Vancomycin susceptibility in methicillin-resistant Staphylococcus aureus is mediated by YycHI activation of the WalRK essential two-component regulatory system

The treatment of infections caused by methicillin-resistant Staphylococcus aureus is complicated by the emergence of strains with intermediate-level resistance to vancomycin (termed VISA). We have characterised a molecular pathway involved in the in vivo evolution of VISA mediated by the regulatory proteins YycH and YycI. In contrast to their function in other bacterial species, we report a positive role for these auxiliary proteins in regulation of the two-component regulator WalRK. Transcriptional profiling of yycH and yycI deletion mutants revealed downregulation of the ‘WalRK regulon’ including cell wall hydrolase genes atlA and sle1, with functional autolysis assays supporting these data by showing an impaired autolytic phenotype for each deletion strain. Using bacterial-two hybrid assays, we showed that YycH and YycI interact, and that YycHI also interacts with the sensor kinase WalK, forming a ternary protein complex. Mutation to YycH or YycI associated with clinical VISA strains had a deleterious impact on the YycHI/WalK complex, suggesting that the interaction is important for the regulation of WalRK. Taken together, we have described a novel antibiotic resistance strategy for the human pathogen S. aureus, whereby YycHI mutations are selected for in vivo leading to reduced WalRK activation, impaired cell wall turnover and ultimately reduced vancomycin efficacy.


Results
Genetic analysis of yycHI mutations in clinical VISA strains. We previously performed whole-genome sequencing of multiple, clinically-derived VISA isolates to characterise the genetic mechanisms of the in vivo evolution of reduced susceptibility to vancomycin in S. aureus 9 . We identified a novel deletion mutation within yycI, which belongs to the walRK/yycHIJ operon. In S. aureus, this operon consists of walR, which codes for a response regulator, walK, which codes for a sensor histidine kinase, yycH and yycI, which encode putative regulatory proteins and yycJ, which codes for a 5′ -3′ exonuclease involved in mismatch repair 20 (Fig. 1A). The yycI mutation was one of 13 found between the VSSA progenitor strain A6224 and VISA strain A6226 that emerged after prolonged vancomycin treatment 9 . The four-nucleotide deletion (ATAA from the position 79-82) resulted in a predicted frameshift from the 27 th amino acid and truncation of the protein from 262 amino acids in the VSSA strain (A6224) to 56 amino acids in the VISA strain (A6226) (Fig. 1B). No mutation was observed for the upstream protein, YycH (Fig. 1C). Intriguingly, in a separate study, Mwangi et al. identified 35 mutations between VISA strain JH9 (A8094 in this study) and the VSSA progenitor strain JH1 (A8090 in this study) including a single nucleotide polymorphism (SNP) in yycH 6 . The G to A substitution at the 81 st nucleotide base was predicted to change the tryptophan at position 27 to a premature stop codon, truncating the protein to ~5% of its original length (453 amino acids) (Fig. 1C). No YycI mutation was observed in the VISA strain (A8094) (Fig. 1B). DNA sequencing confirmed the yycHI mutations in A6226 and A8094. Given the degree of truncation, and the predicted role for the proteins in WalRK regulation, we hypothesized that yycHI mutation likely contributes to reduced vancomycin susceptibility in clinical VISA strains. YycH and YycI influence vancomycin susceptibility. To determine the exact contribution of yycHI to the development of VISA independent of other mutations, chromosomal deletions of yycH (A8090Δ yycH), yycI (A8090Δ yycI) and yycHI (A8090Δ yycHI) were generated within the genetic context of VSSA clinical isolate, A8090. Vancomycin susceptibility was assessed using Etest and population analysis profiling (PAP). A8090Δ yycH and A8090Δ yycI each showed a reduction in vancomycin susceptibility, with a MIC of vancomycin increasing from 1.0 μ g/mL to a maximum of 3.0 μ g/mL (range 2.0-3.0 µg/ml) for both mutants, which is in the non-susceptible range (> 2 μ g/mL) (Table 1), and has been associated with poor patient outcomes 21 . Notably, the MIC for vancomycin of the double mutant (A8090Δ yycHI) was no greater (Table 1) than for each of the single mutants, suggesting that both YycH and YycI proteins are equally essential for vancomycin susceptibility. PAP analyses are the gold standard for assessing VISA and heterogeneous VISA (hVISA), and provide a more sensitive quantification of resistant sub-populations. Each deletion strain had a greater number of cells tolerant to > 2 μ g/mL of vancomycin (Fig. 2), which defines the VISA phenotype and predicts for persistent bacteremia and complicated infection outcomes 3,22 .
