2. Introduction
Staphylococcus aureus is a leading cause of all forms of orthopedic infection including osteomyelitis, septic arthritis, and infections associated with indwelling hardware [
1,
2,
3]. It is becoming increasingly difficult to treat these infections owing to the persistent emergence of strains resistant to key antibiotics. Moreover, these infections are characterized by formation of a biofilm, the presence of which confers a therapeutically relevant level of intrinsic resistance to both conventional antibiotics and host defenses [
4,
5]. Thus, the treatment of
S. aureus orthopedic infections requires long-term intensive antibiotic therapy accompanied by surgical debridement to remove infected tissues and indwelling hardware, and even after such intensive medical and surgical intervention, the recurrence rate is unacceptably high [
1,
2,
3].
These observations have rekindled interest in virulence-associated therapies that could be used to overcome this therapeutic recalcitrance [
6]. The staphylococcal accessory regulator (
sarA) and the accessory gene regulator (
agr) are the two
S. aureus regulatory loci that have been studied extensively in this regard [
6,
7,
8,
9,
10]. The
sarA locus encodes a 15 kDa DNA-binding protein (SarA) that impacts the transcription of many
S. aureus genes including
agr [
11]. Specifically, SarA is required for maximum expression of
agr, activation of which modulates the production of multiple
S. aureus virulence factors in a coordinated fashion [
12]. In this regulatory scenario,
agr is downstream of
sarA, suggesting that inhibition of
agr could disrupt the coordinated production of these virulence factors irrespective of the functional status of
sarA. However,
sarA also serves regulatory functions that are independent of
agr, one critical observation in this regard being that mutation of
sarA results in increased protease production and decreased biofilm formation, while mutation of
agr has the opposite effect on both phenotypes.
Extracellular proteases produced by
S. aureus include aureolysin, SspA (V8), SspB, ScpA, and as many as six
spl-encoded proteases (SplA-F) [
13,
14]. These proteases collectively serve vital roles on behalf of the bacterium by promoting tissue invasion, nutrient acquisition, and avoidance of host defenses [
13,
14,
15,
16,
17,
18,
19]. This suggests that the ability of
S. aureus to survive in vivo and cause disease would be limited in strains unable to produce proteases. Conversely, strains that produce elevated levels of these proteases would be predicted to exhibit increased virulence. However, a derivative of the methicillin-resistant USA300 strain LAC (Los Angeles County clone) that does not produce any of these 10 extracellular proteases is hypervirulent in murine models of bacteremia and osteomyelitis [
13,
20], while strains that produce significantly increased amounts of these proteases are attenuated in the same models [
20,
21,
22].
S. aureus sarA mutants are the prototype example of the latter in that they produce elevated levels these extracellular proteases, and this has been shown to play a significant role in defining the attenuation of
sarA mutants in diverse animal models of infection including osteomyelitis [
20,
22,
23,
24,
25]. Based on this, we used a mass-based proteomics approach to identify proteins that are present in reduced amounts in
sarA mutants by comparison to their isogenic parent strains owing to protease-mediated degradation [
21,
26]. Mutation of other
S. aureus regulatory loci also results in increased protease production, but we confirmed that mutation of
sarA results in a greater increase in protease production, and a greater decrease in virulence, than mutation of any other
S. aureus regulatory locus examined to date [
23,
27,
28]. However, these studies were limited to regulatory mutants that exhibit increased protease production, and as noted above this does not include
agr.
To address this, and to define the impact of
sarA on protease production in an
agr mutant, we used our murine osteomyelitis model to directly evaluate the relative virulence of LAC, isogenic
sarA,
agr, and
sarA/agr mutants, and isogenic derivatives of all four strains unable to produce aureolysin, ScpA, SspA, SspB, or SplA-F [
20]. The results confirmed that
sarA and
sarA/agr mutants are attenuated to a greater extent than the isogenic
agr mutant, and that this is due to, a statistically significant extent, the increased production of these extracellular proteases. This suggests that
S. aureus proteins of interest in the specific context of osteomyelitis could be identified and prioritized based on their reduced abundance in both
sarA and
sarA/agr mutants by comparison to LAC. Indeed, the reduced virulence of these mutants was correlated with reduced biofilm formation and reduced cytotoxicity for osteoblasts and osteoclasts, and these phenotypes were correlated in turn with the reduced abundance of full-length and presumably functional forms of specific virulence factors implicated in these phenotypes including α-toxin, both components of the Panton–Valentine leukocidin (LukF and LukS), the primary
S. aureus extracellular nuclease (Nuc1), and protein A (Spa) [
20]. However, our studies investigating these correlations were limited by the availability of antibodies appropriate for targeted Western blot analysis. To overcome this limitation, we used a mass-based proteomics approach capable of quantitatively distinguishing between full-length and truncated
S. aureus proteins [
21,
26] to assess the global impact of increased protease production on the proteomes of
