Molecular Characterization of Clinical Rel Mutations and Consequences for Resistance Expression and Fitness in Staphylococcus aureus

ABSTRACT

The stringent response (SR) is a universal stress response that acts as a global regulator of bacterial physiology and virulence, and is a contributor to antibiotic tolerance and resistance. In most bacteria, the SR is controlled by a bifunctional enzyme, Rel, which both synthesizes and hydrolyzes the alarmone (p)ppGpp via two distinct catalytic domains. The balance between these antagonistic activities is fine-tuned to the needs of the cell and, in a “relaxed” state, the hydrolase activity of Rel dominates. We have previously shown that two single amino acid substitutions in Rel (that were identified in clinical isolates from persistent infections) confer elevated basal concentrations of (p)ppGpp and consequent multidrug tolerance in Staphylococcus aureus. Here, we explore the molecular details of how these mutations bring about this increase in cellular (p)ppGpp and investigate the wider cellular consequences in terms of resistance expression, resistance development, and bacterial fitness. Using enzyme assays, we show that both these mutations drastically reduce the hydrolase activity of Rel, thereby shifting the balance of Rel activity in favor of (p)ppGpp synthesis. We also demonstrate that these mutations induce high-level, homogeneous expression of β-lactam resistance and confer a significant fitness advantage in the presence of bactericidal antibiotics (but a fitness cost in the absence of antibiotic). In contrast, these mutations do not appear to accelerate the emergence of endogenous resistance mutations in vitro. Overall, our findings reveal the complex nature of Rel regulation and the multifaceted implications of clinical Rel mutations in terms of antibiotic efficacy and bacteria survival.

INTRODUCTION

The stringent response (SR) is a universal bacterial stress response that is classically induced by amino acid starvation (1). The presence of an uncharged tRNA in the ribosomal A-site is sensed by the SR controller, Rel, which responds by synthesizing the signaling molecules guanosine tetra- and pentaphosphate, which are collectively known as (p)ppGpp (2, 3). Through a variety of different mechanisms in different bacteria, (p)ppGpp acts to downregulate most metabolic processes, leading to a cessation of growth. While this was the classical view of the SR, we now know that a basal level of (p)ppGpp is present and essential in all cells and that this “alarmone” acts as a master regulator of almost all aspects of bacterial physiology and virulence (2, 4–7). We also now know that activation of the SR is not binary (on/off), and the cellular concentration of (p)ppGpp is finely tuned to the needs of the cell (8). In Escherichia coli and other gammaproteobacteria, two RelA/SpoT homolog (RSH) proteins, RelA and SpoT, work together to regulate the cellular (p)ppGpp concentration. However, in most bacteria, the basal level of (p)ppGpp is controlled by a single, bifunctional RSH known as Rel (3). The enzymatic N-terminal region of Rel consists of a (p)ppGpp hydrolase domain, which degrades (p)ppGpp into GTP/GDP and pyrophosphate; a (p)ppGpp synthetase domain, which synthesizes (p)ppGpp from GTP/GDP and ATP; and a central three-helix bundle (C3HB) linker region, which joins the two catalytically active domains (Fig. 1A). These two antagonistic activities work together to control the level of (p)ppGpp in the cell (9). The C-terminal region of Rel consists of four domains (Thr-tRNA synthetase, GTPase, and SpoT [TGS], alpha-helical domain [AH], ribosome intersubunit domain [RIS], and apartokinase, chorismate mutase, and TyrA domain [ACT]) that are involved in ribosome binding and regulation of the N-terminal catalytic domains (3, 10, 11). In the absence of stalled ribosomes (i.e., in a “relaxed” state) and in vitro, domains within the C-terminal region of Rel associate with the synthetase domain, thereby limiting synthetase activity and maintaining Rel in a predominantly hydrolase-on/synthetase-off conformation (11, 12). Truncation of the C-terminal domain produces a version of Rel that exhibits higher synthetase activity, while hydrolase activity is unaffected (11, 13).

FIG 1.

