Population Genomics Reveals Distinct Temporal Association with the Emergence of ST1 Serotype V Group B Streptococcus and Macrolide Resistance in North America

ABSTRACT

Identified in the 1970s as the leading cause of invasive bacterial disease in neonates and young infants, group B Streptococcus (GBS) is now also recognized as a significant cause of morbidity and mortality among adults with underlying medical conditions and the elderly. Concomitant with the increasing incidence of GBS invasive disease in adults is the rise of resistance among GBS isolates to second line antibiotics. Previous research shows that among serotype V GBS, one of the most common capsular types causing adult invasive disease, sequence type 1 (ST1), accounts for an overwhelming majority of adult invasive disease isolates and frequently harbors macrolide resistance. In this study, using whole-genome sequencing data from strains isolated in the United States and Canada over a 45-year period, we examined the association of antimicrobial resistance with the emergence of invasive serotype V ST1 GBS. Our findings show a strong temporal association between increased macrolide resistance and the emergence of serotype V ST1 GBS subpopulations that currently co-circulate to cause invasive disease in adults and young infants. ST1 GBS subpopulations are defined, in part, by the presence of macrolide resistance genes in mobile genetic elements. Increased frequency of macrolide resistance-encoding mobile genetic elements among invasive GBS ST1 strains suggests the presence of such elements contributes to GBS virulence. Our work provides a foundation for the investigation of genetic features contributing to the increasing prevalence and pathogenesis of serotype V GBS in adult invasive disease.

INTRODUCTION

Streptococcus agalactiae (group B Streptococcus [GBS]) emerged in the 1970s as the leading cause of invasive disease among young infants in the United States (1). Over the past 2 decades invasive GBS disease incidence has steadily increased among non-pregnant adults typically with underlying chronic disease or advanced age, thus it is a significant cause of morbidity and mortality among both infants and adults (2, 3). Invasive infant GBS disease is characterized by age at onset as an early onset disease (EOD), occurring at birth or up to 6 days, or late onset disease (LOD) between age 7 through 89 days. In both, bacteremia is the most frequent, but pneumonia and meningitis also occur (3, 4). Since the implementation of widespread GBS screening during pregnancy and intrapartum antibiotic prophylaxis in the mid-1990s, GBS EOD has declined by more than 80% (2, 5, 6). However, LOD incidence is unchanged and is nearly equal to that of EOD (2, 7). The most common clinical manifestations in non-pregnant adults are bacteremia without a focus, skin/soft tissue infections and pneumonia (3, 8–11), and nonpregnant adult cases currently account for 90% of the national disease burden. Adult invasive GBS disease is typically observed in patients who are older (mean age ∼60 years), or Black, and the majority of disease occurs in patients with co-morbid conditions such as diabetes, cirrhosis, decubitus ulcer, neurogenic bladder, immune deficiency, and malignancy (11–13). These associations are not exclusive to the U.S. patient population as indicated by studies conducted in other countries (14–16).

Differences in molecular epidemiology have been noted among GBS isolates derived from infant compared with adult invasive disease. GBS serotyping based on structural differences in capsular polysaccharides (serotype; Ia, Ib, II-IX) have been used to classify isolates (17, 18). Serotype III GBS is most frequently encountered in infants accounting for nearly half (47%) of EOD and 73% of LOD (4). In contrast, in the United States disease caused by serotype V is more common in adults (∼15%) compared with infants (<2% in LOD) (2, 19, 20). In a study from Alberta, Canada, similar findings were found regarding serotype prevalence associated with invasive GBS disease in adults (21). At the genetic level, multilocus sequence typing (MLST) has served to further classify GBS isolates into sequence types (ST) that have shown to differ in virulence potential and tropism (22, 23). Most significantly, a hypervirulent genetic lineage of serotype III sequence type 17 (ST17) has disseminated worldwide and is strongly associated with the majority of GBS infant disease, particularly meningitis (23–25). On the other hand, nonpregnant adult invasive disease has been mostly attributed to serotype Ia ST23 and serotype V ST1 strains (23, 25–27), particularly among the elderly (26).

