Early agr activation correlates with vancomycin treatment failure in multi-clonotype MRSA endovascular infections

Abstract

Objectives

Persistent MRSA infections are especially relevant to endovascular infections and correlate with suboptimal outcomes. However, the virulence signatures of Staphylococcus aureus that drive such persistence outcomes are not well defined. In the current study, we investigated correlations between accessory gene regulator (agr) activation and the outcome of vancomycin treatment in an experimental model of infective endocarditis (IE) due to MRSA strains with different agr and clonal complex (CC) types.

Methods

Twelve isolates with the four most common MRSA CC and agr types (CC5-agr II, CC8-agr I, CC30-agr III and CC45-agr I) were evaluated for heterogeneous vancomycin-intermediate S. aureus (hVISA), agr function, agrA and RNAIII transcription, agr locus sequences, virulence and response to vancomycin in the IE model.

Results

Early agr RNAIII activation (beginning at 2 h of growth) in parallel with strong δ-haemolysin production correlated with persistent outcomes in the IE model following vancomycin therapy. Importantly, such treatment failures occurred across the range of CC/agr types studied. In addition, these MRSA strains: (i) were vancomycin susceptible in vitro; (ii) were not hVISA or vancomycin tolerant; and (iii) did not evolve hVISA phenotypes or perturbed δ-haemolysin activity in vivo following vancomycin therapy. Moreover, agr locus sequence analyses revealed no common point mutations that correlated with either temporal RNAIII transcription or vancomycin treatment outcomes, encompassing different CC and agr types.

Conclusions

These data suggest that temporal agr RNAIII activation and agr functional profiles may be useful biomarkers to predict the in vivo persistence of endovascular MRSA infections despite vancomycin therapy.

Introduction

Staphylococcus aureus is the most common cause of endovascular infections, including infective endocarditis (IE), and the second most frequent cause of bacteraemia.^1,2 Invasive infections such as IE, particularly those caused by MRSA strains, are associated with a high mortality (15%–40%).^3–5 In addition, the high rates of clinical failure on vancomycin (even with CLSI-defined ‘susceptible’ strains) and the growing problem of daptomycin non-susceptibility, have further complicated the management of patients with invasive MRSA infections.^5,6

The pathogenesis of S. aureus is largely controlled by regulatory networks composed of: (i) global regulons (e.g. sarA); (ii) two-component regulatory systems (e.g. agr and saeRS); and (iii) downstream effector genes (e.g. hla).^7,8 Central to the orchestration of this regulatory network is the quorum-sensing operon agr. Of note, agr consists of two transcription domains, RNAII and RNAIII, driven by the divergent promoters P2 and P3, respectively.⁹ RNAII encodes a quorum-sensing system and RNAIII is the effector molecule of the agr system. In vitro, agr down-regulates many pivotal cell surface proteins involved in tissue colonization while up-regulating exoproteins associated with host cell damage.⁹ In vitro studies have suggested that the temporal activation of agr is an important determinant of agr functionality; i.e. earlier versus later agr activation correlates with intact versus attenuated function, respectively.^6,10 As most of the available data on agr expression and regulatory impacts have only been obtained in vitro, an important question relates to how such temporal regulatory functions translate to the in vivo scenario, particularly in the context of endovascular MRSA infections that persist versus resolve after antibiotic treatment.

We previously demonstrated a positive correlation between early-onset agr RNAIII activation and resistance to vancomycin treatment in an experimental model of IE.⁶ Although intriguing, these initial studies were somewhat limited: (i) they were restricted to strains from only two clonotypes (CC45-agr I and CC5-agr II); and (ii) they did not account for other genotypic influences on outcomes such as variations of agr types. Thus, it remains unknown whether the relationship between the temporal RNAIII activation profiles and antibiotic treatment outcomes is CC and/or agr type-specific in endovascular infection. Therefore, in the current investigation, we sought to validate and extend upon our prior observations, using a broad range of the most common clonotypes causing invasive MRSA infections such as IE.^11–13 The overarching goals of this project were: (i) to assess key genetic strategies used by MRSA to resist antimicrobial therapy (e.g. early agr activation); and (ii) to identify novel biomarkers that correlate with the phenotype of persistent endovascular MRSA infection.

