Surveillance Data Collection

We collected the annual number of GCGS and SDSE bacteremia cases and hospital admissions during January 2011–December 2020 in 6 hospitals in the Kyoto-Shiga region of Japan. Four hospitals were in Kyoto, Kyoto University Hospital (KUHP), Kyoto City Hospital (KCH), Kyoto Katsura Hospital (KKH), and Kyoto Min-iren Chuo Hospital (KMCH); 2 hospitals were Shiga, Japanese Red Cross Otsu Hospital (JRCOH) and Shiga General Hospital (SGH) (Figure 1). Among the hospitals, 3 changed their methods of identifying β-hemolytic streptococci during the study period, from the biochemical and Lancefield grouping methods to matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry: KUHP in changed in 2016, KCH in 2015, and KMCH in 2019. The other 3 hospitals used the biochemical and Lancefield grouping methods throughout the study period; therefore, we collected the combined incidences of GCGS and SDSE bacteremia.

Bacterial Isolates

We conducted a microbiological and genomic study using the preserved isolates collected from GCGS and SDSE isolates from the bacteremia cases during 2005–2021 in 3 hospitals: KUHP, KKH, and JRCOH. The bacterial isolates from patients’ blood samples were identified as GCGS when isolates showed large colony size (>0.5 mm), β-hemolysis on 5% sheep blood agar plate after overnight culture, and Lancefield grouping C or G. In some hospitals, GCGS isolates were then identified as SDSE by using phenotypic methods such as API 20 Strep (bioMérieux), BBL Crystal (Becton Dickinson Microbiology Systems), Microscan Walkaway System (Beckman Coulter, Inc.), or MALDI-TOF mass spectrometry. The study comprised 146 SDSE isolates collected during 2005–2021 (Appendix 1).

Lancefield grouping was performed using the Prolex Streptococcal Grouping Latex Kit (Pro Lab Diagnostics Inc.). Antimicrobial susceptibility testing was conducted by using the MicroScan WalkAway System according to the manufacturers’ instructions or by broth microdilution by using customized dry plates (Eiken Chemical Co., Ltd). Susceptibility results were interpreted following Clinical Laboratory Standards Institute recommendations (16). Multidrug-resistant (MDR) was defined as nonsusceptibility to >2 antimicrobial agents; for this study, we refer to MDR as nonsusceptibility to erythromycin, minocycline, and clindamycin.

Clinical Data Collection

We retrospectively reviewed the charts of patients with preserved isolates. Using clinical records from each episode, we collected the age, sex, dates of illness onset and sample collection, underlying illnesses, receipt of chemotherapy or immunosuppressive agents, signs, symptoms, laboratory findings at the first visit, clinical source of the bacteremia, infection type (community-acquired or nosocomial: nosocomial infections were defined as clinical signs and symptoms occurring >48 hours after admission), infection severity parameters, outcomes, and relapses.

Whole-Genome Sequencing and Bioinformatic Analysis

We extracted genomic DNA from isolates by using the MagNA Pure 96 DNA and Viral NA Small Volume Kit on a MagNA Pure 96 Instrument (Roche). We prepared a short-read library by using the Illumina DNA Prep Kit (Illumina Inc.) and used NextSeq 1000/2000 P1 Reagent Kit version 3 on the NextSeq platform (Illumina) to generate paired-end reads. We used the Rapid Barcoding Kit (Oxford Nanopore Technologies) to prepare a long-read library from unsheared genomic DNA, which we sequenced by using an R9.4.1 Flow Cell on a MinION device (Oxford Nanopore Technologies). We assembled draft genome sequences from the short-reads and complete genome sequences from the long-reads by using Unicycler pipeline version 0.4.8 with default parameters (17). We annotated assembled genomes by using Prokka version 1.14.6 (17) with default settings and used annotated genomes for downstream analysis. We confirmed isolates as SDSE by comparing 16S rRNA gene similarity with the subsequent contigs.

