Novel Insights into the Classification of Staphylococcal β-Lactamases in Relation to the Cefazolin Inoculum Effect

Cefazolin has become a prominent therapy for methicillin-susceptible Staphylococcus aureus (MSSA) infections. However, an important concern is the cefazolin inoculum effect (CzIE), a phenomenon mediated by staphylococcal β-lactamases. Four variants of staphylococcal β-lactamases have been described based on serological methodologies and limited sequence information. Here, we sought to reassess the classification of staphylococcal β-lactamases and their correlation with the CzIE.

ABSTRACT

Cefazolin has become a prominent therapy for methicillin-susceptible Staphylococcus aureus (MSSA) infections. However, an important concern is the cefazolin inoculum effect (CzIE), a phenomenon mediated by staphylococcal β-lactamases. Four variants of staphylococcal β-lactamases have been described based on serological methodologies and limited sequence information. Here, we sought to reassess the classification of staphylococcal β-lactamases and their correlation with the CzIE. We included a large collection of 690 contemporary bloodstream MSSA isolates recovered from Latin America, a region with a high prevalence of the CzIE. We determined cefazolin MICs at standard and high inoculums by broth microdilution. Whole-genome sequencing was performed to classify the β-lactamase in each isolate based on the predicted full sequence of BlaZ. We used the classical schemes for β-lactamase classification and compared it to BlaZ allotypes found in unique sequences using the genomic information. Phylogenetic analyses were performed based on the BlaZ and core-genome sequences. The overall prevalence of the CzIE was 40%. Among 641 genomes, type C was the most predominant β-lactamase (37%), followed by type A (33%). We found 29 allotypes and 43 different substitutions in BlaZ. A single allotype, designated BlaZ-2, showed a robust and statistically significant association with the CzIE. Two other allotypes (BlaZ-3 and BlaZ-5) were associated with a lack of the CzIE. Three amino acid substitutions (A9V, E112A, and G145E) showed statistically significant association with the CzIE (P = <0.01). CC30 was the predominant clone among isolates displaying the CzIE. Thus, we provide a novel approach to the classification of the staphylococcal β-lactamases with the potential to more accurately identify MSSA strains exhibiting the CzIE.

INTRODUCTION

Staphylococcus aureus comprises major pathogens responsible for a wide variety of infections worldwide, including skin and soft tissue infections, bacteremia, endocarditis, pneumonia and osteomyelitis, among others (1, 2). Recently, the emerging infections program from the Centers for Disease Control and Prevention reported that S. aureus was the second most frequent pathogen in health care infections in 2015 (3). Severe infections caused by these pathogens have been associated with poor clinical outcomes and high rates of mortality (4). Infections caused by methicillin-susceptible Staphylococcus aureus (MSSA) are now more frequently reported than those caused by methicillin-resistant S. aureus (MRSA) in health care facilities of the United States (5–7). Further, in some Latin American countries, the prevalence of MRSA has decreased and invasive infections (e.g., bloodstream) caused by MSSA seem more common than those with MRSA (8).

The Infectious Diseases Society of America (IDSA) recommends β-lactams as the cornerstone of therapy for infections caused by MSSA (9, 10). Cefazolin, a first-generation cephalosporin, has emerged as a first-line option for these infections due to recent data suggesting similar efficacy compared to antistaphylococcal penicillins but better tolerability, ease of administration, and improved toxicity profile (11–13). Indeed, in a large comparative retrospective observational study of MSSA bacteremia performed in the USA that included 119 hospitals in the Veterans Affairs (VA) system, the cefazolin-treated group showed 37% and 23% reductions in 30 and 90 days mortality, respectively, compared with nafcillin- or oxacillin-treated groups (14). As cefazolin becomes front-line therapy for severe MSSA infections, concerns about the impact of the cefazolin inoculum effect (CzIE) have emerged that could limit the therapeutic efficacy of cefazolin in deep-seated MSSA infections. The CzIE is defined as a marked increase in the MIC to cefazolin when using a high bacterial inoculum (10⁷ CFU/ml, instead of the standard 10⁵ CFU/ml) (15, 16). This phenotype correlates with the presence of particular types of β-lactamases, although the mechanistic basis of this phenomenon is unknown (17). Furthermore, the CzIE has been associated with clinical failures of cefazolin therapy in patients with severe infections (18, 19). In a prospective study of MSSA bloodstream infections conducted in Argentina (a country where antistaphylococcal penicillins are not available), the presence of the CzIE was common (in approximately 50% of isolates) and correlated with increased mortality in patients with MSSA bacteremia (20). Similarly, patients infected with MSSA isolates exhibiting the CzIE in South Korea were more likely to die at 30 days compared to those infected with strains lacking the CzIE (21).

