For PCR screening of the astA gene, 2,726 E. coli strains isolated from stool specimens of diarrheal children who visited clinics in Kagoshima, Japan, from 2013 to 2020 (referred to as Kagoshima strains; listed in Supplementary Table 1) were used. For the blastn search of astA, 9,065 finished E. coli genomes retrieved from the NCBI database (accessed on 10 May 2022; listed in Supplementary Table 2) were used.

Template genomic DNA for PCR screening and multi-locus sequence typing (MLST) analysis was prepared by the alkaline boiling method from a 1 ml culture grown at 37°C in Lysogeny broth (nacalai tesque). Genomic DNA for whole genome sequencing was purified from a 2 ml overnight culture using the NucleoBond HMW DNA (MACHEREY-NAGEL) according to the manufacturer's instructions.

Detection of the astA gene by PCR was performed using a primer pair (EAST1S (5′-GCCATCAACACAGTATATCC-3′) and EAST1AS (5′-GAGTGACGGCTTTGTAGTCC-3′) (Yatsuyanagi et al., 2002) and KAPATaq EXtra PCR Kit (NIPPON Genetics, Tokyo, Japan). Each reaction mixture (15 μl) contained 1 μl of template DNA, 4.5 μM of each primer and 0.3 U of polymerase. PCR was conducted with initial denaturation for 2 min at 94°C, followed by 30 cycles of 30 s at 94°C, 30 s at 55°C and 30 s at 72°C. PCR products were analyzed by agarose gel electrophoresis using 2 % Agarose S (Nippon Gene).

RAPD-PCR was performed as described by Pacheco et al. (1997) using P1252 (5′-GCGGAAATAG-3′), P1254 (5′-CCGCAGCCAA-3′), and P1290 (5′-GTGGATGCGA-3′) primers and the KAPATaq EXtra PCR Kit (NIPPON Genetics Co., Ltd.). Each reaction mixture (25 μl) contained 5 μl of 5x KAPATaq Extra buffer, 2.5 mM MgCl₂, 300 mM dNTPs, 1 μl of template DNA, 0.4 μM primer and 0.6 U of KAPATaq DNA polymerase. The PCR amplification steps employed were as follows: 4 cycles of 94°C for 5 min, 37°C for 5 min, and 72°C for 5 min, followed by 30 cycles of 94°C for 1 min, 37°C for 1 min and 72°C for 2 min and a final extension step at 72°C for 10 min. After PCR amplification, 10 μl of each PCR product was analyzed by agarose gel electrophoresis using 1.2% agarose S (Nippon Gene).

MLST was performed by PCR amplification and sequencing of seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA) as previously described (Ooka et al., 2012). Alleles of each gene and the sequence types (STs) and clonal complexes (CCs) were assigned using the PubMLST (https://pubmlst.org/organisms/escherichia-spp) (Jolley et al., 2018) and the EnteroBase E. coli/Shigella MLST website (https://enterobase.warwick.ac.uk/) (Zhou et al., 2015), respectively.

Short-read sequencing libraries were prepared using the Nextera XT DNA Sample Prep Kit (Illumina) to obtain paired-end sequences (300 bp × 2) on the Illumina MiSeq platform. Long-read sequencing libraries were prepared using a Rapid Barcoding Kit (Oxford Nanopore Technologies) and sequenced using an R9.4.1 flow cell. After base-calling and demultiplexed using Guppy GPU v3.4.5 (Oxford Nanopore Technologies), long raw reads were filtered based on quality cut-off score of 10, and minimum length of 2,000 bp and trimmed 100 nucleotides from the start of the read using NanoFilt. A hybrid assembly was performed using microPIPE (Murigneux et al., 2021) with long and short reads with default parameters. The complete and draft genome sequences obtained in this study have been deposited in the GenBank/EMBL/DDBJ database under Bioproject no. PRJDB18009 (see Supplementary Table 2 for the list of sequenced strains and their sequencing statuses and accession numbers).

The astA gene in the E. coli genomes was identified by blastn search using the known sequences of astA (V0–V7) (Supplementary Table 3) as queries with cutoff values of 90% nucleotide sequence identity and 60% length match. The BLAST+ source code was downloaded from the NCBI website (https://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/). New astA variants were defined if an identified astA gene showed one or more nucleotide sequence differences compared with all of the eight known sequences.

The nucleotide sequences of astA genes and their 1,000-bp upstream and downstream flanking regions were extracted from the genome sequences listed in Supplementary Table 2. The multiple nucleotide sequence alignments of astA genes and their 1,000 bp upstream and downstream regions and the amino-acid sequence alignment of astA gene products (EAST1 peptides) were constructed using the ClustalW function of MEGAX (Kumar et al., 2018) with default parameters. Multiple sequence alignments were visualized using JalView (V2.11.5.0) (Waterhouse et al., 2009).

