Going Back in Time: Increasing Penicillin Susceptibility among Methicillin-Susceptible Staphylococcus aureus Osteoarticular Infections in Children

ABSTRACT

In the late 1940s to 1950s, Staphylococcus aureus isolates first-gained resistance to penicillin. Recently, some centers have described an increase in the proportion of methicillin susceptible S. aureus (MSSA) which are also susceptible to penicillin (PSSA). There are little data on the frequency of PSSA infections in children. We investigated the prevalence of penicillin susceptibility among pediatric MSSA acute hematogenous osteoarticular infection (OAI) isolates. MSSA OAI isolates were obtained through surveillance studies at Texas Children’s and St. Louis Children’s Hospitals from January 2011 to December 2019. All isolates underwent PCR for blaZ β-lactamase, PVL genes and agr group. All blaZ negative isolates then underwent penicillin MIC determination. blaZ negative isolates with penicillin MIC ≤ 0.125 μg/mL were considered PSSA. Multilocus sequence typing (MLST) was conducted on a subset of isolates. A total of 329 unique isolates were included in the study. The median patient age was 9.2 years (IQR:5.1 to 12.2). Overall, 6.7% of isolates were penicillin susceptible. No PSSA were detected prior to 2015 but increased yearly thereafter. By the final study year, 20.4% of isolates were PSSA (P = 0.001). PSSA were similar to penicillin-resistant MSSA (PR-MSSA) isolates in terms agr group and PVL carriage as well as clinical presentation and outcomes. PSSA were of distinct sequence types compared to PR-MSSA. PSSA appears to be increasing among OAI in U.S. children. Overall, PSSA isolates are associated with a similar clinical presentation as penicillin-resistant isolates. The potential for use of penicillin treatment in PSSA OAI warrants further study.

INTRODUCTION

Staphylococcus aureus is responsible for a range of clinical manifestations, including asymptomatic colonization, mild soft tissue infections, and severe invasive disease. By the late 1940s, shortly after the introduction of penicillin into clinical practice, S. aureus strains emerged which were resistant to penicillin (1, 2). This resistance was mediated through the acquisition of a β-lactamase (penicillinase) encoded by blaZ. Penicillinase producing strains rapidly spread globally both in the hospital and community settings and became the dominant S. aureus infections encountered (3–5). The need for agents active against penicillin-resistant S. aureus prompted the development in the late 1950s of semisynthetic penicillins (or antistaphylococcal penicillins, ASPs, e.g., methicillin, nafcillin) which resist degradation by the staphylococcal penicillinase. This was then later supplemented by the development of first-generation cephalosporins (1GCs, e.g., cefazolin, cephalexin).

Interestingly, a number of centers across the world have recently reported increasing penicillin susceptibility among invasive MSSA infections in adults. The global Sentry Antimicrobial Surveillance Program has noted an increase in reported penicillin-susceptibility among MSSA (PSSA) from 1997 to 2016, peaking at 26% worldwide, although geographic variation existed in prevalence as well as methods for detection (6). A similar, albeit less dramatic, increase in PSSA has been noted in the United States with PSSA accounting for 10.4% of all S. aureus isolates in a Boston study (7).

The incidence and impact of penicillin susceptibility in S. aureus isolates causing infections in children is unknown. Osteoarticular infections (OAIs) including osteomyelitis and septic arthritis are the most common invasive S. aureus infections in children. We investigated the frequency and clinical significance of penicillin-susceptibility among community-associated MSSA OAIs identified through active surveillance studies at two tertiary care children’s hospitals.

RESULTS

Patients.

From January 1, 2011 to December 31, 2019, 329 cases with corresponding viable MSSA isolates met inclusion criteria (267 from Texas Children’s Hospital [TCH] and 62 from St. Louis Children’s Hospital [SLCH]). The median subject age was 9.2 years (IQR;5.1 to 12.2 years) and 123 (37.4%) were of female gender (Table S1 provides clinical and demographic information as well as select microbiologic data). Overall, 10.3% of patients developed orthopedic complications. Complications included chronic osteomyelitis (n = 21), pathological fracture (n = 4), growth arrest (n = 6), avascular necrosis (n = 2) and chronic arthritis (n = 1). Two-hundred and 97 subjects (90.3%) were treated with either nafcillin (n = 46, 13.9%) or a first-generation cephalosporin (n = 251, 76.3%) for definitive therapy. No subject was treated with penicillin, although one was prescribed amoxicillin-clavulanate.

Penicillin susceptibility.

