Prosthetic joint infection (PJI) complicating hip and knee arthroplasty results in medical morbidity, added cost, reoperation, and higher mortality. Reference Boddapati, Fu, Mayman, Su, Sculco and McLawhorn1–Reference Fischbacher and Borens3 The prevalence of hip and knee arthroplasty is increasing due to a rising number of procedures and an aging population. In the United States, ∼10% of women and 8% of men aged in their 70s and 80s have a total knee replacement. Reference Maradit Kremers, Larson and Crowson4 These factors contribute to an overall rising incidence of PJI. Reference Premkumar, Kolin and Farley5
Staphylococcus epidermidis is among the most common causes of PJI in total joint arthroplasty. Reference Mussa, Manciulli and Corbella6–Reference Tsai, Sheng, Lo, Jiang and Chang8 Monomicrobial infections caused by S. epidermidis tend to manifest later than others and to present with more subtle clinical signs. Reference Lowy9,Reference Post, Harris and Morgenstern10 Thus, S. epidermidis infections often present after the 90-day surveillance recommended by the Nationwide Healthcare Safety Network/Centers for Diseases and Control and Prevention (NHSN/CDC) for surgical site infection (SSI), leading to underreporting.
S. epidermidis is a commensal skin organism. PJI is thought to be seeded at the time of surgery or later through injection or hematogenous spread. Reference Sánchez, Benito and Rivera11 A growing number of molecular studies have identified genetic traits that may differentiate pathogenic S. epidermidis from commensal strains. Reference Conlan and Mijares12–Reference Zhang, Ren and Li14
If there are other mechanisms of transmission of S. epidermidis in PJI, such as dominant community strains or nosocomial transmission, they have not been addressed by the existing literature and may be missed by routine surveillance given the indolent clinical presentation. Whole-genome sequencing (WGS) may provide insight into these common but understudied infections. Reference Sundermann, Chen and Miller15–Reference Sundermann, Chen and Miller18 In this study, we performed molecular and clinical assessments of 4 years of consecutive S. epidermidis isolates obtained from prosthetic joints in an orthopedic hospital to better understand the origin of and risk for these infections.
Methods
Whole-genome sequencing was used to analyze a cohort of S. epidermidis isolates obtained from arthrocentesis and surgical joint cultures collected between January 1, 2017, and December 31, 2020. S. epidermidis isolates were collected through routine clinical practice. After receiving approval from the institutional review board, S. epidermidis isolates from patients with 1 or more prior corresponding intra-articular procedures at our hospital were identified using microbiology laboratory records and hospital medical records.
Collection of samples
Clinical specimens were processed in a Clinical Laboratory Improvement Amendments–certified microbiology laboratory. Samples were routinely incubated for 5 days, or 14 days upon clinician request. Microscan (Omron, Renton, WA) was used for strain identification of the clinical isolates and antibiotic susceptibility testing. Isolates were banked in a −80°C freezer as part of routine practices. The identified S. epidermidis samples were thawed and regrown under a laboratory fume hood and for 24 hours at 37°C. Individual colonies were added to thioglycolate broth or chocolate slants and were incubated again for 24 hours at 37°C prior to shipping to Day Zero Diagnostics (Boston, MA) for genomic analysis.
Preparation of bacterial genomic DNA and WGS library
Whole-genome sequencing was performed at Day Zero Diagnostics using the epiXact PRO service. Bacterial DNA from 60 isolates was purified using the DNAeasy Ultra Clean Microbial kit (Qiagen, Hilden, Germany) protocol according to the manufacturer’s guidelines. The quality of extracted DNA of each sample was assessed using Nanodrop (Thermo Fisher Scientific, Waltham, MA) and was quantified using the Qubit dsDNA HS assay kit (Invitrogen, Carlsbad, CA). Genomic DNA libraries were prepared using Nextera tagmentation (Illumina, San Diego, CA) and were ligated with unique, dual-indexed (i5 and i7) primers during amplification. Tapestation 4150 (Agilent) dsDNA assay was used to estimate fragment size. Sample libraries were pooled based on fragment size and concentration for multiplexed sequencing on the Illumina MiniSeq and HiSeqX sequencers. In addition, 2×150-bp paired-end reads were generated and demultiplexed into FastQ-formatted sequencing data.