Penicillin is the antimicrobial agent of choice for treatment of invasive GBS infection in infants and adults (3, 28). In cases of severe penicillin allergy, macrolides, lincosamides (i.e., clindamycin), and vancomycin are considered second line treatment options (29). However, GBS resistance to second line antibiotics has markedly increased since 1996 (30, 31) in the United States and elsewhere, concomitant with the rise in adult invasive disease cases and introduction of intrapartum antibiotic prophylaxis (2, 11, 20, 32). Epidemiological studies from the United States and Canada report a resistance frequency of 3–21% for clindamycin, and of 5–29% for erythromycin in GBS isolates, while other countries report even higher frequencies (6, 31). Increasingly frequent antimicrobial resistance (AMR) in neonatal invasive GBS disease has been associated with acquisition of mobile genetic elements in hypervirulent ST17 strains derived from the same genetic pool as adult invasive disease strains (33). A thorough overview of current understanding of the genetic basis and mechanisms of AMR associated with GBS disease may be found in the review by Hayes et al. (29).

Increased incidence of serotype V GBS invasive disease among non-pregnant adults in the past 25 years has been accompanied by a concomitant increase of AMR in GBS isolates (15, 27, 34–36). The evolution of AMR associated with adult invasive GBS disease has not been fully elucidated. In this study, we explore genetic features of ST1 serotype V GBS contributing to emergence of AMR in adult invasive disease. Our findings show a temporal association between increased AMR frequency and emergence of ST1 subpopulations in adult invasive disease. In addition, we demonstrate that AMR genotype correlates with the population structure of serotype V GBS lineages.

RESULTS

Pre-1992 invasive serotype V GBS strains are more frequently non-ST1 and susceptible to macrolide antibiotics than more contemporary strains.

We previously examined a collection of 229 serotype V GBS strains from the United States and Canada and showed that sequence type (ST) 1 strains predominate (84%) in adult invasive disease (27). The study also showed a relative lack of major recombination or horizontal gene transfer and instead suggested small genetic changes (e.g., single nucleotide polymorphisms, SNPs) contributed to ST1 GBS evolution. However, the study was limited to strains isolated between 1992 and 2012 from adult invasive disease. Using a local GBS surveillance collection of 59 strains from Houston, TX, we identified additional invasive serotype V strains (nonpregnant adult, women during labor or early postpartum and infant) pre-dating the rise of invasive disease in adult humans and increased macrolide use following the introduction of azithromycin in 1991 [(37, 38); Table S1 in the supplemental material]. Using Illumina MiSeq short-read sequence data, we determined the MLST for all isolates to elucidate the ST distribution among this unique data set. We discovered that the proportion of non-ST1 strains derived from invasive disease prior to 1992 (12/32, 37.5%; Table S1) was significantly greater compared to the proportion of non-ST1 strains from 1992 or beyond (1/27, 3.7%; P = 0.002, Fisher’s exact). Most commonly, pre-1992 non-ST1 GBS strains were ST2 (7/32, 21.9%)—a single locus variant of ST1—or ST26 (4/32, 12.5%). Interestingly, the two earliest known serotype V GBS strains from the Center for Disease Control and Prevention (CDC) (SS1168 and SS1172 in Table S1) were also ST2. Moreover, of the 837 serotype V GBS strains (ST-matched to the 59 strains examined) analyzed by the CDC’s Active Bacterial Core surveillance (ABC) from 2015–2017, only 5 were ST2 (5/837, 0.6%). Thus, prior to emergence and recognition of adult invasive disease, serotype V GBS showed much greater strain diversity, suggesting a specific host-pathogen interaction or other factors contributing to the rise of ST1 predominance.

Contemporary serotype V GBS strains derived from adult disease frequently are associated with antimicrobial resistance to macrolides (e.g., erythromycin) and lincosamides (e.g., clindamycin). Inasmuch as invasive serotype V GBS strains used in our analysis were obtained during a period prior to widespread use of macrolides or lincosamides stemming from the introduction of azithromycin (39), we hypothesized that the frequency of resistance to these two antibiotic classes would be lower than observed for contemporary serotype V GBS strains. Recent data from the CDC’s ABC surveillance showed that the majority of macrolide resistance results from erm elements [e.g., erm(A), erm(B), erm(T)], conferring resistance to macrolides, lincosamides, and streptogramin B (MLS_B). A smaller number of isolates contain mef/msrD or lsa elements resulting in macrolide or lincosamide resistance, respectively (20). We tested our hypothesis by determining resistance phenotypes to antimicrobials in common clinical use including macrolides (erythromycin) and lincosamides (clindamycin). As expected, nearly all isolates (53/59, 89.8%) showed resistance to tetracycline primarily attributable to the presence of tet(M) (Table S1 in the supplemental material). However, only a single isolate prior to 1992 was found to have resistance to erythromycin (1/32, 3.1%) compared with 7 in 1992 or later (7/27, 25.9%; P = 0.02, Fisher’s exact). Of the eight strains with erythromycin resistance, six harbored erm(A) and were clindamycin susceptible and two possessed erm(B) conferring both clindamycin and erythromycin resistance (Table S1). Of note, the rate of macrolide resistance due to erm(A) and erm(B) (8/59, 13.6%) in the studied strains was significantly lower than contemporary serotype V GBS from the CDCs 2015–2017 collection (594/837, 71%; P < 0.0001, Fisher’s exact). Overall, the data suggest that higher frequencies of macrolide resistance coincided with increased ST1 prevalence in adult invasive disease.

