Identification and characterization of mutations responsible for the β-lactam resistance in oxacillin-susceptible mecA-positive Staphylococcus aureus

Staphylococcus aureus strains that are susceptible to the β-lactam antibiotic oxacillin despite carrying mecA (OS-MRSA) cause serious clinical problems globally because of their ability to easily acquire β-lactam resistance. Understanding the genetic mechanism(s) of acquisition of the resistance is therefore crucial for infection control management. For this purpose, a whole-genome sequencing-based analysis was performed using 43 clinical OS-MRSA strains and 100 mutants with reduced susceptibility to oxacillin (MICs 1.0–256 µg/mL) generated from 26 representative OS-MRSA strains. Genome comparison between the mutants and their respective parent strains identified a total of 141 mutations in 46 genes and 8 intergenic regions. Among them, the mutations are frequently found in genes related to RNA polymerase (rpoBC), purine biosynthesis (guaA, prs, hprT), (p)ppGpp synthesis (relSau), glycolysis (pykA, fbaA, fruB), protein quality control (clpXP, ftsH), and tRNA synthase (lysS, gltX), whereas no mutations existed in mec and bla operons. Whole-genome transcriptional profile of the resistant mutants demonstrated that expression of genes associated with purine biosynthesis, protein quality control, and tRNA synthesis were significantly inhibited similar to the massive transcription downregulation seen in S. aureus during the stringent response, while the levels of mecA expression and PBP2a production were varied. We conclude that a combination effect of mecA upregulation and stringent-like response may play an important role in acquisition of β-lactam resistance in OS-MRSA.


Genomic analysis of the clinical OS-MRSA isolates.
To determine the genetic background of the strains used in this study, the whole-genome sequences of the 43 clinical OS-MRSA isolates were determined, and their phylogenetic relationships were analyzed by constructing a phylogenetic tree using kSNP3 (Fig. 1). The phylogenetic tree revealed extensive genomic diversity among the isolates, which could be classified into seven main phylogenetic clades. In addition, these isolates could also be grouped into 11 MLST types (ST1, ST5, ST8, ST59, ST89, ST91, ST121, ST338, ST772, ST1516, and ST6217), and they carried four different SCCmec types (II, IVa, IVc, and V). The majority of OS-MRSA isolates were belonged to ST121-SCCmec type V (16 strains, 37%), followed by ST59-SCCmec type V (seven strains, 16%), ST89-SCCmec type V (six strains, 14%), ST89-SCCmec type IVa (three strains, 7.0%), ST8-SCCmec type IVa (two strains, 4.7%), and ST6217-SCCmec type V (two strains, 4.7%). In addition, seven singletons (ST1-SCCmec type IVa, ST5-SCCmec type II, ST59-SCCmec type IVa, ST91-SCCmec type IVa, ST338-SCCmec type V, ST772-SCCmec type V, and ST1516-SCCmec type IVc), each of which comprised 2.3% of all strains, were identified. SCCmec types V (33 strains, 77%) and IVa (eight strains, 19%) were predominant among the OS-MRSA isolates, whereas only one isolate harbored each of SCCmec type II and IVc, respectively. Single nucleotide polymorphisms (SNPs) found in promoter and coding region of mecA are listed in Table 2. The type of SNPs located on mecA promoter region, consisting of MecI/BlaI-binding site (− 19 to − 50) and ribosome-binding site (− 7 to − 11) 24 , and coding region were closely correlated with SCCmec types. For the SNPs in promotor region, SCCmec type II strain JMUB1293 carried an C-30A mutation (A replace C of 30th bases upstream of mecA CDS), all 8 strains belong to SCCmec type IVa had G-7 T mutations, and all strains belong to SCCmec type IVc (1 strain) and V strains (34 strains) carried C-33 T mutations. For the SNPs in mecA coding region, all OS-MRSA strains had a synonymous mutation of C75A, and all SCCmec type V strains carried nonsynonymous mutations of T675A (Ser225Arg), when compared to a prototypic pre-MRSA strain N315, which carries intact mec operon (composed of mecA, mecI, and mecR1) and its mecA gene expression is strongly repressed by mecI 25 .
