Introduction

Staphylococcus species are classified as either coagulase-negative or coagulase-positive based on the ability of the coagulase enzyme to stimulate clot formation in plasma1. Coagulase-negative Staphylococcus (CoNS) species are ubiquitous members of the human microflora, frequently colonizing the skin and mucosal surfaces2. They have been historically considered non- or less pathogenic compared to the more prominent coagulase-positive Staphylococcus aureus1. However, many CoNS species are now considered as major opportunistic pathogens that cause serious infections in humans and animals. They have been implicated in bloodstream infections3, laryngological diseases4, skin and soft tissue infections5, infections of the central nervous system6, and infections from indwelling/implanted medical devices (e.g., catheters, prosthetic implants)7,8. Approximately 8% of all cases involving valve endocarditis are associated with CoNS species, with 25% of cases resulting to mortality9. Increasing prevalence of multidrug-resistant CoNS and resistance to newer antimicrobial agents have compounded the burden of CoNS infections and limits available treatment options10,11.

The evolutionary history of CoNS has not been well characterized because of the historical lack of recognition for their clinical importance and poor taxonomic sampling1,12. They are also often misclassified or overlooked in sampling efforts. Among Staphylococcus species, most investigations of evolutionary processes and genetic diversity have focused on the more well-known S. aureus (e.g., refs. 13,14,15). However, there is growing evidence that recombination has played a major role in CoNS evolution16,17,18. Homologous recombination involves the incorporation of exogenous DNA segment that exhibits high sequence similarity with existing DNA segment in the chromosome19. It can lead to either allelic replacements or the addition of new DNA segments when two flanking regions of high DNA similarity initiate the recombination process and mediate successful horizontal gene transfer20. We recently reported that rates of recombination rates varied among ten CoNS species in a global dataset18. Despite the significance of these findings, our previous inferences were derived from genome sequences from highly variable ecological sources (humans, animals, food products, environment) and geographical origins (seven continents and the International Space Station)18, which can greatly influence patterns and frequencies of recombination.

We hypothesized that CoNS with overlapping ecologies within a defined geographical space will exhibit frequent but structured patterns of horizontal gene transfer through homologous recombination and mobile genetic elements. Here, we sequenced the genomes of 191 CoNS isolates representing 15 species from animals diagnosed with disease in a veterinary diagnostic laboratory. We found that CoNS are an important reservoir of diverse phages, plasmids, and mobilizable genes conferring resistance to 13 antimicrobial classes. We identified frequently recombining genes, variation in accessory genomes, and patterns of DNA donation and receipt within and between CoNS species. Our findings provide a firm understanding of CoNS evolution and inform surveillance efforts and design of future treatment strategies for CoNS infections.

Results

Species and genomic diversity of CoNS from diseased animals

A total of 191 CoNS isolates from diseased animals were retrieved from specimens sent to the New Hampshire Veterinary Diagnostic Laboratory (NHVDL) in the New England region of the United States (Fig. 1a, Supplementary Data 1). The isolates were sampled from 2017 to 2020 (Fig. 1b) from 10 animal species (Fig. 1c). Most isolates were obtained from cats (n = 69), dogs (n = 42), cows (n = 27), and horses (n = 23). Canine and feline hosts harbored the highest species diversity of CoNS, with 12 and 11 CoNS species, respectively (Fig. 1c). However, this is likely due to sampling bias in our dataset that heavily favored these two animal hosts. In vitro screening revealed the presence of 22 methicillin-resistant isolates, representing 11.52% of the dataset (Fig. 1d).

Fig. 1: Sampling characteristics of the 191 CoNS genomes.
figure 1

a Geographical sources of CoNS species. Pie charts show the proportion of CoNS species per state. Numbers in parenthesis indicate the number of CoNS genomes per state. Two isolates came from animals from Alabama but were diagnosed in New England are not shown here. b Distribution of CoNS isolates and species per year of sampling. c Distribution and frequency of CoNS isolates per animal host. d Proportion of methicillin-resistant and methicillin-susceptible isolates per CoNS species determined by in vitro testing. Numbers on top of bars in panels b and c indicate the number of genomes in that group. For visual clarity in panel d, only the first three letters of the species name are shown.