The classification of β-lactamases has traditionally been based on either the functional characteristics of the enzymes (22) or their primary structure (23). An initial classification of staphylococcal β-lactamases, based on serological methods (24), categorized four variants (A, B, C, and D). Subsequently, using the Ambler classification (based on sequence), it was evident that staphylococcal β-lactamases belonged to class A, with variant enzymes differing in the kinetic properties of hydrolysis for particular substrates that included some cephalosporins and nitrocefin (25, 26). Indeed, β-lactamases type A and D appear to be more efficient at hydrolyzing cefazolin and nitrocefin, while types B and C show higher catalytic efficiencies against cephalothin (27). Subsequently, the presence of critical amino acid at positions 128 and 216 of the enzyme (close to the active site) were thought to be sufficient to identify the kinetic differences among the four variants of the staphylococcal β-lactamase (28, 29).

Here, using a genomic approach, we sought to reassess the classification of staphylococcal β-lactamases based on their impact on the CzIE. Using a large collection of isolates of MSSA recovered from bloodstream infections in Latin America (where a high prevalence of the CzIE has been reported) (8, 16), we show that the enzymes exhibit a higher degree of diversity than previously thought and that there are specific allotypes associated with the CzIE. Furthermore, we identify specific residues that highly correlate with the CzIE.

Our findings are likely to provide new insights into the classification of these staphylococcal enzymes, along with an updated framework to evaluate the impact of these determinants in the CzIE.

RESULTS

Bloodstream MSSA isolates from Latin American hospitals exhibit high frequency of the cefazolin inoculum effect.

To investigate the prevalence of the cefazolin inoculum effect (CzIE) in MSSA from Latin America, a total of 690 contemporary MSSA bloodstream isolates were included in the study. The collection encompassed 641 isolates recovered from a prospective multicenter study performed in 9 Latin American hospitals between 2010 and 2014 (8), and 49 more recent isolates from blood cultures in one Colombian hospital (2018 to 2019). Overall, all 690 MSSA isolates were susceptible to cefazolin at standard inoculum (MIC₉₀ 1 μg/ml). However, when cefazolin MICs were determined using a high bacterial inoculum, a high proportion of isolates (40%) exhibited the CzIE with MIC₉₀ up to ≥128 μg/ml (P = <0.0001). The distribution of the frequency of the CzIE in MSSA from the participant hospitals in each country is shown in Table 1. As mentioned above, the overall rate of the CzIE among MSSA bloodstream isolates recovered in Latin American hospitals was 40% (278/690) ranging from 32% to 54%. Brazil and Chile had the lowest frequency, with 32% of MSSA exhibiting the CzIE.

TABLE 1.

Distribution of the frequency of the CzIE in MSSA recovered from bloodstream infections in Latin American Hospitals

Country	No. of MSSA isolates	Frequency of CzIE
		No. of isolates (%) with CzIE	No. of isolates (%) without CzIE
Argentina	89a	48 (54)	41 (46)
Colombia	196	78 (40)	118 (60)
Ecuador	72	30 (42)	42 (58)
Chile	88	28 (32)	60 (68)
Perú	69	26 (38)	43 (62)
Brazil	77	25 (32)	52 (68)
Guatemala	63	28 (44)	35 (56)
Mexico	11	4 (36)	7 (64)
Venezuela	25	11 (44)	14 (56)
Total	690	278 (40)	412 (60)

^aPreviously reported in Miller et al. (20).

BlaZ types A and C are the most predominant β-lactamases associated with the CzIE in MSSA from Latin America.