Overall, 300 isolates had blaZ detected by PCR (91.2%); of blaZ-negative isolates, 22/29 had penicillin MICs ≤ 0.125 μg/mL resulting in 22 being considered PSSA (6.7%). Among penicillin-resistant isolates with negative blaZ PCR, penicillin MICs ranged from 4 to 8 μg/mL. In our study group, PSSA were first observed in 2015, followed by an annual increase in the proportion of isolates which were PSSA such that by 2019, 20.4% of isolates were penicillin-susceptible (P = 0.002, R²=0.778) (Fig. 1). There was no significant change in subject age or demographics across the study period. Overall, a higher proportion of isolates at SLCH were PSSA compared to TCH (12.9% versus 5.2%, P = 0.04).

FIG 1.

Trends in PSSA during the study period. The frequency of PSSA across all sites increased during the study period, linear regression P = 0.002, R² = 0.778.

Subjects with PSSA were slightly older than children with penicillin-resistant MSSA (PR-MSSA) infections, although this did not achieve statistical significance (11.8 years [IQR: 2.9 to 13.7] versus 9.1 years [IQR:5.1 to 12.1], P = 0.08) (Table S1). PSSA and PR-MSSA infections were comparable in terms of patient demographics and clinical manifestations. There was a numerically higher rate of complications among PSSA (18.1% versus 9.8%, P = 0.26).

For PSSA, MICs to β-lactam agents are displayed in Fig. 2. Notably the penicillin MIC₉₀ (0.125 μg/mL) was lower than that for oxacillin (0.25 μg/mL) or cefazolin (0.5 μg/mL). There was no statistically significant correlation between penicillin and ampicillin MICs (Spearman r = 0.36, P = 0.09) (Table S2). None of the PSSA isolates were resistant to clindamycin compared to 36/307 (11.7%) PR-MSSA (P = 0.15). PSSA had slightly lower MICs to cefazolin and oxacillin compared to PR-MSSA (P = 0.001 each) (Fig. 3A and B).

FIG 2.

β-lactam MICs in PSSA. MICs to penicillin, ampicillin, oxacillin, cefazolin, and cephalexin were determined by broth dilution for all PSSA.

FIG 3.

Cefazolin and oxacillin MICs in PSSA versus PR-MSSA. Cefazolin and oxacillin broth dilution MICs were conducted for PSSA and PR-MSSA. (A) Cefazolin MICs in PSSA were lower than PR-MSSA, P = 0.001. (B) Oxacillin MICs were lower in PSSA than PR-MSSA, P = 0.001.

Fifty random blaZ-positive isolates underwent penicillin MIC determination. The median penicillin MIC among blaZ-positive isolates was 8 μg/mL; a single blaZ-positive isolate had a penicillin MIC = 0.125 μg/mL. Among the subset of isolates with penicillin MICs determined (n = 79), the blaZ PCR was concordant with the MIC in 89.9%.

β-Lactam time-kill studies.

PSSA strains were subjected to time-kill studies using previously described methods (8) with oxacillin (2.5 μg/mL), penicillin (1.25 μg/mL), and ampicillin (1.25 μg/mL), respectively, as well as an antibiotic-free control. Comparable log₁₀ reductions in viable bacteria at 24 h of incubation were noted between oxacillin (median log₁₀ reduction 2.35, IQR:1.8 to 3), penicillin (2.5, IQR:1.9 to 3.2), and ampicillin (2.3, IQR:1.8 to 3.1).

Molecular characteristics.

PSSA and PR-MSSA were similar with respect to Panton Valentine leucocidin (PVL) carriage and accessory gene regulator (agr) group (Table S1). Twenty-two PSSA and 44 PR-MSSA were characterized by multilocus sequence typing (MLST). The 66 studied isolates were diverse comprising 22 different sequence types (ST) including one novel ST (ST7664). A high degree of similarity among PSSA strains was noted (Fig. 4). ST1 along with single and double locus variants (SLV, DLV) disproportionately contributed to PSSA (7/22, 31.8%, versus 4/44 PR-MSSA, 9.1%, P = 0.03); four of these PSSA strains were isolated at TCH and three from SLCH. The next most common ST among PSSA, ST8, occurred with similar frequency in PR-MSSA (6/22, 27.3% versus 8/44, 18.2%, P = 0.5). In contrast, the most common ST among PR-MSSA was ST30 (14/44, 31.8%) which was not present among PSSA (P = 0.003).