We identified multilocus sequence types (MLSTs) from the assembled contigs by using a BLASTn search (https://blast.ncbi.nlm.nih.gov) against the S. dysgalactiae MLST database in pubMLST (https://pubmlst.org/organisms/streptococcus-dysgalactiae) and then assigned the MLST to a clonal complex (CC) by using global optimal eBURST (goeBURST). We assigned CCs by a single-locus variation from a founding sequence type (ST). We determined emm types via BLAST search against the emm sequence database curated by the Centers for Disease Control and Prevention (https://www2.cdc.gov/vaccines/biotech/strepblast.asp). We detected AMR genes by using abricate (https://github.com/tseemann/abricate) and the Comprehensive Antibiotic Resistance Database (https://card.mcmaster.ca) (18). We predicted virulence factors of SDSE genomes by using BLASTP against the Virulence Factor Database (http://www.mgc.ac.cn/VFs) (19) set A, a core dataset that covers genes associated with experimentally verified virulence factors, and a cutoff E-value of 1E−5.

We submitted the raw Illumina and MinION read data to the International Nucleotide Sequence Database Collaboration (https://www.insdc.org) under BioProject accession no. PRJDB12179. We summarized the assembly statistics and general genomic information (Appendix 1).

Phylogenetic Analyses

We used Roary version 3.13.0 (20) and default parameters to identify the core genomes of the 146 SDSE isolates. We generated a maximum-likelihood phylogenetic tree from the core genome alignment by using RAxML-NG version 1.0.3 and a general time-reversible plus gamma distribution DNA substitution model with 100 bootstrap replicates (21). We used FigTree version 1.4.4 (https://github.com/rambaut/figtree) and iTOL version 6 (22) to visualize phylogenetic trees.

Tn916-Like Integrative and Conjugative Element Analysis

We extracted sequences of the Tn916-like integrative and conjugative elements (ICEs) from the assembled contigs by using a BLASTn search (23) against the Enterococcus faecalis Tn916 reference sequence (GenBank accession no. U09422.1). We manually analyzed the region structures by using the Artemis Comparison Tool (24).

Statistical Analyses

We used χ² and Fisher exact tests to compare categorical variables, as appropriate, and used Mann-Whitney U tests to analyze continuous, nonparametric data. To assess the time trends in GCGS/SDSE bacteremia incidence during 2011–2020, we calculated the incidence rate ratio and significance of the data by comparing the later cohort (2016–2020) with the early cohort (2011–2015) and considered a 2-sided p<0.05 statistically significant. We performed all statistical analyses using R version 3.6.0 (The R Foundation for Statistical Computing, https://www.r-project.org).

The Ethics Committee of Kyoto University Graduate School and Faculty of Medicine approved the study (approval no. R3240). The study was conducted in accordance with the principles expressed in the Declaration of Helsinki.

Temporal Trends in SDSE Bacteremia

The Kyoto-Shiga Region is a metropolitan area in Japan with a population of ≈2.5 million within a range of 70 km. We collected data on 398 episodes of bacteremia caused by GCGS and SDSE in 6 hospitals in the Kyoto-Shiga region during 2011–2020 (Figure 1). We investigated the temporal trend in combined incidence of GCGS and SDSE bacteremia per 10,000 admissions (Figure 2). The incidence of GCGS and SDSE bacteremia increased from 2.77/10,000 admissions in 2011 to 8.77/10,000 admissions in 2020. We noted a statistically significant increase in GCGS and SDSE bacteremia incidence between the first and last 5-year periods (incidence rate ratio = 2.03, p = 0.03). During 2018–2020, the average annual incidence of bacteremia reached 13.3 cases/10,000 admissions in KKH and 19.7 cases/10,000 admissions in KMCH.

