.0070) (Fig. 1C), which lead to avirulence but not through the loss of virB, virF, and ori. This demonstrated that the introduction of ccdAB and gmvAT reduced pINV loss overall, making less frequent events more easily detectable. However, pINV loss was still the main genetic event leading to avirulence in ccdAB 1 /gmvAT 1 and occurred in almost 50% of CR 2 bacteria (Fig. 1B). Loss of virB and/ or virF, were also detected at low frequency in both strains (Fig. 1C). Therefore, the ancestral loss of ccdAB and gmvAT from pINV Ssonn does not fully explain the high rate of pINV loss in S. sonnei.
Polymorphisms in VapBC contribute to pINV loss in S. sonnei. To further understand the factors responsible for the high rate of pINV loss in S. sonnei, we examined the sequence of the plasmid for TA systems. pINV Ssonn is predicted to encode a RelBE which is absent in pINV Sflex and a VapBC TA system (32). The RelBE TA system is encoded near the origin of replication and a closely related system stabilizes p307 in E. coli (see Fig. S1A and B in the supplemental material) (33). Analysis of whole-genome sequences from a global collection of 132 S. sonnei isolates (19) showed that pINV had been lost from most of the isolates; however, RelBE was found on pINV from all 43 S. sonnei isolates that contained the pINV sequence (Table S1). Although RelBE is a functional TA system, we found that it does not contribute to pINV maintenance under laboratory conditions ( Fig. S1C and D).
Therefore, we focused our attention on the VapBC TA system, which is sufficient and necessary to maintain pINV Sflex and the only TA system found on pINV from both S. sonnei and S. flexneri (17). The VapC toxin cleaves the initiator tRNA, tRNA fMet , preventing initiation of translation, while the VapB antitoxin binds to and blocks the activity of VapC (34,35). VapBC expression is autoregulated by conditional cooperativity through binding of heterocytomeric VapBC complex to two sites in the vapBC promoter, known as operator sites (vapO), resulting in transcriptional repression when VapB is in excess (34,36). When pINV is lost following cell division, PSK occurs as VapB is readily degraded by Lon and cannot be replaced, leaving VapC free to cleave tRNA fMet (35)(36)(37).
The promoter region of vapBC, including the vapO sites, was identical in S. flexneri and S. sonnei (Fig. S2A). Comparison of the predicted amino acid sequences of this TA system from S. flexneri M90T (accession number AL391753) and S. sonnei 53G (accession number NC_016833) revealed single amino acid differences in VapB and VapC ( Fig. S2B and C). The polymorphic residue in VapB (T 58 A, S. flexneri versus S. sonnei) is located toward the C terminus of the antitoxin, which is involved in neutralizing VapC, while the polymorphic site in VapC (K 32 R, S. flexneri versus S. sonnei) is not located near its active site, in the self-dimerization interface, or its interface with VapB ( Fig. S2B and C) (34). To determine if the vapBC sequences in S. flexneri M90T and S. sonnei 53G are representative of these species, we examined available S. flexneri and S. sonnei genome sequences for vapBC. The predicted amino acid sequence of VapBC in S. flexneri M90T and S. sonnei 53G were found in most isolates belonging to these species (83% and 78%, respectively) ( Fig. 2SB and C). Therefore, we subsequently refer to VapB T 58 as VapB Sflex , VapB A 58 as VapB Ssonn , VapC K 32 as VapC Sflex , and VapC R 32 as VapC Ssonn .
