Characterization of Atypical Shiga Toxin Gene Sequences and Description of Stx2j, a New Subtype

ABSTRACT

Shiga toxin (Stx) is the definitive virulence factor of Shiga toxin-producing Escherichia coli (STEC). Stx variants are currently organized into a taxonomic system of three Stx1 (a, c, and d) and seven Stx2 (a, b, c, d, e, f, and g) subtypes. In this study, seven STEC isolates from food and clinical samples possessing stx2 sequences that do not fit current Shiga toxin taxonomy were identified. Genome assemblies of the STEC strains were created from Oxford Nanopore and Illumina sequence data. The presence of atypical stx2 sequences was confirmed by Sanger sequencing, as were Stx2 expression and cytotoxicity. A strain of O157:H7 was found to possess stx1a and a truncated stx2a, which were originally misidentified as an atypical stx₂. Two strains possessed unreported variants of Stx2a (O8:H28) and Stx2b (O146:H21). In four of the strains, we found three Stx subtypes that are not included in the current taxonomy. Stx2h (O170:H18) was identified in a Canadian sprout isolate; this subtype has only previously been reported in STEC from Tibetan Marmots. Stx2o (O85:H1) was identified in a clinical isolate. Finally, Stx2j (O158:H23 and O33:H14) was found in lettuce and clinical isolates. The results of this study expand the number of known Stx subtypes, the range of STEC serotypes, and isolation sources in which they may be found. The presence of the Stx2j and Stx2o in clinical isolates of STEC indicates that strains carrying these variants are potential human pathogens.

INTRODUCTION

Escherichia coli is a genetically diverse bacterium commonly present in a wide range of environments (1, 2). E. coli may possess specific virulence genes, which can cause infectious disease in humans. There are six recognized enteric pathotypes of E. coli, including Shiga toxin-producing E. coli (STEC) (3). All enteric E. coli pathogens are a significant cause of illness and mortality, but in regions with strong public health infrastructure, STEC is considered the most significant E. coli pathotype. This is due to the high infectivity of some STEC strains (4–7), the risk of potentially life-threatening illness with long-term sequela (8), and limited treatment options (9).

STEC is defined by the potential to express one or more Shiga toxins (Stx), as determined by detection of the toxin protein or possession of the gene stx (10, 11). STEC strains are not a clonal group and are diverse serologically and genetically, as Stx is typically encoded within the bacterial chromosome as part of the lysogenic stx phage, resulting in high horizontal mobility of stx genes (12). Consequently, individual STEC strains may possess additional virulence factors, including those which define other E. coli pathotypes (3, 13, 14), such as the locus of enterocyte effacement (LEE) (15), locus of adhesion and autoaggregation (LAA) (16), and the enteroaggregative plasmid (17).

The Stx of E. coli was first identified by Konowalchuk et al. (18), who reported that E. coli isolated from patients suffering enteric illness produced a protein cytotoxic to Vero cells. This toxin was designated verocytotoxin, and only later was it identified as a homolog of the Stx of Shigella dysenteriae (19). Thus arose the synonymous terminology of verocytotoxin-producing E. coli (VTEC) for STEC. Stx is composed of two protein subunits, A and B, which form a complex composed of a single A subunit coupled to a B subunit pentamer (20). The B subunit pentamer interacts with cell surface receptors to trigger uptake of the A subunit, which, once inside the cell, terminates protein synthesis by cleavage of a specific adenine residue from the 28S rRNA of the 60S ribosomal subunit (20).

On the basis of amino acid sequence and serological reaction, Stx was initially divided into two types, Stx1, which differs only in a single amino acid from the Stx of S. dysenteriae, and Stx2, which has 50 to 60% amino acid identity with Stx1 (21–23). Subsequent characterization of Stx from a variety of strains of STEC and other bacteria revealed the existence of Stx subtypes within Stx1 and Stx2. These subtypes have been organized into a taxonomic system based on variation in amino acid sequence, distinguishing three Stx1 (a, c, and d) and seven Stx2 (a, b, c, d, e, f, and g) subtypes (23).

STEC infections can result in a range of patient outcomes, including asymptomatic infection, uncomplicated diarrhea, bloody diarrhea (BD), and life-threatening hemolytic uremic syndrome (HUS) (8, 24). The categorization of the specific Stx subtypes possessed by STEC isolates provides important information to epidemiologists, risk assessors, and other public health officials, as the different Stx subtypes vary in their epidemiological association with patient outcomes. The current consensus is that two subtypes, Stx2a and Stx2d, are associated with a greater likelihood of BD and HUS, but all subtypes are associated with diarrhea, some more so with BD, and some with hospitalization (11). Though individuals under 5 and over 65 are at greater risk of developing HUS, the relationship between Stx subtype and patient outcomes is not fully understood, in part because of limited study of the role of underlying host conditions. Stx subtypes vary in their cytotoxicity and the rate of translocation across the intestinal epithelium (25, 26). However, other virulence factors and genetic markers also correlate with patient outcomes, and their presence may contribute to the severity of disease (10, 11, 14).

Since the establishment of the currently accepted Stx subtype taxonomy in 2012, six additional Stx subtypes have been proposed, Stx1e (27), Stx2h (28), Stx2i (29), Stx2k (30), Stx2l (11), and Stx2m (31) (see Table 1 for the accession numbers of type sequences). STEC strains are diverse, and there is a constant potential for mutation and diversification of stx phages. Since a new Stx subtype can be defined with a difference of as little as 97.7% in amino acid identity (23), it seems likely that other additional Stx subtypes exist. In this paper, we describe STEC isolates from food and clinical sources possessing atypical stx sequences and propose a position for these variant sequences within the Stx taxonomy of Scheutz et al. (23).

