Identification and characterization of the chromosomal yefM-yoeB toxin-antitoxin system of Streptococcus suis

Toxin-antitoxin (TA) systems are widely prevalent in the genomes of bacteria and archaea. These modules have been identified in Escherichia coli and various other bacteria. However, their presence in the genome of Streptococcus suis, an important zoonotic pathogen, has received little attention. In this study, we describe the identification and characterization of a type II TA system, comprising the chromosomal yefM-yoeB locus of S. suis. The yefM-yoeB locus is present in the genome of most serotypes of S. suis. Overproduction of S. suis YoeB toxin inhibited the growth of E. coli, and the toxicity of S. suis YoeB could be alleviated by the antitoxin YefM from S. suis and Streptococcus pneumoniae, but not by E. coli YefM. More importantly, introduction of the S. suis yefM-yoeB system into E. coli could affect cell growth. In a murine infection model, deletion of the yefM-yoeB locus had no effect on the virulence of S. suis serotype 2. Collectively, our data suggested that the yefM-yoeB locus of S. suis is an active TA system without the involvement of virulence.


Results
Identification of the yefM-yoeB locus in S. suis. To identify the YoeB toxin homologs, a BlastP search against the proteins annotated in the genome of S. suis SC84 39 was performed, using the YoeB sequences of S. pneumoniae R6 and E. coli MG1655 as query sequences. Both searches revealed an open reading frame (SSUSC84_1817) encoding an 85 amino acid protein sharing 75% and 50% identity with the S. pneumoniae and E. coli YoeB toxins, respectively. Hence, the protein was termed YoeB. The YefM antitoxin encoded by SSUSC84_1818 was identified by the same method. BlastP analysis showed that the S. suis YefM has 79% and 29% amino acid sequence identity with YefM from S. pneumoniae and E. coli, respectively. Multiple sequence alignments further revealed that 1) the S. suis YefM-YoeB system shares high level of homology with that from S. pneumoniae and E. coli; 2) YoeB processes several conserved residues required for its activity (Glu46, Arg65, His83 and Tyr84) 35 (Fig. 1). Protein homology modeling using CPHmodels predicted the structures of S. suis YefM and YoeB. The secondary structure of YefM is proposed to consist of four α -helices and three β -sheets (Fig. 1a), while that of YoeB contains two α -helices, five β -sheets and a coil (Fig. 1b). To determine whether the yefM-yoeB locus was universally present in the genomes of S. suis, BlastN analysis was performed using the 21 complete genomes of S. suis available in the National Centre for Biotechnology Information database as of 31 May 2015. The results confirmed that all strains harbour the yefM-yoeB locus, except for strain D12, a serotype 9 S. suis (see Supplementary Fig. S1 online).
A genetic structure analysis revealed that yefM is located upstream of yoeB, and the two genes are separated by one nucleotide, apparently arranged in a bicistronic operon (Fig. 2a). BPROM analysis of the upstream region of the yefM gene (about 300 bp) identified the putative − 35 and − 10 regions, which are located in the intergenic region between the yefM gene and its upstream gene (Fig. 2a). To assess whether yefM and yoeB are co-transcribed in S. suis, a reverse transcription polymerase chain reaction (RT-PCR) analysis was performed. Reverse transcriptase was used to synthesize cDNAs and the resulting cDNAs were PCR amplified using primer pair A1/T2. The individual yefM and yoeB genes were also amplified using primer pairs A1/A2 and T1/T2, respectively. As shown in Fig. 2b, the PCR products Scientific RepoRts | 5:13125 | DOi: 10.1038/srep13125 were of the expected sizes for yefM (261 bp), yoeB (258 bp) and yefM-yoeB (520 bp), all consistent with that of the genomic DNA. No PCR products were evident in the negative controls, in which the reverse transcription was performed without the enzyme, therefore eliminating possible DNA contamination. These results demonstrated that in S. suis, yefM and yoeB are actively co-transcribed, thus forming a bicistronic operon.

Construction of a selective expression vector to characterize the toxin-antitoxin systems in E. coli.
