Measuring beta-lactam minimum inhibitory concentrations in Staphylococcus aureus in the clinical microbiology laboratory: pinning the tail on the donkey

ABSTRACT

Significant shortcomings have been identified in standard methods of susceptibility testing in bacteriological media, not only because the media fails to recapitulate the in vivo environment, but susceptibility testing itself fails to capture sub-MIC effects that significantly attenuate bacterial virulence properties. Until susceptibility testing conditions better recapitulate the in vivo environment, attempts to establish the quantitative relevance of beta-lactam MIC using current clinical microbiology standards in Staphylococcus aureus infections will likely prove unsuccessful.

INTRODUCTION

Staphylococcus aureus is one of the most common pathogens causing bacteremia and is a leading cause of bacterial endocarditis, particularly in developed countries (1, 2). While the superiority of beta-lactam antibiotics such as anti-staphylococcal penicillins (ASPs, e.g., nafcillin, oxacillin, and flucloxacillin) and cefazolin over vancomycin is undeniable, mortality in methicillin-susceptible S. aureus (MSSA) bacteremia is still approximately 10%, despite the prompt administration of beta-lactams to which the organism is “susceptible” to according to accepted laboratory standards (3–5).

The long-standing debate in selecting ASPs vs cefazolin as the optimal therapy for MSSA bacteremia has heated up again. In past years, documented case reports of cefazolin failure driven by an inoculum effect raised concerns about utilizing cefazolin in high-inoculum infections such as documented endocarditis (6–8). More recently, retrospective studies appear to show that the more reliable activity of ASPs captured at the individual case level appears to be offset by their adverse drug reactions when administered over longer treatment courses at the population level, perhaps rendering cefazolin a better choice overall (9–11).

Collateral concerns raised alongside the cefazolin inoculum effect are MSSA with borderline oxacillin resistance (BORSA) driven by hyper-beta-lactamase production. Such strains are characterized by the absence of mecA, the gene encoding penicillin-binding protein 2a (PBP2a), with oxacillin MICs 1–8 mg/L, a range that straddles the oxacillin <2-mg/L susceptibility breakpoint (12).

In order to assess the clinical relevance of BORSA among MSSA, Hess et al. performed a recent analysis of clinical outcomes in patients with MSSA bacteremia, with outcomes stratified by receipt of ASP vs cefazolin in relation to oxacillin MIC <1 mg/L vs 1–2 mg/L (13). The authors noted no significant differences in their composite outcome of 30-day mortality, bacteremia clearance within 72 h, and clinical failure requiring salvage therapy (13). In this paper, we will address the significant limitations of using beta-lactam MIC to quantitatively predict outcomes.

Bacteriologic versus physiologic media

Methicillin resistance in S. aureus requires the presence of the mecA gene, which encodes penicillin-binding protein 2a (PBP2a). The gold standard detection methods for methicillin resistance in S. aureus in PCR for mecA or latex agglutination detection of PBP2a. Methicillin resistance is defined by the Clinical Laboratory Standards Institute (CLSI) as an oxacillin MIC ≥4 mcg/mL, and MICs of ≤2 mcg/mL are considered susceptible (7, 8). Oxacillin is a semisynthetic penicillin that has replaced methicillin in susceptibility testing due to its greater stability and the fact that methicillin is no longer commercially available. NaCl enhances the expression of methicillin resistance phenotype in the presence of mecA. In order to increase sensitivity in detecting the presence of mecA phenotypically, susceptibility testing to oxacillin is done in the presence of NaCl, either by the addition of 2% NaCl to cation-adjusted Mueller Hinton broth or by using an agar screening for methicillin-resistant S. aureus (MRSA) using oxacillin 6 mg/L and 4% NaCl. Cefoxitin is a greater inducer of PBP2a expression and therefore is frequently included in susceptibility testing assays and disk diffusion assays as another sensitive assay for detecting the expression of the mecA phenotype among S. aureus.