Genetic reconstitution of yycH and yycI in situ returned the MIC for vancomycin to within the susceptible range (1.0 μ g/mL, Table 1) and each reconstituted strain produced a similar vancomycin population profile to that of the vancomycin-susceptible parent strain (Fig. 2). Together, these data confirm that both YycH and YycI are important for vancomycin susceptibility in S. aureus.
Deletion of yycH or yycI from VSSA strain A8090 did not recreate the MIC of vancomycin for the clinical VISA isolate A8094 (8.0 μ g/mL). This suggests that additional mutations in A8094 contribute to the VISA phenotype for this strain and supports the theory that VISA emerge as a result of cumulative mutations during prolonged vancomycin exposure. In addition, the in vivo evolution of VISA in A8094 and A6226 was associated with co-resistance to daptomycin in the absence of daptomycin exposure (Table 1). To determine if yycHI mutation contributed to the daptomycin-nonsusceptible phenotype of these isolates, daptomycin susceptibility testing was performed including PAP analyses. No change in susceptibility was observed for A8090Δ yycH or A8090Δ yycI when compared to A8090 (Table 1, PAP data not shown). It is likely that mutations present in A8094 and A6226 other than those in yycHI contributed to daptomycin resistance in these isolates.

YycHI positively regulates WalRK in S. aureus.
To determine how YycHI contributes to antibiotic susceptibility and to understand the impact of these proteins upon WalRK, we compared the transcriptomes of A8090, Δ yycH and Δ yycI using RNA-Seq. Transcriptomes were sequenced using the Illumina sequencing platform yielding an average total read count of 6.67 million reads per sample. The transcriptomes of Δ yycH and Δ yycI were highly similar (only six differentially expressed genes representing 0.2% of the transcriptome [Supplementary Table S1]), suggesting that both proteins function within the same system and are equally important in S. aureus physiology (Fig. 3). When the transcriptome of Δ yycH or Δ yycI was compared to the progenitor strain A8090, a total of 42 genes showed altered expression (≥ 2-fold) in at least one of the deletion strains. These included 17 downregulated genes (Supplementary Table S2 MICs for vancomycin were determined 4 times independently for A8090 and its mutant derivatives (A8090Δ yycH, A8090Δ yycI, A8090Δ yycHI, A8090::yycH and A8090::yycI). b Previously referred to as JH1 6 . c Previously referred to as JH9 6 .
genomes (KEGG) pathways, 40% of the genes were predicted to be involved in cell wall, membrane and envelope biosynthesis (Fig. 3A). Furthermore, 21% of genes were predicted to be involved in staphylococcal infection and defence (Fig. 3A). Quantitative digital PCR of select genes confirmed these RNA-Seq data ( Supplementary Fig. S1).
WalR is a critical transcriptional regulator in S. aureus and has been shown to influence the expression of genes involved in cell wall turnover, which explains its role in the development of VISA, as well as a diverse range of genes important for amino acid biosynthesis, central metabolism and virulence [14][15][16] . However, experimental confirmation of WalR binding to the promoter sequence of specific genes and induction of their expression has only been confirmed for five genes, each of which encode for proteins with autolytic activity (sle1, atlA, ssaA, isaA and lytM) 14,15 . Based on work in Bacillus, we hypothesised that YycHI binds to WalK and alters its regulatory activity 18,19 . To determine the direction of YycHI-mediated regulation of WalRK in S. aureus, we interrogated our RNA-seq data for the expression of validated WalR regulated genes. We showed that except for lytM, the remaining four genes (sle1, atlA, ssaA and isaA) were downregulated in both the Δ yycH and Δ yycI deletion mutants when compared to their progenitor strain ( Fig. 3B and Supplementary Table S2). Furthermore, whilst upstream WalR binding is yet to be confirmed, walRK induction increases the expression of four CHAP-domain containing amidase/peptidoglycan hydrolase genes (SaurJH1_0704, SaurJH1_0796, SaurJH1_2370 and SaurJH1_2642) 14 . Of these, SaurJH1_0704 and SaurJH1_2370 were significantly downregulated in Δ yycH and Δ yycI ( Fig. 3B and Supplementary Table S2). We also showed, similar to that observed in Bacillus 19 , that YycH has a growth-phase dependent impact on gene expression that is most pronounced at early stationary compared to mid-exponential phase ( Supplementary Fig. S2A). These transcriptomic changes suggest that in contrast to that seen in Bacillus sp., YycHI positively regulates the WalRK TCRS in S. aureus. Importantly, the transcription of walRK was not different for either Δ yycH or Δ yycI when compared to the progenitor strain A8090.
