Molecular Evolution and Adaptation of Livestock-Associated Methicillin-Resistant Staphylococcus aureus (LA-MRSA) Sequence Type 9

ABSTRACT Livestock-associated methicillin-resistant Staphylococcus aureus (LA-MRSA) sequence type 9 (ST9) has emerged and disseminated in Asia. It is associated with colonization or infection in both humans and animal hosts; however, the genetic factors underpinning its adaptation to animal and human population remain to be determined. Here, we conducted a genomic analysis of 191 ST9 S. aureus genomes collected from 12 different countries, including 174 genomes retrieved from public databases and 17 sequenced in this study. In silico spa typing, staphylococcal cassette chromosome mec (SCCmec) typing, and antimicrobial resistance and virulence gene mining were conducted, and the temporal phylogenetic signal was assessed by Bayesian inference. Our results point toward a human methicillin-susceptible S. aureus (MSSA) origin of ST9 that evolved approximately 2 centuries ago. Three major genetic events occurred during ST9 host shift from human to animals: the loss of the immune evasion cluster genes (scn, chp, and sak), which were reported to contribute to virulence in human infections, the acquisition of the SaPIbov4-like element-encoding vwb gene, which is an animal-specific virulence factor responsible for the clotting of animal plasma, and the acquisition of antibiotic resistance genes, including SCCmec, quinolone resistance-determining region (QRDR) mutations, and a multidrug resistance genetic element (MDRST9). Evidence of direct transmission of animal-adapted strains to human hosts also suggest that transmission could potentially reshape the resistance and virulence genetic pool in these isolates. The rapid clonal expansion of MDR ST9 strains in mainland China and Taiwan highlights the increasing need for effective surveillance of antibiotic consumption in animal husbandry to control antimicrobial resistance spread. IMPORTANCE Staphylococcus aureus sequence type 9 (ST9) is the main LA-MRSA clone spreading in the Asian continent. It can colonize and cause mild to severe infections both in animal and humans. Previous work described its genotypic characteristics; however, the molecular history of global spread of ST9 strains remains largely unclear. We conducted a detailed analysis of genomic evolution of global ST9 strains and identified key genetic changes associated with its adaptation to specific hosts. Our results suggest that the ST9 clone originated from human-adapted strains, which lost genes related to the evasion of the immune system. The introduction of ST9 strains in animal populations was aligned with the acquisition of animal-specific virulent factors and mobile elements harboring multiple antimicrobial resistance genes, especially in isolates from mainland China and Taiwan.

IMPORTANCE Staphylococcus aureus sequence type 9 (ST9) is the main LA-MRSA clone spreading in the Asian continent. It can colonize and cause mild to severe infections both in animal and humans. Previous work described its genotypic characteristics; however, the molecular history of global spread of ST9 strains remains largely unclear. We conducted a detailed analysis of genomic evolution of global ST9 strains and identified key genetic changes associated with its adaptation to specific hosts. Our results suggest that the ST9 clone originated from human-adapted strains, which lost genes related to the evasion of the immune system. The introduction of ST9 strains in animal populations was aligned with the acquisition of animalspecific virulent factors and mobile elements harboring multiple antimicrobial resistance genes, especially in isolates from mainland China and Taiwan.

RESULTS
General characteristics of ST9 strains. To probe the molecular evolution of ST9 strains, we analyzed publicly available genomes at the time of the study (January 2021) (n = 174) and 17 additional genomes from our collection, including ST9 strains collected globally between 1941 and 2019 from human and animal sources. These ST9 strains were isolated from swine (n = 110), human (n = 34), bovine (n = 12), meat (n = 11), other animal (n = 10), and unknown (n = 13) sources. The isolates were mostly collected in mainland China (n = 140), followed by Taiwan (n = 8), Ghana (n = 7), Germany (n = 6), the United States (n = 6), Czech Republic (n = 2), and Argentina, Colombia, Australia, Switzerland, and the United Kingdom (one isolate each). The 140 isolates from mainland China were collected from 13 of 31 different provinces, autonomous regions, or municipalities, with wide geographic distributions (see Fig. S1 in the supplemental material).