Materials and methods

Bacterial strains and growth conditions

In the current study, we selected two CC5-agr II, four CC8-agr I and two CC30-agr III MRSA clinical isolates; in addition, we chose four previously sequenced reference strains of S. aureus (one for each CC type, including CC45) as relevant genetic controls (B. N. Kreiswirth, unpublished data; Table 1). These strains were from a multinational clinical trial collection and other sources¹⁴ and were pre-selected based on their relative δ-haemolysin activities as outlined below. Unless otherwise stated, all S. aureus strains were grown at 37°C either in tryptic soy broth (TSB; Difco) or TSB agar plates. Stocks were kept at –80°C in brain heart infusion (Difco) broth supplemented with 10% glycerol.

Table 1.

Genotypic characteristics of and vancomycin MICs for study isolates

Strains	CC	agr group	SCCmec	spa type (motif)	Source	Location (reference)	Year	Vancomycin MIC (mg/L)
Clinical isolates
26997	5	II	IV	2 (TJMBMDMGMK)	blood	USA	2009	1.0
31190	5	II	II	2 (TJMBMDMGMK)	sputum	USA	1996	1.0
29439	8	I	IV	1037 (YHGFMBO)	SSTI	Colombia	2008	1.0
30279	8	I	VIII	1 (YHGFMBQBLO)	SSTI	USA	2010	2.0
31082	8	I	IV	139 (YGFMBLO)	blood	USA	2009	2.0
30568	8	I	IV	7 (YHGCMBQBLO)	blood	USA	2010	2.0
33367	30	III	IV	19 (XKAKAOMQ)	wound	Romania	2005	1.0
17130	30	III	II	16 (WGKAKAOMQQQ)	SSTI	USA	2007	1.0
Reference strains
36096	5	II	IV	2 (TJMBMDMGMK)	blood	USA	2010	2.0
19069 (USA300-FPR3757)	8	I	IV	1 (YHGFMBQBLO)	abscess	USA	ND	2.0
22033 (WBG10049)	30	III	IV	19 (XKAKAOMQ)	ND	Australia	1999	2.0
21314	45	I	MSSA	XKAKBEMBKB	blood	France	2002	2.0

ND, no data available; SSTI, skin and soft tissue infection.

agr, SCCmec, spa, CC typing and agr locus sequence

The agr group was determined by a multiplex PCR assay described elsewhere,¹⁵ while the SCCmec type was examined using a multiplex real-time PCR published previously.¹⁶ The spa typing was conducted according to a protocol described before; the corresponding clonal complex (CC) of each strain was inferred based on the spa types.¹⁷ Full-length agr locus sequencing was performed as previously described.⁶

MICs, kill curves and population analyses

The determination of vancomycin MICs was conducted by broth microdilution as recommended by the CLSI.¹⁸ In vitro, vancomycin kill curves and population analyses were carried out to detect vancomycin tolerance and heterogeneous vancomycin-intermediate S. aureus (hVISA), respectively.¹⁹ All in vitro experiments were performed at least twice for each strain on different days.