TABLE 1.

Novel Stx subtypes reported since 2012

Subtype	GenBank accession no.	Source or reference
stx1e	KF926684	27
stx2h	CP022279	28
stx2i	FN252457	29
stx2j	MZ571121	This study
stx2k	KC339670	30
stx2l a	AM904726	29
stx2m	SAMEA6873236 (BioSample)	31
stx2o	MZ229604	This study

^aA provisional designation was proposed by Lacher et al. (29) for this Stx subtype as Stx2h (GenBank accession no. AM904726), identical to the subtype stx2e-O8-FHI-1106-1092 (23). As Stx2h had already been published (28), this particular subtype has a new designation, Stx2l.

MATERIALS AND METHODS

Bacterial strains.

Details of the bacterial strains sequenced and characterized in this study are presented in Table 2. These strains were selected for characterization of their stx gene when they generated ambiguous results for stx subtype determination by either conventional PCR (23) or in silico prediction from whole-genome sequence data generated with Illumina technology (data not shown). STEC strains CFIAFB20160170 (O157:H7), CFIAFB20120266 (O170:H18), and CFIAFB20140388 (O158:H23) were isolated from food samples by the Canadian Food Inspection Agency laboratories in the course of routine surveillance activities and have not been epidemiologically linked to human illness. STEC strains BMH-17-0027 (O8:H28) and BMH-17-0036 (O146:H21) were isolated from wheat flour implicated in an outbreak of STEC O121:H19, but have not been epidemiologically linked to human illness (7). Strain 03-3638 (O85:H1) is a human clinical isolate supplied by Matthew Gilmour of the Public Health Agency of Canada. STEC strain PNUSAE005447 (O33:H14) was isolated from a human stool sample by the Minnesota Department of Health.

TABLE 2.

Nucleotide sequences of STEC isolates included in this study (NCBI BioProject accession nos. PRJNA218110 and PRJNA735700)

Isolate	Serotype	Origin (yr)	Stx subtype(s)	NCBI accession no.
				Stx	ONT SRA	Illumina SRA	Genome
CFIAFB20160170	O157:H7	Pork (2016)	Stx1a, truncated Stx2a		SRR14805000	SRR14750051	JAHMAB000000000
BMH-17-0027	O8:H28	Flour (2017)	Stx2a	MZ229605	SRR14804998	SRR14750049	CP076704-CP076705
BMH-17-0036	O146:H21	Flour (2017)	Stx2b	MZ229607	SRR14804999	SRR14750050	JAHMAA000000000
CFIAFB20120266	O170:H18	Sprouts (2010)	Stx2h	MZ229606	SRR14804997	SRR14750046	CP076706-CP076708
CFIAFB20140388	O158:H23	Lettuce (2016)	Stx2j	MZ229608	SRR14804996	SRR14750047	CP076709-CP076713
PNUSAE005447	O33:H14	Human (2016)	Stx2j	MZ571121		SRR5122334	AASRCM000000000
03-3638	O85:H1	Human (2003)	Stx2o	MZ229604	SRR14804995	SRR14750048	CP076714-CP076717

All strains were stored at −80°C either on CryoStor beads (VWR, Mississauga, Ontario, Canada) or in brain heart infusion (BHI; Becton, Dickinson [BD], Mississauga, Ontario, Canada) broth with 15% glycerol. Cultures were recovered by plating onto BHI agar (BD) and incubated overnight at 35°C. Broth cultures of strains were prepared by overnight incubation in 10 mL of BHI broth at 35°C.

DNA extraction.

High-molecular-weight DNA was extracted from bacterial cells by phenol-chloroform-isoamyl alcohol (PCI) with ethanol precipitation (32). The purified DNA was resuspended in 100 μL of 10 mM Tris-HCl. The DNA was processed to remove RNA and DNA fragments smaller than 1,500 bp using solid-phase reversible immobilization beads (Ampure XP beads; Beckman Coulter, Indianapolis, IN) according to the protocol of Hosomichi et al. (33). The DNA concentration in the resulting solution was quantified using a Qubit fluorometer and the double-stranded DNA (dsDNA) high-sensitivity assay kit (Fisher Scientific, Ottawa, Ontario, Canada) according to the manufacturer’s instructions. Samples were stored at 4°C for same-day analysis and stored long term at −20°C. A single DNA preparation was created for each strain and used for all sequencing.

Illumina and Oxford Nanopore sequencing.

Paired-end Illumina sequencing (2 × 300 bp, v3 reagent kits) was performed on a MiSeq using NexteraXT-prepared libraries according to the manufacturer’s instructions (Illumina Inc.). For Oxford Nanopore sequencing, DNA libraries were constructed using the rapid barcoding sequencing kit (catalog no. SQK-RBK004) and sequenced on a 1D MinION (R9.4, FLO-MIN106 flow cell) for up to 24 h as per manufacturer protocols (Oxford Nanopore Technologies). Signal processing, basecalling, demultiplexing, and adapter trimming were performed with MinKNOW (v18.07) software.

Sanger sequencing.

