Selection, Recombination, and Virulence Gene Diversity among Group B Streptococcal Genotypes

Abstract

Transmission of group B Streptococcus (GBS) from mothers to neonates during childbirth is a leading cause of neonatal sepsis and meningitis. Although subtyping tools have identified specific GBS phylogenetic lineages that are important in neonatal disease, little is known about the genetic diversity of these lineages or the roles that recombination and selection play in the generation of emergent genotypes. Here, we examined genetic variation, selection, and recombination in seven multilocus sequence typing (MLST) loci from 94 invasive, colonizing, and bovine strains representing 38 GBS sequence types and performed DNA sequencing and PCR-based restriction fragment length polymorphism analysis of several putative virulence genes to identify gene content differences between genotypes. Despite the low level of diversity in the MLST loci, a neighbor net analysis revealed a variable range of genetic exchange among the seven clonal complexes (CCs) identified, suggesting that recombination is partly responsible for the diversity observed between genotypes. Recombination is also important for several virulence genes, as some gene alleles had evidence for lateral gene exchange across divergent genotypes. The CC-17 lineage, which is associated with neonatal disease, is relatively homogeneous and therefore appears to have diverged independently with an exclusive set of virulence characteristics. These data suggest that different GBS genetic backgrounds have distinct virulence gene profiles that may be important for disease pathogenesis. Such profiles could be used as markers for the rapid detection of strains with an increased propensity to cause neonatal disease and may be considered useful vaccine targets.

Group B Streptococcus (GBS) is a leading cause of neonatal sepsis, pneumonia, and meningitis (51) and causes infections in pregnant women, nonpregnant adults, and the elderly with underlying medical conditions. Maternal GBS colonization is a main risk factor for neonatal disease, and roughly 20 to 40% of pregnant women are colonized (14, 23). Colonization rates of up to 31% and 34% have been documented in young men (4) and nonpregnant women (4, 42), respectively, whereas a rate of 22% has been observed in individuals over 65 years of age (18). GBS has also been identified as the cause of bovine mastitis in up to 45% of symptomatic bovines (30). Nine distinct polysaccharide capsule types (serotypes) are known, and the serotype distribution varies by population.

The genetic diversity of GBS populations has been studied using a variety of different methods, including restriction fragment length polymorphism (RFLP) (24), ribotyping (5, 25), pulsed-field gel electrophoresis (49), multilocus enzyme electrophoresis (MLEE) (45), random amplification of polymorphic DNA (36), restriction digestion pattern (RDP) typing (53), and multilocus sequence typing (MLST) (28). By utilizing methods that focus on conserved genetic changes within GBS strains, virulent GBS clones that have diversified genetically can be identified. Both MLEE and MLST can distinguish the major GBS serotype III clones associated with neonatal invasive disease as sequence type 17 (ST-17) in the MLST system (28, 29, 38) or electrophoretic type 1 in the MLEE system (45). This clone is also evident in the RDP system as RDP-III (53).

A recent study of 75 GBS strains representing different sources and STs reported that the ST-17 lineage is relatively homogeneous and contains a unique set of surface proteins (9). Homogeneity within a GBS lineage that is significantly associated with neonatal disease is likely important for disease pathogenesis, though few studies have been conducted to identify specific differences in virulence characteristics between lineages. Similarly, the roles of selection and recombination in the generation of STs, as well as known virulence genes, have only recently been explored and require further investigation (9a). Here, we assess the genomic diversity of GBS strains representing a variety of common clonal genotypes, examine evidence for selection and recombination, and evaluate the extent of DNA polymorphism and allelic variation in several putative virulence genes.

MATERIALS AND METHODS

Bacterial strains.

We selected 94 GBS strains that represented different multilocus STs, origins, and capsular serotypes for analysis (Table 1); the most common STs observed in other MLST studies (3, 29, 38) were overrepresented. Thirty-one strains were well-characterized controls obtained from the University of Michigan (n = 11), ATCC (n = 1; ATCC 12403 [22]), GlaxoSmithKlein (n = 6) (10), or the Channing Laboratory (n = 11) (56). The six bovine strains were isolated from milk expressed by bovines with mastitis in the United States (n = 3) (52) or the United Kingdom (n = 3) (2). Invasive strains from neonates (n = 16), adults (n = 5), and elderly patients (n = 3) were previously collected via population-based surveillance in Alberta, Canada (15), while the 35 colonizing strains were recovered from a cohort of pregnant women in Calgary, Canada (14).

TABLE 1.