Genomic analysis
The single nucleotide polymorphism (SNP)–based clonality and WGS analysis included species identification, in silico multilocus sequence typing (MLST) typing, and phylogenomic analysis, along with genotypic assessment of the prevalence of specific antibiotic resistance genes, virulence genes, and other relevant genes. Kraken software was used for species identification. Samples with high level contamination (>10%) with other species were discarded from the clonality analysis. MLST was inferred in silico using the PubMLST database. Reference Wood and Salzberg19 Samples were grouped by MLST and, for each group, a reference genome of matching MLST was identified from RefSeq and was used for read mapping and SNP calling. The de novo assemblies from each group were also used as reference to verify SNP calling or when no closely related reference was identified. Mapping metrics included percent reference genome covered (>95%) and percent mapped reads (>80%). SNP calling metrics included Phred score (>20) minimum depth (≥10 reads) and major allele frequency (>90%). A pairwise SNP analysis was performed within each MLST group. Insertions and deletions were not included in the SNP analysis. Recombination events were identified using Genealogies Unbiased By recomBinations In Nucleotide Sequences (GUBBINS) software. Reference Croucher, Page and Connor20 Cluster isolates were defined as 2 or more isolates with a pairwise distance of <25 SNPs. Reference Coll, Raven and Knight21
The phylogenomic analysis was performed using an external reference genome (Genbank: GCA-002850315v1). The presence of genomic features of interest, both genes and markers, was assessed in each sample using the BLAST version 2.11 algorithm and sample assemblies. We used iTOL V6 for phylogenomic tree visualization. Reference Letunic and Bork22
Epidemiologic data
Hospital electronic medical records were manually reviewed for patient demographics and environmental data including age, sex, and patient’s town. The following surgical and procedure data were considered: operating room (1–16), personnel in procedure, inpatient unit, date of procedure, surgical instruments and implants, prosthesis versus native joint, revision versus primary surgery, and days between prior intervetion and culture positivity. PJI was determined using the Musculoskeletal Infection society (MSIS) criteria. Reference Parvizi, Zmistowski and Berbari23,Reference Parvizi, Gehrke and Chen24 Additional reviews of surgical and clinical data including personnel, equipment and supplies, and hospital space were performed to compare cluster cases and noncluster cases. Univariate analyses were performed using SAS version 9.4 software (SAS Institute, Cary, NC). P < .05 was considered statistically significant.
Results
Between 2017 and 2020, 11,358 hip arthroplasties and 13,406 knee arthroplasties were performed. During this time, the deep and organ-space SSI incidence rate and SIRs remained at historical baselines for the hospital (Supplementary Fig. 1 online). Also, 60 S. epidermidis joint isolates were identified from 44 patients, including 35 patients with prosthetic joints. Of these 35 patients, 19 (54.3%) met MSIS criteria for PJI, and 9 (22.0%) met NHSN/CDC criteria for hip arthroplasty or knee arthroplasty. Of those who met the MSIS PJI criteria but not the NHSN/CDC organ-space SSI criteria, 6 patients presented outside the 90-day surveillance window, 2 patients did not meet the clinical criteria for infection, and 3 patients had no prior surgery at our institution.
Figure 1. Flowchart of isolate selection. Hexagons represent isolates that were excluded. In total, 48 isolates were included in the final analysis (48 for phylogenomic analysis, 45 for clonality analysis). Of these 48 isolates, 3 were not included in the clonality analysis because of multistrain contamination. *From patients with at least 1 prior procedure in the same joint at our hospital. **Based on a SNP difference of 0; all patients from duplicate isolates met MSIS criteria for PJI.