δ-Haemolysin activity

δ-Haemolysin activity was assayed by cross-streaking the test strains perpendicularly to RN4220, a strain that is a hyperproducer of β-haemolysin but that does not produce α-haemolysin.⁶ Strains SH1000 and SH1001 (the agr mutant of SH1000) were used as positive and negative controls, respectively. δ-Haemolytic activity was denoted by an enhanced area of haemolysis at the intersection of the RN4220 streak and the test strain streaks as described.⁶ In addition, the δ-haemolysin activity of each strain was semi-quantitatively scored by two investigators involved in this study who were blinded to the isolate numbers, as follows: non-detectable, weak and strong. All experiments were conducted at least three times on separate days.

agrA and RNAIII transcription by quantitative RT–PCR

It is well accepted that transcription from P2 and P3 occurs in a strictly AgrA-dependent manner. However, recent studies have identified that core virulence factors in various S. aureus strains depend either on RNAIII or AgrA for regulation.²⁰ Thus, in the current study, we defined the temporal transcription profiles of both agrA and RNAIII (as surrogate biomarkers of global agr operon activation) by qRT–PCR analyses.⁶ In brief, overnight cultures of S. aureus were diluted 100-fold in fresh TSB medium and grown at 37°C in a shaking incubator. Cells were harvested at 2, 3, 4, 8 and 24 h of growth, representing the early-, mid-, late-, post-exponential phases, and the stationary phase, respectively. Quantitative real-time PCR was carried out using an ABI Prism 7000 instrument (Applied Biosystems) and the SYBR green PCR master mix (Applied Biosystems). Reaction mixtures were prepared using the 100 nM primers listed in Table 2. A well-characterized gene, gyrB, was used as an internal control.^6,21 Relative target gene expression was calculated as the differences in cycle thresholds (ΔC_T) (gyrBC_T – target gene C_T) for all samples.²¹ qRT–PCR experiments were performed using at least two biological replicates, with each one tested in triplicate.

Table 2.

Primers used in this study

Primer	Sequence (5′–3′)	Reference
gyrB-F	CGCAGGCGATTTTACCATTA	6
gyrB-R	GCTTTCGCTAGATCAAAGTCG	6
RNAIII-F	GCCATCCCAACTTAATAACCA	6
RNAIII-R	TGTTGTTTACGATAGCTTACATGC	6
agrA-F	CGAAGACGATCCAAAACAAAG	48
agrA-R	ATGTTACCAACTGGGTCATGC	48

Experimental rabbit IE model

It was important to assess the potential translatability of our in vitro findings in a realistic and relevant infection model in vivo. A catheter-induced model of IE affecting the aortic valve in rabbits was used to study the composite metrics of virulence and responsiveness to vancomycin therapy among the S. aureus strains used in the study.^6,22 The animals were maintained in accordance with the American Association for Accreditation of Laboratory Animal Care criteria. The Animal Research Committee (IACUC) of the Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center approved all the animal studies.

To assess the relative intrinsic virulence of the strains, infected animals (∼10⁵ cfu, representing the ID₉₅ inoculum) were euthanized at 24 h after intravenous challenge by a rapid intravenous injection of sodium pentobarbital (200 mg/kg; Abbott Laboratories). The cardiac vegetations, kidneys and spleen were then removed under sterile conditions and quantitatively cultured.^6,22 The tissue densities of S. aureus were calculated as the mean log₁₀ cfu/g (±SD).

To assess the response to vancomycin therapy, animals with IE were randomized at 24 h post-infection to receive either no therapy (controls) or vancomycin therapy (15 mg/kg, intravenously, twice daily for 3 days). This is a standard effective dose of vancomycin in the experimental model of IE caused by vancomycin-susceptible staphylococcal strains.^6,22 At 24 h after the last dose of vancomycin, the animals were sacrificed and the target tissues removed and quantitatively cultured as described above.

In vivo population analysis

An in vivo population analysis assay was performed to determine whether the in vivo emergence of the hVISA phenotype might explain the subsequent impact on either the virulence of the organism or the responsiveness to the antibiotic. A volume of 10 μL of vegetation homogenate was plated directly onto TSA agar plates containing serial concentrations of vancomycin (0.125–8 mg/L) as described above.

In vivo assessment of δ-haemolysin activity

Similar to the hVISA analyses above, it was important to test whether δ-haemolysin activity among the infecting strains was altered during in vivo passage. Thus, an in vivo assessment of δ-haemolysin production was performed by cross-streaking vegetation homogenates perpendicularly to RN4220 as described above.