Clinical stringent response-activating mutations shift the antagonistic activities of Rel in favor of (p)ppGpp synthesis. (A) Schematic diagram of Rel domain architecture. Amino acid numbering on the top refers to SaRel. Abbreviations: TGS, Thr-tRNA synthetase, GTPase, and SpoT domain; AH, alpha-helical domain; RIS, ribosome intersubunit domain, otherwise known as conserved cysteine domain (CC) or zinc finger domain (ZFD); and ACT, apartokinase, chorismate mutase, and TyrA domain, otherwise known as RNA recognition motif domain (RRM). (B) Rates of hydrolase and synthetase activity exhibited by wild-type and mutant versions of full-length SaRel. Data shown are the mean of three replicates; error bars represent the SEM. Asterisks above bars indicate statistically significant differences from the wild type as determined by a one-way ANOVA with Dunnett’s multiple-comparison test (**, P ≤ 0.01; ****, P ≤ 0.0001; ns, not significant [P > 0.05]). (C and D) Cartoon representations of the N-terminal catalytic region of SaRel (as predicted by AlphaFold2) and BsRel (PDB accession no. 6YXA), respectively. The different domains are colored blue (hydrolase), yellow (synthetase), and gray (central three-helix bundle). F128, L152, and catalytic residues are shown as sticks. The first seven residues of the SaRel AlphaFold2 structure represent an alternative start site and have been deleted. (E) Sequence alignment of the N-terminal catalytic region of SaRel and BsRel. Shading indicates the degree of amino acid conservation (black, conserved; gray, conservative substitution; white, not conserved). The red and green circles indicate the sites of the two clinical mutations (F128 and L152, respectively).

(p)ppGpp and the SR have been implicated in a variety of virulence mechanisms in diverse pathogens, and there is now growing evidence to suggest that they also play an important role in antibiotic resistance and tolerance (4). Expression of β-lactam resistance among clinical isolates of methicillin-resistant Staphylococcus aureus (MRSA) is typically heterogeneous and low level; the population, as a whole, exhibits a relatively low MIC, but subpopulations of cells exist that exhibit very high levels of resistance (14). Activation of the SR by exposure to subinhibitory concentrations of mupirocin (an isoleucyl tRNA synthetase inhibitor) or the serine analog serine hydroxamate has been shown to induce high-level, homogeneous expression of β-lactam resistance in strains of MRSA (15–18). Furthermore, whole-genome sequencing of cells from highly resistant subpopulations has repeatedly revealed truncation mutations affecting Rel (16, 17, 19, 20). We have previously shown that two Rel mutations (F128Y and L152F) identified in clinical isolates of S. aureus and Enterococcus faecium associated with persistent bacteremias confer multidrug tolerance when introduced into a model strain of S. aureus (21). Antibiotic tolerance describes the ability of a bacterial population to survive in the presence of otherwise lethal concentrations of a bactericidal antibiotic without exhibiting an elevated MIC, usually as a consequence of reduced growth rate and/or longer lag time (22). While tolerance itself contributes to antibiotic treatment failures (23), it is also thought to act as a precursor to the emergence of endogenous resistance mutations (24–26). Therefore, it is plausible that Rel mutations that confer tolerance may also promote resistance development. The prevalence and importance of Rel mutations among clinical isolates associated with difficult-to-treat infections are only just now beginning to emerge, with two further recent reports of staphylococcal clinical isolates bearing nonsynonymous mutations in rel (27, 28).

In this study, we interrogate the effects of clinical Rel mutations at the molecular level (on the opposing catalytic activities of Rel) and the cellular level in terms of resistance expression, resistance development, and bacterial fitness. We show that these mutations dramatically reduce the rate of (p)ppGpp hydrolysis catalyzed by Rel and shift the overall balance of Rel catalysis in favor of (p)ppGpp synthesis. We also demonstrate that while these mutations confer significant antibiotic tolerance, they do not increase the mutation rate of S. aureus or accelerate the emergence of endogenous resistance mutations in vitro. Finally, we show that these clinical Rel mutations have a multitude of cellular consequences that have the potential to affect the success of antibiotic chemotherapy, including induction of high-level homogeneous expression of β-lactam resistance and a significant fitness advantage in the presence of antibiotics.

RESULTS

Molecular basis of elevated (p)ppGpp conferred by clinical Rel mutations.

Introduction of a clinical Rel mutation (F128Y or L152F) into S. aureus results in a higher cellular concentration of (p)ppGpp, a significant growth defect, and multidrug tolerance, with the F128Y mutation conferring a more pronounced phenotype than L152F (21). Using the available crystal structure of the N-terminal catalytic region of Rel from Streptococcus dysgalactiae (SdRel), these mutation sites were previously mapped to the C3HB of Rel, which has been shown to be important for hydrolase activity (9). Therefore, we and others hypothesized that these mutations cause elevated cellular (p)ppGpp levels by reducing hydrolase activity (21, 29, 30). However, recent structural and mutational analysis of Rel from Bacillus subtilis (BsRel) and RelA from E. coli (EcRelA) indicates a more complex role for this linker region in the allosteric activation and regulation of the two opposing catalytic activities (12, 31). Therefore, we set out to experimentally investigate the effect of the F128Y and L152F mutations on Rel hydrolase and synthetase activity. We introduced these two mutations into an E. coli expression construct that encodes full-length S. aureus Rel (SaRel; residues 1 to 729 with an N-terminal 6×His tag) (11) and used enzyme-coupled assays to determine the rates of Rel hydrolysis and synthesis (Fig. 1B). As previously reported (11, 13), wild-type SaRel exhibited far greater hydrolase activity than synthetase activity under the assay conditions used. In contrast, the F128Y mutant enzyme exhibited a dramatic decrease in hydrolase activity compared with the wild type and no significant change in synthetase activity, leading to an overall shift in favor of (p)ppGpp synthesis. The L152F mutant enzyme also displayed a reduction in hydrolase activity, coupled with a small but significant increase in synthetase activity. These changes result in a more even balance between hydrolase and synthetase activity, with a slight preference toward (p)ppGpp synthesis.