For Sanger sequencing, the coding sequence of the stx gene was amplified with a set of two primers for flanking regions of the coding region of stx specific for each strain (Table 3). The PCR was performed in an Eppendorf Mastercycler Pro thermocycler. Each 25-μL reaction mixture contained 2.5 μL of 10× high-fidelity PCR buffer (Fisher Scientific), 0.2 mM deoxynucleoside triphosphate (dNTP) mix (10 mM; Qiagen), 2 mM MgSO₄ (50 mM; Fisher Scientific), 0.2 μM each primer, and 0.5 U of Platinum Taq DNA polymerase high fidelity (5 U/μL; Fisher Scientific). The thermocycler conditions were as follows: one cycle of 94°C for 2 min, 35 cycles of 94°C for 1 min, 58 to 65°C for 1 min, and 72°C for 1.5 min, and one cycle of 72°C for 2 min. BMH-17-0036 had an annealing temperature of 58°C, BMH-17-0027 had an annealing temperature of 65°C, and the rest of the samples had an annealing temperature of 64°C. The amplification product was purified with QIAquick PCR purification kit (Qiagen) according to the manufacturer.

TABLE 3.

Primers used for Sanger sequencing

Primer ID	Sequence (5′→3′)	Location in stx	Strain(s)
Stx2-1-1	TTATATCTGCGCCGGGTCTG	Flanking	BMH-17-0027, BMH-17-0036, CFIAFB20120266, 03-3638
Stx2-1-2	TATATATCTGCGCCGGGTCTG	Flanking	CFIAFB20140388
Stx2-2-1	AATGCCATGACCAGAGAGGC	Internal	CFIAFB20120266, CFIAFB20140388, 03-3638
Stx2-2-2	AATACAATGACCAGAGATGC	Internal	BMH-17-0027, BMH-17-0036
Stx2-3	GAGTGAGGTCCACGTCTTCC	Internal	BMH-17-0027, BMH-17-0036, CFIAFB20120266, CFIAFB20140388, 03-3638
Stx2-4-1	CAACTGACTGAATTGTGACATTGCT	Flanking	BMH-17-0036, CFIAFB20140388, 03-3638
Stx2-4-2	TAGCTAGCTGAATTGTGACATTGCT	Flanking	CFIAFB20120266
Stx2-4-3	CAACTGACTGAATTGTGATACAGAT	Flanking	BMH-17-0027

The amplified product was subjected to bidirectional cycle sequencing on an Applied Biosystems 3500 genetic analyzer (Applied Biosystems, Waltham, MA, USA), using the BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems) as recommended by the manufacturer. To ensure double coverage across the coding region, the two flanking primers were used in conjunction with two internal primers (Table 3). The number of cycles in the thermocycler program was increased to 35 from 25 to account for the long sequence length. Amplified products were purified using Wizard MagneSil Green (Promega, Madison, WI, USA) cleanup system prior to sequencing.

Bioinformatic analysis.

stx2 subunit sequences for ABI consensus UGENE v33, along with manual trace analysis were confirmed using FinchTV v1.3.1 (Geospiza, Inc.) (34) and Prokka v1.14.5 (bacterial prediction default settings) (35) annotation of NCBI reference genomes.

Phylogenetic trees for Stx2A, Stx2B, and concatenated Stx2A/2B subunits were generated using MUSCLE v3.8.31 (36) alignments in RAxML (v8.2.12, 1,000 bootstraps, PROTGAMMAWAG model) (37) and a custom Python script utilizing ETE Toolkit v3.1.1 for visualization (38). stx2 pairwise nucleotide alignment scores were generated with the BioPython v1.74 globalxx aligner (39) and were then visualized using the seaborn v0.9.0clustermap method (https://seaborn.pydata.org/). The amino acid sequences of the Stx2 holotoxin were also analyzed by the unweighted pair group method using arithmetic averages (UPGMA) by neighbor-joining cluster analysis, and evolutionary unrooted trees were created from maximum parsimony cluster analysis using 100 bootstrap simulations as previously described (23). The amino acid sequences were analyzed for sequence motifs that would support the phylogenetic analyses (Fig. 1). The full nucleotide sequences, including the intergenic region, were analyzed by the same procedure to evaluate the possible differences between nucleotide and amino acid sequences. Discrepancies between the neighbor-joining and the maximum parsimony cluster analyses of the amino acid sequences were resolved using the evolutionary unrooted tree from maximum parsimony and compared to nucleotide analyses in order to assign subtypes and variants. The novel sequences were used to search the National Center for Biotechnology Information (NCBI) Nucleotide collection (nr/nt database) (15 September 2021) by Nucleotide Basic Local Alignment Search Tool (BLASTn; version 2.12.0) analysis.

FIG 1.

Amino acid sequences of the complete Stx2 holotoxins (top numbers), including the signal peptide regions and the complete A and B subunits and mature toxins (bottom numbers) of the 103 variants within the 12 subtypes of Stx2. Serine (S) at position 291 and glutamate (E) at position 297 in the mature toxin defining the activatable tail are underlined. Subtypes Stx2b, Stx2g, and 4 variants of Stx2e share these two amino acids but differ either in the C-terminal end of the A₂ subunit (Stx2b) or in the B subunit (4 variants of Stx2e and Stx2g). (Box) Motifs that, combined, could determine the activatable property of Stx2d. These motifs are also found in the two Stx2k subtypes. *, minor single substitutions in one or more variants within each of the 12 subtypes. The first amino acids in the mature holotoxin consensus sequence of subunits A and B are underlined. (a) Stx2b variants comprise two subgroups with 5 variants (Stx2b-ONT-I7606, Stx2b-O111-PH, Stx2b-O174-031, Stx2b-O128-24196-97, and the new variant, Stx2b-O146-BMH-17-0036) in one group and the other 12 variants in the other group. (b) Stx2e-O8FHI-1106-1092 has been renamed Stx2l-O8FHI-1106-1092. A similar variant was found in a Danish patient in 2010, Stx2l-O65-1610T27873. (c) A subgroup of three Stx2a variants (Stx2a-O113-CL-3, Stx2a-O104-G5506, and Stx2a-O8-VTB178) differs from 18 other Stx2a variants by S, E, and aspartate (D) in positions 313, 319, and 362. (d) Only four Stx2e variants (Stx2e-OR-TS09-07, Stx2e-O26-R107, Stx2e-O139-S1191, and Stx2e-O101-E-D43) have the S in position 313. (e) The three Stx2j subtypes have a 12-nucleotide deletion toward the end of 3′ end of subunit A. (f) Six Stx2b variants lack the last two amino acids, asparagine (N) and aspartate (D).