Next, we investigated whether the polymorphisms in S. sonnei VapBC affect pINV loss. We generated strains of S. sonnei containing pINV with S. sonnei vapBC (S. sonnei VapBC Ssonn ), or with the locus replaced with vapBC from S. flexneri (S. sonnei VapBC Sflex ), or a chimeric vapBC (i.e., S. sonnei VapB Sflex C Ssonn and S. sonnei VapB Ssonn C Sflex ). A sonnei 53G lacking pINV. Each dot represents the result for a single colony and shows the percentage of all bacteria with plasmid loss. Solid line, mean 1 standard error of the mean (SEM) (n = 9 biological replicates). **, P # 0.01; *, P # 0.05; n.s. not significant; Wilcoxon signed-rank test (nonparametric comparison of median of VapBC1/empty pSTAB to 1). (B) The percentage of colonies that are CR 2 of wild-type S. flexneri or S. sonnei (VapBC Ssonn ), or S. sonnei with pINV harboring vapBC Sflex (VapBC Sflex ), vapB Sflex C Ssonn (VapB Sflex C Ssonn ) or vapB Ssonn C Sflex (VapB Ssonn C Sflex ) after growth at 37°C for ;25 generations; solid line, mean 1 SEM (n = 9 biological replicates). **, P # 0.01; n.s. not significant; Wilcoxon signed-rank test (nonparametric comparison of median of strain/S. flexneri to 1). (C and D) The number of CR 2 colonies possessing pINV but lacking specified plasmid genes were assayed by multiplex PCR and shown as a percentage of all colonies. Loss of the entire plasmid is inferred by loss of the origin of replication (C). "Other" refers to CR 2 colonies that contain virB, virF, and the origin of replication (D). The value of each category is indicated inside the corresponding bar portion. Eight independent CR 2 colonies were obtained from each of three biological replicates (i.e., 72 colonies were analyzed per strain). ****, P # 0.0001; ***, P # 0.001; **, P # 0.01; *, P # 0.05; n.s. not significant; Wilcoxon signed-rank test (nonparametric comparison of the median of strain/S. flexneri or S. sonnei to 1, per specific event). l.o.d., limit of detection. chloramphenicol resistance cassette (cat) was inserted downstream of vapBC in all strains to ensure retention of pINV before the start of experiments. As before, we quantified the emergence of CR 2 colonies after ;25 generations by growth for 16 h at 37°C in the absence of chloramphenicol. There was a significantly higher number of CR 2 bacteria arising from S. sonnei VapBC Ssonn and S. sonnei VapB Sflex C Ssonn compared with wild-type S. flexneri (P = 0.0039) (Fig. 2B). In contrast, there was no difference in the number of CR 2 bacteria emerging from S. sonnei VapBC Sflex or S. sonnei VapB Ssonn C Sflex compared with wild-type S. flexneri (P = 0.8203 and P = 0.4961) (Fig. 2B).
We also used multiplex PCR to investigate whether the single amino acid substitutions in VapBC altered the molecular events that lead to avirulence. Of note, the presence of vapC from S. flexneri led to a significant reduction in pINV loss as a cause of avirulence compared with S. sonnei vapC (S. sonnei VapBC Ssonn versus VapBC Sflex , P = 0.0039; S. sonnei VapBC Ssonn versus VapB Ssonn C Sflex , P = 0.0039) (Fig. 2C). In contrast, replacement of vapB from S. sonnei with vapB from S. flexneri had less effect on pINV loss (S. sonnei VapBC Ssonn versus VapB Sflex C Ssonn , P = 0.0391) (Fig. 2C). Therefore, the presence of S. flexneri vapC significantly decreased pINV loss, which was consistent with the results obtained with pSTAB. Furthermore, results demonstrated that, in addition to pINV loss, loss of virB and/or virF could also lead to avirulence in S. sonnei (Fig. 2D).
Taken together, these results demonstrated that single amino acid substitutions in the VapBC TA system influenced the frequency of pINV loss in S. sonnei with the substitution in VapC (K32R) having a major effect. However, replacement of the vapBC in S. sonnei with vapBC from S. flexneri made pINV loss undetectable in S. flexneri, suggesting that the genetic background also influenced VapBC function in S. sonnei in addition to single amino acid substitutions.
The effect of single amino acid substitutions on VapBC structure and function. Next, we investigated the mechanism(s) by which the amino acid substitutions in (B and C) The number of CR 2 colonies possessing pINV but lacking specified plasmid genes were assayed by multiplex PCR and shown as a percentage of all colonies. Loss of the entire plasmid is inferred by loss of the origin of replication (B); "other" refers to CR 2 colonies that contain virB, virF, and the origin of replication (C). The value of each category is indicated inside the corresponding bar portion. Eight independent CR 2 colonies were obtained from each of three biological replicates (i.e., 72 colonies were analyzed per strain). The value of each category is indicated inside the corresponding bar portion. ****, P # 0.0001; n.s. not significant; one-sample t test (parametric comparison of mean test/control strain to 1) and Wilcoxon signed-rank test (nonparametric comparison of the median of strain/control to 1, per specific event). l.o.d., limit of detection.
