SigB Is a Dominant Regulator of Virulence in Staphylococcus aureus Small-Colony Variants

Staphylococcus aureus small-colony variants (SCVs) are persistent pathogenic bacteria characterized by slow growth and, for many of these strains, an increased ability to form biofilms and to persist within host cells. The virulence-associated gene expression profile of SCVs clearly differs from that of prototypical strains and is often influenced by SigB rather than by the agr system. One objective of this work was to confirm the role of SigB in the control of the expression of virulence factors involved in biofilm formation and intracellular persistence of SCVs. This study shows that extracellular proteins are involved in the formation of biofilm by three SCV strains, which, additionally, have a low biofilm-dispersing activity. It was determined that SigB activity modulates biofilm formation by strain SCV CF07-S and is dominant over that of the agr system without being solely responsible for the repression of proteolytic activity. On the other hand, the expression of fnbA and the control of nuclease activity contributed to the SigB-dependent formation of biofilm of this SCV strain. SigB was also required for the replication of CF07-S within epithelial cells and may be involved in the colonization of lungs by SCVs in a mouse infection model. This study methodically investigated SigB activity and associated mechanisms in the various aspects of SCV pathogenesis. Results confirm that SigB activity importantly influences the production of virulence factors, biofilm formation and intracellular persistence for some clinical SCV strains.


Introduction
Staphylococcus aureus is a widespread Gram positive pathogen that causes serious problems in health care settings and in the community because of antibiotic resistance and high morbidity and mortality rates [1,2]. This bacterium has the ability to cause both acute and chronic infections in several organs and expresses numerous virulence factors that are involved in a wide range of pathogenic processes [3]. It is thought that particular sets of virulence factors are required for specific steps during disease progression or for infection of specific organs.
The expression of S. aureus virulence factors is controlled by complex regulatory networks and as a function of the bacterial population density and environment. During in vitro growth, the transition from the expression of adhesins and other cell-surface proteins to that of exoproteins (hemolysins, proteases and nucleases) involves the activation of the quorum-sensing agr system that influences virulence gene expression and mainly depends on a regulatory RNA, RNAIII [4,5]. Although the functionality of the agr system has been reported to be important for the pathogenesis of S. aureus in several experimental infection models, it is possible that low agr activity could be advantageous during specific diseases [4] such as chronic pulmonary infections in CF patients [6].
Several S. aureus components are involved in the regulation of virulence factors including two-component signal transduction systems and regulatory factors, which allow the bacterium to adjust its genetic expression program as a function of its surrounding [3]. The alternative transcription factor sigma B (SigB) is known to affect the expression of several genes encoding virulence factors and stress-response proteins, and seems to counterbalance the influence of the agr system on the expression of virulence factors. The activity of SigB peaks early during the stationary phase of growth [7] and some genes (e.g. asp23) are expressed as a direct function of SigB activity. SigB is also known to support the expression of some genes in early exponential growth (e.g. fnbA) while having a negative effect on the expression of other genes encoding exoproteins (e.g. hla, sspA, nuc1) in late exponential growth [8]. Importantly, many genes regulated by SigB are not preceded by a SigB promoter and the effect of SigB on the expression of those genes is likely to be mediated by downstream regulators [8]. The agr system and SigB are known to be interconnected [9].
Biofilms are microbial communities embedded in a matrix of extracellular polymeric substances (EPS) that can adhere to either biological or non-biological substrates. Importantly, biofilms are deemed to be involved in persistent infections. The formation of a biofilm is initiated by the adhesion of bacteria to a substrate and involves a maturation phase characterized by intercellular aggregation and the acquisition of an architecturally complex tridimensional structure. Several mechanisms of biofilm formation exist in S. aureus [10]. While some staphylococcal strains use an icadependent polysaccharide matrix, the ability to form biofilms seems to involve ica-independent mechanisms in several other strains [11][12][13]. In many cases, intercellular aggregation has been shown to be mediated by cell-surface proteins such as protein A, biofilm-associated protein (Bap) or FnBPs [10]. Extracellular DNA (eDNA) also seems to be an important EPS in many bacterial biofilms and S. aureus possesses mechanisms to incorporate DNA into biofilm matrixes [14][15][16]. On the other hand, the dispersal of biofilms involves mechanisms that allow the detachment of bacterial cells or biofilm fragments and the dissemination of the pathogen in the environment or the host [10,17]. The agr system and SigB are among the numerous regulators of virulence known to modulate biofilm formation and dispersion by controlling the expression of both adhesion and dispersion factors (e.g. exoproteases and nucleases) [10].
S. aureus is now considered a facultative intracellular pathogen. Although the precise role of intracellular infections is still an object of debate for staphylococci, it has been speculated that the intracellular persistence of bacteria may confer protection against the host immune system and extracellular antibiotics [18,19]. The best characterized mechanism of cell invasion used by S. aureus involves the formation of a fibronectin bridge between bacterial FnBPs and the host a 5 b 1 integrin, but FnBPs-independent mechanisms also exist [19]. Once S. aureus is within a host cell, several scenarios were reported to occur and it is likely that a wide range of different outcomes are possible according to bacterial strains and cell types [19]. Nevertheless, it has been demonstrated that the agr system is required to escape a phagolysosome [20,21] and that both the agr system and SigB influence the induction of host cell death [22,23].
Small-colony variants (SCVs) constitute a subpopulation of oxidative phosphorylation-deficient bacteria that grow slowly and differ from prototypical strains in various aspects. SCVs usually are less pigmented, less susceptible to aminoglycoside antibiotics, have altered biochemical properties and low hemolytic activity [18,24]. Importantly, SCVs are usually recovered from several different types of persistent infections [24] such as those found in the lungs of cystic fibrosis (CF) patients [24,25]. In experimental infection models, SCVs have the ability to cause infections and to persist although they appear less virulent [24]. The appearance and role of SCVs during infections are still unclear but some results suggest that this phenotype is more difficult to eradicate with antibiotic treatment in vivo [26,27]. An outstanding study has recently demonstrated that SCVs emerge during long-term infections and are associated with a weaker immune response [28]. The contribution of specific virulence factors in the establishment of SCV infections remains an incompletely investigated research field.
The production of virulence factors is altered in SCVs [25,29,30]. The expression profile of virulence genes in some SCVs has been attributed to an atypical activity of global regulators where SigB activity is constitutive and that of the agr system is low [25,29]. More precisely, SigB and agr activities seem to explain the constitutive expression of some cell-surface proteins and the down-regulation of several exoproteins in those SCVs [31,32]. However, the relative roles of SigB and agr that may explain some of the characteristics of SCVs are still speculative and are only based on interpretations of gene expression profiles. Interestingly, some SCV strains produce relatively higher amounts of biofilms than prototypical strains [32,33] and persist more efficiently inside non-phagocytic host cells [34], which both, alone or together, may explain the association of these bacteria with chronic infections. Despite that, molecular mechanisms involved in the ability of SCVs to form biofilms and to persist within host cells have not been thoroughly investigated although it was shown that SigB may contribute to these phenotypes in SCVs [25,32]. The aim of this study was to investigate the role of SigB-dependent mechanisms in the various aspects of the pathogenesis of SCVs. Our results underline the major role played by SigB in the pathogenesis of SCVs through its influence on virulence factors production, biofilm formation and intracellular persistence.