Statistical analysis

To compare tissue counts of S. aureus in vancomycin-treated versus untreated control animals, univariate analyses were performed using the Student's t-test. P values of <0.05 were considered statistically significant.

Results

In vitro susceptibility to vancomycin

Vancomycin MICs for the study strains were all in the susceptible range as outlined by the CLSI guidelines (Table 1). In vitro population analyses of the isolates revealed no hVISA phenotypes [Figure S1 (A1, B1 and C1), available as Supplementary data at JAC Online; data not shown for the four reference strains]. Killing curve assays demonstrated no evidence of in vitro vancomycin tolerance (data not shown).

In vitro δ-haemolysin activities

A total of two CC5, three CC8, one CC30 and one CC45 isolates produced strong zones of haemolysis due to δ-haemolysin activity (Figure S2). In contrast, the other five strains exhibited weak or non-detectable δ-haemolysin production (Figure S2).

agrA and RNAIII transcriptions

The initiation of transcription of RNAIII strictly relies upon the expression of agrA, in an RNAII-dependent manner. Our data support these observations as the transcription of agrA was observed in all strains with RNAIII expression except strain 30568 (Figure 1a and b for agrA and RNAIII expression, respectively). In addition, a rapid increase in RNAIII transcription occurred in the post-exponential growth phase, concomitant with a cell-density-dependent expression of the agr operon (Figure 1b). More importantly, the seven strains with strong δ-haemolysin activity (one CC5, two CC8 and one CC30 clinical isolate, and three of the four reference strains) exhibited a relatively early onset of RNAIII expression, beginning at 2–3 h of growth (Figure 1b). The other five strains had very low-level or no RNAIII transcripts detectable before 4 h (Figure 1b). These data indicated that the early onset of RNAIII transcription correlated with strong δ-haemolysin production, independent of the CC/agr genotype. As the growth rates for all the study strains were similar, the observed differences in the onset of RNAIII expression were not due to differences in growth dynamics (data not shown).

Figure 1.

In vitro expression of agrA (a) and RNAIII (b) in all S. aureus strains studied at 2, 3, 4, 8 and 24 h of growth. Data were obtained by RT–PCR and relative transcript levels of agrA and RNAIII represent the mean (+SD) of at least two biological replicates (fold changes versus gyrB). Strains with strong δ-haemolysin activity are underlined.

agr locus sequence analyses

We sequenced the entire agr locus of the 12 study strains (including RNAII, RNAIII and their respective promoter regions; Table 3). The results in two CC5-agr II strains (exhibiting strong δ-haemolysin activity and an early onset of RNAIII activation) revealed identical agr locus sequences but these differed from the agr sequence reference strain N315 by only one nucleotide between agrD and agrC. The agr sequence of another CC5-agr II strain, 31190 (having weak δ-haemolysin activity and a late onset of expression of RNAIII), differed by one nucleotide from these two strains within the intergenic region between hld and agrB. However, it is not clear whether this agr mutational event was related to the different δ-haemolysin activity observed in this strain set. Two CC8-agr I (31082 and 30568) and two CC30-agr III (33367 and 22033) strains with different δ-haemolysin activities had the same agr sequence as their respective reference strains. Taken together, these results indicate that no consensus genetic determinants in the agr locus were identified as being predictive of agr functionality crossing different CC and agr genetic types.