VapBC affect its function. VapBC forms a hetero-octameric complex that binds the operator sites in the promoter sequence (vapO) when VapB is in excess (34,36). To establish whether the amino acid differences affect the overall architecture of the VapBC complex, we set up crystallization trials in the presence of vapO-containing DNA. However, the crystals obtained contained protein only, enabling us to determine the atomic structure of VapBC from S. sonnei to establish whether the amino acid differences affect the overall architecture of the VapBC complex ( Fig. 4A; Table S2). Alignment of the S. sonnei and S. flexneri (PDB accession number 3TND) VapBC hetero-octamers demonstrated that they are highly similar (root mean square deviation of 0.460 Å over all Ca atoms) ( Fig. 4A and Table S2), indicating that the substitutions did not significantly alter the complex.
To determine the influence of the VapC K 32 R polymorphism on its toxicity, we expressed each version of VapC under the control of an arabinose-inducible promoter in S. sonnei lacking pINV. An inactive version of the toxin, VapC D 7 A, was included as a negative-control (Fig. 4B) (35). Expression of vapC Sflex (encoding VapC K 32 ) or vapC Ssonn (encoding VapC R 32 ) led to a similar reduction in bacterial viability from 1 h postinduction onwards (P = 0.6867, one-way ANOVA, pBAD33::VapC Ssonn versus pBAD33:: VapC Sflex ) (Fig. 4B). Because the VapC K 32 R polymorphism was neither in the catalytic site nor at its interface with VapB, it could be involved in VapC cleavage of tRNA fMet . Therefore, to determine the influence of the VapC K 32 R polymorphism on VapC activity, we measured the cleavage of tRNA fMet , the target of the toxin, using an in vivo cleavage assay as previously described (35). VapC from S. sonnei or S. flexneri was expressed under an arabinose-inducible promoter in E. coli, and the cleavage of tRNA fMet was followed by Northern blotting. There was no detectable difference in the cleavage of tRNA fMet by VapC Ssonn and VapC Sflex (Fig. 4C).
We also investigated whether the VapC K 32 R substitution altered the capacity of either VapB Sflex or VapB Ssonn to act as an antitoxin, although this residue was not located at the TA interface. VapC was expressed under the control of an arabinose-inducible promoter in S. sonnei lacking pINV with either homologous or heterologous vapB coexpressed under the control of its native promoter. There was no significant difference in the ability of either VapB to neutralize either VapC (P = .0.9999, one-way ANOVA) (Fig. 4D).
In conclusion, we found that the polymorphisms in VapBC that influence pINV maintenance did not affect the structure of the VapBC complex, VapC toxicity, or the ability of VapB to act as an antitoxin.
IS-mediated deletions in pINV Ssonn lead to avirulence. pINV loss is not the only event that can result in avirulence in S. sonnei. Multiplex PCR of avirulent bacteria emerging from S. sonnei strains showed that some CR 2 colonies retained pINV but lost virB and/or virF (Fig. 1C and 2D), which is similar to S. flexneri (26). Of note, there were marked differences in the frequency of these events in S. flexneri strains compared to S. sonnei strains (Fig. 2D). Loss of virB alone dominated in wild-type S. flexneri as previously shown (26) but was infrequently detected in S. sonnei (S. flexneri versus S. sonnei VapBC Ssonn and S. sonnei VapB Ssonn C Sflex P = 0.0078; S. flexneri versus S. sonnei VapBC Sflex and S. sonnei VapB Sflex C Ssonn , P = 0.0039) (Fig. 2D). In contrast, loss of virF only and loss of virF with virB were rarely detected in S. flexneri but were relatively common in S. sonnei (Fig. 1C, 2D, and 3C).