In light of the new structural information available, we reanalyzed the location of the two mutation sites within Rel using the crystal structure of BsRel (truncated after the TGS domain) (12) and an AlphaFold2 model of SaRel (32, 33) (Fig. 1C and D). These two proteins share 65% sequence identity across the crystallized BsRel fragment, and both F128 and L152 are conserved (Fig. 1E). In both structures, F128 is located within helix α7, which is part of the core hydrolase domain and precedes the C3HB (helices α8 to α10). When BsRel is in its relaxed, hydrolase-on/synthetase-off conformation, helix α7 is thought to be stabilized by the association of the C-terminal TGS domain with the synthetase domain and contribute to the allosteric regulation of Rel hydrolase activity (12). Indeed, mutation of R125 and M127 in BsRel (R125 and L127 in SaRel) led to a large reduction in hydrolase activity (and a small increase in synthetase activity) (12). Therefore, the significant decrease in hydrolase activity exhibited by our F128Y mutant enzyme is in line with the data reported for BsRel and supports the importance of helix α7 in modulating the activity of the hydrolase domain. Furthermore, the F128Y mutation did not affect hydrolase activity in a SaRel_CAT construct that lacks the entire regulatory C-terminal region, including the TGS (data not shown). This indicates that the F128Y mutation does not directly affect the hydrolase domain but, rather, influences the allosteric regulation of hydrolase activity.

In contrast to F128, L152 is located outside the main hydrolase domain and within the C3HB, either at the C-terminal end of helix α8 (SaRel model) or in the loop joining helices α8 and α9 (BsRel structure). Based on the structure and mutagenesis of SdRel, the C terminus of helix α8 and the α8/α9 loop region contribute important residues and H-bonding partners to the guanine binding cleft of the hydrolase active site, including T151 (9). Therefore, substitution of L152 for a bulkier phenylalanine would be expected to induce conformation shifts in this region of the protein that could affect hydrolase activity. Unfortunately, repeated attempts to express and purify an L152F mutant of the Rel_CAT construct were unsuccessful, so we were unable to compare the effect of this mutation on the regulated and unregulated forms of SaRel. It is interesting to note that the L152F mutation in SaRel also caused a small but significant increase in synthetase activity, even though residues within helix α8 and the N-terminal region of helix α9 are not thought to contribute to the synthetase active site directly (9). Roghanian et al. have recently shown that pppGpp binds to an allosteric site between the hydrolase and synthetase domains of EcRelA and “activates” (p)ppGpp synthesis by promoting the coordination of substrates in the synthetase active site for efficient catalysis (31). This pppGpp binding site includes the C terminus of helix α8 and the N terminus of helix α9. In particular, R161 in EcRelA (R150 in SaRel) is predicted to participate in the coordination of pppGpp, and mutation to alanine resulted in a 15-fold reduction in affinity. Therefore, the increased synthetase activity of the L152F mutant SaRel enzyme may be related to activation of synthetase activity via this allosteric pppGpp binding site. Overall, it is clear that activity and regulation of the two opposing activities of Rel—and subsequent cellular concentration of (p)ppGpp—can be affected by mutations within multiple different regions of the protein.

Clinical Rel mutations affect the expression of β-lactam resistance but do not accelerate the emergence of endogenous resistance.