Temporal emergence of ST1 subpopulations is associated with high frequency macrolide resistance.

Recently, examination of GBS strains derived from diverse animals and humans showed extensive bacterial adaptation through gene gain/loss that was host specific (40). That study confirmed previous findings on the frequency of tetracycline resistance (∼70% of all strains) and overall AMR was most common in strains of human origin. However, while the evidence supported antibiotic resistance being associated with GBS expansion in humans, the signal was strongest for genes conferring erythromycin [e.g., erm(B) and erm(A)] but not tetracycline resistance [e.g., tet(M)]. Encouraged by this and the data indicating a temporal correspondence between increased prevalence of ST1 GBS and higher frequencies of macrolide resistance, we sought to further examine the relationship between ST1 emergence and macrolide resistance. Given that ST1 is the predominant ST among contemporary serotype V strains and the close phylogenetic relationship to ST2 (27), our analysis focused on these two STs. In addition to the 55 newly sequenced ST1/2 strains (Table 1), we included in our analysis previously sequenced ST1 serotype V GBS (n = 195; 1992-2012) (27) and 757 ST1 serotype V GBS publicly available from the CDC ABC’s surveillance system (2015–2017) (Table 2 and Table S2 in the supplemental material). Phylogenetic relationships were inferred based on single nucleotide polymorphisms (SNPs) relative to the reference strain SS1 (CP010867). ST2 serotype V GBS strains formed a distinct cluster (clade ST2) apart from the main ST1 population (clade ST1) (Fig. 1A). Maximum likelihood phylogeny followed by Bayesian clustering (BAPS) revealed that the main ST1 population was distributed into 3 subclades (SC1, SC1a, and SC2) (Fig. 1B). ST1 GBS strains were nearly evenly distributed between SC1 (502/997, 50.3%) and SC2 (495/997, 49.6%). SC1a strains represented a distinct lineage within SC1 strains and accounted for 14% of the ST1 population (140/997).

TABLE 1.

Serotype V GBS strains analyzed in this study

Locale	Yr	Age group	Disease	No.	ST1/ST2	Reference
Houston, TXa	1972–1996	Infant/perinatalb	Invasive	55	48/7c	This study, 49
Houston, TXd	1992–2012	Adult	Invasive	117	117/0	27
Canada	1992v2012	Adult	Invasive	76	76/0	27
U.S.e	2015v2017	Adult/Infant	Invasive	759	757	20, 61, 62
Total				1005

^aComplete strain list provided in Table S1 in the supplemental material. BioProject PRJNA556442.

^bPerinatal defined as women with invasive GBS disease during labor or the first 24 h postpartum.

^cSS1168 (1977) and SS1172 (1978) are ST2 strains originally described by Wilkinson (62).

^dBioProject PRJNA274384.

^eWhole-genome sequencing performed by the Centers for Disease Control and Prevention Active Bacterial Core Surveillance (BioProject PRJNA355303).

TABLE 2.

Occurrence of resistance genes in ST1 serotype V GBS

Genea	SC1b (%) (n = 362)	SC1a (%) (n = 140)e	SC2 (%) (n = 495)e	Total (%) (n = 997)	P valuec
tet(M)	307 (84.8)	117 (83.6)	479 (95.6)	904 (90.7)	NS
erm(A)	80 (22.1)	104 (74.3)	3 (0.6)	187 (18.8)	<0.0001
erm(B)	11 (3.0)	1 (0.7)	473 (95.6)	485 (48.6)	<0.0001
lsa(C)	1 (0.3)	39 (27.9)	6 (1.2)	46 (4.6)	<0.0001
Total	92 (25.4)	116 (82.9)d	476 (96.2)d	684 (68.6)d	<0.0001

^aThe genes erm(A), erm(B), and lsa(C) accounted for 99% of macrolide/lincosamide resistance in the ST1 population. The genes erm(T) (n = 6) and erm(D) (n = 1) were found rarely.

