|Home | About | Journals | Submit | Contact Us | Français|
The subtilase cytotoxin (SubAB) is an AB5 toxin described in certain Shiga toxin (Stx)-producing Escherichia coli (STEC) strains that usually lack the locus for enterocyte effacement (LEE). We report for the first time the production of SubAB by two Stx-negative E. coli strains, isolated from unrelated cases of childhood diarrhea. The characterization of the SubAB-coding genes showed a 90% nucleotide sequence similarity with that of the prototype subAB, located on the virulence plasmid of the STEC O113 strain 98NK2 (pO113). In both strains, subAB was physically associated with tia, an invasion genetic determinant of enterotoxigenic E. coli. The strains were negative for the saa gene, encoding an adhesin located on pO113 and present in many of the SubAB-positive strains described so far. PCR screening of 61 STEC and 100 Stx-negative E. coli strains in our collection revealed the presence of subAB in five LEE-negative STEC strains but not in the Stx-negative strains. subAB was contiguous to tia in three of the positive strains, which were all negative for saa. These results indicate that SubAB production is not restricted to STEC and suggest that a subAB-tia putative pathogenicity island is involved in the dissemination of subAB genes, as an alternative to plasmid pO113.
Diarrheagenic Escherichia coli may produce AB5 toxins constituted by an A subunit containing the catalytic activity and a pentamer of B subunits, usually involved in the binding to a cellular receptor. E. coli AB5 toxins directly involved in the induction of illness include Shiga toxins (Stx) (28) and the heat-labile enterotoxins, related to cholera toxin (27). The subtilase cytotoxin (SubAB) is the prototype of a recently discovered AB5 cytotoxin family produced by E. coli strains associated with human disease and that also produce Stx (19).
SubAB is constituted by a 35-kDa A subunit showing a subtilase-like serine protease activity and by five B subunits, related to a putative exported protein from Yersinia pestis and forming a pentamer which mediates binding to glycan receptors terminating in α-2-3-linked N-glycolylneuraminic acid on the target cell surface (2). This toxin was first described in the Stx-producing E. coli (STEC) strain 98NK2, associated with a small outbreak of hemolytic uremic syndrome (HUS) in South Australia (19). This strain belonged to serotype O113:H21 and did not possess the locus for enterocyte effacement (LEE), which governs the attaching-effacing mechanism of intestinal adhesion and represents a common feature of STEC strains associated with severe human disease (7).
In the 98NK2 strain, SubAB is encoded by two closely linked, cotranscribed genes (subA and subB [subAB]) located on a large, conjugative virulence plasmid designated pO113. This plasmid is also characterized by the presence of saa (20), a locus encoding the production of an autoagglutinating adhesin whose role in the colonization of the host intestinal mucosa has been hypothesized (20).
Since their first description in 2004, SubAB-coding genes have been identified in other LEE-negative STEC strains (4, 18), and it has been hypothesized that the toxin may contribute to the pathogenesis of the disease induced by these strains by a synergistic action with Stx (18, 19). SubAB causes a cytopathic effect (CPE) on Vero cells (19), and upon intraperitoneal injection in mice, it induces the typical features of the Stx-associated HUS, including extensive microvascular damage, thrombosis, and necrosis in brain, kidneys, and liver (30). Despite these features, the involvement of the cytotoxin in STEC-associated HUS in humans has not been demonstrated, and the information on the prevalence of SubAB-coding genes in STEC strains associated with severe human disease and in other pathogroups of E. coli is still scanty.
During routine Vero cell cytotoxicity assays conducted in our laboratory on E. coli strains of human, animal, and food origins, we observed that culture supernatants of two strains isolated from children with uncomplicated diarrhea induced a CPE resembling but distinct from that produced by Stx. The CPE was not neutralized by Stx-neutralizing antisera, and the strains were negative in PCR assays aiming at the identification of Stx-coding genes. In this paper, we demonstrate that the CPE induced by these two strains is governed by SubAB-coding genes and describe for the first time the presence of these genes and the production of the cytotoxin in Stx-negative E. coli strains associated with human disease. We also identify the presence of SubAB-coding genes in some LEE-negative STEC strains in our collection and show that, in both Stx-negative and Stx-positive strains, SubAB-coding genes can be located close to the tia gene (GenBank accession no. U20318), an invasion genetic determinant previously described in enterotoxigenic E. coli (ETEC) (6).