Number of GBS strains representing 38 STs by capsule (cps) genotype, strain origin, and isolation date

STa	No. of invasive strains				No. of colonizing strains			cps genotype(s)	Date range	Geographic range
	Neonate (n = 29)	Adult (n = 7)	Elderly (n = 3)	Bovine (n = 6)	Adult (n = 1)	Pregnant women (n = 38)	Unknownb (n = 10)
1 (n = 20)	2	4	1		1	10	2	5, 5v, 6, 7, 8	1992-2002	United States, Canada, Japan
2 (n = 6)	3	1	1			1		2, 4, 4v	1970s, 1996-2001	Israel, United States, Canada
6 (n = 1)	1							1b	1930s	Unknown
7 (n = 1)	1							1a	1930s	Unknown
8 (n = 5)	2					2	1	1b, 5	1996-1999	United States, Canada
10 (n = 1)						1		1b	2000	Canada
12 (n = 6)			1			5		1b, 2	1999-2002	Canada
14 (n = 1)							1	6	Unknown	Unknown
17 (n = 6)	5						1	3	1973-1980s, 1995-1996	United States, Canada
19 (n = 7)	2					4	1	2, 3	1930s, 1998-2000	United States, Canada
22 (n = 3)	1					2		2	1999, unknown	United States, Canada
23 (n = 9)	5c	1				3		1a, 2, 3, NT	1930s, 1995-2000	United States, Canada
24 (n = 1)							1	1a	Unknown	United States
25 (n = 1)	1							1a	1930s	United States
26 (n = 2)							2	5	1970s, unknown	United States
28 (n = 2)		1				1		2	1999-2000	Canada
31 (n = 1)	1							3	1999	Canada
32 (n = 1)						1		3	2000	Canada
33 (n = 1)	1							3	1999	Canada
34 (n = 1)	1							3	1995	Canada
36 (n = 1)						1		3	1999	Canada
49 (n = 1)						1		1a	1999	Canada
61 (n = 1)				1				3	1991-1995	United States
67 (n = 1)				1				2	Unknown	United Kingdom
72 (n = 1)				1				2	Unknown	United Kingdom
76 (n = 1)				1				2	Unknown	United Kingdom
88 (n = 1)						1		1a	2000	Canada
91 (n = 1)				1				3	1991-1995	United States
103 (n = 1)						1		1a	1999	Canada
106 (n = 1)						1		3	1999	Canada
110 (n = 1)							1	5	Unknown	Unknown
130 (n = 1)						1		8	1999	Canada
148 (n = 1)	1							3	2001	Canada
356 (n = 1)				1				2	1991-1995	United States
411 (n = 1)	1							2	1997	Canada
412 (n = 1)	1							1a	1996	Canada
414 (n = 1)						1		4	1999	Canada
444 (n = 1)						1		1a	2000	Canada

^aGBS genome strains included in the study represent ST-10 (NEM316) (22), ST-7 (A909) (55), and ST-23 (2603V/R) (56), while three of the five incomplete genome strains represent ST-6 (H36B) (55), ST-17 (COH1) (55), and ST-19 (18RS21) (55).

^bUnknown; strains with missing information, typically well-characterized control strains.

^cOne strain was isolated from an older child, though the exact age is not known.

MLST capsule genotyping.

MLST was applied to all 94 GBS strains by amplifying and sequencing PCR fragments for seven housekeeping genes (adhP, atr, glcK, glnA, pheS, sdhA, and tkt) as described previously (41). Consensus sequences were trimmed in SeqMan (DNAStar, Madison, WI), and the GBS database (http://pubmlst.org/sagalactiae/; 27) was used for allele and ST assignments. A previously described (40) PCR-based RFLP assay determined the polysaccharide capsule (cps) genotype.

Uncovering allelic variation in cspA, gbs2018, scpB, and sip.