Formation of biofilms by SCVs may involve extracellular proteins
Mechanisms involved in biofilm production by SCVs were first investigated in presence of glucose and NaCl and by comparing biofilm formation for the genetically-related CF07-L (prototypical) and CF07-S (SCV) strains co-isolated from a CF patient. Glucose and NaCl are known to stimulate the production of protein-and polysaccharide-dependent biofilms in S. aureus, respectively [11]. Figure 1A and 1B show that the SCV CF07-S strain formed more biofilm than the normal CF07-L strain in the presence of increasing amounts of glucose (as shown previously [32,33]), but not in the presence of NaCl. Detachment assays using proteinase K confirmed that extracellular proteins are involved in biofilms produced by three different SCV isolates from CF ( Figure 1C). SCVs CF03-S and CF1D-S were already shown to produce greater amounts of biofilm than their normal co-isolated counterparts [33] and were here used to confirm results obtained with strain CF07-S. In order to support the hypothesis that polysaccharides are not involved in the greater biofilm formation ability of CF07-S, qPCR was performed to compare the expression of icaC in CF07-L and CF07-S from planktonic cultures and no statistically significant difference was accordingly observed ( Figure 1D). Menadione was also used to supplement and restore normal growth levels for CF07-S during experiments measuring the expression of icaC at different phases of growth. It was previously shown that menadione abolished the production of biofilm by CF07-S [32]. Figure 1D confirmed that restoring the oxidative phosphorylation and the metabolic status of CF07-S by menadione supplementation does not alter the expression of icaC (P.0.05). This section supports the hypothesis that extracellular proteins are involved in the formation of biofilms by the three SCV strains tested.

SCVs may have low biofilm-dispersing proteolytic activity
It is known that proteinaceous biofilms are dispersed by mechanisms involving extracellular proteases [17]. Experiments were thus performed to evaluate the proteolytic activity of SCVs in comparison to their normal counterparts and demonstrated that SCVs have no or low proteolytic activity (Figure 2A). Two types of plates supplemented with non-fat dry milk (BHI glucose and MH agar) were used to confirm that results are reproducible in different media. It is noteworthy that the low (or null) proteolytic activity of SCVs was still observed on either media even after 48 hours of incubation ( Figure 2A and data not shown). Also, the presence of the serine protease inhibitor PMSF increased biofilm production for strain CF07-L, but not for the already high producer CF07-S ( Figure 2B), thus confirming the association between proteolytic activity and biofilm dispersion in the CF07 strain background. Accordingly, the expression of the V8 serine protease gene (sspA) was down-regulated in CF07-S in comparison to the prototypical CF07-L strain in stationary phases of growth ( Figure 2C). No statistically significant difference was detected between CF07-L and CF07-S in the expression of the protease gene aur ( Figure S1). Moreover, the transcripts for both splA and splC were not detected in the CF07 background (data not shown), which suggests that serine protease-like proteins from the spl operon are not contributing to biofilm dispersion in these strains. These results suggest that the proteolytic dispersion of biofilm is not an active process in these SCVs.