We performed initial classification of the β-lactamases using the published definition that involves amino acid residues at positions 128 and 216 of BlaZ. β-lactamase was deemed to be nontypeable if one of the residues present in positions 128 and 216 of the four types (A to D) was not identified (27, 28). From the total set of 690 bloodstream MSSA isolates, whole-genome sequencing (WGS) analysis showed that 49 isolates displayed incomplete blaZ sequences, harbored a premature stop codon in blaZ, or exhibited heterogeneous reads suggesting a diverse population. Further, two isolates were identified as the novel S. argenteus species. Thus, those isolates were excluded from further genomic analysis. Among the remaining 641 MSSA isolates, 517 (81%) carried the blaZ gene. Type C BlaZ was the most predominant β-lactamase (37%), followed by types A and B (33% and 29%, respectively). Of note, type D BlaZ was the least frequent enzyme, identified only in 5 MSSA.

Table 2 shows the correlation between the type of β-lactamase and the presence of the CzIE in 517 MSSA. As mentioned above, all isolates were susceptible to cefazolin at the standard inoculum rate (MICs ≤8 μg/ml). The most common types associated with the CzIE were type A (134/172; 78%) and type D (4/5; 80%). Type B had a poor correlation with the CzIE, only present in 10% of isolates (15/149), whereas CzIE was identified in approximately 50% of MSSA harboring BlaZ type C (Table 2). Interestingly, we found a statistically significant higher value in the cefazolin geometric mean (GM) MIC in isolates harboring type A versus type C BlaZ (56.82 μg/ml versus 34.06 μg/ml, respectively; P = 0.0002). Also, the 15 isolates that harbored type B BlaZ and exhibited the CzIE had higher cefazolin MICs (GM 73.52 μg/ml) compared to those of types A and C.

TABLE 2.

Cefazolin MIC (μg/ml) at high inoculum and standard inoculum according to BlaZ type in MSSA from Latin America^a

Type of BlaZ (no. of isolates)	Parameter	CzIE present			CzIE absent
		No. isolates (%)	Standard	High	No. isolates (%)	Standard	High
Type A (n = 172)	GM MIC	134 (78%)	0.85	56.82b	38 (22%)	0.72	4.51
	Range		0.25 to 2	16 to ≥128		0.5 to 2	0.5 to 8
	MIC90		1	≥128		1	8
	MIC50		1	64		0.5	4
Type B (n = 149)	GM MIC	15 (10%)	1.04	73.52b	134 (90%)	0.76	1.36
	Range		0.5 to 8	16 to ≥128		0.25 to 4	0.25 to 8
	MIC90		2	≥128		1	4
	MIC50		1	≥128		1	1
Type C (n = 190)	GM MIC	100 (53%)	0.91	34.06b	90 (47%)	0.73	3.61
	Range		0.5 to 4	16 to ≥128		0.25 to 2	0.5 to 8
	MIC90		2	≥128		1	8
	MIC50		1	32		1	4
Type D (n = 5)	GM MIC	4 (80%)	0.70	32c	1 (20%)	1	4
	Range		0.5 to 1	16 to ≥128		—	—
	MIC90		1	≥128		—	—
	MIC50		0.5	16		—	—
Nontypeable (n = 1)	MIC	—	—	—	1	1	2

^aGM MIC, geometric mean of cefazolin MIC; standard inoculum, 10⁵ CFU/ml; high inoculum, 10⁷ CFU/ml. —, not applicable.

^bSignificance P = <0.0001 by Mann-Whitney U test.

^cSignificance P = 0.0286 by Mann-Whitney U test.

Specific allotypes of BlaZ are associated with the CzIE.