To characterize the toxin-antitoxin systems in E. coli, a selective expression vector was constructed as previously described 40 . A DNA fragment containing the araC gene and the promoter P BAD was amplified from the pBADhisA plasmid, digested with the Xho I and Hind III enzymes, and then cloned into pET-30a, an expression vector induced by isopropyl β -D-thiogalactopyranoside (IPTG), to generate the selective expression vector, designated pETBAD (see Supplementary Fig. S2 online). Plasmid pET-BAD has five unique restriction sites for cloning and possesses the IPTG-inducible promoter P lac and the arabinose-inducible promoter P BAD , thus expression can be induced by IPTG and/or arabinose (see Supplementary Fig. S2 online).

Overproduction of YoeB inhibits cell growth in E. coli which can be alleviated by YefM.
To determine whether the yefM-yoeB locus is indeed an active TA system, the yefM and yoeB genes were cloned separately as well as together into the pBADhisA expression vector. The plasmids were introduced into E. coli Top10 cells, and the transformants were selected in LB agar plates with 0.2% D-glucose (repressed conditions of P BAD ). E. coli Top10 cells harbouring the corresponding plasmids were grown in LB medium, and 0.2% D-glucose or L-arabinose was added at time zero. In the presence of 0.2% D-glucose, Top10 cells harbouring the pBADhisA-yefM and pBADhisA plasmids showed no major difference in growth, while cells carrying the pBADhisA-yoeB plasmid showed a moderate growth defect (Fig. 3a). In the case of inductive conditions (0.2% L-arabinose), E. coli Top10 cells harbouring the pBADhisA-yoeB plasmid exhibited drastic growth inhibition, while cells harbouring other two plasmids showed only moderate reductions in their OD 600 value (Fig. 3b). Surprisingly, under both repressed and inductive conditions, E. coli Top10 cells harbouring the pBADhisA-yefM-yoeB plasmid exhibited obvious growth inhibition (Fig. 3), Even so, Top10 cells carrying the pBADhisA-yefM-yoeB plasmid showed much better growth than that carrying pBADhisA-yoeB (Fig. 3b), indicating that YoeB-induced growth inhibition could be alleviated by YefM.
We further investigated the toxic and antitoxic effect of the TA components using the selective expression system constructed here. In the selective expression plasmid, the IPTG-inducible promoter P lac and the arabinose-inducible promoter P BAD control the expression of YefM and YoeB, respectively (Fig. 4a). Thus, E. coli BL21 (DE3) cells harbouring the pETBAD-yefM Ssu -yoeB plasmid could express the S. suis YefM and/or YoeB upon induction with IPTG and/or L-arabinose. As shown in Fig. 4b, the E. coli BL21 (DE3) cells exhibited considerable growth inhibition in the presence of L-arabinose. In contrast, only moderate growth inhibition was observed in the presence of IPTG or IPTG and L-arabinose together.  coli Top10 cells harbouring the plasmid pBADhisA-yefM, pBADhisA-yoeB, pBADhisA-yefM-yoeB and pBADhisA were diluted 1:1000 in LB-ampicillin. Each culture was then divided into two equal volumes. The first half served as the control, to which 0.2% D-glucose was added (a), 0.2% L-arabinose was added to the second half to induce expression of the target gene (b). Culture growth was evaluated by measuring the OD 600 every hour. The data shown are averages with standard deviations for the results from three independent experiments. These results indicated that the protein encoded by the yoeB gene is a toxin against E. coli and that the protein encoded by the yefM gene could counteract the toxicity. Therefore, the S. suis yefM-yoeB locus works as a typical TA system. YoeB Ssu -induced growth inhibition in E. coli that could be alleviated by YefM Spn , but not by YefM Eco . In S. pneumoniae and E. coli, the toxicity of YoeB could be counteracted only by its cognate antitoxin 28 . To test whether there was cross-complementation between non-cognate YefM and the S. suis YoeB, E. coli BL21 (DE3) cells were transformed with plasmid pETBAD-yefM Spn -yoeB and pETBAD-yefM Eco -yoeB (Fig. 4a). As seen in Fig. 4c, induction of YoeB Ssu resulted in a drastic reduction in OD 600 value, whereas coinduction of YefM Spn and YoeB Ssu alleviated the growth inhibition in E. coli, indicating that the toxic effect of YoeB Ssu was counteracted by coexpression of YefM Spn . However, coinduction of YefM Eco and YoeB Ssu did not neutralize the YoeB Ssu toxicity (Fig. 4d). In addition, induction of YefM Ssu , YefM Spn and YefM Eco also displayed an effect on growth inhibition in E. coli ( Fig. 4b-d).
