We validated isolates collected during 2019 and during January–February 2021. We also included a subset of isolates from 1986–2018 (S. pneumoniae) and 2000–2018 (GAS and GBS) because phenotypic testing showed those isolates were nonsusceptible to >1 antimicrobial drug. AIP had determined antimicrobial drug MICs, S. pneumoniae serotypes, and GAS emm types previously by using phenotypic microbiologic methods. We cultured isolates according to a previously established protocol (Appendix). We extracted genomic DNA and prepared DNA libraries by using Nextera DNA Flex with 96 dual indices (Illumina, https://www.illumina.com). We pooled libraries and performed WGS by using a MiSeq instrument and MiSeq v2 500 cycle reagent kit (Illumina). We used bioinformatic pipelines developed and validated by the CDC Streptococcus Laboratory (https://github.com/BenJamesMetcalf) (Appendix) (5–7).

We compared MICs from phenotypic testing with MICs predicted by WGS. We considered results to be preliminarily discordant when MICs between phenotypic and WGS methods differed by >2 dilutions. We retested isolates with discordant results by using Etest (bioMérieux, https://www.biomerieux.com), BD BBL Sensi-Disc for the D-Zone (BD, https://www.bd.com), or other disk diffusion methods (BD). We confirmed results were discordant when the original phenotypic result agreed with follow-up testing but continued to differ from the WGS prediction by >2 dilutions.

For GAS isolates, we compared emm type results from Sanger sequencing with WGS assignments. We extracted DNA again from isolates that had preliminary discordant results and tested all 3 extracts (original extract, extract for WGS, and reextraction) by using Sanger sequencing. We confirmed those results were discordant if the Sanger sequencing results continued to differ from the WGS assignment.

For S. pneumoniae isolates, we compared phenotypic serotype results with WGS assignments. We tested those isolates with preliminary discordant results by using the Immulex Pneumotest (SSI Diagnostica A/S, https://ssidiagnostica.com) with Quellung reaction confirmation. We confirmed results were discordant when the follow-up testing agreed with the original serotype result and continued to differ from the WGS assignment.

We tested 130 GBS (82 from 2019/2021, 48 historical), 217 GAS (169 from 2019/2021, 48 historical), and 293 S. pneumoniae (203 from 2019/2021, 90 historical) isolates. Initial comparisons showed a preliminary concordance of 99.4% (891/896) for GBS, 99.2% (1,288/1,298) for GAS, and 98.7% (3,470/3,517) for S. pneumoniae. Follow-up testing confirmed the WGS prediction for 4/5 (80%) GBS isolates, 10/10 (100%) GAS isolates, and 46/47 (97.9%) S. pneumoniae isolates. Final concordance was 99.9% (895/896) for GBS, 100% (1,298/1,298) for GAS, and 99.98% (3,516/3,517) for S. pneumoniae.

Initial phenotypic analysis showed 192 nonsusceptible results for GBS isolates (Table 1). Two GBS isolates exhibiting high (MIC ≥8 μg/mL) levofloxacin resistance contained amino acid substitutions in the GyrA subunit of DNA gyrase (S81L) and the ParC subunit of topoisomerase IV, of which 1 had S79F and 1 had S79Y. Tetracycline nonsusceptibility was most often caused by the tetM determinant (n = 37); the tetO determinant accounted for the other 7 instances. The 5 preliminary discordant results for GBS occurred with erythromycin (n = 2) and clindamycin (n = 3) (Table 2); follow-up testing confirmed the WGS prediction for 4/5 (80%) isolates. Most combined erythromycin and clindamycin nonsusceptibility was associated with the presence of the 23S rRNA methylase genes, ermTR (n = 42) or ermB (n = 19). All ermB-positive isolates were constitutively clindamycin resistant, whereas 12/42 (28.6%) ermTR-positive isolates were inducibly clindamycin resistant. Erythromycin resistance and clindamycin susceptibility (M phenotype) were detected in 13 isolates with the mef-positive genotype. Two isolates contained the lsaC gene, which confers resistance to clindamycin. One GBS isolate with a discordant result contained the lsaC gene but was phenotypically sensitive to clindamycin. Three isolates constitutively resistant to erythromycin and clindamycin contained multiple resistance mechanisms, 1 each of ermB plus lsaC, ermB plus mef, and lsaC plus mef.

