The 155 strains, representing 28.7% of the 541 genomes analyzed, grouped in G2-Sg1-Yw, were isolated between 2001 and 2020. Most strains in this group were from Australia (23.9%; n = 37/155), the Philippines (19.4%; n = 30/155), and Argentina (16.1%; n = 25/155). Among the strains with reported isolation sites (58.1%; n = 90/155), the majority (63.3%; n = 57/90) were recovered from skin/soft tissue infections (SSTI), and 15.6% (n = 14/90) from blood. Nearly all strains in this subgroup are MRSA (99.4%; n = 154/155), with SCCmec type IVc predominating (87.7%; n = 135/154). Other SCCmec subtypes were detected at lower frequencies, suggesting multiple gain and loss events of SCCmec into this subgroup, similar to what was observed in G1-Pk. Among the typable genomes, the predominant spa type was t019 (Table 1). This subgroup is characterized by a cluster of strains from SWP (e.g., strain WBG10049), marked with green stars in Figure 1 and Supplementary Figure S1. SWP and Argentinian strains shared a common ancestor in 1967 (95% HPDI 1932–1988) (Figure 2), consistent with McAdam et al. (2012) findings that SWP strains shared a common ancestor in 1967 (95% HPDI 1952–1984).

Key evolutionary events in the specialization of G1-Pk and G2 include changes in GIs and the acquisition, by the strains in G2-Sg1-Yw, of an unnamed plasmid (Acc: CP053637.1) that carries the mphC gene associated with macrolide resistance and the cadD gene for cadmium resistance. The loss of a novel S. aureus pathogenicity islands (SaPI), SaPITokyo11212-like, occurred during the split between G2-Sg1-Yw and the other G2 subgroups around 1888 (95% HPDI 1823–1920); thus, it is present only in genomes G1-Pk and G2-Sg1-Yw (Table 3).

The SaPIs of CC30 strains were previously detailed by McGavin et al. (2012), who identified four main pathogenicity islands (SaPI-1 to 4). In fact, the SaPI1 identified by those authors shows higher nucleotide identity and coverage with SaPITokyo11212, as described by Suzuki et al. (2015). SaPITokyo11212 contains the ear and seb genes, where seb encodes enterotoxin B, which causes food poisoning, and ear encodes an exoprotein predicted to act as a superantigen. However, the deletion of the ear gene did not significantly impact a mouse model (Singh et al., 2017). The novel SaPITokyo11212-like lost the seb gene and the adjacent transposase IS256 (Figure 6). Therefore, the absence of virulence-associated genes may explain why this island was not conserved in other G2 subgroups. SaPI2, which carries the tst gene, was present in most genomes within G2-Sg2-Pr. In all tst-positive strains, the tst gene was harbored by SaPI2-like elements similar to those found in MN8 ST30 strain, which carry tst as the sole virulence gene (McGavin et al., 2012). SaPI3 was absent from all ST30 genomes. SaPI4, which also lacks virulence genes, was identified exclusively in G2-Sg2-Pr, a group containing MN8-related strains (Figure 6).

A region with hypothetical open reading frames (ORFs) named “proteins near ESAT” showed greater conservation among genomes of G2 subgroups (particularly G2-Sg1-Yw and G2-Sg3-Gn) compared to G1-Pk (Figure 7). In S. aureus, the ESAT-6-like secretion system (ESS), comprising 12 or more genes, facilitates type VII secretion (TVIIS) across the bacterial cell wall. ESS is crucial for secreting virulence factors during infections, influencing host immune responses by altering cytokine production (Anderson et al., 2016). However, the specific role of proteins located near ESAT region remains to be determined.

The lukSF genes encoding PVL in the yellow subgroup (76.8%; n = 119/155) primarily belong to the H2a haplotype (98.3%; n = 117/119), like G1-Pk. However, unlike G1-Pk strains, where the main phage carrying the PVL haplotype H2a was φPVL108, most strains in G2-Sg1-Yw carry PVL by φPVL-CN125 (73.1%; n = 87/119), with only a few associated with φPVL108 (16.8%; n = 20/119) (Table 2). Previous studies have reported similar phenomena of different phages carrying the same PVL haplotype (Di Gregorio et al., 2021; Coombs et al., 2020). Indeed, Di Gregorio et al. (2021) have also identified φPVL-CN125 as being associated with the prominent ST30 Argentinian cluster ARG-4, which encompassed genomes that clustered within G2-Sg1-Yw in our phylogenetic tree. Like G1-Pk, this subgroup carried mostly IEC type-B (93.5%; n = 145/155) (Figure 4). Strains of this subgroup carry vSaα with a more complete lpl locus (Figure 3) and vSaβ type-III. Additionally, 92.9% (n = 144/155) of all genomes in this subgroup were classified as sraP-B, featuring a variable region of 3,238 bp, as illustrated in Figure 5.