Introduction of the S. suis yefM-yoeB system into E. coli could affect cell growth. It seemed likely that introduction of the S. suis yefM-yoeB system into E. coli affects cell growth, since E. coli Top 10 cells carrying the pBADhisA-yefM-yoeB plasmid showed considerable growth inhibition under both repressed and inductive conditions. To test the hypothesis, the yefM and yoeB genes were cloned together into the pET-30a expression plasmid. When introduction of the pET30a-yefM-yoeB and pET-30a plasmids into E. coli Trans5α and Top10 strains, cells carrying pET30a-yefM-yoeB showed an obvious growth defect compared with cells carrying pET-30a (Fig. 5a). However, when introduction of the two plasmids into E. coli BL21(DE3) strain, no major difference in growth was found (Fig. 5a).
The same experiments were carried out with the pSET2-yefM-yoeB plasmid. In this plasmid, the S. suis yefM-yoeB locus is under the control of its own promoter. As shown in Fig. 5b, E. coli Trans5α cells harbouring the pSET2-yefM-yoeB plasmid exhibited a remarkable growth defect compared with that harbouring the empty plasmid. The growth inhibition effect was even more severe when the plasmid was transformed into Top10 strain, as cells carrying the pSET2-yefM-yoeB plasmid showed growth arrest over a period of 12 hours (Fig. 5b). However, BL21(DE3) strain harbouring pSET2-yefM-yoeB showed only a slight defect in growth.
Taken together, the results clearly demonstrated that introduction of the S. suis yefM-yoeB system into E. coli could affect cell growth.

Construction and microbiological characterization of the ΔyefM-yoeB mutant.
To investigate the functions of the yefM-yoeB locus in S. suis 2, an isogenic yefM-yoeB knockout mutant of S. suis 2 strain SC19, termed Δ yefM-yoeB, was constructed through homologous recombination (Fig. 6a). To rule out the possible polar effect and introduction of a second mutation during the construction of Δ yefM-yoeB, we generated a complementation strain, designated CΔ yefM-yoeB using the E. coli-S. suis shuttle vector pSET2 41 . The resulting mutant and complementation strains were confirmed by PCR (Fig. 6b), RT-PCR (Fig. 6c) and direct DNA sequencing (data not shown).
The effects of yefM-yoeB deletion on the basic microbiological properties of S. suis 2 were investigated in terms of morphology, haemolytic activity and in vitro growth. The cell morphologies of the Δ yefM-yoeB mutant, WT and CΔ yefM-yoeB strains were examined under light microscope using Gram staining. However, no obvious differences were found (see Supplementary Fig. S3a online). When inoculated on sheep blood agar plates, the three strains showed similar haemolytic activity (see Supplementary  Fig. S3b online). The growth kinetics of the mutant strain were compared with those of the WT and complementation strains by measuring the optical density at 600 nm (OD 600 ) every hour. We found that the growth kinetics of Δ yefM-yoeB were essentially identical to those of the WT and CΔ yefM-yoeB strains (Fig. 7), indicating that the yefM-yoeB locus of S. suis 2 plays no role in growth in vitro.