Because the emergence of avirulence in S. flexneri usually occurs following recombination between homologous copies of ISs (26), we investigated whether the loss of virB and/or virF in S. sonnei resulted from IS-mediated deletions. We examined pINV Ssonn for pairs of ISs that occur in the same orientation and flank the T3SS PAI and/ or virF. Using primers binding external to selected ISs, we amplified and sequenced potential IS-mediated deletions. This revealed that three pairs of ISs could contribute to avirulence in S. sonnei. To examine whether the deletions leading to avirulence in S. sonnei were due to the presence of distinct ISs on pINV compared with S. flexneri, we compared the content of ISs in pINV from these two species. In S. flexneri M90T pINV (pWR100), ISs accounted for 53% of open reading frames, and ISSfl4, IS1294, and IS600 were responsible for most T3SS PAI deletions (Fig. 5B) (26,39). pINV Sflex harbored a total of 29 different ISs, consisting of 12 ISs types. IS1294 was the most abundant IS (present in six copies) with five copies of IS629 and four of IS600 on the plasmid, respectively. Only four ISs were present in a single copy ( Fig. 5C and D). In contrast, only ;23% of open reading frames in pINV Ssonn were intact ISs with 26 ISs belonging to 13 different types. IS600 was the most abundant IS in pINV Ssonn with five copies on the plasmid. The remaining ISs were present in up to three copies per plasmid, and seven were only present in a single copy (Fig. 5C and D). Therefore, there was a substantial difference in the repertoire and organization of ISs on the two plasmids with pINV Sflex harboring multiple copies of more ISs than pINV Ssonn .
Taken together, these results demonstrated that the number of ISs and their location influenced the different recombination events leading to avirulence in S. sonnei and S. flexneri.

DISCUSSION
Here, we investigated the molecular mechanisms of pINV maintenance in S. sonnei and the reasons for the frequent emergence of avirulence (17). While three functional TA systems (VapBC, GmvAT, and CcdAB) are found on pINV Sflex (17,21,40,41), pINV Sonn only has two TA systems, VapBC and RelBE, both of which are close to the origin of replication. While a RelBE homolog can maintain plasmids in E. coli (33,42), this system does not contribute to the maintenance of pINV Ssonn under the conditions tested (see Fig. S1 in the supplemental material). Aside from the repertoire of TA systems, we showed that polymorphisms in VapBC and the number, type, and organization of ISs on pINV determine plasmid loss and plasticity in S. sonnei.
We confirmed that ancestral deletions affecting two TA systems, GmvAT and CcdAB, increase the loss of pINV Ssonn (17). However, even when GmvAT and CcdAB were introduced into pINV Ssonn , ;50% of avirulent bacteria still resulted from pINV loss (Fig. 1B). VapBC (also referred to as MvpAT) is the only TA system present in pINV from both S. flexneri and S. sonnei and plays a critical role in maintaining pINV Sflex (17,21,40,41). Given this, we hypothesized that differences in VapBC could also contribute to the high pINV loss in S. sonnei. We found that single amino acid polymorphisms in VapBC, particularly the substitution in VapC, have a marked effect on the function of this TA system in the model plasmid pSTAB and pINV.
The K 32 R substitution in VapC could affect pINV loss by several mechanisms during PSK. Of note, this change in VapC is not in the active site of the toxin, which is required for tRNA fMet cleavage (35) or in the region that interacts with VapB or in the toxin dimerization region (Fig. 4A) (34,36). To investigate how the VapC polymorphism influences pINV loss, we determined the atomic structure of the S. sonnei VapBC complex and found that the amino acid substitutions do not affect the overall structure of VapBC (Fig. 4A). Furthermore, there was no detectable change in the toxicity of VapC, its ability to cleave tRNA fMet , or its susceptibility to being neutralized by VapB (Fig. 4B to D). However, these assays might not be sufficiently sensitive to detect subtle yet important differences in VapC activity and/or VapB antitoxicity or interactions of VapBC with its operator DNA (Fig.  S2A) (36). For PSK, VapC must be liberated from VapB. Therefore, it is also possible that the VapC K 32 polymorphism reduces its affinity for VapB, increasing levels of free toxin in bacteria and enhancing PSK. Alternatively, the change in VapC might affect the degradation of VapB by the cellular proteases. Further studies are underway to examine these possibilities and the impact of the other VapBC sequence variants we identified ( Fig. S2B and C) on the activity of this TA system.
ISs are critical for the architecture and evolution of plasmids in Shigella (39,43). Indeed, recombination between ISs is likely to have resulted in the acquisition of the T3SS PAI (43). The spontaneous emergence of avirulence in S. flexneri is largely mediated by deletions of the T3SS PAI following intramolecular recombination between homologous copies of ISs (26,31,44). We found that some avirulent S. sonnei also arise following deletions of PAI-associated genes, but these events involve different ISs than S. flexneri because pINV from these two species harbor distinct IS profiles (Fig. 5C and  D). Additionally, deletions involving the T3SS and its regulators may go undetected in S. sonnei because of the high level of pINV loss (Fig. 2C). As ISs can act as the substrate for horizontal gene flow, differences in ISs in pINV Ssonn and pINV Sflex might also affect the repertoire of T3SS effectors in the two species.