^bST1 subclade (SC) as defined in Fig. 1 with total number of strains within each subclade indicated.

^cObserved frequencies of the given resistance gene significantly different from expected between ST1 subclades. P value determined by Chi-square.

^dIncludes strains with >1 resistance gene in SC1a (n = 28) and SC2 (n = 6).

^eBold italic text indicates the highest resistance gene frequency among ST1 subclades.

FIG 1.

Phylogenetic reconstruction of serotype V GBS. (A) Neighbor-joining phylogenetic tree of ST1 and ST2 serotype V GBS. Individual nodes are colored by year as defined in the legend. Dashed line demarcates ST2 GBS strains. (B) Neighbor-joining tree as in panel A magnified to show only ST1 GBS strains. Subclades (SC1, SC1a, and SC2) are shaded and labeled. Nodes are colored by year as defined in the legend.

Examination of the distribution of strains between subclades in the ST1/2 GBS population suggested an uneven temporal distribution with early strains more likely to reside in SC1 (Fig. 1A and B). Prior to 2000, SC1 strains accounted for 90% of the ST1 population (64/71). However, in our collection beginning in 2000, we observed an abrupt increase in SC2 and to a lesser extent SC1a strains such that after 2015 more than 75% of strains from the CDC ABC belonged to SC2 or SC1a (Fig. 2). To better define temporal relationships, we performed a Bayesian analysis to infer ancestral dates using BactDating on the entire ST1/2 GBS population. BactDating is computationally less expensive and more amenable to large data sets than BEAST and results in similar estimates (41). The estimated substitution rate for the ST1 population was 1.6 substitutions per core genome year⁻¹ (0.8 × 10⁻⁶ substitutions per site year⁻¹)—a rate very similar to other streptococci determined using BEAST (42). The most recent common ancestor for the entire population (ST1 and ST2) was estimated at 1876 (root date, Fig. 3). The SC1 population, and therefore ST1 GBS, was predicted to have emerged in 1966 followed by SC2 in 1981 and finally SC1a in 1984 (Fig. 3). The small date ranges indicate a high degree of confidence with respect to subclade emergence. Overall, these data demonstrate a strong temporal signal in serotype V ST1 GBS.

FIG 2.

Serotype V ST1 subclades as a proportion of the population over time. Total number of strains in each year interval is given above individual bars. The year interval 2010–2014 includes Canadian strains (*) and 2015–2017 includes all sequenced strains from the CDC ABC (**).

FIG 3.

Rooted phylogenetic tree of 1005 ST1/ST2, serotype V GBS strains. Individual branch dates determined through BactDating are labeled and confidence interval given in parentheses. Nodes are colored by year as defined in the legend. Subclade designation (Clade ID), tetracycline resistance genotype (TET), macrolide-lincosamide-streptogramin genotype (MLS), and in silico amplicon designation (further defined in Fig. S1 in the supplemental material) are defined in the legend and indicated in the vertical bars to the right of the tree.

We next determined the resistance gene content of the ST1 GBS population using both short-read sequencing and de novo assembly information (see Materials and Methods). Distribution of resistance genes including frequencies within individual ST1 subclades are summarized in Table 2. As expected, tetracycline resistance as a result of tet(M) was common occurring in 90% (904/997) of strains and was not significantly different between ST1 subclades (Fig. 3). Consistent with macrolide resistance contributing to ST1 GBS evolution, we observed significant differences in the distribution of all macrolide resistance genes among ST1 subclades (Tables S1 and S2 in the supplemental material, Fig. 3). Macrolide resistance due to erm(B) was significantly more abundant in SC2 strains compared to either SC1 or SC1a with over 95% erm(B)-positive (472/495, 95.6%) (Table 2, Table S1, and Fig. 3). Closer examination of tet(M) distribution determined that distinct alleles, differing in 10 nucleotides (99.5% identity), corresponded to strains in SC1 and SC2 (Fig. 3), suggesting a possible link to erm(B). The macrolide resistance gene erm(A) was significantly more likely to be found in SC1 (80/362, 22.1%) and SC1a (104/140, 74.3%) strains compared with SC2 in which erm(A) was identified rarely and always occurred with erm(B) (Table 2, Table S1, and Fig. 3). Finally, 46 ST1 isolates were found to contain lsa(C)—the majority of which (39/46, 84.7%) formed a distinct subpopulation within SC1a (Fig. 3). Of note, SC1 strains are closest to the root of the phylogenetic tree (Fig. 3) estimated to have emerged in 1966 and demonstrated the lowest rate of macrolide resistance. In contrast, the highest rates of macrolide resistance were observed in SC1a and SC2 (Table S2 and Fig. 3)—subclades that were predicted to emerge much later when macrolide antibiotics were more widely used in clinical practice due to the improved efficacy and patient tolerance exhibited by azithromycin (39). Thus, together the data indicate a strong association between temporal ST1 subclade emergence and resistance to macrolide antibiotics.

