Bacterial Strains, Culture Conditions, DNA Preparations, Serotyping, and Pilus Typing

We used 37 serotype IV GBS strains collected during 2009–2012 by the Toronto Invasive Bacterial Disease Network, a population-based surveillance system for invasive bacterial diseases in metropolitan Toronto and the Peel region in Ontario, Canada (total population under surveillance ≈5.5 million persons in 2011). Laboratory-based surveillance involves the 28 hospitals providing care to and the 25 laboratories processing sterile site cultures for residents of the population area; laboratory personnel submit all GBS isolates from sterile sites to the central study laboratory. The sample was composed of all available serotype IV GBS strains determined by latex agglutination and corresponded to 6.2% of GBS identified during the study by the program (8). Only 1 strain per patient was included in the collection; there was no clear evidence of localized outbreaks caused by serotype IV GBS during the collection period. Strains are listed in Technical Appendix Table 1.

Bacteria were grown at 37°C in an atmosphere of 5% CO₂ on Columbia agar plates supplemented with 5% sheep blood. Genomic DNA was obtained by using the QIAamp DNA Minikit (QIAGEN, Toronto, Ontario, Canada) according to the manufacturer’s protocol for gram-positive organisms. The presence of pilus islands PI-1, PI-2a, and PI-2b was determined by using PCR as described (22).

WGS, Bioinformatics Methods, and Phylogenetic Analysis

WGS libraries were prepared for all 37 isolates by using Nextera XT Kits (Illumina, San Diego, CA, USA) and sequenced as paired-end reads with either Illumina HiSeq 2500 (100 bp + 100 bp) or MiSeq (150 bp + 150 bp) instruments (data are available at the Sequence Read Archive under accession no. SRP040805). Parsing of the multiplexed sequencing reads and removal of barcode information was performed by using onboard software. STs were derived directly from short reads by using SRST2 (https://github.com/katholt/srst2). Ambiguous or novel alleles were confirmed by PCR and Sanger sequencing.

Neighbor-joining phylogenetic trees (1,000 bootstrap replications) were then generated by using SplitsTree4 (23) and the concatenated sequences of the 7 loci used in the GBS multilocus sequence typing (MLST) scheme. For WGS-based phylogenies, we sequenced to closure the genomes of ST-459 strain NGBS061 and ST-452 strain NGBS572 (GenBank accession nos. CP007631 and CP007632, respectively) by using single-molecule real-time (SMRT) sequencing (Pacific Biosciences, Menlo Park, CA, USA). Two SMRT cells of sequence were generated for each isolate, which generated >220 Mb of data in reads >3 kb (37,218 reads, average length 6.1 kb for NGBS061; 54,803 reads, average length 5.6 kb for NGBS572). In brief, we used HGAP version 2 (24) to correct the long reads and Celera Assembler 7.0 (25) to assemble the corrected reads, followed by 2 rounds of polishing with Quiver (https://github.com/PacificBiosciences/GenomicConsensus). Coverage of final assemblies in reads >3 kb was 84× for NGBS061 and 111× for NGBS572.

To assess base-calling accuracy in the Pacific Biosciences assembly, Illumina short reads for the 2 isolates were aligned to their respective assemblies by using BLAT (26). Both genome assemblies were completely concordant with full-length perfectly aligning Illumina short-reads, except for NGBS061 in a region of increased AT richness, which had reduced Illumina coverage. For consistency, both genome assemblies were formatted to begin at the first nucleotide of the intergenic region immediately preceding the gene encoding the chromosomal replication initiation protein (dnaA; Genbank accession no. YP-328728).

Illumina short reads of all other isolates were aligned to the reference genomes, and coverage relative to the reference genome was calculated by using Mosaik (https://code.google.com/p/mosaik-aligner/). Polymorphisms were identified by using VAAL (27). A matrix file containing genotypes of all strains at each polymorphic locus was created from the VAAL polymorphism output data by using a custom script. Then, for each individual strain, single-nucleotide polymorphisms (SNPs) were concatenated in order of occurrence relative to the genome of the corresponding reference strain (see Results). Neighbor-joining phylogenetic trees (1,000 bootstrap replications) were then generated with SplitsTree4.

