Novel determinants of antibiotic resistance: identification of mutated loci in highly methicillin-resistant subpopulations of methicillin-resistant Staphylococcus aureus.

We identified mutated genes in highly resistant subpopulations of methicillin-resistant Staphylococcus aureus (MRSA) that are most likely responsible for the historic failure of the β-lactam family of antibiotics as therapeutic agents against these important pathogens. Such subpopulations are produced during growth of most clinical MRSA strains, including the four historically early MRSA isolates studied here. Chromosomal DNA was prepared from the highly resistant cells along with DNA from the majority of cells (poorly resistant cells) followed by full genome sequencing. In the highly resistant cells, mutations were identified in 3 intergenic sequences and 27 genes representing a wide range of functional categories. A common feature of these mutations appears to be their capacity to induce high-level β-lactam resistance and increased amounts of the resistance protein PBP2A in the bacteria. The observations fit a recently described model in which the ultimate controlling factor of the phenotypic expression of β-lactam resistance in MRSA is a RelA-mediated stringent response. IMPORTANCE It has been well established that the level of antibiotic resistance (i.e., minimum concentration of a β-lactam antibiotic needed to inhibit growth) of a methicillin-resistant Staphylococcus aureus (MRSA) strain depends on the transcription and translation of the resistance protein PBP2A. Here we describe mutated loci in an additional novel set of genetic determinants that appear to be essential for the unusually high resistance levels typical of subpopulations of staphylococci that are produced with unique low frequency in most MRSA clinical isolates. We propose that mutations in these determinants can trigger induction of the stringent stress response which was recently shown to cause increased transcription/translation of the resistance protein PBP2A in parallel with the increased level of resistance.

INTRODUCTION

It is generally agreed that the appearance of methicillin-resistant Staphylococcus aureus (MRSA) strains among clinical isolates represents the single most serious blow to the chemotherapy of S. aureus infections, since the unique resistance mechanism carried by all MRSA strains provides protection against the single largest family of antibacterial agents—the β-lactam antibiotics (1). Since their first appearance in clinical specimens in 1960, this resistance mechanism has made its way into a large variety of S. aureus lineages and diverse clones of MRSA have spread throughout the globe to cause serious and often life-threatening infections both in hospitals and in the community (2–5).

Most MRSA strains carry an identical—acquired—genetic determinant mecA (6, 7) which is part of a mobile genetic element (staphylococcal cassette chromosome mec element [SCCmec]) (8) inserted into the S. aureus chromosome at a unique chromosomal site. mecA encodes a protein, PBP2A, a peptidoglycan transpeptidase with extremely low affinity for the entire large family of β-lactam antibiotics (9), and the presence of this protein plays a critical role in allowing MRSA strains to continue synthesis of peptidoglycan and bacterial growth in the presence of high concentrations of β-lactam antibiotics. A model for the mechanism of action of PBP2A on the molecular level has been proposed (10).

In contrast to the common molecular mechanism of resistance, individual MRSA clinical isolates differ widely in their susceptibility to β-lactam antibiotics with individual MRSA strains presenting methicillin MIC values as low as a few µg/ml up to several hundred µg/ml depending on the particular MRSA clone, and this variation in resistance level cannot be explained by transcriptional regulation of mecA through the activity of regulatory elements such as the mecl and mecR1 or blal and blaR1 genes (11, 12).

A detailed examination of the β-lactam susceptibility of cultures of MRSA strains presents an even more complex and intriguing picture. MRSA grown from single-cell inocula produce cultures that are highly heterogeneous with respect to their antibiotic susceptibility with most cells (more than 99%) showing only moderate- or low-level resistance often close to the MIC values of methicillin-susceptible isolates. On the other hand, the same cultures also contain bacteria with an extremely high level of resistance—in the MIC range of several hundred µg/ml—and the frequency of such highly resistant cells in a given culture (10−4 to 10−5) appears to be specific for the particular MRSA clone. This phenomenon has become known as the “heterogeneous” phenotype. It was first recognized and described in 1960, in the microbiological analysis of the historically first MRSA infection by Jevons who was surprised to recover two MRSA populations with widely different methicillin MIC values from a patient with an MRSA infection (13). The methicillin MIC of the majority of the bacteria was 2 µg/ml, but upon prolonged incubation of the specimen, more bacteria were recovered with a much higher antibiotic MIC value in the range of several hundred µg/ml.

