Genomic epidemiology of antimicrobial resistance in Staphylococcus aureus: Analysis of 1,152 complete genomes

Study design and framework

This study employed a retrospective, cross-sectional genomic surveillance design to investigate the prevalence, distribution, and temporal trends of AMR genes in S. aureus genomes. The research framework was designed as a comprehensive genomic epidemiological analysis utilizing publicly available whole genome sequencing data to characterize the contemporary landscape of AMR in S. aureus populations globally.^[20]

The study design incorporated both descriptive and analytical components: Descriptive analysis to characterize the overall prevalence and distribution of AMR determinants, and analytical assessment to examine temporal trends, geographic trends, co-occurrence patterns, and associations between different resistance mechanisms. This approach enables comprehensive surveillance of AMR patterns while identifying emerging resistance trends that may inform public health policy and clinical practice guidelines.

The target population comprised all complete S. aureus genome assemblies deposited in the National Center for Biotechnology Information (NCBI) RefSeq database between January 2020 and April 2025. Only genomes flagged as “complete” were included to ensure the highest data quality and minimize the risk of missing AMR genes located in unsequenced gaps. Genomes with NCBI quality control flags for issues such as “CheckM contamination” ≥5% or “CheckM completeness” ≤90% were excluded. The search was performed using the NCBI taxonomy ID for S. aureus to ensure comprehensive retrieval.

Sample size determination

No formal sample size calculation was performed a priori, as the study aimed to include all qualifying genomes available in the database at the time of data collection. The final sample size was determined by the number of complete S. aureus genome assemblies meeting the inclusion criteria, ultimately comprising 1,152 unique genomes. This sample size provides adequate statistical power for detecting AMR gene prevalence rates, temporal trends, and co-occurrence patterns while ensuring robust estimates for rare resistance determinants.^[21]

AMR gene detection and analysis

AMR gene identification was performed using AMRFinderPlus version 3.12.8 with database version 2024-07-22.1, a validated bioinformatics tool developed by NCBI for comprehensive detection of AMR determinants.^[16,22] The software was run with default parameters and the Staphylococcus species-specific database. The analysis pipeline detected both chromosomally located and plasmid-associated AMR genes. The output was processed to define key study variables, including the presence/absence of AMR genes, drug class resistance profiles, and the mechanism of resistance (acquired gene vs. point mutation). MDR was defined as acquired resistance to at least one agent in three or more different antimicrobial classes.

Statistical analysis and data visualization

Comprehensive statistical analyses were conducted using the Python programming language with specialized libraries for data manipulation and analysis, supplemented by R statistical software for advanced genomic epidemiology analyses.^[23,24] Descriptive statistics were calculated for overall AMR gene prevalence, including the frequency of detection for individual genes, drug classes, geographic variation, and resistance mechanisms across the genome collection.

Temporal trend analysis was performed to examine changes in AMR gene prevalence over the study period (2020–2025) based on genome release dates. Co-occurrence analysis was conducted to identify patterns of multiple AMR genes within individual genomes, including calculation of pairwise gene associations and generation of co-occurrence matrices for the most prevalent resistance determinants within the same genome assembly.

Inferential statistics, including the calculation of 95% confidence intervals (CIs) for prevalence estimates, were applied where applicable. Where available, geographical metadata from the submitter institute information was used to stratify the analysis by country and continent.

Dataset characteristics and overall AMR prevalence

A total of 1,152 complete S. aureus genome assemblies met the inclusion criteria and were successfully analyzed using AMRFinderPlus.

Of the 1,152 genomes analyzed, 1,086 (94.27%, 95% CI [92.78, 95.47]) harbored at least one AMR gene, while only 66 genomes (5.73%) showed no detectable AMR determinants. This exceptionally high prevalence of AMR genes underscores the widespread distribution of resistance mechanisms within contemporary S. aureus populations globally. The analysis identified 135 unique AMR gene symbols across 25 distinct drug classes and 39 drug subclasses, demonstrating the remarkable genetic diversity of resistance mechanisms present in this pathogen. Acquired resistance genes (AMR element subtype) predominated with 8,297 detections compared to 2,534 point mutations (a 3.3:1 ratio), indicating that horizontal gene transfer remains the primary mechanism for AMR dissemination. Based on our definition, 789 isolates (68.5% of AMR-positive genomes) were classified as MDR.

Distribution and frequency of AMR genes

The most frequently detected AMR gene was tet(38), present in 1,085 genomes (99.91% of AMR-positive genomes), indicating near-universal distribution of tetracycline resistance mechanisms. The beta-lactamase regulatory genes blaI and blaZ were detected in and the fosfomycin resistance gene fosB was detected in significant numbers [Table 1].