TA systems can stabilize local sequences by preventing IS-mediated deletions (26). Our results suggested that GmvAT might reduce IS1294-mediated loss of virF in S. flexneri because this system is located 7.4 kb from virF on pINV Sflex within a region bounded by two copies of IS1294 (39). The absence of gmvAT from pINV Ssonn might explain our finding that avirulent strains lacking virF only were more common in S. sonnei than S. flexneri (Fig. 2D). This was further supported by the decrease in virF loss observed when we introduced gmvAT into its native site in pINV Ssonn (Fig. 1C).
S. sonnei is a highly successful pathogen despite having an unstable virulence plasmid when grown in the laboratory (14)(15)(16). pINV may be highly stable in S. sonnei in vivo, with environmental signals in the intestinal tract enhancing pINV maintenance. However, the emergence of avirulent bacteria does generate phenotypic heterogeneity, which might be beneficial for the survival of virulent and/or avirulent S. sonnei in certain circumstances. For example, rapidly replicating avirulent bacteria could act as an immune decoy in the intestinal tract or provoke local inflammatory responses and modify the local microenvironment in favor of virulent bacteria (45). Of note, pINV from S. sonnei harbors genes responsible for the biosynthesis of the O antigen which is incorporated into the lipopolysaccharide (LPS) and capsule of the bacterium (14,15,46,47). Therefore, loss of pINV in S. sonnei also results in loss of the O antigen and capsule, which could enhance inflammatory responses to infection (47,48).
Our findings could facilitate molecular studies of S. sonnei and aid the development of vaccines and animal/human challenge models (49). Vaccines based on live attenuated S. sonnei strains can be adversely influenced by pINV loss (18,(49)(50)(51)(52), which might affect other vaccine strategies, including those based on outer membrane vesicles by reducing levels of T3SS effectors and O antigen, which are both critical immunogens (20). However, stabilization of pINV in live vaccine candidates would require attenuation of virulence through additional genetic changes (49). Finally, detailed knowledge of plasmid dynamics in S. sonnei might offer novel approaches to target the plasmid maintenance systems to combat the threat posed by this important human pathogen.
Construction of strains. Lambda Red recombination was employed to construct mutants (53). Approximately 1 kb of sequence upstream and downstream of the gene interest was used to flank an antibiotic resistance cassette. Fragments were amplified by PCR (primers used in this study are shown in Table S5) then ligated into pUC19 (54) using NEBuilder HiFi master mix (New England Biolabs). The resulting plasmids were transformed into E. coli DH5a and used as the template to generate approximately 1 mg of linear DNA by PCR, which was transformed by electroporation into S. sonnei 53G expressing the recombinase from pKD46 (53). Bacteria were plated onto CR-TSA plates containing appropriate antibiotics and incubated overnight at 37°C. Strains were checked by PCR and sequencing. For construction of S. sonnei ccdAB 1 /gmvAT 1 , VapBC Sson /WT, VapBC Sflex , VapB Sflex C Ssonn , VapB Ssonn C Sflxe and S. flexneri, the cat cassette was introduced at nucleotide 204,058 in S. sonnei 53G pINV (available at https://www .ncbi.nlm.nih.gov/nuccore/NC_016833.1?report=genbank) and at site 191,776 bp in S. flexneri M90T (accession number AL391753), respectively (26). For S. sonnei ccdAB 1 /gmvAT 1 , ccdAB and gmvAT were introduced into pINV Ssonn at sites corresponding to their positions in S. fexneri (17).
For the construction of pET28a-VapBC Ssonn , vapB and vapC were amplified from S. sonnei genomic DNA using the primers described in Table S5 and then inserted into pET28a digested with NdeI and XhoI using NEBuilder HiFi master mix.