Clinical Characteristics of Patients with SDSE Bacteremia

We investigated 146 SDSE bacteremia episodes by using preserved isolates from 133 patients at KUHP, KKH, and JRCOH. The study included 118 isolates collected from KUHP, KKH, and JRCOH, representing 50.6% (118/233) of all cases from the 3 hospitals during 2011–2020. KUHP preserved consecutive isolates throughout the study period (2005–2021), but JRCOH and KKH did not. In detail, JRCOH preserved consecutive isolates during 2015–2021 but only preserved 1 isolate during 2013–2014. Similarly, KKH preserved 8 of 20 isolates collected in 2021.

Among 133 patients with SDSE bacteremia, most were 80–89 years of age (Table 1). Many patients had underlying conditions, predominantly cerebrovascular disease (28.6%), solid organ cancer (27.1%), and dementia (21.3%). We noted recurrent SDSE bacteremia in 9 patients who had a total of 13 episodes (Appendix 2 Table 1).

Of the 146 episodes in 133 patients, the most common sources of bacteremia were cellulitis (50.7%) or primary bacteremia (18.5%) (Table 2). Of 142 episodes for which detailed data were available, 7 (4.9%) involved STSS and 11 (7.7%) required vasopressor use; of 143 patients, 10 (7.0%) died during hospitalization.

Of 146 SDSE bacteremia episodes, most (136, 93.2%) were community-acquired; 10 (6.8%) were nosocomial (Appendix 2 Table 2). Of the 10 patients with nosocomial episodes, 7 had active malignancy, 5 received chemotherapy, and 3 had febrile neutropenia. Four patients had multiple pathogen infections: 2 episodes were primary bacteremia occurring in patients with neutropenia, the other 2 catheter-related bloodstream infections. In the neutropenic patients with primary bacteremia, Escherichia coli or viridans streptococci were isolated from each patient. In the patients with catheter-related bloodstream infections, Staphylococcus aureus or S. epidermidis was isolated from each patient. Two nosocomial-onset patients died while hospitalized.

AMR and Virulence Profiles of SDSE Isolates

Of the 146 SDSE isolates identified, 138 were classified as Lancefield group G, 6 as group C, 1 as group A, and 1 as untypeable; 25 STs were represented, and a goeBURST analysis further classified the 25 STs into 7 CCs and 9 singletons. CC17 (34.2%) was the most prevalent, followed by with CC25 (28.1%), ST525 (11.0%), and CC29 (8.9%). Prevalence of CC17, CC25, and CC29 did not change substantially over time, but after ST525 was first isolated in 2016, its prevalence rate increased significantly, from 1.7% (1/58 isolates) during 2005–2017 to 20.0% (15/88 isolates) during 2018–2020 (p = 0.001) (Table 3; Appendix 2 Figure). Conversely, the prevalence of nonmajor CC and ST isolates (other than CC17, CC25, CC29, and ST525) decreased significantly from 27.6% (16/58 isolates) during 2005–2017 to 11.4% (10/88 isolates) during 2018–2020 (p = 0.015) (Appendix 2 Figure). Patients harboring ST525 strains were significantly older than were patients harboring non-ST525 strains (median age 87.0 years vs. 80.5 years; p = 0.007).

We classified emm-type isolates into 19 groups (Appendix 2 Table 3). The most prevalent emm type was stG6792 (28.1%), which was predominant in CC17, followed by stG245 (20.0%), found only in CC25; stG485 (9.6%), found in CC29 and CC128; and stG840 (11.0%) found only in ST525.

