Ethics Considerations

The human isolates used in our study were recovered as part of routine clinical diagnostic testing; thus, ethics approval and informed consent were not required. All data were handled anonymously. Sample collection was in accordance with the European Parliament and Council decision for the epidemiologic surveillance and control of communicable disease through the European Antimicrobial Resistance Surveillance Network (https://www.ecdc.europa.eu/en/activities/surveillance/EARS-Net/Pages/index.aspx). Slaughterhouse samples were part of the routine control practices for evaluation of good hygiene practices and programs to assure meat safety (European Parliament and Council regulation no. 853/2004).

Bacterial Isolates

The global S. saprophyticus collection we used included 299 isolates from humans collected in 7 countries during 1997–2017: 286 from UTIs, 12 from invasive disease, and 1 from colonization (Appendix 1 Table 1). We also analyzed the genomes of S. saprophyticus for 38 isolates from 5 other countries: 35 isolates from human UTIs (8), 2 from human hand swabs (8), an isolate from Byzantine Troy (8), and ATCC 15305 (9), a previously investigated human UTI-causing isolate.

The local collection included isolates collected in Lisbon, Portugal, during 2016–2017: 128 human UTI isolates collected in 3 hospitals and 104 slaughterhouse isolates collected from equipment, pork samples, workers’ hands, and a pig’s rectum. In addition, we included 5 isolates from animals and 12 isolates from food used in other studies (8) (Appendix 2).

Whole-Genome Sequencing and Assembly

Phylogenetic Analysis

Figure 1

Figure 1. Maximum-likelihood tree of Staphylococcus saprophyticusisolates recovered from human infections and colonization globally, 1997–2017. The tree was constructed by using 9,134 SNPs without recombination. Among analyzed isolates, 321 were...

Figure 4

Figure 4. Growth rate of Staphylococcus saprophyticusclonal lineages in tryptic soy broth (TSB) and in different concentrations of female sex hormones. All assays were performed in triplicate and each experiment...

We performed whole-genome sequencing (WGS) on MiSeq (Illumina, https://www.illumina.com) and MinIon nanopore (Oxford Nanopore, https://nanoporetech.com) platforms, as described (10) (Appendix 2). We separately analyzed global population and local epidemiology of S. saprophyticus and their phylogeny by using single-nucleotide polymorphisms (SNPs). We identified SNPs by mapping the draft genomes to a reference genome, S. saprophyticus ATCC 15305 (GenBank accession no. AP008934.1) by using the web-based CSI phylogeny version 1.4 (11) with the default parameter, but we disabled the minimum distance between SNPs in the parameter. We used Gubbins version 2.3.4 (12) with default parameters to concatenate SNPs and remove recombination regions. We reconstructed the phylogenies by using RAxML version 8.2.4 (https://github.com/stamatak/standard-RAxML) and generalized time-reversible nucleotide substitution with gamma correction model with 100 bootstrap value. We visualized the maximum-likelihood trees by using Interactive Tree of Life (ihttps://itol.embl.de) (Figures 1–3; Appendix 2 Figure 3). Recombination to mutation (r/m) ratios detected by using Gubbins were calculated as the average r/m of isolates in the entire collection and separately for each lineage by using as reference closed genomes of KS40 for lineage G and KS160 for lineage S, both obtained on the MinIon platform (Oxford Nanopore).

Statistical Analyses

We used Prism 6.0 (GraphPad, https://www.graphpad.com) to compare the means of 2 groups. We used a 2-tailed unpaired Mann-Whitney test or χ² test for comparison and considered p<0.05 statistically significant.

Lineages of S. saprophyticus Causing UTIs

Among study isolates, we first analyzed the diversity of S. saprophyticus causing UTIs by using genomic data of 321 human UTI isolates collected from 8 countries on 4 continents during 1997–2017. From the SNPs initially detected, 42% arose from recombination events in the population, corresponding to a mean genome-wide r/m of 1.5:1, meaning that high accumulation of SNPs was due to recombination rather than mutation in UTI strains. The maximum-likelihood tree constructed from the 9,134 SNPs without recombination defined 2 lineages, which we called G and S (Figure 1). Most (74%, 236/321) UTI isolates were from lineage G and differed by 0–4,318 SNPs with an average nucleotide identity (ANI) of 98.5%–99.999%, whereas S isolates (26%, n = 85/321) were slightly less distantly related, differing by 0–3,540 SNPs with an ANI of 99.3%–99.991% (Appendix 2 Figure 1).