sarA and
sarA/
agr mutants and to identify
S. aureus proteins present in reduced amounts in both attenuated mutants owing to protease-mediated degradation.
4. Results
Characterization of mutants. We first evaluated our results in the context of the mutations that define each of the strains included in these studies. For instance, although SarA is an intracellular protein it was detected at comparable levels in conditioned medium (CM) from overnight (16 h) stationary phase cultures of LAC and its protease-deficient mutant, it was not detected in CM from
sarA or
sarA/
agr mutants or their protease-deficient derivatives (
Table 1). AgrA is also an intracellular protein, and it was not detected in CM from LAC or any of its isogenic regulatory mutants irrespective of their ability to produce extracellular proteases. However, delta toxin (δ-toxin), which is encoded within the regulatory RNA associated with
agr (RNAIII) that is produced under the positive regulatory control of AgrA was detected at comparable levels in LAC and its protease-deficient derivative, but not in CM from the
agr or
sarA/
agr mutants (
Table 1). The abundance of δ-toxin was also reduced in CM from the LAC
sarA mutant by comparison to CM from LAC. This is consistent with studies demonstrating that
sarA is required for maximum expression of
agr [
11,
12]. However, the abundance of δ-toxin was restored in the protease-deficient
sarA mutant, demonstrating that the increased production of extracellular proteases further limits the abundance of δ-toxin at a post-translational level.
Similarly, the only
S. aureus proteins encoded outside the
agr operon and known to be produced under the regulatory control of AgrA are the genes encoding phenol-soluble modulins (PSMs) [
30], and all alpha class PSMs (PSMα1, PSMα2, PSMα3, and PSMα4) were absent in CM from the LAC
agr and
sarA/
agr mutants and their protease-deficient derivatives (
Table 1). PSMs were present at comparable levels in LAC and its protease-deficient mutant, and absent or only present at very low levels (≤0.6 average spectral counts) in the isogenic
sarA mutant. However, the abundance of all four alpha class PSMs was increased in CM from the protease-deficient
sarA mutant. This confirms the importance of
agr in PSM production, and the importance of
sarA in limiting proteases to facilitate their accumulation.
The protease-deficient mutants have null mutations in all of the genes encoding aureolysin, ScpA, SspA, SspB, and SplA-F [
26], so it would be anticipated that none of these proteases would be detected. We did detect some of these proteases in CM from the protease-deficient
sarA mutant and, to a lesser extent, the protease-deficient derivative of LAC (
Table 1). However, the abundance of these proteases even in CM from the protease-deficient
sarA mutant was very low (≤1.2 average spectral counts across five biological replicates), and none were detected in the protease-deficient
agr or
sarA/
agr mutants. Our previous phenotypic characterization of these mutants also confirmed that protease activity was limited to the same degree in all four protease mutants [
20].
The primary strain in which proteases were detected in a protease mutant was the protease-deficient
sarA mutant. This is relevant in that every extracellular protease other than SspB was present in an increased amount in CM from the
sarA mutant by comparison to CM from LAC (
Table 1). This suggests that the low levels detected in CM from the protease-deficient
sarA mutant are background noise. However, the observation that SspB, the abundance of which was decreased in CM from
sar, agr and
sarA/
agr mutants, was detected in CM from the protease-deficient
sarA mutant contradicts this conclusion. At the same time, we previously demonstrated that spectral counts for full-length SspB are decreased in CM from a
sarA mutant, but total spectral counts as assessed independently of mass are increased [
26]. This is consistent with our previous demonstration that transcription from all four protease-encoding genes and operons is increased in
sarA mutants [
23], and it confirms that mutation of
sarA results in the increased production of all 10 of the primary extracellular proteases produced by LAC. By comparison to LAC, the abundance of all proteases was reduced in the LAC
agr mutant (
Table 1).
Most proteases were present in CM from the
sarA/
agr mutant in amounts comparable to CM from LAC and higher than those observed in the isogenic
agr mutant (
Table 1). However, the abundance of most was reduced in CM from the
sarA/
agr mutant by comparison to the isogenic
sarA mutant. The exception was aureolysin, which was most abundant in CM from the
sarA/
agr mutant. This likely accounts for the increase in overall protease activity in CM from the
sarA/
agr mutant relative to CM from LAC, while the fact that none of the other proteases were increased likely explains why overall protease activity was decreased in the
sarA/
agr mutant by comparison to the
sarA [
20]. Importantly, aureolysin is one of two proteases shown to contribute to the greatest extent to the attenuation of
sarA mutants in osteomyelitis [
24]. The other is staphopain A (ScpA), the abundance of which was increased to a greater extent in the
sarA mutant than the isogenic
sarA/
agr mutant (
Table 1). In fact, the abundance of ScpA was not significantly increased in CM from the
sarA/
agr mutant by comparison to CM from LAC. Mutation of the genes encoding these same proteases (
aur and
scpA) was also shown to replicate the hypervirulent phenotype of a LAC protease null mutant in a murine sepsis model [
13].