Most contemporary clinical isolates of MRSA express β-lactam resistance in a similar heterogeneous fashion (14). Plotting the number of bacteria capable of forming colonies against the concentration of the antibiotic in the agar plates produce phenotypic profiles called population analysis profiles (PAPs), and the shape of the PAP is characteristic for the particular MRSA strain (15, 16). The PAP was subsequently shown to be a unique phenotypic marker of MRSA clones—highly reproducible in chronologically distinct isolates of the same MRSA lineage (17). The presence of highly methicillin-resistant cells in cultures of MRSA is of obvious relevance both for the detection of MRSA in clinical specimens and also for therapeutic options (18–20).

The stability of PAP for a given MRSA clone indicates that the heterogeneous composition of MRSA cultures is genetically controlled, i.e., the highly resistant subpopulations of bacteria must carry mutations in some genetic determinants that are “wild type” in the majority of less-resistant cells of the same clone.

Full genome sequences of a large number of MRSA strains are now available in the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov). However, most of the sequenced strains show heterogeneous antibiotic resistance, and therefore, the sequence information available relates only to the genetic makeup of the “cell majorities” in each of these particular MRSA clones and provide no information about unique mutated genes that must be present in highly resistant subpopulations carried by each one of the sequenced MRSA strains.

The purpose of this study was to identify mutated genes associated with the rare highly resistant bacteria that are produced with a low frequency (10−4 to 10−5) during growth of four historically early MRSA lineages belonging to the archaic clone of MRSA (17). The four strains selected for the study carry the mecA determinant on SCCmec type I cassettes that do not contain an active form of the mecI or mecR regulatory genes (12). The β-lactam antibiotic (methicillin or oxacillin) MIC values for the majority of bacteria in these four strains were in the range of 2 up to 12 µg/ml of oxacillin. When these four “parental” MRSA strains were plated on oxacillin-containing agar for PAPs, rare colonies (10−4 to 10−5) capable of growing on agar containing 100 µg/ml of the antibiotic were also detected. Cultures of such colonies—named “H*R”—(for homogeneous and high-level resistance)—produced highly and homogeneously resistant populations of bacteria with MIC values in the range of several hundred µg/ml.

Chromosomal DNA was prepared from the H*R cultures along with DNA prepared from the corresponding heteroresistant “parental” (majority) cells followed by full genome sequencing. Mutated genes unique to the particular H*R culture were identified by comparison to the status of the same gene in the “parental” culture.

Mutations in 27 genes and 3 intergenic sequences were identified in the highly resistant H*R derivatives recovered from the four heteroresistant “families” of MRSA. While the mutated genes represent a range of functional categories, we suggest that a common feature of these mutations may be their capacity to induce a stringent stress response in the bacteria. This proposal is consistent with recent evidence that identified a key role of the relA gene complex in defining the level of β-lactam resistance in laboratory models of MRSA strains (20, 21). As experimental evidence strongly suggested this, each of the four heteroresistant “parental” MRSA strains described in the present communication could be made to change their mode of expression of resistance from heterogeneous to homogeneous by the use of mupirocin, i.e., by experimentally inducing the stringent stress response in the bacteria (22–30).