Moreover, whole-genome sequencing demonstrated that 34 of the 43 (79%) OS-MRSA isolates carried a complete bla operon (Table 1), which could be classified into two genotypes, namely bla operon-1 and bla operon-2, based on the nucleotide sequences. These two operons shared nucleotide identities of 94% for blaZ, 92% for blaR1, and 94% for blaI. Twelve strains (28%) carried bla operon-1, and all but one (JMUB217) bla operon-1 was located on plasmids. Meanwhile, 23 isolates (53%) carried intact bla operon-2 in their chromosomes. JMUB217 carried both bla operons on its chromosome. An incomplete bla operon-2 lacking blaZ but having intact blaR1 and blaI was present in isolates JMUB1301 and JMUB1313. The absence of blaZ in the bla operons of these isolates was confirmed by PCR (data not shown). Lastly, seven isolates (16%) lacked a bla operon.
Influence of bla operon on reduced susceptibility to oxacillin in OS-MRSA. A previous study suggested that blaI expression levels were associated with reduced oxacillin susceptibility in OS-MRSA isolates 26 .
To understand how the bla operons affect oxacillin susceptibility in the tested OS-MRSA strains, mutants with knockout of β-lactamase repressor gene blaI were generated and their effect on the oxacillin susceptibility was analyzed. Our whole-genome sequencing analysis showed that bla operon were carried by 36 of the 43 (84%) OS-MRSA isolates, and the bla operons could be classified into two genotypes, bla-1 and bla-2. The OS-MRSA JMUB217 (ST772, V, mecI − , blaI-1 + , blaI-2 + , OXA = 0.75 µg/ml) carried both blaI-1 and blaI-2, therefore, we deleted either one or both of their blaI to generate single and double blaI-knockout mutants, and their MICs of penicillin G and oxacillin were determined (Fig. 3A). Knockout of blaI-1 or blaI-2 alone did not significantly affect MICs of the penicillin G and oxacillin, whereas the double knockout could raise MIC of penicillin G significantly from 1.5 to 8 µg/mL, but of oxacillin slightly from 0.75 to 2 µg/mL. Similar to the results of MIC determination, knockout of blaI-1 or blaI-2 alone did not affect the levels of mecA expressions and PBP2a production,  5   JMUB1308  ND  ND  ND  -+  +  +  Chromosome   JMUB1311  ND  ND  ND  -+  +  +  Chromosome   JMUB1291  ND  ND  ND  -+  +  +  Chromosome   JMUB1305  ND  ND  ND  -+  +  +  Chromosome   JMUB1285  ND  ND  ND  -+  +  +  Chromosome   JMUB1289  ND  ND  ND  -+  +  +  Chromosome   JMUB1284  ND  ND  ND  -  www.nature.com/scientificreports/ www.nature.com/scientificreports/ whereas the double knockout enhanced mecA expression and PBP2a production but to a lesser extent (Fig. 3B). These results indicated that the influence of blaI on oxacillin susceptibility is limited.

Contribution of increased mecA expression to reduced oxacillin susceptibility in OS-MRSA.
To understand the role of the identified mutations in reduced oxacillin susceptibility, OS-MRSA strain JMUB217 and its oxacillin-resistant mutants were used as representative strains for further study because 24 mutants carrying 26 variants in 11 genes and an intergenic region were derived from the JMUB217. In addition, oxacillin MICs for the 24 mutants ranged widely (1.5-256 µg/mL), and were highly elevated compared to their parent strain JMUB217 (0.75 µg/mL). A sequential experiment was carried out. First, a mecA-overexpressing mutant was created to investigate whether changes in mecA expression affects oxacillin susceptibility in OS-MRSA. A vector pKAT containing mecA and its native promoter was introduced into JMUB217 to generate the mecAoverexpressing mutant JMUB217 (pmecA), and MIC determination found that the generated mutant exhibited increment of oxacillin MIC from 0.75 to 12 µg/mL ( Fig. 3; Table 3). Next, mecA-knockout mutant JMUB217 (∆mecA) was generated and its oxacillin MIC was measured. The oxacillin MIC decreased slightly from 0.75 to 0.38 µg/mL in this mecA-deleted mutant ( Table 3), indicating that the presence of mecA itself confers a low level oxacillin resistance. Moreover, overexpression of mecA in the mecA-deleted mutant JMUB217 (∆mecA) resulted in an increment of the oxacillin MIC to 12 µg/mL, similar to that of the mecA-overexpressing mutant JMUB217 (pmecA). Finally, a set of mecA-knockout strains from three oxacillin-resistant mutants (JMUB217-11, JMUB217-22, and JMUB217-24), carrying mutations of RpoC P358L , RpoB G645H , and RpoC G498D , respectively, were generated and their oxacillin MICs were determined. The results found that their MICs of 4, 32, and 256 