Patterns of Variations in Escherichia coli Strains That Produce Cytolethal Distending Toxin

Strains, medium, and growth.Twenty E. coli strains that carry cdt genes as determined by PCR analysis (31) were used in this study. Nineteen of the strains were isolated from humans; one, strain F5953-98, was isolated from a calf. All of these strains but one were isolated from diverse geographical locations in the United States and Canada over the last several years and include several different serotypes (Table 1). Strain 3260-97 was originally isolated in Bangladesh. All of the strains, with the exception of H1, H19, and 2P6, were sent to the Centers for Disease Control and Prevention for serological or epidemiological analysis. Strains H1, H19, and 2P6, blood culture isolates from adult humans with urosepsis, were kindly provided by James R. Johnson and have been described by Johnson and Stell (17). E. coli strain S5 was used as a control in some experiments. It was kindly provided by Eric Oswald.

E. coli cultures were grown in L medium at 37°C at 250 rpm for 18 h. Strains were tested for CDT production according to the methods of Pickett et al. (30, 31). All 20 strains produced CDT titers between 30 and 200; any differences did not correlate with CDT type, nor were they statistically significant.

PCR.An initial multiplex PCR analysis to determine the presence of diarrheagenic E. coli virulence genes was carried out as described previously (38). This analysis detects the presence of the estI, eltI, stx₁, stx₂, eae, cnf1, Ehly, ipaH, uidA, and bfpA genes and the EAF plasmid. Additional PCR tests were done to analyze for the presence of cnf2, estII, and F17 fimbriae (2, 27, 28). eae subtypes were determined by PCR analysis (1, 10, 22, 33).

PCR detection of cdt genes began with the use of the degenerative primers VAT2 and WMI1, according to the method of Pickett et al. (31). These primers are based on two highly conserved regions in cdtB and will amplify, in E. coli strain 9142-88, a 510-nucleotide cdtB fragment encoding 170 amino acids (30, 31).

Additional PCR analysis of the cdt genes included use of the reactions described by Okuda et al. (25) for detection of sequences related to the type I CDT sequence found in E. coli strain 6468/62 (34) and the type II CDT sequence from strain 9142-88 (30), as well as the PCR described by Clark et al. (3) for the detection of type III CDT sequences, based on that found in E. coli strain S5 (29).

Our type I- and type II/III-specific PCR analysis of the cdtB gene described in this work used four sequence-specific primers. Two of these primers were based on the type I cdtB sequence, and two were based on the type II cdtB sequence (30, 34). The type I primers were type I-F, 5′-GCTTCTGCAACGTT(G,T)ACTGAG-3′, and type I-R, 5′-AATCGCAAGATTTACTCTGTTAGCG-3′, and the type II primers were type II-F, 5′-TCAACGGCTGTA(A,G)ATACAGG-3′, and type II-R, 5′-CAGAGCAAGATTGACTCTTCCACCAAG-3′. These forward and reverse primers are based on amino acids 44 to 50 and 86 to 94 (Fig. 1), respectively. The forward primers were used in combination with the reverse primer WMI1; the reverse primers were used in combination with the forward primer VAT2. PCRs and cycle parameters were as described by Eyigor et al. (9), except that the annealing temperature was 47°C. All PCR primers were synthesized by Integrated DNA Technologies (Coralville, Iowa).

• Open in new tab
• Download powerpoint

FIG. 1.

Alignment of partial CdtB sequences. A comparison of the sequences of a portion of the predicted sequence of CdtB from five E. coli isolates is shown. The names of the strains are shown at the left. Strain 3032-96 makes a type III CDT, strains 3260-97, 3102-97, and 9142-88 make a type II CDT, and strain 3015-99 makes a type I CDT. The amino acid sequences are numbered continuously at the right. The initial valine shown is residue number 25 of the CdtB sequence of E. coli strain 9142-88 (30).

Randomly amplified polymorphic DNA (RAPD) PCR was performed using the Ready-to-Go RAPD analysis kit according to the manufacturer's instructions (Amersham Pharmacia Biotech, Piscataway, N.J.). All six primers from this kit as well as primers 1247 and 1281 used for analysis of E. coli strains by Madico et al. (23) were tested for use in these studies. Primers 1, 2, and 3 from the Ready-to-Go kit proved to give the best range and distribution of product sizes.

Cloning and sequencing.Total DNA was isolated from E. coli strains using a QIAamp tissue kit (Qiagen Inc., Valencia, Calif.), after which the TOPO TA cloning kit (Invitrogen Corp., Carlsbad, Calif.) was used to amplify and clone a portion of the selected strain's cdtB gene. The degenerative sequences, VAT2 and WMI1, described above were used as primers in the amplification. At least two clones originating from each strain were sequenced (SeqWright, Houston, Tex.).

Nucleotide sequence accession numbers.The nucleotide sequences of the partial cdtB genes have been assigned GenBank accession numbers AY423896 , AY423897 , AY426340 , and AY426341 .