Deletion of the yefM-yoeB locus has no effect on S. suis 2 virulence in mice. To assess the role of the yefM-yoeB locus in the pathogenesis of S. suis 2, we performed an experimental infection model in CD1 mice. As an initial comparison of virulence, groups of ten CD1 mice were inoculated intraperitoneally with 6 × 10 8 CFU of the WT, Δ yefM-yoeB, CΔ yefM-yoeB strains or PBS. Most mice infected with S. suis strains developed typical clinical signs of S. suis 2 infection, including depression, lethargy, weakness, prostration and rough coat hair during the first 72 h post infection. Ultimately, the survival rates of mice in the WT, Δ yefM-yoeB and CΔ yefM-yoeB groups were 50%, 60% and 30%, respectively (Fig. 8a). By contrast, all mice inoculated with PBS remained healthy and survived. No significant difference in survival rates was observed between the Δ yefM-yoeB group and the WT group (P = 0.6793),  A competitive-infection assay was adopted to further compare the abilities of the WT strain and the Δ yefM-yoeB mutant to establish infection. Four mice were inoculated intraperitoneally with a 1:1 mixture of the WT and mutant bacteria. Mice were sacrificed to collect blood, brain, heart, liver, spleen, lung and kidney samples 24 h after inoculation. Bacterial cells recovered from various tissue samples were analysed by colony PCR to determine the competitive index (CI). The result showed that for each tissue, the mean CI values were approximately 1 (Fig. 8b), suggesting that the mutant and WT strains have similar abilities to colonize the tissues.
Taken together, these results indicated that the yefM-yoeB locus is not involved in the virulence of S. suis 2.

Discussion
TA systems have attracted an increasing concern in recent years because of their abundance in the genomes of bacteria and archaea on the one hand, and the limited of knowledge of their physiological functions on the other. In E. coli K12, at least 33 TA systems have been identified, with several being well characterized 1 . However, only one TA module, SezAT, has been described in S. suis, yet its function has not been demonstrated 24 . The yefM-yoeB module is one of the best studied TA systems and has been described in various bacteria, including E. coli 27,37 , S. pneumoniae 28,29 , M. tuberculosis 30   Mice inoculated with PBS served as the control. Survival data were analysed with the log-rank test. No significant difference was observed between the Δ yefM-yoeB group and the WT group or the CΔ yefM-yoeB group. (b) Competitive index of Δ yefM-yoeB against the WT strain. Four female CD1 mice were inoculated intraperitoneally with a mixture of Δ yefM-yoeB and WT at a ratio of 1:1. At 24 h post-infection, blood, brain, heart, liver, spleen, lung and kidney samples from the mice were collected and plated. The Δ yefM-yoeB:WT ratio in these samples was determined by analysing 70-90 colonies by colony PCR. The competitive index (CI) was determined as the mutant:WT ratio in the samples divided by the ratio in the inoculum. A CI value of 1 indicates equal competitiveness. Mean CI values from four mice were compared to 1 using the two-tailed paired t test to determine whether the difference in competitiveness was significant. No statistically significant difference was found. In this study, we showed that the chromosomally encoded yefM-yoeB locus of S. suis is an active TA system with yoeB encoding the toxin and yefM encoding the cognate antitoxin. This is not surprising, as this system shows considerable similarity to the YefM-YoeB system from S. pneumoniae and E. coli. Like most TA systems, the yefM and yoeB genes are co-transcribed. Upstream of the yefM gene, there is an intergenic region of 75 nucleotides, which may act as the promoter region. Overproduction of the YoeB toxin in E. coli Top10 and BL21 (DE3) cells both resulted in toxic effects commonly linked to toxin activity. YoeB homologs were identified as endoribonucleases that inhibit translation by cleaving mRNA, either in a ribosome-dependent or -independent manner 32,35,43 . We therefore reasoned that S. suis YoeB inhibits cell growth via a similar mechanism.
The toxic effect of S. suis YoeB toxin could be neutralized by both the cognate YefM Ssu and the heterologous YefM Spn , but not by the E. coli counterpart, consistent with the fact that YefM Ssu shares higher levels of identity with YefM Spn (79% identity versus 29% for YefM Eco ). As YefM Spn and YefM Ssu display a high level of sequence homology, we speculated that YefM Spn could bind to and neutralize the YoeB Ssu . The lack of cross-complementation between YefM Eco and YoeB Ssu suggested that there is no favourable interaction between the two heterologous proteins. Similar behaviour has also been reported for YoeB Eco , whose activity could be alleviated by both the cognate YefM Eco and the antitoxin Axe of E. faecium, but not by YefM Spn 27,28 . A previous study showed that overexpression of YefM Eco displayed toxicity in E. coli at high expression levels 44 , consistent with our observation that YefM Ssu , YefM Spn and YefM Eco had an effect on growth inhibition in E. coli.