The F128Y and L152F mutations partially activate the SR in S. aureus by shifting the balance of catalysis toward (p)ppGpp synthesis, which we have previously shown leads to antibiotic tolerance (21). However, (p)ppGpp and SR activation have also been implicated in various aspects of antibiotic resistance acquisition and expression (4). β-Lactam resistance expression in S. aureus is multifactorial, being dependent on the presence of several “auxiliary genes” (in addition to the presence of mecA or one of its homologs) (34). Among these auxiliary genes, rel has been identified as a crucial contributor to the expression of β-lactam resistance within highly resistant subpopulations. Premature stop codons that truncate Rel after the enzymatic N-terminal region confer elevated cellular (p)ppGpp content and homogeneous high-level β-lactam resistance (16, 17, 19). Therefore, we set out to determine whether F128Y and L152F would have a similar effect by introducing them into the methicillin-resistant strain USA300 LAC. As we observed with Newman, both Rel mutations imposed a growth defect on LAC, and the defect conferred by the F128Y mutation was slightly greater than that induced by L152F (see Fig. S1 in the supplemental material). In terms of resistance expression, isolates of the USA300 lineage have previously been shown to exhibit heterogeneous β-lactam resistance (19), and we observed the same phenotype with our wild-type strain (Fig. 2). The introduction of either Rel mutation led to an overall higher level of oxacillin resistance, as evidenced by Etest (Fig. 2A), and a shift toward a more homogeneous resistance expression profile by population analysis (Fig. 2B). The calculated area under the curve values for the mutants were significantly higher than those of the wild-type and complemented strains (P < 0.05, one-way analysis of variance [ANOVA] with Dunnett’s post-test). In accordance with its more pronounced SR-activated phenotype, the F128Y mutant exhibited a higher level of β-lactam resistance than the L152F mutant, with >50% of the population surviving at oxacillin concentrations up to 3 μg/mL compared with 1 μg/mL for the L152F mutant and 0.4 μg/mL for the wild-type and complemented strains. The MIC for ceftaroline against S. aureus USA300 LAC was unaffected by the mutations (data not shown).

FIG 2.

Expression of oxacillin resistance in wild-type USA300 LAC and Rel mutants. (A) Oxacillin Etest results for USA300 LAC, two Rel mutants, and their complemented counterparts. Results shown are representative of three repeat experiments. (B) Oxacillin population analysis profiles for wild-type, mutant, and complemented strains. Overnight cultures were diluted appropriately and plated on increasing concentrations of oxacillin. Colonies were counted after 48 h incubation. Data shown are the mean of three independent replicate cultures, each derived from a different starting colony. Error bars, where visible, represent the SEM.

As well as resistance expression, (p)ppGpp and SR activation have long been associated with increased bacterial mutability (35) and the emergence of resistance (4, 36, 37). Furthermore, tolerance-conferring mutations have been shown to precede resistance mutations in response to antibiotic exposure, both in in vitro evolution experiments with Escherichia coli (24) and during in-host evolution of S. aureus (26). Therefore, we investigated the rate of endogenous resistance development in our most SR-activated and tolerant mutant, F128Y. We exposed the F128Y mutant, alongside wild-type Newman, to repeat cycles of ciprofloxacin or daptomycin exposure and compared the rate of MIC increase; ciprofloxacin and daptomycin are both bactericidal antibiotics to which the F128Y mutant exhibits tolerance (21). The SR-activated mutant did not develop endogenous resistance faster than the wild type, with both strains taking 6 to 8 days for the MIC to exceed the susceptibility breakpoint (Fig. S2). We also performed fluctuation assays with wild-type Newman, F128Y, and the F128Y-complemented strain with rifampin and fusidic acid in order to compare their mutation rates. For both antibiotics, the mutation rate did not differ significantly between strains (Table S1).

SR activation provides a fitness advantage in the presence of antibiotics.

A number of reports of in-host evolution of clinical isolates bearing Rel mutations have noted that mutated and unmutated isolates appear to coexist at different points during infection (28, 30). This suggests that the consequences of Rel mutations in terms of fitness may fluctuate between cost and benefit, depending on the host environment (e.g., antibiotic exposure). We have previously shown that introduction of either the Rel F128Y or L152F mutation into S. aureus induces a significant growth defect and concomitant tolerance to five different classes of bactericidal antibiotics (21). Therefore, we hypothesized that these mutants would be outcompeted by the wild type in the absence of antibiotic in vitro but have a selective advantage in the presence of bactericidal antibiotic. To test this, we set up competition cocultures in broth containing no antibiotic or supra-MICs of daptomycin or ciprofloxacin (Fig. 3). Cocultures were inoculated with ~1 × 10⁷ CFU/mL of each competing strain, one of which was novobiocin resistant (NOV^r) due to the presence of an R144I mutation in gyrB (38). To account for any potential fitness effect conferred by the gyrB mutation, the strain that was NOV^r was varied between replicates. In the absence of antibiotic, both mutants were gradually outcompeted by the wild type, making up only ~25% of the population after 6 and 24 h (Fig. 3). In line with the less severe growth defect exhibited by the L152F mutant compared with the F128Y mutant (21), it took longer for the wild-type strain to outcompete the L152F mutant than F128Y (4 h versus 2 h). As expected, both mutants were also outcompeted by their complemented strains, and the wild-type and complemented strains maintained an ~50:50 ratio when cocultured (Fig. S3). When the two mutants were cocultured together in the absence of antibiotic, they remained equally prevalent for 2 to 4 h, but by 6 h, the L152F mutant had outcompeted the F128Y mutant and made up ~80% of the population by 24 h.