µg/ mL decreased to 0.38 µg/mL, similar to the level of the mecA-knockout mutant JMUB217 (∆mecA) ( Table 3). These results indicated that mecA expression is a key factor for promoting reduced oxacillin susceptibility in OS-MRSA. www.nature.com/scientificreports/ Correlation of the levels of mecA expression and PBP2a production with oxacillin MIC in mutants with reduced oxacillin susceptibility. To examine the correlation between PBP2a production and oxacillin susceptibility in laboratory mutants, PBP2a agglutination assay was performed on 11 JMUB217derived mutants (Fig. 5A). Despite the increase of oxacillin MIC, PBP2a production was not significantly changed in these mutants. Next, mecA expression of JMUB217-derived mutants in the presence and absence of oxacillin was measured (Fig. 5B). The results showed that the natural expression of mecA significantly increased in 9 of 11 mutants with 1.3 to 1.9-fold change. However, the mecA expression levels were still lower than those of OR-MRSA (Fig. 2B). Since bla operons of JMUB217 induced mecA expression (Fig. 3), the mecA expression levels were measured in the presence of low concentration of oxacillin (0.1 µg/mL). Results showed that oxacillin induction significantly upregulated the mecA expression level in wild-type (2.8-fold) as well as the resistant mutants (1.3 to 2.9-fold). When compared with wild-type strain, the mecA expression levels induced by oxacillin were significantly increased in three of 11 mutants (JMUB217-11, -23, -24). Interestingly, the resistant mutant with the highest oxacillin MIC did not display the strongest mecA expression in both presence and absence of oxacillin. As seen in Fig (Table S3). In concordance with qRT-PCR data (Fig. 5B), the results of transcriptome analysis showed that mecA expression was significantly induced by oxacillin in both the mutants and parent strain, while the differences in the basal mecA expression between the mutants and its wild-type was small (log 2 -fold change < 1) as shown in Table S3.
A Venn diagram analysis showed that two rpoC mutants JMUB217-11 and JMUB217-18 shared 217 DEGs, of which 64 or 153 genes were commonly upregulated or downregulated, respectively (Fig. 6A). Two rpoB mutants JMUB217-20 and JMUB217-22 shared 297 DEGs with 168 up-and 129 down-regulated genes (Fig. 6B). In case of all five strains carrying mutation of either rpiA or rpoC or rpoB, 13 genes were up-and 15 genes down-regulated commonly ( Fig. 6C; Table S4). Among the commonly regulated genes, upregulation of tryptophan biosynthesis genes (trpBDEFG) and downregulation of nucleotide transporter and biosynthesis genes (pyrRP, hisIG). These genes were known to be related with Rel Sau /RSH-dependent stringent response mediated by amino-acid deprivation in S. aureus 27 . Rel Sau is a bifunctional (p)ppGpp synthase/hydrolase and induces the classic stringent response by accumulation of (p)ppGpp 28 . In addition, downregulation of purine biosynthesis genes such as xprT, purF and guaB, which were usually seen in the stringent response were observed in the four rpoC and rpoB mutants (Table S4; Fig. 7I).
In addition to the alteration of expression of genes directly related to the stringent response, the mutants with reduced oxacillin susceptibility also exhibited downregulation of genes involved in protein quality control and tRNA synthesis. Notably, clpP, clpX, and ftsH were significantly downregulated in mutants with reduced oxacillin susceptibility (Fig. 7B). Moreover, 16 out of 25 tRNA genes were also downregulated in at least one of the five mutants (Fig. 7C). These changes in gene expression might contribute to oxacillin resistance, as previous studies described that deficiencies of protein quality control and tRNA synthesis were associated with the stringent response and β-lactam resistance [29][30][31] .
Transcriptome profiles of the rpoBC and rpiA mutants also showed alteration in gene expression relevant to the peptidoglycan biosynthesis, for example, upregulation of mecA and sgtB, and downregulation of murBJY, femABX, pbp4, ftsW, and rodA (Fig. 7L). Furthermore, changes in the expression of genes involved in autolysis of S. aureus were observed in strains carrying mutations of rpoBC and rpiA, for example, lytM and sle1 were upregulated, whereas lytH, isaA, and ssaA were downregulated (Fig. 7D). All these changes in combination with mecA expression and stringent-like response might direct bacterial metabolism towards acquisition of oxacillin resistance in the resistant mutants.