RESULTS

The majority of clinical MRSA isolates express β-lactam resistance in a heterogeneous fashion. As a first attempt to better understand the genetic basis of this phenomenon, we selected four genetically closely related and historically early MRSA strains which had very similar heteroresistant phenotypes, as indicated by the virtually superimposable population analysis profiles (PAPs) (Fig. 1). The four strains included UK13136, the historically first MRSA, isolated in 1960 in the United Kingdom (13) named in our study as the “parental” strain of family A. This strain has already been characterized by molecular techniques (31). The second MRSA strain, ST63/458, was also isolated in the United Kingdom in 1963 and was the “parental” strain of family B in our study. Both of these strains are of sequence type ST250 and carry an SCCmec type I. The two additional parental strains, E2125 (parental strain of family C) and strain E4278 (parental strain of family D), were both isolated in Denmark in 1964 and 1967, respectively, and both were ST247 carrying SCCmec type I. All four strains belonged to the “archaic” clone of MRSA (17).

FIG 1 

Heterogeneous expression of β-lactam antibiotic resistance in four MRSA strains. (A to D) Population analysis profiles for strains UK13136 (A), UK63/458 (B), E2125 (C), and E4278 (D) are shown by the lines with solid squares. H*R isolates capable of growing in the presence of 100 µg/ml oxacillin were picked from the agar plates as indicated by an asterisk. Population analysis profiles of cultures of H*R isolates are shown by the lines with empty squares.

Figure 1 shows the rather similar PAPs of the four “parental” MRSA strains; in each strain, the majority of bacteria had relatively low oxacillin MIC values between 2 and 12 µg/ml. However, each culture also contained—with a low frequency of about 10−5—highly resistant (mutant) subpopulations of bacteria with an oxacillin MIC of ≥400 µg/ml.

Overnight cultures of the four parental strains were grown in tryptic soy broth (TSB), and 1-ml portions of the turbid overnight cultures were used to prepare the “parental” DNAs for sequencing. Aliquots of the rest of the overnight parental cultures were plated for population analysis, and 20 of the rare highly resistant colonies (named “H*R”) that appeared on the agar plates supplemented with 100 µg/ml oxacillin (see the asterisks in Fig. 1), were picked from the progeny of each of the four parental cultures. The H*R colonies were resuspended in TSB and restreaked on tryptic soy agar (TSA), and cultures of the 10 H*R colonies were grown in TSB, retested for oxacillin resistance by PAP, and used to prepare DNA for sequencing. Eventually, the DNA sequence of each H*R colony was compared to the sequence of the corresponding “parental” culture in order to identify the genes that were mutated in the H*R cultures.

On the basis of this comparison, mutations in 3 intergenic sequences and 27 genes were identified in the stably resistant H*R colonies recovered from the four heteroresistant MRSA strains (Table 1).

TABLE 1 

Mutated loci in highly resistant (H*R) isolates identified in heteroresistant MRSA strains