Full size image

Whole genome sequencing of the isolates yielded high-quality genomes (Supplementary Data 1, Supplementary Fig. 1). Genome sizes ranged from 2.17 Mbp to 3.13 Mbp (mean = 2.5 Mbp). The mean number of contigs was 34 (range = 11–70), while the mean N50 was 329,792 bp (range = 70,109–1,390,089 bp). We assessed species delineation using the average nucleotide identity (ANI) of all orthologous genes shared between any two genomes with a 95% cutoff to define a species21 (Supplementary Fig. 2 and Supplementary Data 2). We also built a maximum likelihood phylogenetic tree based on 735,124 single nucleotide polymorphisms (SNPs) detected in the 1.27 Mbp alignment of 1317 core genes (Fig. 2a). We identified a total of 15 CoNS species. The most common species were Staphylococcus felis (n = 39 genomes), Staphylococcus simulans (n = 31), Staphylococcus xylosus (n = 30), and Staphylococcus epidermidis (n = 25), whereas the least common species were Staphylococcus succinus (n = 2), Staphylococcus agnetis (n = 3), and Staphylococcus coagulans (n = 3). However, we also detected 54 genomes that showed taxonomic discrepancies between the initial species identification using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) and the genome-based species delineation. Nonetheless, within each named CoNS species, isolates from different geographical sources, animal hosts, or year of sampling were intermingled within the phylogeny.

Fig. 2: Phylogenetic relationships and pan-genome characteristics of the 191 CoNS genomes.
figure 2

a Midpoint-rooted maximum likelihood phylogenetic tree based on sequence alignment of 1317 core genes. Tree scale represents the number of nucleotide substitutions per site. Outer rings show the in vitro methicillin susceptibility, sampling site (U.S. state), year of sampling, and animal host (outer to inner rings). Branches are colored according to species. b Sizes of core genomes and accessory genomes per CoNS species. Only the 12 most common species are shown. (Left) Box plots showing the number of accessory genes at 25th, 50th, and 75th percentile. Black dots show the number of accessory genes outside the interquartile range. (Right) Bar plots showing the number of core genes in each CoNS species.

Full size image

For those species represented by more than three genomes, we estimated their core and accessory genome variation (Fig. 2b). Species with the highest number of core genes were Staphylococcus pseudoxylosus (n = 2433 genes), S. xylosus (n = 2322) and Staphylococcus equorum (n = 2271), whereas Staphylococcus hominis (n = 1860), S. felis (n = 1944) and Staphylococcus chromogenes had the lowest number of core genes (n = 1958). The number of accessory genes were highly variable across different species. Staphylococcus lugdunensis had the least number of accessory genes (range = 118–189; median = 161), whereas accessory gene content was highest in S. pseudoxylosus (range = 308–494; median = 363), S. equorum (range = 296–463; median = 359), Staphylococcus warneri (range = 196–424; median = 350) and S. simulans (range = 168–415; median = 313). Furthermore, core genome SNPs computed against the ANI values revealed a strong negative correlation (R = −0.97 to −1, p < 0.001) between genome pairs across the different CoNS species (Supplementary Fig. 3).

Genetic determinants of antimicrobial resistance (AMR) and virulence in CoNS

We carried out an in silico screening of all genomes to identify acquired genes that contribute to AMR and virulence. We detected a total of 43 distinct genes that confer resistance to 13 classes of antimicrobial compounds (Fig. 3 and Supplementary Data 3). Among all species, S. epidermidis harbored AMR genes associated with 12 of the 13 antimicrobial classes (Fig. 3a). Other species that carried high number of AMR genes per antimicrobial class were S. hominis, S. simulans, and S. warneri, with at least one genome of each species carrying AMR genes associated with nine, eight, and seven antimicrobial classes, respectively. We did not detect any AMR genes in S. agnetis, S. chromogenes, and S. coagulans. The CoNS community has a diverse assemblage of specific AMR genes and combinations of AMR genes (Fig. 3b). The most frequently detected AMR genes were blaI (n = present in 50 genomes from seven species), blaZ (n = 42 genomes from six species), tetK (n = 29 genomes from nine species), and fosB (n = 24 genomes of S. epidermidis). The highest number of AMR genes per genome were detected in two S. epidermidis isolates from dogs (n = 14 and n = 13 AMR genes) and in S. warneri (n = 13 AMR genes) isolated from a horse. Overall, genes associated with beta-lactam resistance were most prevalent in the CoNS dataset (Fig. 3c).

Fig. 3: Phylogenetic distribution of AMR, heavy metal resistance, and virulence genes in CoNS.
figure 3

a Proportion of genomes within each species that carry at least one gene that confers resistance to each of the 13 antimicrobial classes. Only the 12 most common CoNS species are shown. QAC—quaternary ammonium compounds. b Proportion of the most frequent AMR genes and gene combinations per antimicrobial class. Each bar stack corresponds to the antimicrobial class listed above panel (a). c Number of CoNS genomes and species harboring at least one resistance gene from each antimicrobial class. d Number of CoNS genomes and species harboring heavy metal resistance genes. e Number of CoNS genomes and species harboring virulence genes.