Seventy-nine percent (151/191) of ST9 genomes harbored mecA, and the most predominant staphylococcal cassette chromosome mec (SCCmec) type was XII or XII-like (135/151). These SCCmec XII-like variants either harbored additional ccrA1/ccrB2 (n = 2) or lacked mec class C2 and ccrC2 (n = 11) (Fig. S2). SCCmec XII or XII-like elements were detected in isolates from both animal (bovine, chicken, meat, and swine) and human sources. Eight isolates harbored SCCmec IV (four from meat samples), and five contained hybrid SCCmec IV1XII (all from meat samples). Two strains harbored SCCmec V (2/191) and were of human origin. Selected genetic characteristics of human and animal isolates are presented in Table 1.
We also evaluated the presence of the immune evasion cluster (IEC) genes encoding staphylococcal complement inhibitor (scn), chemotaxis-inhibiting protein (chp), and staphylokinase (sak). These genes were found in f Sa3 prophages and were reported to be a major mechanism of human-specific adaptation contributing to the increased virulence of ST398 (20,21). Importantly, we found that the IEC genes were located in an ;42-kb prophage similar to phage P282 sequence, which truncated the hlb gene, the common insertion site for f Sa3 (Fig. S3). In total, 15 isolates were found to harbor IEC, nine of human origin and six of unknown origin.
Staphylococcal type Va genomic island in ST9 genomes. A previous study showed that ST9 strains harbor a Saa genomic island, which carries a SaPIbov4-like element, in addition to the staphylococcal superantigen-like (ssl1 to ssl11) and lpl tandem genes (22). This Saa was designated as a type Va genomic island based on the structure comparison with previously described type I to IV genomic islands and six additional novel Saa genomic islands (type VI to XI) (22). The SAPIbov4-like element encodes a von Willebrand binding protein (vwb), an important animal-related virulence factor that can cause bovine and caprine plasma clotting (22). To investigate whether the type Va genomic island was conserved in different ST9 strains, we analyzed its presence and sequence variation in the 191 genomes. We found that all isolates carry a Saa structure, and 86.4% (165/191) of genomes contain the SaPIbov4-like element in Saa. The sequence alignment showed that an ;14-kb SAPIbov4-like element was inserted downstream of the glutamine synthase gene (guaA) and upstream of the aminoglycoside transferase gene (aadE) (Fig. 1). This SAPIbov4-like element had high frequencies (.90%) in swine (102/110), bovine (11/12), chicken (2/2), livestock farm (7/7), and meat (11/11) samples but a lower frequency (25/34; 73.5%) in human samples.
Mutations associated with quinolone resistance in S. aureus were also commonly found in ST9 strains, with 89.0% (n = 170) presenting the double mutation gyrA_S84A/ S84L/S84V and parC_S80F in the quinolone resistance-determining regions (QRDRs). A small number of isolates carried rpoB mutations, including H481N (4.71%; n = 9), which was reported to promote the emergence of stable rifampin-resistant small-colony variant (SCV) subpopulations with reduced susceptibility to vancomycin and daptomycin (25). Other rare rpoB mutations (I527M and S529L) were detected in two isolates.
Phylogeographical context and comparative genomics of ST9 strains. Next, we implemented a Bayesian phylogenetic inference to decipher the global evolutionary history of ST9 LA-MRSA and to identify key genetic changes associated to its adaptation to human and animal populations. Core genome analysis identified 6,955 core SNPs across 191 ST9 strains. The BactDating model of temporal phylogenetic signal showed convergence and was significantly better than the randomized dates model, with effective Orange arrows, hypothetical proteins; gray arrows, integrase and excisionase genes; yellow arrows, transposases; red arrow, vwb gene; light blue arrows, staphylococcal complement inhibitor; purple arrows, nucleotidyltransferase; dark blue arrows, staphylococcal superantigenlike protein genes; pink arrows, restriction-modification system; green arrows, lipoprotein-like genes. Direct repeats are underlined. population sizes of greater than 200 (a, m, and s were .200). BactDating estimates that the most recent common ancestor (MRCA) of ST9 strains was around 1826 (95% confidence interval [CI], 1588 to 1912), approximately 200 years ago. The estimated mutation rate is 4.7 (95% CI, 3.6 to 5.9) single nucleotide polymorphisms (SNP)/genome/year, which is similar to the rates in other S. aureus lineages (26).