Hybrid genome assemblies were created using both Nanopore and Illumina reads. Briefly, Illumina read quality was assessed by FastQC v0.11.8 (40), and BBduk v38.71 (41) was used to remove Illumina adapters, 3′ Q scores less than 20, and reads smaller than 15 bp. Nanopore data set quality was analyzed using NanoPlot v1.20.0 (42), and reads less than 10 kb were removed using Filtlong (v0.2.0). Filtered Illumina and Nanopore reads were assembled using Unicycler v0.4.8 (43) in normal mode. The Unicycler hybrid pipeline first used SPAdes (44) to produce a short read-only assembly, and then miniasm (45) and Racon (46) were used to create bridges between contigs using the long-read data before multiple rounds of sequence inspection and improvement with Pilon (47). BLASTN (v2.9.0) (48) analysis of hybrid genomes using stx sequences (A and B subunits, as well as the complete stx cassettes) confirmed single copies of stx₁ and stx2 in each genome. For all software, default parameters were used except where otherwise specified.

The serotype of isolates was determined by comparison assemblies to the SerotypeFinder database (software version 2.0.1 [27 July 2020], database version 1.0.0 [24 September 2020]) CGE server (https://www.dtu.dk/) for Escherichia coli (49).

Predictions of the secondary structure of the A subunits of Stx2a and Stx2j were created with RaptorX (standard parameters, February 2017; http://raptorx.uchicago.edu/) (50).

Stx expression.

Stx2 expression by all STEC strains, except PNUSAE005447, was determined by an enzyme-linked immunosorbent assay (ELISA). Single colonies were inoculated to 10 mL of BHI broth and incubated overnight at 35°C. An aliquot of the BHI culture was then tested with the Shiga toxin 2 ELISA (microtiter plate) kit (Eurofins Abraxis Inc., Warminster, PA) according to the manufacturer’s instructions. STEC O157:H7 ATCC 35150 (stx1a and stx2a) was included as a positive control for Stx2a expression. Absorbance values of unknowns above 0.215 were considered positive for Stx. As expression of prophage-encoded Shiga toxin can be induced by exposure to the antibiotic mitomycin C (51), strains which tested negative for Stx (A < 0.215) or produced a weak positive response (A < 00.6) were induced by inoculating 1 mL of overnight culture into 14 mL of BHI with mitomycin C (Sigma) at 0.25 μg/mL. The BHI with mitomycin C culture was then incubated overnight at 35°C prior to ELISA analysis. All experiments were conducted in duplicate.

The Shiga toxin activity of culture supernatants from all STEC strains except PNUSAE005447 (O33:H14 stx2j) and CFIAFB20160170 (O157:H7 truncated stx2a) was detected by Vero cell assay (VCA) (52).

Data availability.

All data for this study (Raw sequence data, genome assemblies, ABI consensus sequences) are available from NCBI under BioProject accession no. PRJNA735700. The stx type sequences have also been linked to the PulseNet Escherichia coli and Shigella genome sequencing (BioProject accession no. PRJNA218110).

RESULTS

Sequencing plan.

Different sequencing technologies vary in their throughput, read length, and basecalling error rate (53). Genome misassemblies and artifacts are known to occur in genomes with repeat regions or homologous sequences that exceed the read length (54), including the potential generation of a single consensus sequence from multiple gene copies. Since STEC often carries multiple copies of the stx gene, including different alleles (23, 55), multiple sequencing technologies (Illumina, Oxford Nanopore Technologies, Sanger) were used to minimize the generation of artifacts and to ensure accurate sequencing of stx sequences. Sequencing with Nanopore generated long-read data, with lower fidelity, that could be used to guide assembly of higher-fidelity, short-read data generated by Illumina technology. The resulting genome assembly was used to confirm how many complete or partial stx sequences were present in each genome and to develop primers for Sanger sequencing of individual stx sequences.

Sequence analysis.

Analysis of the hybrid assemblies confirmed that, with the exception of STEC O157:H7 strain CFIAFB20160170, the genomes of the STEC strains studied encoded only a single copy of the gene stx. Initial analysis of Illumina data alone for CFIAFB20160170 indicated the presence of a single atypical stx sequence, but subsequent analysis incorporating the Nanopore data determined this to be an artifact. The final determination was that the genome of CFIAFB20160170 encodes a single complete stx1a, with an additional partial A subunit of stx2a, which is truncated 55 bp at the 3′ end by an insertion at nucleotide position 906.

The remaining STEC strains were found to possess stx sequences that were either previously unreported variants of established stx subtypes or novel subtypes not included in the established taxonomy. Nucleotide sequences of these stx genes were determined by Sanger sequencing (Table 3). Annotated amino acid sequences for the A subunit, B subunit, and signal peptides of the novel and established Stx are provided in Fig. 1. The number of nucleotides and amino acids in the A subunits, intergenic region sequences, and B subunits are compared in Table 4.

