Sub-inhibitory concentrations of oxacillin modify the expression of agr locus in Staphylococcus aureus clinical strains belonging to different clonal complexes

Background The ability of Staphylococcus aureus to invade tissues and cause an infectious disease is the result of a multi-factorial process supported by the huge number of virulence factors inherent to this microorganism tightly regulated by the accessory gene regulator (agr). During antimicrobial therapy bacteria may be exposed to sub-inhibitory concentrations (subMICs) of antibiotics that may trigger transcriptional changes that may have an impact on the pathogenesis of infection. The objective of this study was to investigate the effect of oxacillin sub-MICs on agr system expression as the key component in the regulation of virulence in methicillin-susceptible (MSSA) and -resistant S. aureus (MRSA) strains. Furthermore, we studied the genetic basis of the agr locus and their potential association with the expression levels. Methods We have examined the expression of RNAIII and agrA mRNA as biomarkers for agr expression in the presence and absence of oxacillin subMICs in 10 MSSA and 4 MRSA clinical strains belonging to 5 clonal complexes (CC45-agrI, CC8-agrI, CC5-agrII, CC15-agrII and CC30-agrIII) causing endovascular complications. The DNA sequences of agr locus were obtained by whole genome sequencing. Results Our results revealed that exposure to subMICs of oxacillin had an impact on agr locus expression modifying the relative levels of expression with increases in 11 strains and with decreases in 3 strains. Thereby, the exposure to subMICs of oxacillin resulted in higher levels of expression of agr in CC15 and CC45 and lower levels in CC30. We also observed the presence of mutations in agrC and agrA in 13/14 strains with similar mutation profiles among strains within individual CCs except for strains of CC5. Although, agr expression levels differed among strains within CCs, the presence of these mutations was associated with differences in agr expression levels in most cases. Conclusions Changes in agr expression induced by exposure to oxacillin subMICs should be considered because they could lead to changes in the virulence modulation and have an adverse effect on the course of infection, especially in certain clonal complexes. Electronic supplementary material The online version of this article (10.1186/s12879-018-3088-7) contains supplementary material, which is available to authorized users.


Background
Staphylococcus aureus remains the major cause of endovascular infections with high morbidity and mortality rates that may reach 40% in case of infective endocarditis [1]. Even though most S. aureus genotypes exhibit the capacity to cause invasive disease, several studies have reported the association of particular clonal complexes (CC5, CC15 and CC30) with endovascular complications such as infective endocarditis [2][3][4].
The ability of S. aureus to produce infection involves a wide variety of virulence determinants, such as exotoxins, adhesins and immune evasion genes, constituting a gene framework that is under the control of global regulators, most notably the quorum-sensing accessory gene regulator (agr). Under conditions of high cell density, agr is responsible for the increased expression of many toxins and degradative exoenzymes, and decreased expression of several colonization factors. This regulation is important for the timing of virulence factor expression during infection [5]. Thereby, when adhesion to host tissue is crucial, cell density and agr expression are low, resulting in an increase in the surface virulence factors that are required for the colonization process. Once infection is established, the increased cell density results in higher agr expression that leads to production of toxins which trigger the host inflammatory response [6,7]. The agr system up-regulates the production of secreted virulence factors such as exotoxins, and downregulates the production of cell associated virulence factors [5,[8][9][10].
The agr locus consists of two divergent transcriptional loci, RNAII and RNAIII, driven by promoters P2 and P3, respectively [8,9]. The P2 operon encodes four genes, agrB, agrD, agrC and agrA. The agrA and agrC genes constitute a two-component signaling system, of which agrC is the receptor, and agrA is the response regulator [11,12]. Upon activation by agrC-dependent phosphorylation, agrA binds to the P2 promoter region for RNAII and the P3 promoter region for RNAIII, the effector of target gene regulation that also encodes delta-haemolysin, the expression of which serves as a surrogate for agr functionality [13,14]. Moreover agrA directly up-regulates transcription of the phenol-soluble modulins operon (psmα and psmβ) whose role in the pro-inflammatory response has already been demonstrated [6,7,12,15]. Several studies have reported that different mutations in the agr operon, mainly within the agrC or agrA genes, lead to decreased RNAIII expression and reduced virulence [9,[16][17][18]. However, it is known that strains with dysfunctional agr are more likely to cause persistent infection and be associated with poor outcome [19][20][21].
Despite the emergence of methicillin resistant S. aureus (MRSA), beta-lactams and specifically antistaphylococcal penicillins (nafcillin, oxacillin, cloxacillin and dicloxacillin), remain the usual treatment for the management of invasive infections caused by methicillinsusceptible S. aureus (MSSA). However, because of impaired diffusion and distribution of antibiotic, not all bacteria are exposed to the required lethal concentrations of bactericidal agents and therefore are likely subjected to sub-MIC antibiotic effects on the expression of virulence factors [7,[22][23][24][25][26][27][28][29][30]. Thereby, the effect of sub-MICs oxacillin on phenol-soluble modulins and RNAIII have been studied [27,28], showing a decrease in mRNA level in both virulence determinants in USA300 strain. However, increase in spa and lukE mRNA levels under exposure to subMICs of oxacillin, cephalotin and penicillin were reported by Subrt et al. [29]. Moreover, Rasigade et al. revealed that subMIC of oxacillin, moxifloxacin and linezolid led to increase fnbA/B mRNA levels [30]. Collectively all this studies add more evidence that, that suboptimal concentrations of different antibiotics modulate the expression of virulence factors in S. aureus, and therefore may exert an influence over the pathogenesis of infection [23]. Nevertheless, limited data exist showing the effect of oxacillin in MSSA clinical strains. Thereby, the aim of the present study was to investigate the impact of sub-inhibitory concentrations (subMICs) of oxacillin on the agr operon expression, as a key element of virulence regulatory network, in a collection of S. aureus clinical strains belonging to the main clonal complexes related with invasive diseases. In addition, we analyzed whether the presence of mutations in agr locus could be associated with differences in agr gene expression.