The average age of the study participants was 61 years, and 22.2% were female. Moreover, 37.8% of the isolates were obtained from hip joints, and the rest were obtained from knee joints. Of 60 samples, 25 (41.7%) were polymicrobial, resulting in 7 multispecies or multistrain contaminated samples for which data cleanup was not possible. After removal of these 7 samples as well as 5 duplicate isolates (same infection, multiple specimen at different time points), 48 samples were ultimately included for phylogenomic analysis, and 45 (93.7%) of these samples were included in the clonality analyses (Fig. 1). In total, 25 distinct sequence types (STs) were identified. Among these 45 samples, 2 (4.4%) had no ST match in PubMLST.
Genomic sequencing revealed common genes in all isolates, including virulence and biofilm genes sepA and cap, and the staphylococcal multidrug resistance gene smr. Of the 48 isolates, 31 isolates (64.6%) had the mecA gene, but only 29 isolates (60.4%) had phenotypic oxacillin resistance. Also, 5 isolates (10.4%) were phenotypically susceptible to oxacillin yet harbored mecA, and 3 isolates (6.2%) showed phenotypic resistance despite not having mecA (Fig. 2). mecC was not observed in any isolate.
Figure 2. Phylogenetic structure and sequence types (left panel) of the Staphylococcus epidermidis population. Right panel showing gene presence or absence inferred from whole-genome sequencing de novo assemblies. Green cell corresponds to a full gene hit, and white cells to no hit. Only isolates included in both clonality analysis and phylogenetic analysis are shown. Note. AMR, antimicrobial resistance.
Epidemiologic data revealed that most of these patients had a significant prior healthcare exposure (Table 1). Among the 45 patients, 17 (37.8%) had 2 or more same-joint surgeries prior to a positive culture, and 11 patients (28.9%) had 2 or more aspirations prior to a positive culture.
Table 1. Univariate Table of Select Patient Demographics, Healthcare History, and Infection information.
Note. MSIS, Musculoskeletal Infection Society; PJI, prosthetic joint infection.
a Refers only to procedures performed on the joint of Staphylococcus epidermidis sample collection.
Among 45 strains included in the clonality analysis, we identified 6 distinct clusters of isolates. Cluster isolates accounted for 14 (31.1%) of all S. epidermidis joint isolates sequenced (Fig. 3). Also, 3 (21.4%) of the clonal S. epidermidis isolates and 1 (3.2%) of the nonclonal S. epidermidis isolates were obtained from native joints, and the rest were obtained from prosthetic joints. Each cluster of S. epidermidis clonal isolates comprised 2–4 different patients. The average number of days between the first and last clonal isolate in each cluster was 136 days (range, 25–330).
Figure 3. Single-nucleotide polymorphism (SNP) distance values for all pairs of samples of matching sequence types (STs). Only samples belonging to STs with >1 sample are shown. Blue cells correspond to pairwise distance of <25 SNPs, indicating clonality between isolates. For samples identified as part of a cluster, they are designated by the cluster number (eg, C2 for cluster 2).
Further epidemiologic investigation showed that some of these clonal isolates had ties to aspirations or operative procedures occurring during a similar timeframe (Fig. 4). For several clusters, procedures were within weeks of each other, and others were months apart. None of the cluster cases shared common surgical instruments. One cluster involved 2 patients with the same anesthesiologist for surgeries that were performed 2 months apart. Some cluster isolates seemed to have no overlapping connection. Overall, no convincing factor was identified to account for hospital transmission, but aspirations in close chronological proximity appeared to be a relevant factor in some cases. In all cases, clusters appear to have been self-limited and resolved with standard infection control practices.
Figure 4. Healthcare visits for patients with clonal Staphylococcus epidermidis isolates. Each row represents 1 patient. The bottom row of each cluster’s figure shows the time scale by month. Arrows indicate months that are consolidated in the figure. Positive cultures are shown only for the specified joint where the clonal isolate was taken from. Yellow circles indicate another positive culture from the same joint that was not a clonal Staphylococcus epidermidis isolate.