TABLE 4.

Number of nucleotides and amino acids in the A subunits, intergenic region sequences, and B subunits in subtypes of Shiga toxin, Shiga toxin 1, and Shiga toxin 2^a

Toxin subtype(s)	No. of nucleotides (no. of amino acids) in A subunit	Intergenic region(s) (no. of nucleotides in sequence)	No. of nucleotides (no. of amino acids) in B subunit
stx, stx1a, stx1d	948 (315)	gggggtaaa (9)	270 (89)
stx1c	948 (315)	ggggggtaaab (10)	270 (89)
stx2a, stx2c, stx2dc	960 (319)	aggagttaagY,d aggtgataagce (11)	270 (89)
stx2b	960 (319)	caggagttaaat,f ctggagttaaat,g cgggagttaaath (12)	264 (87)
stx2e	960 (319)	aaggagttaaga (12)	264 (87)
stx2f	960 (319)	cagggggtgaat (12)	264 (87)
stx2g	960 (319)	aaggagttaagc,i aaggagttaagtj (12)	270 (89)
stx2h	960 (319)	caggagttaaac (12)	264 (87)
stx2i	960 (319)	aaggagttaagc (12)	270 (89)
stx2j	948 (316)	tagtgaatgacaggagttaaac (22)	264 (87)
stx2k	960 (319)	aggagttaagt (11)	270 (89)
stx2l	960 (319)	aggagttaagt (11)	270 (89)

^aIntergenic sequences were not submitted for stx_2c-O157-020324 (GenBank accession nos. AY739670 and AY739671) and stx_2d-C-freundii-LM76 (GenBank accession nos. AY739670 and AY739671).

^bOnly two of the four analyzed sequences contain the intergenic region.

^cstx2d-C-freundii-LM76 has an insertion of aat at positions 190 to 192, resulting in a total of 963 bp in the A subunit.

^dY, variable nucleotide. t, 13 stx2a, 16 stx_2c, and 16 stx2d variants; c, 8 stx2a, 1 stx_2c (stx_2c-O157-E32511), and 1 stx2d (stx2d-O22-KY-O19) variant. One stx2c (stx2c-O157-020324) did not contain the intergenic region.

^estx2a-ONT-EBC210.

^f14 variants.

^gstx2b-O8-S-9.

^hstx2b-O118-EH250.

ⁱThree variants.

^jstx2g-Out-S-8.

The stx2 sequence of BMH-17-0036 constitutes a previously unreported variant of stx2b. The Stx2b of BMH-17-0036 has an amino acid identity of 99.5 to 99.7% with three stx2b type sequences (AJ567995.1, X65949.1, and L11078.1) and an amino acid identity of 97.3 to 98.5% with other stx2b type strains (23) but is unique in having isoleucine instead of valine at position 62 of the A subunit.

The stx2 sequence of BMH-17-0027 is a previously unreported variant of stx2a. The nucleotide and amino acid sequences of stx2a and stx2d are highly homologous, but the two subtypes are differentiated by the presence of an amino acid motif that is associated with the Stx2d property of activation (23). This motif is absent in the BMH-17-0027 Stx2a (Fig. 1). The BMH-17-0027 stx2a possesses nucleotide identity of 97.5 to 98.4% with stx2a type sequences and 96.5 to 97.7% identity with stx2d type sequences (23). Amino acid identity is 96.8 to 98.3% with Stx2a type sequences and 96.8% to 98% with Stx2d type sequences.

Strain CFIAFB20120266 was determined to carry stx2h, with the stx sequence having 100% nucleotide identity with the stx2h type sequence, CP022279.

The stx2 sequences of strains CFIAFB20140388 and PNUSAE5547 have 100% nucleotide identity and are designated stx2j. The stx2 sequence of strain PNUSAE005447 has not been previously published but has been previously reported as a new stx subtype by Xiong Wang of the Minnesota Department of Health and was designated Stx2j-O33-5447 (11). The closest nucleotide identity to stx2j is 91.9% with stx2h-O102-STEC299 (CP022279). Stx2j has its closest amino acid identity with Stx2k-O159-12GZSW01 (94.1%) and similarities up to 94.5% to some of the Stx2d variants. A 12-nucleotide deletion toward the end of the 3′ end of subunit A was found in both stx2j-positive strains so that the number of nucleotides for the A subunit was 948. Consequently, Stx2j lacks the SLYTTGE amino acid motif associated with activatability at the end of the A subunit. The B subunit of stx2j has the same length (264 bp) as the B subunits of stx2b, stx2e, and stx2f. The spacer between the A and B subunit was found to be 22 nucleotides, which is greater than all other previously reported stx2 subtypes that have spacers of 11 or 12 nucleotides. A search of NCBI’s nucleotide sequence database using BLASTn found one similar toxin sequence (GenBank accession no. CP027437.1) from a human stool isolate of E. coli O101:H6 that was 99.92% identical to the stx2j sequence from strains CFIAFB20140388 and PNUSAE005447 (56). The two new Stx2j identical sequences were designated Stx2j-O33-5447 and Stx2j-O158-CFIAFB20140388, respectively. The sequence from GenBank accession no. CP027437.1 was designated Stx2j-O101-2012C-4221 and differs from the other two Stx2j sequences by the substitution of serine with proline at position 311 (Fig. 1).