Initial phenotypic analysis showed 190 nonsusceptible results for GAS isolates (Table 1). Three isolates with the S81F substitution in ParC had intermediate levofloxacin resistance; 1 isolate with high (MIC ≥8 μg/mL) levofloxacin resistance contained amino acid substitutions in both GyrA (S81Y) and ParC (D85N) protein subunits. We identified preliminary discordant GAS results when comparing tetracycline (n = 1), clindamycin (n = 8), and erythromycin (n = 1) (Table 2); follow-up testing confirmed the WGS predictions for all 10 of those isolates. All 32 isolates nonsusceptible to tetracycline contained the tetM determinant. Most combined erythromycin and clindamycin nonsusceptibility was associated with the presence of ermTR (n = 71), ermB (n = 5), or ermT (n = 1) gene determinants. All ermB-positive and ermT-positive isolates were constitutively clindamycin resistant, whereas 29/71 (40.8%) ermTR-positive isolates were inducibly clindamycin resistant. Three isolates constitutively resistant to erythromycin and clindamycin contained >1 resistance mechanism, ermT plus ermTR (n = 2) and ermTR plus lsaC (n = 1).

We completed emm typing by Sanger sequencing for 201 GAS isolates, identifying 37 emm types. The initial comparison between Sanger sequencing and WGS showed 199/201 (99.0%) emm type concordance. Sanger sequencing of a third DNA extraction from both isolates confirmed 100% WGS concordance.

Initial phenotypic analysis showed 467 nonsusceptible results for S. pneumoniae (Table 3). At initial comparison, 13 isolates were nonsusceptible to quinupristin/dalfopristin and 24 isolates were nonsusceptible to rifampin; follow-up testing confirmed the WGS prediction for all 37 isolates (Table 2). One isolate exhibited high (MIC ≥8 μg/mL) fluoroquinolone resistance and had amino acid substitutions in both GyrA (S81F) and ParC (S79F) subunits. All 15 isolates containing the chloramphenicol acetyl transferase (cat) gene were resistant to chloramphenicol, and all 35 isolates nonsusceptible to tetracycline contained the tetM determinant. Three isolates without a WGS-identified tetracycline resistance mechanism were phenotypically nonsusceptible to tetracycline at initial testing; follow-up testing confirmed the WGS prediction for all 3 isolates.

We detected nonsusceptibility to trimethoprim/sulfamethoxazole in 111 S. pneumoniae isolates; 62 of those had an insertion of 1 or 2 codons within the folP gene, conferring intermediate drug resistance. The remaining 49 isolates were double mutants, leading to full resistance consisting of a folA gene mutation (I100L amino acid substitution) and insertion of 1 or 2 codons within folP. Most combined erythromycin and clindamycin nonsusceptibility was associated with the presence of ermB (n = 16); all of those were constitutively clindamycin resistant. An additional 3 isolates were ermB positive, phenotypically erythromycin nonsusceptible, but were not tested phenotypically for clindamycin nonsusceptibility. Ten isolates constitutively resistant to erythromycin and clindamycin were ermB and mef positive; 41 isolates with the M phenotype were mef positive. Two isolates were initially discordant for erythromycin sensitivity; 1 was mef positive but phenotypically sensitive, and 1 did not contain a resistance mechanism but was phenotypically nonsusceptible. Follow-up testing confirmed the WGS predication for both isolates.

Most S. pneumoniae isolates had MIC predictions related to penicillin-binding protein (PBP) gene types that indicated sensitivity to ceftriaxone (n = 271), meropenem (n = 247), penicillin (n = 218), and cefotaxime (n = 38) (Table 3). Penicillin and cefotaxime MIC predictions were concordant with phenotypic testing for all S. pneumoniae isolates. One initial discordant result for meropenem and 3 initial discordant results for ceftriaxone were observed; all 4 isolates were phenotypically sensitive, but WGS predicted nonsusceptibility (Table 4). Follow-up testing confirmed the WGS prediction for 3/4 (75%) isolates; 1 isolate was predicted to be nonsusceptible to ceftriaxone but was sensitive according to both the initial and additional phenotypic testing.