The A5 pipeline was used for de novo assembly of Illumina-sequenced GBS strains (28). Areas of recombination of the ST-196 strains NGBS447 and NGBS472 were defined by using BratNextGen software (29) run with 20 iterations and 100 replicates, using a p value of 0.05 as the significance cutoff. Genome visualizations were created by using BRIG (30) and edited by using Adobe Illustrator (Adobe Systems Inc., San Jose, CA, USA).

Antimicrobial Drug Susceptibility

MICs for penicillin, ampicillin, cefotaxime, levofloxacin, tetracycline, erythromycin, and clindamycin were determined by using the agar dilution method according to the Clinical and Laboratory Standards Institute (31). We also used SRST2 and a database of 1,913 variants of genes encoding antimicrobial drug resistance (https://github.com/katholt/srst2) to test for genetic determinants of antimicrobial drug resistance in genomes of serotype IV strains.

Isolate Source and Age Distribution of Patients

The 37 serotype IV GBS isolates used in this study represent 6.2% of 600 GBS strains isolated in the Toronto–Peel metropolitan area of Ontario during 2009–2012 (8). One isolate was collected in 2009, 15 isolates in 2010, 9 isolates in 2011, and 12 isolates in 2012. Most isolates were from invasive cases in adults (19–59 years of age, n = 14) or older adults (>60 years of age, n = 21) patients. We did not observe serotype IV GBS in neonates with early-onset disease (<7 days of age). However, 1 isolate was obtained from a child with late-onset disease (7–89 days of age), and an additional strain was isolated from a child 188 days of age. Most (34/37) serotype IV isolates were collected from blood; only 1 isolate was collected from soft tissue, and 2 isolates were collected from synovial fluid.

Genetic Diversity of Serotype IV GBS Strains

Figure 1

Figure 1. Inferred genetic relationships of invasive serotype IV group B Streptococcus (GBS), Toronto, Ontario, Canada. A) Neighbor-joining phylogenetic tree constructed by using the concatenated sequences of the 7 gene loci (sdhA, adhP,...

To assess genetic diversity of serotype IV GBS strains in our collection, we sequenced the genomes of all 37 strains and used SRST2 to derive MLST information directly from the short-read WGS data. We identified 6 STs representing 3 clonal complexes (CCs), which is indicative of a relatively high level of genetic diversity among our isolates. Two isolates were ST-291 included in hypervirulent CC17; 11 isolates were ST-452 (CC23); 2 isolates were ST-3 (CC1); and 1 isolate was a novel ST-682 included in CC1. Most (19/37) strains belonged to ST-459 (CC1). We also identified 2 strains of ST-196, a closely related single locus variant (sdhA gene) of ST-459. The 2 STs comprising most strains (ST-452 and ST-459) were also the most genetically distant, as shown by neighbor-joining phylogenetic analysis of concatenated sequences of alleles at all MLST loci (Figure 1, panel A).

To further understand genetic differences between these 2 STs, we arbitrarily chose 1 ST-452 strain (NGBS572) and 1 ST-459 strain (NGBS061) and sequenced their genomes to closure by using Pacific Biosciences SMRT technology. The genomes were circular chromosomes of 2,061,426 bp and 2,221,207 bp, respectively (Technical Appendix Figure 1). We did not find evidence of any associated plasmids in either isolate. However, we used an SMRT sequencing protocol with a DNA size-selection step for DNA fragments >20 kb, which might have removed small plasmids that could have been present. The difference in genome size between the strains was explained primarily by the smaller number and size of mobile genetic elements found in NGBS572 than in NGBS061 (Technical Appendix Figure 1). The core genomes of NGBS572 and NGBS061 (i.e., ≈1.9-Mbp portion of the genome lacking mobile genetic elements that is conserved in gene content between both strains) differed by 13,158 biallelic SNPs and 852 insertions or deletions (indels).