The mutants with reduced oxacillin susceptibility tend to slow growth. Mutations in genes involved in the stringent response were reported to be associated with slower growth rate 32 . In addition, some β-lactam-resistant mutants generated in vitro were also found to have slow growth rates or a phenotype of persistent infection [31][32][33][34] . To investigate whether the mutations identified in the oxacillin-resistant mutants affect cell growth, the doubling time of JMUB217-derived mutants was measured. Seven of eleven mutants showed significantly slower growth speed (doubling time 36.5 ± 0.7 min to 57.7 ± 1.1 min) compared to their parent strain www.nature.com/scientificreports/ (32.7 ± 1.4 min). In contrast, three of them had similar doubling time (JMUB217-11, 34.1 ± 0.3 min; JMUB217-24, 34.8 ± 0.4 min; JMUB217-9, 35.3 ± 0.9 min) and JMUB217-7 grew faster than wild type (29.0 ± 0.3 min). Surprisingly, growth of the JMUB217-21 carrying GuaA I249fs mutation was very slow (347.1 ± 9.3 min).

Intracellular ATP level in the mutants with reduced oxacillin susceptibility. Some reports exam-
ining intracellular ATP levels of S. aureus in relation to stress responses found that lower cellular ATP production was associated with bacterial tolerance to several environmental stresses such as salt, cold, and antibiotics, and it could also induce the conversion of bacterial cells into persistent forms, including small colony variants 35,36 . Our transcriptomics study with the resistant mutants illustrated that several genes involved in purine biosynthesis and folate biosynthesis were clearly downregulated (Fig. 7I,J), which was similar to the findings of Cassels et al. 37 . However, some genes involved in ATP biosynthesis were significantly upregulated (Fig. 7J). To analyze whether the mutations of oxacillin-resistant mutants affect ATP biosynthesis, the intracellular ATP levels of 23 mutants and their parent strain JMUB217 were measured. The results showed that the intracellular ATP was increased in the mutants compared to the parent strain, and there was good correlation between the levels of intracellular ATP and oxacillin MIC with correlation coefficient of 0.6047 (p = 0.0017) (Fig. 8).
The presence of OS-MRSA is currently a challenge in the clinical management of staphylococcal infections and requires great attention because it is prone to be misidentified as MSSA via routine β-lactam susceptibility testing 10,12,14 . Indeed, the majority of the OS-MRSA isolates used in this study were initially identified as MSSA according to the oxacillin susceptibility profile provided by the original laboratory despite carrying mecA. Similarly, susceptibility testing using cefoxitin, a stronger inducer of the mecA regulatory system than oxacillin that is used to detect methicillin resistance 42 , failed to accurately identify OS-MRSA (Table 1). These observations suggest that a combination of oxacillin and cefoxitin susceptibility tests, as recommended elsewhere 43 , or detection of mecA will be more reliable for the identification of MRSA. Despite being phenotypically susceptible to oxacillin, β-lactam resistance can easily be induced in OS-MRSA 20,21,44 . The mechanisms regulating oxacillin susceptibility in S. aureus appear to differ depending on the types of mutations and genetic basis of the individual isolates. Chen et al. reported that mutations in the MecIbinding site of the mecA promoter downregulated the expression of PBP2a and increased the susceptibility of ST59-SCCmec type V strains to oxacillin 45 . Meanwhile, they demonstrated that mutation of the ribosome-binding site of mecA in an ST59-SCCmec type IV strain attenuated its oxacillin resistance. Nevertheless, these mutations affect only oxacillin resistance in the strains of ST59 background, whereas mutations in the same locus barely affected the β-lactam resistance levels of isolates with different genetic backgrounds, such as COL (ST250) and CH482 (ST45) 24 . Conversely, mutations in the mecA coding region conferred oxacillin resistance to OS-MRSA strains isolated in the US 21 . These studies suggested that the inclusion of OS-MRSA strains with a diverse genetic backgrounds is crucial for providing a comprehensive insight into understanding the mechanism of oxacillin resistance in OS-MRSA.