Discussion
This 4-year study of consecutive S. epidermidis joint isolates revealed a high degree of antimicrobial resistance and previously undescribed level of hospital transmission of S. epidermidis. S. epidermidis is one of the most common causes of PJI, yet relatively little is known about the transmission dynamics and molecular epidemiology of this organism. This study revealed several novel findings, including multiple clusters of hospital-acquired cases accounting for 14 (31%) of the 45 cases during a period of overall low surgical-site infection rates.
In addition, many isolates contained genetic resistance to common hospital disinfectants and antibiotics, perhaps yielding a competitive advantage in the hospital environment. These findings are supported by similar study results from Sweden. Researchers compared strains of S. epidermidis isolated from prosthetic joints with nasal strains. They found a high incidence of antimicrobial resistance in the infection isolates, but not the nasal isolates, indicating either selection of commensal resistant strains or acquisition of new strains during hospitalization. Reference Månsson, Bech Johannesen, Nilsdotter-Augustinsson, Söderquist and Stegger25,Reference Månsson, Tevell and Nilsdotter-Augustinsson26 In our study, a comprehensive outbreak investigation failed to identify specific epidemiologic links between the clonal S. epidermidis isolates although close temporal relationships in the care were observed. This finding suggests the possibility of reservoirs of S. epidermidis in healthcare settings. Because all analyses occurred in 2021 (1–5 years after the organisms were originally obtained) and no specific people, places, or objects were identified as possible modes of transmission, we did not attempt to search for these strains of S. epidermidis through WGS of random samples of the environment or staff.
Each cluster included 2–4 patients. Cluster cases differed by very few SNPs (Fig. 3), even when detected months apart. This finding suggests slow diversification and/or strain persistence, compared to the molecular clock and mutation rates of organisms such as S. aureus. Reference Datta, Yelin and Hochwald27,Reference Cabrera-Contreras, Santamaría and Bustos28 Compared to noncluster cases, cluster cases did not differ in any of the variables assessed (Table 1). However, both cluster and noncluster S. epidermidis cases were highly exposed to health care, which may have been a byproduct of our inclusion criteria. Clusters had a mean of 136 days (range, 25–330) between the first and last isolates in the cluster, and in some cases patients presented with infection months after the suspected transmission event or procedure. Thus, detection of these clusters was challenging in routine infection control surveillance. Indeed, during this time frame, the hospital had standardized infection ratios for hip and knee arthroplasties, which were statistically lower than national average, and there were no clinically detected PJI outbreaks from any organism.
All clusters were self-limited, suggesting that routine infection control practices were ultimately effective at eliminating the putative reservoir in all clonal cases. In hospital-acquired clonal bacterial spread, the possible modes of transmission of these isolates were 3-fold: (1) transmission from a common environmental reservoir to patients, (2) transmission from one patient to another (likely via an environmental reservoir or common healthcare worker), and (3) contamination of the microbiology laboratory. Reference Conlan and Mijares12,Reference Talbot, Jacko and Petit29 A fourth possibility, endogenous acquisition in patients who acquired a common strain from community source outside the hospital, is less likely given the disparate geographic regions of cluster patients and diverse sequence types (STs) in the series. Importantly, the study inclusion criteria did not require that the patient meet clinical criteria for prosthetic joint infection, allowing better detection of clonal isolates. This further highlights the importance of universal infection control practices including environmental decontamination, equipment disinfection and sterilization, and hand hygiene.