Analysis with PHASTER (57, 58) of the prophage regions encoding Stx2j in the three genomes found them to be variable. Predicted prophage regions are identified in Table 5. The prophage in CFIAFB20140388 was predicted to be 59.4 kb in length with 136 coding regions, the prophage in PNUSAE005447 was 61.2 kb with 69 coding regions, and the prophage in CP027437.1 was predicted to be 47.9 kb in length with 68 coding regions. Alignment of these prophage sequences indicated that there was a conserved region of approximately 48 kb among these prophages (Table 5). The conserved region surrounds the Shiga toxin genes, and the only major variability among the isolates is the insertion of a tnpA gene encoding an IS200/IS605 family transposase in CFIAFB20140388 (Fig. 2).

TABLE 5.

Prophage conserved region of stx2j-positive STEC

Isolate	Accession no.	Start positions	End positions	% query coveragea	% identitya
CFIAFB20140388	CP076709.1	2203650, 2216188	2263011, 2264221	100	100
PNUSAE005447	AASRCM010000009.1	61712, 66655	122952, 113977	98	99.88
2012C-4221	CP027437.1	86316, 84653	38453, 36600	99	99.48

^aRelative to the conserved region of CP076709.1, which was used as the query sequence.

FIG 2.

(A) Region of conserved prophage sequence in STEC encoding stx2j. (B) Gene organization within the hatched box is expanded.

Strain 03-3638 was determined to carry a previously unreported stx subtype, designated stx2o, whose characterization will be reported in a separate paper (F. Scheutz, unpublished data).

Stx expression.

All STEC strains in this study were confirmed to express Stx2 as determined by ELISA (Table 6). Strains CFIAFB20160170 (Stx2a truncated) and CFIAFB20120266 (Stx2h) gave strong positive reactions without induction (absorbance > 0.6). Strains BMH-17-00027 (Stx2a) and 03-3638 (Stx2o) gave weak positive reactions (absorbance > 0.215) without induction. For two strains, BMH-17-0036 (Stx2b) and CFIAFB20140388 (Stx2j), Stx2 expression was not detectable without induction with the antimicrobial mitomycin C. Absorbance levels were still relatively low (0.215 to 0.600) for induced BMH-17-0027 and CFIAFB20140388. Supernatants from the STEC strains tested (BMH-17-0027 Stx2a, BMH-17-0036 Stx2b, CFIAFB20120266 Stx2h, CFIAFB20140388 Stx2j, and 03-3638 Stx2o) were positive for Shiga toxin activity in the Vero cell assay.

TABLE 6.

Stx2 expression as determination by enzyme-linked immunosorbent assay^a

Isolate	Stx2 subtype	Absorbance at 450 nmb
		Uninduced		Induced
		Replicate 1	Replicate 2	Replicate 1	Replicate 2
CFIAFB20160170	2a (Truncated)	1.216	1.170	NDc
BMH-17-0027	2a	0.264	0.228	0.263	0.250
BMH-17-0036	2b	0.105	0.104	1.160	1.271
CFIAFB20120266	2h	3.189	3.263	ND	ND
CFIAFB20140388	2j	0.166	0.139	0.240	0.232
03-3638	2o	0.464	0.448	0.676	0.693
ATCC-35150	2a	6.108	6.045	3.770	3.868
Kit positive control	NAc	0.697	0.698	0.825	0.750
Brain infusion broth	NA	0.107	0.120	0.116
Kit negative control	NA	0.105	0.116	0.109

^aShiga toxin 2 ELISA (microtiter plate); Eurofins Abraxis Inc. Warminster, PA, USA.

^bAbsorbance at 450 nm greater than 0.215 indicates the presence of Stx2 (per the manufacturer’s instructions).

^cND, not done; NA, not applicable.

Secondary structure of the A subunits of Stx2a and Stx2j.

Images of the predicted three-dimensional structure of the A subunits of Stx2a and Stx2j are presented in Fig. 3. Predicted secondary structures have P values of 2.93 × 10⁻¹⁰ for Stx2a and 3.81 × 10⁻¹⁰ for Stx2j. The truncation of the C terminus of Stx2j by three amino acids is predicted to result in a shortening of the alpha helix at the terminus.

FIG 3.

Predicted structure of the A subunit of Stx2a and Stx2j. Red box indicates the C terminus of the protein.

DISCUSSION

In this study, we characterized the stx genes from seven E. coli isolates identified as possessing stx genes other than the 10 established subtypes (1a, 1c, 1d, 2a, 2b, 2c, 2d, 2e, 2f, and 2g) (23). Strain CFIAFB20160170 (O157:H7) was determined to possess stx1a and a partial stx2a sequence that encodes a truncated A subunit. It is notable that Stx2 was detected in this strain by ELISA, though this strain cannot express a functional Stx2a holotoxin due to the absence of the 2a B subunit. The stx2a truncation may reduce the potential of the strain to cause severe illness, as observed in an outbreak of STEC O157:H⁻ with stx1a and a truncated stx2c (59). The initial indication that strain CFIAFB20160170 (O157:H7) possessed a novel stx sequence was apparently an artifact of short-read sequencing and assembly. This is a risk whenever multiple copies of a gene are present in the same cell, as assembly algorithms may place reads with high homology into the same sequence position. This highlights the importance of using validated pipelines for stx gene identification to mitigate the risk of misidentification when multiple copies of the gene are present (55, 60). Complementing short-read data with long-read sequencing or the use of read-mapping strategies should be considered whenever the presence of multiple gene copies, or fragments of the gene sequence (61), in a chromosome is a possibility.

