Diagnostic and commensal Staphylococcus pseudintermedius genomes reveal niche adaptation through parallel selection of defense mechanisms

Abstract

Staphylococcus pseudintermedius is historically understood as a prevalent commensal and pathogen of dogs, though modern clinical diagnostics reveal an expanded host-range that includes humans. It remains unclear whether differentiation across S. pseudintermedius populations is driven primarily by niche-type or host-species. We sequenced 501 diagnostic and commensal isolates from a hospital, veterinary diagnostic laboratory, and within households in the American Midwest, and performed a comparative genomics investigation contrasting human diagnostic, animal diagnostic, human colonizing, pet colonizing, and household-surface S. pseudintermedius isolates. Though indistinguishable by core and accessory gene architecture, diagnostic isolates harbor more encoded and phenotypic resistance, whereas colonizing and surface isolates harbor similar CRISPR defense systems likely reflective of common household phage exposures. Furthermore, household isolates that persist through anti-staphylococcal decolonization report elevated rates of base-changing mutations in - and parallel evolution of - defense genes, as well as reductions in oxacillin and trimethoprim-sulfamethoxazole susceptibility. Together we report parallel niche-specific bolstering of S. pseudintermedius defense mechanisms through gene acquisition or mutation.

Fig. 1. Niche adaptation is not apparent via total encoded gene architecture.

a Core gene phylogeny of 493S. pseudintermedius isolates. Cohort, MLST, household (if applicable), and ARG burden are indicated by concentric color strips. b, c Jaccard dissimilarity by accessory gene content represented by b principal coordinate analysis ordination and c beta diversity compositions of within-group and between-group Jaccard distance. White, light gray, and dark gray violins indicate within cohort and niche, niche-only, and across cohort and niche pairwise comparisons, respectively, between 94 animal diagnostic, 163 human diagnostic, 90 pet colonizing, 33 human colonizing, and 113 environment isolates. Box plots extend from the 25th to 75th percentile with median line displayed, and whiskers extend to 1.5 x inter-quartile range. d S. pseudintermedius pangenome composition. e Total gene content breakdown by COG category.

Fig. 2. Diagnostic isolates are distinguished by greater ARG burden and phenotypic resistance.

a Antibiotic resistance gene content and antibiotic susceptibility testing results, organized by antibiotic class. Only isolates with ≥2 ARGs (n = 188) are shown. b AST interpretation for commonly-prescribed antibiotics (Abbreviation TMX, trimethoprim-sulfamethoxazole). Asterisks indicate a cohort is significantly more resistant or susceptible to the respective antibiotic relative to all other isolates (q < 0.05, Chi-square). c ARG count for diagnostic and household isolates (n = 255 and 238 isolates, respectively; p = 3.452e-09, two-sided Mann–Whitney). Box plot extends from the 25th to 75th percentile with median line displayed, and whiskers represent 10–90 percentiles. d Presence of mecA among isolates (n = 493), distributed by total ARG count. mec-positive isolates are significantly overrepresented among high ARG burden isolates (≥9 ARGs) relative to low ARG burden isolates (2–8 ARGs).

Fig. 3. Syntenic gene structures continue to distinguish select diagnostic isolates following strain correction.

a Pairwise comparisons of ANI and coverage within lineage clusters. Most comparisons reside in the top-right quadrant (>99.999% ANI, >98% coverage), indicating most isolates of the same lineage are members of the same strain cluster. b Distribution of strain clusters, by number of isolates within each cluster and cohort of origin (n = 56 clusters comprising 184 isolates). c Genes found by GWAS to be overrepresented among diagnostic or household isolates and strain cluster representatives. Only isolates or strain clusters with at least one GWAS-identified, overrepresented gene are displayed. Blue and red brackets denote a CRISPR-Cas system and resistance gene cluster, respectively. d, e Representative appearances of the d Tn5405-mediated resistance gene cluster and e type-IIIa CRISPR-Cas system among diagnostic isolates.

Fig. 4. Household isolates share more spacers with isolates within their niche than with diagnostic isolates, agnostic of pairwise ANI.

a Co-occurrence matrix of spacers shared at 100% ANI between only the 17 mecA-positive isolates with Type-IIIa CRISPR systems. Bubble size and saturation reflect total quantity of spacers and percent of spacer repertoire shared between two isolates, respectively. HI_0003-HI_0186 and HI_0035-HI_0177 are two examples of isolate pairs that fall into lineage but not strain clusters, yet have virtually identical spacer repertoires. b Co-occurrence matrix of spacers shared at 100% ANI between all CRISPR-positive isolates. Isolate names and intra-cohort pairwise comparison bubbles are colored by cohort. c Pairwise ANI by shared spacer count. d Pairwise shared spacer counts of household isolates with other household isolates (intra), and with diagnostic isolates (inter) (n = 52 and 104 household and diagnostic isolate and cluster representatives, respectively; q = 0.00120, FDR-corrected Mann–Whitney). Barplot and error bars represent mean + SD.

Fig. 5. Persisting strains accrue mutations in defense mechanism genes amidst household MRSA decolonization.

a Observance of strain clusters across households, faceted by subcohort. MRSA decolonization is highlighted and months without sample collections are noted in gray. Samples collected during the highlighted HOME2 intervention were collected immediately prior to MRSA decolonization and are considered baseline samples. Samples collected during the highlighted SHINE intervention were collected during MRSA decolonization. Bubble size corresponds to the number of isolates per strain cluster found in a collection month. b Defense mechanism genes as a percent of total genes accruing synonymous SNSes only (SSO) and non-synonymous SNSes (NSS), compared against the expected distribution of a reference S. pseudintermedius genome (q = 0.0456, FDR-adjusted one-sided Binomial test). c Number of genes accruing non-synonymous SNSes in multiple clusters, over 10,000 iterations. 9500 iterations are captured to the left of the dashed vertical line. d Schematic representation of the five NSS genes mutated in multiple strain clusters. COG key: [M] Cell wall/membrane/envelope biogenesis, [P] Inorganic ion transport & metabolism, [S] Function unknown, [V] Defense mechanisms. e AST results for oxacillin and trimethoprim-sulfamethoxazole (q = 0.009 and 9.878e−05, respectively, FDR-corrected two-sided unpaired T-test) for Cluster 03 isolates before and during household MRSA decolonization.