The phylogenetic tree shown in Fig. 1 is based on the amino acid sequence of BlaZ. We identified two main clusters. One cluster harbored 19 allotypes, which mostly belonged to type A (11/19) and 6, 1, and 1 allotypes were correlated with types C, B, and D, respectively, in this cluster. The second cluster harbored 10 allotypes and most (8/10) of them belonged to type B. We found 43 different substitutions in the predicted amino acid sequence of BlaZ among isolates with the CzIE versus those lacking the CzIE (Fig. 1). The most frequent allotypes were BlaZ-1 (n = 163) and BlaZ-2 (n = 100). BlaZ-1 (from type C BlaZ) and BlaZ-2 (from type A BlaZ) showed approximately 50% and 96% of correlation with the CzIE, respectively. Indeed, BlaZ-2 showed a high and clear association with the CzIE. Further, these allotypes had cefazolin MICs at the high inoculum value of ≥64 μg/ml in 31% and 75% of BlaZ-1 and BlaZ-2 isolates, respectively (Fig. 2). Of note, BlaZ-3 and BlaZ-5 allotypes (both Type B but unrelated phylogenetically) were highly associated with the lack of the CzIE.

FIG 1.

BlaZ allotypes in methicillin-susceptible S. aureus from Latin America. Maximum likelihood phylogenetic tree of the BlaZ allotypes is shown on the left. Multiple sequence alignment showing the deduced amino acid sequences of BlaZ in 517 genomes of MSSA from Latin America. Forty-three amino acid positions within BlaZ β-lactamase harboring substitutions enabled the differentiation of 29 allotypes (first column). The predicted signal peptide sequence is highlighted in gray, and the β-lactamase domain of the enzyme is highlighted in yellow. The sequence from the reference isolate ATCC 29213 is highlighted in blue; amino acids highlighted in red are the substitutions present in each allotype. The second column shows the classical BlaZ type (NT, nontypeable, due to Lys at residue 119 and Ser at residue 207). The third and fourth columns show the number of isolates showing the cefazolin high-inoculum effect (CzIE) or lacking it (No CzIE). The last three rows show the association of each amino acid substitution with the CzIE; green boxes highlight positions that show statistically significant association with the CzIE. Positions in blue were not associated with the CzIE (P = ≤0.01, Z-test).

FIG 2.

Cefazolin MIC values at high inoculum in relevant allotypes of BlaZ. Distribution of cefazolin MIC values at high inoculum in relevant allotypes of BlaZ Type A, Type B, and Type C. Blue colors represent MIC values ≥16 μg/ml using high inoculum (darker shade of blue color shows a higher MIC value). Shades of green represent susceptible MICs (≤8 μg/ml) (darker shades represent lower MICs).

Amino acids in three positions of BlaZ were significantly associated with the CzIE.

We generated an alignment of 517 diverse sequences of BlaZ, including the sequences of two well-characterized strains of S. aureus harboring Type A β-lactamase (ATCC 29213 and TX0117 that are negative and positive controls for the CzIE, respectively) (19). Results of the alignment identified three amino acid substitutions (A9V, E112A, and G145E) that each had a statistically significant association with the CzIE (P = <0.01) (Fig. 1). A9V is located in the signal peptide and was present in 216 genomes that exhibited the CzIE, compared to 131 genomes lacking the phenotype. E112A was detected among 115 MSSA genomes exhibiting the CzIE, compared to only 5 genomes without the CzIE. Finally, the G145E substitution was identified in 237 MSSA isolates with the CzIE versus 125 lacking this phenotype. Of note, the presence of a lysine residue in position 119 of BlaZ was negatively correlated with the CzIE, with 135 isolates lacking the CzIE compared to 15 isolates exhibiting the CzIE (Fig. 1). In order to identify SNPs potentially associated with the CzIE, we performed the same analysis using the blaZ gene sequences. A total of 45 different alleles were identified and two SNPs showed potential association to the CzIE (Table S1 in the supplemental material). In concordance with the BlaZ analysis, these SNPs (C26T and A335C) corresponded to the A9V and E112A substitutions, respectively.

MSSA CC30 was the most predominant genetic lineage exhibiting the CzIE.