The E. coli strains used in this study were part of the culture collection of the Istituto Superiore di Sanità, and many of them have been described in previous studies (12, 13, 16). Serotyping of SubAB-positive strains was performed by the International Escherichia and Klebsiella Centre (WHO) at the Statens Seruminstitut (Copenhagen, Denmark) or by the Laboratory of Gastrointestinal Pathogens at the Health Protection Agency (London, United Kingdom). Twenty-six E. coli strains from healthy human subjects that are included in the ECOR collection were also included in the study (14).
The E. coli strains were grown in tryptone soy broth (TSB) (Oxoid, Garbagnate Milanese, Italy) at 37°C for 18 h, and culture supernatants were tested by the Vero cell cytotoxicity assay as previously described (3). The cytotoxic titer of a supernatant was defined as the highest dilution inducing CPE after incubation with the Vero cell monolayer for 3 days at 37°C. The presence of Stx was assessed using Stx-neutralizing antisera prepared in rabbits as previously described (3).
The presence of virulence genes associated with the main pathogroups of diarrheagenic E. coli strains was assessed by PCR. The intimin-coding eae gene was detected as previously described (16); stx1 and stx2 were detected according to Rüssmann et al. (22) and Persson et al. (21), respectively. The other virulence genes considered were the enterohaemolysin-coding gene, e-hly (24); saa (17), the catalase-peroxidase katP (1); the type II secretion system-associated gene, etpD (25); the invasion plasmid antigen-coding gene, ipaH (11); the enteroaggregative E. coli antiaggregation protein transporter gene, aat (previously reported as CVD 432) (26); heat-stable enterotoxin 1-coding gene, astA (23); the STEC O157 gene encoding the lymphostatin homologue, toxB (29); and the enteropathogenic E. coli factor for adherence, efa1 (13). Due to the large dimensions of toxB and efa1, the 5′ and the 3′ regions of both genes were separately amplified. The heat-labile and heat-stable enterotoxin-coding genes (elt and est, respectively) were amplified using the primer pairs LTH1/2 and STI1/2, respectively (5, 15). The primer pairs used and the amplification conditions were those described in the respective papers. Amplification products were analyzed on ethidium bromide-stained agarose gels. Primer pairs used for detection of subA, subB, and tia and for the characterization of the subAB locus are listed in Table Table11.
Long PCR experiments were carried out using the TripleMaster PCR system (Eppendorf, Hamburg, Germany) under the conditions described by the supplier.
A library of mutants was prepared from strain ED 591 by random transposon mutagenesis using the EZ::TN <Kan2>Tnp transposome kit (Epicentre, Madison, Wisconsin) under the conditions described in the manual supplied.
Single mutant colonies were inoculated into the wells of microtiter plates containing 150 μl of TSB (Oxoid, Garbagnate Milanese, Italy) supplemented with 50 μg/ml kanamycin and incubated at 37°C for 18 h. Plates were then centrifuged at 2,250 × g in an Eppendorf 5804R centrifuge, and 20 μl of supernatant from each well were inoculated onto Vero cell monolayers as described elsewhere (3).
The mutant negative for toxin production in the Vero cell assay was subjected to genomic sequencing to identify the insertion site of the EZ::TN<Kan2> transposon. Total DNA was prepared from a 5-ml liquid culture by using the Puregene genomic DNA purification kit (Gentra Systems, Big Lake, MN) under the conditions indicated in the manual supplied. Sequencing was performed using the ABI Prism Big Dye terminator cycle sequencing kit version 1.1 (Applied Biosystems, Foster city, CA), according to the manufacturer's instructions. The primers used for sequencing (10) were supplied in the EZ::TN<Kan2>Tnp transposome kit (KAN-2 FP-1 and KAN-2 RP-1). Analyses of sequence data were made using the DNASIS MAX software version 2.0 (Hitachi Software Engineering Co., Ltd.).