An interesting observation was that E. coli Top10 cells harbouring the pBADhisA-yefM-yoeB plasmid showed an obvious growth defect under both repressed and inductive conditions. Similar experiments were then performed with other plasmids and strains. Plasmids containing the S. suis yefM-yoeB system were introduced into different strains of E. coli. Except for BL21(DE3) strain carrying the pET30a-yefM-yoeB plasmid, all tested strains harbouring the plasmids containing the S. suis yefM-yoeB system showed growth inhibition. A previous study revealed that the YefM-YoeB complex forms a 2:1 heterotrimer 35 . We speculated that the yefM-yoeB system could be expressed even under repressed conditions and that YefM expression was not enough to counteract YoeB, thus leading to growth inhibition in E. coli. Since pET-30a and E. coli BL21(DE3) strain constitute a precise inducible expression system, it is possible that the yefM-yoeB locus on pET-30a was not expressed without inducer, therefore growth inhibition was not observed for BL21(DE3) carrying pET30a-yefM-yoeB under normal growth conditions. The speculation agrees with the result that BL21(DE3) carrying the pET30a-yefM-yoeB plasmid exhibited a growth defect under inductive conditions.
To investigate the functions of the yefM-yoeB locus in S. suis 2, a knockout mutant and the corresponding complementation strain were constructed. No obvious differences between the WT and the yefM-yoeB deletion mutant were found in terms of their cell morphology, haemolytic activity on blood agar plates, and in vitro growth. The potential role of TA systems in bacterial pathogenesis has been neglected for a long time. More and more studies have revealed that TA systems are involved in bacterial pathogenicity and host-pathogen interactions 17,18 ; therefore, we evaluated the effect of this TA system on the pathogenesis of S. suis 2 using a murine infection model. Survival curves of mice and the competitive-infection assay both demonstrated that deletion of the yefM-yoeB locus had no role in the pathogenicity of S. suis 2. It was reported that Yersinia pestis lacking the hicB3 antitoxin is virulence-attenuated; however, the mutant lacking the whole hicA3B3 locus is fully virulent 45 . In contrast, the toxins ChpK and MazF but not the antitoxins ChpI and MazE are involved in the virulence of L. interrogans during infection 18 . Future experiments should evaluate the involvement of the individual genes of the yefM-yoeB locus in the virulence of S. suis 2.
It should be noted that TA systems play important roles in the physiology of cells, including biofilm formation and multidrug resistance 25 . The effect of the yefM-yoeB module on biofilm formation, stress tolerance and formation of persister cells will be explored in future studies. It is proposed that TA systems are potential targets for antibiotics 46 . Given the fact that S. suis YoeB can inhibit the growth of E. coli considerably, we are planning to examine the effect of YoeB on S. suis. If a similar toxic effect is observed, a multivalent strategy to synthesize an inhibitor that interacts with the YefM antitoxin and frees the toxin YoeB to inhibit bacterial growth could be promising for the development of new antibiotics.
In conclusion, the yefM-yoeB locus was identified as a new TA system of S. suis. The present study clearly demonstrated that the yoeB gene encodes a toxin that can inhibit the growth of E. coli. Specifically, the toxicity of S. suis YoeB could be alleviated by the cognate S. suis YefM and heterologous S. pneumoniae YefM. More importantly, we reported that introduction of the yefM-yoeB TA system into E. coli could affect cell growth. In addition, deletion of the yefM-yoeB locus had no effect on the virulence of S. suis 2.