Two STEC strains were found to possess previously unreported variants of established stx2 subtypes (BMH-17-0027 O8:H28 stx2a; BMH-17-0036 O146:H21 stx2b). The remaining strains possessed novel stx2 subtypes, one of which, stx2h (CFIAFB20120266), has been previously described (28), and two, stx2j (CFIAFB20140388 O158:H23 and PNUSAE005447 O33:H14) and stx2o (03-3638 O85:H1), have not been previously described. STEC strains carrying these stx genes were confirmed to express Stx2 by ELISA and to be cytotoxic to Vero cells. These findings confirm that the novel stx2 genes reported in this study can be expressed by the host cells and are cytotoxic. The low absorbance readings of the Stx2 ELISA, following induction, of strains BMH-17-0027 (Stx2a) and CFIAFB20140388 (Stx2j) may have resulted from either low affinity of the ELISA antibodies for these particular Stx2 proteins or lower toxin expression levels.

The primary question raised by the discovery of novel Stx subtypes is their significance to human health. Individual strains of STEC differ in their probability of being involved in severe human illness and the probability of being involved in large outbreaks. The factors determining the involvement of specific STEC strains in outbreaks are not fully understood but possibly involve ecological factors which determine the probability of human exposure and the probability of the strain successfully colonizing an exposed host (infectivity). It is unclear whether the Stx subtype plays a significant role in determining the host range of STEC strains. STEC with stx2e is associated with pigs, in which they cause edema disease (62). However, such associations must be considered cautiously. Stx2f was considered strongly associated with pigeons following the first report of the subtype to use the designation Stx2f (63). However, the first report of the Stx2f, as SLTIIva, was from an infant diarrheal isolate of STEC (64), and there have been multiple clinical isolations of Stx2f-positive STEC (65–67). A recent study from the Netherlands found no evidence that pigeons are a source of human exposure to stx2f-positive STEC, as the isolates from pigeons and humans have different serotypes and molecular subtypes (68). Regarding STEC carrying stx2h, stx2j, and stx2o, the available evidence is that human exposure can occur. The only previous report of stx2h has been in STEC O102:H18 isolates from Tibetan marmots (28). Our discovery of an stx2h-positive STEC O170:H18 strain, isolated from sprouts in Canada, indicates that stx2h is not restricted to STEC with a specific host and geographic range. The stx2j-positive STEC reported in the study is diverse in both serotype and isolation source (O158:H23 lettuce, O33:H14 human, and O101:H6 human). Further, evidence of the potential diversity of stx2j-positive STEC is the variability in the three prophage sequences identified in this study that indicate that these phages do not have a recent common origin. The potential for stx2o-positive STEC to cause disease is indicated by the origin of the strain described in this paper as a human clinical isolate. Future isolation of STEC carrying these novel subtypes may indicate an association with specific hosts or environments. As stx subtyping becomes routine for STEC isolates, evidence regarding the diversity of stx subtypes in STEC isolated from different sources may accumulate (69–71).

There is an emerging consensus that the stx subtype is an important component in the hazard characterization of STEC isolates in combination with the presence of genes involved in host colonization (i.e., eae and aggR), previously reported pathogen subtypes and epidemiological evidence (10, 11, 72). Specifically, the presence of stx2a and stx2d is correlated with a greater probability of BD and HUS (10, 72). STEC carrying other stx subtypes have also been isolated from patients suffering BD or HUS, including stx2f (66, 67, 73, 74). Reports of HUS involving stx1c, stx2b, and stx2e are rare and typically involve individuals whose health status is compromised by other conditions (75–78). Subtypes stx1d, stx1e (Enterobacter cloacae), stx2g, and stx2k have been isolated from patients with enteric illness but have not been reported as isolates associated with HUS (27, 69, 79–81). The STEC isolates carrying stx2j and stx2o reported in this study were also recovered from patients with enteric illness and have no association with HUS. Two subtypes, stx2h and stx2i, have not yet been reported in STEC clinical isolates, but the absence of disease reports involving these subtypes may result from low prevalence rather than an absence of toxicity to humans or from the failure of currently used methods to detect these subtypes.

Several factors have been identified which may play a role in determining the association of the different Stx subtypes with severe illness. Experiments with Vero monkey kidney cells, human renal proximal tubule epithelial cells, and mice found that Stx2a and Stx2d were significantly more toxic than Stx1, Stx2b, and Stx2c (25). For Stx2a, it has been shown that the A subunit has a higher affinity for ribosomes and a higher catalytic activity than the A subunit of Stx1a (82). However, factors other than cytotoxicity may play an important role. STEC may carry multiple copies of the same, or different, stx genes (55). The presence of additional stx genes may not have an additive relationship to virulence; a study of patient outcomes and presence of stx1 and stx2 in STEC isolates reported that STEC strains carrying both stx1 and stx2 had a lower correlation with HUS than strains with stx2 alone (83). STEC carrying the same stx subtype may differ in their levels of toxin expression (84). Differences in stx expression have been related to differences in the insertion site of the stx phage (85) or the region of the Stx phage that regulates replication of the phage (86).