The genome assemblies of 720 strains, including those of 690 strains from the NCBI database and those of 30 strains obtained in this study (18 finished and 12 draft sequences), were annotated using Prokka (Seemann, 2014), and core genes (n = 1,632) were identified using Roary v3.13.0 (Page et al., 2015) with 90% amino acid sequence identity cut-off. Core gene single nucleotide polymorphisms (SNPs) (n = 51,418) were extracted using the core gene alignment tool in Roary and used as inputs for maximum-likelihood (ML) inference with RAxML v8 (Stamatakis, 2014). The ML tree was displayed and annotated using iTOL v6 (https://itol.embl.de) (Letunic and Bork, 2016). The tree was mid-point rooted and the confidence value of each branch were estimated by bootstrap with 200 replications. Genomes showing no core gene SNPs (14 strains forming nine pairs or groups) were deduplicated (strains excluded are indicated in Supplementary Table 2). Strains used as the references for E. coli phylogroup assignment are also indicated in Supplementary Table 2. Serotypes of strains belonging to phylogroup E were determined by SerotypeFinder 2.0 (Joensen et al., 2015).

To identify potential virulence-associated genes in each genome sequenced Kagoshima strain, we retrieved the core data set of protein sequences from the Virulence Factor Database (VFDB) website (https://www.mgc.ac.cn/VFs/). A blastx search was performed for the genome sequences using the amino acid sequences of the core dataset, with the following parameters: minimum identity of 60%, minimum coverage of 90%, and E-value cutoff of 0.01.

This study was conducted with the approval of the Ethics Committee for Epidemiological Research, Graduate School of Medical and Dental Sciences, Kagoshima University (#190105).

Of the 2,726 Kagoshima strains isolated from children with diarrhea, 185 (6.8%) were positive in the PCR screening for the astA gene (listed in Supplementary Table 1). In a blastn search of the astA gene in 9,065 finished E. coli genomes obtained from NCBI, 690 (7.6%) were found to possess one or more astA genes.

Prior to the MLST analysis of the 185 astA-positive strains, we performed a RAPD analysis to identify genetically related strains among them. As 39 groups of strains with epidemiological links showed identical amplification patterns in each group (data not shown), one strain was selected from each group and used for the MLST analysis. Thus, 139 strains were subjected to the MLST analysis. This analysis revealed that they belonged to diverse STs: 63 STs were identified with ST6196 being as the largest group containing 14 strains (Supplementary Table 1). This result suggests a wide distribution of astA-positive Kagoshima strains in the entire E. coli lineage. Among the 139 astA-positive strains, 30 strains were selected so that they represented the phylogenetic diversity of astA-positive Kagoshima strains as much as possible and subjected to genome sequencing and following whole genome sequence (WGS)-based analyses.

By analyzing the astA genes in the genomes of 720 astA-positive E. coli strains (those of the 30 Kagoshima strains sequenced in this study and the 690 finished genomes obtained from the NCBI database), we identified 31 nucleotide sequence types. They included four of the eight known sequences (V1, V2, V3, and V5 were not detected in this strain set) and 27 newly identified types which were named V8–V34. In addition, various lengths of astA fragments that encode only C-terminal parts of EAST1 were detected in 350 genomes.

As shown in Figure 1B, while many of the variants (23/34 variants) showed 1- or 2-bp difference compared with the V0/prototype, the remaining 11 variants showed 3–12 bp difference. Of these variants, four variants were inactivated by premature stop codons (V7 and V15) or base-changes in the start codon (V22 and V32). Amino-acid sequence alignment of the 31 astA genes that encode full-length EAST1 (Figure 1C) revealed that seven variants encoded EAST1 identical to that encoded by the V0/prototype. In addition, two variant astA genes encoded an identical EAST1 peptide. Thus, 23 variants of EAST1 were identified. We named them EAST1_1 - EAST1_23, of which EAST1_1 corresponds to the EAST1 encoded by the V0/prototype astA gene. Although three EAST1 types contained three, four or five amino-acid substitution compared to the EAST1_1, the remaining types showed one or two amino-acid differences. Of the 38 amino acid residues, 22 were fully conserved, including two of the four cysteine residues present in EAST1_1. These four cysteine residues (Cys-17, 20, 24, and 27) are involved in the formation of two disulfide bridges responsible for the heat stability and, especially, the Cys-17 residue is important for a disulfide bridge integrity for toxicity expression (Uzzau and Fasano, 2000). Of these four cysteine residues, Cys-17 is conserved in all EAST1 types, but in EAST1_3, 4, and 13, one or two amino-acid substitutions occurred at Cys-20 and Cys-27.

