Staphylococcus epidermidis pan-genome sequence analysis reveals diversity of skin commensal and hospital infection-associated isolates

Abstract

Background: While Staphylococcus epidermidis is commonly isolated from healthy human skin, it is also the most frequent cause of nosocomial infections on indwelling medical devices. Despite its importance, few genome sequences existed and the most frequent hospital-associated lineage, ST2, had not been fully sequenced.

Results: We cultivated 71 commensal S. epidermidis isolates from 15 skin sites and compared them with 28 nosocomial isolates from venous catheters and blood cultures. We produced 21 commensal and 9 nosocomial draft genomes, and annotated and compared their gene content, phylogenetic relatedness and biochemical functions. The commensal strains had an open pan-genome with 80% core genes and 20% variable genes. The variable genome was characterized by an overabundance of transposable elements, transcription factors and transporters. Biochemical diversity, as assayed by antibiotic resistance and in vitro biofilm formation, demonstrated the varied phenotypic consequences of this genomic diversity. The nosocomial isolates exhibited both large-scale rearrangements and single-nucleotide variation. We showed that S. epidermidis genomes separate into two phylogenetic groups, one consisting only of commensals. The formate dehydrogenase gene, present only in commensals, is a discriminatory marker between the two groups.

Conclusions: Commensal skin S. epidermidis have an open pan-genome and show considerable diversity between isolates, even when derived from a single individual or body site. For ST2, the most common nosocomial lineage, we detect variation between three independent isolates sequenced. Finally, phylogenetic analyses revealed a previously unrecognized group of S. epidermidis strains characterized by reduced virulence and formate dehydrogenase, which we propose as a clinical molecular marker.

Figure 1

Body sites analyzed. A collection of 71 S. epidermidis isolates from 7 healthy volunteers across 15 body sites were selected for characterization (number of isolates from site in parenthesis).

Figure 2

Commensal S. epidermidis has an open pan-genome. Pan-genome size of S. epidermidis. (a) Gene accumulation curves for pan-genome (squares) and core-genome (circles) as a function of genomes sequenced (N). Error bars are ±1 standard deviation for the 100 simulations. Core gene data are fit by: y = a·exp^-N/b + c. Pan-genome data are fit by: y = a·N^b + c. (b) Partitioning of core and variable genes for each COG. COGs significantly enriched (P < 0.05, Fisher exact test) in variable or core genes are marked with single and double asterisks, respectively. The dashed line indicates the expected proportions.

Figure 3

S. epidermidis variable genome is dominated by genes shared by few strains. Only the variable genome is shown. Gene clusters present in all genomes or present in only a single genome are omitted. The presence of a gene (and its paralogs) is indicated by a black bar. Nosocomial and commensal isolate names are colored red and green, respectively. The dendrogram was generated using complete linkage clustering of the presence/absence data.

Figure 4

S. epidermidis taxon is made up of two groups. Maximum likelihood phylogenetic tree of S. epidermidis based on four-fold degenerate positions of genes in the core genome. The table shows the sequence type (ST) from multi-locus sequence typing. New sequence types and sequence types with a new allele are indicated by single and double asterisks, respectively. The RP62A and ATCC12228 strains are reference strains (NC_002976, NC_004461). Isolates SK135, BCM-HMP0060 and W23144 were sequenced as part of the Human Microbiome Project Reference Genome catalog. Nosocomial and commensal isolate names are colored red and green, respectively. Bootstrap values are shown on each branch. Outgroup (S. aureus) not shown.

Figure 5

fdh gene neighborhood shows co-linearity across staphylococci. The gene neighborhood of fdh. Strains from group A and B are marked with superscript 'a' and 'b', respectively. Genes with conserved function are the same color. Genome alignments were generated using RAST.

Figure 6

Principal components anaylsis of ortholog clustering. The orthologous gene cluster data for 35 genomes was analyzed using principal components analysis. Nosocomial isolates are shown as red circles and ST2 nosocomial isolates are filled red circles. Commensal isolates are black circles. The dashed ovals indicate the phylogenetic groups described in Figure 4 but are not statistically defined. Percentages of variance explained by PC1 and PC2 are 13.8% and 9.1%, respectively.

Figure 7

in vitro biofilm formation is associated with isolation source. in vitro biofilm formation on polystyrene surface of 71 commensal and 28 medical device (nosocomial) S. epidermidis isolates. Commensal isolates are subdivided by the skin type (dry, sebaceous and moist) as defined in Grice et al. [24]. Each point represents the average of four replicates. Statistical significance was calculated using Mann-Whitney, two-tailed.