Published sequences (Table 2 and see Table S1 in the supplemental material) available for sip (surface immunogenic protein), scpB (C5a peptidase), cspA (serine protease), and gbs2018 (surface protein), also known as bibA (50a), were downloaded from GenBank, aligned using the ClustalW algorithm in MegAlign (DNAStar), and examined for variable regions that could be targeted in RFLP and fluorescent RFLP (fRFLP) (34) assays. MEGA4 (54) was used to construct neighbor-joining trees (50) and to classify the different alleles for each gene. Alleles were labeled numerically, and minor variants, which represented alleles that clustered at >98% bootstrap support with a major allele, were labeled with the major allele number followed by a number. After determining how many alleles were present for each gene among the published sequences, an in silico restriction digest analysis was conducted to map the expected number and size of DNA fragments generated following digestion with a specific enzyme (see Table S1 in the supplemental material). A variable region ranging between 2,000 and 4,000 bp was examined in gbs2018, while a 1,920-bp PCR product was examined for sip that utilized primers located in the two flanking genes (gbs0030 and gbs0032 [22]) (Table 2). For scpB, all 1,411 bp were analyzed, but because GenBank strain AF189002 (7) contained a 12-bp deletion that could not be detected via gel electrophoresis, a conventional RFLP assay was not developed. Instead, an enzyme that produced a 5′ overhang, which allows the incorporation of a fluorescently labeled nucleotide, was chosen for use in fRFLP (34). There was insufficient sequence variation among the known alleles of cspA to distinguish them by RFLP, so DNA sequencing was used in its place.

TABLE 2.

Primers^a for PCR-based RFLP analysis and DNA sequencing, fragment sizes, and assay conditions for four genes

Gene	Method	Primers (sequences)	Size (bp)	Annealing temp (°C) (time)	Elongation temp (°C) (time)	Published sequencesb (reference)	Deposited sequences (this study)
cspA	Sequencing	F2391 (5′-CGAAGTTCCTGGTTCAGAAGATT-3′)	574	53 (30 s)	60 (2 min)		FJ752028-52115
		R2964 (5′-TACTGCAGGACGAGCTTTGAAG-3′)
gbs2018	PCRc	O1a (5′-AAAATAAACGTGGTCCTATCCTAATAAA-3′)	2,000-4,000	54 (1 min)	72 (4 min)	AM051290-94 (9); AM183357-67, -72-86 (35)	FJ752117-52155
		O2a (5′-AAAGGCAAAGTTCTGATGAGGTT-3′)
	Sequencing	Same forward as PCR (O1a)	600	54 (30 s)	60 (2 min)
scpB	PCR	F284 (5′-CAGCAACCTCAAAAGCGACTATTA-3′)	1,411	54 (1 min)	72 (1 min)	AF189002 (7)	FJ752116
		R1994 (5′-ACGGTGACTTGTTTGCTGCTATT-3′)
	Sequencingd	F907 (5′-TGATAGTAGCTTTGGGGGCAAG-3′)	1,411	56 (20 s)	60 (2 min)
		R1373 (5′-CTTGCCCCCAAAGCTACTATCA-3′)
sip	PCR	F48 (5′-GCTATTATCAGTCGCAAGTGTTCA-3′)	1,218	54 (1 min)	72 (1 min)	AF151357, -59-62 (10);	DQ914240-42, -44-51, -53-56, -59, -61-63, -65, -67-69, FJ752156-61
		R1265 (5′-GCAGTAA CGCCACCACGAT-3′)
	Sequencinge	F464 (5′-TTGTTTCGCCAATGAAGACATA-3′)	1,920	54 (30 s)	60 (2 min)
		R485 (5′-TATGTCTTCATTGGCGAAACAA-3′)
		F1133 (5′-ACTCTACACAAAATATGGCAGCAA-3′)
		R1156 (5′-TTGCTGCCATATTTTGTGTAGAGT-3′)
		F541 (5′-AGTCAAGCAGCAGCTAATGAACAG-3′)
		R564 (5′-CTGTTCATTAGCTGCTGCTTGACT-3′)

^aPrimers (except gbs2018) are labeled by the gene or locus name, F or R for forward and reverse, and location within the gene in bp from the start codon and were developed based on the NEM316 genome sequence (AL732656) (22).

^bPublished sequences from the genome strains (NEM316 [AL732656] [22], A909 [CP000114] [55], and 2603V/R [AE009948] [56]), and incomplete genome strains (H36B [AAJS00000000] [55], COH1 [AAJR00000000] [55], CJB111 [AAJQ00000000] [55], 18RS21 [AAJO00000000] [55], and 515 [AAJP00000000] [55]) were used, as were additional sequences.

^cSlightly modified primers developed in a prior study (35) were utilized to amplify a gbs2018 fragment between 2,000 and 4,000 bp depending on the number of repeats. Sequence analysis was conducted on the first 600 bp (5′ end); however, for strains sequenced in this study, only single-stranded DNA was examined.

^dSequencing of scpB utilized three primer sets: the original PCR primers, F907 and R1373, and the reverse complement of F907 and R1373. The one strain chosen for scpB sequencing yielded a PCR fragment smaller than the expected size using the original primer set; the sequenced fragment was between F284 and R1994, as there is a 128-bp deletion within the original primer site.

