Population Genomic Molecular Epidemiological Study of Macrolide-Resistant Streptococcus pyogenes in Iceland, 1995 to 2016: Identification of a Large Clonal Population with a pbp2x Mutation Conferring Reduced In Vitro β-Lactam Susceptibility

Resistance to macrolide antibiotics is a global concern in the treatment of Streptococcus pyogenes (group A Streptococcus [GAS]) infections. In Iceland, since the detection of the first macrolide-resistant isolate in 1998, three epidemic waves of macrolide-resistant GAS infections have occurred, with peaks in 1999, 2004, and 2008. We conducted whole-genome sequencing of all 1,575 available GAS macrolide-resistant clinical isolates of all infection types collected at the national reference laboratory in Reykjavik, Iceland, from 1998 to 2016.

modifications (42). Overlap extension PCR was used to introduce the Met 593 Thr substitution into pbp2x of MGAS27213-L 601 P. Primers pbp2x-5=fwd (CAATTGTACAAAACCGTTACGATCCAAG) and pbp2x-5=rev (TAGTAACATACATCAAAAAGTCTGGTTTATC) were used to amplify the pbp2x 5= end. Primers pbp2x-T593-3=fwd (CTTTTTGATGTATGTTACTACGACTAAACCAC) and pbp2x-3=rev (GTGAATACATATCAGTATTTGTGGG TCATC) were used to amplify the pbp2x 3= end, introducing a single A to C nucleotide change in pbp2x codon 593. Primers pBBL740-fwd (GTAACGGTTTTGTACAATTGCTAGCGTAC) and pBBL740-rev (AAATACT GATATGTATTCACGAACGAAAATC) were used to amplify and linearize suicide plasmid pBBL740 by inside-out PCR. The pbp2x 5=-end and 3=-end amplicons were spliced with the linearized pBBL740 amplicon using a NEBuilder HiFi kit (New England Biolabs). The resultant spliced plasmid was transformed into parental strain MGAS27213-L 601 P, and single crossover transformants were selected by plating on Todd-Hewitt broth supplemented with yeast extract (THY) agar with 10 g/ml chloramphenicol. Transformants were screened by genomic DNA PCR amplification and Sanger sequencing using primers pbp2x-5=-fwd and pbp2x-seq (GATGTCTCACCAGGATTCTTTC). Ten confirmed single crossover transformants were pooled, expanded by outgrowth, and then passaged eight times on THY agar plates without chloramphenicol to promote double crossover resolution. Chloramphenicol-sensitive isolates were identified by duplicate plating and screened for the pbp2x-Thr 593 allelic exchange by PCR amplification and Sanger sequencing. The resultant candidate MGAS27213-L 601 P,M 593 T-derived strains were whole-genome sequenced to confirm the lack of spontaneous spurious mutations.
Data availability. Genomic sequencing data for all 1,515 macrolide/erythromycin-resistant isolates were deposited into the National Center for Biotechnology Information Sequence Read Archive under BioProject accession number PRJNA614628, and the assembled sequences for composite elements ⌽29854, ⌽29862, and ⌽29961 were deposited in GenBank under accession numbers MT311967, MT303952, and MT311968, respectively.
No macrolide-resistant GAS isolate was identified until July 1998. Following this, resistant isolates rapidly increased in proportion to a peak of 47.2% of isolates in 1999. Resistant isolates gradually declined in proportion to 5.0% in 2002. A second more modest increase in the proportion of resistant isolates peaked at 19.9% in 2004. A third peak of resistance rose to 44.9% of GAS isolates in 2008 (Fig. 1).
Whole-genome sequencing genetic characterization. To genetically characterize the cohort, all 1,575 available viable erythromycin-resistant GAS isolates were wholegenome sequenced to an average 214-fold depth of coverage (range, 18 to 1,859ϫ) using Illumina paired-end sequencing. Based on the sequence data, 60 of the isolates were excluded from the investigation for reasons such as the isolate not being S. pyogenes, being a duplicate, or being contaminated. The retained 1,515 erythromycinresistant S. pyogenes isolates and their epidemiological and genetic characteristics are listed in Table S1 in the supplemental material. Sequence reads for the isolates assembled on average into 67 contigs summing to 1.82 Mbp with a GϩC content of 38.4%, values which are consistent with closed genomes of S. pyogenes.