Figure 2

Figure 2. Phylogenomic analysis and distribution of Staphylococcus saprophyticusisolates collected from human infections and a slaughterhouse, Portugal, 2016–2017. A) Maximum-likelihood tree of 232 isolates from human infections or slaughterhouse contamination....

S. saprophyticus lineages we identified among UTI isolates worldwide had distinctive features indicative of 2 evolutionary histories. Although strains of both lineages had similar genome size as determined by their closed genomes (lineage G was 2.5 Mb and S 2.6 Mb), isolates from lineage S (n = 6) grew significantly faster in tryptic soy broth incubated at 37°C (μ_average = 0.34 h^–1) than G strains (n = 8) (μ_average = 0.22 h^–1; p = 0.0007) (Figure 4; Appendix 2). Moreover, we separately analyzed r/m of both lineages by using the respective closed genome of strains from each lineage as a reference and found that the mean estimated r/m was 9 times higher in S isolates (r/m = 4.4:1) than G isolates (r/m = 0.5:1). We did not detect a temporal signal when performing regression analysis (r) of tip-to-root distance versus isolation date, either in the entire collection (r = −0.2423) or for separate lineages (for G, r = −0.1314; for S, r = 0.1889) (Appendix). Hence, we could not determine when the 2 lineages diverged. The lack of temporal signal probably results from the limited number of isolates in each time point.

Figure 3

Figure 3. Maximum-likelihood tree depicting the genetic relatedness of Staphylococcus saprophyticusisolates belonging to clonal lineage G recovered from human infections or slaughterhouse contamination, Portugal, 2016–2017. Each node represents a strain....

Figure 5

Figure 5. Genetic determinants that contribute to the distinction of clonal lineages and lifestyle of Staphylococcus saprophyticus. The graph displays determinants that contribute (A) and mediate (B) adaptation of ...

To further compare the 2 lineages, we constructed the pangenome of the 338 human S. saprophyticus genomes and identified 10,222 genes with 85% blastp clustering by using Roary. Among these, 8,351 genes, present in <99% of the genomes, constituted the accessory genome. A gene accumulation plot of all the genes against the genomes sequenced showed that S. saprophyticus has an open pangenome (Appendix 2 Figure 2, panels A, B). The pan-GWAS analysis of the accessory genome indicated that G and S isolates have different genomic content (Bonferroni p<0.05, OR >1). A total of 128 genes were specific/enriched in the G lineage, including those encoding a type I restriction subunit (group_383), a defense mechanism against genetic transfer (15); metabolism of melibiose (ebgA, melB) (16) and inositol (iolE) (17), compounds that are excreted in urine; toxin–antitoxin systems (group_4685, group_5665), involved in stress response (18); and antimicrobial resistance (qacA) (19) (Table 1; Figure 5, panel A; Appendix 2 Table 1).

For S lineage, 237 genes were specific/enriched, some of which are involved in metabolism of ascorbate (ulaA) and thiamine (tenI) (20); induction of phosphate starvation (psiE), which was previously linked to switch on virulence in other uropathogens (21); platelet binding (splE, sdrE) associated with virulence (22); steroid metabolism (group_7190); and resistance to trimethoprim (dfrG) and biocides (qacC) (Table 2; Figure 5, panel A; Appendix 2 Table 1). Additional genes associated with S lineage that could explain its increased growth rate include those linked to lactose and cellobiose metabolism (group_5572 gene) (23) and cell wall hydrolysis (lytN) (24), whereas the prevalence of transposases (group_3547 and group_1828) (25) and phage genes (recT and yueB) (27) could justify its increased recombination rate.

Local and Global Spread of S. saprophyticus Clones Causing UTIs

We observed no time-based clustering of S. saprophyticus UTI isolates in the phylogeny, but we noted some geographic clustering. In particular, 89% (n = 78) of S isolates were found in Portugal and Spain (Figure 1). In addition, we identified clusters containing isolates from a single country. For instance, cluster G3 (1–8 SNPs) contained only isolates from Portugal and cluster G1 (0–63 SNPs) only isolates from Denmark (Figure 1).

We noted a high degree of isolate admixture in the maximum-likelihood tree of our global collection, suggesting that S. saprophyticus isolates of both lineages are disseminated geographically. G strains were distributed most widely, in 11 countries on 4 continents, whereas we found S isolates in only 6 countries. Despite the genetic diversity described, we still found isolates from different countries that differed by only a few SNPs. One pair in cluster G4 that had only 10 SNPs difference was a blood isolate from Barcelona, Spain (KS266), and a UTI isolate from Lisbon (KS135), collected in 2016 (Figure 1).