The expression of the methicillin-resistant phenotype in S. aureus with mecA can be homogeneous, heterogeneous, and borderline. Most clinical isolates are heterogeneous in their expression of methicillin resistance, which means the population has subpopulations with differential expressions of resistance. Under routine growth conditions, ≥99.9% of MRSA appear to be susceptible to beta-lactams. However, if the cells are grown at 30°C to 35°C or in the presence of 6.5% sodium chloride, they become more homogeneously resistant and express beta-lactam resistance at a much higher frequency (14). In addition, the growth of heterogeneous strains in the presence of beta-lactam results in the selection of a homogeneous phenotype. Serial passage of these cells in the absence of antibiotics leads to slow reversion back to the heterogeneous state. Borderline resistance refers to isolates that are at the margin of resistance (termed BORSA: borderline oxacillin-resistant S. aureus). Some strains with borderline resistance possess the mecA gene, but an extremely small subpopulation expresses resistance. Other strains lack the mecA gene where alterations in or overexpression of the other PBPs cause reduced affinity for beta-lactams. Yet, other strains overproduce beta-lactamase, which can by sheer high local amounts result in significant hydrolysis of beta-lactams (e.g., cefazolin) that are generally fairly resistant to hydrolysis (14–16).

As can be garnered from the complexities above in trying to “bring out” the expression of mecA-mediated methicillin resistance using artificial conditions in vitro, it should come as no surprise that any quantification of such results will have low clinical predictive value as none of the above conditions specifically address differences in salt concentrations, temperature, cell density, etc. that the bacterial cells are experiencing under therapy in vivo.

Guidance of antimicrobial therapy using in vitro susceptibility testing took into account bacterial growth that could be inhibited in the presence of antibiotics. While focusing on the bug-drug interaction, these methods did not consider the host environment that the antibiotics would be working in to treat an infection in vivo (17). Cation-adjusted Mueller Hinton broth, a typical media used to grow bacteria for susceptibility testing, lacks the bicarbonate buffer that is present in the body. The presence or absence of bicarbonate can have a profound impact on antibiotic activity not captured in standard assays. Recent work suggests that overlooking the bicarbonate component may leave critical gaps in our appreciation of antibiotic activity in vivo when using bacteriological media that can be rectified by using physiologic media (18–21). We encourage readers to examine the study by Ersoy et al. that demonstrated that antimicrobial susceptibility testing done in physiologic media had a stronger correlation with in vivo activity than testing done in standard media (20). A specific example where this testing can have clinically useful translation is with Pseudomonas aeruginosa because of the fact that fluoroquinolones off the only oral antibiotic option. Recognizing that azithromycin has notable activity against multidrug-resistant P. aeruginosa and Acinetobacter baumannii in vitro using physiologic media that translates into successful therapy in vivo opens much easier avenues of therapy of sinus and other respiratory infections (18–21). When utilizing bacteriological media, this activity is completely missed. Antibiotic susceptibility performed in physiologic media better predicts activity in vivo as assessed in animal models of infection compared to susceptibility testing performed in bacteriological media (20).

Despite the shortcomings of current susceptibility methods, they appear to be accurate the majority of the time (20). Therefore, completely replacing current methods with physiologic susceptibility testing is not only impractical with the complexities of automated systems but likely not justified. Instead, we advocate the development of supplemental pathways of physiologic testing that clinicians and pharmacists can request for in cases refractory to standard treatments or in cases caused by pan-resistant organisms. Such supplemental physiologic susceptibility testing may identify better salvage regimens that are currently out of reach. Options would be to either buffer standard media with bicarbonate or to use pre-made bicarbonate-buffered physiologic media used in tissue culture, such as Roswell Park Memorial Institute 1640 (RPMI-1640). The use of physiologic media should also be deployed in novel antibiotic discovery programs where molecules are screened for activity.

Are MRSA really MRSA?

A key component in physiologic media that accounts for the different antibiotic susceptibility testing results that better align with in vivo efficacy is bicarbonate buffer, the main physiological buffer in vivo (22). Furthermore, bicarbonate allows the assessment of how antibiotics interact with host antimicrobial peptides because it is required in many in vitro peptide-killing assays (22). In the presence of bicarbonate, beta-lactam susceptibility testing of S. aureus reveals that a significant subset of MRSA shows significant increases in beta-lactam susceptibility (23). When evaluating 58 MRSA bloodstream isolates, 44-mM NaHCO₃ resulted in >4-fold reduction in oxacillin MIC in 33% of isolates and cefazolin MIC in 75% of isolates measured by broth microdilution (23). These isolates have been referred to as being “bicarbonate-responsive.” Making these results even more interesting is the fact that when placed in a simulated vegetation model or a rabbit model of endocarditis, the bicarbonate-responsive MRSA can be effectively treated as MSSA with beta-lactams (24, 25). The mechanism for this bicarbonate effect is not completely understood, but it has been shown that bicarbonate causes repression of PrsA, a cell membrane protein that is required for PBP2a full maturation and expression and allowing cell wall synthesis in the presence of conventional beta-lactams (26, 27). This emerging line of evidence undermines the clinical translatability of established paradigms of beta-lactam susceptibility testing and calls for more reliable methods that clinicians can base their decision-making on. We have previously shown that bacterial cytological profiling identified a subset of MRSA that exhibited lysis with beta-lactam exposure, suggesting a possible rapid diagnostic method to identify these bicarbonate response MRSA (28).