Full size image

We also sought to determine the presence of the mecA gene, which encodes an extra penicillin-binding protein that has low affinity to virtually all beta-lactam antibiotics22. The mecA gene is carried by the mobile chromosomal cassette SCCmec23. Results revealed the presence of the mecA gene in 18 genomes (Supplementary Fig. 4 and Supplementary Data 4). Among the 14 structurally distinct SCCmec types (I–XIV) that have been described23, we found that the mecA gene was carried in SCCmec types I (subtype Ia), III (subtype IIIa), and IV (subtypes IVa, IVc, Ivd, Ivg, Ivh) in genomes of S. epidermidis (n = 15), S. hominis (n = 1), S. lugdunensis (n = 1), and S. simulans (n = 1), and S. warneri (n = 1). We did not detect the mecC gene, which is a divergent form of mecA and also mediates beta-lactam resistance24. We found a slight discrepancy between the results of the in silico detection of the mecA gene and in vitro phenotypic testing for methicillin resistance. There were 12 isolates (S. capitis = 1, S. coagulans = 1, S. felis = 1, S. simulans = 5, S. pseudoxylosus = 2, S. xylosus = 2) whose genomes did not contain mecA but were phenotypically tested as methicillin resistant. On the other hand, eight isolates contained mecA, but were phenotypically tested as methicillin susceptible. These conflicting results may be caused by sequencing and laboratory errors, but may also suggest the absence of transcriptional regulatory genes for mecA for these isolates or the existence of alternative mechanisms of methicillin resistance that are yet to be discovered.

We also screened the CoNS genomes for the presence of genes associated with heavy metal resistance. We detected eight genes associated with resistance to arsenic (arsB, arsC, arsR), cadmium (cadD, cadC), copper (mco), and mercury (merA, merT) (Fig. 3d and Supplementary Data 3). Except for S. agnetis, at least one member of each CoNS species harbored a heavy metal resistance gene. Most of these genes were carried by isolates of S. equorum, S. pseudoxylosus, and S. xylosus, which are closely related in the phylogeny. A total of 83 genomes representing 43.23% of the dataset carried at least one gene conferring resistance to arsenic. These ars-carrying genomes represented 12 CoNS species. A total of 49, six, and one genome(s) carried at least one gene associated with resistance to cadmium, copper, and mercury, respectively.

Three virulence genes were detected in the CoNS genomes. The genes esxA, icaC, and hld were present in 14, 30, and 25 genomes, respectively (Fig. 3e). The icaC gene is part of the ica operon which plays an essential role in mediating the formation of biofilm and has been implicated in therapeutic failure of staphylococcal device-related infections25. The esxA gene plays a vital role in the establishment of infection in hosts by facilitating interactions with host receptor proteins26. Both esxA and icaC were detected in genomes from seven species (S. agnetis, S. capitis, S. coagulans, S. delphini, S. epidermidis, S. lugdunensis, S. xylosus). The hld gene is implicated in the lysis of erythrocyte cells27 and were exclusively detected in all members of S. epidermidis in our dataset.

Lineage shapes the accessory genomes of CoNS

We carried out a network analysis based on the Jaccard similarity of accessory gene content to determine if similarity in accessory gene content between any two genomes is influenced by the genetic background (i.e., higher similarity of accessory genomes within each Staphylococcus species regardless of animal source) or by animal host (i.e., higher similarity of accessory genomes within each animal host regardless of Staphylococcus species) (Supplementary Fig. 5). Our results confirmed the former, showing distinct clustering of genomes by species (top panel) regardless of the identity of the host species (bottom panel). There was no observable clustering of the bacterial accessory genomes when mapped by their host of isolation. For example, the largest cluster representing S. felis consisted of isolates from feline, canine, and lapine hosts. This result shows that sharing of accessory genes between CoNS lineages can traverse species boundaries of their animal hosts. We also observed two clusters that contained a mix of two species (S. pseudoxylosus and S. xylosus; S. agnetis and S. chromogenes). These two species pairs are sister clades in the core genome phylogenetic tree (Fig. 2a), which may explain the close similarities of their accessory genomes.