SigB modulates biofilm formation in SCV CF07-S but is not solely responsible for the repression of proteolytic activity
We previously showed that SigB affects the biofilm formation of CF07-S [32]. The activity of SigB was monitored using qPCR targeting asp23, a known marker of SigB activity [35]. Figure 3A confirmed that the expression of asp23 is higher in SCV CF07-S in comparison to the prototypical strain CF07-L in the early exponential growth (P,0.05, ANOVA with Dunnett's posttest). This confirms that a high SigB activity is seen throughout the growth phases in SCV CF07-S because of its metabolic status (i.e. electron transport deficiency), whereas the expression of the SigB marker asp23 only increases toward the stationary phase in normal strains. Noteworthy, the SCV strains CF03-S and CF1D-S are also characterized by a higher SigB activity in comparison to normalgrowing S. aureus bacteria [25] (data not shown). Figure 3B confirms that the formation of biofilm by CF07-S is indeed regulated by SigB by using a sigB mutant strain carrying an empty vector (pFM1) or a sigB expression vector (pFM2). It has been demonstrated that SigB modulates the dispersion of proteinaceous biofilm [36,37] as well as the expression of exoproteases such as Aur and SspA in normal strains [36]. In contrast to that, Figure 3C shows that the proteolytic activity of SCV CF07-SDsigB remains low and is not altered in comparison to SCV CF07-S. However, the expression of sspA in CF07-SDsigB was increased in comparison to that of CF07-S in the stationary phase of growth ( Figure  S2). In order to confirm that the activity of serine-proteases (such as SspA) was not involved in the dispersal of biofilm from CF07-SDsigB, we performed biofilm formation with CF07-SDsigB exposed to PMSF and did not observe any statistically differences in comparison to the untreated control ( Figure S3). Our results thus support the idea that serine-proteases are not involved in the dispersion of the biofilm forms by CF07-SDsigB, although a higher transcription of the sspA gene can be detected in this strain in comparison to CF07-S. Interestingly, the proteolytic activity was modulated in a SigB-dependent manner when CF07-S was Figure 1. Formation of biofilms by SCVs may involve extracellular proteins. The extent of biofilm production was measured by crystal violet staining (A 560 nm ) of normal CF07-L and SCV CF07-S strains as a function of glucose (A) and NaCl (B) concentrations following a 48h incubation period. Statistically significant differences between strains are indicated for each concentration of glucose and NaCl (P,0.05; twoway ANOVA with Bonferonni's posttest, n = 3). (C) Susceptibility of SCVs' biofilms to treatment with proteinase K. Statistically significant differences between control and treated conditions are indicated (*, P,0.05; **, P,0.01; unpaired t test, n = 3). Results are normalized according to the control condition for each strain. (D) Expression ratio of the icaC gene as a function of growth for strains CF07-L, CF07-S and CF07-S in the presence of menadione, which restores normal growth. Results are expressed according to CF07-L in the early exponential phase of growth. No statistically significant difference was revealed between conditions for each growth phase (ANOVA with Dunnett's posttest, n = 3-4). Results are expressed as means with standard deviations. doi:10.1371/journal.pone.0065018.g001 exposed to menadione ( Figure S4). Overall, this section suggests that the regulation of biofilm formation by SigB does not involve proteolytic mechanisms in the SCV CF07-S.
The regulation of the agr system by SigB influences the hemolytic activity and the formation of biofilm in SCV CF07-S The activity of the agr system is influenced by SigB [9,32] and is repressed in SCVs [25,29]. Figure 4A confirms that the expression of RNAIII (the effector of the agr system) is repressed by SigB in the SCV CF07-S. Accordingly, the inactivation of sigB strongly upregulated the a-hemolysin (hla) gene ( Figure 4B), which expression is already known to be influenced by agr activity [3]. Moreover, the inactivation of sigB in this SCV strain triggers an increase in hemolytic activity ( Figure 4C), which is overall controlled by the agr system, although not solely through Hla expression [38,39]. In order to know whether the control of the agr system by SigB is involved in the formation of biofilm by CF07-S, a RNAIII expression vector was introduced in this strain (pFM4), which was expressing over 10-fold more RNAIII transcript than the empty vector control strain when in presence of the transcriptional inducer (data not shown). Figure 4D shows that the overexpression of RNAIII reduced the formation of biofilm by CF07-S (P,0.05, ANOVA with Tuckey's posttest), but not to a level as low as seen with CF07-L. The level of expression of asp23 was evaluated in CF07-S over-expressing RNAIII and it showed that RNAIII does not modulate SigB activity in this SCV. Accordingly, exposure of a SCV strain to overnight culture supernatants from a normal strain with the same agr type did not influence the expression level of asp23 (data not shown). Overall, this section confirms that the influence of SigB on the agr system is unidirectional and influences both hemolysis and biofilm formation in SCVs.
The SigB-dependent expression of fnbA contributes to biofilm formation in SCV CF07-S It has been shown that FNBPs are involved in the formation of proteinaceous biofilms in normal S. aureus strains [12] and that some SCVs strongly expressed fnbA throughout the growth phases because of a sustained SigB activity [31]. Figure 5A confirms that CF07-S strongly expressed fnbA in a SigB-dependent manner throughout growth in comparison to normal bacteria. The inactivation of fnbA decreased the biofilm formation of CF07-S (P,0.05, ANOVA with Tuckey's posttest), but not to the level of CF07-L, which suggests that some additional fnbA-independent mechanisms may also be involved in the biofilm formation of CF07-S ( Figure 5B). However, expression of fnbA in CF07-SDsigB restored biofilm formation to the level of CF07-S ( Figure 5B). These results suggest that fnbA is importantly involved in the biofilm formation of the SCV strain CF07-S.