Methods
Bacterial strains and growth conditions. Bacterial strains and plasmids used in this study are listed in Supplementary Table S1. S. suis strains were maintained on Tryptic Soy Broth (TSB) or Tryptic Soy Agar (TSA; Difco Laboratories, Detroit, MI, USA) with 10% (vol/vol) newborn bovine serum at 37 °C, unless otherwise specified. E. coli strains were cultured in Luria-Bertani (LB) broth or on LB agar at 37 °C. When necessary, antibiotics (purchased from Sigma) were added at the following concentrations: for E. coli, ampicillin, 75 μ g/ml; kanamycin, 25 μ g/ml and spectinomycin, 50 μ g/ml; for S. suis, spectinomycin, 100 μ g/ml.

RNA isolation and RT-PCR analysis.
Total RNA samples were prepared from S. suis cultures using an SV total RNA isolation system (Promega), according to the manufacturer's protocol. RNA concentrations and integrity were determined by UV spectrophotometry and agarose gel electrophoresis, respectively. RT-PCR was carried out using a QuantiTect Reverse Transcription Kit (Qiagen), according to the manufacturer's instructions. For the co-transcription assay, the gene specific primers A1, A2, T1 and T2 were used for RT-PCR analysis (see Supplementary Table S2 online). To identify the mutant and complementation strains, primer pair ATin1/ATin2 was used.
Plasmid Construction. Plasmids were constructed as follows using the primers listed in Supplementary   Table S2. 1. pETBAD. Primer pair BAD1/BAD2 amplified the DNA fragment containing the araC gene and the promoter P BAD from the pBADhisA plasmid. The DNA fragment was digested with the Xho I and Hind III enzymes, and cloned into pET-30a, to generate the selective expression plasmid pETBAD. 2. pBADhisA-yefM and pBADhisA-yoeB. The yefM gene was amplified from the S. suis 2 genome using primer pair yefM1/yefM2. The PCR product was digested with the Xho I and Hind III enzymes, and then cloned into pBADhisA, to generate plasmid pBADhisA-yefM. Plasmid pBADhisA-yoeB was constructed in a similar manner. 3. pBADhisA-yefM-yoeB and pET30a-yefM-yoeB. The yefM and yoeB genes were amplified from the S. suis 2 genome using primer pairs yefM1/R1 and R2/yoeB2, respectively. The two DNA fragments were fused into one fragment using overlap extension PCR. This DNA fragment was digested with the Xho I and Hind III enzymes, and then ligated into pBADhisA, to generate pBADhisA-yefM-yoeB. Plasmid pET30a-yefM-yoeB was constructed in a similar manner, except that the yefM gene was amplified using primer pair yefM3/R1 and the fused DNA fragment was digested with the BamH I and Hind III enzymes. 4. pETBAD-yefM Ssu -yoeB, pETBAD-yefM Spn -yoeB and pETBAD-yefM Eco -yoeB. The yefM Ssu gene was amplified from the S. suis 2 genome using primer pair SsA1/SsA2. The DNA fragment was digested with the Kpn I and EcoR I enzymes, and then cloned into plasmid pETBAD to yield pETBAD-yefM.
The yoeB gene was amplified from the S. suis 2 genome using primer pair SsT1/SsT2. After digestion with the Hind III and Sac I enzymes, the fragment was cloned into pETBAD-yefM, to generate pETBAD-yefM-yoeB. This construct placed the yefM Ssu and yoeB genes under the control of the IPTG-inducible promoter P lac and the arabinose-inducible promoter P BAD , respectively. Thus, IPTG could induce the expression of YefM and arabinose could induce YoeB. The other two plasmids were constructed using the same procedure, except that the yefM Spn and yefM Eco genes were amplified from the S. pneumoniae R6 and E. coli K12 genomes, respectively. 5. pSET2-yefM-yoeB. A DNA fragment containing the yefM-yoeB locus and its predicted promoter was amplified from the S. suis 2 genome using primer pair CAT1/CAT2. After digestion with the Pst I and EcoR I enzymes, the fragment was cloned into pSET2, to generate the plasmid pSET2-yefM-yoeB. 6. pSET4s-Δ yefM-yoeB. Two flanking fragments (LA and RA) of the yefM-yoeB locus were amplified from the S. suis 2 genome using primer pairs LA1/LA2 and RA1/RA2, respectively. After digestion with the appropriate restriction enzymes, the two fragments were simultaneously cloned into pSET4s to generate the knockout plasmid pSET4s-Δ yefM-yoeB.
