Background: Pediatric osteoarticular infections are commonly caused by Staphylococcus aureus. The contribution of S. aureus genomic variability to pathogenesis of these infections is poorly described. 

Methods: We prospectively enrolled 47 children over 3 1/2 years from whom S. aureus was isolated on culture---12 uninfected with skin colonization, 16 with skin abscesses, 19 with osteoarticular infections (four with septic arthritis, three with acute osteomyelitis, six with acute osteomyelitis and septic arthritis and six with chronic osteomyelitis). Isolates underwent whole genome sequencing, with assessment for 254 virulence genes and any mutations as well as the creation of a phylogenetic tree. Finally, isolates were compared for their ability to form static biofilms and compared to the genetic analysis.

Results: No sequence types predominated amongst osteoarticular infections. Only genes involved in the evasion of host immune defenses were more frequently carried by isolates from osteoarticular infections than from skin colonization (p=.02). Virulence gene mutations were only noted in 14 genes (three regulating biofilm formation) when comparing isolates from subjects with osteoarticular infections and those with skin colonization. Biofilm results demonstrated large heterogeneity in the isolates’ capacity to form static biofilms, with healthy control isolates producing more robust biofilm formation.

Conclusions: S. aureus causing osteoarticular infections are genetically heterogeneous, and more frequently harbor genes involved in immune evasion than less invasive isolates. However, virulence gene carriage overall is similar with infrequent mutations, suggesting that pathogenesis of S. aureus osteoarticular infections may be primarily regulated at transcriptional and/or translational levels.

Construction of the Study Cohort: Subjects were prospectively enrolled between June 12^th, 2016, and December 2^nd, 2019. Subjects admitted to our hospital or Pediatric Emergency Department who were under 18 years of age and without known immune deficiencies or post-operative or orthopedic implant-associated infections were eligible for enrollment. Subjects were enrolled from the following four groups of osteoarticular infections: 1.) acute osteomyelitis (symptoms <14 days, normal orthopedic plain films at admission and elevated inflammatory markers, as previously described) 2.) acute septic arthritis (any subject requiring an arthrotomy for suspected septic arthritis with the growth of S. aureus on a culture of blood and/or synovial fluid) 3.) chronic osteomyelitis (symptoms >14 days at admission, abnormal orthopedic plain films at admission and histopathology supporting the diagnosis if available, with normal or mildly elevated inflammatory markers) 4.) concurrent acute septic arthritis and acute osteomyelitis. To better evaluate genomic composition across a spectrum of invasion, S. aureus isolates collected from two groups of controls were utilized: 1.) children with skin and soft tissue abscesses (with sterile blood cultures, if obtained, and no evidence of systemic invasions such as pneumonia or osteoarticular infections) and 2.) uninfected children with asymptomatic skin colonization who were admitted for non-infectious conditions (e.g. febrile seizures, asthma exacerbations). Demographic and clinical information were obtained for all subjects from the electronic medical record, save for healthy, uninfected controls who were promised anonymity.

Microbiological Methodology: For subjects with infection, bacterial isolates from clinical cultures were confirmed as S. aureus via matrix-assisted laser desorption time-of-flight (MALDI-TOF) analysis, and then collected from sub-cultures for sequencing. Multiple isolates may have been collected from the same subject (e.g. if cultures isolated S. aureus from multiple time points during admission), though genomic comparisons unless stated otherwise were based on the initial isolate. For uninfected control subjects, axillary or nasal swabs were collected and plated on mannitol salt agar. Coagulase-positive isolates fermenting mannitol underwent confirmatory MALDI-TOF analysis and confirmed S. aureus isolates were saved for sequencing. All S. aureus isolates were frozen in 10% glycerol stock at -80 degrees until batched analysis. Susceptibility testing (to differentiate methicillin-resistant S. aureus, MRSA, from methicillin-susceptible S. aureus, MSSA) was performed with disk diffusion prior to freezing and confirmed with molecular analysis for the mecA gene. 

Sequencing Methodology: Prior to sequencing, isolates were re-cultivated in tryptic soy broth, mixed in DNA/RNA Shield lysis tubes™ (Zymo Research™), and centrifuged at 10,000 x g for 1 minute. DNA was isolated using the ZymoBIOMICS DNA isolation kit following the manufacturer’s recommended protocol (Zymo Research™). More specifically, the resulting supernatant was added to a Zymospin™ filter, centrifuged at 8000 x g for 1 minute followed by the addition of DNA binding buffer. The resulting mixture was added to a Zymospin™ column, centrifuged at 10,000 x g for 1 minute followed by rinses with DNA wash buffer. This was added to DNAse/RNAse free water and centrifuged at 10,000 x g for 1 minute. DNA was eluted from this using a Zymospin™ filter via centrifugation. The resulting DNA was prepared for Illumina next-generation sequencing using the Illumina Nextera XT DNA library prep kit, per recommended instructions. Completed sequencing libraries were assessed for quality and concentration by gel electrophoresis (Agilent™) and Qubit fluorometric quantitation (Thermo Fisher Scientific™), respectively. Completed libraries were pooled in equimolar ratios and underwent whole genome sequencing via 2x250 bp sequencing using v3 sequencing reagents on an Illumina MiSeq (see supplementary table for the number of sequencing reads).