Analysis of cdtB sequences revealed invariant and variant CDT types.Multiplex PCR analysis for virulence factors commonly associated with E. coli disease indicated that E. coli strains identified as carrying the cdt genes apparently often carry additional virulence markers associated with either diarrheagenic or uropathogenic E. coli (Table 1). Based on these data, a portion of the sequence of the cdtB gene from 10 representative strains was amplified, cloned, and sequenced for comparison to already-known E. coli cdtB sequences. Translations of the partial sequences of five strains' cdtB genes are presented in Fig. 1. Strain 3032-96, whose complete amino acid sequence for the cloned region is shown, has a sequence that is identical to that of the type III CDT-producing strain, S5 (29). Two other strains, F6450-99 and 3139-97, also had this same sequence (data not shown). Another two strains, 3260-97 and 3102-88, had sequences similar to the S5 sequence and to the type II CDT-producing strain, 9142-88, (30), but several amino acid substitutions were present in the sequences in each of these strains (Fig. 1). The figure compares these substitutions to the sequences for both S5 and 9142-88; overall, the four sequences vary at 12 different residues within the 165-amino-acid cdtB fragment shown. Finally, strain 3015-99 represented five strains with identity to the type I strain, 6468/62, whose cdt gene sequence was reported by Scott and Kaper (34). One of these five strains, 3180-98, had a single amino acid substitution, Gly for Arg, at position 89. Overall, our results indicated the presence of three types of sequences that could be grouped into two broad divisions. First, there were those sequences that were identical or nearly identical to the type I sequence reported by Scott and Kaper (34). Secondly, there was a group of strains with partial CdtB sequences identical to that found in the type III CDT strain S5 (29). And lastly, there were two strains with partial CdtB sequences that were closely similar to that of both the type II strain 9142-88 (30) and the type III strain S5 (29) but which had at least five substitutions within the 165-amino-acid region of CdtB when compared to the prototype type II and III strains (Fig. 1). When referring to the latter two groups together, we have called them the typeII/III group. Individually, type II strains include the prototype type II strain 9142-88 as well as the two strains 3260-97 and 3102-97 that were similar but not identical to both 9142-88 and S5. Type III strains are those strains whose sequences were identical to that of S5.

Type-specific PCR for classification of cdt genes.Examination of the different partial cdtB sequences suggested that it would be possible to make PCR primers that were specific for either type I or for type II and III (II/III) cdtB sequences. We used four specific primers in combination with VAT2 or WMI1 to test whether we could use these primers in a screening PCR to determine if an E. coli strain likely had a type I or type II/III cdtB gene. We first tested the primers with several strains for which the sequence in the relevant region had already been determined. In every case, the type I primers, in combination with VAT2 or WMI1 (Table 2), amplified only sequences of the expected sizes from type I strains. Figure 2 shows the results for two known type I strains, 3061-99 and 3015-99. A PCR product was produced with only one of the two primer sets for each strain. For example, type I strain 3061-99 yielded a product with the type I primers (lane 2), but not with the type II/III primer set (lane 8). Similarly, the type II/III primers only amplified sequences of the expected sizes from known type II or type III strains (Table 1; Fig. 2, lane 11; and data not shown). We then tested the primer sets on eight additional CDT producers whose cdtB gene had not been sequenced. Of these strains, two produced a product with the type I primers (Table 2; Fig. 2, lanes 3 and 7) and the remaining six strains produced a PCR product only when tested with the type II/III primers (Table 2 and Fig. 2, lane 10, for strain 3084-98). All of the strains acted in a consistent and reproducible manner, suggesting that these type-specific primers can be successfully used to differentiate type I from type II and type III CDT-producing strains. These primers cannot distinguish between type II and type III sequences.

• Open in new tab
• Download powerpoint

FIG. 2.

PCR products showing CDT type specificity. Primers for lanes 2 to 7 were type I-1R and VAT-2. Primers for lanes 8 to 13 were type II-1F and WMI-1. Lanes: 1 and 14, HaeIII φX174; 2 and 8, 3061-99; 3 and 9, 3142-98; 4 and 10, 3084-98; 5 and 11, 3260-97; 6 and 12, 3015-99; 7 and 13, 3008-99. Selected DNA fragment standard sizes, in base pairs, are shown to the right.

Clark et al. (3) recently reported on the use of three sets of PCR primers for distinguishing between the three types of cdt genes. The PCR primers for detection and amplification of the type I and type II cdt genes were first reported by Okuda et al. (25) and are derived from the cdtA genes of the type I and type II prototype strains, 6468/62 and 9142-88, respectively. The type III primers are taken from the cdtA and cdtC genes of S5, such that the entire cdt coding region is amplified (3, 29). We tested these primers on our CDT-producing strains in order to determine if they would successfully detect and distinguish the different cdt sequences in these strains. The best results were achieved with the type I primers, although these were not without deficiencies. The type I strains 3008-99, 3015-99, 3061-99, H1, H19, and 2P6 all produced a single appropriately sized product (575 bp) with the type I primers. The results for type I strains 2P6, 3008-99, 3015-99, and 3061-99 are shown in Fig. 3A. However, type I strain 3180-98 produced no PCR product, and type I strain 3142-98 produced only a trace amount of product with these primers (Fig. 3A, lanes 9 and 8, respectively). No type II or type III cdt sequences were amplified by these primers.