E. coli growth analysis. E. coli Top10 cells transformed with pBADhisA-yefM, pBADhisA-yoeB, pBADhisA-yefM-yoeB and pBADhisA were cultured overnight in LB broth supplemented with 75 μ g/ mL ampicillin and 0.2% D-glucose. The next day, the four cultures were diluted 1:1000 in LB-ampicillin. Each culture was then divided into two equal volumes. The first half served as the control, to which 0.2% D-glucose was added, while 0.2% L-arabinose was added to the second half to induce expression of the target gene. Culture growth was evaluated by measuring the OD 600 every hour. E. coli BL21(DE3) cells harbouring the respective selective expression plasmids were incubated in LB broth supplemented with 25 μ g/ml kanamycin to an OD 600 of about 0.3. Each culture was divided into four equal parts, to three of which was individually added 0.2% L-arabinose, 1 mM IPTG or both, respectively. The fourth part had nothing added to it and served as a control. These cultures were further incubated and samples were taken every hour to determine the OD 600 .
E. coli strains were transformed with plasmids containing the S. suis yefM-yoeB system or the corresponding empty plasmids. Cells were cultured overnight in LB broth supplemented with antibiotics and diluted 1:1000 in fresh medium. Culture growth was monitored by measuring the OD 600 . For those cells showing growth arrest in LB broth, such as Top10 carrying pSET2-yefM-yoeB, isolated colonies were used as inocula. In parallel, colonies of the same size for strains carrying the corresponding empty plasmids were used. Deletion of the yefM-yoeB locus and functional complementation. Gene knockout mutant of the yefM-yoeB locus was constructed using plasmid pSET4s as described previously 47,48 . The knockout plasmid pSET4s-Δ yefM-yoeB was introduced into the competent cells of S. suis SC19 by electroporation. After two steps of allelic exchange at 28 °C, spectinomycin-sensitive clones were selected to identify the mutants by PCR using primers listed in Supplementary Table S2. The mutants were further confirmed by RT-PCR analysis and direct DNA sequencing.
For complementation assays, the recombinant plasmid pSET2-yefM-yoeB was introduced into the Δ yefM-yoeB mutant by electroporation. The complementation strain CΔ yefM-yoeB was selected with spectinomycin and confirmed using PCR, RT-PCR and DNA sequencing.

Mouse infections. The Laboratory Animal Monitoring Committee of Huazhong Agricultural
University approved all the animal experiments, which were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of Hubei Province, China. Forty female CD1 mice (5-weeks-old) were randomly divided into four groups with 10 mice per group. Mice in Groups 1, 2 and 3 were inoculated intraperitoneally with 6 × 10 8 CFU in 200 μ L PBS of the WT, Δ yefM-yoeB and CΔ yefM-yoeB strains, respectively. Group 4 was injected with 200 μ L PBS, and served as the control group. Mice were monitored daily over 14 days for clinical signs and survival rates.
For the competitive-infection assay, four mice were inoculated intraperitoneally with a mixture of the WT and mutant strains at a ratio of 1:1 (1 × 10 8 CFU). The actual ratio in the inoculum was determined by plating the suspension of each strain before mixing. Mice were sacrificed 24 h after inoculation and brain, heart, liver, spleen, lung and kidney samples were collected, homogenized and diluted for plating. Blood samples were directly diluted for plating. The Δ yefM-yoeB:WT ratios in these samples were determined by analysing 70-90 colonies using colony PCR with primer pair ATout1/ATout2, which yielded PCR products of 597 bp and 1117 bp for the Δ yefM-yoeB and WT strains, respectively. The competitive index (CI) was calculated as the Δ yefM-yoeB:WT ratio in each sample divided by the ratio in the inoculum.
Statistical analyses were carried out using GraphPad Prism 5 (San Diego, USA). Log-rank test was used to analyse the mice survival curves. Two-tailed paired t test was used to analyse the data in the competitive-infection assay. A P value of < 0.05 was considered statistically significant.