^esip sequencing used three primer sets; sip allele 3d required the use of a different primer set (F541 and R564) because of a deletion within the F464 and R485 binding sites.

PCR-based RFLP analyses.

Primers that targeted conserved sequences flanking variable regions identified in the published sequence analysis were developed for each gene (Table 2). Digestions of PCR products specific for gbs2018 and sip were performed in a 30-μl reaction mixture using the following restriction enzymes and conditions: 5 U DdeI at 37°C overnight for gbs2018 and 5 U Taq^αI at 65°C overnight, as well as 5 U BanI, BanII, and HpaI at 37°C overnight, for sip. Conventional RFLP was used to examine the digested PCR products, and band sizes were estimated using Pro-Score/RFLP software version 2.39 (DNA ProScan, Inc., Nashville, TN). Rather than using RFLP, a portion of the cspA gene was sequenced in all of the strains included in the study (Table 2), and an established multiplex PCR protocol (13, 21) was used to identify the different alleles of the Alp protein gene family (bca, alp2, alp3/alp4, epsilon, and rib).

The fRFLP protocol developed for scpB was used to capture the 12-bp deletion identified in the published sequence analysis. First, 15 μl of PCR product was purified with 6 μl of ExoSap-It (USB, Cleveland, OH) to remove unincorporated deoxynucleotide triphosphates and primers. The reaction mixture was incubated at 37°C for 15 min and at 80°C for 15 min for enzyme inactivation and digested overnight with 5 U of DdeI at 37°C. Restriction fragments were fluorescently labeled using the GenomeLab Methods Development Kit (Beckman Coulter, Fullerton, CA) by combining 2 μl restriction digest product, 1.5 μl 10× reaction buffer, 0.1 μl ddUTP, 0.1 μl Taq DNA polymerase, and double-distilled H₂O to 15 μl and incubating the mixture at 60°C for 1 h. The labeled samples were purified with Sephadex G-50 Fine columns, dried using vacuum centrifugation, and resuspended in 10 μl deionized formamide. Finally, after the samples were mixed with CEQ DNA Size Standard 600 (Beckman Coulter), fRFLP was conducted under previously described conditions: a capillary temperature of 50°C, denaturation for 2 min at 90°C, injection at 2 kV for 30 s, and separation at 4.8 kV for 1 h (34). The DNA fragment sizes were visualized using CEQ2000XL software ver. 4.3.9 (Beckman Coulter).

Identifying additional virulence gene alleles.

Because the published sequence data did not represent the extent of variation that could be uncovered in the four virulence genes, DNA sequencing was completed on all GBS strains that yielded unexpected banding patterns in the RFLP analyses. Furthermore, a subset of additional GBS strains were selected for sequencing based on their placement within a neighbor-joining tree (50) derived from the MLST data. The CEQ2000XL DNA sequencer (Beckman Coulter) was used to sequence the gene-specific PCR products, though the number of strains, DNA fragment sizes, and conditions differed by gene (Table 2). Consensus sequences for cspA, gbs2018, and sip were aligned to published sequences with MegAlign using the ClustalW algorithm (DNAStar).

Phylogenetic analysis.

The neighbor-joining (50) and neighbor net (12) algorithms, untransformed distances, and bootstrap confidence values based on 1,000 replications were applied to concatenated MLST and virulence gene sequences using MEGA4 (54) and SplitsTree 4 (26). To test for the role of past recombination in generating allelic variation, the pairwise homoplasy index (11) was calculated in SplitsTree 4 (26).

The action of natural selection on molecular variation was inferred from the number of synonymous substitutions per synonymous site (dS) and the number of nonsynonymous substitutions per nonsynonymous site (dN) estimated by the modified Nei-Gojobori method using MEGA4 (54). In addition to the pairwise homoplasy index test, the genetic algorithm recombination detection (GARD) approach with a β-Γ rate distribution and three rate classes was used to further evaluate recombination and identify recombination breakpoints for each gene (http://www.datamonkey.org/GARD) (32). If breakpoints were identified, then the gene was broken up into segments defined by the breakpoints for further analysis. Testing for positive or negative selection on individual codons within each segment was performed using the single likelihood ancestor counting (SLAC) method (http://www.datamonkey.org) (31), which was also used to calculate the ratio of dN to dS and its 95% confidence interval. Finally, the F84 distance substitution model from DAMBE (http://dambe.bio.uottawa.ca/dambe.asp) (57) examined the level of substitution saturation by plotting the estimated numbers of transitions and transversions against the genetic distance for all pairwise comparisons of the alleles for the locus in question. Saturation was observed when the number of transversions exceeded the number of transitions and the transition frequency reached a plateau.