The rise of penicillin-susceptible S. aureus (PSSA)

The introduction of penicillin into clinical practice in the late 1940s was followed by the emergence of penicillin-resistant Staphylococcus aureus strains harboring penicillinase encoded by plasmid-borne blaZ (29). Paralleling the arms race of the cold war of the second half of the 20th century, an antibiotic “arms race” between humans and S. aureus began, first with the development of semi-synthetic anti-staphylococcal penicillins, and, with the emergence of MRSA, the increased use of vancomycin (30). MIC-defined resistance to vancomycin was slow to emerge. Vancomycin is unique among anti-staphylococcal antibiotics in that it took almost 40 years for a microbiologically defined vancomycin-resistant clinical isolate to emerge, as compared to other anti-staphylococcal antibiotics that required a very short time period. This unique feature suggests a considerable discordance between clinical resistance and microbiologic resistance defined by MIC. Indeed, an increased understanding of the host-pathogen relationship in S. aureus bloodstream infections can explain vancomycin treatment failures being commonplace for serious MRSA bloodstream infections, despite clinical isolates being fully “susceptible” to vancomycin (31).

Despite the emergence of PSSA, penicillin susceptibility testing is inconsistent and archaic depending on the methods utilized. In a recent study assessing disk testing for penicillin resistance, EUCAST methodology utilizing the P1 penicillin disc showed that among the 342 tests on blaZ+ strains, 37 (11%) were classified as susceptible. For those utilizing the P10 penicillin disc CLSI method, 395 blaZ+ tests showed 83 (21%) to be susceptible (32). Penicillin resistance was defined based on zone size and clearance zone edge characteristics around the penicillin disk (32). Many microbiology laboratories do not routinely assess penicillin susceptibility among clinical S. aureus isolates.

Over the second half of the 20th century, penicillin-susceptible S. aureus became an obscurity, representing <5% of clinical strains among large strain collections but, in reality, rarely if ever encountered clinically. However, recent studies show a rise in PSSA, both globally and at the institutional level (33, 34). McNeil et al. recently described the emergence of penicillin-susceptible S. aureus (penicillin MIC <0.125 mg/L) in pediatric osteoarticular infections between January 2011 to December 2019 at two large tertiary care pediatric hospitals in Saint Louis, MO, and Houston, TX (35). No PSSA were identified from 2011 to 2014, but increased yearly thereafter, peaking at 20.4% (by 2019) (35). Children were treated most frequently with a first-generation cephalosporin (76%) or nafcillin (14%).

A review of data at our institution reveals a near doubling in the rates of penicillin susceptibility among S. aureus from the 1990s until the present (Fig. 1). PSSA rates ranged from 8% to 10% from 1994 to 2003 and then slowly incrementally started climbing after 2004, now at 16%–20% range and comparable to the McNeil study (Fig. 1).

Fig 1.

Rates of penicillin-susceptible S. aureus (PSSA) at our institution from 1995 to 2022. The arrow designates the time daptomycin became clinically available in 2004.

Hess et al. did not evaluate the susceptibility of their strains to penicillin to determine whether the presence or absence of beta-lactamase had an impact on oxacillin MIC and outcome (13). This might have enriched their study analysis given that BORSA are derived from the hyper-production of beta-lactamase. Clinical outcomes stratified by the presence or absence of beta-lactamase (outcomes of PSSA compared to MSSA that are penicillin-resistant) would have been of interest, especially given the emergence of PSSA.

It may be worth commenting here on why PSSA may have emerged in the last 20 years. We believe that the clinical introduction of daptomycin and its increased use during that time may have played a role in this emergence. When daptomycin is complexed with calcium, it inserts into the bacterial lipid membrane in an ionic detergent-like manner, causing rapid cellular depolarization and cell death (36). The negative impact of ionic detergent molecules on penicillin resistance in S. aureus was documented by Sonsten and Baldwin in 1972 when they showed that the growth of S. aureus in ionic detergents (alkyl sulfates of >8 carbon atoms) resulted in the loss of penicillinase (14, 37). As a result of these findings, detergent-based methods have been used in the curing of S. aureus of their penicillinase-harboring plasmids.