Triparental mating. To transfer pINV Ssonn into S. flexneri, tri-parental mating was performed using S. sonnei 53G with a cat cassette downstream of vapBC (conferring chloramphenicol resistance) as the donor strain, S. flexneri BS176 (streptomycin resistant) as the recipient, and E. coli containing pRK2013 (conferring kanamycin resistance) as the helper strain (55,56). Strains were grown separately overnight at 37°C in 10 mL LB liquid medium with appropriate antibiotics. Cultures were resuspended in 10 mL PBS after washing and then mixed at a ratio of 1:1:5 of donor, helper, and recipient, respectively, in 100 mL. Bacteria were spotted onto LB agar without antibiotics and incubated overnight at 37°C. Bacteria were harvested from spots and resuspended in 1 mL TSB and plated onto CR-TSA containing chloramphenicol and streptomycin to select for transconjugants.
sacB-neo assays. S. sonnei strains possessing pINV with the sacB-neo cassette in mxiH or pSTAB derivatives (17,37) were grown from frozen stocks at 37°C for 16 h on solid LB medium to reach ;25 generations. On three separate occasions, three colonies were resuspended in 100 mL PBS and serial dilutions plated on solid media with sucrose or with kanamycin only. PAI loss was calculated as the ratio of CFU on plates with sucrose and on plates without sucrose/with kanamycin and shown as a percentage.
CR-binding assays. Shigella spp. were grown at 37°C on CR-TSA plates containing chloramphenicol overnight to obtain single colonies. On three separate occasions, three independent CR 1 colonies were resuspended in a 5 mL TSB liquid medium and incubated at 37°C with shaking at 180 rpm for 16 h (;25 generations). Samples were diluted in PBS and plated onto CR-TSA and incubated overnight at 37°C before counting the number of CR 1 and CR 2 colonies. The proportion of CR 2 colonies was quantified by dividing the number of emerging CR 2 colonies by the total number of colonies (CR 1 and CR 2 ) and expressed as a percentage. Colonies were assessed by visual inspection.
Multiplex PCR. Multiplex PCR was used to detect virF, virB, and ori (the pINV origin of replication). hns was included as a chromosomal control (26). Reactions included Taq polymerase (Sigma-Aldrich) with an annealing temperature of 51.2°C and an extension time of 1.5 min. For each strain, eight CR 2 colonies emerged from each biological repeat following the CR-binding assay were analyzed. For each strain, the percentage of all colonies showing a given gene loss was calculated by first quantifying the percentage of CR 2 colonies with a particular gene loss and then adjusting the result using the percentage of CR 2 colonies in the total bacterial population as described in the CR-binding assay.
Toxicity, anti-toxicity, and VapC-mediated cleavage assays. To assess the toxicity of VapC, the protein was expressed using pBAD33 (57) in S. sonnei 53G lacking pINV. Initially, bacteria were grown at 37°C with shaking at 180 rpm in LB liquid medium with 0.2% glucose (wt/vol) to repress toxin expression. At an optical density (OD) OD 600 ;0.1 cultures were pelleted by centrifugation at 3000 Â g then resuspended in LB liquid medium with 0.2% arabinose (wt/vol) to induce toxin expression and grown at 37°C with shaking at 180 rpm. Aliquots of cultures were taken at time points following induction then serially diluted in PBS and plated onto LB solid media containing 0.2% glucose (wt/vol) to measure bacterial viability.
To assess the ability of VapB to prevent VapC toxicity, VapB from S. flexneri or S. sonnei was expressed from their native promoter on pGM101 in S. sonnei 53G lacking pINV with or without pBAD33::vapC Sflex or pBAD33::vapC Ssonn (17). Expression of VapC was induced as above in bacteria with or without a plasmid containing vapB, and viability was assessed by plating aliquots of cultures to solid media with 0.2% glucose (wt/vol).