A genomic library of strain ED 591 was constructed by using a Lambda ZAP II predigested EcoRI/CIAP-treated vector kit (Stratagene, La Jolla, CA), under the conditions described by the supplier. In detail, 200 ng of total DNA were extracted from strain ED 591 (PureGene; Gentra Systems, Big Lake, MN), partially digested with the EcoRI restriction enzyme, and ligated with 1 μg of the vector by incubation with 2 units of DNA ligase for 18 h at 4°C. Two microliters of the Gigapack III gold packaging extract (Stratagene, La Jolla, CA) were added to the ligase reaction mixture and incubated at room temperature for 2 h. Following addition of 500 μl of SM buffer (5.8 g/liter NaCl, 2.0 g/liter MgSO4·7H2O, 50 mM Tris-HCl [pH 7.5], 0.01% [wt/vol] gelatin) and 20 μl of chloroform, the supernatant containing the phages was titrated in E. coli XL1-Blue MRF′ host bacteria. Amplification of the library was performed to obtain a final titer of 1.5 × 109 PFU/ml.
Plaques were screened using the PCR product obtained with the RTSubABF/RTSubABR primers (Table (Table1)1) as a probe. Sequencing of the positive clone was performed using the ABI Prism Big Dye terminator cycle sequencing kit version 1.1 (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions.
A DNA sequence of 3,888 bp comprising the complete sequences of SubA- and SubB-coding genes and part of the tia locus from Escherichia coli strain ED 591 has been deposited into GenBank under the accession number FJ664545.
Between 1990 and 2008, more than 3,000 E. coli strains of human, animal, and food origins were tested in our laboratory for Stx production by using the Vero cell cytotoxicity assay. The culture supernatants of two strains, termed ED 32 and ED 591, induced the death of the cell cultures, but the CPE was not neutralized by antisera against Stx1 and Stx2. The strains had been isolated from children less than 3 years of age with uncomplicated diarrhea, observed in two different Italian regions in 1990 and 2005, respectively. In both cases, stool samples had been examined for common enteric pathogens (Salmonella, Shigella, Campylobacter, rotavirus) with negative results. The isolates were examined for the presence of Stx-coding genes by PCR, using a panel of primer pairs able to amplify stx1 and the variants of stx2 (21, 22). Both strains gave negative results, thus confirming that the cytotoxin responsible for the CPE did not belong to the Stx family.
Microscopically, the appearance of the CPE was slightly different from that induced by Stx (not shown). The CPE appeared after 48 h of incubation and advanced by time, whereas the CPE induced by Stx was already visible after 24 h. Substitution of the cell culture medium 6 h after inoculation did not affect the outcome of the CPE. The culture supernatants of both strains induced CPE up to a 1:128 dilution. The toxic activity in culture filtrates of the positive strains was destroyed by heating at 75°C for 15 min or at 60°C for 30 min.
The CPE observed on Vero cells and the absence of Stx-coding genes suggested the possibility that ED 32 and ED 591 strains were producing the recently described SubAB cytotoxin (19). Therefore, they were tested for the presence of SubAB-coding genes by PCR, using the primer pairs (Table (Table1)1) described by Paton and colleagues (19). Both strains gave positive results with the primer pair RTsubABF/RTsubABR (Table (Table1),1), which is able to amplify a region bridging the end of subA and the beginning of subB. However, they were negative in PCRs performed with the primer pairs SubAF/SubAR and SubBF/SubBR (Table (Table1),1), deployed to amplify the whole subA and subB genes, respectively.
To demonstrate that the CPE observed was due to a SubAB-like cytotoxin, a library of mutants was prepared from the ED 591 strain by random mutagenesis using the EZ::TN<Kan2> transposon. One thousand eight hundred mutants able to grow on media containing 50 μg/ml kanamycin were obtained and tested by the Vero cell cytotoxicity assay. One of the mutants proved negative and was further analyzed to determine the insertion site of the transposon. Sequence analysis showed that the transposon was located within the coding sequence of the subA gene, thus demonstrating the connection between the CPE observed on Vero cells and the presence of a functional subA gene.