Notably, the methicillin resistance gene mecA was detected in 543 genomes (50% of AMR-positive genomes, 95% CI [47.03-52.97]), indicating that approximately half of the analyzed isolates represented MRSA strains. The mecR1 gene, encoding the methicillin resistance regulatory protein, co-occurred with mecA in 442 genomes, demonstrating the coordinated presence of the complete methicillin resistance regulatory system.

Aminoglycoside resistance genes showed considerable diversity, with ant(9)-Ia ant(6)-Ia and the bifunctional enzyme aac(6’)-Ie/aph(2’’)-Ia in 15–20% of genomes. Macrolide resistance was primarily mediated by erm(A), detected in 218 genomes (20.07%).

The vanA genes, which confer resistance to glycopeptides, were found in 11 genomes (1.01%). Point mutations that can confer oxazolidinone resistance (23S_C2220T, 23S_G2604T, 23S_G2794T) were found in 32 genomes (2.94%).

Drug class resistance profiles

Analysis of resistance by drug class revealed tetracycline as the most prevalent (99.91%), followed by fosfomycin (84.25%) and beta-lactam resistance (80.57%) [Table 2].

Quinolone resistance, primarily mediated by chromosomal point mutations, was detected in 390 genomes, while lincosamide/macrolide resistance accounted for 218 genomes. The high frequency of multiple resistance mechanisms within individual drug classes highlights the genetic redundancy and complexity of AMR in S. aureus.

At the drug subclass level, beta-lactam resistance remained the most frequent category with 2,422 detections, while methicillin-specific resistance accounted for 1,116 detections, confirming the substantial burden of MRSA within the analyzed population.

Resistance mechanism analysis

The analysis distinguished between acquired resistance genes (AMR element subtype) and chromosomal point mutations (POINT element subtype). Acquired resistance genes predominated with 8,297 detections compared to 2,534 point mutations, indicating that horizontal gene transfer remains the primary mechanism for AMR dissemination in S. aureus. This 3.3:1 ratio strongly suggests the continued primary role of horizontal gene transfer, and direct analysis of mobile genetic elements would be needed to confirm this.

Geographical distribution

Analysis by continent showed high AMR prevalence across all regions with sufficient data, with North America (317/328 AMR-positive, 96.6%), Asia (303/315, 96.2%), and Europe (422/461, 91.5%) showing significant AMR burden [Supplementary Table 1]. At the country level, Japan (96/98, 98.0%) and Switzerland (87/89, 97.8%) had the highest prevalence among countries with over 30 genomes [Table 3]. The top resistance genes varied by region; for example, ant(9)-Ia and erm(A) were highly prevalent in genomes from Japan, whereas fosB was more common in European and North American isolates [Supplementary Tables 2 and 3]. The profile of top AMR genes in countries that uploaded at least 10 genomes in the NCBI database during the study period is shown in Figure 1. A heatmap of AMR gene prevalence (%) by country, grouped by continent, is shown in Supplementary Figure 1.

Close

Figure 1:

Global distribution and profile of the Top antimicrobial resistance genes in S. aureus (with countries that deposited at least 10 genomes in the NCBI database during the study period).

Export to PPT

Temporal trends in AMR prevalence

Temporal analysis revealed significant fluctuations in both genome availability and AMR prevalence over the study period [Table 4]. The proportion of AMR-positive genomes increased from 88.68% in 2020 to a peak of 98.43% in 2024, before declining to 88.64% in the partial 2025 dataset. This temporal variation may reflect changes in sequencing priorities, outbreak investigations, or genuine epidemiological shifts in AMR prevalence.

The number of genomes released annually varied considerably, with peak releases in 2024 (318 genomes) and 2022 (246 genomes), while 2020 and 2021 showed lower numbers (106 and 115 genomes, respectively). The 2025 data represent a partial year (through April), accounting for the reduced genome count (176 genomes).

Gene-specific temporal trends showed relative stability for the most prevalent resistance determinants, with tet(38), blaI, and fosB maintaining consistent high frequencies across all years. Similarly, drug class resistance profiles remained relatively stable, with beta-lactam, fosfomycin, and tetracycline resistance consistently ranking as the most prevalent categories.

Co-occurrence patterns of AMR genes

The distribution of AMR gene numbers per genome revealed substantial variation, ranging from single gene isolates to highly MDR strains carrying up to 26 different resistance genes [Supplementary Tables 4 and Supplementary Figure 2]. The median number of AMR genes per genome was 7, with 118 genomes (10.87%) carrying exactly 5 resistance genes, representing the most common resistance profile.