Among the 720 strains analyzed, the most dominant nucleotide sequence type was V22 (217 strains; 30.1%), followed by V0/prototype (146 strains; 20.3%), V6 (68 strains; 9.4%), V27 (46 strains; 6.4%), and V12 (36 strains; 5.0%). The other types were detected in less than 2% of the 720 strains (Table 1). Of the top five types, V22 has been inactivated as mentioned above. V12 encoded the EAST1 identical to that encoded by V0/prototype (EAST1_1). As additional seven variants encoded EAST1_1, the most predominant EAST1 type was EAST1_1, which can be potentially produced by 199 strains.

As for the genomic locations of the astA genes identified in the 720 strains, they were located on chromosome or plasmid. While the locations of the astA genes other than V0/prototype showed some bias toward either chromosome or plasmid, V0/prototype was located almost evenly on chromosome and plasmids and 12 strains carried it on both chromosome and plasmid (Table 1). As astA-bearing plasmids were 28-380 kb in size (data not shown), they are likely single- or low-copy plasmids and many of them are probably transmissible or were previously transmissible.

Interestingly, 28 strains harbored two or three types of potentially active astA genes (encoding a full length EAST1) with various combinations: two types in 27 strains and three types in one strain (Table 2). Of the 15 combinations detected, V0/prototype involved in eight combinations, which was an expected finding from its wide distribution. A more interesting and important finding was that a notable number of V0/prototype-containing strains (43/146) possessed multiple copies of this type of astA gene (Table 1). Of these strains, while 22 strains possessed two copies and seven strains possessed three copies, 14 strains contained five or more copies (up to 11 copies). This was in sharp contrast to other types of astA genes: only six variants, of which two were inactivated ones (V15 and V22), appeared twice or three times in the genomes of only eight strains. These findings indicate that the amplification of astA occurred almost specifically for the V0/prototype.

To investigate how various types of astA identified in the 720 strains are associated with IS1414, we analyzed the sequences flanking each astA gene (1,000-bp sequences upstream and downstream of astA). This analysis revealed that while all were associated with IS1414, many of the associated IS1414 have been decayed by deletion or inactivated by mutations in the transposase gene (Supplementary Figures 1, 2). However, most V0/prototype genes (203/270) were associated with intact IS1414, thus they are transposable as a part of IS element (Supplementary Table 2). The V30 and V31 genes were also associated with intact IS1414, but they were harbored by only one strain, respectively.

We constructed an ML phylogenetic tree of the 720 astA-positive strains based on their core gene sequences to investigate the phylogenetic relationship of the 720 strains and the distribution of the V0/prototype and variant astA genes and astA fragments in this strain set. As shown in Figure 2, the astA genes were distributed in all E. coli phylogroups. Notably, most of V22 (the most prevalent but inactivated variant) were distributed in phylogroup E: 94.5% of V22-positive strains belonged to this phylogroup (Supplementary Table 4). However, this bias was apparently introduced by the presence of a large number of O157:H7 strains in the strain set analyzed: it included 258 O157:H7 strains, of which 202 contained the V22 variant (Supplementary Table 5). Except for this bias of V22, there was no clear association between the astA type and the lineage (phylogroup) of strains. For example, the V0/prototype astA gene was distributed in all phylogroups and the strains harboring multiple copies of V0/prototype were also found in all phylogroups.

We investigated the relationship between clinical symptoms and astA gene variants in 30 Escherichia coli isolates collected in Kagoshima Prefecture that were subjected to whole-genome sequencing in this study. Detailed epidemiological data were available for 14 of the 30 patients. However, isolates of Campylobacter jejuni or norovirus were also detected in 8 of these cases, suggesting that E. coli was not the primary causative agent; thus, these cases were excluded from further analysis. In the remaining 6 cases, only astA-positive E. coli strains were isolated. Among these, five isolates—excluding strain K9291—harbored intact astA variants (V0/prototype or V33), while K9291 carried the inactive V15 variant. Furthermore, a virulence factor search against the VFDB database revealed that only strain K12343 did not possess any potential virulence factors other than the astA gene. The remaining strains carried additional virulence factors, including senB (encoding enterotoxin), the cfa operon (encoding CFA/I fimbriae), or genes encoding effector proteins (EspL, EspR, EspX, EspY) secreted via the locus of enterocyte effacement-encoded type III secretion system (LEE-T3SS). However, none of the five strains possessed the LEE region itself in their genomes, suggesting that the LEE-T3SS is absent and therefore the effector proteins are unlikely to be secreted (Supplementary Table 6).