Phage and plasmid diversity in CoNS

Mobile genetic elements such as phages and plasmids are known to influence the structure and evolution of bacterial genomes28. We therefore investigated the diversity of phages and plasmid replicons in CoNS (Fig. 4a and Supplementary Data 5). Phage DNA was detected in 83.26% (n = 159) of the CoNS isolates. The number of phages per genome ranged from 0 to 12 (median = 2). S. agnetis isolate 4727_3A_S125 from a bird contained the highest number of phages (n = 12), while the genomes of S. simulans isolate 7856_8A from a cow and S. agnetis isolate 6361e_S103 from a bird possessed 10 and nine phages, respectively (Fig. 4a left panel, Supplementary Data 5). Staphylococcal phages are highly diverse and can reach up to 152 Kb in size, occupying a substantial portion of the bacterial host chromosome29. In our study, the total combined genome size of phages was largest in S. felis isolate1657A_S114 from a cat with 545 Kbp representing 20.83 % of its whole genome. This was followed by S. delphini isolate 66A_S182 from a dog (446 Kbp) and S. pseudoxylosus isolate 3581D_S_S66 from a goat (423 Kbp), representing 16.76% and 13.54% of their whole genomes, respectively. However, we do note that there was substantial variation in the total length of phage DNA and number of phages per genome within each species.

Fig. 4: Phage and plasmid replicon diversity in CoNS.
figure 4

a Density plot showing the distribution of concatenated phage sequences sizes per genome per species of CoNS. The boxplot on the right shows the number of phages per genome per species at 25th, 50th, and 75th percentile. Black dots show the number of accessory genes outside the interquartile range. b Violin jitter plot showing number of plasmid replicons detected in each genome per species. Black dots represent the genomes that comprise each species. For both panels, only the 12 most common species are shown.

Full size image

We further screened the concatenated phage sequences for the presence of antimicrobial and virulence genes (Supplementary Data 5). The fusB gene was detected in S. epidermidis (isolates from rodent = 1 and cat = 1), whereas mphC, msrA, and qacA genes were detected in an S. epidermidis isolate from a cat. Furthermore, blaZ and erm44 were detected in S. pseudoxylosus and S. warneri, respectively. Copper resistance gene cadD and virulence resistance gene hld were detected in phage sequences from S. epidermidis isolates 9306A_S137 and 2712_A_S183 from cats.

Analysis of short reads does not provide reliable information on plasmid genomes and associated gene content. Nonetheless, we used the presence of the gene that codes for the plasmid replicon initiator protein (rep) as an indication of a plasmid in the bacterial genome30 (Supplementary Data 6). We detected a range of 1–9 rep genes per genome, with S. epidermidis isolate 4088_3C from a cow carrying the highest number. Members of S. epidermidis, S. succunis, and S. warneri had at least one plasmid replicon detected in their genomes (Fig. 4b). In contrast, no plasmid replicons were detected in members of S. chromogenes and S. coagulans, while fewer representatives from species such as S. delphini (n = 1/5), S. felis (n = 5/39), and S. pseudoxylosus (n = 4/7) contain plasmid replicons in their genomes. Similar to phages, we observed considerable variation in the number and types of rep genes among members of each CoNS species.

Some lineages have acquired more recombined DNA than others

We next sought to determine the impact of homologous recombination in the core and shared accessory genes of CoNS. Using fastGEAR31, we identified a total of 1882 recombination events that occurred within and between species (Supplementary Data 7). The lengths of the recombined DNA segments exhibited an exponential distribution and consists of frequent short recombination events (<500 bp) and rarer large recombination events (Fig. 5a and Supplementary Data 8). The median length of recombined fragments is 242 bp. The largest recombination event was 10,061 bp long and was detected in S. epidermidis. It contained the ebh gene encoding the extracellular matrix-binding protein. Three large recombination events were also detected in S. simulans isolates: a 4591 bp DNA fragment that encompassed the pls surface protein, a 3944 bp that contained that S. aureus surface protein A, and a 3872 bp DNA that contained the YSIRK-signal domain-containing protein. Summing up the total length of recombined DNA per isolate, we found that S. simulans isolates were the most frequent recipients of recombined DNA (mean = 12,434 bp; range = 4902–23,217 bp) (Fig. 5b). Other frequently recombining species were S. hominis (mean = 7022 bp and range = 4378–12,243 bp), and S. xylosus (mean = 5041 bp and range = 1335–9233 bp). For S. simulans, S. hominis, and S. xylosus, these recombination events were detected in 446, 148, 198 genes and which represent 11.53%, 5.04%, and 4.73% of each species’ pan-genome, respectively. Within the same species, the impact of recombination varied among genomes. Certain isolates of S. epidermidis and S. felis had experienced more frequent recombination compared to other isolates within the species. These included isolate 6101a_S160 (S. epidermidis from a cat) with 13,191 bp of total recombined DNA whereas isolates 4602_AN3_S151 and 1405b_S102 (both S. felis from cats) have a total of 13,191 bp and 10,845 bp recombined DNA, respectively. In contrast, the species whose genomes contained the shortest total length of recombined DNA were S. capitis (mean = 445 bp and range = 751–190 bp), S. coagulans (mean = 59 bp and range = 29–120 bp), and S. lugdunensis (mean = 115 bp, range = 16–284 bp). Overall, these results that certain lineages have had a history of more frequent recombination than other lineages.