Table 3.

agr locus sequence, δ-haemolysin and RNAIII activation of the study strains

Strains	CC	agr group	δ-Haemolysin activity	RNAIII activation	agr sequence	Reference strains
Clinical isolates
26997	5	II	strong	early	T→A between agrD and agrC	N315
31190	5	II	weak	late	T→A between agrD and agrC, G→A between hld and agrB,	N315
29439	8	I	strong	early	T→A in agrC (F309I)	USA 300
30279	8	I	weak	late	A→G in RNAIII, T→A in P3 promoter region, A→T in agrB (I171F), T→A in agrC (F309I)	USA 300
31082	8	I	strong	early	T→A in P3 promoter region, T→A in agrC (F309I)	USA300
30568	8	I	none	none	T→A in P3 promoter region, T→A in agrC (F309I)	USA300
33367	30	III	strong	early	RNAIII 386A del, agrC (R55G)	MRSA 252
17130	30	III	weak	late	none	MRSA 252
Reference strains
36096	5	II	strong	early	T→A between agrD and agrC	N315
19069 (USA300-FPR3757)	8	I	strong	early	none	USA300
22033 (WBG10049)	30	III	weak	late	RNAIII 386A del, agrC (R55G)	MRSA 252
21314	45	I	strong	early	none	CA-347 (CP006044)

Intrinsic virulence and responsiveness to vancomycin in the IE model

The intrinsic virulence of the study strains in the IE model (based on achievable target-tissue counts of S. aureus) was similar among the study isolates (Figure 2). Therefore, there was no obvious relationship between the above agr metrics, CC types and intrinsic virulence of the strains studied in this model of infection.

Figure 2.

S. aureus densities in the target tissues in the IE model in the presence versus absence of vancomycin therapy for 3 days (panels a, b and c represent clinical CC5, CC8 and CC30 S. aureus strains, respectively, and panel d represents four reference strains). Each dot represents one rabbit and the horizontal black bars indicate the means of the observations. Strains with strong δ-haemolysin activities are underlined. VAN, vancomycin.

In contrast, our data revealed a direct correlation between temporal agr RNAIII expression in vitro and vancomycin therapy in vivo. For example, the vancomycin therapy of animals infected with the five strains that exhibited a late onset of RNAIII and weak or no δ-haemolysin activity in vitro (clinical isolates 31190, 30279, 30568 and 17130, and reference strain 22033) resulted in uniform and highly significant reductions in MRSA counts in all the target tissues (Figure 2). Thus, we observed a ≥5 log₁₀ cfu reduction in the vegetation counts and ≥3 log₁₀ cfu reduction in the kidney and spleen counts compared with the respective untreated control groups (P < 0.0001 for all strains and all three target tissues). In contrast, animals with IE caused by strains with strong δ-haemolysin activity and an early onset of RNAIII transcription showed one of two outcomes: (i) no response to vancomycin treatment, with residual target-tissue MRSA densities similar to those in the respective untreated controls (the 33367 clinical isolate and the 36096 reference strain; Figure 2); or (ii) a <3 log₁₀ cfu reduction in target-tissue counts (clinical isolates 26997, 29439 and 31082, and reference strains 19069 and 21314; Figure 2). Importantly, there was no significant difference in the weight of the vegetations between the groups (data not shown). Therefore, the microbiological difference in vegetation counts was not impacted by the differences in structure or size of the vegetations.

In vivo population analyses

Figure S1 (A2, B2, B3 and C2) shows the in vivo population analyses from vegetation homogenates of the study clinical isolates surviving vancomycin therapy in the IE model (data not shown for the four reference strains). No subpopulations that were hetero-resistant to vancomycin were observed in the vegetations from any animals (Figure S1; data not shown for the control animals without treatment). We were unable to test for potential hVISA phenotypes in strains highly susceptible to vancomycin therapy in the IE model due to the low residual bacterial counts of the vegetations (<3 log₁₀ cfu/g).

In vivo δ-haemolysin activity

No changes in profiles of δ-haemolysin activity were observed in any isolates after passage through the IE model (Figure S2; data not shown for the four reference strains).

Discussion

S. aureus is a highly adaptable bacterium capable of dynamic changes in its virulence and resistance phenotypes in the face of exposure to host defences or administered antibiotics. As a result of such capabilities, ‘persistent’ endovascular MRSA infections frequently occur despite seemingly appropriate antibiotic therapy. This reality represents a daunting and increasingly life-threatening clinical syndrome. The fundamental phenotypic and genotypic virulence and resistance ‘signatures’ of S. aureus that drive such persistent clinical outcomes are not well defined. Thus, the identification of potential biomarkers that specifically characterize persistent S. aureus strains is urgently needed. The current study was designed to contribute new insights that might help to address this void in knowledge.

