Genomic Characterization of Clostridium perfringens Bacteremia Isolates at a Tertiary Care Center in Japan

1  Corrections to the published abstract

A subset of the bloodstream isolates had been analysed with an incorrect input file, and two datasets had been mixed. All 218 genomes were re-analysed in a single uniform pipeline. The direction of every principal finding is unchanged.

Denominators are those of the frozen re-analysis, as on the poster: bloodstream 25, human non-bloodstream 29. nagL and pfoA carry a Benjamini–Hochberg q across the 18 virulence genes tested; tetA(P) is a resistance gene outside that family, so its Fisher p is uncorrected. † Used in the abstract's calculations but not printed there.

The abstract's tetA(P) figure of 8% was a transcribed odds ratio of 0.086. Its direction is not reversed — carriage is lower in the bloodstream group in both — but the effect collapses to an odds ratio of 0.68 and is not significant.

IDWeek permits errors in an accepted abstract to be indicated during the presentation (Abstract Changes / Edits). The abstract is frozen as submitted; this table is the correction.

2  Robustness checks

• ✓The bloodstream group can be defined either way. Whether the group is the 25 genomes (24 from this study plus one public bloodstream genome) or the 24 isolates of this study alone, the same seven genes reach q < 0.05 with identical carriage proportions. Only the q values shift.
• ✓Assembly quality is not the driver. Re-assembling the 24 isolates moved the contig-count median from 278 to 42 and changed 1 of 480 gene calls.
• ✓Annotation confounding was removed. The earlier input mixed two annotators along the bloodstream / comparison boundary. All 210 genomes in the pan-genome and phylogeny are now annotated with a single version of Prokka.
• ✓The quality gate does not touch the comparison. The 8 genomes excluded by CheckM2 are all animal- or food-source, so the bloodstream and human non-bloodstream groups are intact.
• ✓Redundancy was found by sequence, not by name. MD5 of sequence content caught duplicate genomes that name and accession matching had missed.
• ✓The core-gene count agrees with an external standard. An independently constructed cgMLST scheme for this species defines 1,431 core genes, bracketed by our 99% (1,215) and 95% (1,563) values. The abstract's 538 lies far outside every published figure.

3  Core-gene thresholds

The core-genome definition is not a knife edge. Relaxing the threshold from 100% to 80% of genomes moves the count from 822 to 2,057, and the 99% value used throughout the poster sits in the middle of that range.

Computed from the Roary gene_presence_absence.Rtab for the 210-genome set (23,512 genes total). The 99% column is the value reported on the poster.

4  Node support and core alignment

The poster prints Figure 1 as a cladogram without support values, so they are given here. The tree has 210 tips and 205 internal nodes carrying support.

154 of 205 internal nodes (75%) meet both conventional thresholds (SH-aLRT ≥ 80 and UFBoot ≥ 95). Branch support is therefore strong for three quarters of the tree; the poorly supported nodes lie within the densely sampled, shallow parts of phylogroup III, which is consistent with the reticulation that Figure 2 shows directly.

Alignment accounting

The poster states that the tree was built from 56,596 SNP sites in a 1,066,463 bp core alignment. IQ-TREE reports a total of 1,020,921 sites, because constant sites were supplied as counts of unambiguous A/C/G/T only. The difference is accounted for here so the two figures can be reconciled.

5  Gene tables

5.1  Virulence genes, bloodstream vs non-bloodstream

All 18 genes tested, not only the seven that reached significance. Genes detected in neither group are not listed and were not tested.

Fisher exact test, two-sided; Benjamini–Hochberg across all 18 genes. Odds ratios use a 0.5 continuity correction, so a zero cell still gives a finite value. Shaded rows are the seven with q < 0.05 reported on the poster.