Pairwise co-occurrence analysis identified strong associations between functionally related genes [Table 5]. The most frequent gene pair was blaI + tet(38), co-occurring in 772 genomes (71.07% of AMR-positive genomes), followed by fosB + tet(38) in 682 genomes (62.80%). Beta-lactamase genes showed particularly strong co-occurrence patterns, with blaI + blaZ detected together in 675 genomes (62.16%).

The methicillin resistance genes mecA and mecR1 co-occurred in 442 genomes, representing 81.10% of mecA-positive isolates, indicating the coordinated presence of the complete methicillin resistance regulatory apparatus in most MRSA strains.

Heatmap analysis of the top 15 AMR genes revealed distinct clustering patterns, with beta-lactamase genes forming a highly correlated cluster, while tetracycline and fosfomycin resistance genes showed broad associations with multiple other resistance determinants [Figure 2]. UpSet plot visualization further illustrated the complexity of multi-gene resistance combinations, with the most common combination including tet(38), blaI, fosB, and blaZ [Figure 3].

Close

Figure 2:

Heatmap showing co-occurrence frequency of top 15 antimicrobial resistance genes. AMR: Antimicrobial resistance.

Export to PPT

Close

Figure 3:

UpSet plot showing co-occurrence of top 10 antimicrobial resistance gene combinations.

Export to PPT

Global burden and clinical significance of AMR in S. aureus

This comprehensive genomic surveillance study reveals an alarmingly high prevalence of AMR genes in contemporary S. aureus populations, with 94.27% of analyzed genomes harboring at least one resistance determinant. This prevalence substantially exceeds previous reports from targeted surveillance studies and emphasizes the global magnitude of AMR in this critical pathogen.^[25,26] The near-universal presence of resistance genes in publicly available S. aureus genomes likely reflects a combination of factors, including preferential sequencing of clinically significant isolates, outbreak investigations, and the inherent bias toward characterizing resistant strains in research contexts.

The dominance of tet(38) in 99.91% of AMR-positive genomes is particularly striking and represents a significant shift from historical tetracycline resistance patterns. Unlike the ribosomal protection proteins tet(M) and tet(O) commonly associated with tetracycline resistance in staphylococci, tet(38) encodes an efflux pump mechanism that provides a distinct resistance phenotype.^[27,28] The near fixation of this gene suggests either strong selective pressure from tetracycline use or genetic hitchhiking with other essential resistance determinants.

Methicillin resistance and MRSA epidemiology

The detection of mecA in 50.18% of analyzed genomes provides contemporary insights into global MRSA prevalence. While this proportion may not directly reflect clinical MRSA rates due to sequencing selection bias, it demonstrates the substantial representation of MRSA in research and surveillance contexts. The high co-occurrence rate of mecA and mecR1 (81.10%) indicates that most MRSA strains maintain complete methicillin resistance regulatory systems, suggesting stable inheritance and expression of methicillin resistance.^[29,30]

In the absence of a functional mecR1, the expression of mecA might be constitutive or dependent on alternative regulatory pathways. While some studies suggest that a non-functional mecI-mecR1 system is required for full b-lactam resistance, others have found no clear correlation, with some highly resistant strains possessing an intact regulatory locus.^[31] The absence of mecR1 in some isolates could imply that these strains have evolved alternative mechanisms to ensure stable, high-level methicillin resistance, potentially through mutations in the mecI repressor or other chromosomal factors.

The coordinate presence of beta-lactamase genes (blaZ, blaI, blaR1) in most isolates reflects the layered beta-lactam resistance mechanisms employed by S. aureus. This dual resistance strategy, combining both beta-lactamase production and altered penicillin-binding proteins, provides robust protection against diverse beta-lactam antibiotics and likely contributes to treatment failures in clinical settings.^[9,32]

The detection of vanA and oxazolidinone resistance mutations, even at low frequencies (1.01% and 2.94%, respectively), is a critical public health concern, as these represent resistance to last-resort antibiotics.

Resistance to glycopeptides and oxazolidinones

The vanA and oxazolidinone resistance mutations are a critical public health concern, even at the low frequencies of 1.01% and 2.94% observed in this study. Vancomycin and linezolid are reserved antibiotics for treating severe MRSA infections. The vancomycin resistance in S. aureus through the acquisition of the vanA operon from Enterococci represents a significant therapeutic challenge. Similarly, resistance to oxazolidinones, a synthetic class of antibiotics, threatens one of the few remaining reliable treatments for MDR Gram-positive infections. Given the clinical implications, surveillance and rapid detection of such strains are important to prevent their spread and to guide appropriate clinical management.^[33]

MDR patterns and clinical implications

The extensive co-occurrence of resistance genes, with some isolates harboring up to 26 different resistance determinants, highlights the emergence of XDR S. aureus strains. The strong associations between functionally related genes, particularly within the beta-lactamase system, suggest coordinated regulation and potential physical linkage on mobile genetic elements.^[34]