RESULTS

Genetic diversity of MLST loci.

A total of 38 STs were represented in this study of 94 GBS strains (Table 1). The 38 STs had 72 variable nucleotide sites among all seven concatenated housekeeping gene sequences (3,546 bp) and were defined by 6 to 10 alleles per gene (Table 3). The dS rate ranged from a low of 0.77 for tkt to a high of 1.88 for atr, with an average of 1.46 synonymous substitutions per 100 synonymous sites. The dN rate per 100 nonsynonymous sites was lower, ranging between 0.04 and 0.46 for glnA and tkt (Table 3). Tests for selection operating on the allele sequences at each MLST locus based on the SLAC method found no individual sites (codons) to be under significant negative or positive selection, indicating that the MLST loci are evolving neutrally. Furthermore, phylogenetic network analysis revealed no evidence for recombination in any loci except pheS, which was not confirmed using the GARD approach. Significant recombination, however, was detected for all 38 STs combined (Table 3).

TABLE 3.

Sequence variation in seven MLST loci from 38 STs

MLST locus	Amplicon size (bp)	No. of major alleles	No. of variable sites	dS × 100 (mean ± SE)	dN × 100 (mean ± SE)	dN/dS (95% CI)a	Phi Pb
adhP	498	10	11	1.61 ± 0.66	0.23 ± 0.11	0.20 (0.06, 0.47)	1.0
atr	501	8	12	2.33 ± 0.74	0.26 ± 0.15	0.09 (0.02, 0.23)	1.0
glcK	459	8	10	1.91 ± 0.75	0.23 ± 0.13	0.13 (0.03, 0.33)	1.0
glnA	498	8	9	1.36 ± 0.55	0.21 ± 0.12	0.11 (0.03, 0.29)	1.0
pheS	501	7	8	1.46 ± 0.69	0.32 ± 0.19	0.13 (0.03, 0.33)	0.04c
sdhA	519	6	12	2.58 ± 0.91	0.47 ± 0.23	0.18 (0.07, 0.39)	0.16
tkt	480	8	10	0.89 ± 0.54	0.68 ± 0.30	0.67 (0.31, 1.24)	0.16
Avg	493.7	7.9	10.3	1.73 ± 0.69	0.34 ± 0.18	0.22 (0.08, 0.47)
Totald	3,546	38	72	1.48 ± 0.26	0.20 ± 0.05	0.13 (0.10, 0.18)	6.5 × 10−10

^a95% CI, 95% confidence interval.

^bPhi P, pairwise homoplasy index P value.

^cDespite evidence for recombination using the Phi in SplitsTree (26), a more thorough analysis using the GARD approach (32) failed to identify any recombination breakpoints in pheS.

^dRepresents all 38 STs with concatenated gene data.

Population structure and recombination in MLST loci.

Among the 94 GBS strains, 29 of the 38 STs clustered into one of seven clonal complexes (CCs); nine singleton STs were identified that were not part of a cluster (Fig. 1). The largest cluster, CC-23, contained eight STs that grouped together with 98% bootstrap support, whereas the smallest clusters were comprised of only three STs each that grouped together with >87% bootstrap support (Fig. 1). The bovine strains represented three singleton STs (ST-67, -72, and -356) and one cluster (CC-61) that was not closely related to the CCs containing the human strains (Fig. 1). Of all the CCs, however, CC-61 was most closely related to CC-17, though the bootstrap value at the node separating the two clusters was only 42%. The STs within each CC differed by 1.3 (CC-1) to 4.0 (CC-12) nucleotides, and the overall difference between all seven CCs was only 19.4 nucleotides.

FIG. 1.

Phylogenetic relationships of 38 GBS STs. The numbers at the different nodes represent bootstrap values, and only those relationships with >80% bootstrap support are indicated. The colored circles show the different CCs that contained STs grouping together with >87% bootstrap support. p distance, untransformed distance. Bovine strains are represented in CC-61 and by STs 67, 72, and 356.