Extracellular DNA may be a component of the biofilm formed by SCVs
Extracellular DNA released during autolysis of bacterial subpopulations is now recognized as an important component of biofilm matrices [14,16,17]. We speculated that SigB activity modulates autolysis and the release of extracellular DNA in SCVs. Figure 6A shows that CF07-S bacteria lysed at a rate strikingly higher than CF07-L, but that SigB was not involved in the control of this process. Noteworthy, Triton induced autolysis of strain CF07-L but did not alter that of strains CF07-S and CF07-SDsigB ( Figure S5). The role of extracellular DNA in the formation of biofilm by three SCVs is confirmed on Figure 6B showing the susceptibility of the SCVs' biofilms to treatment with DNase I. Although the release of extracellular DNA by autolysis seemed to be more important in the SCV CF07-S than in its normal counterpart CF07-L and that this DNA release may be involved in the formation of biofilms, this mechanism was not modulated by SigB activity.  The nuclease activity of SCV CF07-S is repressed through a SigB-dependent mechanism Biofilm dispersion was shown to be influenced by a SigBdependent mechanism controlling nuclease activity in normal S. aureus strains [40]. Figure 7A demonstrates that the nuclease activity of each of three SCV strains was lower than that of their normal counterparts. Furthermore, this nuclease activity was controlled by SigB in the SCV strain CF07-S ( Figure 7B). Of the two genes encoding nucleases in the S. aureus genome (nuc1 and nuc2), only nuc1 was shown to be modulated by SigB [40]. Accordingly, Figure 7C and 7D demonstrate that nuc1, but not nuc2, was down-regulated by SigB in the SCV strain CF07-S. This section suggests that the nuclease activity of SCV CF07-S is repressed through the SigB-dependent control of nuc1 expression.
SigB is required for the intracellular replication of SCV CF07-S within epithelial cells The role of SigB in the persistence of SCVs within epithelial cells was next studied because it was previously suggested that SigB may contribute to the infection of host cells by SCVs [25]. Figure 8A shows that SCV CF07-S has the ability to accumulate inside CF-like epithelial cells as previously reported [34]. Interestingly, this ability to replicate over time inside CF-like epithelial cells was not observed for the normal CF07-L strain and was dependent on the activity of SigB. Although SigB provided to SCV CF07-S the ability to accumulate intracellularly over time, SigB was not apparently involved in the internalization process as evaluated by our cell infection protocol ( Figure 8A, 3 h postinvasion time, P.0.05). Initial infection of cells with CF07-SDfnbA was also not different from that of CF07-S although a trend was observed ( Figure 8A, 3 h post-invasion time, P.0.05). Many host cell internalization protocols compare bacteria and strains isolated from the exponential phase of growth. Here, we used SCVs collected from a 20-h old agar plate to perform the cell internalization step, a condition where fnbA should be strongly up-regulated in CF07-S in comparison to CF07-L ( Figure 5A). In such conditions, fnbA-dependent differences in internalization between the strains studied here should have been observed if this gene had been by itself an essential determinant in this process. In fact, we showed that a fnbAB double mutant in strain 8325-4 was drastically less internalized than its parent strain ( Figure S6), which supports the validity of the cell infection protocol used in this study and confirmed the role of at least one or the two FnBPs in internalization of S. aureus in polarized CF-like epithelial cells. Moreover, immunolocalization experiments revealed that some a5b1 integrins were at the apical surface of cells (data not shown). The role of SigB in the ability of SCV CF07-S to accumulate within cells was confirmed in Calu-3 and Calu-3 shCFTR ALTER cells, both of which producing a normal CFTR ( Figure S7). Figure 8B confirms that SigB was involved in the intracellular accumulation of SCV CF07-S 48 h post-invasion and shows that the expression of sigB in CF07-SDsigB complemented the infection. Interestingly, the extent of epithelial cell death 48 h post-invasion did not follow the intracellular accumulation of SCV CF07-S ( Figure 8C). These results suggest that SigB confers on SCV CF07-S the ability to replicate within epithelial cells but that it is not involved in the induction of host cell death.
SigB provides better fitness to SCV CF07-S in the presence of a normal strain in a mouse pulmonary model of infection In order to evaluate if the above-described SigB-dependent phenotypes influence the pathogenesis of SCV infections, a mouse pulmonary infection model was used. We compared pulmonary infections (CFU recovered from the lungs of each mouse) achieved with either the normal strain CF07-L or the SCV strain CF07-S at 48 h post-inoculation in single infections, but also in infections where both normal and SCV strains were combined (i.e. a coinfection). Indeed, SCVs are rarely found alone in clinical samples [24], and a co-infection model may be more representative of the clinical setting. Interestingly, Figure 9A shows that more SCVs are   recovered from lungs during a co-infection with the normal strain CF07-L than during single infections 48 h post-inoculation. This increase in SCV counts was not observed in lungs infected with CF07-L in combination with CF07-SDsigB ( Figure 9B). Importantly, more SCV counts were recovered from lungs of mice infected with the combination of CF07-L and CF07-S in comparison to that observed with the combination of CF07-L and CF07-SDsigB (P,0.5, Mann-Whitney test) ( Figure 9C). This section confirms that SigB may be advantageous to the pathogenesis of SCVs.