Mutation no.	Locus in S. aureus COL	Description[a]	Nucleotide change[b]	Amino acid change	Functional category[c]
1	SACOL0314	Transcriptional regulator (rpiR family)	C349448T	Thr216Ile	b
2	SACOL0403	Transcriptional antiterminator (bglG family)	C409253 Del	Frameshift after Ser461	b
3	SACOL0434	Hypothetical protein 1	C440981T	Gln41Stop	f
4	SACOL0460	IMP dehydrogenase (guaB)	C463203T	Arg310Cys	a
5	SACOL0461	GMP synthase (guaA)	A463937 Del	Deletion after Ser27	a
			C464191T	Pro142Leu
6	SACOL0495	Hypothetical protein 2	G497663T	Asp777Stop	f
7	SACOL0533	Methionyl-tRNA synthetase (metS)	G542898T	Leu285Phe	c
8	SACOL0544	Ribose-phosphate pyrophosphokinase (prsA)	C552272T	Pro291Leu	a
9	SACOL0554	Hypoxanthine phosphoribosyltransferase (hpt)	G562925A	Met1Ile	a
10	SACOL0555	Cell division protein (ftsH)	C565004A	Ala429Asp	e
11	SACOL0562	Lysyl-tRNA synthetase (lysS)	G570441T	Arg11Leu	c
			G570684A	Arg93His
12	SACOL0574	Glutamyl-tRNA synthetase (gltX)	G597328A	Glu439Lys	c
13	SACOL0576	Cysteinyl-tRNA synthetase (cysS)	C599284G	His255Asp	c
14	SACOL0583	Ribosomal protein L11 (rplK)	A603893T	Ile140Val	c
15	SACOL0588	DNA-directed RNA polymerase, β-subunit (rpoB)	C608417T	Ala477Val	b
			T610153A	Tyr1056Asn
16	SACOL0589	DNA-directed RNA polymerase, β′-subunit (rpoC)	G611391A	Arg239His	b
			G611672T	Gly333Cys
			C611927A	Leu418Ile
			A612110T	Ile479Phe
			T612157A	Asp494Glu
			C612859A	Asn728Lys
			C612921A	Ala749Glu
			C613500T	Thr942Ile
			G613517T	Val948Leu
17	SACOL0758	1-Phosphofructokinase (fruK)	G779646 Ins	Frameshift after Ala35	d
18	Intergenic	Hypothetical protein 3/glucose-6-phosphate isomerase	C968358T		f
19	SACOL0991	Oligopeptide ABC transporter, permease (oppB)	G998492A	Asp265Asn	d
20	SACOL1689	GTP pyrophosphokinase (relA2)	G1719144 Del	Frameshift after Met383	a
			C1719536T	Gln255Stop
21	SACOL1710	Valyl-tRNA synthetase (valS)	A1741737T	Asp177Val	c
22	Intergenic	Valyl-tRNA synthetase/DNA-3-methyladenine glycosylase	C1742564 Del		f
23	SACOL1717	Porphobilinogen deaminase (hemC)	C1747855T	Thr265Ile	d
24	SACOL1745	Ribosome binding site of pyruvate kinase (pyk)	C1783697 Del	Deletion of RBS	d
25	SACOL2038	tRNA N6-adenosine threonylcarbamoyltransferase (gcp)	G2098447A	Gly193Asp	c
26	SACOL2072	DEAD box ATP-dependent RNA helicase (srmB)	G2137588A	Gly459Asp	b
27	SACOL2108	Translation factor SUA5 (sua-5)	C2168647G	Pro104Arg	c
28	SACOL2117	Fructose-bisphosphate aldolase (fbaA)	T2177986 Del	Frameshift after Lys167	d
			G2178061A	Gly143Arg
			C2178241T	His83Tyr
29	SACOL2215	Ribosomal protein S13/S18 (rpsM)	G2295683C	Ile308Leu	c
30	Intergenic	Hypothetical protein 4/hydroxymethylglutaryl-CoA reductase	G2617978A		f

CoA, coenzyme A.

C349448T, C at position 349448 changed to T; C409253 Del, deletion of the C at position 409253; G779646 Ins, insertion of G at position 779646.

Functional categories a through g as defined in Table 2.

TABLE 2 

Functional categories of mutations associated with highly resistant (H*R) isolates

Functional category	No. of determinants	Mutation(s)[a]  associated with H*R isolates in the following strain:
		UK13136 (family A)	ST63/458 (family B)	E2125 (family C)	E4278 (family D)
Guanine metabolism (category a)	5	guaA, guaB	guaA, relA2, prsA		relA2, hpt
Transcription (category b)	5	bglG, rpoB, RNA helicase	rpoC		rpiR, rpoB, rpoC
Translation/ribosomal structure (category c)	9	rpsM, lysS	rplK	gcp, cysS, valS	lysS, gltX, metS, sua-5
Transport/metabolism (category d)	5	pyk, hemC	fbaA		oppB, fruK
Cell division (category e)	1	ftsH
Unknown function (category f)	5	HP2, Intergenic		Intergenic	HP1, Intergenic
Total	30

Genes potentially involved with induction of a stringent stress response are indicated in boldface print. The fbaA and oppB genes are downregulated by (p)ppGpp-mediated stringent stress response (33). HP stands for hypothetical protein.