Analysis of the antibiotic resistance gene (ARG) content of the cohort identified 17 different ARGs that were present in 21 different combinations (Table 1). One or more ARG were detected in 1,471 (97.1%) of the isolates, and no macrolide-resistant gene was found in 44 isolates of 13 different emm types (Table S1). Previous publications have shown that emm types 4, 6, 12, and 75 are commonly associated with macrolide resistance. The most prevalent combination of macrolide resistance genes was mef(A) and msr(D), conferring the M resistance phenotype, which was found in 1,369 (90.4%) isolates. Virtually all isolates (1,359/1,369, 99.3%) of the three most prevalent emm types (4, 6, and 12) encode mef(A) and msr(D) (Fig. 3). Reciprocally, virtually all (1,359/1,369, 99.3%) isolates carrying mef(A) and msr(D) in the cohort are of emm types 4, 6, or 12.
The erm(B) gene was found in 69 (4.6%) isolates (43/69 ϭ emm types 11 and 75), and the erm(TR) gene was found in 30 (2.0%) isolates (20/30 ϭ emm77). Thus, 99 (6.5%) of the isolates have an erythromycin rRNA methylase gene conferring the MLS B phenotype. No isolate was found that encoded both a macrolide efflux and an erythromycin resistance methylase gene. Phylogenetic relationships. Within an emm type, the macrolide-resistant mef(A)and msr(D)-carrying isolates are closely genetically related, consistent with the isolates arising from clonal expansion of a recent common progenitor (Fig. 4). The mef(A)-and msr(D)-carrying emm4 (n ϭ 709), emm6 (n ϭ 322), and emm12 (n ϭ 327) isolates across the 1.7-Mbp core chromosome differed pairwise on average by only 11.3, 9.2, and 20.8 SNPs, respectively. For each of these emm types, the few erythromycin-resistant isolates that lacked any detectable ARGs were more genetically distant from the mef(A)-and msr(D)-carrying isolates and appear to represent infrequent sporadic spontaneous resistant mutants.
To investigate the context of the mef(A) and msr(D) genes within the isolate genomes, the de novo-assembled contigs of the emm type 4, 6, and 12 isolates were searched using BLASTn. For each of these emm types, the mef(A) and msr(D) genes were found adjacently encoded on transposon Tn1207.1 inserted into a phage forming a composite MGE like that first described for ⌽10394.4 of macrolide-resistant emm6 strain MGAS10394 (19,43). These elements were all found integrated at the same site in the genome disrupting the comEC gene. Full-length de novo assemblies of the mef(A)-and msr(D)-carrying MGEs were obtained from emm4 strain MGAS29862 ant (6) Accurately detecting SNPs in phages in the S. pyogenes genome is problematic, as most isolates are polylysogenic, which frequently causes cross-mapping of reads and erroneous overcalling of SNPs in phage. Despite this, mapping the wholegenome sequencing reads of the erythromycin-resistant isolates to the mef(A)-and msr(D)-carrying MGEs detected relatively few SNPs. The 709 emm4 isolates differed pairwise by 0.4 SNPs determined relative to ⌽29862, the 322 emm6 isolates by 4.6 SNPs relative to ⌽29961, and 326/328 (99%) emm12 isolates by 0.95 SNPs relative to ⌽29854. The finding that the isolates of the same emm type have mef(A)-and msr(D)-carrying composite MGEs that are nearly sequence invariant is again consistent with the macrolide-resistant isolates stemming from recent clonal expansions.
Comparison of S. pyogenes genomes has identified strain-to-strain differences in MGE content stemming from the dynamic gain and loss of ICEs and phages as the largest source of genetic diversity. As a third measure of relatedness, the MGE content of the isolates was assessed by sequence read mapping relative to a database of known S. pyogenes MGE-encoded integrases (n ϭ 31) and virulence factors (n ϭ 19). This comparison process generates a 50-allele present/absent genotype. Among the 322 erythromycin-resistant emm6 isolates, 309 (96%) have the same inferred MGE content (Table S2) composite MGEs. Our data demonstrate that the erythromycin-resistant emm type 4, 6, and 12 isolates are within their respective emm types, each closely genetically related, consistent with the isolates of each emm type stemming from recent clonal expansions.