Most genomes in our phylogenetic tree clustered within this subgroup (51.9%; n = 281/541). Among the strains with reported isolation region, most were from the USA (30.6%; n = 83/271), the UK (11.8%; n = 32/271), and Australia (11.8%; n = 32/271), with samples collected from 1980 to 2022. Notably over 80% (n = 232/281) of the strains were isolated between 2010 and 2020. Most of the strains were MSSA (93.2%; n = 262/281) and did not possess lukSF, except for one strain (SAO23, Acc.: GCA_028363555.1), which carried the PVL haplotype H1a (Table 2) associated with φ5967PVL. The subgroup exhibited considerable variability in spa types, with the most common being t012 (23.1%; n = 65/281), t021 (15.3%; n = 43/281), and t338 (13.5%; n = 38/281) (Table 1). Among the 19 MRSA strains in this subgroup, various SCCmec subtypes were identified: SCCmecIVc (n = 11), SCCmecIVa (n = 4), and SCCmecIIa/IIb (n = 2), with two strains being untypable. The SraP-C polymorphism serves as a cladistic biomarker for this subgroup with a variable region of 784 bp (Figure 5).

These strains possess vSaα with a more complete lpl locus (Figure 3B) and vSaβ type-III, sharing most virulence-associated factors with other ST30 genomes. However, their genomes have expanded through significant acquisition such as SaPI2, which encodes the tst gene for TSST-1, present in 79.4% (n = 223/281) of the genomes analyzed. This is an important evolutionary trait, as all tst-positive strains, except one, are concentrated in this subgroup.

Studies estimate the annual incidence of toxic shock syndrome (TSS) to be between 0.03 and 0.07 cases per 100,000 population (Adams et al., 2017). Historically, ST30-MSSA strains have been associated with TSS, exemplified by the historical strain MN8 (depicted with a blue star in Figure 1 and Supplementary Figure S1), which is also clustered within this subgroup. Isolated in 1980 during a pandemic of menstrual toxic shock syndrome (mTSS) (Reingold, 1982), MN8 was part of 867 reported cases in the USA, with 492 cases in 1981. Eighty-eight cases resulted in death, leading to a case-fatality ratio of 5.6% among those with known outcomes, including 15 cases in 1981 (case-fatality ratio of 3.3%) (Cibulka, 1983). The incidence of mTSS significantly decreased after the withdrawal of highly absorbent tampons from the market. More recently, the reported incidence of mTSS ranges from 0.03 to 0.50 cases per 100,000 people, with an overall mortality rate of approximately 8% (Berger et al., 2019).

SaPI2 is estimated to have been acquired around 1963 (95% HPDI: 1937–1980) by G2-Sg2-Pr strains, coinciding with the absence of the lukSF gene in the analyzed genomes. The first non-menstrual TSS was described in the USA by Todd et al. (1978), though they speculated that it may have been observed earlier, corroborating the chronology presented in our study. SaPI2 appears to have entered exclusively into G2-Sg2-Pr, except for one strain (AUSMDU00017262, Acc.: GCA_018602235.1) grouped in G2-Sg3-Gn (Figure 1). This suggests a clonal origin for its spread among G2-Sg2-Pr, with its initial entry possibly being a rare event in a common ancestor of this subgroup. The absence of SaPI2 carrying the tst gene in some strains of this subgroup may result from evolutionary pressures, such as the metabolic cost of maintenance, reduced selective advantage, or spontaneous excision (Ubeda et al., 2008). Strains without SaPI2 might also gain a fitness advantage in specific niches, with variability potentially arising from genetic drift or host-pathogen interactions (Baltrus, 2013).

McGavin et al. (2012) identified a deletion in the trpD gene in the MN8 strain, which explains the known tryptophan auxotrophy of mTSS strains and may contribute to the vaginal tropism of tst-carrying strains. Most strains (87.9%; n = 247/281) in G2-Sg2-Pr possess the same trpD allele as the MN8 strains, reinforcing the notion that this deletion has enhanced the host adaptability of strains of this subgroup.