The fact that the abundance of aureolysin was increased in both
sarA and
sarA/
agr mutants while the abundance of ScpA was only increased in the
sarA mutant suggests that the increased production of aureolysin may be particularly important in the pathogenesis of osteomyelitis. Indeed, aureolysin-mediated degradation of alpha PSMs was shown to be a primary factor defining the attenuation of a LAC
saePQRS (
sae) mutant in osteomyelitis [
31]. The reduced abundance of PSMs was also confirmed to contribute to the attenuation of a LAC
sarA mutant, although whether this could be attributed specifically to degradation by aureolysin was not examined [
32].
Our previous Western blots demonstrated that the increased production of extracellular proteases in a LAC
sarA mutant limits the abundance of LukF, LukS, and Nuc1 in their full-length forms with a corresponding increase in the abundance of a truncated but immunoreactive protein [
20]. Using our mass-based proteomics approach, we confirmed the presence of truncated versions of LukF, LukS, and Nuc1 in CM from the
sarA mutant but not the protease-deficient
sarA mutant (
Figure 1), thus confirming the ability to quantitatively distinguish between full-length and truncated proteins. We also confirmed that these proteins, whether in truncated or full-length form, are present in increased amounts in CM from the
sarA mutant and its protease-deficient derivative by comparison to CM from LAC (
Figure 1), confirming that mutation of
sarA results in increased production of LukF, LukS, and Nuc1, but that the increased production of proteases limits their abundance in full-length and presumably functional forms.
Additionally,, we previously concluded that LukF is only present in a truncated form in CM from the
sarA mutant and that this might be a limiting factor in the activity of the bi-component Panton–Valentine leukocidin (PVL) [
20]. Here, we confirmed the presence of this truncated product and the fact that it was absent in CM from the protease-deficient
sarA mutant, but also detected a larger version of LukF that was not apparent by Western blot (
Figure 1). This size of this protein was similar to full-length LukF, but it was nevertheless truncated as evidenced by the loss of an antibody-binding site [
20]. Thus, this does not contradict the suggestion that the absence of full-length LukF may be a limiting factor in the phenotypic impact of PVL, but it does further illustrate the utility of our mass-based approach. This did not appear to be a concern with LukS and Nuc1 in that both full-length and truncated versions of these proteins were detected by both Western blot and in our proteomics analysis (
Figure 1).
Impact of protease production on the proteome of S. aureus sarA, agr, and sarA/agr mutants. The results discussed above are consistent with the expected phenotypes of all mutants included in our comparisons. They also validate our mass-based approach including its ability to distinguish between truncated and full-length proteins. Based on this, we extended our analysis in an unbiased, comprehensive, and antibody-independent fashion. We detected 1039 full-length proteins in conditioned media (CM) from stationary phase (16 h) cultures of LAC (
Supplementary Tables S1 and S2). Only 139 (13.4%) were detected in CM from the
sarA mutant, while 870 (83.7%) were detected in the protease-deficient
sarA mutant (
Figure 2). A total of461 full-length proteins (44.4%) were detected in CM from the
sarA/
agr mutant, with 563 (54.2%) being detected in CM from the protease-deficient
sarA/
agr mutant. The lower number of proteins detected in CM from the
sarA/
agr mutant by comparison to the
sarA mutant is consistent with the observation that protease production is higher in CM from the
sarA mutant than in CM from the
sarA/
agr mutant [
20]. Conversely, the fewer proteins detected in CM from the protease-deficient
sarA/
agr mutant by comparison to the protease-deficient
sarA mutant is consistent with the impact of mutating
agr on protein production. This is further reflected in the reduced number of proteins detected in CM from the
agr mutant (496), which was not increased in CM from the isogenic protease-deficient
agr mutant. Indeed, the
agr mutant was the only regulatory mutant examined in which the number of proteins detected was not increased in the protease-deficient derivative (
Figure 2).