To provide further support to the direction of WalRK regulation by YycHI in S. aureus, we performed functional autolysis assays. It is well established that WalRK is a regulator of autolysis in S. aureus 14 and this is supported by the five validated WalR-regulated genes that all encode proteins with autolytic activity 14,15 . Using Triton X-100 induction, we compared the autolytic rate of Δ yycH and Δ yycI to their progenitor, A8090. Each deletion strain showed impaired autolysis as determined by a reduced decline in OD 600 over time (Fig. 4). Gene reconstitution restored the autolytic rate back to that of the vancomycin-susceptible parent strain (A8090) (Fig. 4). Together, the observed differential gene expression and the functional impact on autolysis suggests that YycHI acts as a positive regulator of WalRK in S. aureus. Deletion of either YycH or YycI leads to downregulation of WalR-regulated genes and WalR-mediated autolysis, which likely contributes to reduced susceptibility to the cell-wall acting antibiotic, vancomycin.
Transcriptional effects of yycHI deletion upon genes associated with host-pathogen interaction. In addition to regulating cell wall turnover, recent reports have shown that WalRK indirectly controls the expression of a number of virulence factors via activation of the SaeSR TCRS 15 . Fittingly, our transcriptomic data highlighted down-regulation of multiple genes that are involved in host-pathogen interaction, a number of which are indirectly regulated by WalR (Fig. 3C) 15 . These included immunoglobulin-binding proteins encoded by spa and sbi, which were downregulated in both Δ yycH and Δ yycI. In addition, fibrinogen-binding proteins encoded by efb and SaurJH1_1235 were less abundant in Δ yycH and Δ yycI, and excreted factors scn and SaurJH1_1240 encoding staphylococcal complement inhibitors were significantly downregulated in Δ yycI (Supplementary  Table S2). In contrast, the entire capsule operon was overexpressed in both Δ yycH and Δ yycI (Fig. 3C), which is commonly reported for VISA strains 12 . YycH, YycI and WalK form a ternary protein complex. Thus far little is known about the staphylococcal YycH and YycI proteins and their interactions. Work in B. subtilis has showed that YycH and YycI are tethered to the cytoplasmic membrane and physically interact at their respective transmembrane domains (TMDs) 18 . Using a bacterial-two hybrid system, we assessed the interactions between YycH and YycI from S. aureus. Full-length YycH was found to interact with full-length YycI using this approach (Fig. 5A). We observed that except for the TMDs of each protein (YycI 1-53 and YycH 1-36 ), the remainder of the protein was dispensable for their interaction (Fig. 5A). YycH/YycI constructs lacking TMDs did not interact with their respective full-length binding partners (Fig. 5A). These data suggest that staphylococcal YycHI interactions are mediated by the TMDs. Remaining genes are predicted to be under the control of WalRK due to the presence of an upstream putative WalR binding site 14,46 . (C) Genes involved in host-pathogen interaction were differential expressed in Δ yycH and Δ yycI. Genes encoding surface proteins, adhesins and complement inhibitors were mostly downregulated, whilst the capsule operon was strongly overexpressed in Δ yycH and Δ yycI. To determine whether YycH/YycI interact with the sensor kinase WalK, we first tested them individually. Similar to our observations for vancomycin susceptibility, YycH or YycI alone were not sufficient to produce a quantifiable interaction with WalK (Fig. 5B), suggesting that YycH and YycI are required as a protein complex before they can adequately interact with WalK. To test this hypothesis, full-length YycH, YycI and WalK were co-expressed in Escherichia coli BTH101; a quantifiable interaction was observed (Fig. 5B).