Potential for altered beta-lactam antibiotic susceptibility. Recently, it was shown that many S. pyogenes clinical isolates with nonsynonymous (amino acid substituting) nucleotide changes in the penicillin-binding protein 2X gene (pbp2x) are associated with reduced susceptibility in vitro to one or more members of the beta-lactam family of antibiotics (12,13). Among the 1,515 macrolide-resistant isolates, 25 pbp2x alleles encoding 10 PBP2X variants were identified (Table S3). Although the pbp2x allele differed from one emm type to another, virtually no allelic variation in pbp2x was found within an emm type for the cohort. That is, in terms of pbp2x allele/PBP2X variants, 712/713 emm4 isolates have the same pbp2x allele/PBP2X variant, 324/324 emm6 isolates are the same, and 327/332 emm12 isolates are the same. The emm4 isolates have the consensus PBP2X wild-type (WT) sequence that is most prevalent among S. pyogenes isolates of multiple emm types (12,44,45). The PBP2X variant of all 324 emm6 isolates have three substitutions (Ile 502 Val, Pro 676 Ser, and Lys 708 Glu), and 327/332 emm12 isolates have a single substitution (Met 593 Thr) relative to the PBP2X WT 751-amino acid sequence. This lack of pbp2x sequence diversity is again consistent with emm type 4, 6, and 12 macrolide-resistant isolates stemming from recent clonal expansions. The susceptibility to penicillin G, ampicillin, and erythromycin of the three predominant PBP2X variants present in the emm 4, 6, and 12 isolates was tested for five isolates of each emm type ( Table 2). The isolates were selected to represent the temporal spread of each emm type corresponding with the three peaks of macrolide-resistant infections. All five emm4 isolates having the PBP2X WT variant were fully susceptible to the beta-lactam antibiotics penicillin G and ampicillin. Despite the five emm6 isolates having a PBP2X variant with three amino acid substitutions relative to the PBP2X WT, they were also fully susceptible to the beta-lactam antibiotics. In contrast, all five emm12 isolates with a Met 593 Thr substitution PBP2X variant had approximately 2-fold increased MICs for both penicillin G and ampicillin. To unambiguously determine if the PBP2X Met 593 Thr substitution alters beta-lactam susceptibility, we constructed an isogenic PBP2X Thr 593 substitution derivative using the parental strain MGAS27213-PBP2X-L 601 P (12). Importantly, whole-genome sequencing confirmed that the constructed derivative strain, MGAS27213-PBP2X-L 601 P,M 593 T, only differs from the parent strain by a single nucleotide change in codon 593 (ATG to CTG) of pbp2x. As anticipated, the parental strain had fully susceptible PBP2X WT penicillin G and ampicillin MIC levels. In contrast, the isogenic PBP2X Met 593 Thr derivative had 2-fold increased MICs ( Table 2). All 15 of the emm4, emm6, and emm12 isolates encoding mef(A) and msr(D) were erythromycin resistant, and both the parental and PBP2X Met 593 Thr derivative strains were erythromycin susceptible.

DISCUSSION
Macrolide-resistant GAS first appeared in Iceland in 1998 and has for most years since been relatively rare, with a yearly incidence typically below 5%. This contrasts with three rapid increases reaching peaks in 1999 (47.2%), 2004 (19.9%), and 2008 (44.9%). These peaks suggested clonal epidemics, now confirmed in this study. The first wave (1998 to 2001) was composed predominantly of emm4 (74%), the second (2004 to 2005) of emm12 (68%), and the third (2007 to 2008) of emm6 (91%). The peaks did not coincide with significant changes in either the type or amount of macrolide consumed over the year preceding the peaks. This suggests that GAS clones can spread rapidly in populations where herd immunity may be low to that particular clone, decline in numbers as herd immunity increases, and be replaced by another newly emerging clone. The results presented here should not be interpreted as macrolide consumption having had no effect on the three epidemic peaks or, alternatively, that significant changes in macrolide usage are not necessary for there to be significant changes in the prevalence of macrolide-resistant GAS infections and the predominant clone causing such infections. Similar results showing a lack of correspondence between macrolide consumption and occurrence of macrolide-resistant GAS isolates in Portugal have been reported, where a decline in erythromycin resistance was associated with the disappearance of isolates belonging to an emm3-ST315 lineage and yet accompanied by a high consumption of macrolides (26). The short time between the emergence in Iceland of the first erythromycin-resistant emm4 isolates in 1998 and the first emm12 isolates in 2000 (Fig. 1), with both contributing to the first macrolide-resistant epidemic wave (1998 to 2001), and the similarity in gene content and synteny of the mef(A)-and msr(D)-carrying elements in these emm types (Fig. S1), raises the possibility that the emergence events are directly related. That is, it is possible that the emm12 lineage progenitor arose through recent horizontal acquisition of the mef(A) and msr(D) composite MGE directly from an Icelandic emm4 donor. Alignment of emm4 ⌽29862 with emm12 ⌽29854 revealed a difference of 668 SNPs. The several hundred-fold greater numbers of SNPs identified for the mef(A) and msr(D) composite MGE inter-emm type versus intra-emm type (Ͻ1 SNP pairwise) is inconsistent with the hypothesis of a recent emm4 to emm12 transmission event and argues for emergence of macrolide-resistant emm4 and emm12 lineages into Iceland not being directly related.