The clade II strains can be divided into three subclades (IIa, IIb, and IIc) and a few singletons (n # 2), with cluster IIc containing the largest number of genomes (n = 147) from mainland China and Taiwan. Clade IIa comprised only MSSA isolates, which were of human origin (5/6), collected in several countries around the world, including Argentina, the United States, and the United Kingdom. The spa types identified were t099, t100, t193, t464, and t587. Interestingly, most clade I and IIa (93.3%; 14/15) strains harbored the aforementioned IEC genes, scn, chp, and sak. The fact that we detected human MSSA isolates harboring the human complement evasionassociated IEC genes in the ancestral clades suggests a possible MSSA human origin of ST9 LA-MRSA clones and the change of human-specific virulence factors during the host shift to animals.
The separation of clade IIb/IIc from clade IIa correlated with the loss of the IEC genes (scn, chp, and sak), followed by the acquisition of the SaPIbo4-like element in the backbone of Saa genomic island (type V) (Fig. 1) and the acquisition of QRDR mutations (parC_S80F) in ;1956 (Fig. 3).

DISCUSSION
We detected three major genetic events along the evolutionary history of ST9: the loss of the IEC genes (scn, chp, and sak), which were reported to contribute to virulence in human infections, the acquisition of the SaPIbov4-like element-encoding vwb gene, which is an animal-specific virulence factor responsible for the clotting of animal plasma, and the acquisition of antibiotic resistance genes, including SCCmec, QRDR mutations and the MDR ST9 genetic elements.
First, all human MSSA isolates from the ancestral clade carried a f Sa3 b-hemolysin-converting prophage, harboring the genes implicated in immune evasion (scn, chp, and sak). In contrast, isolates from clade IIb/IIc carried an intact b-hemolysin gene and were negative for f Sa3, supporting the role of scn, chp, and sak as specific mechanisms of human adaptation. These three genes have a major role in complement evasion: the staphylococcal complement inhibitor (scn) binds to C3 convertases, preventing the activation of all three complement pathways (27); the chemotaxis inhibitory protein (chp) binds to C5aR1 and FPR1, thereby blocking the recognition of C5a and fMLF chemoattractants (28); and the staphylokinase (sak) activates plasminogen into plasmin, which is a serine protease bound to staphylococcal membrane which disrupts opsonization and phagocytosis through degradation of C3b and IgG, and it also blocks the cytolytic effect of human a-defensins (29).
Notably, previous studies showed that these virulence factors were specific for human hosts (30). Among them, SCN inhibits the alternative pathways exclusively found in humans (27), while the human chemotaxis-inhibiting protein had much lower capacity for binding to animal neutrophils, with 30-fold-reduced activation in mouse compared to human neutrophils (28). Similarly, bacterial plasminogen activators (PA) (such as streptokinase or staphylokinase) have a restricted ability to cleave different animal plasminogens (31). In addition, previous studies showed that the b-hemolysinconverting prophages were almost exclusively found in human isolates but absent from animal isolates (32). The results are consistent with previous studies showing that the IEC-harboring f Sa3 b-hemolysin prophages were closely associated with S. aureus strains from humans (33,34). Our results reiterate the loss of IEC is a major molecular event underlying the host shift from humans to animals during the molecular evolution of ST9 strains.