The agr operon plays an important and complex role in the pathogenesis of MRSA infections.^9,23,24 In addition, polymorphisms in agrD and agrC have been utilized to genotypically define the major S. aureus agr groups (I–IV).^25–27 Several previous studies have suggested an association of specific agr groups with distinct clinical syndromes involving S. aureus.^25,26 For example, agr group II has been associated with a poor response to vancomycin therapy in patients with MRSA infections.²⁸ Of note, such linkage of the agr group II genotype to suboptimal treatment outcomes has not been verified in more recent reports using a rabbit model of IE.⁶ Interestingly, in addition to agr genotypes, four CC types (CC5, CC8, CC30 and CC45) are reported to predominate in serious S. aureus infections.^29,30 Moreover, distinct CC types show a differing potential to cause invasive disease.³¹ For instance, CC5 and CC30 have been reported to be associated with more severe infections.¹¹ Furthermore, Miller et al.³² recently correlated the CC22 genotype with cases of IE. Conversely, Feil et al.³³ reported that disease isolates were equally represented across all CC types, suggesting that the link between CC genotype and the propensity to cause specific disease syndromes is at best inconsistent. The above examples illustrate how the genotypic and phenotypic plasticity of S. aureus is likely to afford a variety of survival strategies and advantages in distinct host defence settings. Thus, rather than seeking a single gene or distinct host phenotype that enables persistence, our work seeks to identify and validate more integrated signals of adaptive responses in S. aureus that are associated with virulence and antimicrobial resistance.

Vancomycin has remained a stalwart for the treatment of invasive MRSA infections. We recently studied the relationship of agr RNAIII expression with vancomycin treatment outcomes in experimental IE due to 10 clinical MRSA strains with two distinct genetic backgrounds (CC5-agr II and CC45-agr I). These investigations demonstrated that early-onset agr RNAIII activation correlated with failures of vancomycin treatment in this model.⁶ Therefore, our current studies were designed to further investigate the question of whether this putative correlation between the temporal agr RNAIII activation profiles and the outcome of antibiotic treatment is CC-agr type specific or more generic across CC-agr types. In this regard, our strategy was strengthened by the use of clinical MRSA isolates representing the most common CC-agr types (CC5-agr II, CC8-agr I and CC30-agr III) in comparison with four respective S. aureus reference strains.

Several key insights emerged from this study. We found that the overall temporal transcription of RNAIII was strain dependent but CC/agr type independent. Most importantly, consistent with our previous observation,⁶ we found that early-onset agr RNAIII activation (beginning at 2–3 h of growth) significantly correlated with reduced vancomycin efficacy in experimental IE. The exact mechanism(s) by which early agr activation provides the organism with reduced vancomycin susceptibility in vivo in this model is yet to be defined. However, in this regard, we have recently confirmed that strains exhibiting early RNAIII activation (which are vancomycin non-responders in the IE model) differed substantially from the strains showing later RNAIII expression (vancomycin responders in the same model) in terms of: (i) lower vancomycin binding; (ii) increased intrinsic biofilm formation; (iii) higher survival in the presence of vancomycin within biofilms in the presence or absence of catheters; (iv) enhanced biofilm formation during vancomycin exposure below the MIC; (v) significantly greater damage to endothelial cells;^19,34 and (vi) reduced killing by certain host defence antimicrobial peptides.³⁵ Each of these differences may contribute individually or in combination to the strain-specific vancomycin treatment outcomes in the IE model. Importantly, we performed an in vitro growth assay and found that all the study strains had similar growth rates. In addition, all the study isolates had similar fitness profiles in the IE model: (i) all strains had the same ID₉₅ for the induction of IE (10⁵ cfu/animal); and (ii) the intrinsic virulence of the study strains in the IE model, based on achievable target-tissue counts of S. aureus, was similar among the study isolates (Figure 2). These data indicate that the observed differences in the onset of agr RNAIII expression and treatment outcomes were not due to differences in in vivo growth dynamics. In contrast to our observations, other studies have suggested that agr dysfunction can be associated with decreased vancomycin efficacy and persistent bacteraemia in MRSA infections.^5,27,36–38 For instance, Fowler et al.⁵ reported that δ-haemolysin activity was less common in persistent versus resolving MRSA isolates. However, these differences did not reach statistical significance.⁵ In addition, these conflicting findings may reflect geographical or other epidemiological patterns. Of note, it has been reported that even though the lack of δ-haemolysin production is a phenotypic marker of agr dysfunction,^5,39,40 its absence does not necessarily correlate with the absence of RNAIII transcripts.⁴¹