The mutated genes, their putative functions, and the nature of the nucleotide and amino acid change are listed in Table 1. The following six genes carried multiple mutations: fbaA (3 mutations), (2 mutations), lysS (2 mutations), rpoB (2 mutations), rpoC (9 mutations), and relA2 (2 mutations).

These genes may be involved in (p)ppGpp-mediated stringent stress response: , lysS, and relA2 are directly linked to the synthesis of (p)ppGpp which targets RNA polymerase, the product of rpoB and rpoC (32). The fbaA gene encoding fructose bisphosphate aldolase was reported to be downregulated in (p)ppGpp-mediated stringent stress response induced by serine hydroxamate (33). Of the 30 genetic loci carrying mutations, 27 are expected to alter function either by point mutations or by frameshifts. The remaining 3 determinants were in intergenic sequences with a change in a single nucleotide. Thus, mutations in 27 different genes would appear to be responsible for the increase in the resistance of H*R isolates to oxacillin either singly or through a concerted effect of all mutations.

Table 2 lists functional categories of the mutated genes identified in H*R isolates of the four MRSA families identified by capital letters A through D. Of the 27 mutated genes, 21 were in guanine metabolism (a), in transcription (b), in translation/ribosomal structure (c) and/or in transport (d). Interestingly, in most H*R isolates, mutation in a single gene was sufficient to produce the highly resistant phenotype (Table 3). Thirteen out of 17 genes listed in Table 3 are included in 3 functional categories: four (, prsA, hpt, and relA2) in guanine metabolism; two (rpoB and rpoC) in transcription; and seven (metS, lysS, cysS, valS, gcp, sua-5, and rpsM) in translation/ribosome structure. Each of these mutations would be expected to trigger the stringent stress response and produce high and homogeneous resistance.

TABLE 3 

H*R isolates carrying mutations in a single gene

Mutation no.[a]	Locus in S. aureus COL	Gene	H*R strain	Functional category
5	SACOL0461	guaA	A3, BB9	a
7	SACOL0533	metS	DD8	c
8	SACOL0544	prsA	B5	a
9	SACOL0554	hpt	DD3	a
11	SACOL0562	lysS	AA9	c
13	SACOL0576	cysS	CC3	c
15	SACOL0588	rpoB	DD9	b
16	SACOL0589	rpoC	B4, B8, B9, BB2, BB3, DD6, DD7	b
17	SACOL0758	fruK	DD5	d
20	SACOL1689	relA2	BB8, D3	a
21	SACOL1710	valS	C8	c
23	SACOL1717	hemC	AA2	d
24	SACOL1745	pyk	A5	d
25	SACOL2038	gcp	CC1	c
27	SACOL2108	sua-5	DD1	c
28	SACOL2117	fbaA	B10, BB5, BB6	d
29	SACOL2215	rpsM	A2	c

Mutation numbers as in Table 1.

Twelve H*R isolates each carrying a single mutation (Table 3) were compared to their respective parental strains for the relative amounts of PBP2A. Cells were grown in the presence of 0.5 µg/ml oxacillin to induce the mecA gene, and membrane fractions were prepared for Western blotting. All H*R isolates showed at least 2-fold increase in PBP2A compared to their parental strains (Fig. 2) suggesting that each mutation resulted in the recruitment of increased amounts of PBP2A into the cell membranes.