5.2  Enteric toxin genes

netE, netF, tpeL and netB were found in no bloodstream genome and no human non-bloodstream genome, but readily in animal-source genomes. Because these genes are not uniformly represented across annotation sources, each row states which call it comes from, and both available calls are given where they differ.

netE, netF and tpeL are not represented in VFDB and were assessed from Bakta annotation; netB is reported on the poster from ABRicate/VFDB. The two sources disagree for netB: 51 animal genomes by VFDB against 23 by Bakta. For the four enteric-disease toxin genes the absence from the bloodstream and human groups is nonetheless unanimous across both sources — every cell is zero either way — so the poster's conclusion does not turn on this choice. cpe is listed for context and is the one gene here where the sources differ inside the bloodstream group (VFDB 1 of 25, Bakta 0 of 25; the poster reports the VFDB call throughout). Toxinotype F is the reported typing of the source studies, an independent assignment rather than a gene call, and is given for comparison. Denominators are the 218 genomes used for gene detection; the 8 CheckM2 failures (7 animal, 1 food) carry none of these genes, so including them changes no numerator.

5.3  Resistance genes by source

All 15 resistance genes detected across the 218 genomes. The poster has room for 10; the 5 omitted from it are lnu(D), mef(A), aadE, erm(A), erm(B), each found in one or two animal-source genomes and in no bloodstream isolate.

ABRicate against NCBI AMRFinderPlus, one uniform run over all 218 genomes, collapsed to presence/absence per genome (a gene may hit a genome more than once). optrA and fexA co-occur in a single animal genome (2C45, chicken caecum, China 2018, which also carries erm(A)); tet(M) is in one other (CP-41, peacock, China 1984). Both lie in phylogroup III, the phylogroup that contains every bloodstream isolate of this study. Tetracycline efflux genes are widespread across every source.

5.4  Toxinotype by source

Toxinotypes follow the scheme of Rood et al.⁶; those of the comparison genomes are as reported by Abdel-Glil et al.¹ Every genome carries an assignment, so the columns sum to the group sizes. These are reported typings, not gene calls; see 5.2 for the cpe detections, which give a slightly different count.

6  Isolation sources, as recorded

Every source string exactly as recorded by the reference studies, for all 193 comparison genomes. The poster collapses these into five groups; the original strings are given here because several carry anatomical or clinical detail that the grouping discards.

Every genome in the comparison set was matched to the reference study by accession or by strain name (193 of 193 matched). Four human genomes are recorded only as “Human” with no anatomical site; three of these carry a recorded enteric disease (food poisoning, sporadic diarrhoea, enteritis necroticans / Darmbrand). The one gall-bladder isolate of that collection, NCTC 2544 (ERR1456745), exists only as raw reads with no public assembly and could not be included. The bloodstream group is 25 genomes: 24 from this study and one public genome (2023/00056), which are not part of this table.

7  Analysis set

8  Software and data

Seven of the 218 genomes were annotated with Bakta v1.11.4 against the same reference database; annotation output was verified identical to v1.12.0 across 14,427 CDS (coordinates, protein sequences, gene symbols, products and cross-references all matched exactly).

The single public bloodstream genome is strain 2023/00056 (GCA_963424335.1, BioProject PRJEB65712, BioSample SAMEA114318975), a binary-enterotoxin-producing isolate from a fatal case of pneumonia and septic shock reported by Ben Saïd et al.³ It falls within phylogroup II, not phylogroup III.

The tables in this handout are generated from the frozen analysis by scripts/make_supplement_tables.py and are available as tab-separated files alongside this page, under data/.

References

1. Abdel-Glil MY et al. Sci Rep 2021;11:6756.
2. Abdel-Glil MY et al. Microbiol Spectr 2021;9:e00533-21.
3. Ben Saïd L et al. Microorganisms 2024;12:1095.
4. Okazaki A et al. Int J Infect Dis 2025;151:107358.
5. Kiu R et al. Nat Microbiol 2023;8:1160–75.
6. Rood JI et al. Anaerobe 2018;53:5–10.

Disclosures. S.H. has received honoraria from Denka, Eiken Chemical, KYORIN Pharmaceutical, Meiji Seika Pharma, MSD, Pfizer, Shionogi and Ushio. All other authors report nothing to disclose.

Contact. Aiko Okazaki, MD, PhD · Project Assistant Professor, UTOPIA Center, The University of Tokyo · okazaki-aiko321@g.ecc.u-tokyo.ac.jp

Affiliations. 1 UTOPIA, The University of Tokyo · 2 The University of Tokyo Hospital · 3 Juntendo University · 4 The University of Tokyo · 5 Toho University School of Medicine · 6 Institute of Science Tokyo