Fig. 5: Frequently recombining CoNS lineages.
figure 5

a Histogram showing the distribution of fragment sizes of recombined DNA in CoNS. The longest recombination events are highlighted with arrows. b Total length of recombined DNA per genome. The tree is identical to that in Fig. 2a.

Full size image

Frequent recombination in some donor-recipient partners

We next sought to identify donor-recipient pairs of recombination events to determine if there were pairs that were more frequently exchanging DNA than with others. We identified 1659 and 223 recombination events involving genomes of the same species and between species, respectively (Fig. 6a and Supplementary Data 7). We detected the presence of intra-species recombination in 13 of the 15 CoNS species. The lack of intra-species recombination detected in S. coagulans and S. succinus is likely due to the low number of genomes in these species. S. simulans had the highest number of recombination events (n = 719), although this may be due to more frequent samples from this species. Nonetheless, these recombination events were detected in 3867 genes, indicating that 18.59% of the species pan-genome has been affected by recombination originating from other members of its own species. Other species that have high number of intra-species recombination events included S. felis (n = 202 recombination events in 163 genes), S. xylosus (n = 195 recombination events in 164 genes), S. epidermidis (n = 179 recombination events in 139 genes), S. hominis (n = 160 recombination events in 138 genes), and S. chromogenes (n = 69 recombination events in 62 genes). The most frequently recombining genes between isolates of the same species and that have known functions were the surface protein pls (21 recombination events in S. simulans) and iron-regulated surface determinant protein B isdA (9 events in S. simulans). Recombination was also detected in hypothetical proteins, with 12 and 10 recombination events in S. epidermidis and S. felis, respectively.

Fig. 6: Intra- and inter-species recombination in 191 CoNS genomes.
figure 6

a Chord diagram showing the recombination events within and between members of each CoNS species. Each colored block in the outer ring represents a CoNS species, while the length of block represents the number of genomes for that species in which recombination was detected. b Chord diagram showing the recombination events in CoNS genomes from different animal hosts regardless of CoNS species. For visual clarity, only animal hosts with the highest number of isolates sampled are include in which recombination was detected. Recombination shown here includes recombination events detected in all CoNS species from the three animal hosts. For both panels, each line in the center connects a pair of donor and recipient genomes. For each line, the recipient genome is indicated by the end of the line ending nearer the outside of the plot and the potential donor cluster is indicated by the end of the line nearer the center of the plot. For inter-species recombination, the color of the connecting line is based on the color of the recipient genome.

Full size image

We also identified donor-recipient partners in inter-species recombination events, although these were not as frequent as intra-species recombination (Fig. 6a). All 15 CoNS species were involved in inter-species recombination. S. simulans was the most frequent DNA donor and DNA recipient, although this is likely due to this species having the highest number of genomes in our dataset. Regardless, we identified a total of 88 pairs of two CoNS species which are linked by at least one recombination event between them. The most frequent recombination partners were the sister taxa S. pseudoxylosus and S. xylosus, which were both frequent donors and recipients of each other. We identified 42 recombination events involving S. pseudoxylosus donors and S. xylosus recipients, whereas 11 recombination events involving S. xylosus donors and S. pseudoxylosus recipients. We also detected eight recombination events between S. epidermidis (donor) and S. hominis (recipient). Between species, the most frequently shared genes involved four recombination events involving the catalase gene [S. xylosus (donor) and S. delphini (recipient)] and site-specific integrase gene [S. warneri (donor) and S. epidermidis (recipient)] coding sequences, whereas there were three recombination events for the sharing of arsenite efflux transport membrane protein arsB between S. epidermidis (donor) and S. hominis (recipient). Overall, these results revealed that certain donor and recipient lineages exchange DNA more often with each other than they do with other lineages.

To further investigate the frequency of recombination of CoNS from within the same animal host species or between host species, we mapped all recombination events in any CoNS genome from the three most common animal groups sampled (Fig. 6b). These included 69 genomes from felines, 42 genomes from canines, and 27 genomes from bovines, representing a total of 72.25% (n = 138 genomes) of our CoNS dataset. We identified a total of 587 recombination events that occurred in 431 genes. The number of recombination events between isolates regardless of CoNS species that were from the same animal host were 190, 91, and 59 for feline, bovine, and canine hosts, respectively. The number of recombination events between isolates regardless of CoNS species that were from different animal hosts were: 63 between canines (donor) and felines (recipient), 59 between felines (donor) and canines (recipient), 41 between canines (donor) and bovine (recipient), 33 between bovine (donor) and canines (recipient), 29 between bovine (donor) and felines (recipient), and 22 between felines (donor) and bovine (recipient). These results show that recombination between CoNS genomes is frequent regardless of the identity of the animal host.