Correlations between protein abundance and virulence. LAC
sarA and
sarA/
agr mutants are more attenuated in our osteomyelitis model than an isogenic
agr mutant, and virulence is restored to a significant extent in a protease-deficient derivatives of both mutants [
20]. This suggests that proteins that are present in decreased amounts in CM from
sarA and
sarA/
agr mutants owing to protease-mediated degradation would be of interest in the pathogenesis of osteomyelitis. We used a log
2 fold change >2.0 and
p value <0.05 to identify these proteins. Because statistical analysis is not possible when considering a protein that was not detected in one of the strains included in the comparison, we also identified proteins that were not detected in CM
sarA and
sarA/
agr mutants but were detected in CM from both of their protease-deficient derivatives (
Supplementary Tables S1 and S2).
The ability to identify some proteins of potential interest in CM from the
sarA/
agr mutant was limited by the influence of
agr on protein production. For instance, PSMs contribute to the pathogenesis of osteomyelitis and are present in reduced amounts in CM from
sarA mutants owing to protease-mediated degradation [
32], but the impact of proteases on the abundance of PSMs is not apparent in CM from
sarA/
agr mutant because PSMs are not made owing to the
agr mutation (
Table 1). In fact, 543 of the 1039 proteins detected in CM from LAC (52.3%) were not detected in CM from the isogenic
agr mutant, and virtually all of these were detected at very low levels if at all in CM from the protease-deficient
agr mutant (
Supplementary Table S3). The abundance of an additional 426 proteins was decreased to a significant extent in CM from the
agr mutant by comparison to CM from LAC (
Supplementary Table S4), with the remaining 70 proteins being detected in CM from LAC and its
agr mutant in comparable amounts (
Supplementary Table S5). None of the 426 proteins that were present in decreased amounts in CM from the
agr mutant were significantly increased in CM from the protease-deficient
agr mutant. This demonstrates that the reduced abundance of proteins in the
agr mutant is a function of limited production rather than protease-mediated degradation. Nevertheless, because
sarA and
sarA/
agr mutants are attenuated in our osteomyelitis model owing at least in part to protease-mediated degradation [
20], identifying proteins with reduced abundance in
sarA and
sarA/
agr mutants remains important.
A total of 45 proteins were present in a significantly increased amount in CM from the protease-deficient
sarA/
agr mutant by comparison to the
sarA/
agr mutant itself, with an additional 182 being detected in CM from the protease-deficient
sarA/
agr mutant but not in CM from the
sarA/
agr mutant (
Supplementary Table S6). Thus, the abundance of 227 proteins was limited to some degree by increased protease production in the
sarA/
agr mutant (
Figure 3). The number of proteins limited by increased protease production in the
sarA mutant was much greater, but this would be expected given the influence of
agr on protein production and the observation that protease activity in the
sarA mutant is higher than in the isogenic
sarA/
agr mutant [
20].
Of the 227 proteins found to be limited by extracellular proteases in CM from the
sarA/
agr mutant, 3 were undetectable in CM from both the
sarA mutant and its protease-deficient derivative (
Supplementary Table S6), thus excluding them from the common pool of proteins present in reduced amounts in both
sarA and
sarA/
agr mutants owing to protease-mediated degradation. Ten of the remaining 224 were present in a significantly increased amount in CM from the protease-deficient
sarA mutant by comparison to CM from the
sarA mutant, while 214 were detected in CM from the protease-deficient
sarA mutant but not in CM from the
sarA mutant itself (
Supplementary Table S6). These results are consistent with the hypothesis that the increased production of extracellular proteases in
sarA mutants limits the abundance of proteins that are potentially important in the pathogenesis of osteomyelitis, and that it does so irrespective of the functional status of
agr, suggesting that these 224 proteins may be of particular interest in this regard.
Impact of eliminating protease production on the proteome of LAC. While our primary focus was on identifying proteins that are present in reduced amounts owing to increased protease production in
sarA and
sarA/
agr mutants, it was previously reported that eliminating the ability produce extracellular proteases increases the virulence of LAC itself in a murine sepsis model [
13]. Since it might be anticipated that this could be attributed to an increase in the abundance of some of the same virulence factors that were limited by protease production in
sarA and
sarA/
agr mutants, we also identified proteins that were present in increased amounts in CM the protease-deficient derivative of LAC by comparison to CM from LAC itself. However, the only two proteins we identified were ArcB and MreC. MreC was among the proteins present in reduced amounts in CM from both the
sarA and
sarA/
agr mutants owing to protease-mediated degradation, but ArcB was not (
Supplementary Table S6).