FIG 3.

Coculture competition assays between wild-type Newman, the Rel F128Y mutant, and the Rel L152F mutant in the absence and presence of antibiotic. Cocultures were inoculated with ~1 × 10⁷ CFU/mL of each strain (one of which was novobiocin-resistant [NOV^r]) into either drug-free TSB or TSB containing daptomycin (and Ca²⁺) or ciprofloxacin. Viable counts were performed at the times indicated. Data shown are the mean of four independent cocultures for each drug condition; in half of the replicates, the wild-type was NOV^r and the mutant was unmarked, and vice versa. Error bars represent the SEM. Lowercase letters above bars indicate statistically significant differences between means (multiplicity-adjusted P < 0.05) as determined by one-way ANOVA with Tukey’s multiple-comparison test (i.e., bars bearing the same letter have means that are not significantly different from one another). Raw CFU per milliliter data are shown in Fig. S4 in the supplemental material.

When the wild-type and F128Y mutant were inoculated into broth containing ciprofloxacin, F128Y quickly became the dominant strain, constituting ~80% of the population after 2 h. This dominance was maintained for at least 6 h, but by 24 h, the two strains were equally prevalent. A very similar trend was observed when the wild type and L152F mutant were competed in the presence of ciprofloxacin. In both cases, the return to an ~50:50 ratio by 24 h is consistent with the death of the mutant “catching up” with that of the wild type somewhere between 12 and 24 h, as we have previously observed in monoculture time-kill experiments (21). When the two mutants were competed against each other in the presence of ciprofloxacin, the F128Y mutant gradually outcompeted the L152F mutant, which is in line with its lower rate of killing in monoculture (21). The selective advantage of the two mutants over the wild type was also apparent in the presence of daptomycin. The F128Y mutant quickly outcompeted the wild type and remained the most prevalent strain for at least 6 h; we were unable to assess the ratio of strains in cocultures after 24 h of daptomycin exposure due to too few cells surviving. The L152F mutant also outcompeted the wild type, but, in general, the dominance of L152F over the wild type was not as great as that of F128Y (e.g., 67% versus 85% at 6 h for F128Y and L152F, respectively; P = 0.0043, unpaired two-tailed t test). When the two mutants were cocultured together under daptomycin exposure, they maintained a ratio close to 50:50 over the course of the experiment. Overall, it is clear that the fitness relationship between the wild-type and Rel mutant strains is complex and highly dependent on antibiotic exposure.

DISCUSSION

There is an increasing body of evidence to suggest that (p)ppGpp and the stringent response play an important role in bacterial virulence and contribute to persistent infections in the absence of antibiotic resistance, in particular, in S. aureus (4, 7, 21, 27–30, 39). The cellular concentration of (p)ppGpp in S. aureus (and most other pathogens) is tightly controlled by the opposing catalytic activities of Rel and its regulatory C-terminal domains (11, 12). Within an unstressed cell, Rel exists largely in a hydrolase-on/synthetase-off state, which results in a low basal level of (p)ppGpp. We have shown that two reported clinical Rel mutations that increase the cellular concentration of (p)ppGpp in S. aureus do so by shifting the balance of Rel catalysis in favor of (p)ppGpp synthesis. The F128Y mutation does this by drastically reducing the hydrolase activity of Rel, whereas the L152F mutation simultaneously affects both hydrolase and synthetase activity. The activity, allosteric activation, and regulation of Rel are incredibly complex and involve residues located within the core catalytic domains, C3HB, and C-terminal regulatory region. New clinical Rel mutations have been identified by whole-genome sequencing in recent years, including L533F in an isolate of Staphylococcus epidermidis associated with relapsing endocarditis and breakthrough bacteremia (27). D134Y, A301T, E384K, and V670G mutations have also been found in S. aureus isolates associated with persistent bacteremia (28). Some of these fall within the hydrolase and synthetase domains themselves (D134Y, A301T, and E384K), while others are located within the C-terminal regulatory region (L533F and V670G). Experimental testing would be required to determine whether these mutations affect the balance of Rel activity and the cellular (p)ppGpp concentration. Interestingly, a comparison of (p)ppGpp content between pairs of matched S. aureus isolates associated with persistent or resolving bacteremia revealed significantly higher (p)ppGpp levels in the persistent bacteremia isolates (39). It is clear that while Rel is an essential protein that is central to many important aspects of bacteria physiology, it is readily amenable to mutations that affect its catalytic balance.