Discussion
Although the role of SCVs in chronic infection is still controversial, more and more studies support the clinical association between these variants and persistent infections. It is possible that the SCV phenotype helps colonization of the host by reducing tissue destruction and activation of the host immune system [24,28]. However, the fact that some SCVs can form greater amounts of biofilm, that they are able to persist within nonphagocytic host cells and that they can revert to the normal phenotype is worrisome considering that these phenotypes might be involved in chronic infections [18,41,42]. This study better defines some of the molecular mechanisms underlying the ability of some SCV strains to form biofilms and to persist within nonphagocytic host cells.
One hypothesis to explain the difference in the expression of virulence genes seen between normal and SCV strains is that the slow growth of SCVs does not allow auto-inducing peptides to reach the extracellular concentration required to activate the agr system, which is known to trigger the expression of several virulence factors down-regulated in SCVs [3,4,24]. However, results of this study show that SigB activity is dominant over that of the agr system in the strain SCV CF07-S. Indeed, whereas the deletion of sigB led to overexpression of RNAIII, overexpression of RNAIII did not influence the expression level of asp23 in this strain. Queck et al. [5] demonstrated that, although RNAIII is involved in the regulation of most agr-dependent virulence factors, agr-dependent genes with metabolic-related functions are regulated through an RNAIII-independent pathway. Hence, it would be relevant to evaluate the effect of agrA and agrACDB overexpression in the SCV strain CF07-S in order to better study the role of agr in the pathogenesis of this SCV. The construction of an agr/sigB double mutant in the CF07-S background should also help to investigate the interaction between these two regulators. However, results from this study strongly suggest that SigB activation is upstream to any putative agr role in the SCV CF07-S. Because several genes of the SigB regulon are thought to be indirectly influenced by SigB [8], future studies should also aim at revealing the role of regulators or regulatory networks activated by SigB in the pathogenesis of SCVs. Interestingly, it was previously shown that the expression of sarA is up-regulated through a SigBdependent mechanism in CF07-S [32]. Some cell-surface protein genes are actively expressed by some SCVs (e.g. fnbA, clfA and sceD) whereas several secreted exoprotein genes are repressed [25]. Interestingly and at least for fnbA, this upregulation is apparently constitutive and observed throughout the growth of some SCVs [31]. In addition to transcriptional and translational regulation, it is known that secreted proteases are used by S. aureus to degrade adhesins and to modulate adhesion to host tissues [43]. The low proteolytic activity of SCVs we observed in this study thus supports the hypothesis that the SCV phenotype promotes long-term colonization of host tissues. However, the role of cell-surface proteins (other than FnBPs) in the pathogenesis of SCVs remains to be investigated more thoroughly.
Although the ability of some SCVs to form biofilms is known [32,33,44,45], almost all studies on SCVs focused on their intracellular persistence [18,24]. Nevertheless, the role of biofilms in chronic infections must not be overlooked, especially in the context of pulmonary infections in CF patients [41,42]. Until now, mechanisms involved in biofilm formation by SCVs were incompletely understood, although some results suggested that polysaccharide intercellular adhesin (PIA) was a component of their biofilm matrix [44,45]. However, the roles of SigB, agr and FnBPs in biofilm formation of normal strains [10] rather suggested that extracellular proteins may be involved in biofilm formation by SCVs. In the three SCV strains used in this study, we showed that proteins, and to some extent extracellular DNA, are important for biofilm formation and that, at least for strain CF07-S, overexpression of the ica operon did not seem to explain the greater production of biofilm in comparison to its normal counterpart CF07-L. However, these results would benefit from a more direct characterization of the components constituting the biofilm produced by CF07-S. Interestingly, it was reported that proteinaceous biofilms are usually devoid of PIA [11][12][13].
Biofilm production is lower in a sigB mutant of SCV CF07-S [32]. A role for fnbA in the SigB-dependent biofilm formation of this SCV strain is shown by the present study, but a fnbA mutation did not completely abolish biofilm formation. As autolysis may be influenced by SigB [46,47], and because extracellular DNA has been shown to be involved in the formation of a FnBPs/SigBdependent biofilm [46], it was speculated that a SigB-dependent control of autolysis mediates the release and incorporation of DNA into the biofilm matrix of SCV CF07-S. However, although it was found that the rate of autolysis was drastically higher in the SCV strain CF07-S in comparison to its normal counterpart, SigB did not appear to modulate autolysis in SCVs. Interestingly, it was shown that the proton motive force influences the induction of autolysis [48], which may suggest that the mechanism that controls autolysis in this SCV is upstream to SigB. It was also reported that SigB prevents biofilm dispersion by repressing nuclease activity [40], which is a mechanism we also observed in the SCV CF07-S. Studying the molecular mechanisms controlling extracellular nuclease activity in SCVs should be of interest as SigB was also described to activate nuc1 expression through the SpoVGdependent pathway [49]. Overall, SigB positively controls biofilm formation by at least two different mechanisms in the SCV CF07-S: it supports the expression of fnbA and represses nuclease activity.
According to this study, the SigB-dependent expression of fnbA in SCV CF07-S importantly influences biofilm formation and less so the host cell infection process. However, once the SCV CF07-S is internalized, this study demonstrates that SigB is required for replication and survival within both non-CF and CF-like epithelial cells. It was initially thought that this intracellular accumulation of SCVs was mediated through a SigB-dependent repression of the induction of host cell death, a mechanism that had already been described in other strains [22,23]. Instead, we observed that while the normal strain CF07-L was more cytotoxic than the SCV strain CF07-S, the deletion of sigB did not influence the cytotoxicity of CF07-S in a statistically significant manner. It is thus possible that SigB influences the expression of genes involved in the survival of the SCV CF07-S inside host cells or in the intracellular trafficking of this SCV. The role(s) of the SCV phenotype during S. aureus infections remain to be fully investigated. Moreover, it is not known whether biofilm formation and intracellular persistence are important aspects of the pathogenesis of these strains. The role played by different virulence factors and regulators in vivo are even less well known. Indeed, there is no study addressing the importance of specific virulence factors in the pathogenesis of SCVs using experimental models of infections. This study suggests that the colonization of lungs by the SCV CF07-S is helped by co-infection with a normal S. aureus strain and perhaps explains why SCVs are rarely recovered alone from clinical specimens [24]. Whether or not tissue damage and inflammation induced by these coinfections are different from that of single infections will be important aspects to investigate in future studies. The in vivo dependency of this SCV strain toward the normal phenotype may also explain why the large-scale chromosome flip-flop inversion involved in the reversible switching mechanism of SCV formation recently described for one strain is self-organized to maintain both phenotypes in the bacterial population [50]. This study also supports the hypothesis that SigB is important for the pathogenesis of the SCV CF07-S during lung infection. Additional studies will be necessary to identify the specific SigB-dependent mechanisms involved such as adhesion to host tissues, biofilm formation, intracellular persistence or stress resistance.
Based on the evidence presented here for a central role of SigB in the pathogenesis of SCV CF07-S, therapeutic approaches targeting SigB activity, perhaps in concert with other traditional antibiotics, might offer new alternatives to tackle or prevent chronic infections caused by S. aureus. Recently, a small drug-like SigB inhibitor active against Listeria monocytogenes and Bacillus subtilis was described [51]. Unfortunately, in our hands, we observed no effect for this SigB inhibitor on the expression of asp23 in the SCV strain CF07-S (data not shown). Nonetheless, screening of compound libraries for inhibitors of S. aureus SigB activity might lead to the development of novel therapies interfering with the pathogenesis of SCVs.