We found highly diverse genetic lineages among the 641 MSSA isolates, with eight clonal complexes (CC1, CC5, CC8, CC15, CC22, CC30, CC45 and CC97) identified and accounting for 79% of the genomes (Fig. 3). CC8 (n = 110, 17%), CC5 (n = 103, 16%), CC30 (n = 97, 15%), and CC1 (n = 92, 14%) were the most common genetic lineages without specific geographical clustering (Fig. S1). The MSSA isolates that exhibited the CzIE (n = 253) clustered into seven clonal groups, CC1 (n = 34), CC5 (n = 28), CC8 (n = 44), CC15 (n = 8), CC22 (n = 1), CC30 (n = 89), and CC45 (n = 9). CC30 was the predominant clone among isolates displaying CzIE, with 35% out of 253 of the isolates belonging to sequence type 30 (ST30). In addition, BlaZ-2 (the allotype highly associated with the CzIE, see above) was the most frequent allotype among the CC30 isolates displaying the CzIE (94% of the isolates; 84 out of 89) (Fig. 2 and Fig. 3). The MSSA lacking the CzIE clustered into six CCs, CC1 (n = 58), CC5 (n = 75), CC8 (n = 66), CC15 (n = 29), CC45 (n = 28), and CC97 (n = 23). Of note, CC97 was exclusively found in isolates lacking the CzIE (Fig. 2 and Fig. 3) with 52% (12 out of 23) isolates lacking the blaZ gene (83% of the CC97 isolates clustered into ST97).

FIG 3.

Phylogenetic tree of methicillin-susceptible S. aureus with blaZ allotypes from Latin America. Mid rooted maximum likelihood phylogenetic tree from the core genome of 641 MSSA isolates. Yellow, red, and blue shadows over the branches of the tree show the most frequent clonal complexes (CC) and the gray shadow shows CC97 within the sample. The inner ring shows the isolates with and without the CzIE. The middle ring shows the typical BlaZ characterization. The outer ring shows the BlaZ allotypes for each isolate.

DISCUSSION

Our comprehensive study with isolates collected in various countries of Latin America suggest a high overall prevalence (40%) of the CzIE. Of note, as reported before (20), the highest frequency was in Argentina, a country where cephalosporins are used as first-line therapy for MSSA infections and isoxazolyl penicillins are not available. Since the sample only represents a few hospitals in each country, we cannot generalize our findings to the entire region. However, previous studies (16), including a previous larger sample of hospitals in four countries of Latin America and infections other than bacteremia, are consistent with our results. The frequency of the CzIE seems to be lower in other countries such as the United States, South Korea, and Japan (30–32), although a limited number of isolates have been reported. Moreover, we have demonstrated that WGS-based approaches prove to be valuable tools for the study of the genetic basis of the CzIE.

Using a genomic approach, we offer new insights into the classification of staphylococcal β-lactamases in relation to the CzIE. The original classification of such enzymes was based on serological methodologies, which grouped the enzymes in four distinct categories (24). Biochemical and limited DNA sequencing studies (25, 28, 29, 33–35) further supported this classification, which continues to be widely used. The molecular approach to classify S. aureus BlaZ considered only two amino acid residues (128 and 216, corresponding to residues 217 and 220 in Fig. 1) using only a partial sequence of BlaZ (instead of the full 281 amino acid residues). These two residues are located close to the active site of the enzyme and substitutions (using site-directed mutagenesis) in these two positions in type A, C, and D β-lactamases resulted in kinetic differences (27). However, this categorization was performed using a limited number of staphylococcal β-lactamases belonging to only three strains, considered “prototypical” isolates (27). Further, this classification implies that only the enzymatic kinetic characteristics are critical to explain the phenotype, ignoring regulatory and processing regions within blaZ (i.e., the signal peptide sequence, which has been shown to be important for the release of BlaZ into the extracellular milieu) (35). Our results, using a large data set of isolates and considering the entire BlaZ sequence (including the signal peptide) did not identify the same residues as statistically associated with the CzIE, suggesting that factors other than kinetic properties may have a critical role in this phenotype.

Our phylogenomic analyses yielded three important findings. First, we found a marked degree of diversity in BlaZ from clinical MSSA isolates that was not previously captured using the classification based on serotypes. Using this approach, we were able to identify specific BlaZ allotypes associated with the presence or absence of the CzIE. Indeed, BlaZ-2 was the dominant allotype showing a clear association with the CzIE. In contrast, BlaZ-3 was the most abundant allotype associated with the absence of the CzIE. Although the associations were statistically significant, the presence of these allotypes in strains with and without the CzIE confirms that many other factors, apart from the pure enzymatic activity, play an important role in the CzIE. Nonetheless, our results provide proof of principle that the allotype rather than the type of β-lactamase could be a more accurate tool for identifying strains with a likelihood of exhibiting the CzIE.