Supplementary Table: Summary statistics for S. aureus genome assemblies. Assemblies were generated with short read data in Unicycler.

Bioinformatic and Phylogenetic Methodology: For analysis of virulence genes, FASTA sequences were identified for 254 virulence genes (genes taken from a published compilation and a supplementary literature search). Sequencing data of S. aureus were aligned using BWA 0.7.17 using S. aureus reference genome NCT8325 downloaded from NCBI. Binary alignment map (BAM) files were sorted and indexed using Samtools 1.9. BCFTools 1.9 was used to count allele frequency from the BAM files. Transcriptome information of S. aureus was downloaded from GenBank as CP000253.1 general feature file and converted to gene transfer format (GTF) using GFF Utilities. Then FeatureCounts was used to count reads aligned to genes. Proportion tests were used to assess for a proportional difference of variants between case and control groups. Adjusted p < 0.05 was considered statistically significant.

For phylogenetic analysis, raw sequencing reads were trimmed with Trim Galore using default settings. Assemblies were created with Unicycler (Supplementary table). Sequence types were determined using ARIBA. Forty-seven isolates (one from each patient) were included for phylogenetic analysis. For these 47 isolates, a core genome alignment was created with Roary. A maximum likelihood phylogeny was built from the core genome alignment with IQ-TREE using 5000 ultrafast bootstraps and a GTR+G model of nucleotide substitution. Phylogenies were visualized using GGTREE. Branches were analyzed by year, source of the sequenced isolate, the presence of the mecA gene, and the type of infection. Given that the traditional classification of the types of osteoarticular infection as either septic arthritis, acute osteomyelitis, or chronic osteomyelitis may be somewhat arbitrary and not reflective of a continuum of infection (e.g. both septic arthritis and chronic osteomyelitis may arise as complications of acute osteomyelitis), a severity of illness score was calculated for subjects with acute osteomyelitis as previously described for assessment of phylogenetic relatedness and disease severity. 

Static Biofilm Assay: Static biofilm assays were conducted using a modified method of Cassat et al. that we recently described. Briefly, 96-well plates were coated overnight at 4 °C with pooled human plasma (IPLANAC; Innovative Research, Novi, MI). Overnight cultures in duplicate for each strain were grown in TNB [trypticase soy broth (Becton, Dickinson and Company, Sparks, MD) with 0.5% w/v dextrose (VWR Analytical, Radnor, PA) and 3% w/v NaCl (Fisher Scientific, Waltham, MA)] at 37 °C with 220 rpm. Overnight cultures were OD_600nm matched to within 0.05 and then diluted 1:200 %v/v in fresh media. Coated wells were gently washed with phosphate-buffered saline (PBS) and then inoculated with six technical replicates per biological replicate. PBS in coated wells served as a negative stain control. Plates were then incubated statically for 24 h at 37 °C. The non-adherent culture was aspirated, washed twice with PBS, and then and the wells were fixed with 100 % v/v ethanol. Ethanol was removed and the plate was allowed to dry for 10 min. Biofilm was stained with 0.1 % w/v crystal violet (Sigma-Aldrich) for 2 minutes and then aspirated and washed twice with PBS. The stain was eluted with 100 % v/v ethanol by shaking for 10min and then diluted 1:10 in 100 % v/v ethanol for OD_595nm measurement. A USA300 S. aureus MRSA isolate (AH1263) and its isogenic agr-deletion mutant (AH1292) were included on each plate as internal controls and biofilm comparators.

Statistical Methodology: Descriptive statistics including counts and frequencies were used to profile participant characteristics, including the type of osteoarticular infection. For categorical variables, chi-square tests were calculated using Fisher’s exact test for cell sizes less than five. For continuous variables, means, medians, and interquartile range (IQR) were assessed. In addition to analysis of the distribution of individual genes between types of infection and controls, genes were also grouped into families according to putative function (toxin, adhesins, antibiotic resistance, immune evasion, proteases, hemolysins/leukocidins/hyaluronidases) as described in the literature, and the mean proportion positive for each family was calculated. Assignment of genes to a family was based upon putative functions listed on the website www.uniprot.org, a recently published review on the topic and supplementary literature review. Differences in mean distribution between osteoarticular infections vs. healthy controls and vs. skin abscess controls were calculated using a t-test. Descriptive statistics were also employed to evaluate gene carriage in isolates from separate sources in the same patient (e.g. bone and blood cultures) and isolates from the same patient serially over time. All statistical tests were two-sided. To decrease the likelihood of false positive findings given the large number of statistical comparisons undertaken, the Benjamini and Hochberg correction was used and reported as the final p-value. The quantity of biofilm production for each included bacterial isolate was compared between skin soft tissue infections, bone and joint infections and healthy controls, and to the biofilm comparators and internal controls AH1263 and to AH1292. Biofilm quantities were evaluated on a log scale to accommodate non-normal distributions. Mixed models were estimated by accounting for repeated measures. All statistical tests used a two-side alpha value of .05. Analyses were conducted using Statistical Analysis Systems (SAS) software, v. 9.4 (Cary, N.C.).

Microsoft Excel, .tsv.