A genomic library obtained from total DNA of the ED 591 strain was screened by Southern hybridization using the RTsubABF/RTsubABR PCR product as a probe. One clone of about 3.7 kb in length was positive, and sequence analysis showed that it contained the complete subA gene and most of the subB gene at one extremity. The sequence of the whole subB gene was obtained by PCR amplification of total DNA from the ED 591 strain, using the primer pair SubB_up/SubB_end_R deployed on the subB gene sequence from pO113 (GenBank accession no. AF399919) (Table (Table1).1). The alignment of the subA and subB sequences from ED 591 with the corresponding sequences in pO113 returned a homology of 90% at the nucleotide level. The alignment of the translated amino acid sequences showed 93% identity and 96% positive substitutions for the SubA protein, and 93% identity and 94% positive substitutions for the SubB protein (Fig. (Fig.1).1). The 3.7-kb clone also contained a portion of the tia gene, encoding an invasion determinant described in ETEC strains (6), interrupted by an IS2 element (Fig. (Fig.2).2). Consistently, a tia-specific PCR amplification conducted using total DNA from the ED 591 strain with the primer pair tia_sense/tia_lo (Table (Table11 and Fig. Fig.1)1) yielded an amplicon of about 1.8 kb (Fig. (Fig.3),3), the size expected for the 650-bp tia sequence (GenBank accession no. U20318) interrupted by a 1.2-kb IS2 element. This DNA trait, located about 800 bp upstream of subA, is not present in the sequence of the pO113 (GenBank accession no. AF399919), suggesting that, in ED 591, subAB may be located in a different genomic region.
A set of PCR primers was used to characterize the locus encoding the SubAB toxin of strain ED 32 (Fig. (Fig.22 and Table Table1).1). The primers were either described in the literature or deployed on the sequence of the subAB locus obtained for strain ED 591. PCR mapping showed that the locus encoding the SubAB cytotoxin of strain ED 32 was similar to the locus present in strain ED 591, with the exception that, in the former, the tia gene was not interrupted by the IS2 element (Fig. (Fig.2).2). This finding was confirmed by sequencing the PCR product obtained from strain ED 32 by using the primers SubA2 and tia_lo (Fig. (Fig.22).
A panel of 61 Stx-positive and 74 Stx-negative E. coli strains from the culture collection of our laboratory was examined for the presence of subAB genes by PCR, using the primer pair RTsubABF/RTsubABR. The STEC strains belonged to serogroups O157 (17 strains), O26 (12 strains), O111 (eight strains), O103 (three strains), and O113 (eight strains) and to 10 other different serogroups (13 strains). The Stx-negative strains included 43 enteropathogenic E. coli, 10 ETEC, 13 enteroaggregative E. coli, and eight cytotoxic necrotizing factor 1-producing E. coli strains. Twenty-six nonpathogenic E. coli strains isolated from healthy humans and that are part of the ECOR collection (14) were also included in the study. All the Stx-negative strains were also negative in the PCR assay, while five STEC strains of human origin, namely ED 81, ED 97, ED 99, ED 186, and ED 424 (Table (Table2),2), gave positive results. Strains ED 81 and ED 424 had been isolated from cases of uncomplicated diarrhea, while the remaining were from an HUS case (ED 186) and from household contacts of cases of HUS (ED 97 and ED 99).
The seven subAB-positive strains were investigated by PCR for the presence of tia and other virulence genes associated with diarrheagenic E. coli, and their features are reported in Table Table2.2. All the strains were negative for the intimin-coding eae gene, but four possessed the enterohemolysin-coding gene, which is considered a hallmark of the presence of large virulence plasmids associated with STEC O157 and other pathogenic STEC strains (9). However, these isolates were negative in PCR assays specific for other virulence determinants often located on those plasmids (1, 25). Besides the two Stx-negative strains, three of the five STEC strains were positive for tia, which was also present in only two of the 56 subAB-negative STEC strains, belonging to serogroups O111 and O103. All the subAB-tia-positive strains were also positive in long PCR experiments conducted using the primer pair subB_end_R/tia_lo, yielding amplicons of about 2.8 kb. This result indicated that subAB and tia were located contiguously in all the strains with an architecture similar to that observed in strain ED 591, apart from the presence of the IS2 element interrupting tia in the latter (Fig. (Fig.2).2). The two tia-negative strains (ED 99 and ED 186) were positive for the saa locus, which was conversely absent in the tia-positive strains.
SubAB produced by E. coli is an AB5 toxin able to induce a CPE on Vero cells (19) and the typical features of the Stx-induced HUS in mice (30). Till now, the production of SubAB or the presence of subAB genes has been reported only for certain STEC strains (4, 18, 19). This paper, to the best of our knowledge, represents the first description of the production of SubAB and of the presence of subAB genes in Stx-negative strains of E. coli. Our data also confirm that SubAB cytotoxins are active on Vero cells (19), since we showed that the CPE induced by culture supernatants of the Stx-negative strain ED 591 was suppressed by the interruption of the subA gene by insertional mutagenesis.