Tn916-like MGE integrated at a common locus determine the majority of resistance genotypes in ST1 GBS.

Previously, we resolved the complete genome of the ST1 GBS strain SS1 (25). We identified five loci (RDF.1-RDF.5) in which mobile genetic elements (MGE) were integrated into the ST1 genome. RDF.2 was a composite MGE harboring both prophage-like and Tn916-like elements including tet(M) (Fig. 4A). Given that Tn916-like MGE are also known to contain additional resistance genes, we sought to determine the MGE associated with macrolide/lincosamide resistance genes in the ST1 GBS population. All macrolide resistance-encoding genes identified in the ST1 isolates analyzed were found adjacent to the Tn916-like component of RDF.2. Macrolide resistance conferred by erm(B) among SC2 strains resulted from integration of an erm(B)-containing cassette into Tn916 (Tn3872) while the small number of strains with lsa(C) resulted from cassette integration into Tn916 at a different locus (Fig. 4B). In contrast, erm(A)-positive strains resulted from a larger insertion between rlmD of the core genome and Tn916 (Fig. 4B). Consistent with erm(A) residing on an independent MGE from the Tn916-like component of RDF.2, strains positive for erm(A) were significantly more likely to be tet(M)-negative (30/187, 16%) compared to erm(B) (2/486, 0.4%) or lsa(C)-positive strains (0/46) (P < 0.0001, Fisher’s exact).

FIG 4.

Mobile genetic elements (MGE) associated with antimicrobial resistance genes in the serotype V, ST1 GBS population. (A) Structure of RDF.2 from the reference strain SS1. The Tn916 genetic element is shaded. (B) Magnification of Tn916 from RDF.2 in panel A showing integration sites of additional gene content. Individual antimicrobial resistance genes are labeled.

We next confirmed the presence of RDF.2-associated variation among the ST1 population. Using in silico PCR (see Materials and Methods), we generated amplicons representing the Tn916-like elements associated with RDF.2 from 908 (90.1%) strains of the ST1 population. Amplicon clustering based on a Mash distance matrix and visual inspection revealed four main groups in which groups 2 and 4 formed distinct subgroups (Fig. S1 in the supplemental material). We observed a significant correlation with the amplicon cluster, ST1 subclade, and presence/absence of antimicrobial resistance genes (Fig. S1 and Fig. 3). For example, of the 478 ST1 designated as SC2 and for which amplicons were generated, 409 (85.6%) were amplicon cluster 4D (Fig. S1 and Fig. 3). In comparison, no cluster 4D amplicons were found within SC1a and only 3 were found within SC1 (3/330, 0.7%; P < 0.0001, Fisher’s exact). Moreover, of the 412 ST1 strains with cluster 4D amplicons, 410 (99.5%) were erm(B) positive.

It is possible that additional mobile gene content apart from the variation within RDF.2 is associated with ST1 population structure. Thus, we performed a pangenome analysis to differentiate core versus accessory gene content within the ST1 GBS population. A total of 8,834 genes defined the pangenome of which 1,393 (15.8%) were considered core (shared by >99% of strains). To identify those genes most likely associated with subclade-defining MGE, the pangenome accessory (non-core) gene data set was pruned to include only those genes found between 5% and 95% of strains (n = 476; see Materials and Methods). Comparing frequency of individual accessory gene presence, we identified 191 genes significantly overrepresented within a subclade (TableS3, Fig. S2 in the supplemental material). Among the most highly significant genes were those associated with Tn916-like elements and variation with RDF.2 (Fig. 4 and Fig. S2). Additional gene content was identified but did not include subclade-defining features. However, absence of long-read sequencing data precludes definitive identification of MGE within the population. Together, these data demonstrate that the major gene content differences in ST1 GBS subpopulations are primarily associated with variation among Tn916-like elements within RDF.2.