FIG 2 

Determination of PBP2A in membranes of H*R derivatives carrying single mutations. A group of H*R isolates carrying single mutations in a variety of genes associated with high-level oxacillin resistance (Table 3) were analyzed by SDS-PAGE and by testing the relative amounts of PBP2A by Western blotting with a monoclonal antibody prepared against PBP2A. Three isolates (A2, A5, and AA2) from family A, four isolates (B5, BB6, BB8, and BB9) from family B, and five isolates from family D (D3, DD1, DD3, DD5, and DD6) were included in the analysis. Lanes A, B, and D show the SDS-PAGE profiles and the Western blot analysis of PBP2A in these parental isolates. Lanes Am, Bm, and Dm contain parental samples in which the relative amounts of PBP2A were estimated in the presence of mupirocin. The single mutations carried by the H*R derivatives of family A were as follows: rpsM in lane A2, pyk in lane A5, and hemC in lane AA2. The single mutated genes analyzed in family B were prsA in lane B5, fbaA in lane BB6, relA2 in lane BB8, and in lane BB9. The mutations analyzed in members of family D were relA2 in lane D3, sua5 in lane DD1, hpt in lane DD3, fruK in lane DD5, and rpoC in lane DD6. The M lanes contain molecular size markers (100, 70, 55, and 45 kDa).

DISCUSSION

Inspection of Table 1 through 3 and the figures indicates that a large number and different kinds of mutations can profoundly influence the phenotypic expression of oxacillin resistance in the four heteroresistant MRSA strains. Determinants include genes in cell division as well as genes associated with various aspects of intermediary metabolism. Such a diversity of genetic determinants is reminiscent of the large number of “auxiliary genes” (or fem genes) identified earlier as determinants essential for the optimal expression of high and homogeneous resistance in MRSA strains (34, 35). As a hypothesis to account for the polygenic nature of this phenomenon, it was proposed that the expression of antibiotic resistance involves a bacterial stress response (36).

This proposition, originally suggested to explain the multigenic nature of homogeneous oxacillin resistance, seems to also fit the mechanism of heterogeneous resistance analyzed in this communication. In a recent study, we described identification of the critical role that the S. aureus relA gene plays in the phenotypic expression of oxacillin resistance (20). RelA protein plays a central role in the control of biosynthetic activities through its catalytic product—ppGpp and pppGpp—that can interact with and regulate the ribosomal protein synthesis machinery.

The large number and functional diversity of the determinants that influence heterogeneous expression of resistance as described in this communication would fit the model in which a stress response—specifically, the stringent stress—is the ultimate controlling factor of the phenotypic expression of oxacillin resistance in MRSA (20, 21).

Figure 3 shows a modification of the relA model to indicate how diverse genetic determinants identified in the present communication could impact on the level of antibiotic resistance through specific interactions with the relA-controlled RNA polymerase system. In a recent communication (21), we began to test whether the physiological level of antibiotic resistance (i.e., the oxacillin MIC value) is paralleled by the cellular amounts of the mecA gene product PBP2A. In the model systems described in reference 21, increase in the MIC value was accompanied by a parallel increase in the cellular amounts of PBP2A, and increased amounts of PBP2A were also detected in the H*R derivatives described in this communication (Fig. 2).

FIG 3 

Model for the triggering of the stringent stress response by mutations identified in the highly resistant (H*R) isolates. A schematic model for the postulated effect of H*R mutations on the relA-controlled stress response of S. aureus is shown. Functional categories of mutations are defined as in Table 2. HD, hydrolase; SYN, synthetase; TGS, a domain named after three enzymes that contain it [threonyl-tRNA synthetase (ThrRS), GTPase, and guanosine-3′,5′-bis(diphosphate) 3′-pyrophosphohydrolase (SpoT)]; ACT, the domain named after three proteins that carry it [aspartate kinase, chorismate mutase, and prephenate dehydrogenase (TyrA)].

