Clinical context and antimicrobial susceptibility
The 23 S. epidermidis isolates analyzed in detail here were cultured from clinical specimens collected from deep sternal wounds, central line-related sepsis or infected implanted medical devices such as joint prostheses. Diagnostic culturing has been performed in clinical microbiology laboratories serving healthcare in Perth, Australia; in Umeå, Sweden; or in Östersund, Sweden (Supplementary Table S1).
The antimicrobial susceptibility of the 23 S. epidermidis isolates was tested using a panel of 31 antimicrobial drugs, and the isolates were found to be resistant to multiple antimicrobial agents, as shown in Fig 1.
The 23 isolates were generally resistant to cefoxitin and other β-lactam antimicrobial agents and to at least 4 out of 13 additional distinct antimicrobial classes (fusidic acid, mupirocin, macrolides/clindamycin, tetracyclines, quinolones, gentamicin/tobramycin, rifampicin, chloramphenicol, nitrofurantoin, daptomycin, linezolid, folate pathway inhibitors, and glycopeptides). Broth dilution MIC values determined using the EUCAST methodology are detailed in Supplementary Table S2.
Genome phylogenies
We identified the Illumina short-read sequenced isolates as belonging to ST2 (n=11), ST215 (n=11) and ST22 (n=1) as defined by the S. epidermidis MLST scheme (ref Thomas). By combining Illumina and Nanopore long-read sequencing data, we finished a selected ST215 genome (clinical isolate SE14) by reconstructing a 2,676,785 bp circular chromosome and a 2,326 bp circular plasmid. The average depth of the chromosome coverage was 871× for Illumina and 78× for Nanopore, and the plasmid coverage was 340× and 54×, respectively. We saved all reads in excess not mapping to the two contigs (approximately 1,000 (0.2%) nanopore reads and 200.000 (0.16%) Illumina reads) but found no evidence for additional plasmids by reassembly. We calculated a complete S. epidermidis whole-genome phylogeny based on multiple alignments using 1361 publicly available completed genomes representing 58 STs to place ST215 within the total known genomic diversity of the species (Fig. 2A). ST215 clustered with the previously described major genomic cluster A in a phylogeny with two additional major clusters, C and B, reproducing previous findings (23, 24). In accordance with a prior study, the ST2 genomes exhibited a paraphyletic pattern characterized by two distinct groups of ST2 genomes that did not share a recent common ancestor. To put all 23 isolates sequenced in the context of multiple described healthcare-associated STs, we calculated an additional genome phylogeny based on short-read data with a set of 60 S. epidermidis genomes, including STs commonly described from healthcare (Supplementary Table S3) (23). We mapped the short reads on a 1,613,625-nucleotide core and found 2,456 single-nucleotide polymorphisms, resulting in a tree that reproduced the division of the major clusters A, B, and C and showed that 11 ST215 genomes and 11 ST2 genomes clustered tightly but separately within major cluster A (Fig. 2B).
Hereafter, we analyzed and compared all the ST2 and ST215 genomes generated for this study to explore differences and similarities between the two genetic lineages, which are the two most common multidrug-resistant S. epidermidis lineages in hospitals in Sweden.
Analyses of homologous recombination in ST215 and ST2
By constructing unrooted phylogenetic networks using Splits Tree3.2 software, we found a greater network among the genomes assigned to ST2 than among the ST215 genomes, suggesting that recombination events caused more conflicting phylogenetic signals for ST2 (Fig. 3).
To further investigate recombination events in the genomes of the ST2 and ST215 isolates, we used BratNextGen software and detected recombination events, which are graphically displayed in Fig 3. There were nine recombination events in the ST2 genome analysis, with an average size of 30 544 bp, two of which partially overlapped, suggesting recombination hotspots in this region (Fig. 4). According to the analysis of the ST215 genomes, 11 distinct events were detected, with an average size of 46,245 bp and no genomic location overlap.
We identified antibiotic resistance and virulence genes in the recombining nucleotide stretches and in parts of the genome with no recombination signal by BRATNextGen analysis (Supplementary Table S4). Antibiotic resistance genes (ARGs) found in the recombining genomic regions included bacA, norA, evgA, gyrA, msbA, glpT, and tetR, and the virulence genes in these regions included capA, capB, capC, capD, sspB, gehD, aae, and sspA (Supplementary Table S5). Among the ST2 genomes, ARGs were more common in the recombinant regions than expected by chance (p = 0.008), while virulence genes were not significantly more common. In contrast, among the ST215 genomes, virulence genes were more common in the predicted recombinant regions than they were by chance (p = 0.021), while ARGs were not significantly more common in these regions. Alignments of the ST2 and ST215 genomes have been posted as open science data in the European general-purpose open-access research data repository Zenodo hosted by CERN (25).