Intriguingly, none of the strains from this subgroup carry PVL phages. Previous studies using the agr knockout strain TB4, along with complemented and overexpressed mutants using pSK9067, have demonstrated that the Agr system serves as a negative regulator of tarM, which encodes a glycosyltransferase responsible for the α-N-acetylglucosamine modification of the major S. aureus phage receptor, wall teichoic acids (WTA) (Yang et al., 2023). In fact, most WTA in S. aureus are adorned with β-N-acetylglucosamine by the TarS enzyme (Mistretta et al., 2019). Consequently, inhibition of the Agr system can hinder infections by bacteriophages that require a higher proportion of β-N-acetylglucosamine (Yang et al., 2023). Additionally, they showed that different ST types, including ST30 (strain MN8), were equally infected by the phage tested, suggesting the presence of similar phage receptors. Therefore, it is reasonable to suppose that the agrC mutation might render G2-Sg2-Pr strains resistant to PVL bacteriophages. An interesting future approach would be to test this hypothesis by complementing agr-RNAIII in an agrC mutant G2-Sg2-Pr strain to assess PVL phage infection. Remarkably, the only tst-positive strain outside the G2-Sg2-Pr cluster was found in the G2-Sg3-Gn. Unlike the tst-positive strains from G2-Sg2-Pr, which do not have lukSF genes and show a mutated agrC allele, this strain also carries the lukSF genes for producing PVL. Thus, we analyzed the agrC of this isolate and found a WT allele, which aligns with the premise above. It is well known that Agr inhibition decreases bacterial virulence in abscess model. Despite this, dysfunctional Agr strains can still cause severe invasive infections, including bloodstream infections (BSI). Moreover, this condition may confer advantages to hospital pathogens, as Agr impairment is associated with enhanced bacterial intracellular survival and persistent bacteremia (Das et al., 2016).

The prevalence of ST30-MSSA in G2-Sg2-Pr, particularly among the vulnerable age group of children under 2 years old, raises concerns due to their increased risk for S. aureus pneumonia (Neto et al., 2020). This association should raise important concerns within the medical community in countries where these clones are prevalent. Studies also indicate silent dissemination of tst-positive MSSA-ST30 in ICU patients and staff in Greece, posing a risk of hospital transmission and severe infections (Papadimitriou-Olivgeris et al., 2017). Additionally, research has shown a high prevalence of ST30-MSSA among children and infants (Achermann et al., 2015; Liang et al., 2021). This silent spread results from the focus of most epidemiological surveillance on identifying MRSA strains in hospitals.

Another notable aspect of this subgroup is the prevalence of the sea gene and the decrease of chp compared to G1-Pk, G2-Sg2-Yw, and G2-Sg3-Gn. The increased presence of IEC type-A (46.3%; n = 130/281), carrying sea, as the most prevalent type in this group is relevant, especially considering that IEC-B is generally more prevalent in S. aureus (Ahmadrajabi et al., 2017; Figure 4). Strains causing BSI that carry both sea and tst have been reported (Nienaber et al., 2011). Furthermore, studies using rabbit models of tampon-associated vaginal infections found that strains with both tst+/sea + exhibited greater virulence, indicating that the combination may enhance the pathogenic potential of these ST30 strains (Nienaber et al., 2011).

Moreover, 91.8% (n = 258/281) of genomes in this group lack the toxin-antitoxin system toxM, consistent with the findings of McGavin et al. (2012) regarding its absence in some CC30 genomes. Thus, it is possible that this absence may relate to a niche adaptation of this subgroup (McGavin et al., 2012). SaPI4, which lacks known virulence genes, was detected only in this subgroup. The genomes in this subgroup harbor a newly identified conjugative transposon (designated here as Tn1421), which carries a transposase of the ISSau5 and two ORFs annotated as staphylococcal secretory antigen-like (SsaA-like proteins) (Supplementary Table S1). SsaA is a peptidoglycan hydrolase known to be highly immunogenic. In S. epidermidis, SsaA has been linked to exacerbating nosocomial infections (Lang et al., 2000). Global gene expression studies have linked the S. aureus ssaA gene to biofilm production, suggesting that SsaA-like proteins may serve as virulence factors contributing to the success of these strains (Resch et al., 2005).