DISCUSSION

Previously we determined that the emergence of serotype V GBS causing invasive disease in non-pregnant adults from the United States and Canada primarily involves ST1 strains with a close phylogenetic relationship (27). Here we have expanded our whole-genome sequence analysis to include serotype V strains collected prior to 1992 and more contemporary isolates (2015–2017). Our findings indicate that prior to emergence and recognition of widespread GBS adult invasive disease in the 1990s there was greater diversity of STs and more susceptibility to macrolide antibiotics among serotype V strains. In contrast, contemporary serotype V GBS isolates are significantly more likely to be ST1 and to harbor macrolide resistance. Interestingly, similar trends in the frequency of macrolide resistance have been observed in other contemporary studies (21, 23, 43–45). Our temporal analysis shows that the emergence of macrolide resistant serotype V ST1 strains coincides with the emergence of adult GBS invasive disease. In addition, differences in the distribution of macrolide resistance genes and associated MGE correlates with the presence of distinct subpopulations of co-circulating ST1 serotype V GBS that together currently represent the majority of GBS invasive disease in non-pregnant adults.

The identified ST1 subpopulations (SC1, SC1a, SC2 clades) are defined, at least in part, by distinct macrolide resistance-encoding elements. Notably, the SC1 clade strains exhibit the lowest frequency of macrolide resistance and stands in contrast to the high frequency resistance among the more recently emerged SC2 and SC1a clades. In addition, the high frequency of macrolide and lincosamide resistance characteristic of contemporary ST1 strains was not detected among our collection of pre-1992 serotype V strains. Interestingly, our temporal analysis showed that ST1 serotype V GBS strains were nearly evenly distributed between SC1 (50.3%) and SC2 (49.6%) subclades and demonstrate multiple subpopulations of AMR GBS strains co-circulating together over time. Even though more recent subclades (SC2, SC1a) represent more than 75% of contemporary ST1 serotype V GBS isolates (2015-2017), we did not observe a complete replacement of a specific subclade. Increased use of macrolides in pregnant and nonpregnant adults around the mid-1990s may have contributed to ST1 subpopulation emergence, but other uncharacterized selective forces (e.g., core chromosomal gene polymorphisms and gene content differences) among ST1 strains also likely play an important role.

All macrolide resistance-encoding genes identified in SC1, SC1a, and SC2 clades are found in the vicinity of a Tn916-like component of the RDF.2 mobile genetic element previously identified in ST1 GBS (27). The SC2 clade strains encode the erm(B)-containing Tn3872 mobile genetic element that is commonplace among human oral and upper respiratory tract streptococcal species (46–49). The presence of Tn3872 has been associated with the replacement of STs prevalent among adult invasive pneumococcal disease isolates from Spain by macrolide resistant STs between 1999 and 2007 (47). When first detected in GBS serotype V invasive disease strains isolated from neonates and adults in the United States between 1999 and 2000, the presence of Tn3872 suggested that acquisition of this element occurred through horizontal gene transfer from pneumococci (50). SC1 and SC1a strains were significantly more likely to encode the erm(A) resistance gene integrated between rlmD and the Tn916-like element. In addition, a significant proportion of SC1a strains encode resistance to clindamycin mediated by lsa(C), one of the genes identified as responsible for the LS_A (lincosamide-streptogramin A) resistance phenotype first described in a vaginal GBS serotype III isolate (51). The first analysis of the LS_A phenotype among GBS in the United States detected high frequency (75%) of lsa(C) in invasive disease isolates from non-pregnant adults (52).

The association of macrolide resistance-encoding genes with the RDF.2 element of ST1 strains and the ST1 prevalence among invasive serotype V strains is suggestive of a contribution of this mobile genetic element to GBS virulence. However, the contribution of the overall RDF.2 gene content to GBS global gene expression and virulence is still unknown. Intriguingly, our previous study found the RDF.2 region encodes a putative surface adhesin (alpha-like protein, AlpST-1) whose contribution to virulence and invasiveness remains to be explored (27). Recent literature (53, 54) evaluating potential GBS vaccine target candidates found Alp and pilus proteins are highly represented in adult invasive disease strains. In our previous analysis, we identified additional chromosomal loci to be under highly selective pressure in serotype V GBS strains (27), including loci encoding proteins involved in polysaccharide capsule production, regulators of pilus expression and members of two component gene regulatory systems whose contribution to ST1 virulence merit further investigation. Further confounding ST1 evolution are several examples of serotype switching including large-scale recombinations (55). The contribution of capsular switch events to ST1-related invasive disease and evolution remains to be elucidated. Thus, although GBS serotype V ST1 emergence in invasive disease can be in part explained by acquisition of novel AMR elements, further work is needed to determine the contribution of additional chromosomal loci under selective pressure with specific ST1 subclades.