The association of Stx subtypes and severe illness may be in part correlation, rather than causation. A study of STEC isolates from the Netherlands found that specific stx subtypes correlated positively, or negatively, with other virulence-associated genes (14). The authors proposed that the association of stx subtype and patient outcomes might be a consequence of the assemblage of virulence genes associated with that stx subtype among certain STEC linages (14). This hypothesis aligns with the correlation of severe disease with both STEC with stx2a and eae and eae-negative STEC with stx2d (87). A report that the rate of Stx2a translocation across a cell monolayer differs between STEC strains implies the existence of mechanisms independent of the toxin protein that can amplify toxin impact (26). A further consideration is the potential role of other genes carried by Stx phages in STEC pathogenicity and the ability of Stx phages to influence gene expression in other portions of the bacterial chromosome (88). The presence of a conserved region flanking the Shiga toxin genes in the stx2j prophages is consistent with the Shiga toxin operating as part of a group of genes of integrated function and selective value. Finally, since stx and many other E. coli virulence genes are encoded on mobile genetic elements (3), there is a possibility that horizontal gene transfer will create high-risk STEC strains possessing stx subtypes that are not currently associated with BD or HUS.

Consequently, it would seem a reasonable precaution to assume that STEC carrying any stx-subtype have the potential to cause enteric illness. Since the presence of stx, or production of Stx, are the only traits that can differentiate STEC from other E. coli, it would be prudent to use methodologies for STEC testing which are capable of detecting as wide a range of the toxin subtypes as possible. Unfortunately, many established PCR- and serologically based methods exclude some subtypes (89–91). Laboratories should respond to the discovery of new Stx subtypes by periodically reviewing the inclusivity and exclusivity of methods for the detection of stx or Stx and consider the adoption of new methodologies if subtypes considered of concern are excluded. For nucleotide-based methods with known primer and probe sequences, such evaluations can be made by in silico prediction of binding to nucleotide sequences if type strains are not available. This is not possible for methods based on protein detection or nucleotide-based methods that use proprietary primers or probes, and so, such methods must be evaluated for the ability to detect Stx or stx from type strains. Ideally, information on Stx or stx subtype inclusivity should be provided by method originators and manufacturers.

Comparison of the stx2j nucleotide (Fig. 4) and Stx2j amino acid (Fig. 5) sequences with the sequences other subtypes indicates that stx2j and Stx2j form a distinct group, closer to stx2f and Stx2f than the other subtypes included in the analysis. If the Stx2 variants and subtypes are considered points in a design space of potential functional Stx2 proteins, then Stx2j looks like an intermediate form. This implies that other undiscovered Stx2 proteins closer yet in homology to Stx2f exists and that the diversity of Stx proteins is greater than currently recognized.

FIG 4.

stx2 subtype nucleotide comparison using maximum parsimony tree. stx2a, stx2b, stx2c, stx2d, stx2e, stx2f, and stx2g adapted from Scheutz et al. (23), stx2h from reference 28, stx2i from reference 29, stx2j from this study, stx2k and stx2o from reference 30, stx2l from reference 11, and stx2m from reference 31. Branch lengths are only shown for lengths longer than 10.00.

FIG 5.

Stx2 subtype amino acid comparison using maximum parsimony tree. Stx2a, Stx2b, Stx2c, Stx2d, Stx2e, Stx2f, and Stx2g adapted from Scheutz et al. (23), Stx2h from reference 28, Stx2i from reference 29, Stx2j and Stx2o from this study, Stx2k from reference 30, Stx2l from reference 11, and Stx2m from reference 31. Branch lengths are only shown for lengths longer than 3.00.

The primary challenge to predicting the potential diversity of Stx proteins is the lack of understanding of the function of Stx in promoting the survival of Stx phages and the bacteria that host them. Specifically, what is the selective advantage that the toxin imparts to the phage population? What compensation does the bacterial host population receive from carrying the phage in its chromosome if individual cells will be killed by phage lysis triggered by physiological stress? Stx phages are phylogenetically and morphologically diverse but have a common genome structure and lysogenic lifestyle (86, 92). This diversity indicates that Stx phages have probably arisen from the convergent evolution of multiple phage lineages, adapting to similar ecological niches following the incorporation of Stx genes. The diverse range of asymptotic animal hosts of STEC (2) and the carriage of Stx phages by bacteria that are not associated with human disease (88) do not support the hypothesis that Stx production evolved to induce disease processes in primates or other hosts. It has been proposed that Stx production confers protection to bacterial populations from protozoan predation (93), and, by implication, to other predators by use of phagocytosis, such as macrophages. Experimental evidence supporting this hypothesis has been presented (93, 94). If this is the case, the evolution of Stx phages and Stx proteins has not been a process of adaption to greater effectiveness as a virulence factor but may represent the outcome of competing selection pressures, including optimizing protection of the bacteria/phage mutualism from phagocytosis and diversification for promoting phage survival in multiple hosts and ecological niches. If this hypothesis is correct, a greater diversity of Stx phages and Stx proteins is awaiting discovery.

Conclusions.

This study expands the known diversity of stx carried and expressed by E. coli. The diversity of stx may lie unrecognized in previously collected isolates that require genome sequencing to uncover. However, we also observed that artifacts generated by short-read sequencing of strains with multiple stx genes may falsely indicate a novel stx, and so, suspected novel genes sequences should be confirmed by complementary sequencing platforms that allow resolution of genome structure.

Two of the novel Stx subtypes identified, Stx2j and Stx2o, were found in clinical isolates, indicating that strains carrying these variants are potential human pathogens. This study is also the second report of E. coli with stx2h, indicating this subtype has a potentially worldwide distribution.

The potential diversity of Stx is obscure, as is its ecological function in phages and their host bacteria, but it seems probable that the known diversity of the toxin will broaden with further research. This has implications for the reliability of methods of analysis for STEC. It may be advisable for laboratories to verify the inclusivity and exclusivity of their methods for STEC as new Stx subtypes are reported. Changes in methodology to support the detection of a wider range of toxin subtypes may be required.