Although oxacillin resistance in OS-MRSA might be caused by increased mecA expression, the exact mechanism triggering mecA overexpression is unknown. The structure of bla operon is highly homologous to the mec operon 46 , and quite a high portion (84%) of OS-MRSA isolates analyzed in this study carried bla operon (Table 1; Fig. 1), which suggest that bla operon might influence oxacillin susceptibility in OS-MRSA. However, deletion of the repressor gene blaI from an OS-MRSA strain JMUB217 resulted in only slight increases of the oxacillin MIC. In addition, mecA levels were not uniformly increased in a set of JUMB217-derived oxacillin-resistant mutants compared to the parent strain JUMB217, as determined by qRT-PCR. Therefore, we postulated that oxacillin resistance in OS-MRSA strains involves a more complex regulatory pathway than simply direct mecA signaling.
The stringent stress response governed by the alarmone (p)ppGpp is involved in the β-lactam resistance of MRSA 34,47 . Both our whole-genome comparative analysis and RNA-seq analysis demonstrated that many mutations identified in the resistant mutants (Fig. 4) and their altered gene transcriptions (Fig. 7) were associated with depletion of pathway relevant to the purine biosynthesis, protein quality control, and tRNA synthesis, which was very similar to the massive transcription downregulation seen in S. aureus during the stringent response. However, the expression profiles of stringent response elements in oxacillin-resistant mutants derived from this study were not indicative of the classic stringent response elicited by mupirocin treatment 48 . During the classic stringent response, the cellular stresses resulting from amino acid starvation and mupirocin exposure induce the accumulation of uncharged (deacylated) tRNA 49 . The uncharged tRNA in turn binds to the A (aminoacyl-tRNA) site of the 70S ribosome and activates Rel Sau to produce (p)ppGpp 50 . However, in some Gram-positive bacteria like Bacillus subtilis, (p)ppGpp does not directly regulate RNA polymerase (RNAP). Rather, (p)ppGpp synthesis reduces intracellular GTP levels, subsequently leading to the induction of the stringent response 51,52 . Hence, mutations in genes involved in glycolysis, pentose phosphate biosynthesis, folate synthesis, and purine biosynthesis might mimic the (p)ppGpp-mediated reduction in intracellular GTP levels and induce "stringentlike response", as evidenced by the downregulation of genes responsible for GTP biosynthesis (purine and folate biosynthesis, pentose phosphate biosynthesis, and glycolysis) (Fig. 7). In addition, gene mutations identified in this study included many stringent response elements, and most of them were previously reported to be associated with conversion of hetero-to homo-resistance against β-lactam, such as genes associated with RNA polymerase (RNAP; rpoB 31,53 and rpoC 31,54 ), purine biosynthesis (guaA 31  www.nature.com/scientificreports/ β-lactam resistance in the oxacillin-resistant mutants of OS-MRSA. It remains to be further studied, however, how this response leads to substantial metabolic changes towards acquisition of the resistance. OS-MRSA is considered problematic in the clinical setting because the strain is prone to develop high-level β-lactam resistance during the course of antibiotic treatment 20,44 . Because the targets of antibiotics are generally essential proteins in bacteria, the acquisition of antibiotic resistance is usually associated with a fitness cost 56 . In S. aureus, slow-growth phenotypes, including the formation of small colony variants, are associated with tolerance to antibiotics [57][58][59][60] . Contrarily, some of the mutations identified in the JMUB217 strain altered its oxacillin susceptibility without affecting its doubling time. This suggested that the mutations conferring reduced oxacillin susceptibility in OS-MRSA may incur only small fitness costs because of the complementary upregulation of ATP synthase genes. The increased expression of ATP biosynthesis genes was supported by the positive correlation between oxacillin MICs and intracellular ATP levels (Fig. 8), which might explain the easy acquisition of oxacillin resistance in OS-MRSA. Nonetheless, chromosomal mutations in rpoBC and other genes involved in purine biosynthesis were identified in slow VISA strains 61,62 , indicating that the fitness cost of mutations may depend on the genetic background of individual strains.
This study aimed to understand the genetic pathways associated with oxacillin resistance in OS-MRSA isolates from diverse genetic backgrounds. Our results suggested that OS-MRSA was rendered oxacillin-resistant by a combination effect of stringent-like response (a stress response similar to the classic stringent response) and subsequent expression of antibiotic resistance genes (e.g., mecA, bla operon). The relatively low fitness cost of the mutations may fuel the easy selection of oxacillin-resistant OS-MRSA mutants during the course of antimicrobial treatment.