Given the apparent prevalence of elevated (p)ppGpp and Rel mutations among staphylococcal clinical isolates, it is important to understand the cellular consequences in relation to antibiotic therapy. We found no association between elevated (p)ppGpp and mutation rate or the emergence of endogenous resistance mutations with our F128Y mutant. Previous studies that linked the SR to resistance development were performed with E. coli and Pseudomonas aeruginosa (36, 37), which utilize a different downstream SR cascade to S. aureus. In proteobacteria, SR-mediated modulation of transcription occurs primarily via binding of (p)ppGpp to RNA polymerase, which also affects DNA replication rates; in contrast, (p)ppGpp does not interact with RNA polymerase in S. aureus, and transcriptional changes occur indirectly as a result of decreased cellular GTP or binding of (p)ppGpp to riboswitches (2). Therefore, the lack of increased mutability in our SR-activated mutant may be related to its alternate pathway of SR metabolic control. Indeed, a study of the SR and mutability in B. subtilis (which utilizes a similar SR cascade to S. aureus) reported no difference in the rate of streptomycin resistance between rel⁻ and rel⁺ strains (40). Aside from the SR, the F128Y mutation confers antibiotic tolerance (21), and other tolerance-conferring mutations have been shown to precede the emergence of resistance mutations in S. aureus (26). We saw no evidence of this in our F128Y mutant and can only speculate that this difference may be due to variation in the underlying mechanisms behind tolerance in these different strains (none of the tolerance mutations that have been reported to precede resistance mutations occurred in Rel).

While the F128Y mutation does not affect resistance development, it does influence resistance expression in S. aureus. Acquisition of the β-lactam resistance determinant mecA by strains of S. aureus is of major clinical importance, as it confers resistance to most β-lactam antibiotics, and its optimal expression is controlled, at least in part, by Rel (16, 17, 19, 20). Both the F128Y and L152F mutations induced high-level, homogeneous β-lactam resistance when introduced into USA300 LAC. This finding agrees with previous studies in which elevated (p)ppGpp led to homogeneous expression via increased transcription of mecA and production of PBP2A (15, 17, 18). In a clinical setting, the significance of homogeneous β-lactam resistance expression is not clear, as such a strain would be readily identified as resistant and a different antibiotic prescribed. However, cases of oxacillin-susceptible MRSA (OS-MRSA)—where isolates possess mecA but are phenotypically susceptible to β-lactams—are increasing worldwide (41). These isolates readily become resistant upon exposure to β-lactams at rates similar to those associated with spontaneous resistance to rifampin (~10⁻⁷) (42). A recent survey of the genetic basis of this resistance found that mutations that truncate Rel after the synthetase domain are common among these evolved resistant strains (43). In fact, a Rel truncation mutation alone is sufficient to increase the oxacillin MIC by 24-fold. Therefore, exposure of OS-MRSA strains to β-lactams has the potential to select for SR-activating Rel mutations, which will simultaneously induce β-lactam resistance expression and multidrug tolerance. Conversely, antibiotic exposure that selects for Rel mutations that confer tolerance would inadvertently induce β-lactam resistance. This relationship may go some way to explaining the poorer clinical outcomes associated with bacteremia caused by OS-MRSA compared with MRSA, including a higher occurrence of persistent bacteremia and more frequent failure of vancomycin monotherapy (with the need for combination salvage therapy) (44).

Finally, we investigated the effect of both the F128Y and L152F mutations on bacterial fitness, both in the presence and absence of antibiotic. In accordance with their known growth defects (21), both mutants were rapidly outcompeted by the wild-type in the absence of antibiotic. In the presence of an antibiotic to which the mutants are tolerant, however, both mutants demonstrated a significant fitness advantage and quickly outnumbered the wild type. During an infection, a bacterial population is exposed to cyclical and variable concentrations of antibiotic, which would be expected to differentially and temporally favor the growth of the tolerant or wild-type strain (with supra-MICs favoring a tolerant but growth-impaired mutant and sub-MICs favoring the wild type). Therefore, our fitness data are in line with two reports of clinical infections caused by bacteria bearing Rel mutations that have noted reisolation of the wild-type strain after emergence of the Rel mutant (28, 30). It is clear from these clinical examples and our own data that the selective advantage conferred by Rel mutations is multifaceted and situation dependent and does not always outweigh the growth defect of the mutant. An important next step in assessing the overall importance and clinical consequences of Rel mutations in relation to antibiotic efficacy will be their evaluation in in vivo models of infection.

MATERIALS AND METHODS

Reagents and antibiotics.