Admittedly, our temporal analysis is limited by an uneven temporal and geographic distribution of strains with greater sampling over time. Nonetheless, the narrow date intervals observed in our temporal analysis indicate a strong temporal signal with respect to subclade emergence. Additionally, the pre-1992 GBS strains include isolates derived from infants. However, ST prevalence in non-pregnant adult invasive disease prior to 1992 is largely unknown, as failure to recognize adult invasive disease cases at that time meant that such strains did not represent a significant proportion of GBS-related disease surveillance isolates. Interestingly, a study by Blumberg et al. recognizing the emergence of serotype V isolates, indicated that strains collected from infant and adult cases of invasive GBS disease during 1992–1993 surveillance period shared the same ribotype (11), suggesting similar serotype V STs were circulating in adult and young infant populations pre-1992. Furthermore, a study by Teatero et al. (33) determined that a substantial number of serotype III, ST17 strains causing neonatal and adult invasive disease between 2009 and 2012 in Toronto, Canada were interspersed and clustered closely in a phylogenetic tree, providing strong evidence that adult and infant disease strains originate from the same genetic pool. Serotype V strains contribute the largest proportion of adult invasive GBS disease in the United States and Europe, but serotype VI–IX and Ia strains appear to constitute as large or larger proportions of adult invasive disease in Asia and Latin America respectively (53). Although most antimicrobial resistance in adult invasive GBS disease has been identified in ST1, serotype V strains (20), resistance has been identified in other serotypes and STs (56), meaning surveillance and genomic analysis of other serotype and sequence types is necessary to identify associations of resistance with invasive disease.

In summary, the current study represents one of the most comprehensive temporal analyses to date of a single GBS capsular type causing adult invasive disease. Our data suggest that widespread macrolide use led to the appearance of serotype V ST1 subpopulation emergence and co-circulation in adult invasive disease. Undoubtedly, genetic differences beyond AMR contribute to the ongoing evolution of ST1 subpopulations and pathogenesis of GBS disease—an area of ongoing investigation.

MATERIALS AND METHODS

Bacterial strains, growth conditions, and chromosomal DNA extraction.

Serotype V GBS strains used in this study are listed in Table S1 in the supplemental material. All strains were derived from invasive disease cases (infants <3 months of age, pregnant women during labor or within 24 h postpartum [perinatal] and adults), minimally passaged, and stored at −80°C. Strains were grown from frozen stocks on tryptic soy agar supplemented with defribrinated sheep blood (Remel, ThermoFisher, Lenexa, KS) and incubated at 37°C with 5% CO₂. Cells were collected for DNA extraction from plates for individual strains after assuring pure culture. Chromosomal DNA extraction was performed as have previously described (27) using the DNeasy blood and tissue kit (Qiagen, Valencia, CA) modified for Gram positive bacteria. Quantity and quality were assessed using a NanoDrop instrument and Qubit prior to proceeding to whole-genome sequencing.

Whole-genome sequencing and bioinformatic methods.

Short-read sequencing libraries were prepared using the Nextera Flex (Illumina, San Diega, CA) sequencing kit for 59 serotype V GBS strains (Table S1) as per the manufacturer’s instructions and 300-bp paired-end sequence obtained using an Illumina MiSeq instrument. A custom in-house pipeline was used to identify polymorphisms (single nucleotide and insertions/deletions) relative to the reference strain SS1 (accession number CP019867) including raw, short-read sequences obtained from the sequence read archive (SRA) at NCBI. Short-read sequences were trimmed using Trimmomatic v 0.38 and error correction performed using SPAdes v 3.12.0 (57). Corrected reads were subsequently mapped to the reference genome using SMALT v 0.7.6 (https://www.sanger.ac.uk/science/tools/smalt-0) and polymorphisms identified using freebayes v 1.2.0. All polymorphisms were filtered for depth of coverage (≥15-fold coverage), polymorphism frequency (≥75%), and (phred score ≥10). Resultant variant call format files were processed using the custom python scripts prephix and phrecon (https://github.com/codinghedgehog). Maximum likelihood phylogeny using biallelic SNP loci was estimated using RAxML (v 8.2.11). Node support was evaluated using 100 bootstrap replicates. Phylogenetic trees generated by RAxML were adjusted for recombination using ClonalFrameML (58). The program hierBAPS (59) was used to define subpopulations. Strains with evidence of recombination were excluded from further analysis (n = 95) based on Bayesian analysis (hierBAPS) and visual inspection using output derived from ClonalFrameML. BactDating (41) was used for estimation of ancestral dates with the recombination corrected ML phylogeny by ClonalFrameML as input and default parameters. Repeat date estimates were performed with increased chain length (up to 1 million) but significant improvements over default were not observed (data not shown). Subsequent visualization and manipulation of trees including associated metadata was performed using CLC Genomics Workbench v 20. De novo assembly using Illumina short-read raw data was performed using SPAdes (57).