yycHI mutations associated with clinical VISA strains disrupt the interaction between YycH, YycI and WalK. In addition to yycHI mutations described in this report, numerous diverse walK point mutations have been described in VISA strains 16,17 . To date, it is unclear how these mutations impact on the activity of WalRK or the relationship with YycHI. As such, we next investigated the impact of VISA associated walK/yycHI mutations upon the interaction between YycH, YycI and WalK using our bacterial two-hybrid system. WalK proteins containing amino acid substitutions associated with diverse VISA isolates were assessed including G233D from JKD6008, Q371Δ from A8392, and R263C/S273N from A8118 (Table 1). These WalK mutations had no effect on the interaction with YycHI (data not shown). We next assessed the impact of yycHI mutations from the clinical VISA strains A8094 and A6226 (Fig. 1). The interaction between YycH and YycI was reduced 6-fold when the mutated YycH from A8094 was assessed, and the interaction was abolished when the mutated YycI from A6226 was included in the analysis (Fig. 5C). Most importantly, the interaction with the kinase, WalK was significantly reduced when mutated YycH or YycI proteins from either A8094 or A6226 were included (Fig. 5D). Together, these data suggest that reduced vancomycin susceptibility in our clinical VISA strains is due to impaired interaction of the YycHI protein complex with WalK, leading to reduced expression of the WalR regulon and reduced WalR-mediated autolysis. A proposed model for this is shown in Fig. 6.

Discussion
Defining the complete molecular pathway to reduced vancomycin susceptibility in S. aureus has remained challenging owing to the heterogeneous nature of the genetic mutations associated with VISA. We, and others have used whole-genome sequencing of clinical isolates to guide targeted genetic manipulation experiments to determine the exact contribution of select genes to reduced susceptibility to vancomycin, independent of other mutations 9,16 . By engineering yycH and yycI deletions from a vancomycin-susceptible strain, we have described for the first time the contribution of these genes to the development of reduced vancomycin susceptibility and we confirmed their importance by restoring the vancomycin-susceptible phenotype via gene reconstitution (Fig. 2).
The function of YycH and YycI in S. aureus is typically inferred from their role in Bacillus sp. where they negatively regulate the WalRK ortholog YycFG 18,19 . WalRK/YycFG is highly conserved across low GC content gram-positive bacteria, and is essential in S. aureus most likely due to its crucial role in regulating cell wall metabolism 23,24 . Due to its role in cell wall maintenance, it is not surprising that WalRK mutation is commonly described for VISA 16,17 . Two distinct mechanisms have been identified that infer a role for WalRK in the evolution of VISA. Firstly, IS256 insertion in the promoter of the operon has been shown to reduce vancomycin susceptibility and secondly, point mutations within walK and walR contribute to vancomycin tolerance in clinical VISA strains 16,25,26 . What remains unclear, however, is the impact of these mutations upon the activity of WalRK. Jansen et al. showed that IS256 insertion led to increased expression of walRK whereas more recently, McEvoy et al. showed that upstream IS256 insertion decreased expression of the TCRS 25,26 . Similarly, transcriptional profiling of VISA-associated walK and walR point mutants revealed no clear directionality for the activity of the TCRS based on contrasting expression of isaA and atlA, each of which are induced by WalR 16 . In this study, we have described a third mechanism of WalRK-mediated VISA development, via mutation to auxiliary proteins YycHI. Along with reduced vancomycin susceptibility, deletion of yycH or yycI resulted in reduced autolysis and downregulation of genes under the transcriptional control of WalRK, including atlA, sle1 and CHAP domain amidase/peptidoglycan hydrolase genes SaurJH1_0704 and SaurJH1_2370 (Fig. 3B). Taken together these data suggest that in contrast to Bacillus sp., YycHI serves as an activator of WalRK in S. aureus and that the VISA phenotype associated with yycHI mutation correlates to reduced activity of the system, as opposed to enhanced activity, which was previously proposed 6,9,27 . The regulatory impact of YycHI in B. subtilis was found to be as a result of direct interaction with the kinase, YycG 18,19 . A report investigating YycH and YycI from S. aureus failed to provide evidence of an interaction with WalK using a detergent-micelle-model 28 . Here, we have used a bacterial two-hybrid system to show that YycH or YycI in isolation was not sufficient to produce an interaction with WalK but when all three proteins were present, a ternary protein complex formed (Fig. 5B). This correlates with what has been described for B. subtilis using immunoprecipitation assays, whereby YycH only co-precipitated with YycG when YycI was present and vice versa 29 . YycH and YycI appear indispensible for the regulation of WalK, which explains why minimal transcriptional or phenotypic differences were observed between the yycH, yycI and yycHI deletion strains in this study. In prokaryotes, a number of TCRSs are regulated by interaction with auxiliary proteins. For example, the PII protein from E. coli acts as a repressor of NtrBC by blocking the autokinase domain and enhancing the phosphatase activity of the sensor kinase NtrB 30 . In addition, SaePQ serves as a repressor of the staphylococcal virulence regulator SaeSR by activating the phosphatase activity of the sensor kinase SaeS 31 . Whilst we have shown a physical interaction between YycHI and WalK, further experiments are required to understand exactly how YycHI serves to regulate WalRK in S. aureus.