Phylogenetic network analysis clearly separated the CCs identified in the neighbor-joining phylogeny (Fig. 2). CC-23 was located most distant from the other CCs, while the CC-17 lineage was adjacent to the bovine-specific CC-61 lineage in both the network and neighbor-joining analyses (Fig. 1 and 2). The network also revealed that a significant degree of recombination has played an important role in the generation of GBS STs, particularly in the sample of singleton STs evaluated. There were too few parsimonious-informative sites in the MLST genes from each CC to detect recombination within CCs, and no recombination was detected (P = 0.29) between CC-17, -23, and -61 (Fig. 2), which represent the predominant clusters at the bottom of the neighbor-joining tree (Fig. 1). After grouping the two singleton STs (ST-22 and ST-32) with CC-17, -23, and -61, however, there was evidence for recombination (P = 4.2 × 10⁻⁴). Similarly, recombination also was not detected (P = 0.16) between the predominant clusters at the top of the neighbor-joining tree (Fig. 1), though after the singletons ST-36, -103, and -356 were included, recombination was apparent (P = 9.0 × 10⁻⁴).

FIG. 2.

Neighbor net analysis revealed a complex network with extensive recombination (shown as parallelograms) between 38 GBS STs. The circles highlight the CCs identified in the neighbor-joining phylogeny. p distance, untransformed distance.

Virulence gene diversity.

Several major alleles were identified for cspA (n = 12), gbs2018 (n = 6), sip (n = 8), and scpB (n = 4) in this population of 38 GBS STs using a combination of PCR-based RFLP assays and sequencing (Table 4). Many of these alleles have not been previously described, including scpB allele 4, which contains a 128-bp deletion. The most polymorphic gene was gbs2018. Considerably more variation was observed in the 2,000- to 4,000-bp gbs2018 PCR fragment than was expected based on the in silico analysis. Therefore, major allele assignments for gbs2018 were made based on the published sequence data from 20 strains representing six previously identified alleles (9); four minor variants were identified in these sequences, as well (see Table S1 in the supplemental material). Sequencing gbs2018 in an additional set of 19 strains did not identify any new alleles; however, a range of zero to eight variants were found per allele by performing RFLP of the larger PCR fragment in the remaining 55 strains. In all, 12 gbs2018 variants were found (Table 4), and five strains with gbs2018 allele 5 contained a 4-bp insertion. The average number of nucleotide differences between all six gbs2018 alleles in the first 600 bp of sequence was 44.9, with a total of 495 (83%) variable nucleotide sites (Table 4). The phylogenetic network constructed from the gbs2018 sequences revealed several parallel paths of mutation, which is indicative of phylogenetic incompatibilities in the early divergence of alleles and therefore provides evidence for recombination (Table 4). Significant recombination was also detected in the GARD analysis, as numerous recombination breakpoints were found. Despite this, a saturation analysis demonstrated that the observed substitutions within gbs2018 were saturated, thereby rendering all other phylogenetic inferences invalid, including both the SplitsTree and GARD analyses. This saturation level is likely due to the high degree of variation and large number of internal repeats previously identified within gbs2018 (35), which makes sequence alignments difficult.

TABLE 4.

Sequence variation in three putative virulence genes among 94 GBS strains representing 38 STs

Gene	Size (bp)	No. of major alleles	No. of minor variants	No. of variable sites	Phi Pa	dN/dS (95% CI)b				No. of negatively selected sites (location)
						Full gene	Segment 1	Segment 2	Segment 3
cspA	574d	12	2	18	0.57	0.15 (0.06, 0.29)				1
sipc	1,311	8	7	43	7.9 × 10−4	0.22 (0.14, 0.32)	0.42 (0.18, 0.82)	0.41 (0.20, 0.74)	0.05 (0.01, 0.13)	1 (segment 3)
gbs2018	600	6e	11	495	0.00

^aPhi P, pairwise homoplasy index P value.

^b95% CI, 95% confidence interval.

^cThe GARD approach found evidence for two recombination breakpoints in sip at 653 bp and 987 bp, which were used to define the three sip gene segments used in the SLAC method analysis.

^dA larger fragment (1,490 bp) of cspA was also evaluated in 26 strains. Only the data for the shorter fragment (574 bp) are given, as the results for the two fragments were similar in each category.

^egbs2018 allele 6 was not observed in GBS strains analyzed in this study but was identified previously (AM051294) (9).