S. aureus strains collection
A total of 14 S. aureus strains (10 MSSA and 4 MRSA) from two multi-center investigations of infective endocarditis [31] and catheter-related bacteremia [32] were included in the present study. The strains were selected as representatives of the CCs that were most frequently detected in the original investigations: CC5 (ST5 and ST125), CC15, CC30, CC45 and CC8 (ST8 and ST72). We randomly selected two MSSA strains of each CC, as well as methicillin resistant (MRSA) strains that belonged to the same CCs if they were available (Table 1).

Sub-inhibitory concentrations of oxacillin and growth curves
Strains were exposed to subMICs of oxacillin in order to observe agr gene expression under this condition. To select the optimal subMIC for each strain, we monitored growth by measuring the optical density of broth cultures exposed to ½, ¼ and 1/8 of the respective MIC for each strain as previously determined by E-test (bioMérieux).
We defined the subMIC as the highest concentration below the MIC that showed no major growth defects over the entire growth curve relative to a control culture without oxacillin. [27] For each strain, the MIC and subMIC of oxacillin are shown in Table 1. Oxacillin was added to cultures at the time of inoculation (1:100 from precultures) into 10 ml of tryptic soy broth (TSB) and incubated at 37°C in a shaking incubator at 250 rpm. Growth under the same conditions was monitored by determining the optical density at 600 nm (OD600) and it was measured at the end of the incubation to determine the impact of antibiotic exposure over bacterial density (Additional file 1: Figure S1).

RNA isolation
Isolated colonies were resuspended in TSB and adjusted to 0.5 McFarland and incubated at 37°C overnight. Fresh TSB (10 ml in 50-ml conical flask) was inoculated with 100 μl of overnight broth cultures (1:100 dilution) and incubated at 37°C in a shaking incubator at 250 rpm in the presence or absence of the appropriate subMIC of oxacillin. Growth was monitored by determining the optical density (OD600). RNA was isolated after growth for 20 h (stationary phase) using the Qiagen Rneasy Mini kit according to the manufacturer's instructions and after addition of 2 volumes of RNA protect reagent (Qiagen, N.V).