Additionally, neither ArcB nor MerC were more abundant in CM from the protease-deficient derivative of LAC that was found to be hypervirulent in a murine sepsis model [
13]. Moreover, none of the six proteins that were identified in this study (LukA, PSMα4, Sbi, SEK, SPIN, and a putative chitinase encoded by SAUSA300_0964) were identified in our proteomic comparisons of LAC and its protease-deficient derivative (
Supplementary Tables S1 and S2). However, the abundance of all six was reduced in CM from the
sarA mutant, and other than Sbi the abundance of these proteins was increased to a statistically significant extent by eliminating the production of extracellular proteases (
Figure 4). The abundance of Sbi, LukA, SAUSA300_0964, and PSMα4 was also reduced in CM from the
sarA/
agr mutant, although the reduced abundance of LukA and PSMα4 is likely due to the
agr mutation (
Table 1).
Impact of sarA and agr on oxacillin resistance. We also found that the penicillin-binding proteins PBP1, PBP2, PBP2A, and PBP3 were detected in CM from LAC but not in CM from its isogenic
sarA mutant, and that the abundance of all four PBPs was increased in CM from the protease-deficient
sarA mutant (
Figure 5). Mutation of
sarA increased susceptibility to oxacillin and was increased even more in
sarA/
agr and
agr mutants (
Figure 6). However, eliminating the ability to produce extracellular proteases did not impact the susceptibility of any of these regulatory mutants. These results are consistent with a previous report concluding that the impact of
sarA on oxacillin resistance is a function of decreased transcription of the gene (
mecA) encoding PBP2A [
33]. Irrespective of the mechanism(s) involved, this suggests that targeting
sarA could have the added benefit of enhancing the ability to overcome the therapeutic recalcitrance of osteomyelitis, particularly when caused by methicillin-resistant strains.
Impact on factors involved in invasion of the osteocyte lacuna-canalicular network (OLCN). The reduced abundance of PBP3 may also be relevant in the pathogenesis of osteomyelitis because the absence of PBP3 has been correlated with a decreased ability of
S. aureus to invade the OLCN [
34]. Mutation of the gene encoding the primary
S. aureus autolysin (
atl) also limits OLCN invasion, and like PBP3 the abundance of Atl was reduced in CM from
sarA and
sarA/
agr mutants. However, unlike PBP3, the abundance of Atl was restored to a statistically significant extent in CM from the protease-deficient
sarA and
sarA/
agr mutants (
Figure 7). Since invasion of the OLCN has been proposed to provide
S. aureus with a protective niche in bone, this suggests another mechanism by which targeting
sarA could enhance the ability to overcome the therapeutic recalcitrance of osteomyelitis to conventional antibiotic therapy.
Impact on Atl on exoprotein abundance. It was recently demonstrated that mutation of
atl results in the increased abundance of extracellular LukAB owing to changes in peptidoglycan cleavage [
35]. Atl was essentially undetectable in CM from the
sarA mutant, so it might be anticipated that the phenotype of the
sarA mutant would mimic the phenotype of an
atl mutant. Since eliminating protease production restored the abundance of Atl in the
sarA mutant, this leaves open the possibility changes in Atl rather than changes in protease production account for the increased abundance of at least some proteins in CM from the protease-deficient
sarA mutant by comparison to CM from the
sarA mutant itself. To test this, we used Western blots to examine the abundance of α-toxin, LukF, and LukS in CM from LAC, its
sarA and protease-deficient
sarA mutants, and an isogenic
atl mutant. The results confirmed our previous findings [
20] with respect to the decreased abundance of these proteins in their full-length form in CM from the
sarA mutant and their increased abundance relative to LAC in CM from the protease-deficient
sarA mutant (
Figure 8). They also demonstrated that the abundance of α-toxin and LukS, but not LukF, was increased relative to LAC in CM from the
atl mutant. However, the abundance of all three exoproteins was still lower in CM from the
atl mutant than CM from the protease-deficient
sarA mutant (
Figure 8).
5. Discussion
Staphylococcus aureus produces a diverse array of virulence factors and causes an equally diverse array of infections. This has complicated attempts to identify specific virulence factors that can be studied to gain a better understanding of pathogenesis mechanisms or perhaps even be targeted to prophylactic and/or therapeutic advantage. One alternative is to focus on regulatory elements that modulate the production of multiple
S. aureus virulence factors, and the two regulatory targets that have been investigated most extensively in this regard are
sarA and
agr [
6,
7,
8,
9,
10]. Although
sarA is known to function through an agr-dependent pathway [
11,
12,
36,
37], the impact of mutating
sarA and
agr is often different. One example of specific relevance in this report is their opposite impact on the production of extracellular proteases [
38]. Specifically, mutation of
agr results in decreased protease production, while mutation of
sarA has the opposite effect.