We assessed AMR rates according to CC and ST (Table 4). No isolates showed nonsusceptibility to penicillin G, ceftriaxone, or meropenem. Nonsusceptibility rates were 31.5% to erythromycin, 30.8% to minocycline, and 25.5% to clindamycin; prevalence of isolates nonsusceptible to those 3 antimicrobial agents increased significantly, from 8.6% (5/58 isolates) during 2005–2017 to 21.6% (19/88 isolates) during 2018–2021 (p = 0.042). Of note, all 16 ST525 isolates were MDR and were resistant to those 3 antimicrobial agents. For CC25, nonsusceptibility rates were 29.3% for erythromycin, 29.3% for clindamycin, and 39.0% for minocycline. Nonsusceptibility rates were much lower for CC17 and CC29: 18.0% for CC17 and 7.7% for CC29 for erythromycin; 4.0% for CC17 and 7.7% for CC29 for minocycline; and 16.0% for CC17 and 7.7% for CC29 for clindamycin. Further AMR gene analysis showed that ermB was the most prevalent (20.5% of all isolates) resistance gene for macrolide resistance and was predominantly found in CC25 and ST525 isolates, and ermA (8.9%) was mainly found in CC17 isolates (Table 5). In addition, tetM was the predominant gene for tetracycline resistance (31.5%), and 23 of 24 MDR isolates possessed both tetM and ermB.

We assessed prevalence of the virulence factor–associated genes according to CC (Appendix 2 Table 4). All SDSE isolates contained virulence factor–associated genes, including fbp54, lmb, scpA/scpB, slo, ska, and hasC. Prevalence of other virulence-related genes, such as sda, speG, and pilus island 1–associated genes, differed between CCs. speG was significantly associated with STSS (p = 0.026). However, other virulence genes did not have statistically significant correlation with severe disease (defined as STSS, necrotizing fasciitis, or need for vasopressors) or in-hospital death (Appendix 2 Table 4).

Phylogenic Analysis of SDSE Isolates

We visualized a phylogenetic tree of the SDSE isolates based on the 39,258 concatenated single nucleotide polymorphisms of the core genome (Figure 3). Phylogenetic analysis of 146 SDSE strains from Japan revealed no geographic relationships among specific clades or relationships with the year of isolation. However, the CCs reflected phylogenetic relationships, and the phylogenetic tree showed the clonal accumulation of clindamycin, minocycline, and erythromycin resistance in the ST525 clade and a part of the CC25 clade. The isolates possessed both tetM and ermB AMR genes.

Structure and Genetic Characterization of Tn916-Like ICEs on SDSE Genome

To investigate the genetic elements contributing to the clonal accumulation of tetM and ermB, we performed further sequence analysis by using the complete genome sequences. The tetM and ermB genes were located on the Tn916-like ICEs inserted into the chromosomal backbone. The genomes of 35 (24.0%) of the 146 strains harbored Tn916-like ICEs (Table 5). The dendrogram based on prevalence of coding sequences in the elements showed that SDSE-associated Tn916-like ICEs were mainly divided into 2 distinct groups (Figure 4). Group A included all 16 ST525 isolates and 1 ST17 (CC17) isolate (Figure 5); group B included all 17 ST127 (CC25) isolates and 1 ST17 isolate (Figures 5, 6).

We conducted a further detailed structural and sequence comparison of the Tn916-like ICEs. In group B, ST17_B had an identical structure to the original Tn916, and a hypothetical protein was inserted on the intergenic region between open reading frame (ORF) 20 and ORF21 in all ST127 isolates (Figures 5, 6). In group A, ST17_A had an identical structure to Tn3872, which carries Tn917 with ermB inserted in the form of a splitting ORF9 on Tn916, and all ST525 isolates had an additional hypothetical protein inserted similar to those in the ST127 isolates (Figures 5, 6). In the ST127 isolates with ermB on a sequence region other than Tn916-like ICE, ermB was sandwiched between 2 IS1216E genes and inserted into a specific site (Figure 5). The Tn916-like ICEs were clonally distributed in a phylogenic tree; each type of Tn916-like ICE showed the distribution pattern along with phylogenetic relationships and specific insertion sites on the genome (Figure 6). To further investigate Tn3872-like ICE in ST525, we performed a BLASTn search against the National Center for Biotechnology Information database using parameters of >90% query coverage and >95% identification. We found almost identical genetic structure with highly conserved sequences in other streptococci and nonstreptococci bacteria (Appendix 2 Table 5).