SRST2 was used for molecular typing of individual strains including capsule (20) and sequence type with raw short-reads as input. Two independent methods were used to identify resistance genes in GBS strains. First, the resistance gene identifier (RGI) software (v 5.1.0) and the associated comprehensive antibiotic resistance database (https://card.mcmaster.ca; May 2020) was used to determine resistance gene content from raw short-read data. In addition, we used individual de novo SPAdes assemblies and BLAST to search the ResFinder database (v 4.1, May 2020). Discrepancies were resolved by visual inspection of the de novo assembly.

In silico PCR and pangenome analysis.

A bespoke in silico PCR perl script in_silico_pcr.pl (https://github.com/egonozer/in_silico_pcr) was used to create amplicons at the rlmD (23S methyltransferase insertion site of RDF.2 (primer_F: 5′-TAAAAATTCAACCAGTAGACTTATTCCCAATGAC-3′) as well as within RDF.2 downstream of the Tn916-like coupling sequence found in reference consensus assemblies (primer_R: 5′-TATCGCCTTTTGTGACCATTTTTATGAAATTTTTT-3′) using 998 ST1, draft, short-read assemblies. Because of the complex nature of segmental duplications arising from repeat structures common in MGEs, 90/998 (9.9%) amplicons were not extracted for further downstream analysis. A Mash pairwise distance matrix was generated from the 908 putative RDF.2 amplicon sequences. A principal-component analysis (PCA) was performed using the prcomp() R function on the pairwise distance matrix with PC1 and PC2 explaining 95.5% of variance (data not shown). The within groups sum of squares based on a max of 15 k-means clusters extracted from the pairwise distance matrix was used to determine the optimal number of clusters (n = 4). Clusters designations derived by PCA were overlaid on the Mash tree to further identify subclusters within the in silico generated amplicons (Fig. S1).

A pangenome analysis was performed on the ST1 population (n = 998) using Roary (60). Genes were subsequently pruned after being predicted and identified in greater than or equal to 5% of the total population (n = 50) and less than or equal to 95% of the total population (n = 955 yielding a total of 476 accessory genes (5.4%, 476/8834). Excluding singleton genes as well as genes shared by the majority of the population should reveal genetic signatures that are shared within clade and nested subclade structures. A Fisher’s exact test with a false discovery rate adjusted P value was calculated for the accessory genome subset proportions across each of the three subclades identified previously (i.e., SC1, SC1a, and SC2). Using a stringent FDR adjusted P value < 0.001, 191 accessory genes with statistically significant proportion differences across the three subclades were compared for gene presence/absence using a heatmap. The proportion of genes detected in reference Tn916/Tn3872, Tn916-like transposons, and within the full-length RDF.2 ICE among these 191 accessory genes was determined using a blastn query of representative genes in the pangenome matrix against RDF.2 genome references generated from the 11 complete GBS assemblies referenced in Table S2. In addition, accessory genes within the lsa(C) transposon were queried against the short-read assembly of SRR506123.

Statistics.

Chi-square was used to compare frequencies of antimicrobial resistance between groups. Where appropriate, Fisher’s exact test was used in place of Chi-square. All statistical analyses were performed using GraphPad/Prism v 9. P values < 0.05 were considered significant for all tests following correction of multiple comparisons.

Data availability.

Raw sequence data generated as part of this study are available under BioProject PRJNA556442. In addition, data from BioProjects PRJNA177823–PRJNA177840, PRJNA274384, and PRJNA355303 were used in this study.