We have extensively investigated the impact of increased protease production in
S. aureus sarA mutants and demonstrated that mutation of
sarA results in a greater increase in protease production, and a greater decrease in biofilm formation, than mutation of any other
S. aureus regulatory locus identified to date, and that this is true in diverse clinical isolates including both methicillin-sensitive and methicillin-resistant strains [
23,
27]. These collective studies led us to propose that the ability of
sarA to repress the production of extracellular proteases is a key mechanism of post-translational regulation in
S. aureus that prevents these proteases from compromising the
S. aureus proteome [
21,
22,
23,
24,
25,
26]. However, none of these earlier studies considered the functional status of
agr, which is an important consideration given that spontaneous
agr mutants are known to arise in vivo and may in fact promote the transition between acute and chronic forms of
S. aureus infection including osteomyelitis [
39,
40,
41,
42].
To address this, we used our osteomyelitis model to assess the relative virulence of sarA, agr, and sarA/agr mutants generated in the USA300 strain LAC, and the results confirmed that
sarA and
sarA/agr mutants are more attenuated in our osteomyelitis model than the isogenic
agr mutant [
20]. This suggests that
sarA is not only a viable target for therapeutic intervention but is perhaps the preferred target by comparison to other regulatory loci including
agr. The observation that the reduced virulence of both
sarA and
sarA/agr mutants can be attributed to a significant extent to the increased production of extracellular proteases also suggests that a proteomics approach could be used to go beyond the regulatory loci themselves to identify and prioritize specific virulence factors of interest based on a correlation between reduced virulence and the reduced abundance of these virulence factors. In fact, to the extent that our collective studies have also demonstrated that mutation of
sarA results in the increased production of specific
S. aureus virulence factors, but limited accumulation of these virulence factors in their full-length and presumably functional forms owing to protease-mediated degradation [
20], it suggests that a proteomics approach capable of quantitatively distinguishing between full-length and truncated proteins is the only way to accomplish this important task.
Using this approach, we identified 224 full-length S. aureus proteins that are present in reduced amounts in CM from LAC sarA and sarA/agr mutants by comparison to CM from their isogenic protease-deficient derivatives. Many of these proteins were present at very low levels, and in fact most (214) were only identified because they were not detected in CM from sarA and sarA/agr mutants but were detected in CM from their protease-deficient derivatives. The absence of individual proteins in some samples precludes statistical analysis, but we included these among the 224 proteins of potential interest because it is not possible to know how much of a given protein is required to have a phenotypic effect.
The studies we report are an extension of previous studies that used the same mass-based proteomics approach to examine the correlation between protein abundance and virulence in our osteomyelitis model using six isogenic LAC mutants that differed in the functional status of
saeRS and
sarA relative to each other [
21]. This was based on the demonstration that
saeRS and
sarA act synergistically to limit protease production [
43] and that mutation of
saeRS also limits virulence in an osteomyelitis model [
21]. In fact, we demonstrated that LAC and its constitutively active
saeRS variant (
saeC) are more virulent that isogenic
saeRS,
sarA,
saeRS/
sarA, and
saeC/
sarA mutants, all of which are attenuated to a comparable degree. Proteomic analysis led to the identification of 114 proteins that were significantly more abundant in CM from both virulent strains by comparison to all 4 attenuated strains. Of these 114 proteins, 25 were among the 224 proteins identified in this study (
Supplementary Table S6).
Since mutation of
saeRS also results in increased protease production, albeit to a lesser degree than mutation of
sarA [
43], it might be anticipated that the overlap between these two datasets would be larger. However, most of the 224 proteins we identified in this study were identified only because they were undetectable in CM from
sarA and
sarA/
agr mutants and detectable in CM from their protease-deficient derivatives, and in our previous study we excluded proteins from further consideration if they were not detected in CM from all 6 strains [
21]. Moreover,
sae also functions in a coordinated fashion with
agr to regulate the production of
S. aureus virulence factors in response to different environmental signals [
44]. This further illustrates the complex and highly interactive nature of
S. aureus regulatory circuits and the critical need to maintain an appropriate balance between protein production and protein degradation. This also provides additional evidence supporting the hypothesis that limiting protease production is a key factor in the post-translational remodeling of the
S. aureus virulence factor repertoire and suggests that the 25 proteins (
Supplementary Table S6) that were identified in our collective studies based on correlations with virulence may warrant prioritization even among the 224 identified in this report.