Another MGE, possibly an integrative plasmid of about 24 kb carrying two IS1272 insertions, was specific to this ST30 subgroup. This IS is typically found in SCCmecIV, serving as a biomarker for differentiating SCCmec types (Boye et al., 2007). The presence of IS1272 insertions outside SCCmec may lead to false positives for SCCmecIV (Viana et al., 2022). This MGE also contains genes associated with tolerance to cadmium, zinc, mercury, and arsenic, as well as genes encoding copper-transporting ATPase (copB) and multicopper oxidase (mco), which are typically mobilized together. Only certain S. aureus strains carry copB-mco, primarily those in CC22, CC30, and CC398 (Zapotoczna et al., 2018). These genes have been shown to enhance bacterial growth under copper stress and improve survival in macrophages and human blood (Zapotoczna et al., 2018), indicating that the entry of this MGE could significantly influence the evolution, spread, and pathogenesis of TSST-1-producing strains.

A unique insertion of ISSau2 was reported by McGavin et al. (2012) downstream of saeRS in the MN8 strain, which they speculated could disrupt the transcriptional terminator of saeRS, increasing gene expression and compensating for agrC mutations. However, in our study, this insertion was sporadic among ST30 isolates, predominantly occurring in G2-Sg2-Pr (7.1%; n = 20/281) compared to other ST30 strains (1.9%; n = 5/260), suggesting a rare and convergent evolutionary event, as they independently acquire this element. This subgroup exhibits a similar antimicrobial susceptibility pattern to other G2 subgroups, except for methicillin, as most isolates are MSSA, and a higher frequency of ermA gene (40.2%; n = 113/281) (Supplementary Figure S2), the most common non-β-lactam antibiotic resistance gene reported among CC30 strains (Goudarzi et al., 2020; Viana et al., 2021).

This subgroup includes 14.2% of the genomes analyzed (n = 77/541), primarily from Brazil (50.6%; n = 39/77), and several European countries (35.1%; n = 27/77). Of the 42 Brazilian ST30 genomes analyzed, only three clustered in G2-Sg2-Pr, while all others, including the Brazilian strains sequenced by our team, were MRSA and grouped in G2-Sg3-Gn. These strains account for 18.2% of CC30-SCCmecIV (n = 109) in a collection of 600 MRSA clinical strains isolated between 2014 and 2017 in Rio de Janeiro, originating from various clinical sources, including respiratory secretions, BSI, SSTI, and nasal swabs (Viana et al., 2021). The genome sizes of the 37 strains sequenced by us ranged from 2,792,580 bp to 2,963,657 bp, with GC content between 32.6 and 32.8%. The number of detected ORFs varied from 2,668 to 2,690, with most isolates carrying SCCmec type IVa and exhibiting variable spa types (Supplementary Table S1).

The SraP-D polymorphism, located in a variable region of 1916 bp, serves as the cladistic biomarker of this subgroup (Figure 5). There is also significant diversity in spa types, with t021 (24.7%; n = 19/77) and t318 (11.7%; n = 9/77) being the most prevalent (Table 1). Most genomes in this subgroup are MRSA (83.1%; n = 64/77), primarily associated with subtypes IVa (64.1%; n = 41/64) and IVc (17.2%; n = 11/64) of SCCmec (Table 1). Among the 68 strains with isolation source reported, mostly were from SSTIs (38.2%; n = 26/68) and colonization sites (33.8%; n = 23/68). PVL genes were detected in most strains (85.7%; n = 66/77) from haplotypes H2b (69.7%; n = 46/66) and H1a (30.3%; n = 20/66) (Table 2). However, O’Hara et al. (2008) associated haplotype R with MRSA and H with MSSA strains. This discrepancy could be due to the absence of ST30 strains in their study. They also indicated the haplotype H2 as the likely ancestor of H3 and H1 variants. In our study, PVL-H1a haplotype was detected in the more terminal subgroups of G2. Most PVL-positive strains from G2-Sg3-Gn (53.0%; n = 35/66) carry the PVL phage φ5967PVL (Table 2), which is considered a PVL phage likely not descended from earlier bacteriophages such as φPVL108 (Zhang et al., 2011).