The second major event in the evolution of ST9 was the acquisition of the SaPIbov4-like element in the backbone of the Saa genomic island (type V). This SaPIbov4-like element harbored an animal-specific virulence factor gene, the vwb gene, which encodes a homolog of the von Willebrand factor binding protein (35). Several vwb alleles showed species-specific coagulation activities in animals through unique N-terminal motifs which activate bovine and equine prothrombin, as an important animal host adaptation mechanism (36). However, the effect of SaPIbov4-encoding vwb on plasma clotting has been evaluated only in bovine and caprine plasma (22,36). Its function in porcine plasma has not been fully studied, and future work is needed to understand its role in swine pathogenesis. Our analysis showed the acquisition of vwb in isolates from clade IIb and IIc, which contained mainly animal isolates. The SaPIbov4-like element was acquired by several spa types, initially in MSSA spa t1430 and later in MRSA spa t899 from clade IIb and clade IIc. This element also carried a second scn variant, which has 52.1% similarity to scn (SCIN-A) encoded in f Sa3. A previous study reported the identification of an equine scn (eqSCIN) variant in prophage f Saeq1, detected in different lineages of S. aureus exclusively isolated from horses (37). Remarkably, this variant inhibits C3 convertases from horses but also is a potent inhibitor of human and pig complement (37). However, the role of this scn in interfering complement function from different human and animal hosts remains unclear and deserves future studies.
The third notable event was the acquisition of multiple resistance genes, including SCCmec, QRDR mutations, and the MDR ST9 element(s). Interestingly, the acquisition of SCCmec and QRDR mutations correlated with the emergence of spa t899 in ST9 strains. The phylogenomic analysis revealed an ancestral clade composed of MSSA human isolates of diverse spa types and a most recent clade composed of MRSA animal isolates of predominantly spa t899. In addition to our analysis, previous molecular typing reports of MRSA isolates collected from pigs and pig industry-related workers in China have also showed the predominance of ST9 and ST9 single-locus variant (SLV) spa t899 strains (6,10,13,14). In contrast to our findings of spa t899-SCCmec XII predominance among ST9, isolates from previous work were spa t899-SCCmec III and SCCmec IV. These findings also support a human MSSA origin of ST9 MRSA, with independent acquisition of SCCmec elements in the background of different spa types.
Several of those genes, including aa(69)-Ie/aph(299)-Ia, blaR1, blaZ, tet(L), lnu(B), and lsa(E), were located in the MDR ST9 chromosomal region(s). Our phylogenetic analysis showed that the acquisition of MDR ST9 region(s) was exclusively found in clade IIc from isolates from mainland China and Taiwan. A BLAST search of this element against NCBI database failed to detect similar sequences from other S. aureus clones, suggesting the MDR ST9 may originate through the molecular evolution of ST9 strains. Interestingly, the b-lactamase genes (blaI, blaR, and blaZ) and the arsenic resistant genes (arsR, arsB, and arsC) were found in plasmids from ancestor clade I strains (Fig. 2). Examination of MDR ST9 identified multiple genes with insertion elements (IS256, IS6, and ISL3) and a transposon (Tn552) (Fig. 2). We therefore hypothesized that MDR ST9 may have originated from the chromosomal integration of plasmid-borne genes (such as blaI, blaR, blaZ, arsR, arsB, and arsC) along with the acquisition of additional antimicrobial resistance genes encoding tetracycline (tetL) and aminoglycoside [aac(69)-Ib/aac(69)-II] resistance, through IS-or transposon-mediated transposition, as a result of evolution against the increased antibiotic selection pressures. This clone appeared to be widely disseminated in China, and genetically similar isolates were also reported recently in food surveillance for antimicrobial resistance from raw meat products in Hong Kong (38).