Although a relatedness cutoff is not defined yet for S. saprophyticus, the low number of SNPs observed between strains from the same and different countries is below the relatedness cutoff of 10–40 SNPs for most bacterial species (28). The apparent relatedness we noted implies that UTI isolates from different patients in the same country and in different countries are highly related and could belong to a cross-border chain of transmission. This finding challenges the assumption that S. saprophyticus causing UTIs were mainly endogenous (29).

Genetic Relatedness of S. saprophyticus from Slaughterhouses and UTIs

Pork is the most frequently consumed red meat in Europe (30) and is often contaminated with S. saprophyticus (1). We found that 35% of slaughterhouse samples (from meat, equipment, workers’ hands, and a live pig) were contaminated with S. saprophyticus.

To understand whether S. saprophyticus causing UTIs could be related to S. saprophyticus in pork, we compared 104 isolates collected from a slaughterhouse against 128 isolates collected from human UTIs in Lisbon during 2016–2017. Among the 104 isolates from the slaughterhouse, 39 (37.5%) were collected from meat, 32 (30.8%) from equipment, 32 (30.8%) from workers’ hands, and 1 (≈1%) from a live pig. SNP-based phylogenetic analysis with a tree-rooted at the midpoint showed that most (91%; 95/104) slaughterhouse isolates belonged to lineage G (Figure 2, panels A, B) and that a strain from slaughterhouse equipment was at the base of this lineage (bootstrap 100). In addition, the phylogenetic reconstruction including isolates from this study and other isolates from production and companion animals (including 2 pigs, 2 bovine, and 1 canine) and food (8) showed that most (3/5) animal isolates clustered together at a basal clade of lineage G (bootstrap 100) (Appendix 2 Figure 2). Some clusters in the phylogenetic tree (e.g., G9) had slaughterhouse isolates at the base and UTI isolates at the tip. However, we also observed the opposite (e.g., G11), tree clusters with slaughterhouse isolates at the tip and UTI isolates at the base (Figure 2, panel A). Moreover, 41% of G strains included the antimicrobial resistance gene tetK (p<0.0001) (Figure 2, panel A), which is associated with resistance to tetracycline, an antimicrobial drug extensively used in animal production (31).

Phylogenetic reconstruction of all isolates from Lisbon based on SNPs provided additional examples of admixture of isolates recovered from different sampling sites in the slaughterhouse and from the slaughterhouse and humans. Isolates from meat were frequently intermixed with isolates from equipment and colonized workers as observed in cluster G6, wherein stains differed by only 1–65 SNPs (Figure 2, panel A; Figure 3). Likewise, cluster G9 included isolates from the slaughterhouse that were intermixed with human UTI isolates. Isolates collected from slaughterhouse equipment differed by only 19 SNPs from human UTI isolates. Likewise, human UTI isolates differed from meat isolates by 25 SNPs and from isolates of colonized workers by 26 SNPs. In addition, slaughterhouse and human UTI isolates had the same antimicrobial resistance profile, exhibiting resistance to fosfomycin, fusidic acid, and tetracycline (Appendix 1 Table 1). We could not ascertain any epidemiologic link between the workers in the slaughterhouse and UTI patients, due to data protection limitations, but contact with meat production previously has been identified as a risk factor for UTI (6).

The only live-pig isolate was intermixed in cluster G8 with isolates from meat, slaughterhouse workers, and UTI patients (Figure 3), having 307–658 SNPs difference. The admixture of strains in the tree suggests the existence of frequent cross-transmission within the slaughterhouse and between the slaughterhouse and humans.

Insufficient disinfection procedures probably contributed to the high transmission rate of S. saprophyticus within the slaughterhouse, as demonstrated by the highly related strains (<11 SNPs) on dirty and clean equipment surfaces and similar strains (<16 SNPs) isolated 12 months apart. Carriage of the antimicrobial resistance gene, qacA, by all G strains could justify the observed unsuccessful cleaning procedures (Figure 2, panel A).

The accumulation of substitutions and genetic distance evidenced by the phylogenetic analysis suggest that isolates from slaughterhouses and food probably are the primary sources of S. saprophyticus G strains (Figure 3). Transmission probably occurs more frequently from the slaughterhouse and food to humans; however, we cannot ascertain directionality due to the lack of temporal signal.

Evidence Supporting the Human Origin of Lineage S

In contrast to isolates belonging to lineage G, which were mostly from UTIs and the slaughterhouse, S isolates were almost exclusively of human origin (97%; n = 66/68), either from UTIs (n = 59) or human colonization (n = 7) (Figure 2, panels A, B). When we reconstructed the phylogeny of all isolates in this and other studies (8) (Appendix 2 Figure 3), a human isolate was at the base of the S lineage. The only 2 S isolates seen in animals were from nonhuman primates. Furthermore, a resistance determinant for trimethoprim (dfrG), which routinely is used to treat human UTIs, was associated with this lineage (18%; p<0.05) (Figure 2, panel A; Table 2).