The predominance of acquired resistance genes over chromosomal mutations (3.3:1 ratio) emphasizes the continued importance of horizontal gene transfer in AMR evolution. This finding has significant implications for infection control, as mobile resistance elements can rapidly disseminate between bacterial populations and species.^[13,35]

The high frequency of aminoglycoside resistance genes, particularly the diversity of mechanisms represented by ant(9)-Ia, ant(6)-Ia, and aac(6’)-Ie/aph(2’’)-Ia, reflects the complex selective pressures imposed by aminoglycoside use in clinical settings. The bifunctional enzyme aac(6’)-Ie/aph(2’’)-Ia is particularly concerning as it confers resistance to both aminoglycosides and some beta-lactams, potentially limiting therapeutic options.^[10,36]

This study identified macrolide resistance, primarily mediated by the erm(A) gene (38.49% of AMR-positive genomes), as a significant component of the AMR landscape in S. aureus. The erm genes encode for methylases that alter the ribosomal target site, conferring broad resistance to macrolides, lincosamides, and streptogramin B (MLSB phenotype). The high prevalence of erm(A) is consistent with other studies that have shown it to be a dominant macrolide resistance determinant in S. aureus, particularly in MRSA isolates.^[37]

The extensive co-occurrence of AMR genes observed in this study is not a random finding. Specific combinations have strong clinical and policy implications by creating multilayered resistance phenotypes that can severely compromise therapeutic options. The frequent co-occurrence of mecA with beta-lactamase genes (blaZ, blaI, blaR1) is particularly concerning. This combination confers a remarkable, twotiered resistance to beta-lactam antibiotics. From a policy perspective, this finding emphasizes the urgent need for diagnostic strategies that can rapidly identify both resistance mechanisms simultaneously. Infection control policies must be rigorously enforced for strains harboring this combination.

Temporal dynamics and surveillance implications

The temporal analysis reveals concerning trends in AMR prevalence, with the proportion of resistance-positive genomes reaching 98.43% in 2024. While this may partially reflect sampling bias toward clinically relevant isolates, the sustained high levels of resistance across all study years indicate a persistent AMR burden in S. aureus populations. The relative stability of individual gene frequencies suggests established resistance patterns rather than emerging epidemics, which may indicate endemic circulation of MDR lineages.^[38,39]

The fluctuations in genome availability across years likely reflect varying research priorities, funding cycles, and public health responses to specific outbreaks or surveillance initiatives. For example, the lower genome submissions in 2020-2021 could have been due to the COVID-19 outbreak, where sequencing priorities had changed to address the ongoing pandemic. The substantial increase in genome releases during 2024 may indicate enhanced surveillance efforts or responses to emerging resistance concerns.

Genomic surveillance and its public health applications

This study demonstrates the power of leveraging publicly available genomic data for AMR surveillance, providing insights that complement traditional phenotypic surveillance systems. The ability to identify resistance genes, predict phenotypic resistance, and track temporal trends through genomic analysis offers significant advantages for public health monitoring and clinical decision-making.^[40,41]

However, the limitations of sequence-based surveillance must be acknowledged. The preferential inclusion of clinically significant or outbreak-associated isolates in public databases may overestimate AMR prevalence compared to population-based sampling. In addition, the correlation between genotypic resistance predictions and phenotypic resistance testing requires ongoing validation, particularly for novel resistance mechanisms or gene variants.^[42,43]

Study limitations and future directions

Several limitations should be considered when interpreting these findings. Our analysis was constrained by the targets included in the AMRFinderPlus database. While critically important, the complex, polygenic nature of resistance in VISA/hVISA, involving mutations in genes like walK/R, graR/S, is not fully covered in the current database version. The non-random sampling approach, while comprehensive for available genomes, may not represent true population-level AMR prevalence due to sequencing selection bias toward resistant or clinically significant isolates. The lack of comprehensive metadata for many genomes limited the ability to perform detailed epidemiological analyses by infection source, phenotypic, or clinical context. Our analysis is based on in silico genomic prediction, and phenotypic validation was beyond the scope of this study. Stratification by isolation source (e.g., clinical vs. environmental) was not feasible due to inconsistent and incomplete metadata. Furthermore, our analysis of population structure did not extend to MLST or SCCmec typing, which was beyond the defined scope of this investigation but remains an important avenue for future research.

Future surveillance efforts should prioritize systematic, population-based genomic sampling to provide more accurate estimates of AMR burden. Integration of genomic surveillance with clinical outcome data would enhance understanding of the relationship between specific resistance genotypes and treatment efficacy. In addition, expanded international collaboration in genomic surveillance could provide more comprehensive global perspectives on AMR evolution and transmission patterns.