FIG 4.

eBurst of PSSA and select matched PR-MSSA. PSSA are depicted in blue and PR-MSSA in gray. Numbered nodes correspond to individual sequence types (ST); the size of individual nodes corresponds to the relative abundance of isolates of a given ST. Nodes connected by a single line represent single locus variants (SLV).

DISCUSSION

Since the 1950s to the 1960s, the overwhelming majority of S. aureus strains have been resistant to penicillin, such that many clinical laboratories in the United States do not report or even perform penicillin-susceptibility testing. This stems from the large preponderance of penicillin resistance in most contemporary U.S. studies, as well as challenges related to testing. blaZ expression may be constitutive or inducible such that simple MIC determination/phenotypic testing may under detect potentially penicillin resistant strains (9), which could ultimately result in penicillin treatment failure. Older literature suggests that 16% of S. aureus isolates with penicillin MICs below the susceptibility breakpoint are capable of producing β-lactamase (9). Phenotypic tests for β-lactamase production in S. aureus, such as nitrocefin disc testing, have suboptimal sensitivity; false-negative rates with the nitrocefin discs range from 17% to 100% (10–12). Other assays, such as penicillin zone edge observation, have poor interobserver reliability with only 22% concordance in one study (11). In a survey of certified clinical labs in New York, only 15.4% performed β-lactamase testing in S. aureus (13). In fact, the American Heart Association guidelines for the management of infective endocarditis (IE) in adults recommend against penicillin G treatment of S. aureus IE that is reported as penicillin-susceptible, given the inability of many clinical laboratories to accurately characterize this (14).

In a recent U.S. study utilizing a combination of automated broth dilution and phenotypic β-lactamase detection methods (nitrocefin and zone edge), the investigators described increasing penicillin-susceptibility among S. aureus (7). The use of penicillin for the treatment of MSSA could represent a step forward for therapy, allowing for the use of a low-cost narrow spectrum agent. We present, to the best of our knowledge, the first data describing the frequency and clinical characteristics of PSSA OAI in children.

While PSSA only comprised 6.7% of all isolates, we observed a significant temporal trend during the study period. By the final year, >20% of isolates were PSSA at each center. However, there were differences in the overall frequency of PSSA at both sites, suggesting a component of geographic variation. In a study of MSSA bloodstream isolates in Quebec from 2010 to 2015, 28% were penicillin susceptible (15). Investigators in Scandinavia reported the frequency of penicillin susceptibility among MSSA bloodstream isolates ranged from 29% to 57% by year (16). In contrast, in a study conducted in Boston during an overlapping time period, only 10.4% of isolates were PSSA (7). Taken together, these findings support a degree of regional variation in PSSA prevalence.

The reasons for the re-emergence of PSSA in our study population are unclear. One could hypothesize that increasing incidence of PSSA may be related to select advantageous clones. Studies of bloodstream isolates from a center in Sweden suggest that PSSA may be related to clonal expansion of certain strain types, particularly clonal complex (CC) 5 and CC45 (10). We found that PSSA disproportionately belong to ST1 and associated variants, accounting for 31.8% of penicillin-susceptible isolates, supporting clonal expansion as a partial explanation for rising PSSA in our population. Of note, a much higher rate of penicillin-susceptibility was noted in the Swedish study (31%), and these investigators included both health care- and community-associated isolates, potentially contributing to the differences in STs from our cohort. Both of these studies should be contrasted with work among PSSA bloodstream isolates in China which revealed a very high degree of genetic diversity with no clonal complex constituting more than 17% of isolates (17). Of note, blaZ-negative MRSA strains are described in the literature (18), raising questions about the potential fitness cost associated with blaZ.

Cases of PSSA and PR-MSSA OAI were similar with respect to clinical presentation and epidemiology, making it impossible to distinguish between these based on clinical risk factors alone. It is worth acknowledging that our study did not capture data on preceding antimicrobial use, and thus, it is unclear how antibiotic pressure may have impacted penicillin susceptibility. While it is conceivable that the use of penicillin antibiotics for other infections (such as otitis media [19]) could have selected for PR-MSSA, given the overall preponderance of penicillin-resistance in S. aureus in our population as well as globally, the effect size of such a phenomenon would likely be small.