We next investigated genetic diversity within the ST-452 and ST-459 groups. Both groups were relatively homogeneous. On average, ST-452 strains differed from the ST-452 reference strain NGBS572 by only 45 SNPs and 10 indels in the core genome. On average, ST-459 strains had only 68 SNPs and 16 indels relative to the core genome of the ST-459 reference strain NGBS061. Neighbor-joining phylogenetic trees constructed by using concatenated SNPs in the core genome showed additional but limited diversity among isolates within each ST and some clustering within ST-459 (Figure 1, panels B, C).

Other STs in our collection had either 1 or 2 strains. Thus, we chose not to sequence these genomes to closure to assess genetic diversity but instead identified polymorphisms relative to the ST-452 and ST-459 reference strains. Both ST-3 strains showed a comparable number of polymorphisms against each reference strain (Technical Appendix Table 2). Similar results were observed for each of the ST-291 strains (i.e., both strains differed from the ST-452 and ST-459 reference strains by a comparable number of polymorphisms) (Technical Appendix Table 2). In contrast, the 2 ST-196 strains greatly differed between themselves in the number of polymorphisms separating them from the reference strains (Technical Appendix Table 2). One ST-196 strain, strain NGBS447, had 384 core SNPs and 60 core indels relative to the ST-459 reference strain, where the second ST-196 strain, NGBS472, had 1689 core SNPs and 124 core indels relative to the ST-459 reference strain. Further analysis discovered that the excess polymorphisms in NGBS472 were not evenly distributed across the genome. Instead, some discrete regions of the genome showed an overabundance of polymorphisms, which is suggestive of recombination.

Using BratNextGen, we precisely defined these regions of recombination (Technical Appendix Figure 2). When these regions of recombination were excluded from the analysis, strain NGBS447 had 107 SNPs and 16 indels relative to the ST-459 reference strain and NGBS472 had 281 SNPs and 128 indels relative to the same reference genome. Genome regions having undergone recombination in strain NGBS472 included genes encoding major virulence factors, such as the 2-component system csrRS (32) and pili (33). Pili in GBS are encoded by genes found in 2 distinct genome locations and organized in so-called pilus island 1 (PI-1) and PI-2. PI-2 has 2 variants (PI-2a or PI-2b) (33). One of the ST-196 strains had a pilus profile of PI-1 + PI-2a, and the other had a profile of PI-1 + PI-2b (Technical Appendix Figure 3). Except for these ST-196 strains, strains of the different STs had consistent pilus island profiles (Technical Appendix Figure 3).

Antimicrobial Drug Susceptibility of Serotype IV GBS Strains

Figure 2

Figure 2. Susceptibility to erythromycin, clindamycin, and tetracycline among serotype IV group B Streptococcus and sequence types (STs), Toronto, Ontario, Canada. All ST-459 strains and 1 ST-291 strain were resistant to erythromycin; we...

All serotype IV GBS strains were susceptible to all β-lactams, vancomycin, and levofloxacin. We observed resistance to erythromycin and clindamycin in 20/37 (54%) and 19/37 (51%) of the isolates, respectively; 23 (62%) of the isolates were resistant to tetracycline. The proportion of resistance and susceptibility, as well as 50% MICs and 90% MICs, are shown in Technical Appendix Table 3. All ST-459 isolates were resistant to clindamycin and erythromycin; ST-452 isolates were susceptible to these antimicrobial drugs (Figure 2; Technical Appendix Table 4). Strains of other STs were susceptible to these 2 drugs, except 1 of the ST-291 strains, which was resistant to erythromycin but susceptible to clindamycin (Figure 2; Technical Appendix Table 4). Macrolide resistance in ST-459 strains is probably attributable to the action of ermTR, a gene that was detected in all ST-459 strains (Technical Appendix Table 1). We also identified the ermT gene in the genome of the erythromycin-resistant ST-291 strain (Technical Appendix Table 1). Resistance to tetracycline correlated with presence of the tetM gene (Technical Appendix Table 1) and was observed among strains of all STs except ST-452. However, we identified 3 ST-459 strains that were susceptible to tetracycline (Figure 2; Technical Appendix Table 4). One of these had the tetM gene, but genome data showed a 7-nt deletion in the gene, which was predicted to result in early termination of translation. We did not detect tet genes in the other 2 ST-459 strains (Technical Appendix Table 1).