Reverse transcription and qRT-PCR
Expression of both the agrA and RNAIII genes was analyzed by quantitative reverse transcriptase PCR (qRT-PCR) as surrogate biomarkers of global agr operon activation. Expression levels of the housekeeping gyrB gene were used to normalize differences in mRNA quantification. In brief, 1 μg of total RNA was transcribed into cDNA using a Transcriptor First Strand cDNA Synthesis Kit (Roche Applied Science). Quantitative real time PCR was carried out on a LightCycler® 480 with LightCycler® 480 FastStart DNA Master Plus SYBR Green I (both Roche Applied Science). Previously described primers were used to amplify the RNAIII, agrA and gyrB transcripts [28,33]. The experiments were performed using three biological replicates, each tested in triplicate. For each reaction, the ratio of the target gene (RNAIII or agrA) to gyrB transcripts was calculated based on the difference in cycle thresholds (2 Ct_gyrB-Ct_target_gene ) for all strains to obtain RNAIII and agrA relative gene expression levels in stationary phase in the presence and absence of the subMIC of oxacillin [33,34]. The results were expressed as the n-fold difference in levels of transcription of the 2 Ct_gyrBoxa-Ct_agrAoxa and 2 Ct_gyrBoxa-Ct_RNAIIIoxa relative to 2 Ct_gyrBcontrol_Ct_agrAcontrol and 2 Ct_ gyrBcontrol-Ct_RNAIIIcontrol , respectively.

Genotyping analysis of agr locus by whole genome sequencing
The DNA sequences of agr locus were obtained after that assembling of reads obtained from whole genome sequencing of all strains included in this study before of exposure to oxacillin subMICs. The sequencing was performed on the Illumina MiSeq platform (150 bp paired-end reads) using TruSeq DNA PCR free kits (Illumina, CA, USA). These agr nucleotide sequences of the entire agr locus were compared with those of agr group reference strains -agr I: NCTC 8325, agr II: N315; agr III: MW2-by multiple alignments using the Geneious server (Geneious 10.0.6).

Statistical analysis
In order to yield more accurate and reliable summary statistics continuous variables were expressed as mean and standard deviation (SD) from the expression data obtained of three biological replicates each tested in triplicate. These values were analyzed by Student's t or Wilcoxon test, as appropriate. Significance was defined as p < 0.05. Data were stored and analyzed using SPSS software version 15.0 (SPSS, Chicago, IL, USA).

Results
Expression of agr locus in the presence of oxacillin subMICs in stationary phase and its relationship with the CC, agr type and methicillin resistance The exposure to subMICs of oxacillin in stationary phase resulted in a trend to higher RNAIII transcript levels than in absence of the antimicrobial agent for 11/14 strains being statistically significant for 5 strains of them (p < 0.05) (Fig. 1). Only SA_123 (CC45) (p: 0. 735), SA_107 (CC30) (p: 0.028) and SA_170006 (CC5) (p: 0.043) exhibited reduced RNAIII expression ( Table 2). We observed a similar pattern of relative gene expression for agrA, with increased levels of agrA mRNA in the presence of oxacillin subMICs for all strains, being statistically significant for 5 of them (p < 0.05), except SA_80001 (CC30) (p: 0.028), SA_107 (CC30) (p: 0.345) and SA_103 (CC5) (p: 0.465) which showed lower levels of expression ( Fig. 1, Table 2, Additional file 2: Table S1). In addition, we observed that in four strains (SA_123, SA_103, SA_170006 and SA_80001), the change in the expression of RNAIII and agrA did not move in the same direction. However, these differences in the expression values for both agrA and RNAIII, in the presence and absence of oxacillin subMICs, were not statistically significant.
In order to investigate the association between agr locus expression and methicillin resistance profile we studied the agr locus expression in the strains classified as MSSA and MRSA and we found higher agrA and RNAIII levels of gene expression for both groups in the presence of subMICs of oxacillin than in absence of this antibiotic, although this trend was not statistically significant (Table 3). When we analyzed by CC, we detected an increase in relative gene expression for RNAIII and agrA in the presence of oxacillin for both CC45 (p = 0.005 and p = 0.000, respectively) and CC15 (p = 0.005 and p = 0.001, respectively). Nevertheless, CC30 showed a decrease in gene expression for both RNAIII and agrA in the presence of subMICs of oxacillin, but the difference was not statistically significant.
With regard to agr type, agrII strains showed higher expression levels under oxacillin exposure for both RNAIII and agrA (Table 3).