In this study, we assessed specific genes that may give S. epidermidis a competitive advantage in orthopedic infections. Of the genes selected, the most important findings were the presence of mecA and smr. MecA was present in 31 (64.6%) of 48 isolates, including 5 isolates that were phenotypically susceptible to β-lactams. Further investigation is required to better understand the significance of mecA positive, cefazolin-sensitive isolates, and implications for surgical prophylaxis and treatment. In addition, 100% of the isolates contained the smr gene, encoding an efflux pump that confers decreased susceptibility to quaternary ammonia compounds and other cationic substrates including common hospital disinfectants and potentially biguanides such as chlorhexidine. Reference Wassenaar, Ussery, Nielsen and Ingmer30 In addition, we also observed a partial match for qacA and qacB, additional resistance determinants for quaternary ammonia compounds and biguanides. This finding raises a genetic basis of concern for emerging resistance to chlorhexidine in staphylococi, supporting similar findings from other studies in Staphylococcus spp. Reference Månsson, Bech Johannesen, Nilsdotter-Augustinsson, Söderquist and Stegger25,Reference Månsson, Tevell and Nilsdotter-Augustinsson26,Reference Nogueira, Cidral and Lopes31–Reference Van den Poel, Saegeman and Schuermans34 In our arthroplasty population, our hospital has utilized a chlorhexidine and mupirocin decolonization protocol for preoperative arthroplasty patients with S. aureus colonization, as well as routine chlorhexidine decolonization for 2 days prior to surgery for all patients since 2005. Surgical skin preparation is achieved by either chlorhexidine/alcohol or iodine/alcohol. Reference Kim, Spencer and Davidson35 Despite this longstanding protocol, mupirocin resistance (mupA) was not observed.
The STs represented in this study were diverse. STs previously identified as being associated with pathogenicity, including ST2 and ST215, were present, but the majority of STs in this study were from strains not previously linked with virulence or pathogenicity. Furthermore, no one strain dominated over others in this study. Prior research has also linked ses genes to commensal strains of S. epidermidis. Reference Ortega-Peña, Vargas-Mendoza and Franco-Cendejas36,Reference Qi, Jin and Duan37 Our data do not support that ses genes are linked to pathogenicity or commensal strains in orthopedic infections.
This study was limited by the retrospective nature of the study design, which did not allow for real-time analysis of environmental or staff S. epidermidis strains. Thus, specific reservoir sources could not be determined. MSIS criteria were used to identify patients with PJI, but data were collected retrospectively and, in some cases, clinicians may not have ordered inflammatory markers or histology depending on clinical presentation. This factor could have artificially lowered the number of PJI in the sample. As part of our inclusion criteria, we only included patients with a prior procedure at our institution. We may have underestimated clonality if patients outside our inclusion criteria acted as a transmission reservoir. Furthermore, SNP thresholds for defining clonality are not well established for S. epidermidis, and variable strain clock rates and sample collection times can affect thresholds. Reference Stimson, Gardy, Mathema, Crudu, Cohen and Colijn38 Nevertheless, the low SNP threshold used in this study is a conservative cutoff that falls below observed SNP distances of other staphylococci collected in studies measuring host diversity. In addition, we selected only a handful of genes deemed to be important in S. epidermidis pathogenicity; it is possible that other genes play a role but were not analyzed. Finally, additional investigation is needed to determine phenotypic resistance patterns associated with smr genes and partial qacA/B genes in these S. epidermidis isolates.
In conclusion, we identified a high degree of antimicrobial resistance genes in S. epidermidis isolates from prosthetic and native hip and knee joint cultures. In addition, S. epidermidis joint isolates were part of a clonal cluster in 14 (31.1%) of 45 cases. Clusters were generally small (ie, 2–4 patients), separated by months in some cases, and they usually did not meet NHSN criteria for SSI, making detection unlikely in routine clinical and infection prevention practices. As WGS becomes more readily available, it will be a valuable tool in detecting low-level hospital transmission. In this study, clusters appear to be self-limited, demonstrating benefit to universal infection control practices.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2023.253
Acknowledgments
Financial support
This study was funded by the Paul Fremont-Smith Jr Fund for Infection Prevention Research.
Competing interests
Mohamad Sater, Emma Briars, Ian C. Herriott, and Miriam Huntley are Day Zero Diagnostics employees. Both remaining authors report no conflicts of interest relevant to this article.