Discussion

We present a comparative population analysis of 191 CoNS genomes sampled from diseased animals diagnosed in a veterinary diagnostic laboratory located in the northeast region of the United States. Our findings suggest that genetic exchange within and between species, though frequent, is not random nor indiscriminate. First, while often misclassified or overlooked in sampling efforts1, CoNS are important reservoirs of diverse phages, plasmids and mobilizable genes encoding AMR, heavy metal resistance, and virulence. The most frequently exchanged genes through homologous recombination were those associated with surface proteins, which may facilitate common strategies of different CoNS species to adapt to the same animal hosts. Second, we identified frequent exchange of DNA between certain donor-recipient partners, suggesting a structure to the genetic flux in which specific lineages act as hubs of gene sharing. Third, we detected frequent recombination between CoNS strains and species regardless of their animal host species, indicating that ecological barriers to recombination32 can be surmounted in co-circulating lineages. We interpret these findings as implying that frequent but structured patterns of recombination exist within and between CoNS species, which are driven by their overlapping ecological niches and geographical proximity.

The consequences of horizontal gene transfer followed by homologous recombination are vast. It is known to influence a myriad of evolutionary and population processes, including levels of standing diversity, niche expansion, spread of resistance and virulence determinants, and rapid adaptive changes in response to new or fluctuating environmental conditions33,34. Horizontal gene transfer can also result to the emergence of novel genetic variants or hybrids with unique phenotypes such as multidrug resistance, hyper-virulence, and altered transmissibility35,36. However, current models of gene transfer via recombination incorporate the null expectation that recombination is a homogeneous process across the species37,38. Such models lead to classifying bacteria as either clonal and panmictic species. In clonal species (e.g., Mycobacterium tuberculosis39), genetic variation between strains is generated mainly by mutation and emerging under neutral evolution through the random birth and extinction of lineages. In a panmictic species (e.g., Helicobacter pylori40), rates of recombination approach those found in sexual eukaryotes. High rates of recombination can theoretically prevent the separation of distinct daughter lineages, creating indistinct or fuzzy boundaries between them41,42. Although many studies have generated crucial insights into the nature and frequencies of recombination between bacterial species34,37,38, it is often assumed that all strains recombine at a uniform frequency and randomly across the entire species.

Our findings revealed that frequencies of recombination can vary dramatically even between strains of the same species, with a unique pattern of exchange for different strains and lineages. This pattern has been also observed in major bacterial pathogenic species43,44,45. Some strains donate or receive DNA more often than others46,47, while some strains tend to preferentially recombine with specific partners44,48. In the current study, we show that such variable patterns of recombination exist not only within species but also between species. The role of the animal host is particularly important in shaping patterns and frequencies of inter-species recombination in CoNS, as has also been demonstrated in the genus Campylobacter32. Differences in recombination frequencies also suggest that lineages respond to selective pressures in different ways. Such variation also implies that recombination itself can evolve in response to natural selection49,50 and can occur quickly on an evolutionary timescale51,52.

Bacterial strains that contain large tracts of recombined DNA53 are particularly intriguing because unpredictable phenotypes with considerable public health threats may arise. These strains reflect a mode of evolution that proceeds in jumps (or saltation) brought about by sudden and large genomic modifications and that have profound phenotypic consequences54,55,56. Frequently recombining strains may be likened to bacterial “hopeful monsters”, whereby most cases of rapid and dramatic changes to the genome are likely to be deleterious, but may occasionally be highly adaptive54,55. Such strains that contain large segments of recombined DNA have been reported in Streptococcus pneumoniae44, Klebsiella pneumoniae57, and Vibrio anguillarum58. The five genomes in our study that contain large tracts of recombined DNA as well as a few highly recombining genomes in S. epidermidis and S. felis seem to emerge as “hopeful monsters.” The origins of these hyper-recombinants and the factors that facilitate their emergence remain unclear, but close phylogenetic relationships, similar ecological niches, and geographical proximity can amplify opportunities for frequent and large-scale recombination events to occur. Investigations on the long-term evolutionary trajectory and phenotypic characteristics of hyper-recombining CoNS will have profound implications to effectively managing CoNS infections.