We observed very low penicillin and ampicillin MICs among PSSA, much lower than for oxacillin or cefazolin. According to current Clinical and Laboratory Standards Institute (CLSI) guidance (20), susceptibility to penicillinase-labile penicillins as a group (e.g., ampicillin) in S. aureus can be inferred from susceptibility to penicillin assuming the penicillin MIC ≤ 0.125 μg/mL and tests for β-lactamase are negative. Strictly speaking, however, no ampicillin breakpoints exist for S. aureus. Using in vitro time-kill experiments, we noted a similar bactericidal effect among penicillin, ampicillin and oxacillin in contemporary PSSA. Such findings argue in favor of routinely testing for and subsequent reporting of penicillin susceptibility in MSSA and utilizing penicillin when possible. Furthermore, while intravenous penicillin requires every 4 to 6 h dosing in children and can be costly, it is possible that ampicillin (or even amoxicillin) could serve as a more economical treatment for PSSA infections. It is even conceivable that testing for blaZ could be incorporated into multiplex rapid diagnostic platforms allowing for the early identification of presumptive PSSA. Of note, in our study group, nine isolates did not have blaZ detected by PCR but had elevated penicillin MICs. Such a result may be explainable by polymorphisms at the blaZ locus (21), which could have led to false-negative PCR results and emphasizes the need for both optimized PCR and MIC testing.

The precise impact that penicillin-susceptibility in-and-of-itself has on clinical outcomes in S. aureus is unclear. In previous retrospective studies in both Canada and Sweden, there were no differences in mortality between adult patients with bacteremia caused by PSSA versus PR-MSSA after adjusting for confounders (15, 16). We observed a numerically higher rate of orthopedic complications associated with PSSA isolates, although this did not achieve statistical significance. As penicillin susceptibility was not reported by the clinical laboratories of the participating centers in our study, this information was unavailable to treating clinicians and consequently no subjects were treated with penicillin. Thus, while outcomes were comparable between PSSA and PR-MSSA, it unclear how penicillin treatment of PSSA may have modified outcomes. In a retrospective observational study in Australia, the treatment of adults with PSSA bacteremia with benzylpenicillin was associated with reduced mortality relative to flucloxacillin treatment (22). In studies from the 1970s, peak concentrations in synovial fluid of children with septic arthritis treated with oral cephalexin (25 mg/kg), ampicillin (50 mg/kg), or penicillin (50 mg/kg) were 11.3, 11.9, and 1.9 μg/mL, respectively (23). Based on these historic drug levels combined with our MIC data, there may be at least a hypothetical PK advantage to using penicillins (with peak concentrations >15-fold the MIC) over first-generation cephalosporins for PSSA OAI. It should be mentioned, however, that in other observational studies in Denmark and Australia, treatment of adults with PSSA bacteremia with penicillin had similar outcomes as semisynthetic penicillins (dicloxacillin, flucloxacillin) or first-generation cephalosporins (24, 25); thus, the value of penicillin versus other agents in this clinical scenario remains controversial. Clinical trials are needed to reevaluate the role of penicillin/ampicillin in the management of PSSA infections, albeit these trials may be difficult to conduct.

There are some additional limitations to this study which should be acknowledged. While there was a high proportion of PSSA in the final study years, given the overall relatively small number of PSSA isolates, our study was underpowered to detect subtle differences in clinical presentation between PSSA and PR-MSSA. The retrospective, observational nature also limits the degree to which outcomes can be associated with penicillin-susceptibility. Furthermore, the unequal contribution of cases by the participating institutions may have potentially biased findings. As all PR-MSSA were not subjected to MLST, the ability to evaluate associations between ST and penicillin-susceptibility is limited. This is partially ameliorated, however, by the matching of PSSA and PR-MSSA by year and site of isolation. Finally, as this study focused on osteoarticular infections, the findings may not be generalizable to other S. aureus infections in children. However, given that isolates were recruited from two large tertiary pediatric centers and that OAIs represent the most common invasive S. aureus infections in children, penicillin-susceptibility is likely prevalent in other staphylococcal infections and in other regions of the country.

In conclusion, penicillin-susceptibility is increasing among S. aureus OAIs in children, although geographic variation exists. PSSA and PR-MSSA OAIs have a similar clinical presentation. Further studies are needed to better understand the optimal management of invasive PSSA infections in children.

MATERIALS AND METHODS

Patients and isolates.