X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside), chloramphenicol, GDP, GTP, ADP, ATP, NAD⁺, and NADH were purchased from Bio Basic Inc. (Markham, ON, Canada). Daptomycin and ppGpp were obtained from Toronto Research Chemicals (Toronto, ON, Canada) and Jena Bioscience (Jena, Germany), respectively. All other antibiotics, enzymes, and reagents, unless otherwise stated, were from MilliporeSigma (Burlington, MA).

Bacterial strains, plasmids, and growth conditions.

S. aureus strains Newman and USA300 LAC were gifts from Michael Murphy (University of British Columbia) and Paul Kubes (University of Calgary), respectively. The Newman mutants F128Y and L152F and their complemented counterpart (F128Y comp and L152F comp) have been reported previously (21). Selection and sequencing of novobiocin-resistant mutants of these strains are detailed below. Escherichia coli Stellar (TaKaRa Bio USA, Mountain View, CA) was used for all cloning, while E. coli IM08B (45) was used to prepare methylated plasmid for transformation into S. aureus. Plasmids pCG511, pCG512 (11), and pIMAY-Z (45) were gifts from Christiane Wolz (University of Tüebingen) and Ian Monk (University of Melbourne, Australia), respectively. S. aureus strains were routinely cultured in tryptic soy broth/tryptic soy agar (TSB/TSA) at 37°C and stored long term at −80°C with 8% (vol/vol) glycerol.

Allelic exchange and mutagenesis.

The amino acid sequence of Rel is identical between S. aureus strains Newman and USA300 LAC (GenBank accession number HUW68_RS08510); therefore, the F128Y and L152F mutations were introduced into LAC and complemented using our Newman allelic exchange constructs and protocol (21). The F128Y and L152F mutations were introduced into the Rel expression constructs pCG511 and pCG512 by site-directed mutagenesis using Phusion polymerase (New England Biolabs, Ipswich, MA) and the following primers: F128Y fwd, 5′-ACAACAAGCTGAAAATCATCGCAAGTTATATATTGCGATTGC CAAAG-3; F128Y rev, 5′-CTTTGGCAATCGCAATATATAACTTGCGATGATTTTCAGC TTGTTGT-3′; L152F fwd, 5′-ATAATATGCGTACCTTCAAAGCCATGCCGCGCG-3′; and L152F rev, CGCGCGGCATGGCTTTGAAGGTACGCATATTAT-3′.

Selection and sequencing of novobiocin-resistant strains.

Novobiocin-resistant (NOV^r) derivatives of Newman strains were selected by plating ~10¹⁰ cells on TSA containing 1 μg/mL novobiocin and incubating at 37°C for 48 h as previously described (38). The mutations conferring NOV^r in different colonies were determined by PCR amplification and sequencing of gyrB using CloneAmp HiFi PCR premix (TaKaRa Bio USA, Mountain View, CA) and the following primers: gyrB fwd, 5′-GAAATTAT AAAGTAACAGAAAGCGATGG-3′, and gyrB rev, 5′-GTAATTCAGCCATCAAGAGTTCC-3′.

Rel enzyme kinetics.

Wild-type and mutant Rel proteins were expressed and purified as previously described (11). Purified proteins were assessed for purity by SDS-PAGE analysis, exhaustively dialyzed against Rel buffer (20 mM HEPES, pH 7.0, 20 mM NaCl, 1 M KCl, 20 mM MgCl₂, and 30% [vol/vol] glycerol) at 4°C, and concentrated using an Amicon stirred ultrafiltration unit fitted with a 10-kDa-molecular-mass-cutoff cellulose membrane. Protein concentrations were determined using extinction coefficients calculated by ProtParam on the ExPASy server (46). The hydrolase activity of Rel and its mutants were determined in an enzyme-coupled assay as previously described (47) in which the GDP produced by Rel upon hydrolysis of ppGpp is linked to the oxidation of NADH to NAD⁺. Hydrolase reactions contained 1 mM phosphoenolpyruvate, 200 μM NADH, 0.55 U lactate dehydrogenase, 0.85 U pyruvate kinase, 250 μM ppGpp, and 2 μM Rel in hydrolase assay buffer (50 mM Tris, pH 7.0, 200 mM NaCl, 10 mM MgCl₂, 1 mM MnCl₂, and 20 mM KCl). Reactions were performed in triplicate in 96-well plates at 37°C in a SpectraMax M5 plate reader, with continuous monitoring of the oxidation/production of NADH at 340 nm. Initial rates were converted to micromolar per minute using an extinction coefficient of 4,900 M⁻¹cm⁻¹ for NADH that was experimentally determined in hydrolase assay buffer. The synthetase activity of Rel and its mutants was compared using the AMP-Glo assay kit (Promega Corporation, Madison, WI). Reaction mixtures containing 100 μM ATP, 100 μM GTP, and 2 μM Rel in synthetase assay buffer (50 mM Tris, pH 7.0, 200 mM NaCl, 20 mM MgCl₂, and 20 mM KCl) were set up in a 384-well white plate and incubated at 37°C for 30 min. Reactions were terminated by the addition of AMP-Glo reagent I and incubated at room temperature for 60 min to degrade any remaining ATP and convert the AMP produced by Rel into ADP. AMP detection solution was then added, the plate was incubated again at room temperature for 60 min, and, finally, the luminescence was read in a SpectraMax M5 plate reader. Luminescence values were converted to AMP concentrations by comparison with a standard curve of AMP generated in synthetase assay buffer. All statistical analyses were performed in GraphPad Prism 6.08.