If the stringent stress response is the central controlling element of the level of oxacillin resistance in MRSA, then one would expect that artificial triggering of the stress response would convert heterogeneously resistant MRSA to highly and homogeneously resistant cultures. In an effort to test this, we repeated the population analysis of the four heteroresistant “parental” strains in the presence of sub-MICs of mupirocin, an agent capable of inducing stringent stress. Figure 4 demonstrates that each one of the four heteroresistant parental MRSA strains described in this communication would exhibit high and homogeneous resistance if the phenotype was assayed in the presence of sub-MICs of mupirocin added to the oxacillin-containing agar plates. Identical results were obtained when serine hydroxamate, an inhibitor of seryl-tRNA synthetase, was used instead of mupirocin. These observations may open up so-far untested avenues for the design of antibacterial agents that could influence resistance level of MRSA through a novel type of intervention.

FIG 4 

Conversion of the heterogeneous population analysis profiles of four MRSA strains to high and homogeneous resistance by induction of the stringent stress response. Population analysis profiles (PAPs) of the four MRSA cultures (strains in Fig. 1A to D) (solid squares), PAPs of H*R derivatives (empty squares), and PAPs determined on agar plates on which the antibiotic was supplemented by sub-MICs (0.03 µg/ml) of mupirocin (solid circles). Asterisks indicate the concentration of oxacillin (100 µg/ml) at which the H*R colonies were picked.

MATERIALS AND METHODS

Aliquots (1 µl) of the four heteroresistant “parental” cultures of UK13136 (parental strain of family A), ST63/458 (parental strain of family B), E2125 (parental strain of family C), and E4278 (parental strain of family D) were inoculated into 5 ml of tryptic soy broth (TSB) and incubated at 37°C overnight with agitation. Portions (1 ml) of the overnight cultures were removed to prepare chromosomal DNAs representing the majority of cells (poorly resistant cells) of these cultures. The overnight cultures were diluted with TSB, and population analysis profiles (PAPs) were done on tryptic soy agar (TSA) plates containing increasing concentrations of oxacillin (Fig. 1). CFU were counted after 48-h incubation of the plates at 37°C. Twenty medium-size colonies capable of growing on TSA plates containing 100 µg/ml oxacillin were picked from the PAP plates of each of the four “parental” MRSA. These colonies were named “H*R” for homogeneous and high-level oxacillin resistance. H*R colonies were recovered from the plates with 1-µl loops and dispersed into Eppendorf tubes containing 200 µl of TSB. Portions (1 µl) from each Eppendorf tube were streaked onto a TSA plate which was incubated at 37°C for 48 h. The H*R isolates were next passaged three times onto fresh TSA plates, after which the isolates were retested for resistance level by Etest and population analysis. A total of 42 H*R isolates with high-level and homogeneous oxacillin resistance (oxacillin MIC of ≥400 µg/ml) were inoculated into 5 ml of TSB, incubated at 37°C with agitation overnight, and used to prepare H*R DNAs. The 42 H*R isolates included 10 colonies of UK13136, 12 of UK63/458, 10 of E2125, and 10 of E4278.

The antibiotic resistance profiles of the four heteroresistant strains were also determined by including sub-MICs of mupirocin in the antibiotic-containing plates used for population analysis (37, 38). Mupirocin is a known inducer of the stringent stress response in bacteria.

Genome sequencing.

Sequencing libraries were prepared according to previously published methods (39–41). Samples were run on an Illumina HiSeq 2000 sequencer operated according to the manufacturer’s instructions with 100 cycle paired-end runs. Data for the samples have been deposited in the European Nucleotide Archive (see below).

Detection of variations between H*R isolates and the “parental” strains.

The sequence of chromosomal DNAs isolated from the H*R colonies was compared to the DNA sequence of the corresponding “parental” strain in order to identify in the H*R isolates mutated loci that may be associated with the high-level oxacillin-resistant phenotype of these clones. This was done using three different approaches to call only high-confidence variants.

The first two approaches are based on de novo assemblers, which are capable of detecting variants (single nucleotide polymorphisms [SNPs]), insertions, and deletions while building the contigs. Both methods completely ignore reference genomes while calling variants between “parental” and H*R isolates. However, to make comparison between the methods easier, the results were mapped back to reference strain COL (GenBank accession number CP000046).