In contrast, both sip and cspA were similar to the MLST loci, with fewer variable nucleotide sites and low dS and dN rates that were within the range of and slightly greater than the average for the seven MLST genes (Table 4). The complete sequence analysis of 41 strains identified eight major sip alleles. Four of the eight major alleles were identified from 14 previously published sequences (Table 2); one minor variant (sip-3.2), which contained a 4-bp deletion, was observed in sip allele 3. Sequencing sip in an additional 27 strains revealed four novel alleles (sip-5, sip-6, sip-7, and sip-8) and six minor variants (sip-1.2, sip-1.3, sip-1.4, sip-3.3, sip-3.4, and sip-8.2) that contained single nucleotide polymorphisms and/or deletions ranging in size from 60 to 120 bp. In addition, the sip-5 allele contained a 6-bp insertion at nucleotide 733. For cspA, 12 major alleles were identified among the 88 strains sequenced during the course of the study; only 3 of the 12 alleles were present among the eight genome strains with published sequence data available (22). Despite the large number of variants found, the 12 cspA alleles contain only 18 variable sites, while the 8 sip alleles contain 43 variable sites (Table 4). The phylogenetic networks for both genes (see Fig. S1 in the supplemental material) demonstrated parallel paths, though only sip had significant evidence for recombination (Table 4); the GARD analysis identified two recombination breakpoints in sip. No breakpoints were identified for cspA, indicating a lack of evidence for recombination, a finding that is consistent with the neighbor net analysis. Because of the recombination detected in sip, as well as the lack of evidence for saturation, the dS and dN rates were calculated independently for each gene segment using SLAC analysis (Table 4).

Clonal distribution of virulence gene alleles.

In general, some CCs are more homogeneous than others with regard to the distribution of specific gene alleles (Fig. 3; see Table S2 in the supplemental material). For example, only CC-1 strains have more than one gbs2018 allele, while the other CCs have strains with only one. The distribution of gbs2018-1 suggests that lateral gene exchange is likely responsible for the presence of allele 1 in more divergent genetic backgrounds (Fig. 3C). Additional evidence for lateral gene exchange was identified for sip-1, scpB2, the epsilon alpha-like protein gene, and cspA1, as each allele is located in multiple independent lineages (Fig. 3). In contrast, evidence for vertical inheritance was observed for the alpha C (bca) and rib alpha-like protein genes, cspA2, gbs2018-2, and sip4, as these alleles are present in closely related CCs consistent with the neighbor-joining phylogeny (Fig. 3). Some alleles also have a limited distribution across CCs, with the most notable lineage (CC-17) containing three unique alleles for sip, cspA, and gbs2018 (Fig. 3). CC-61 has unique alleles for cspA and gbs2018, while the PCR assays for scpB were negative (Fig. 3); scpB also was not found in the three singleton bovine strains. The utilization of previously described scpB primers (16) that are internal to those used in this study failed to amplify a fragment in all six bovine strains.

FIG. 3.

Frequencies and distribution of major alleles representing five GBS virulence genes by CC ranked in order according to the neighbor-joining phylogenetic tree (Fig. 1). The frequencies of alleles specific to alpha C protein genes (A), cspA (B), gbs2018 (C), scpB (D), and sip (E) are indicated by different colors. Only STs that were part of a CC were plotted, as singleton STs located near each CC typically shared the same allele profile and thus should not be grouped together as one category.

DISCUSSION

Little is known about the genetic diversity and virulence characteristics of common GBS STs that have been found in multiple populations worldwide. This study of the genetic diversity of GBS strains representing a variety of clonal genotypes confirmed the finding that different genetic backgrounds have distinct virulence gene profiles that may be important for disease pathogenesis (9). This finding is corroborated by comparative genomic data indicating that GBS has an “open pangenome” compared to other pathogens (9a, 55), thereby influencing the overall level of diversity. Here, we examined the population structure of 38 distinct GBS STs and found that the MLST genes appear to be evolving neutrally, since no individual codons are under negative or positive selection, as inferred from dN/dS ratios. The pairwise homoplasy index also indicated that the individual MLST loci lack evidence for recombination, providing additional support (28) for the use of these housekeeping genes in detailed phylogenetic analyses.

The neighbor-joining phylogeny grouped the 38 STs into seven distinct CCs with >80% bootstrap support (Fig. 1); prior studies (3, 8) identified similar clusters. The overall level of diversity is low; however, this sample of STs does not represent the complete diversity within the GBS population, as over 440 STs have been described to date (27). Despite the low diversity level observed in the MLST loci, the extent of networking at the base of each CC in the neighbor net analysis reveals a range of genetic exchange from low to high, as marked by the MLST alleles. CC-23 appears to have a low rate of exchange, CC-7 and CC-12 have medium rates, and CC-17 and CC-61 have high recombination rates (Fig. 2). These findings suggest the hypothesis that genotypic diversity is attributable to recombination between GBS genotypes in nature and that some genotypes are likely recombining at a higher frequency than others. This variation in the level of recombination may be due to regulatory differences or differences in the exchange levels of large DNA fragments (9a) among genotypes.