Growth curves.

Growth curves were determined in 25 mL TSB in 125-mL Erlenmeyer flasks in a shaking water bath at 37°C and 180 rpm. Overnight cultures (50 μL) derived from different colonies were used to inoculate triplicate flasks. At the indicated time intervals, 100 μL samples were transferred to a flat-bottom 96-well plate, and the optical densities at 600 nm (OD₆₀₀) were read using a SpectraMax M5 plate reader. OD₆₀₀ readings were corrected for a path length of 0.28 cm. Specific growth rates (μ) were determined from the slope of ln(OD₆₀₀) versus time during exponential phase, and converted into doubling times using the equation ln(2)/μ. Lag times were found by extrapolating the slope of the exponential phase back to the initial OD₆₀₀.

Etest and population analysis profile determinations.

Oxacillin and ceftaroline Etest strips (bioMérieux, Saint-Laurent, QC, Canada) were applied to Mueller-Hinton agar (MHA) plates, inoculated with cultures grown overnight in Mueller-Hinton broth (MHB) according to the manufacturer’s instructions, and incubated at 37°C for 24 to 48 h. For population analysis profiles, strains were grown overnight in MHB at 37°C, diluted in phosphate-buffered saline (PBS), and plated on MHA containing increasing concentrations of oxacillin. Plates were incubated at 37°C for 48 h before counting. Area under the curve calculations were performed using GraphPad Prism 6.08.

Time-to-resistance assays.

The number of cycles of drug exposure and recovery required to increase the MIC of daptomycin or ciprofloxacin above the susceptibility breakpoint was determined for wild-type Newman and the F128Y mutant. For daptomycin, the method of Friedman et al. (48) was followed, with daptomycin concentrations ranging from 0.188 to 5 μg/mL and six replicates per strain. For ciprofloxacin, a time-kill and recovery method was employed. Exponentially growing bacteria were exposed to ciprofloxacin at 4× MIC for 2 h and then pelleted, resuspended in drug-free MHB, and allowed to recover overnight. The MIC was determined following each recovery step, and 12 replicates were performed per strain.

Fluctuation assay.

The mutation rates of Newman wild type, F128Y, and F128Y comp were determined via fluctuation assays (49) with fusidic acid and rifampin. A single colony of each was used to inoculate 5 mL TSB and grown overnight at 37°C with shaking. This culture was then diluted 10⁻⁵ in TSB, 100 μL aliquots were placed into 36 wells of a round-bottom 96-well microplate, and the plate was sealed with Breathe-Easy sealing membrane. Following overnight incubation at 37°C, the entire 100 μL from 16 wells was spread plated on TSA containing 1 μg/mL fusidic acid or 0.5 μg/mL rifampin. The cultures from the remaining four wells were diluted to 10⁻⁷ in PBS and plated on TSA for viable counting. Viable counts were determined after overnight incubation at 37°C, while drug plates were incubated for 48 h before counting. The number of mutations per culture, m, for each strain-drug pairing was calculated using the Jones median estimator (50) and divided by the mean viable count (N_t) to give the mutation rate, μ (49). The 95% confidence limits for the median were calculated as previously described (49).

Coculture competition assay.

Overnight cultures of competing strains were inoculated (5 μL) into the same 5 mL TSB (containing 8 × MIC ciprofloxacin or daptomycin and Ca²⁺ when indicated) and grown at 37°C with shaking as previously described (51); in each coculture, one strain was NOV^r. Culture samples were taken immediately after inoculation and at intervals thereafter, diluted in PBS, and plated in triplicate on TSA and TSA containing 1 μg/mL novobiocin. Following incubation of plates at 37°C for 16 h, colonies on the two types of agar were recorded; when colony numbers were similar, counts were corroborated by replica plating colonies from TSA plates onto TSA containing novobiocin. For each strain pairing and drug condition, four independent cocultures were performed using different overnight cultures as inocula; in half of these replicate cocultures, the mutant strain was NOV^r, and in the other half, the wild-type or complemented strain was NOV^r.