Paralleling our experimental model outcomes, clinical treatment failures with vancomycin occur fairly commonly even with ‘vancomycin-susceptible’ S. aureus strains.⁴² Some studies have suggested that an in vivo evolution of the hVISA phenotype during therapy can be responsible for such clinical failures.⁴³ However, in the current study, we could not confirm hVISA development in vivo during vancomycin therapy.

Consistent with previous studies from our laboratories⁶ and others,^44–46 we found no common or consensus SNPs within the agr locus that predicted its temporal activation profiles and/or functionality in vitro, or vancomycin responsiveness in experimental IE crossing different CC/agr types. In support of our findings, other investigations have also found no differences in the agr sequence between clinical pairs of parental vancomycin-susceptible S. aureus strains and hVISA/VISA strains evolving during antibiotic-persistent infection.^45,46 Taken together, these findings suggest that strain-specific gain-of-function or loss-of-function readouts of genetic determinants in the agr locus cannot yet be linked to agr functionality and vancomycin treatment outcomes. This observation also implies that genotypic factors beyond or interacting with the agr locus may play a role in adaptive response phenotypes with respect to virulence and antibiotic resistance outcomes in vivo.

Our investigations have several limitations. For example, hVISA strains were not studied in our model; such strains may be very important as they have been associated with high frequencies of treatment failure and persistent bacteraemia.^43,47 In addition, additional S. aureus isolates from patients with known endovascular infections should be evaluated to better understand the precise factors responsible for persistent outcomes in this clinical context. Moreover, based on current findings, we do not know whether the relationship between the RNAIII activation profiles and the outcome of antibiotic treatment is ‘vancomycin specific’ or more broad range against other antimicrobials. Although this is beyond the scope of the present study, future studies should examine this potential correlation.

In summary, we have demonstrated that: (i) early agr RNAIII activation, although not ‘causal’,⁶ is a ‘biomarker’ or ‘signature’ predictive of an antibiotic-persistent outcome; and (ii) factors outside the agr locus are responsible for the genetic linkage between temporal agr RNAIII activation and the outcome of vancomycin treatment in the IE model. The exact mechanism(s) that correlate with the failure of vancomycin treatment remain to be fully elucidated. Studies including whole-genome sequencing, susceptibilities to the peptides involved in host defences against antimicrobial agents (e.g. platelet microbicidal proteins and human neutrophil peptide 1), biofilm formation, endothelial cell damage, vancomycin binding, etc. are in progress in our laboratories to better define these mechanism(s).

Funding

This work was supported in part by grants from the National Institutes of Health (grants R21AI097657 to Y. Q. X., RO1AI-39108 to A. S. B. and AI111 to M. R. Y.); and the U.S. Department of Defense (grant W81XWH-12-2-0101 to M. R. Y.).

Transparency declarations

None to declare.

Supplementary data

Figures S1 and S2 are available as Supplementary data at JAC Online (http://jac.oxfordjournals.org/).