The TMDs of YycH and YycI appeared to be important for protein-protein interactions (Fig. 5A). The yycH mutation associated with the clinical VISA strain, A8094 resulted in TMD truncation, which impacted on the interaction with YycI, as well as complex formation with WalK (Figs 1 and 5). Similarly, the yycI mutation associated with the clinical VISA strain, A6226 resulted in a frameshift within the TMD that led to similar effects upon interaction with YycH and WalK (Figs 1 and 5). Thus, based on current evidence, yycHI VISA mutations impact on WalRK activation as a direct result of reduced interaction with the sensor kinase WalK. When compared to infections caused by VSSA, VISA are more likely to be associated with prolonged bacteremia and less likely to cause acute clinical instability such as septic shock 32,33 . In the current study, whilst we did not observe altered expression of major staphylococcal toxins, we did observe differential transcription of genes that impact on interaction with the host (Fig. 3C). Surface protein A (coded for by spa) is a trigger for platelet aggregation, which stimulates the production of platelet microbiocidal proteins 34 . Deceased expression of spa, as seen for our yycH/I deletion strains, likely results in reduced stimulation of platelet microbiocidal proteins that are important for host-defense against S. aureus bloodstream infection. As a corollary, increased expression of the capsule operon contributes to resistance to opsinophagocytosis by human polymorphonuclear leukocytes and enhanced bacterial persistence in a murine model of S. aureus infection 35 .
In keeping with the notion of YycHI as an activator of WalRK, the expression of Ig-binding B domain containing protein (sbi), complement inhibitors (scn, SaurJH1_1240) and fibrinogen binding proteins (efb, SauJH1_1235) were downregulated in this study, each of which have been shown to be positively regulated upon WalR overexpression 15 . Constitutive activation of WalR has also been shown to promote bacterial clearance and increase neutrophil recruitment in vivo 15 . Taken together, it is plausible that mutations to yycHI and reduced WalRK activation may contribute to a 'stealth' strategy coined by Gardete et al. whereby VISA alter the expression of virulence determinants to facilitate immune evasion and persistence within the context of infection 27 ; a hypothesis that requires further exploration.
In summary, our data highlight a novel role for YycH and YycI to the development of reduced vancomycin susceptibility in S. aureus. YycHI mutations associated with VISA directly impacted upon interaction with the sensor kinase WalK and yycH/yycI deletion resulted in reduced expression of genes under its control. This indicates a positive role for YycHI proteins in regulating the WalRK TCRS in S. aureus.

Methods
Bacterial strains, plasmids and culture conditions. S. aureus strains were grown in Heart Infusion (HI) broth or agar and are listed in Table 1. E. coli strains were grown in LB broth or agar and are listed in Table  S4. For the selection of strains carrying plasmids (listed in Table S4), media was supplemented with ampicillin (100 μ g/mL), carbenicillin (100 μ g/mL), kanamycin (50 μ g/mL) or chloramphenicol (25 μ g/mL).