Ethics Statement
The animal experiments were conducted following the guidelines of the Canadian Council on Animal Care and the institutional ethics committee on animal experimentation of the Faculté des Sciences of Université de Sherbrooke. The institutional ethics committee on animal experimentation of the Faculté des Sciences of Université de Sherbrooke specifically approved this study (protocol FM2009-01).

Bacterial strains, cell lines and growth conditions
Strains used in this study are listed in Table 1. CF03-L/CF03-S, CF07-L/CF07-S and CF1A-L/CF1D-S are related pairs of strains co-isolated from CF patients, which respectively have a normal (-L, for large colony) and a SCV (-S, for small colony) phenotype. The genetic relatedness of each strain among pairs was confirmed by the analysis of multiple loci with a variable number of tandem repeats (VNTR) [33]. Multilocus sequence typing (MLST) was also previously performed for strains CF07-L and CF07-S [34]. Except where otherwise stated, S. aureus strains were grown in brain heart infusion (BHI) broth (BD, ON, Canada). Whenever required, chloramphenicol (10 mg/ml) (ICN Biomedicals, Irvine, CA), erythromycin (10 mg/ml) (Sigma, Oakville, Ontario, Canada) and tetracycline (10 mg/ml) (Sigma) were added to the media. CdCl 2 (Sigma) was used to induce the expression of genes under the control of the P cad -cadC promoter [52]. CdCl 2 was usually used at a wide range of concentrations for each of the assays performed in this study although only results obtained with the optimal concentration of CdCl 2 are shown in figures.