To assess the ability of VapC to cleave tRNA fMet , E. coli MG1655 containing pBAD33::VapC Sflex or pBAD33:: VapC Ssonn were grown exponentially in LB liquid medium at 37°C with shaking at 180 rpm. At an OD 600 of ;0.4, toxin expression was induced by the addition of 0.2% arabinose (wt/vol). Samples (1 mL) were collected before (0 min) and 2, 4, and 6 min after addition of arabinose then immediately mixed with 125 mL of 5% phenol in ethanol on ice to prevent further RNA degradation. Samples were harvested by centrifugation at 3000 Â g for 5 min at 4°C, and RNA was extracted using the hot phenol method as previously (58). Total RNA (2.5 mg) was denatured in formamide and separated on a denaturing 8% polyacrylamide gel (19:1) containing 8 M urea buffered in Tris-borate-EDTA (TBE). The RNA was transferred to a Zeta-Probe membrane (Bio-Rad) by electroblotting. Membranes were prehybridized in hybridization buffer (0.9 M NaCl, 0.05 M NaH 2 PO 4 , 0.05 M EDTA, 5Â Denhardt's solution [Thermo Fisher Scientific], 0.5% SDS and 550 mg salmon sperm DNA, pH 7.4) for 30 min at 42°C before the addition of the DNA probe. The probe was generated by 59 phosphorylation of 30 pmol of tRNA fMet specific DNA oligonucleotide (59-CTTCGGGTTATGAGCCCGACGAGCTA) with 30 mCi [ 32 P]-ATP using 1 mL T4 polynucleotide kinase (Thermo Fisher Scientific) in a total volume of 20 mL according to manufacturer's instructions. Hybridization was continued overnight at 42°C. To reduce nonspecific hybridization, membranes were washed in 2ÂSSC (0.3 M NaCl, 0.03 M Na 3 C 6 H 5 O 7 ) with 0.1% SDS at room temperature. Cleavage of tRNA fMet was visualized by phosphorimaging.
Protein purification and crystallography. pET28a-VapBC Ssonn was transformed into E. coli C41. Following growth at 37°C to an OD 600 of ;0.8, expression was induced with 1 mM IPTG and cultures incubated for a further 3 h before harvesting by centrifugation at 5000 Â g for 10 min. VapBC was purified as described previously (34). In brief, 150 mM VapBC Ssonn octamer was combined with doublestranded DNA (59-ACAATAGATATACACAAGACATATCCAC-39) resuspended in H 2 O using a ratio of 1:1.2 of VapBC Ssonn octamer to DNA. The VapBC Ssonn :DNA mixture was dialyzed in a solution of 25 mM Tris, pH 8.0, for 5 hours at room temperature in a Slide-A-Lyzer cassette (ThermoFisher Scientific) with a 3.5 kDa cutoff. Crystals were grown using the sitting drop method in 0.1 M ammonium sulfate, 0.3 M sodium formate, 0.1 M sodium cacodylate (pH 6.5), 3% g-PGA, and 5% PEG 4000 at a ratio of 0.4:0.6 of protein to mother liquor. Crystals obtained were formed of VapBC Ssonn alone. Crystals were briefly incubated in a solution of crystallization buffer supplemented with 40% ethylene glycol followed by flash freezing in liquid nitrogen. Data were collected on beamline I04-1 at Diamond Light Source and indexed, scaled, and reduced using 2Â dials (59) within ISPyB (60). The structure was solved by molecular replacement using PHASER (61) within CCP4i (62) with the structure of VapBC Sflex (accession number 3TND), (34) as the starting model. Iterative manual rebuilding and refinement using Coot (63) and PHENIX (64) led to the models described in Table S2.
Statistical analyses. All data were analyzed using GraphPad Prism (version 7.4). The following statistical analyses were used as indicated in the legends of figures or the text: t test, one-sample t test/ Wilcoxon signed-rank test, one-way ANOVA, and two-way ANOVA. For t test, data were first analyzed for normal distribution using the Shapiro-Wilk test, and, depending on the result, parametric (Welch's t test) or nonparametric t test (Mann-Whitney test) were performed. For one-sample t test/Wilcoxon signedrank test analysis, data were log-transformed and divided by the respective control strain. The Shapiro-Wilk test was then used to test for normal distribution. If data were normally distributed, a parametric one-sample t test was performed to compare the mean of data of each strain to 1 because 1 is the result of the ratio control/control or to the limit of the detection (LOD = 0.00001, i.e., 0.001%), which was equal to 25 log LOD in the absence of the control strain. If the data were not normally distributed, Wilcoxon signed-rank test was performed, comparing the median of data of each strain to 1 or the LOD. For oneway ANOVA, Kruskal-Wallis, Dunn's multiple-comparison test was performed. For two-way ANOVA, Sidak's or Tukey's multiple-comparison test was performed.
Data availability. Coordinates and structure factors have been deposited in the Protein Data Bank (accession number 6SD6).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.9 MB.