In this study, we analyzed the 243 E. albertii genome sequences used in our previous study (Ooka et al., 2019). Detailed information of the strains is shown in Supplementary Table 1. The strain information for the 92 E. albertii strains used for the evaluation of EAH-genotyping PCR is provided in Supplementary Table 2.

In previously sequenced E. albertii genomes, the fliC gene has been found to be located between fliA and fliD (Ooka et al., 2015). In the first-step analysis, as it is known that fliC sequences are highly diverse in E. coli (Ratiner et al., 2010), the fliC-containing regions of the 243 E. albertii genomes were identified by blastn search using the fliA and fliD sequences of E. albertii strain CB9786 as queries, with an E-value threshold of 0.01. The fliC gene of each strain was then manually annotated with in silico Molecular Cloning Genomics Edition software version 7.29L (IMC-GE; In Silico Biology, Japan). For the strains not identified the fliC gene in the first-step analysis, we performed the second-step analysis by blastn search using the fliC sequences identified in the first step analysis or the primer sequences designed for EAH-genotyping PCR, which was described below, as queries, with an E-value threshold of 0.01.

Multiple alignment of nucleotide sequences of the fliC gene of E. albertii strains and amino acid sequences of the flagellin of E. albertii and Salmonella Typhimurium strain SJW1103 was prepared using GENETYX (version 15.0.1). After identical sequences showing no SNPs were deduplicated, the nucleotide sequence alignment of E. albertii fliC genes with those of 53 known E. coli H-serotypes (H1–H56, but missing H13, H22, and H50; Supplementary Table 3) was performed using the ClustalW program in MEGA (version 7.0.26) with the default parameters (Kumar et al., 2016).

The phylogenetic analysis of fliC genes was performed with MEGA using the obtained multiple alignment. A phylogenetic tree was reconstructed by the neighbor-joining (NJ) method with the p-distance model. Bootstrap analysis with 1000 replicates was performed to assess the significance of internal branching. The core-gene SNP-based maximum-likelihood (ML) phylogenetic tree of the 243 E. albertii strains was constructed previously (Ooka et al., 2019) using RAxML v8 (Stamatakis, 2014) and displayed and annotated using iTOL v4¹ (Letunic and Bork, 2016).

Based on the variation in the sequences of the E. albertii fliC genes, we designed four pairs of PCR primers to specifically detect each of the four E. albertii H-genotypes (EAHgs) (Table 1). As a positive control for PCR and a genetic marker of E. albertii, one primer pair targeting an E. albertii-specific region (E_al_1_NF/NR primers) (Ooka et al., 2019) was also included in the primer set. Template DNA for PCR was prepared by the alkaline boiling method. KOD -Multi&Epi- DNA polymerase (TOYOBO, Osaka, Japan) was used for PCR. Each reaction mixture (25 μl) contained 1 μl of template DNA, each primer at 1 μM, and 0.5 U of polymerase. PCR was performed with 25 cycles of 94°C for 2 min for initial denaturation, followed by 10 s at 98°C, 30 s at 60°C, and 60 s at 68°C. The PCR products were analyzed by agarose electrophoresis using 2% agarose S (Nippon Gene, Tokyo, Japan).

Among the 243 E. albertii genomes examined by the first and second step analysis, intact and partial sequence of the fliC genes were identified in 231 and 9 genomes and no sequences were detected in the remaining 3 genomes (Supplementary Table 1). Through the clustering analysis of the 231 of intact fliC gene sequences, we identified a total of 73 sequence types with one or more SNPs (Supplementary Table 1), among which 42 were singletons, and 31 were composed of sequences from multiple genomes (named clusters C1–C31).