Second, we were able to identify particular amino acid residues that were significantly associated with the CzIE. The E112A and G145E substitutions were highly associated with allotypes that exhibited the CzIE. Interestingly, none of these amino acid substitutions are located in the active site of the enzyme, so the functional effect of these changes is unknown. The A9V substitution was located in the signal peptide, suggesting that the processing of the enzyme could also play an important role in the development of the CzIE. As mentioned above, a previous study showed that a single amino acid substitution (serine to proline) at position 22 of BlaZ ( in the signal peptide) resulted in changes in the cleavage of the membrane-bound enzyme, thereby increasing the release of BlaZ into the extracellular medium (36), which could affect the phenotype.

Third, our study found that CC30 was the predominant lineage exhibiting the CzIE, grouping the majority of isolates harboring the BlaZ-2 allotype. Two different studies from South Korean hospitals reported a similar association between ST30 and type A isolates exhibiting the CzIE (33, 37). Of note, S. aureus strain TX0117, a well-characterized clinical isolate of type A BlaZ from a patient with therapeutic failure to cefazolin and exhibiting the CzIE, also belongs to ST30 (19). The CC30 lineage has been associated with both severe and persistent infections and also reduced pathogenesis in animal models, which is related to acquired mutations in bacterial regulators and virulence genes (38). Moreover, genomic analyses of representatives of S. aureus clinical isolates from the predominant CC30 clones identified a polymorphism (C55R) in the histidine kinase of the accessory gene regulator (Agr) quorum-sensing system, which was acquired by contemporary isolates of this clone during evolution (38). Interestingly, an association between dysfunction of the quorum-sensing Agr system and the CzIE was previously described (37). Therefore, it is tempting to speculate that this lineage harbors unique genomic attributes that may contribute to the development of the CzIE.

There were several limitations of our study. First, we did not investigate the kinetic profiles of the different allotypes of β-lactamases. Further, although the MIC values are important from a clinical point of view, in this work we used cefazolin MICs as a surrogate of the kinetic characteristics of β-lactamases. Thus, in order to establish if allotype classification of BlaZ could correlate with the CzIE and predict the phenotype in clinical isolates of S. aureus, enzymatic analysis would need to be performed. Second, we evaluated only isolates collected from bloodstream infections. Thus, our findings may not be applicable to MSSA recovered from other sources (i.e., deep-seated abscesses, osteomyelitis, pneumonia, endocarditis, and other infections with high bacterial loads). Finally, clinical outcomes were not available for analysis, and thus correlation with clinically relevant phenotypes would be of paramount importance in future research.

In summary, we provide a novel approach for the classification of the staphylococcal β-lactamases, based on a comprehensive collection of isolates in a region where the CzIE is highly prevalent. Furthermore, our results support using a classification based on BlaZ allotypes that could more accurately predict the CzIE among MSSA clinical isolates, although the CzIE seem to be a multifactorial phenomenon. Further analyses focused on the mechanistic basis of the CzIE are warranted.

MATERIALS AND METHODS

Bacterial isolates.

We included 641 isolates of MSSA from a prospective multicenter study of S. aureus bacteremia performed in nine Latin American countries between 2010 and 2014 (8). Additionally, 49 bloodstream isolates collected in one hospital in the city of Pereira (central Colombia) between 2018 and 2019 were also included. The identification of all the isolates was confirmed using multiplex PCR assays, as previously described (8).

MIC and evaluation of the cefazolin inoculum effect.

MICs of cefazolin were determined with standard (5 × 10⁵ CFU/ml) and high (5 × 10⁷ CFU/ml) bacterial inocula by the broth microdilution method, as previously described (16, 39). The maximum antibiotic concentration tested for cefazolin was 128 μg/ml. The CzIE was defined as an increase in the MIC to ≥16 μg/ml when using the high inoculum. All MIC determinations were performed in triplicates, and three independent researchers performed the observation of the results at 24 h. Controls included S. aureus TX0117, a high-level producer of BlaZ type A β-lactamase (exhibiting the CzIE), S. aureus ATCC 29213, a producer of small amounts of BlaZ type A β-lactamase (lacking the CzIE), and S. aureus ATCC 25923, a BlaZ-negative strain without the CzIE (16, 39). Geometric mean (GM) MICs were calculated and compared by nonparametric statistics by the Mann-Whitney U test using GraphPad Prism version 8.2.0. P values of ≤0.05 were regarded as statistically significant.