IEC-B predominates in this subgroup (64.9%; n = 50/77), albeit at a lower frequency than in G2-Sg1-Yw, where nearly all genomes possess this IEC (Figure 4). IEC-A is present in 20.8% (n = 16/77) of G2-Sg3-Gn genomes, indicating the presence of the sea gene. The vSaα of this subgroup shows greater similarity to G2 subgroups than to G1-Pk, which carries only the lpl4 gene (Figure 3). In this subgroup, the mecA gene was found in 83.1% (n = 64/77) of genomes. Besides mecA, the dfrG gene was identified in 33.8% (n = 26/77) of genomes in this subgroup, excluding the Brazilian strains (Supplementary Table S2). This gene encodes dihydrofolate reductase which confers trimethoprim resistance. This is concerning, as sulfamethoxazole-trimethoprim is often recommended for treating uncomplicated skin infections, including those caused by CA-MRSA (Esposito et al., 2017).

Only the genomes of G2-Sg3-Gn carried a novel SaPI, designated here as SaPI-14-021. This SaPI is located between positions 798,755 and 813,890 in the ST30-MRSA strain CR14-021 (Acc.: GCA_021012415.1), consisting of 26 coding regions. It integrates into an att site (TCCCGCCGTCTCCAT), identical to that of SaPIm4, and shares the same integrase sequence (Novick and Subedi, 2007). SaPI-14-021 shares 75% nucleotide identity with the SaPI (fhuD) (Khokhlova et al., 2015) and 79% with SaPI68111 (Li et al., 2011), suggesting it could result from a recombination of these two elements (Figure 8).

Notably, SaPI-14-021 lacks the fhuD gene, a key virulence marker of SaPI (fhuD), and instead contains an ORF that encodes a putative ferric-siderophore transport system protein (tonB-like) with no similarity with fhuD. This ORF was annotated using RASTtk (BV-BRC v. 3.39.10).³ Mosaicism involving pathogenicity islands is a well-documented phenomenon. Uehara et al. (2019) reported a mosaic pathogenicity island named SaPISaitama2, which includes the tst, sec, and sel genes, originating from the recombination of SaPIm4 and SaPIm1. Similarly, Iwao et al. (2012) described the SaPIj50 island as a mosaic of SaPIm4, PI T0131, and SaPIm1, carrying the tst, sec, sel, and bla genes. SaPI-14-021 was found in most strains in G2-Sg3-Gn (80.5%; n = 62/77), particularly in a more terminal subset (Figure 9). A search of the NCBI database revealed its presence also in the genomes of strain 13420, an ST30-SCCmecIV vancomycin-intermediate strain from Brazil (2009) (Chamon et al., 2020), and strain ER09001.3 [ST1472-(CC30)-SCCmecIV], isolated in the USA in 2017 (Caban et al., 2020).

SaPI-14-021 likely entered a basal ancestor of this subgroup around 1953. In the more terminal subset of G2-Sg3-Gn, it was present in nearly all strains, except for three, which share a more recent ancestor with the other strains in this subset, dating back to 1983 (95% HPDI 1969–1992) (Figure 9). Limited genome sequencing prevented a more comprehensive tracing of the evolutionary history of these strains. However, tree analysis suggests that the Brazilian strains likely originated in Europe, as the more basal genomes were from European countries (Figure 9). The tree structure also suggests that the island was likely acquired by the G2-Sg3-Gn genomes only once, was lost in some strains, but became highly conserved within the subset of Brazilian strains. Additionally, the most basal genome in the G2-Sg3-Gn cluster, UP678, also carries this island in the same location as the Brazilian strain CR14-021, further supporting this conclusion (Figure 9).

Another notable finding specific to this subgroup is the presence of a cadmium resistance plasmid, pNTUH_6457, found in 85.7% (n = 66/77) of strains (Acc.: LC377539, DDBJ database). Interestingly, this plasmid differs from the one associated with cadmium resistance in group G1-Pk, indicating a convergent pattern for cadmium resistance by distinct MGE. Additionally, an unnamed plasmid, present in 42.9% (n = 33/77) of these genomes—primarily in the second subset of strains that includes Brazilian isolates—carries 18 ORFs, including one coding for a penicillin-binding protein, one homologous to the mecR1 gene, and another coding for an efflux ABC transporter. This may indicate a potential link between resistance and fitness/virulence, as efflux systems have been shown to be crucial for virulence and bacterial adaptability in various bacterial species (Olivares Pacheco et al., 2017; Wang-Kan et al., 2017; Gaurav et al., 2023).