Another possibility is that the emergence and expansion of the macrolide-resistant emm types contributing to the three epidemic waves that occurred in Iceland from 1998 to 2008 was driven by changes in antibiotic usage. Antimicrobial consumption of macrolides in Iceland was fairly constant from 1997 to 2009, with mean annual outpatient usage ranging from 1.85 defined daily doses per 1,000 inhabitants per day in 1998 to 1.25 in 2009 (46). Over this period, there was a gradual decrease in the use of short-acting macrolides (i.e., erythromycin) and a corresponding increase in the use of intermediate-acting (i.e., clarithromycin) and long-acting (i.e., azithromycin) macrolides, but no year-to-year dramatic shifts occurred (Fig. S2). The detected macrolideresistant S. pyogenes isolates increased 12.8-fold from 1998 to 1999 (from 34 to 434 isolates, nearly all emm4) and increased 4.3-fold from 2007 to 2008 (from 65 to 281 isolates, nearly all emm6). The lack of any substantial change in macrolide usage corresponding with these dramatic increases in detected macrolide-resistant isolates is inconsistent with the emergence and expansion being driven by antibiotic selective pressure.
Although beta-lactam susceptibility testing was not done for all of the Iceland macrolide-resistant emm12 isolates, it is likely that all 327 isolates that have the PBP2X Met 593 Thr amino acid substitution have reduced beta-lactam susceptibility with ϳ2fold increased MICs for penicillin G and ampicillin. This idea is supported by the findings that there were 2-fold increased penicillin G and ampicillin MICs for the five tested emm12 isolates temporally spread over the first (1999) and second (2004) peaks of macrolide-resistant infections and that the isogenic PBP2X Met 593 Thr substitution in the emm89 genetic background that demonstrated this single nonsynonymous A to C nucleotide change in pbp2x/single amino acid Met to Thr change in PBP2X is sufficient to increase penicillin G and ampicillin MICs 2-fold. It needs to be made clear that none of the isolates tested had MICs meeting the in vitro definition for penicillin or ampicillin resistance (EUCAST clinical breakpoint tables v10.0: benzylpenicillin resistant, Ͼ0.25 g/ml). It is noteworthy that the PBP2X Met 593 Thr substitution is (along with the PBP2X Pro 601 Leu [12,47]) only the second PBP2X amino acid change to be experimentally proven to reduce S. pyogenes beta-lactam susceptibility. A molecular understanding of how the PBP2X Met 593 Thr change alters beta-lactam susceptibility requires further investigation and would be aided by determination of an S. pyogenes PBP2X crystallographic structure.
One prevailing argument for why all bacteria have not evolved/acquired polymorphisms conferring resistance to any given antibiotic is that such resistance mutations result in organisms that are of reduced fitness in an environment that lacks that antibiotic (48)(49)(50)(51). In such an environment, bacteria with fitness-reducing resistance mutations are, over time, out-competed by more fit, susceptible bacteria and consequently become less prevalent/go extinct in the population. To our knowledge, the identification in Iceland of the closely genetically related 327 emm12 macrolideresistant isolates is the largest population identified of S. pyogenes clinical isolates with a PBP2X substitution conferring reduced beta-lactam susceptibility that are clearly recent clonally related descendants. Given that most of the isolates in this cohort come from pharyngitis patients, this indicates that S. pyogenes strains with some beta-lactam susceptibility-altering mutations in pbp2x are sufficiently fit to be readily transmitted and cause abundant pharyngitis. This finding contradicts the recent analysis of Hayes et al. (45) of PBPs among 9,667 GAS isolates, which found that "while heavy antibiotic selective pressure may select for mutations in the PBPs, there currently is no evidence of such mutations becoming fixed in the S. pyogenes population." The identification of a large number of naturally occurring GAS strains with mef(A) and msr(D) M phenotype macrolide resistance, in conjunction with a pbp2x nonsynonymous mutation producing a peptidoglycan synthesis transpeptidase (PBP2X) that confers reduced beta-lactam susceptibility, is concerning given that beta-lactams and macrolides are the first and second antibiotics of choice for treating S. pyogenes infections. Such strains are potentially stepping stones along the evolutionary path to true beta-lactam-resistant GAS. The use of either beta-lactam or macrolide antibiotics could provide the selective environment that favors the survival of such strains, increasing the opportunity for the incremental accumulation of additional resistanceenhancing polymorphisms. This emphasizes the need for beta-lactam susceptibility monitoring of GAS and the need for a vaccine to prevent GAS infections.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.6 MB. SUPPLEMENTAL FILE 2, XLSX file, 0.4 MB.