We had the opportunity to study an MSSA isolate TX0117 and its penicillin-cured derivative TX0117c. TX0117c demonstrated a consistent fourfold increase in daptomycin MIC compared to its parent strain from 0.125 to 0.5 mg/L (38). Resistance to killing by cationic antimicrobial peptides produced by the innate immune system as well as reduced susceptibility to daptomycin (a de novo antimicrobial peptide when complexed with calcium) may be a collateral phenotype PSSA and another manifestation of the daptomycin/peptide vs beta-lactam seesaw effect. Supporting this hypothesis is that McNeil et al. noted almost a twofold higher rate of complications (consisting of chronic osteomyelitis, pathologic fracture, growth arrest, avascular necrosis, or chronic arthritis) among children with PSSA as compared to penicillin-resistant S. aureus (PRSA) (18.1% vs 9.8%), a numerical difference but without sufficient numbers to achieve statistical significance in their small study (35). It has been well-established that relative resistance to cationic antimicrobial peptides, such as platelet microbicidal proteins (PMP), is associated with metastatic infections and complicated S. aureus bacteremia (39). The loss of daptomycin susceptibility in S. aureus is accompanied by higher resistance to killing by host cationic antimicrobial peptides, consistent with cross-resistance between these exogenous and endogenous antibiotics (40).

Given that penicillin-susceptibility may be a marker of virulence in endovascular S. aureus infections, we encourage labs to take the extra steps to quantify penicillin MIC in documented endovascular infections. Further investigation of these isolates is warranted. Clinician awareness of penicillin-susceptible S. aureus endocarditis would potentially offer potential therapeutic cues, including the avoidance of daptomycin monotherapy given that the pathway to daptomycin resistance may have already begun before daptomycin is even used. Such cases would potentially be treated with combination therapy. Given the shortcomings of disk testing, we recommend that labs serious about adopting penicillin therapy for PSSA use PCR for blaZ testing, although they are not commercially available.

Beta-lactam activity goes far beyond the MIC

Favorable outcomes associated with beta-lactams as compared to non-beta-lactam drugs, and adverse outcomes associated with being deprived of beta-lactam therapy through a penicillin allergy, are being increasingly recognized (15, 41). As we have reviewed previously, these established clinical observations, particularly when it comes to treating serious S. aureus infections, speak to the adjunctive indirect antibacterial properties that beta-lactams exert in vivo (41). These adjunctive properties do not involve direct killing or growth inhibition of S. aureus by beta-lactams. However, beta-lactams render S. aureus more vulnerable to host defense killing through cell wall modifications and reduce expression of surface-associated virulence factors collectively referred to as microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) (41). One important cell surface S. aureus virulence factor, protein A, has been shown to depend on lipoteichoic acid (LTA)-rich regions of the septal membrane for expression (16). Beta-lactams, recognized to release LTA from S. aureus, may undermine protein A and potentially other MSCRAMMs expression on the bacterial surface, attenuating bacterial virulence (42, 43).

Beta-lactam induction of inflammation may enhance clearance of endovascular-based infections or prevent the establishment of infection, in part through LTA and cell wall components (42–45). Cytokines are produced by leukocytes upon S. aureus stimulation, and a lack of pro-inflammatory cytokine production promotes worse outcomes in mice with S. aureus infection (46). A lack of IL-1 beta response is a marker of persistent S. aureus bacteremia and elevated concentrations of IL-10 markers of mortality (47). IL-1 beta is a potent inducer of T-lymphocytes and drives the production of several pro-inflammatory cytokines in response to pathogens such as S. aureus (48). Recent work suggests that beta-lactam effects on alpha-toxin- (49, 50) and O-acetyl transferase-mediated lysozyme resistance (51, 52), as well as other exotoxin expression mediated through PBP1 binding (53), may enhance host IL-1 beta production, thereby promoting more rapid bacterial clearance. In addition, bacterial cell wall that is synthesized in the presence of beta-lactams by MRSA and demonstrated reduction in cross-linking generates a greater IL-1 beta response by macrophages (54).