Genotypic analysis of agr locus
After observing differences in the agr expression among the strains, we intended to know whether these strains harbored any mutations that might potentially be responsible for the agr functionality. In general, sequencing of the agr locus revealed similar mutation profiles among strains belonging to the same CC (Table 3). Strains from CC45-agrI showed up to 9 nonsynonymous mutations in the agrB region and from 2 to 3 non-synonymous mutations in agrC, including missense mutations in SA_123 (417 Glu > Lys) and SA_   SA_170006. The two strains from CC15-agrII did not show any differences between them, with both exhibiting the same missense substitution at position 136 in agrA (136, Lys > Arg). No differences were found between the two CC30-agrIII strains, which showed the same genotypic profile with missense mutations in the coding region of agrB (49 Ile > Leu and 115 Ile > Leu) and agrC (55 Gly > Arg) ( Table 4).

Relationship between mutation profile and expression of agr system
After analyzing both agrA and agrC genetic profile, we aimed to investigate the potential effect of these mutations on the expression of agr in stationary phase. Overall there were strain-to-strain differences in the relative levels of expression (Fig. 2), which were largely independent of CC, with only strains in CC5 all exhibiting similar levels of expression. Notably, the strains in CC5 also had the overall lowest levels of RNAIII and agrA expression (p = 0.024 and p = 0.016, respectively).
Although it is difficult to determine the influence of mutations on the agr locus expression because of the variability found among strains, we studied whether the presence of the different mutations could have some impact on agr expression. In general, although some mutations detected did not seem to directly affect the expression of the agr system, the majority of them were associated with changes on the expression levels of RNAIII and/or agrA (Table 5). Thus, the mutation 42 Val > Gly found in the agrC in strains of CC45 and SA_ 190006 and SA_70002 of CC8 was associated with high expression levels for both RNAIII and agrA (specifically 5-and 6.9 fold, respectively). Likewise, mutations 262 Asp>Ala and 258 Phe > Ile detected in SA_190006 and SA_70002 (CC8) showed a 7.3-fold increase in RNAIII expression levels and a 5.8-fold increase in agrA expression levels (p < 0.05). In contrast, SA_180015 (CC8, ST8) and SA_80004 (CC5) both exhibited a deletion in agrC exhibited 93.5-and 148-fold decrease in RNAIII expression levels, and 9.2-and 15.1-fold decrease in agrA expression values for both strains, respectively (p < 0.001). The other CC5 strains also had low expression values for both RNAIII and agrA and, although we did not find any deletions, there were non-synonymous mutations in agrC in SA_103 (96Ser > Leu) and SA_170006 (121I > T) that were associated with 15.9-and 40.3-fold decrease in RNAIII expression and 4.3 and 6.7 in agrA expression, respectively (p < 0.001). However, SA_180009 of CC5 that also exhibited low expression did not have any mutations throughout the entire agr operon. We observed small and nonsignificant changes in the expression of RNAIII by the presence of mutation detected in agrA (136 Lys > Arg) in strains of CC15 (p = 0.478). Nevertheless, this mutation was linked to low expression values for agrA gene with a 2.3-fold decrease (p < 0.001). Finally, for CC30 strains, the mutation detected in agrC (55Gly > Arg) seems to be responsible of the low means values of expression obtained for both RNAIII and agrA (3.2-and 1.80-fold decrease, respectively) (p < 0.05).