Materials and methods
Bacterial strains and growth conditions. A total of 43 OS-MRSA isolates from various clinical samples were collected from routine clinical laboratories in hospitals across Japan and Taiwan between 1998 and 2015 (Table S1 [63][64][65] ). Mueller-Hinton broth (MHB; Becton Dickinson Co., Ltd., Sparks, MD, USA) and tryptic soy broth (TSB; Becton Dickinson) were used to culture S. aureus, whereas Escherichia coli was grown in Luria-Bertani (LB; Becton Dickinson) medium. In some experiments, antibiotics were added to the medium at the following concentrations: ampicillin (Nacalai Tesque, Inc., Kyoto, Japan) at 100 µg/mL for E. coli, tetracycline (Nacalai Tesque) at 5 µg/mL for S. aureus, and chloramphenicol (Nacalai Tesque) at 10 µg/mL for S. aureus and E. coli. For preservation, bacterial cells were cultivated on tryptic soy agar (TSA; Becton Dickinson) and incubated at 37 °C upon receipt. A single colony was then selected and grown overnight in TSB at 37 °C. The overnight culture was aliquoted and stored at − 80 °C in 50% glycerol (Wako Pure Chemical Industries, Ltd., Tokyo, Japan) until use. mecA detection via PCR. DNA was extracted from OS-MRSA isolates grown overnight on TSA plates using MightyPrep reagent (Takara Bio Inc., Shiga, Japan) in accordance with the manufacturer's instructions. PCR was then performed on the extracted DNA using Quick Taq HS DyeMix (Toyobo Co., Ltd., Osaka, Japan). A primer pair (mecA-F and mecA-R, Table S5) was used to amplify a 519-bp region of mecA. The thermal cycling conditions included initial denaturation at 94 °C for 2 min followed by 30 cycles of 94 °C for 30 s, 55 °C for 30 s, and 68 °C for 1 min. Finally, the amplified products were electrophoresed on 1% agarose gel, stained with ethidium bromide, and visualized using AE-6933FXES Printgraph (Atto Co., Tokyo, Japan). PBP2a production. PBP2a was extracted from colonies grow on MHA and was detected using the MRSAscreen latex agglutination test (Denka, Seiken Co. Ltd., Tokyo, Japan) according to the manufacturer's instructions.

Antibiotic susceptibility test. Oxacillin and cefoxitin
Isolation of mutants with reduced susceptibility to oxacillin from parent OS-MRSA strains. To isolate mutants with reduced oxacillin susceptibility, all 43 OS-MRSA parent strains were exposed to oxacillin according to the E-test method as described for susceptibility testing. Briefly, the oxacillin E-test was performed on OS-MRSA strains inoculated onto MHA plates. A single colony growing inside the inhibition zone after 24-48 h of incubation was randomly picked and sub-cultured in TSB for 24 h at 37 °C. The overnight culture was then serially diluted tenfold with 0.9% NaCl and spread onto a TSA plate. A single colony growing on the TSA plate was again randomly selected and inoculated into TSB for preservation in 50% glycerol at − 80 °C. The oxacillin susceptibility of the stocked cells was determined again using the E-test method to discriminate mutant colonies from persister colonies. The cells exhibiting higher oxacillin MICs were selected as oxacillin-reduced susceptibility mutants, which were then used for subsequent analysis.
Whole-genome sequencing. Genomic DNA was extracted from OS-MRSA and its mutants using the phenol-chloroform method and purified using a QIAamp DNA mini kit (Qiagen, Hilden, Germany) following previously developed methods 66  www.nature.com/scientificreports/ sequencing as previously described 66,67 . Briefly, a mate-pair library was prepared using a Nextera mate-pair library prep kit (Illumina, Inc., San Diego, CA, USA), and sequencing was performed using an Illumina MiSeq platform with the MiSeq reagent kit version 3 (Illumina). The mate-paired reads of OS-MRSA were trimmed using the FASTQ toolkit version 2.2.0 to generate high-quality reads and assembled using Velvet Assembler version 1.2.10 to construct genome scaffolds. The generated genomic sequences were finally annotated using Microbial Genome Annotation Pipeline (https ://www.migap .org/). Meanwhile, the genomic sequences of in vitroselected mutants with reduced oxacillin susceptibility were determined by sequencing paired-end reads as previously described 68 . The paired-end library was prepared using a Nextera XT library prep kit and sequenced using the Illumina MiSeq platform with the MiSeq reagent kit version 3. The paired-end reads of each mutant were mapped against the genomic sequences of their corresponding parent OS-MRSA strains, and mutations were detected using CLC Genomics Workbench version 9 (CLCbio, Qiagen, Valencia, CA, USA). Mutations identified in each mutant were verified by Sanger sequencing using the Applied Biosystems 3130xl genetic analyzer (Thermo Fisher Scientific, MA, USA).