In addition to
sarA and
saeRS, mutation of several other
S. aureus regulatory proteins has been implicated in protease production, and this suggests that additional proteomic comparisons could further enhance the ability to prioritize specific targets for further study. To date, we have examined the impact of 21
S. aureus regulatory loci implicated in protease production and biofilm formation. The results demonstrate that mutation of
sarA results in a greater increase in protease production than mutation of any of these other regulatory loci, but also suggest that
codY,
sigB, and
rot warrant further consideration in this regard [
22,
25,
27]. It is also important to extend such comparisons to other clinically relevant strains. Indeed, the methicillin-sensitive, USA200 strain UAMS-1 is virulent in our osteomyelitis model, and mutation of
sarA attenuates this virulence owing to increased protease production, but LAC and UAMS-1 differ in significant ways beyond their methicillin-resistance status including their ability to produce important cytolytic including α-toxin and the Panton–Valentine leucocidin [
23,
24,
25,
32]. This suggests alternative pathogenesis pathways to the same end, thus making it important to prioritize based not only common proteins impacted by mutation of different regulatory loci but also on common proteins impacted in diverse clinical isolates of
S. aureus.
Additionally, comparison of datasets obtained independently of each other can be informative, but direct comparisons using the same experimental and statistical methods are clearly preferred, particularly given the sensitivity of current proteomics technology. Indeed, while our primary focus was on identifying proteins that were present in reduced amounts in both
sarA and
sarA/agr mutants owing to protease-mediated degradation, we included CM samples from a protease-deficient derivative of LAC itself, and we did not identify any of the six proteins that Gimza et al. [
13] found were more abundant in a protease-deficient derivative of LAC by comparison to LAC itself. This could potentially be explained by differences in the specific experimental approaches employed in these two studies, and this accounts for our efforts to fully validate our proteomics approach including its ability to distinguish between truncated and full-length proteins. We also employed a different model to assess virulence, and we did not observe a statistically significant increase in virulence with the protease-deficient derivative of LAC in the specific context of osteomyeltis, but we did demonstrate what we believe are clear trends suggesting increased virulence in the protease-deficient derivative of LAC [
20]. Moreover, whether virulence was significantly increased in osteomyelitis vs. sepsis is irrelevant in the context of this discussion since both proteomic comparisons were performed using CM from stationary phase cultures grown in vitro in tryptic soy broth.
There were, however, other differences with respect to sample preparation and analysis that could impact the results. For instance, Gimza et al. [
13] standardized cultures based on optical density before precipitating proteins and standardizing samples for comparative analysis based on overall protein content. In contrast, we examined CM after standardization based on optical density without further manipulation based on the logic that standardizing based on protein content might mask important differences owing to the global impact of extracellular proteases. Both studies also used a mass-based approach after resolution by SDS-PAGE, but we used gradient gels and focused on a narrower size range to distinguish between the abundance full-length and truncated proteins [
21,
24,
26]. It is reasonable to suggest that such differences could impact the results, but it is difficult to envision how they can fully explain this discrepancy since we also did not identify any of these six proteins even after removing the size restriction and examining total abundance in any form. Thus, we do not have a definitive explanation for this discrepancy, but this nevertheless further illustrates the need to make direct comparisons between multiple strains and multiple isogenic mutants proven to exhibit significant differences in virulence, and to make these comparisons using consistent methods that have been validated to the greatest extent possible. Given that our proteomics analysis confirmed the identity of all mutants included in our comparisons, and also validated recognized phenotypes beyond the regulatory mutations themselves, we believe the experimental approach we describe is appropriate to address these issues moving forward.
Indeed, while we did not find that any of the six proteins identified by Gimza et al. [
13] were more abundant in CM from the protease-deficient derivative of LAC by comparison to CM from LAC itself, we did find that the abundance of four of these six proteins were reduced in CM from the attenuated LAC
sarA and
sarA/
agr mutants. In some cases, the reduced abundance of these proteins in CM from the
sarA/agr mutant could not be attributed to increased protease production, but this does not detract from the possibility that the reduced abundance of these proteins in CM from
sarA and
sarA/agr mutants contributes to their reduced virulence. Indeed, this further illustrates the importance of
agr in enhancing the production of
S. aureus virulence factors, and the importance of
sarA in concomitantly repressing the production of extracellular proteases to maintain a critical balance in the abundance of these virulence factors that is to the best advantage of
S. aureus in specific in vivo environments including bone.
Identifying critical virulence factors is important because these factors contribute to in vivo survival and ultimately to the damage to the host associated with ongoing infection. To the extent that it compromises the local vasculature, this damage contributes to the therapeutic recalcitrance of
S. aureus orthopedic infections to conventional antibiotic therapy. However, the limited abundance of these virulence factors is not the only relevant observation in defining this therapeutic recalcitrance. Indeed, penicillin-binding proteins including PBP2A were also identified in this report. This is consistent with previous reports demonstrating that mutation of
sarA increases β-lactam susceptibility [
33]. These reports concluded that the increased β-lactam susceptibility of
sarA mutants was correlated with reduced
mecA transcription, and our results demonstrating that the increased susceptibility to oxacillin observed in a
sarA mutant is not impacted by increased protease production support this conclusion. There is also a report demonstrating that SarA binds the
mecA promoter [
45], further suggesting a protease-independent transcriptional mechanism. However, irrespective of the mechanism involved, it is important to note that mutation of
sarA increased oxacillin susceptibility to a level comparable to the breakpoint minimum inhibitory concentration (MIC) for a methicillin-resistant strain (≤2.0 µg/mL).