Most population prevalence studies of GBS disease have found that several common GBS STs (e.g., ST-1, -12, -19, -17, and -23) occur most frequently, although these studies also report additional singleton STs or variants (3, 8, 29, 38, 43). It is therefore possible that the more prevalent STs represent the oldest GBS lineages or that there have been one or more selective sweeps that have removed genotypes with “unfavorable” characteristics and promoted those with “favorable” ones. Furthermore, recombination may be responsible for the generation of closely related STs within a given CC, as well as the singletons located on distinct branches of the phylogenetic tree. Because shifts in the type of GBS strains important for disease have been observed (6, 48), the high level of recombination among different genotypes combined with the divergence of successful clones could potentially contribute to additional shifts, thereby influencing disease prevention protocols, particularly vaccination. Several GBS vaccine candidates have been investigated, including the polysaccharide capsule (1), alpha (47) and beta (46) C proteins, and Sip (39), which are promising alternatives to antibiotic chemoprophylaxis of women during childbirth. It is particularly important, however, for such vaccines to be multivalent and to have an ability to recognize multiple genetic variants in order to effectively combat GBS disease.

In this study, the high level of recombination observed among all STs (Fig. 2) may be important for the generation of novel virulence gene alleles, and it correlates with the degree of lateral gene exchange observed for some genes (e.g., alpha-like protein genes and scpB) (Fig. 3). Indeed, it is possible that lateral gene exchange is responsible for the distribution of some virulence gene alleles across divergent genotypes, but in other cases, vertical inheritance may contribute to the distribution of specific alleles among closely related genotypes. In some genes, however, there is evidence for both lateral and vertical gene transmission, as certain alleles are found in closely related genotypes (e.g., gbs2018-2 and sip-2), while others are found in divergent genotypes (e.g., gbs2018-1 and sip-1) (Fig. 3).

The CC-17 lineage, which contains several unique virulence gene alleles (e.g., cspA3, gbs2018-3, and sip-2) and is relatively homogeneous (9), appears to have diverged independently, with an exclusive set of virulence characteristics. The same is true for CC-61, the lineage comprised only of bovine strains. Such virulence characteristics were likely initially acquired through point mutation or recombination, as was demonstrated by the high level of gene exchange for MLST loci of both lineages (Fig. 2). In general, is possible that the success of both the CC-17 and CC-61 lineages in separate hosts can be attributed to subsequent and independent divergence. This hypothesis is supported by evidence demonstrating that GBS strains from humans and bovines represent distinct populations (2, 9, 19, 52,) and vary in the distribution of putative virulence genes (e.g., scpB [16, 17, 20, 58]). Moreover, ST-17, a member of CC-17 that is associated with neonatal disease (3, 29, 37, 38, 43) and meningitis (43), was found to have evolved from a bovine ancestor (2). Although recombination appears to be important in the generation of new virulence gene alleles (e.g., sip), there was no statistical evidence for positive selection using the SLAC method within sip or cspA. This is an unexpected finding for two surface proteins anticipated to be under immunogenic pressure, as such surface proteins are often highly variable, with mosaic allele structures and evidence of positive selection (33, 44). For sip, several large in-frame deletions were identified in some alleles, and therefore, it is possible that the generation of deletions is a way of creating protein structural variants that may allow GBS to evade host immune system responses. Such differences may limit the efficacy of sip as a single antigen to be used in GBS vaccines, although there was no variation in sip-2, the allele found exclusively among the neonatal disease-associated CC-17 lineage. Previous work on Sip has demonstrated that it elicited protective host immune responses (10), and it was one of four antigens identified in a multiple-genome scan that increased the survival rate of mice challenged with GBS (39). Despite this, differences in surface accessibility (39) and Sip antibody responses (10) have been observed, and such differences could be attributable to allelic variation within sip. The same is true for cspA and gbs2018, as the CC-17 lineage was comprised of strains with only cspA3 and gbs2018-3 (Fig. 3); ST-17 strains were found to have a single allele of gbs2018 in prior studies, as well (9, 35). Together, these findings suggest that the independent divergence of the CC-17 lineage has likely led to the dissemination of strains with virulence gene characteristics important for neonatal disease development. Such findings, however, require further investigation, and future studies should focus on examining the genetic diversity in a larger subset of GBS genotypes from numerous sources. In the meantime, specific virulence gene alleles (e.g., sip-2, gbs2018-3, and cspA3) can be used as markers to rapidly identify GBS infection with a strain belonging to the highly virulent CC-17 lineage.