Genomic analysis

For GPS isolates, raw sequencing data and de novo assemblies were downloaded from http://www.pneumogen.net/gps/. GPS de novo assemblies were generated using a Velvet pipeline as previously described [31]. The remaining sequencing data were downloaded from the NCBI SRA database (See S1 File for a list of accession numbers). If only draft assemblies were available, paired-end 150 bp reads were generated using the BBMap’s RandomReads script. De novo assemblies for non-GPS isolates were generated using SPAdes v3.10 as previously described [18,32]. Assemblies were then annotated using Prokka v1.12 [33] and pangenome analysis was conducted using Roary [34]. Assemblies constructed using simulated reads were checked for variation in assembly metrics, annotation, and SNP diversity, in order to detect any evidence of sequencing and assembly errors. Variants for 13 polymorphic pneumococcal protein antigens—including pneumococcal surface proteins C and A (pspC and pspA)—were determined by mapping raw reads to an antigen variant database using SRST2 as previously described [27]. Quality filtered and trimmed reads were mapped to S. pneumoniae OXC141 (NCBI Reference Sequence: {"type":"entrez-nucleotide","attrs":{"text":"NC_017592","term_id":"387756560","term_text":"NC_017592"}}NC_017592), a Clade I, serotype 3 ST180/CC180 carriage isolate, using SMALT v0.7.6. SNPs were called using SAMtools v1.3.1 [35], and were filtered requiring QUAL>50, depth of coverage ≥5 and a minimum alternate allele frequency ≥0.75 [12,36]. Gubbins v2.1 was used to assess recombination, and coding sequences (CDSs) impacted by recombination blocks were annotated and plotted using Circos [37,38]. Last, we assessed non-synonymous mutations and recombination in the comCDE operon, which has previously been implicated in the low competence of CC180 compared to other pneumococcal lineages [18].