Mapping of genes encoding antimicrobial resistance and virulence
The presence or absence of genes described as ARGs or virulence genes were mapped in short-read genome data of the 11 ST215 isolates and 11 ST2 isolates (Supplementary Table S1). We identified 30 ARGs, 26 of which were assigned to specific antimicrobial agent classes and 35 of which were putative virulence genes. Fifty of these genes, 22 ARGs (mecA, mecR1, blaZ, norA, dfrC, bacA, ileS, blal, murA, carA, fabl, parC, evgA, mgrA, msbA, rpsL, gyrA, srmB, glpT, tetR, ykkC, and aac(6´)-le-aph(2´´)-la) and 29 virulence genes (sitA, sitB, sitC, graR, graS, SE0415, SERP0296, geh1-Lipase, sepA, aae, atle, embp, ebp, capA, capB, capC, capD, dltA, dltB, dltC, dltD, mprF, sspB, gehC, gehD, sdrF, sdrG, sdrH, and sspA), were present among all the ST2 and ST215 genomes (Supplementary Table S5). Among the genes that differed in presence among any pair of isolates, some patterns were related to a specific genetic background (ST), time, or place, as outlined in Table 1. The fusB gene encoding fusidic acid resistance was present in all the isolates except for the ST2 isolates from Umeå and Östersund in Sweden (the SE07-SE11 isolates). The gene tetK, encoding resistance to the tetracycline class of antibiotics, appeared to have been lost over time in the ST215 isolates (SE16-SE23) and was not present in any of the ST2 isolates. The virulence genes aap, icaA, icaB, icaC icaD and icaR were present in all the ST2 isolates but were consistently absent among the ST215 isolates (Table 1). In the completed genome of the ST215 isolate SE14, the antimicrobial resistance gene ermC was found to be located on the plasmid, while the other resistance genes of SE14 were on the chromosome. None of the putative virulence genes in the SE14 genome were on the plasmid, and all of them were on the chromosome.
Table 1. Antibiotic resistance and virulence genes differing between 22 S. epidermidis genomes of ST2 or ST215.
Isolate ID
|
Sequence type
|
|
Presence (+) or absence (-) of gene
|
|
|
|
Antibiotic resistance genes
|
Virulence genes
|
|
|
msrA
|
qacA
|
qacB
|
fusB
|
ermC
|
mupA
|
tetK
|
cat
|
aap
|
icaA
|
icaB
|
icaC
|
icaD
|
icaR
|
SE01
|
ST2
|
+
|
+
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE02
|
ST2
|
+
|
+
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE03
|
ST2
|
+
|
+
|
+
|
+
|
-
|
-
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE04
|
ST2
|
+
|
+
|
+
|
+
|
-
|
-
|
-
|
-
|
+
|
+
|
-
|
+
|
+
|
+
|
SE05
|
ST2
|
+
|
+
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE06
|
ST2
|
+
|
+
|
+
|
+
|
-
|
-
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE07
|
ST2
|
-
|
+
|
+
|
-
|
-
|
-
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE08
|
ST2
|
-
|
+
|
+
|
-
|
+
|
-
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE09
|
ST2
|
-
|
+
|
+
|
-
|
+
|
-
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE10
|
ST2
|
-
|
+
|
+
|
-
|
-
|
-
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE11
|
ST2
|
-
|
+
|
+
|
-
|
-
|
-
|
-
|
-
|
+
|
+
|
+
|
+
|
+
|
+
|
SE12
|
ST215
|
-
|
+
|
+
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE13
|
ST215
|
-
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE14
|
ST215
|
-
|
+
|
+
|
+
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE16
|
ST215
|
-
|
+
|
+
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE17
|
ST215
|
-
|
-
|
-
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE18
|
ST215
|
-
|
+
|
+
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE19
|
ST215
|
-
|
-
|
-
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE20
|
ST215
|
-
|
-
|
-
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE21
|
ST215
|
-
|
-
|
-
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE22
|
ST215
|
-
|
+
|
+
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
SE23
|
ST215
|
-
|
-
|
-
|
+
|
-
|
-
|
+
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
To investigate the agreement between antibiotic ARGs in the genome and phenotypic antimicrobial resistance, the presence or absence of a gene (or presence/absence of a resistance mutation in gyrA, parC, or rpsL) was matched with the antimicrobial susceptibility testing results. In general, we found good agreement by analyzing 26 genes that we could assign to a specific antimicrobial agent class (Fig. 5).
SNP accumulation over time among the ST215 genomes
After removing the recombined nucleotide stretches, we found a correlation between the clinical sampling date and the accumulation of nucleotide substitutions in the ST215 core genomes isolated from 1995 to 2008 by applying a Monte Carlo-based test and by a Mantel test (P = 0.042 and P = 0.01, respectively) (Supplementary Fig. S1). In contrast, testing of the ST2 genomes indicated no time dependence.