Additionally, PBP3 and the primary
S. aureus autolysin Atl were also present in low levels in CM from
sarA and
sarA/
agr mutants, and the abundance of both was significantly restored by eliminating protease production. The impact of protease production on the abundance of Atl and PBP3 may be particularly relevant to both the pathogenesis and therapeutic recalcitrance of
S. aureus osteomyelitis. Specifically, mutation of
atl and
pbp3 has been shown to limit septic implant loosening and abscess formation in the medullary cavity, while mutation of
pbp3 has been correlated with reduced peri-implant osteolysis, reduced osteoclast activity, reduced invasion of the OLCN, and reduced production of the activator of nuclear factor kappa-B ligand (RANKL), a central modulator of the balance between osteoblast and osteoclast activity in bone remodeling [
34,
46]. Thus, these observations may be particularly relevant in the specific clinical context of
S. aureus orthopedic infections.
Finally, one unexpected result we observed is the reduced abundance of SarA in CM from the
agr and protease-deficient
agr mutants (
Table 1). It has been demonstrated that
sarA is required for maximum expression of
agr, but we are unaware of any reports suggesting that
agr enhances expression of
sarA. Our studies were performed using a null mutant that does not produce SarA, but different levels of SarA production in different strains has been correlated with relative levels of protease production [
47]. Thus, if this were a true regulatory effect with a phenotypic impact, it would be anticipated that the limited production of SarA in the
agr mutant would result in increased protease production, and this was not the case for any of the 10 extracellular proteases (
Table 1). Based on the observation that the abundance of Atl was reduced in CM from the
agr and protease-deficient
agr mutants, one plausible explanation is that the reduced abundance of SarA in CM from
agr mutants reflects changes in the abundance of Atl leading to changes in cell wall turnover and release of intracellular proteins, particularly since our comparisons were performed using CM from stationary phase cultures.
The abundance of SarA was comparable in CM from the
agr and protease-deficient
agr mutants, as was the abundance of Atl and SarA (
Table 1). Atl abundance was also lower in
sarA and
sarA/
agr mutants by comparison to the isogenic
agr mutants, and in both mutants it was significantly increased in the isogenic protease-deficient strains. This leaves open the possibility that differences in the abundance of SarA and presumably other proteins in CM from
sarA and
sarA/agr mutants by comparison to their protease-deficient derivatives is a function of the abundance of Atl. Indeed, it was recently demonstrated that mutation of
atl results in the increased abundance of extracellular LukAB owing to changes in peptidoglycan cleavage [
34]. Additionally, Atl was essentially undetectable in CM from the
sarA mutant, so it might be anticipated that the phenotype of the
sarA mutant would mimic the phenotype of an
atl mutant. LukA and LukB (SAUSA300_1975 and SAUSA300_1974, respectively) were undetectable in CM from
sarA and
sarA/
agr mutants but readily detectable in CM from their protease-deficient derivatives (
Supplementary Table S2). This is consistent with the hypothesis that the abundance of these proteins in CM from
sarA mutants is limited directly by protease-mediated degradation, but it could also be explained at least to some degree by the limited abundance of Atl in
sarA and
sarA/agr mutants.
However, despite the reduced abundance of Atl in CM from the sarA and sarA/agr mutants by comparison to their protease-deficient derivatives, the abundance of LukF, LukS, and Nuc1 was increased to a comparable degree in both regulatory mutants irrespective of their ability to produce proteases, with the only protease-dependent difference being whether these proteins were present in truncated or full-length and presumably functional forms. Additionally, we confirmed by Western blot that, while mutation of atl does result in an increase in the abundance of some exoproteins (e.g., α-toxin and LukS), the abundance of these proteins, as well as LukF, remains below that observed in an isogenic protease-deficient sarA mutant. Thus, while we cannot exclude the possibility that changes in the abundance of Atl contribute to differences in the abundance of some of the 224 proteins defined as protease dependent by our comparisons, we do not believe they can account for our cumulative results to an extent that detracts from the conclusion that increased protease production plays a phenotype-defining role in sarA and sarA/agr mutants that includes reduced virulence in the clinical context of osteomyelitis even if differences in the abundance of Atl, which are themselves protease dependent, do in fact contribute to the proteome phenotype of these mutants.