Quellung serotyping was completed for 258 S. pneumoniae isolates, which comprised 30 serotypes. During the initial comparison, 2 discordant results were observed, likely related to an isolate mixup during phenotypic testing; additional testing confirmed the WGS-assigned result for both, indicating 100% concordance.

Despite data published by the CDC’s Streptococcus Laboratory supporting the accuracy of WGS-based antimicrobial drug susceptibility predictions (5–7), some collaborators have not pursued WGS workflows because they are uncertain those predictions are accurate (K.M. Miernyk, unpub. data). The antimicrobial susceptibility data shown here provide further evidence to address those concerns. For the 3 Streptococcus spp., we found 99.96% (5,709/5,711) concordance between phenotypic testing and genomic predictions. In addition, the WGS predictions were more accurate than phenotypic testing. Of 62 initial discordant results, WGS was confirmed to be correct for 60 (97%) of those.

Antimicrobial drug susceptibility data are needed to inform patient treatment and develop population treatment guidelines. IPD disproportionately impacts Alaska Native persons. AIP’s IBDS indicated the IPD rates associated with serotypes not targeted by the licensed 13-valent pneumococcal conjugate vaccine (PCV13) were 27.2/100,000 Alaska Native persons and 6.7/100,000 non–Alaska Native persons during April 2010–December 2013 (4). Similarly, the percentage of persons carrying non-PCV13 serotype S. pneumoniae in their nasopharynx that was nonsusceptible to erythromycin or penicillin increased significantly from 16.8% (n = 709) during 2008–2011 to 26.5% (n = 1,466) during 2012–2015 (11). This finding suggests Alaska Native persons might be at increased risk for IPD caused by S. pneumoniae that are resistant to commonly used antimicrobial drugs. Capsular S. pneumoniae serotyping data are necessary to evaluate vaccines for IPD prevention. Finally, GAS emm typing data predict the M protein serotype (12), a potential target for vaccines to prevent invasive GAS infections. Traditional phenotypic methods are time consuming and require many different specialized technical skills, reagents, and consumables. The WGS workflow described here provides those data for 48 samples simultaneously by using 1 method. In addition, as the COVID-19 pandemic revealed, supply chains can be unreliable. WGS provides data in a single workflow, enabling a more streamlined process with fewer consumables and reagents.

AIP has not previously had the capacity to consistently characterize resistance mechanisms in any Streptococcus spp. bacteria. To better elucidate changes in macrolide resistance after PCV13 introduction, we briefly used PCR to characterize ermB and mef macrolide resistance mechanisms in S. pneumoniae (13). However, we have not investigated those mechanisms in GAS or GBS collected from persons in Alaska, which could be of particular importance for GAS. Candidate GAS vaccines target the M protein (14), and >275 known M types exist. Therefore, vaccine pressure on targeted M types could affect circulating strains of GAS, which has been observed for S. pneumoniae after PCV13 introduction (15). Macrolide nonsusceptibility is not uncommon for GAS, and it will be critical to understand whether changes in nonsusceptibility after vaccine introduction are caused by expansion of existing strains, by introduction of new strains, or by some other mechanism.

In conclusion, our antimicrobial susceptibility, serotype, and emm type validation data confirm the accuracy of WGS-based predictions for GAS, GBS, and S. pneumoniae when performed at AIP. The single WGS workflow is more efficient than multiple workflows needed for phenotypic testing. WGS pipelines identify previously unknown genotypic mechanisms for nonsusceptibility of Streptococcus spp. isolates collected in Alaska and provide additional data, such as GBS serotypes and multilocus sequence types. WGS also provides a genetic sequence for every isolate, which is available for future investigations. In 2022, AIP transitioned all IBDS workflows for Streptococcus spp. to WGS, and we continue to improve those processes. We have decreased cost by sequencing more extracts on a flow cell and have been increasing local analysis capabilities and data storage for more rapid and local pathogen detection. We also perform biannual phenotypic testing on a subset of isolates to monitor for new resistance genes. Future work includes validating a WGS workflow for GAS that can be used in remote field settings, enabling AIP to provide more rapid outbreak response.