Construction of the phylogenetic tree.
To construct the OS-MRSA phylogenetic tree, kSNP3 69 , available at https ://sourc eforg e.net/proje cts/ksnp/, was first used to identify single nucleotide polymorphisms (SNPs) in the whole-genome sequencing data of OS-MRSA strains. The k-mer size was set to an optimum length of 13 as estimated by Kchooser for extracting SNPs from the sequence data. A maximum parsimony tree was then constructed using the majority of the SNPs present in at least 75% of the genomes. The generated phylogenetic tree was visualized using FigTree ver.

Determination of intracellular ATP levels.
The parent OS-MRSA strains and the laboratory-selected mutants were cultured overnight in MHB at 37 °C with agitation at 150 rpm. The overnight cultures were adjusted to an OD 600 of 0.2 in MHB, and 100 µL of the OD-adjusted culture were inoculated into 10 mL of MHB. The cultures were grown at 37 °C with agitation at 25 rpm in an automatic temperature gradient rocking incubator. One mL of each mid-exponential phase culture (OD 600 = 0.5) was then transferred to a clean 1.5-mL tube and immediately centrifuged at 15,000 rpm for 1 min at 4 °C to pellet cells. After centrifugation, the cell pellet was stored immediately at − 80 °C until analysis. To determine intracellular ATP levels, a BacTiter-Glo Microbial Cell Viability Assay kit (Promega, WI, USA) was used according to the manufacturer's instructions. Briefly, the cell pellet was resuspended in 1 mL of MHB, and 25 µL of the cell suspension were mixed with an equal volume of BacTiter-Glo Reagent in a 384-well opaque plate (Iwaki, Tokyo, Japan) and incubated at room temperature for 5 min. The luminescence was then read on an EnVision 2104 Multilabel Reader (Perkin Elmer, Waltham, MA, USA). The ATP concentration was determined with reference to an ATP standard curve prepared from ATP disodium salt hydrate (A2383, Merck KGaA, Darmstadt, Germany). ATP disodium salt was dissolved in distilled water, yielding 1 µM ATP standard solutions. Serial tenfold dilutions of the ATP standard solution were created using MHB to prepare diluted standards that were then used to generate the standard curve. The intracellular ATP concentration of each sample was presented as the mean of three independent experiments performed using three biological replicates.

RNA extraction.
Overnight cultures of the parent OS-MRSA strains and the laboratory-selected mutants were adjusted to an OD 600 of 0.4. The OD-adjusted cultures were then diluted 1:100 in 1 or 10 mL of MHB for qRT-PCR and RNA-seq analysis, respectively. Each culture was grown to the early log-phase (OD 600 = 0.3) before treatment with a final concentration of 0.1 μg/mL oxacillin or equal volume of distilled water (control) for 1 h (qRT-PCR) or until OD 600 = 0.6 (RNA-seq). After oxacillin treatment, the bacterial cells were harvested by centrifugation at 15,000 rpm for 1 min at 4 °C (qRT-PCR) or at 8000 rpm for 5 min at 4 °C (RNA-seq). Pelleted cells were resuspended in 600 μL (qRT-PCR) or 6 mL (RNA-seq) of TE buffer (

RNA-seq analysis.