Our results should be interpreted within the confines of the study limitations. First, we acknowledge limitations associated with sampling bias. Each CoNS species was disproportionately represented by variable number of genomes. Our dataset is also heavily biased to favor some animal groups more than others, with more isolates recovered from feline, canine, and bovine hosts. These two sources of bias paint an uneven characterization of genome content per species, including the frequencies and patterns of recombination within and among CoNS species and genomes. Future research should also spotlight CoNS-specific genetic elements, especially those that may contribute to their ecological adaptation, drug resistance, and pathogenesis. A broader and systematic sampling strategy will be instrumental in uncovering novel CoNS species and lineages from an even more diverse range of animal species and their mechanisms of DNA mobility. This is especially true at the human-animal interface where pathogen transmission is most likely to occur. It will also help in establishing the host range of each CoNS species and in identifying those species that are restricted to a single animal species.

In summary, our results underscore the impact of frequent but skewed patterns of homologous recombination, mobile genetic elements, and accessory gene sharing in shaping the diversity, genome dynamics, and host adaptation of CoNS. These data will be important in developing effective surveillance, diagnostic, and treatment options for CoNS infections.

Methods

Bacterial sampling

The CoNS culture collection consisted of 191 isolates that were retrospectively sampled from September 2017 through March 2020. A total of 18, 67, 87, and 19 isolates were collected in 2017, 2018, 2019, and 2020, respectively. Isolates were obtained as culture swabs from routine clinical specimens submitted to the NHVDL, New Hampshire, USA. Isolates were mostly recovered from body sites such as ear (n = 35), animal milk (n = 26), skin (n = 24), wounds (n = 17), and urine (n = 16). The clinical specimens were received from multiple veterinary practices from the New England region of the country, consisting of the states of Connecticut, New Hampshire, Maine, Massachusetts, and Vermont. We also included two isolates where the animals were from Alabama (located in the southern part of the country) but were diagnosed in New England. These were included in our analysis because it was unknown whether they were infected while they were in New England. All isolates were from animals with confirmed clinical infections. No live vertebrates were used in this study; hence, the NHVDL was exempt from the International Animal Care and Use Committee (IACUC) approval process. Pure isolates were cultured in commercially prepared tryptic soy agar with 10% sheep red blood cells and brain heart infusion broth. Preliminary identification of CoNS species was carried out using a Brucker Biotyper MALDI-TOF MS in accordance to the manufacturer’s instructions. The machine determines proteomic fingerprint of the organism under query and matches it against two libraries of reference spectra RUO library 6903(V6) and 7311(V7) available in the Bruker MBT Compass (Bruker Daltonics, Bremen, Germany). All isolates were stored in DMSO solution in −80 °C. Associated metadata information for each isolate, including date of sample collection, location, and isolation source, are included in Supplementary Data 1.

Methicillin susceptibility screening

In vitro screening for cefoxitin and oxacillin resistance was carried out using the Kirby Bauer disc diffusion technique. Because there are no breakpoint guidelines in the most current Clinical and Laboratory Standards Institute (CLSI) for CoNS species, we used the CLSI breakpoint guidelines for cefoxitin which is used as the official predictor of methicillin resistance for S. aureus59. For isolates identified as methicillin resistant, we determined the presence of the penicillin-binding protein PBP2 using a commercial latex agglutination test (MASTALEX MRSA Latex Kit, MAST, UK) following manufacturer’s guidelines. We confirmed the presence of the mecA gene by screening the genome sequence of each isolate (described below).

DNA extraction and whole genome sequencing

Isolates were grown overnight in brain heart infusion broth at 37 °C. Genomic DNA from overnight cultures were then extracted using the Zymo Quick DNA Fungal/Bacterial Miniprep Kit (Irvine, California) following manufacturer’s instructions. DNA quantity was measured using a Qubit fluorometer (Invitrogen, Grand Island, NY) according to manufacturer’s instructions. Genome sequencing was carried out using the NextSeq 2000 platform at the Microbial Genome Sequencing Center (now SeqCenter) in Pittsburgh, Pennsylvania in 2022. Sample libraries were prepared using the Illumina DNA Preparation kit and IDT 10 bp UDI indices following the manufacturer’s instructions. Sequencing generated paired-end reads (2 × 151 bp) on multiplexed libraries. Demultiplexing, quality control, and adapter trimming were performed with the Illumina bcl-convert (v3.9.3).