SGA v0.9.19 (42) commands “preprocess” and “index” were run using default settings. Variants were called using “graph-diff” with k-mer (−k) = 61 and min-discovery-count (−x) = 10. Cortex v1.0.5.15 (43) was run using the provided workflow pipeline using the joint variant discovery with k-mers between 31 and 63 (43).

The third approach is based on mapping. Each “parental” isolate and corresponding H*R isolates were mapped against strain COL using SMALT v.0.7.4 (http://www.sanger.ac.uk/resources/software/smalt). High-quality SNPs were called as described previously (44). Detection of indels was carried out using GATK (45). The obtained variant call format files (VCFs) were processed using an in-house script to remove variant calls due to the use of the reference genome COL in order to find only differences between the “parental” strain and H*R isolates. Variants found by all three methods were finally checked manually in order to carry on further analysis with high-quality variants.

Preparation of staphylococcal membrane proteins.

Membrane fractions were prepared from isolates belonging to families A, B, and D following the method described previously (21, 38) with slight modification. S. aureus strains were grown at 37°C in 200 ml of TSB in the presence of 0.5 µg/ml of oxacillin to induce transcription of the mecA gene. The strains analyzed included three H*R isolates from family A (A2, A5, and AA2), four H*R isolates from family B (B5, BB6, BB8, and BB9), and five H*R isolates from family D (D3, DD1, DD3, DD5, and DD6). Each analysis included the SDS-PAGE profiles of the corresponding parental strain: strain UK13136 for family A, strain ST63/458 for family B, and strain E4278 for family D. The relative amounts of PBP2A were determined in each of the isolates using Western blotting. The SDS-PAGE profiles and relative amounts of PBP2A were also compared for each of the parental strains with and without mupirocin (0.03 µg/ml) added to the growth medium. All cultures were harvested at an optical density at 620 nm (OD620) of 0.5, washed, and resuspended in 3 ml of 20 mM Tris-HCl (pH 7.6) containing 1× Halt protease inhibitor cocktail (Thermo Fisher Scientific, Inc.), 10 mM MgCl2, 100 µg/ml lysostaphin, 50 µg/ml lysozyme, 50 µg/ml DNase I, and 50 µg/ml RNase A. The cells were incubated at 37°C for 30 min and disrupted by sonication with pulse of 40% output for 5 min. The supernatants were transferred to fresh ultracentrifuge tubes after centrifugation at 7,000 × g for 20 min. Membrane fractions were collected by centrifugation at 100,000 × g for 1 h. The collected membranes were resuspended in 20 mM Tris-HCl (pH 7.6) and stored at −70°C. The concentration of total membrane proteins was determined by the bicinchoninic acid (BCA) assay.

Western blotting.

Western blotting with a monoclonal antibody prepared against PBP2A was used to determine PBP2A in membrane preparations as described previously with a few modifications (21, 38). The membrane proteins (50 µg for families A and B and 100 µg for family D) were loaded on the polyacrylamide gel (8% or 10% resolving gel; 4% stacking gel) for SDS-PAGE. The rabbit anti-PBP2A antibody was used as the primary antibody with dilution of 1:15,000, and the secondary horseradish peroxidase (HRP)-conjugated anti-rabbit antibody (0.5 mg/ml; PerkinElmer) was diluted to 1:10,000. ChromPure human IgG Fc fragment (Millipore) was added to the blocking solution at a final concentration of 3 µg/ml in order to prevent the antibodies from nonspecific binding. Pierce enhanced chemiluminescence (ECL) 2 (Thermo Fisher Scientific, Inc.) substrate was used for visualization of PBP2A bands with X-ray film exposure.

Nucleotide sequence accession numbers.

Data for the genome sequencing samples have been deposited in the European Nucleotide Archive under the sample numbers ERS157365, ERS157381, ERS157396, ERS157409, and ERS157425 to ERS157449.