Genetic manipulation. The E. coli/S. aureus shuttle vector pKOR1 was used to generate in-frame deletions of yycH (A8090Δ yycH), yycI (A8090Δ yycI) and yycHI (A8090Δ yycHI) as described previously 36 . Approximately one kilobase up-and down-stream of each gene was amplified using primers listed in Table S4. Products were cloned into pKOR1 using BP Clonase (Invitrogen), then used to transform E. coli DH5α (NEB). Each plasmid was passaged through S. aureus strain RN4220 before being electroporated into A8090 37 . Integration of the vector was stimulated by growing cells in the presence of chloramphenicol at 42 °C followed by vector excision at 30 °C (no antibiotic). Counter-selection was achieved using 1 μ g/mL anhydrotetracycline (ATc).

Antibiotic susceptibility testing and autolysis assays. Minimum inhibitory concentrations (MICs)
were determined using Etest (bioMérieux) as per manufacturer's specifications. To assess for antibiotic heteroresistance, population analysis profiling was performed using the method of Wootton et al. 9,40 .
Autolytic activity was assessed using Triton X-100 induction 41 . S. aureus cells were grown to an optical density at 600 nm (OD 600 ) of 0.8, chilled then pelleted by centrifugation. Cells were washed once with PBS then resuspended in PBS (pH 7.4) supplemented with 0.05% Triton X-100 to an OD 600 of 1.0. Autolysis was defined as the decline in OD 600 over time.
RNA extraction, sequencing and digital PCR. For RNA sequencing, total RNA was extracted from early stationary phase (OD 600 7.0, Fig. S2B), as the impact of YycHI upon WalR-dependent gene expression in Bacillus sp. was most pronounced at this growth stage 19 . RNA was stabilized in situ using RNAlater (QIAGEN) and samples were frozen at − 80 °C. S. aureus cells were thawed at ambient temperature then disrupted using a Precellys 24 tissue homogenizer (Bertin Technologies). RNA was purified using an RNeasy mini kit (QIAGEN), with on-column DNAse I (QIAGEN) digestion. RNA from two independent extractions was forwarded to the Beijing Genomics Institute for sequencing using Illumina HiSeq 2000. Reads were mapped to the A8090 genome (GenBank accession NC_009632) and gene expression was quantified using the Reads Per Kilobase Per Million (RPKM) method 42 . Differential gene expression between strains was determined using NOISeq 43 . A fold change ratio of > 2.0 with a q-value ≥ 0.8 was considered significant as described elsewhere 43 . To validate the differential gene expression data generated by RNA sequencing, digital PCR was performed for sle1, SaurJH1_2073, capL, spa and efb (primers listed in Table S5). cDNA (500 ng) was synthesized using Superscript III reverse transcriptase (Invitrogen). Digital PCR was performed as described by the manufacturer (Bio-Rad) for three biological replicates and gene expression was normalized against gyrB expression. Bacterial two-hybrid (BTH) analysis. Bacterial two-hybrid vectors were constructed and assays performed as described previously (Table S4) 44 . Coding sequences of proteins of interest were fused to either T25 (pKT25) or T18 (pUT18C) of adenylate cyclase. Each construct was verified by DNA sequencing. To determine if protein fragments interacted, combinations of pKT25 and pUT18C based plasmids were co-transformed into E. coli strain BTH101, which is adenylate cyclase deficient. Interaction was observed via functional complementation of adenylate cyclase and expression of the lac reporter 44 . The relative strength of interaction was quantified by measuring β -galactosidase activity using o-nitrophenol-β -galactoside (ONPG) as a substrate, as described elsewhere 45 . β -galactosidase assays were performed in triplicate for 4 biological replicates. Significance was determined by Mann-Whitney U test with a significance level of P < 0.05.
Published during review process. While this paper was being reviewed, a manuscript was published by Poupel O et al. (Poupel et al., 2016 doi: 10.1371/journal.pone.0151449) that supported the key findings from this study. Deletion of yycHI from the S. aureus HG001 genetic background (vancomycin MIC 3 µg/ml) resulted in a decrease in vancomycin susceptibility (MIC 4 µg/ml) which was restored upon complementation. The authors concluded that the YycHI complex does not act as a repressor of WalRK in S. aureus by analyzing autolysis and the expression of atlA, sle and saouhsc00773 in yycH and yycI deletion mutants during the exponential phase of growth (OD 600 nm 1.0). By assessing global gene expression at a later point in the growth cycle (OD 600 nm 7.0) we have now shown that in contrast to Bacillus, YycHI induces the expression of genes under the transcriptional control of WalRK in S. aureus.