An example of this was highlighted in a case whereby the combination of daptomycin plus nafcillin resulted in prompt clearance of MRSA bacteremia within 24 h after persisting for 21 days on first, second, and third-line therapies and no identified source control solutions (55). Notably, the MRSA strain was “resistant” to both daptomycin (MIC 2 mg/L) and nafcillin (128 mg/L). Testing of this isolate in physiologic media Roswell Park Memorial Institute (RPMI) medium still resulted in oxacillin MIC 32 mg/L, suggesting that the prompt bacteremia clearance was mediated by more than just direct antimicrobial activity. Indeed, a detailed study of this MRSA in the laboratory demonstrated enhanced vulnerability to killing by neutrophils and cathelicidin, as well as daptomycin, when it was grown in the presence of nafcillin at concentrations less than 1/50th MIC. Such low concentrations had no effects on growth rate, and these profound effects would have been entirely unnoticed in standard media (55).

These effects have recently been used in combination therapy strategies in treating MRSA bacteremia with various beta-lactam drugs (56–59). Similar effects have been shown with ampicillin-resistant vancomycin-resistant Enterococcus faecium (VRE), whereby the addition of ampicillin, which alone demonstrated no activity, markedly potentiated the activity of daptomycin and host innate peptides in the killing of VRE (60). It appears that beta-lactams, by enhancing the killing of S. aureus, Enterococcus species, and potentially other organisms, provide a means by which bacteremia may be cleared more efficiently by boosting the activity of either the other antibiotics in a combination regimen (i.e., daptomycin or vancomycin) and/or the innate immune system. Therefore, profound clinically relevant effects of beta-lactams on S. aureus at concentrations well below the MIC compromise the ability to predict drug efficacy and therapeutic outcome by beta-lactam MIC alone.

We refer readers to our recent review on beta-lactam adjunctive properties in the context of patients labeled penicillin-allergic who unfortunately are unnecessarily deprived of these benefits by being given non-beta-lactam antibiotics (41).

Conclusions

This review summarizes the limitations of current in vitro antimicrobial susceptibility testing in the clinical laboratory that compromise the quantitative value of oxacillin MIC in predicting therapeutic outcomes with beta-lactam therapy. A graphic summary is presented in Fig. 2. With what has emerged recently revealing shortcomings of MIC testing in bacteriologic media, we can confidently say that these results were expected. Indeed, if S. aureus MIC to beta-lactams (and even vancomycin) were the sole or even the most significant determinant of treatment outcomes, clinicians would have had all the antibiotics they needed to treat serious MSSA and MRSA infections by the early 1970s along with good source control. Vancomycin covers >99% of MRSA and ASPs, and cefazolin covers 100% of MSSA, so according to MIC criteria, no new drugs would be needed. We know this cannot be further from the truth. Yet, the antibiotic regulatory approval process hinges almost exclusively on MICs performed in bacteriologic media superimposed on the easiest-to-perform clinical studies like soft tissue and urinary tract infections to evaluate and introduce new antibiotics. Unfortunately, just like bringing more food items to the supermarket does not directly translate into cooking better food, the current one-dimensional antibiotic approval process has not necessarily advanced the treatment of serious MRSA and MSSA infections. In order to improve treatment efficacy and improve clinical outcomes, the triad of microbiologists, scientists, and clinicians must work toward building understanding and clinical translation of the host-pathogen-drug triad. This will require significant adjustments in how antimicrobial susceptibility testing is done, how infections are risk-stratified, and how antibiotics are utilized.

Fig 2.

Graphic summary of how conventional beta-lactam MIC testing in standard microbiological media misses critical beta-lactam attributes in vivo, significantly compromising its quantitative clinical value. Key beta-lactam properties not appreciated in bacteriologic media susceptibility testing are sublethal cell envelope changes by rendering bacteria more vulnerable to host innate immunity (left), enhanced pro-inflammatory cell signaling fostering bacterial clearance (second from left), diminished bacterial-host cell interaction (second from right), and effects of bicarbonate buffer (right).

We believe that current clinical trial paradigms of assessing antibiotics need more granularity to define optimal therapy. This will require (i) assessment of antibiotics in more physiologic media; (ii) incorporating the host immune response in assessing antibiotic function; (iii) understanding the effect of antibiotics on bacterial virulence; and (iv) dynamic clinical trial designs that incorporate risk stratification and temporal aspects of antimicrobial therapy initiation along the continuum of infection. While these factors may appear out of reach, gradual integration alongside current practices seems the most practical. However, unless progress is made in these areas, we will continue to bring new antibiotics to market but make little progress in improving outcomes. For example, when one realizes that mortality due to MRSA bacteremia is generally the same today as 40 years ago, there is a lot of work to be done.