Genome assembly, quality check, annotation, and species designation

We used the shovill v.1.1.0 pipeline to assemble all paired-end reads (https://github.com/tseemann/shovill). Shovill uses the SPAdes assembly algorithm60 but alters several pre- and post-assembly steps to generate similar and high-quality assembly results in less time. We used the --trim option to enable adapter trimming and improve assembly. To assess quality of the assembled genomes, we used QUAST v.5.0.261. We excluded assemblies with >200 contigs and an N50 < 40,000 bp. We used CheckM v.1.1.3 to determine only those genomes with level of completeness of >90% and contamination of <5%62. We calculated the genome completeness (mean = 99.31%; range: 97.20–99.81%) and genome contamination (mean = 0.9%; range = 0 to 4.44%), which were all within the genome quality standards recommended by CheckM (Supplementary Data 1 and Fig. 1). To determine genomic relatedness and delineate species boundaries, we calculated the genome-wide Average Nucleotide Identity (ANI) for every possible pairs of genomes using fastANI v.1.3221. ANI refers to the mean nucleotide identity of orthologous pair of genes that are shared between a pair or collection of microbial genomes21. We used the >95% ANI threshold to confirm species identification. The draft genomes were annotated using Prokka v.1.14.663 and polished with Bakta64.

Pan-genome analysis and phylogenetic tree reconstruction

To characterize the pan-genome, we used PIRATE v.1.0.4, a fast and scalable platform for clustering orthologous gene families in bacteria65. Briefly, gene orthologs are clustered over the default identity threshold values ranging from 50% to 98% (50, 60, 70, 90, 90, 95, 98) sequence identity using CD-HIT66. A gene presence and absence matrix were produced in tab separated file using the PIRATE supplement Rtab.pl script. Gene sequences were aligned using MAFFT67. Sequence alignments of core genes (i.e., gene families present in >95% of genomes) were concatenated to generate the core genome alignment. Single nucleotide polymorphisms (SNPs) were extracted from the core genome alignment using SNP-sites68. The core SNP alignment was used as input for building a maximum likelihood phylogenetic tree using RAxML v.8.2.1269. We used the general time reversible model for nucleotide substitution70 under the GAMMA model of rate heterogeneity. Phylogenetic trees were visualized and annotated using figtree v.1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/) and Interactive Tree of Life IToL71.

Accessory gene network analysis

The gene presence or absence matrix generated using PIRATE, excluding the core genes, were used as input in GraPPLE (https://github.com/JDHarlingLee/GraPPLE). Briefly, a Jaccard similarity coefficient based on the number of shared genes over the total number of genes across a pair of genomes was calculated using the pw_similarity.py script. Metadata were added using the metadata_to_layout.py script. Clustering by accessory gene content similarity was visualized as networks using Graphia72. Accessory genome similarity was clustered using MCL - Markov algorithm. Edges of networks were transformed using the k-nearest neighbor algorithm (k = 5).

In silico detection of antimicrobial resistance genes, virulence genes, SCCmec, plasmid replicons, and phages

Genome assemblies were screened for the presence of acquired antimicrobial resistance genes and heavy metal resistance genes using the National Centre for Biotechnology Information’s (NCBI) AMRFinderPlus v.3.10.23 and its accompanying NCBI-compiled AMR database73. We also screened the genome assemblies for the presence of virulence determinants using ABRicate v.1.0.1 (https://github.com/tseemann/abricate) containing the Virulence Factor Database (VFDB)74. We also used ABRicate to search for the rep gene that codes for the plasmid replicon initiator protein (rep) against the Plasmid Finder database30. We used staphopia-sccmec75 to carry out in silico detection and classification of SCCmec. Staphophia-sccmec uses a primer-based approach where assemblies are aligned against SCCmec typing primers75. Samples with a perfect match are assigned an SCCmec type. We used VirSorter2 (https://github.com/jiarong/VirSorter2)76 to determine phage diversity in CoNS genomes. For each genome, the length of phage DNA regions was summed to give total length in each genome.

Inference of homologous recombination

Using the sequence alignments of individual core genes and shared accessory genes, we inferred recent and ancestral recombination events used fastGEAR31 (Supplementary Data 9). FastGEAR identifies lineages in the alignments and implements a Hidden Markov Model to compare polymorphic sites occurring in individual strain and compare them to other polymorphic sites occurring in members of its own lineage as well as strains from other lineages. The output of fastGEAR is parsed into HERO (Highways Enumerated by Recombination Events) (https://github.com/therealcooperpark/hero), a pipeline implemented in Python to visualize donor-recipient strain pairs in recent recombination events identified by fastGEAR. Visualization of recombination events was carried out using Circos v.0.69-877.

Statistics and reproducibility

To test the significance of the inferred recombinations and identify false-positive results, we used a diversity test implemented in fastGEAR31. This is based on a simple binomial test which computes a Bayes factor (=1) that measures how different the SNP density changed between the DNA fragment in question compared to its background (lineage). We calculated the Pearson correlation coefficient implemented in R package ggpubr v0.4.0 to determine the association between core genome SNPs and ANI of every genome per species. We used a p-value threshold of <0.001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.