Genetic analyses of the SubAB-positive, Stx-negative strains, ED 32 and ED 591, showed that the nucleotide sequences of the subA and subB genes were identical in the two strains and 90% similar to that of the corresponding genes present in the pO113 plasmid of strain 98NK2 (19). We do not know if this sequence polymorphism may reflect differences in the biological properties of the toxin. However, it does not have implications for diagnostics, since the PCR primers designed by Paton and colleagues (18, 19) on the sequence obtained from the prototype strain 98NK2 were able to identify the presence of the subAB genes of the Stx-negative strains described in this study.
All the SubAB-positive STEC strains described so far, despite belonging to different serogroups, had in common the lack of the LEE locus (4, 18) and, in many cases, the presence of saa, a locus encoding the production of an autoagglutinating adhesin possibly involved in the colonization of the host intestinal mucosa (20). Both strains ED 32 and ED 591 were negative for saa, while their subAB genes were contiguous to a gene homologous to tia, a genetic determinant responsible for the production of a 25-kDa outer membrane protein required for the invasion of intestinal epithelial cells, described in ETEC strains and associated with the pathogenesis of ETEC-induced disease (6). The tia gene is not present in the sequence of the pO113 plasmid (GenBank accession no. AF399919), and this suggests that the subAB genes of the Stx-negative strains described in this study may be framed by a different DNA region.
PCR screening of a collection of E. coli strains belonging to different pathogroups and of nonpathogenic strains from the ECOR collection (14) revealed the presence of subAB only in five STEC strains. Since the strains tested were selected to represent different serogroups and virulence gene assets, no conclusions can be drawn on the prevalence of SubAB in STEC isolated in Italy. On the other hand, the negative PCR results, together with those from the routine use of the Vero cell assays performed in the past years in our laboratory, indicate that SubAB production is rare among Stx-negative strains, even if further studies are needed to clarify this issue.
As far as the presence of additional virulence genes was concerned, all the subAB-positive strains detected were negative in PCR assays specific for the intimin-coding eae gene, thus confirming that the presence of subAB appears to be restricted to LEE-negative E. coli strains (4, 18). However, in a difference from previous reports (4, 18, 19), we found the presence of the saa gene in only two of the five SubAB-positive STEC strains identified in this study, while the remaining three were positive for the tia gene, as were the two Stx-negative strains. The presence of tia has been occasionally reported in EPEC and enteroaggregative E. coli (6); however, it does not appear to be a common feature in STEC, since it has been detected in only two of the 56 subAB-negative STEC strains investigated in this study. PCR mapping and sequence analyses also showed that in the three STEC strains, tia and subAB were physically associated in the same DNA region (Fig. (Fig.2),2), which could represent a putative pathogenicity island (PAI) (8). This assumption is supported by the GC content of the DNA region (48.4%) (GenBank accession no. FJ664545), which is lower than the average GC content estimated for the E. coli K-12 genome (50.8%).
All the tia-positive strains did not possess saa, which was conversely present in the two tia-negative STEC strains (Table (Table2).2). Since saa has been described in the pO113 plasmid, this observation, together with the absence of tia in the pO113 nucleotide sequence, suggests that the putative subAB-tia PAI is not located in this type of plasmid. This assumption has been confirmed by Southern hybridization experiments showing that a subAB probe, prepared by PCR using the RTsubABF/RTsubABR primer pair, did not hybridize with the purified plasmids obtained from strains ED32 and ED 591 but gave positive signals with their total DNA (data not shown).
In conclusion, we provide the first evidence that SubAB cytotoxins can be produced by Stx-negative E. coli and can be associated with human disease even in the absence of Stx production by the infecting strain. The nucleotide sequences of the cytotoxins produced by the two strains identified in this study were identical but slightly different from those of the prototype SubAB described by Paton et al. (19). Moreover, the subAB genes of the Stx-negative strains were associated with tia and not with saa, which was present in many of the subAB-positive STEC strains described so far (2, 4). The subAB genes were contiguous to tia also in three of the five subAB-positive STEC strains identified in this study. The observation of this new gene cluster suggests that, alternative to plasmid pO113, subAB genes may be vehiculated by a putative PAI and that multiple evolutionary events may have occurred during the dissemination of these genes.
This work was partially supported by grants from the European Community Network of Excellence Med-Vet-Net (contract FOOD-CT-2004-506122) and from the Italian Ministry of University and Research (SAFE-EAT project).
Published ahead of print on 25 November 2009.