DNA manipulation and plasmid construction
Recommendations from the manufacturers of kits were followed for chromosomal DNA isolation (Sigma), plasmid DNA isolation (Qiagen, ON, Canada), extraction of DNA fragments from agarose gels (Qiagen) and purification of PCR products and digested DNA fragments (Qiagen). The exception was the use of supplemental lysostaphin (Sigma) at 200 ug/ml to achieve efficient lysis of S. aureus cells. The oligonucleotides (Sigma) used as primers for cloning and plasmid construction are listed in Table 2. Primers were designed to add restriction sites upstream and downstream of the amplified products. PCRs were performed using the Vent DNA polymerase (NEB, Pickering, ON, Canada) or PfuUltra highfidelity DNA polymerase (Agilent Technologies, Mississauga, ON) and cycling times and temperatures were optimized for each primer pair. Plasmid constructs were generated using E. coli DH5a (Invitrogen, Burlington, ON, Canada), restriction enzymes (NEB) and the T4 DNA ligase (NEB). Plasmid constructs were validated by restriction digestion patterns and DNA sequencing before electroporation in S. aureus RN4220 and in final host strains. Plasmids used in this study are listed in Table 1.

Generation of CF07-SDfnbA
An isogenic mutant of the CF07-S strain was constructed, in which the fnbA gene was disrupted by the insertion of the ermA cassette by homologous recombination. The temperature-sensitive pBT2-DfnbA::ermA plasmid was used combined to a strategy optimized for gene disruption in the SCV CF07-S [31,54]. The knockout of fnbA in strain CF07-SDfnbA was confirmed by PCR.

Proteolysis, hemolysis and nuclease activity
Bacterial suspensions (0.5 McFarland standard) prepared for each strain were spotted (2 ml) on BHI agar containing 0.25% glucose or Mueller-Hinton (MH) agar (BD) supplemented with 20 g of non-fat dry milk per liter, MH agar supplement with 5% sheep blood (Oxoid, Nepean, ON, Canada) or DNase test agar (BD). Plates were observed after 24 and 48 h of incubation at 35uC. The three main hemolysins of S. aureus (Hla, Hlb and Hld) are active on sheep blood agar plates [38].

Quantitative real-time PCR (qPCR)
For each qPCR experiment, cultures grown overnight were used to inoculate BHI broth at an A 600 nm of 0.1 (100 ml of culture in a 250-ml erlenmeyer). The bacteria were then grown at 35uC with shaking until they reached a specific growth phase. The A 600 nm values for the early and the late exponential growth phases were, respectively, 0.4 and 3.0-3.5 for the prototype strains and, 0.25 and 0.9-1.0 for the SCV strains [25]. The stationary phase samples were taken after 12 h of growth. RNA extraction, cDNA synthesis and qPCR were performed as previously described [25,31,34] using the primers described in Table 2. The relative expression ratios were calculated by using the cycle threshold (C t ) of the housekeeping gene gyrB (n-fold expression = 2 2DCt , where DC t represents the difference between the C t of the gene studied and the C t of gyrB for each strain).

Biofilm formation
Strains were incubated on BHI agar or BHI agar 0.25% glucose for 18 h at 35uC. BHI ( Figure 1A and 1B) or BHI 0.25% glucose broths were then adjusted to a 0.5 McFarland standard for transfer into wells of a flat-bottom 96-well polystyrene microtiter plate containing a half volume of the same medium and 26 the final concentration of the tested compound (glucose, NaCl, phenylmethylsulfonyl fluoride [PMSF] or CdCl 2 ) in a final volume of 200 ml. PMSF (Sigma) was used at a final concentration of 100 mM. For biofilm dispersal assays, two-day old biofilms were incubated for 2 h with proteinase K (Sigma) and DNase I (NEB) at a final concentration of 100 mg/ml and 104 U/ml, respectively. The plates were incubated at 35uC for a total of 48 h and the amount of residual biofilm was revealed as previously described [32,33].

Autolysis
Overnight cultures were used to inoculate BHI broths at an A 600 nm of 0.1. Bacteria were then grown with shaking at 35uC for 2.5 h (100 ml of culture in a 250-ml erlenmeyer), collected, washed and suspended in a 0.05 M glycine buffer (pH 8.0). This bacterial suspension was mixed 1:1 with 0.05 M glycine (pH 8.0) supplemented or not with Triton X-100 at a final concentration of 0.05% (initial bacterial suspension to flask volume ratio of 1 to 4). The decrease in A 600 nm was monitored as a function of time in order to evaluate bacterial lysis and was expressed relative to the initial turbidity. Bacterial suspensions were maintained at 35uC and agitated throughout the experiment, and were vortexed before each measurement of turbidity. Similar results were also obtained using 0.05 M Tris-HCl (pH 7.2) (data not shown).