Moreover, we also detected evidence of interhost transmission of ST9 strains. Within clade IIc, at least one cluster of isolates from pig and human sources showed very close core SNPs (,20), in support of the likelihood of the transmission of the ST9 strains between human and animals (Fig. 3). These isolates were recently described in a pork production chain and were found to be carried by pigs and pig workers in several farms (40). Although we could not determine the direction of transmission, both animal-to-human and human-to-animal transmission could be possible (10,11,16,17). Though not as common as animal-to-human transmission, human-to-animal transmission (i.e., reverse zoonosis) has been documented in S. aureus infections in livestock or companion animals (39). Interestingly, we detected one isolate carrying the IEC (scnchp-sak) in this clade IIc. This strain was isolated from a patient with a bloodstream infection (BSI) without livestock contact and showed high in vivo and in vitro virulence, with virulence genetic profiles closely related to those of human-associated ST9 MSSA (12). Nonetheless, this isolate carried the SCCmec XII, MDR ST9 , and the type V genomic island with the SaPIbov4-like element. Our results suggest that this isolate may have independently acquired human-specific factors (IEC) and caused severe infection in humans, molecular evidence of an MDR animal-adapted strain (identified by MDR ST9 and SaPIbov4-like elements) that could obtain hypervirulence through horizontal gene transfer. This finding warns of the potential emergence of multidrug-resistant and hypervirulent LA-MRSA strains.
Our results resembled the evolution of CC398, another successful LA-MRSA lineage, which originated in human as MSSA, spread to livestock, and later acquired methicillin resistance. In our study, the evolutionary analysis of ST9 genomes points to a human MSSA origin of ST9, which lost the IEC genes (sak, chp, and scn). The introduction of ST9 stains in animal populations was aligned with the acquisition of the SaPIbov4-like element, SCCmec (IV and XII), and multidrug resistance. The animal-adapted ST9 LA-MRSA strains showed the ability to infect humans, and the transmission from animal to human hosts could potentially reshape the resistance and virulence genetic pool in these isolates. The rapid clonal expansion of MDR ST9 in China and Taiwan highlights the increasing need for effective surveillance of antibiotic consumption in animal husbandry to control antimicrobial resistance spread.

MATERIALS AND METHODS
ST9 genomes. Seventeen ST9 S. aureus isolates, collected in China from three provinces between 2011 and 2016, were included in this study. An additional 174 ST9 S. aureus genomes were retrieved from the NCBI whole-genome sequence database or short-read archive comprising publicly available genomes at the time of the study (January 2021). The accession number, host, isolation source, geographical origin, and genotype data are listed in Data Set S1.
Dating of ST9 strains. Filtered reads from each isolate were mapped to the S. aureus ST9 reference genome (strain QD-CD9; accession number CP031838) by Snippy 4.4 (https://github.com/tseemann/ snippy) using default settings. For genome assemblies downloaded from the NCBI WGS database, 10 million 150-bp paired-end reads were simulated using the wgsim (https://github.com/lh3/wgsim) algorithm from SAMtools (50) and were mapped to the reference genome using Snippy. Prophages were predicted using PHASTER (51), and repeated regions were examined using MUMmer (52). SNPs among prophages and repeated regions were excluded, as they reflect horizontal gene transfer events or are unable to be resolved by short-read sequencing. The recombination analysis was then performed using Gubbins v3.0.0 (53). The BactDating R package (54) was used to estimate node dates of ST9 strains. The recombination-corrected tree from Gubbins output (53) and the isolation time were used as the inputs in BactDating v1.1 (54), using a mixed model with 10 8 iterations to ensure that the Markov chain Monte Carlo (MCMC) simulation was run for long enough to converge (the effective sample sizes of the inferred parameters a, m, and s were .200). Three BactDating replicates and one with a randomized tip date were conducted, and the convergence was evaluated with the Gelman diagnostic using the coda R package. The temporal signal significance was determined by comparing the first replicate model to the model with randomized tip date using the model compare function of the BactDating package (54). The resulting BactDating tree was then annotated using iTOL v5 (55).
Data availability. The raw reads of the 17 ST9 S. aureus genomes sequenced in this study were deposited in GenBank under BioProject accession no. PRJNA354234.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. DATA SET S1, XLSX file, 0.