Discussion
In the current study, we found that sub-lethal concentrations of oxacillin have an impact on the levels of agr gene expression in a selection of clinical MSSA and MRSA strains belonging to the main clonal complexes that were associated with invasive disease in Spanish hospitals, namely CC5, CC30, CC15, CC45 and CC8 [2][3][4]. These differences between the presence and absence of oxacillin subMICs resulted in increases or decreases in the levels of gene expression that were strain-dependent. In addition, we observed that the strains generally shared different mutations in agr locus according to the CC and that these mutations had an impact on the expression values of agr.
A recent review has shown that sub-lethal concentrations of different antibiotics such as beta-lactams are able to modulate the expression of virulence genes [23]. Our study shows a strain-dependent effect of oxacillin subMICs on the agr operon expression measured as transcriptional changes in the RNAIII and /or agrA mRNA in the clinical strains included in this study. We observed an increase relative to the control condition without oxacillin in 11/14 S. aureus clinical strains being statically significant for 5 of them. Additionally, strains with low expression levels under exposure to oxacillin subMICs were also detected. Because of the agr operon plays an important role in S. aureus virulence the exposure to subMICs of oxacillin could have significant consequences on the pathogenesis of infection. The initiation of transcription of RNAIII strictly relies upon the expression of agrA, in a RNAII-dependent (agrBDCA) manner [8,9,11]. Our data support these observations since the transcription of agrA was observed in all strains with RNAIII expression. However, in the presence of oxacillin subMICs, most strains showed the same direction of change in expression values for both genes, except for 4 strains that showed divergent expression values in RNAIII and agrA, although not statistically significant. This unexpected result could be due to other factors affecting on agrA expression. It is important to take into account the wide network of global (sarA, sarU, sarX, codY) and other virulence regulators (mgrA, sigB, saeSR) that have a direct or indirect effect on the agr [35,36], and specifically on the P2 and P3 promoters to which agrA binds [37], or on agrC-agrA interaction [38]. When we stratified the expression results by CC, we observed higher expression values of agr locus in presence of oxacillin subMICs than in absence of this antimicrobial agent for both CC15 and CC45. However, the relatively  Total number of single nucleotide polymorphism including synonymous and non-synonymous mutations small number of strains affords a substantial chance for statistical bias and does not allow performing proper multivariate analyses. On the other hand, methicillin resistance has been associated with a decrease in RNAIII expression under exposure to suboptimal concentrations of different antibiotics [28,34,39]. Our results do not seem to support this hypothesis, and only one MRSA strain (SA_170006) showed a reduction in the expression values relative to the control. However, taking into account that from the 4 MRSA (mecA positive) strains, only one is defined as an MRSA according to CLSI breakpoints (SA_170006) [40]. This subtle difference could be impacting on the correlation MSSA/MRSA with RNAIII/agrA expression. Indeed, previous studies [28,34] suggested that the expression of mecA and either its potential to subtly affect peptidoglycan structure or its interaction with other cell wall-associated proteins prevent the auto-inducing peptide (AIP) from being detected. These results in an unresponsive agr system that could be directly attributed to induced levels of mecA expression because of RNAIII expression was repressed as mecA expression increased. Although the role of agr system for pathogenesis in animal model is well established [41,42] and agr-defective mutants have shown an attenuated virulence and agr blocking agents exhibit anti-infective properties in   [43], several studies have reported that agr defective mutants are frequently recovered from patients with bacteremia, where the mutants are associated with persistent infection and poor outcome [9,[19][20][21]. According to previous reports we observed that 13/14 strains harbored both synonymous and non-synonymous mutations throughout the agr operon, especially in agrC and agrA [9,17,18]. Moreover, we have observed that while the genotypic analysis revealed similar mutations patterns accordingly to CC, the agr expression was clearly strain-dependent. Some of these mutations were associated to changes in agr gene expression in strains with and without a specific mutation. Deletions found in SA_180015 (CC8) and SA_ 80004 (CC5) in agrC were related with low expression levels and might be responsible for reduced gene expression in these strains. Both these deletions have previously been detected in strains with dysfunctional agr [9,17,18]. Strains of CC5 exhibited the most consistent agr gene expression although this group had the most divergent agr sequences within any CC, all mutations detected in this CC were associated with decreases expression values. Finally, our observations showed that CC30 exhibited low levels of agr gene expression with the presence of a previously identified mutation in agrC [18,44]. We recognize that the attempt to correlate RNAIII / agrA expression with mutations within the agr operon without complementation experiments is problematic, since in most cases the sample size for any given mutation was one, making it difficult to draw strong correlations between mutations and agr expression. Even so, these results, collectively, indicate that mutations in agr have an impact on agr expression and could be correlated with levels of agr gene expression for strains belonging to the same CC. Nevertheless, although different studies have shown that mutations in agr-defective strains do not persist in natural populations and that inactivating mutations have a relatively short life, other limitation of this study is that we do not have the results from sequencing of the agr operon after exposure to antibiotic, and therefore, we do not know the influence that this selective pressure could exert on mutations in the agr locus. Furthermore, it is possible that mutations detected in the agr system may have an effect not only on the expression on the agr locus itself, but also on other virulence genes regulated by agr. However, although these results could be confirmed by complementation experiments with the wild type gene, we cannot ignore the potential effect that other regulators may also have on agr operon expression independently of the mutations detected [24,29]. Our findings offer a valuable starting point for further studies that analyze gene expression changes induced by oxacillin subMICs from a whole transcriptomic approach would help to provide a more accurate overview of genes and regulators that compose the virulome framework in S. aureus and that may be affected under this condition.