To perform RNA-seq analysis, ribosomal RNAs (rRNAs) in total RNA preparations of the JMUB217 strain and its mutant derivatives were first depleted using a Ribo-Zero rRNA Removal Kit (Bacteria) from Illumina. Double-stranded cDNA was then synthesized using a PrimeScript Double Strand cDNA Synthesis Kit (Takara Bio). The generated cDNA served as the template for constructing the paired-end library using a Nextera XT library prep kit, and the library was subsequently sequenced using the Illumina MiSeq platform and the MiSeq reagent kit version 3. RNA-seq analysis was performed using CLC Genomics Workbench version 9, and the RNA-seq reads were aligned to the reference genomes of the parent strain JMUB217. Gene expression was normalized by calculating the reads per kilobase per million mapped reads, and differentially expressed genes were identified using Baggerly's test (β-binomial test) with false discovery rate correction. Genes with adjusted p < 0.01 were considered to be significantly differentially expressed.
Construction of mecA-and blaI-knockout mutants. To construct mecA and blaI-knockout mutants of the JMUB217 strain, the pKFT markerless gene deletion system was used as previously described 71 . Briefly, to delete mecA, 1-kb upstream and downstream flanking sequences of the target gene were amplified by PCR using the primer sets SacI-mecAKO-UP-2/mecA_fPCR_UP and PstI-mecAKO-UP/mecA_fPCR_DN (Table S5), respectively, with KOD FX Neo (Toyobo). Then, second-round PCR was performed using the first-round PCR products as templates with the primer set SacI-mecAKO-UP-2/PstI-mecAKO-UP. The second-round PCR products and pKFT were digested with the restriction enzymes PstI and SacI (Takara Bio) and ligated using Ligation high ver. 2 (Toyobo), generating the plasmid pmecAKO. pmecAKO was transformed into E. coli DH5α, and the transformed cells were plated on LB agar with 100 µg/mL ampicillin. Regarding the generation of blaI-knockout mutants, DNA fragments containing blaI-1 (locus tag: JMUB217_1395) or blaI-2 (locus tag: JMUB217_2048) were first amplified with the primer sets BlaI-1-1/BlaI-1,2-2 and BlaI-2-1/BlaI-1,2-2 (Table S5), respectively. The PCR fragments and pKFT were then digested using the restriction enzymes BamHI and PstI (Takara Bio) and ligated using Ligation high ver. 2. The ligated DNA fragments were independently transformed into E. coli DH5α, and the transformed cells were plated on LB agar with 100 µg/mL ampicillin. The plasmids were extracted, and second-round PCR was conducted using the primer set BlaI-1,2-3/BlaI-1-4 for blaI-1 knockout and BlaI-1,2-3/BlaI-2-4 for blaI-2 knockout. The self-ligated PCR fragments (pblaI-1KO and pblaI-2KO) were again individually transformed into E. coli DH5α, and transformed cells were plated on LB agar with 100 µg/mL ampicillin. Afterwards, all three plasmids (pmecAKO, pblaI-1KO, and pblaI-2KO) were extracted from the E. coli DH5α transformants and transformed into E. coli BL21. The plasmids extracted from E. coli BL21 were subsequently electroporated into S. aureus JMUB217 and mutants with reduced oxacillin susceptibility as previously described 72 , and the cells were cultured on TSA with 5 µg/mL tetracycline at 30 °C. An isolated colony was then grown overnight in 4 mL of TSB containing 5 µg/mL tetracycline at 30 °C. Single crossover was performed by growing the overnight culture on TSA with 5 µg/mL tetracycline at 43 °C. Then, double crossover was performed by incubating the single crossover mutant on TSA at 30 °C. The double crossover event was confirmed by PCR and Sanger sequencing.

Complementation of mecA.
To generate a mecA-complemented mutant, a DNA fragment containing wild-type mecA from strain JMUB217 was amplified using the primers SmaI-mecAcomp-F-pKAT and SalI-mecAcomp-R-pKAT (Table S5). The PCR fragment and pKAT were digested with SmaI and SalI (Takara Bio) and ligated using Ligation high ver. 2. The ligated DNA fragment was transformed into E. coli DH5α, and the transformed cells were plated on LB agar with 10 µg/mL chloramphenicol. Finally, the complementation plasmid was extracted and electroporated into the JMUB217 strain 72 .
Statistical analysis. All statistical analyses were performed using Prism 8 (GraphPad Software, San Diego, CA, USA). Statistical comparison was carried out using the Student's t-test whereas the correlations between variables were calculated using the non-parametric Spearman's correlation coefficient (r s ). Statistical significance was denoted with a P value of < 0.05.

Data availability
The raw sequence data have been deposited in DNA Data Bank of Japan (DDBJ) under accession number DRA009699 and DRA009727. www.nature.com/scientificreports/