To determine whether S isolates could have originated in humans, we grew isolates from both lineages in the absence and presence of human physiologic concentrations (15–350 pg/mL) of estradiol (32), a female hormone commonly found in urine and the vagina, and in different pH values mimicking the stomach (pH 2.5), vagina (pH <4.5) (33), skin (pH 5.5), and urine (pH 4.5–8.0) (34). The growth rate of S isolates did not change significantly at the highest physiologic concentration of estradiol specific to humans (0.34 h^–1 vs. 0.33 h^–1), but the growth rate for G strains decreased by 59% at this concentration (0.22 h^–1 vs. 0.13 h^–1; p = 0.0007) (Figure 4, panels A–C; Appendix 2). All isolates grew at all pH levels assayed, except for pH 2.5. At pH 4.5 and 5.5, isolates of both lineages had similar growth rates, but S isolates had a higher growth rate than G isolates at pH 8, although this difference was not statistically significant (p = 0.133). These results suggest that lineage S isolates are more adapted than lineage G strains to high estradiol concentrations found in women, but not found in other female animal hosts, such as pigs, bovine, or canines (35). GWAS also identified a gene involved in steroid metabolism, 3-β hydroxysteroid dehydrogenase (HSD), associated with lineage S (Table 2; Figure 5, panels A, B). Steroids such as estradiol are primary signaling molecules for host–microbe interactions (26) and involved in the interconversion of active and inactive steroid hormones (26). Presence of HSD could be an adaptive evolution to colonization of the bladder, a hormone-rich environment. Evidence supports a human (primate) origin for lineage S, but studies sampling a wider range of ecologic sites and geographic regions are needed.

UTIs and S. saprophyticus Dissemination among Humans in the Community

To explore dissemination of S. saprophyticus causing UTIs in the community through human-to-human contact, we analyzed genomic data of the 128 UTI isolates from outpatients of 3 hospitals in the Lisbon area. Transmission of S. saprophyticus from lineage G and S between persons in the community was apparent, as demonstrated by the high relatedness of strains from UTI patients at different hospitals. In the G cluster, G10 isolates had 9–24 SNPs difference, and in the S cluster, S2 isolates differed by 6–64 SNPs (Figure 2, panel A). However, due to data protection regulations, we could not ascertain whether the patients were epidemiologically linked. Results suggest that patients might have acquired these strains from the same reservoir or that direct and indirect cross-transmission could have occurred in the community.

Disease Signatures among S. saprophyticus Populations

Several virulence factors, including urease, have been described in S. saprophyticus (36), but the basis of pathogenicity in this species is mainly unknown. We used a pan-GWAS approach to compare the genetic content of 128 isolates from human infections to 104 isolates recovered from a slaughterhouse, all collected in Lisbon during 2016–2017. We identified 6 genes that appear to be associated with an increased pathogenic potential in S. saprophyticus (Figure 5, panel A; Appendix 2 Tables 2–5). These genes included those encoding an SplE-like protein and a gene cluster encoding a complete accessory secretory system associated with a serine-rich adhesion, similar to SraP. Previous studies have described highly homologous secretory system (>97% nucleotide identity) associated to serine-rich proteins that bind platelets, including SraP in S. aureus (37) and UafB in S. saprophyticus (38).

We also found other genes associated with infection that encoded phage proteins. An analysis on 128 UTI isolates using PHASTER (39) identified 5 phages in 48 (38%) strains. The phages were 38–125 Kb and had <50% identity with any known phage. The genomic vicinity of the identified phages varied in the chromosome, suggesting that lysogenic conversion was not the mechanism involved in pathogenicity of these strains; instead, hypothetical phage genes could be crucial for pathogenicity. In addition, genes encoding resistance to the antimicrobial drugs trimethoprim (dfrG), lincosamide (lnuA), streptogramin B (erm44v), or to macrolides (mphC-msrA) also were associated with strains from human infections (Figure 2; Figure 5, panel B; Appendix 2 Tables 2–5).

Another factor highly associated with human infections was the occurrence of recombination, which was 6 times higher in isolates from UTIs (r/m 1.7:1) than from isolates associated with colonization or contamination (r/m 0.3:1). This finding suggests recombination might be a strategy of S. saprophyticus to evade the host immune system, as described elsewhere (40).