The phylogenetic analysis of the 73 fliC sequences identified in E. albertii with 53 fliC sequences of known E. coli H-serotypes as references revealed that the fliC genes of E. albertii formed a monophyletic branch, separate from those of E. coli (Figure 1A). To obtain more detailed information on the sequence variation in E. albertii fliC genes, we performed a phylogenetic analysis of only the E. albertii fliC genes (Figure 1B). This analysis revealed that the E. albertii fliC genes can be divided into four distinct groups, in which the nucleotide sequence identities between the groups were less than 90%, and those within each group were over 97%. We defined these four groups as the genotypes of E. albertii fliC genes and named them E. albertii H-genotypes 1–4 (EAHg1–EAHg4). In addition, by the primer screening analysis for the nine strains with partial sequences of the fliC gene, all nine strains could be genotyped into either of the four EAHgs. Together with the results of in silico analysis of 231 genome-sequenced strains, the most dominant type among the 243 E. albertii genomes was EAHg4 (109 strains; 44.9%), followed by EAHg1 (59 strains; 24.3%), EAHg3 (50 strains; 20.6%), EAHg2 (22 strains; 9.0%), and no fliC gene (3 strains; 1.2%) (Table 2).

Amino acid sequence comparison of the representative fliC-encoded flagellin of the four genotypes and Salmonella Typhimurium strain SJW1103 revealed that while the sequences of the D0 and D1 domains were highly conserved, those of the D2 and D3 domains were variable (Supplementary Figure 1).

We designed a multiplex PCR primer set (Table 1) based on the sequence variation in the fliC genes of the four EAHgs. A forward universal primer was designed based on the highly conserved sequences of all four EAHgs, and reverse primers were designed based on the variable regions of each EAHg (Supplementary Figure 2), so that the amplicons generated with the combinations of the universal forward primer and the reverse primers exhibited a ladder pattern that ranged from 167 to 515 bp. One E. albertii-specific primer pair (E_al_1_NF/E_al_1_NR) (Ooka et al., 2015) was included in the primer set as a marker to detect E. albertii and as a positive control for PCR. The examination of the primer set in four strains representing each of the four EAHgs confirmed that the primer set yielded PCR products of the expected sizes for each genotype (Figure 2). To validate the PCR-based genotyping system, we performed the comparison of in silico and PCR-based genotyping on the same dataset of genomes. In this analysis, only 42 strains were used because the remaining 201 strains are not available in our laboratories. As shown in Supplementary Figure 3, all of the 42 strains exhibited the same results between in silico and the PCR-based genotyping. In addition, we applied the mixed DNA samples containing four strains with different EAH-genotypes as a template for the system to evaluate the specificity and obtained the primer set yielded PCR products of the expected sizes for four genotypes (Supplementary Figure 4).

To further evaluate the performance of the system, we determined the H-genotypes of 92 E. albertii strains using this system. These strains were isolated from diarrheal patients and birds in various regions of Japan. In this analysis, we were able to genotype all strains (Supplementary Table 2). Similar to the results of in silico analysis of the 240 genome-sequenced strains, EAHg4 (48 strains; 43.6%) was found to be predominant, followed by EAHg1 (29 strains; 26.4%), EAHg2 (18 strains; 16.4%), and EAHg3 (15 strains; 13.6%) (Table 2). These results suggest that our system is useful for the H-genotyping of E. albertii and that the four genotypes cover the most of the diversity of H-genotypes in the E. albertii lineage.

Finally, we investigated the relationship between the phylogeny of the genome-sequenced strains used in this study and the distribution of the four EAHgs and the 40 Escherichia albertii O-genotypes (EAOgs), identified in our previous study (Ooka et al., 2019), in these strains by mapping H- and O-genotype information in a core-gene sequence-based ML phylogenetic tree of the strains (Figure 3). This analysis revealed that although very closely related strains shared the same H-genotype, each EAHg appeared in multiple sublineages in both clades 1 and 2, suggesting relatively frequent within-species transfer of fliC genes in E. albertii. In addition, there is no correlation between the combination of the H- and O-genotypes and their phylogenetic relationship.