Cases were identified through two prospective surveillance studies ongoing at Texas Children’s Hospital (TCH, affiliated with Baylor College of Medicine, Houston, TX) and St. Louis Children’s Hospital (SLCH, affiliated with Washington University School of Medicine, St. Louis, MO) the details of which have been described previously (26, 27). Briefly, S. aureus isolates identified by the clinical microbiology laboratory at each institution are subcultured and then transferred to the respective research laboratories; corresponding clinical and demographic data are abstracted from the medical record. Cases of community-acquired MSSA acute hematogenous osteomyelitis and acute hematogenous septic arthritis (collectively referred to as osteoarticular infections, OAI) with viable isolates from January 1, 2011 to December 31, 2019 were eligible for inclusion. The diagnosis of OAIs was based on the constellation of physical examination findings, radiologic reports, and microbiology studies (28). Cases associated with a duration of symptoms on presentation >28 days, open/penetrating trauma, orthopedic hardware, chronic wounds or decubiti, or OAI developing after surgery and/or cases in children with medical comorbidities were excluded. The characteristics of a subset of these subjects has been presented in other publications (29, 30). Medical records for all subjects were reviewed. The institutional review boards of Baylor College of Medicine and Washington University School of Medicine approved this study.

Definitions.

Duration of bacteremia refers to the number of calendar days with a positive blood culture (29). For patients who underwent surgical management of OAI, delayed source control was regarded as >3 calendar days from the time of admission until first surgical source control (29). Definitive therapy was considered the antibiotic which patients were prescribed at time of hospital discharge or the final in-hospital antibiotic in patients who received all care as inpatients. Orthopedic complications included progression to chronic osteomyelitis (29), pathological fracture, limb length discrepancy, avascular necrosis, chronic dislocation, and/or chronic arthritis (29, 30). Chronic osteomyelitis was considered the presence of sequestrum or permeative lucencies on radiographs, persistent signs/symptoms or the readmission for management of osteomyelitis after at least 4 weeks of antimicrobial therapy (29).

Microbiology studies.

All isolates were transferred to the Edward O. Mason Infectious Diseases Research Laboratory (IDRL) at TCH for further study. For purposes of this study, only one isolate per patient was included. Isolates did not undergo serial passage. The clinical microbiology laboratories of TCH and SLCH do not report penicillin-susceptibility among MSSA, and thus, penicillin-susceptibility status was unknown to treating clinicians. All isolates underwent PCR for blaZ in the IDRL using previously published primers (11). All isolates negative for blaZ were further subjected to penicillin MIC determination using a broth dilution assay (20). Isolates which were blaZ negative and had a penicillin MIC ≤ 0.125 μg/mL were regarded as PSSA (20); all others were regarded as penicillin-resistant MSSA (PR-MSSA). A random subset of PR-MSSA had also penicillin MICs determined. All isolates underwent PCR for the presence of genes associated with Panton Valentine leucocidin (PVL) as well as accessory gene regulator (agr) group classification using a multiplex assay (31–33). Susceptibility testing to clindamycin, trimethoprim-sulfamethoxazole, and cefoxitin were performed by the clinical laboratories in the routine course of care. All isolates underwent cefazolin and cephalexin MIC determination using broth dilution in the IDRL (30); vancomycin MICs were determined with Etest. Identified PSSA and a subset of PR-MSSA (matched based on center and year of isolation 2:1 with PSSA) underwent MIC determination for oxacillin in the IDRL using broth dilution; additionally, all PSSA had ampicillin MICs determined using this methodology. ATCC29213 was used as the control in susceptibility testing.

β-Lactam time-kill studies.

PSSA strains were subjected to antibiotic time-kill studies using similar methods to those described previously (8). We hypothesized that penicillin would have similar in vitro activity against PSSA as oxacillin. Briefly, 10⁸ CFU of the organisms of interest were inoculated into tryptic soy broth containing either oxacillin (2.5 μg/mL), penicillin (1.25 μg/mL), or ampicillin (1.25 μg/mL) as well as a control containing no antibiotics. The broth culture was allowed to incubate at 37°C with gentle agitation. A 100-μL aliquot was obtained at 0 h, 4 h, and 24 h of incubation from each culture, serially diluted, and streaked on blood agar plates, which were in turn incubated overnight at 37°C after which colony counts were performed. The recovery of organisms was compared between cultures incubated with oxacillin versus penicillin versus ampicillin. Data were presented as the log₁₀ reduction in CFU/mL relative to control.

Multilocus sequence typing.

Multilocus sequence typing (MLST) was conducted on all PSSA and a subset of PR-MSSA (matched 2:1) using previously described methods (34). STs were assigned based on a publicly available database (https://pubmlst.org/organisms/staphylococcus-aureus). Minimum spanning trees were constructed using Phyloviz (www.phyloviz.net).

Statistical analyses.

Categorical variables were compared with Fisher exact test; continuous variables were analyzed with Wilcoxon rank sum. Temporal trends in rates of PSSA were examined with linear regression. Statistical analyses were conducted using STATA ver. 16 (STATA Corp., College Station, TX).