Cell infection assays
Cell infection assays were performed as previously described with few modifications [34]. Briefly, cells were seeded on 12-well Transwell plates (,1.5610 5 cells/insert) and cultured for 9 to 10 days in an air-liquid system. The complete medium in basal compartments was replaced by the invasion medium (1% FBS and no antibiotics) 18 h before assays. Inocula were prepared by suspending bacteria grown for 20 h on BHI agar plates in ice-cold phosphate-buffered saline (PBS). The use of bacteria harvested CF03-L Normal strain co-isolated with CF03-S [33] CF07-S SCV strain isolated from a CF patient [25] CF07-L Normal strain co-isolated with CF07-S [33] CF1D-S SCV strain isolated from a CF patient [33] CF1A-L Normal strain co-isolated with CF1D-S [33] CF07-SDsigB CF07-S DsigB::ermA; Em r [31] CF07-SDfnbA CF07-S from agar plates allowed to drastically decrease the occurrence of normal-growing revertants in inocula [34]. Inocula were then washed three times in ice-cold PBS and suspended in the invasion medium supplemented with 0.5% bovine serum albumin (BSA) at a density of approximately 4610 8 CFU/ml. Cells were washed twice with PBS and 250 ml of the bacterial suspension was apically added to each insert, which yielded a multiplicity of infection of 100:1. Invasion was allowed for 3 h, and then inserts were emptied and washed three times with PBS. Invasion medium supplemented with 20 mg/ml of lysostaphin or 20 mg/ml of lysostaphin/50 mg/ ml of gentamicin (Sigma) was then added to both apical and basal compartments to kill extracellular bacteria and the cells were incubated for the indicated time. When only lysostaphin was added to the invasion medium, cells were washed once again with PBS and the invasion medium supplemented with lysostaphin was replaced at 24 h post-invasion and/or 1 h before cell lysis to ensure that no bacteria survived or replicated outside cells. Results obtained using both protocols were similar. Whenever required, CdCl 2 was added to the supplemented invasion media. At the end of the incubation time and following three washes with PBS, cells were detached with 100 ml of 0.25% trypsin and lysed for 10 min by the addition of 400 ml of water containing 0.05% Triton X-100. Lysates were serially diluted 10-fold and plated on agar for CFU determination.

Cytotoxicity assays
Apical cell culture supernatants from cells infected for 48 h (see Cell infection assays protocol) were harvested. Cells and bacteria  were removed from supernatants by two subsequent 5 min centrifugations at 300 and 68006 g, respectively. Samples were stored at 4uC for a maximum of 2 days before use and host cell lysis was evaluated based on the amount of lactate deshydrogenase (LDH) activity as recommended by the manufacturer (Roche, Laval, Qc, Canada).

Mouse model of pulmonary infection
Colonies grown on BHI agar plates for 20 h were used to prepare bacterial suspensions in cold phosphate-buffered saline (PBS, Sigma). Bacterial suspensions were washed twice and suspended in cold PBS to the number of CFUs required for infection (1610 8 CFUs). The test bacteria (CF07-L, CF07-S and CF07-SDsigB) were used alone or combined (CF07-L with CF07-S or CF07-L with CF07-SDsigB). For a given experiment, the quantity of bacteria in the mixed inoculum was equivalent to the sum of each inoculum prepared for individual strains. CD-1 female mice (22-24 g, Charles River Canada) were inoculated by intratracheal instillation of 50 ml of the bacterial inoculum using a sterilized 250-ml glass syringe (Hamilton Company) to which was attached a slightly bent 24G feeding needle (Fine Science Tools, Canada). Then, using an otoscope equipped with a speculum (model 21700, Welch Allyn), the larynx was located and the feeding needle was inserted into the trachea for more or less one centimeter. While still looking through the otoscope to maintain the needle in position, the inoculum was slowly instilled. After 48 h of infection, both lungs were harvested and homogenized (Kinematica Polytron homogenizer) in 2 ml of PBS. The CFU content was evaluated by plating logarithmic dilutions of homogenates on TSA and enumerating normal and SCV colonies. The number of SCV colonies was confirmed on TSA containing 2 mg/ml of gentamicin. At least two independent experimental infections were carried out and data were combined. Figure S1 Expression ratio of the aur gene as a function of growth for strains CF07-L and CF07-S. QPCR results are expressed according to CF07-L in the early exponential phase of growth. No statistically significant difference was revealed for each growth phase (unpaired t test, n = 4-5). Results are expressed as means with standard deviations. (TIF) Figure S2 Expression ratio of the sspA gene as a function of growth for strains CF07-S and CF07-SDsigB. QPCR results are expressed according to CF07-S in the early exponential phase of growth. A statistically significant difference between both strains was revealed for the stationary growth phase (**, P,0.01; unpaired t test, n = 3-5). Results are expressed as means with standard deviations. (TIF) Figure S3 Effect of PMSF on the biofilm formation of CF07-SDsigB. Relative biofilm formation of CF07-SDsigB in the presence of the serine-protease inhibitor PMSF following 48 h of incubation. Results are normalized according to the unexposed condition. No statistically significant difference was revealed (ANOVA with Dunnett's posttest, n = 3). Results are expressed as means with standard deviations. (TIF) Figure S4 Effect of menadione on the proteolytic activity of CF07-L, CF07-S and CF07-SDsigB. BHIg milkagar plates supplemented with 3 mg/ml of menadione were incubated for 48 h at 35uC. (TIF) Figure S5 Effect of Triton-X100 on the autolysis rate of strains CF07-L, CF07-S and CF07-SDsigB. Autolysis of strains CF07-L (A), CF07-S (B) and CF07-SDsigB (C) as a function of time exposed or not to 0.05% Triton-X100. Results are expressed as percentages of the initial turbidity for each condition.