Genome sequencing, assembly, and phylogenetic analysis.

Genomic DNA was obtained from overnight cultures treated with lysostaphin at 37°C for 30 min, using the DNAeasy blood and tissue kit (Qiagen). DNA libraries were prepared using the NexteraXT DNA sample preparation kit and sequenced on the Illumina platform HiSeq or MiSeq at 150-bp or 300-bp paired-end reads, respectively. Raw reads were trimmed using Trimmomatic (40) specifying NexteraXT adapters and default parameters for the Illumina platform. Genome assemblies were performed with SPAdes v3.13 (41) and annotated using RAST (42). Low-quality genome assemblies were discarded (n = 13) due to contamination or low coverage. Phylogenetic analysis of the genome assemblies was carried out using the core genome (ortholog genes present in at least 95% of the genomes), obtained with Roary (43). Orthogroups were individually aligned with MUSCLE (44) and then concatenated to obtain a matrix. The tree was built using RAxML (45) with a GTR+Γ model of heterogeneity. The phylogenetic trees were plotted using iTOL (46). Sequence type (ST) and clonal complex (CC) were assigned based on the S. aureus MLST database (https://pubmlst.org/).

BlaZ allotype definition.

From the assembled genomes, the sequence of the blaZ operon was extracted based on BLASTN (47) and searches against the reference sequence of ATCC 29213 blaZ operon (NCBI accession no. LHUS02000010 complement [4218 to 7297]) were performed. Isolates with incomplete sequences or missing sequence of the blaZ operon were further evaluated by mapping the sequencing reads to the reference sequence and obtaining a consensus using CLC Genomics Workbench v8.5. As mentioned above, we excluded 49 genomes and the final sample used in the analysis included only isolates with a complete blaZ sequence (n = 517). Initial BlaZ typing was performed based on the amino acids at positions 128 and 216 (119 and 207 based on our numbering that includes the signal peptide, Fig. 1) into types A, B, C, and D (28). If one of these residues was not present it was defined as nontypeable. Then, a BlaZ allotype was defined as each unique amino acid sequence found among the evaluated genomes. An allele number (designated BlaZ-0) was assigned to the first allotype using the amino acid sequence of the reference strain S. aureus ATCC 29213 (accession no. ODV53133.1). Subsequently, changes in sequence were assigned new numbers starting with the most frequent allotypes among the sampled genomes. A BlaZ phylogenetic tree was built to identify clusters among the allotypes, and it was performed over a multiple sequence alignment of the allotypes with MUSCLE (44). The tree was generated with RAxML (45) using the protein sequence of BlaZ and automatic substitution matrix determination with a Γ model of heterogeneity. Allele determination of the blaZ gene was also performed over the 517 genomes. Gene sequences were aligned with MUSCLE and unique alleles were determined, an allele number was assigned (starting from blaZ-0 for the sequence of ATCC 29213), and subsequent alleles were numbered accordingly to the frequency.

Association of BlaZ allotypes and blaZ alleles with the cefazolin high-inoculum effect.

The frequency of MSSA isolates exhibiting and lacking the CzIE was obtained for each BlaZ allotype and each blaZ allele. A multiple sequence alignment of the sequences was performed with MUSCLE (43). For each amino acid substitution among allotypes and for each nucleotide change among the alleles, we obtained the frequency of isolates harboring the substitution and exhibiting or lacking the phenotype. We performed a Z-test to evaluate differences in the proportion for each amino acid substitution or nucleotide change, with a statistical significance of P = ≤0.01. The test was performed using python module Stats models version 0.8.0 (48).

Data availability.

Sequencing information for this study was submitted to the NCBI GenBank database under the BioProject numbers PRJNA580194 and PRJNA595347.