• Open in new tab
• Download powerpoint

FIG. 3.

PCR products produced by type I, type II, or type III CDT-producing strains. (A) PCR products amplified from the indicated E. coli strains with the type I-specific primers of Okuda et al. (25). Lanes: 1, 3014-99; 2, 3139-97; 3, F6450-99; 4, 2P6; 5, 3008-99; 6, 3015-99; 7, 3061-99; 8, 3142-98; 9, 3180-98; 10, 123-bp ladder. (B) PCR products amplified from E. coli strains with the type II-specific primers of Okuda et al. (25). Lanes: 1, S5; 2, 3014-99; 3, F5971-99; 4, F6448-99; 5, F5953-98; 6, 3084-98; 7, 3148-97; 8, 3032-96; 9, 3260-97; 10, 3102-97; 11, 9142-88; 12, 123-bp ladder. (C) PCR products amplified from E. coli strains with the type III-specific primers of Clark et al. (3). Lanes: 1, S5; 2, 9142-88; 3, F5971-99; 4, F6448-99; 5, F5953-98; 6, 3148-97; 7, 3139-97; 8, F6450-99; 9, 3032-96; 10, 123-bp ladder. (D) PCR products amplified from E. coli strains with CNF2-specific primers. Lanes: 1, S5; 2, 9142-88; 3, F5971-99; 4, F6448-99; 5, F5953-98; 6, 3148-97; 7, 3139-97; 8, F6450-99; 9, 3032-96; 10, 123-bp ladder. Sizes of relevant base pair markers are at the right.

The type II and type III reactions had additional problems. As expected, the type II primers successfully amplified the appropriate region of cdtA from 9142-88 (Fig. 3B, lane 11). In addition, these type II primers also produced a significant amount of amplified product from strains 3084-9 and 3102-97 (Fig. 3B, lanes 6 and 10). The type III primers amplified the cdt genes from strains F5953-98, F6450-99, and 3032-96 (Fig. 3C, lanes 5, 8, and 9) as well as from the prototype strain, S5 (lane 1). These results for both the type II and type III primers are in agreement with our sequencing results for strains 3102-97, 3032-96, and F6450-99; no cdt genes from strains F5953-98 and 3084-98 have been sequenced. However, neither the type II nor type III primers could amplify any product from the type II/III strain 3014-99 (Fig. 3B, lane 2, and data not shown). Similarly, the type II primers produced only faint products and the type III primers produced no products with the type II/III strains 3148-97, F5971-99, and F6448-99 (Fig. 3B and C). The faint products that these latter three strains produced, as well as the product produced from the known type II strain 3260-97, with the type II primers were less distinct than the product produced from strain S5, the prototype type III strain (Fig. 3B, compare lane 1 to lanes 3, 4, 7, and 9). Strain 3139-97, a known type III strain, produced no product with the type III primers (Fig. 3C, lane 7). In other words, 6 of the 12 type II or type III strains either were undetected by these primers or, in the case of some results with the type II primers, the amplified products appeared to be the result of non-type-specific amplification, since the amount of product was less than that produced by the type III control strain. These results indicated that the type II-specific primers are not specific and will amplify cdt sequences in some type III strains and that some type II and type III strains are not detected by either the type II or type III primers.

Correlation of strain characteristics and CDT type.Analysis of the strains' known virulence factors and serotypes in relation to CDT type revealed some interesting associations (Table 1). Of the eight strains identified as encoding type I CDT, five were serogroup O2 and, of these, three were O2 nonmotile strains. Three of the strains, H1, H19, and 2P6, were initially identified as CDT producers in a study by Johnson and Stell (17) of E. coli strains associated with urosepsis, and certain O2 serotypes are well documented to be relatively common causes of this type of extraintestinal infection. However, the other two O2 strains with a type I CDT were also not stool isolates; 3008-99 was isolated from the urine of a 77-year-old female, and strain 3015-99 was isolated from the blood of a 40-year-old female. The other three type I strains, 3180-98, 3061-99, and 3142-98, were all isolated from human stools but, in common with the O2 serogroup type I strains, they did not encode Shiga toxin, enterohemolysin, or intimin. However, only the O2 serogroup type I strains possessed pap genes as assessed by PCR (reference 17 and data not shown). These results do not show that all CDT type I strains form a distinct clone, but they do indicate that at least a subset of type I CDT-producing strains appear to be associated with E. coli extraintestinal infections and that serogroup O2 strains that produce CDT likely possess type I CDT genes. In general, type I CDT-producing strains share the distinction of not having any of several known E. coli diarrheagenic-associated virulence markers (Table 1).

The type II CDT-producing strains varied from the type III strains based both on their cdtB fragment sequence and upon their complement of known virulence markers. There were three type III strains that encoded Stx1 and CNF2 but not intimin (Table 1; Fig. 3D); these are the same three strains that produced a product with the type III primers based on the cdt sequences in strain S5 (Fig. 3C). These strains appear to be similar to some of the bovine strains described by De Rycke et al. (7) in that they encode CDT type III and CNF2 but not F17 fimbriae (data not shown), although two of these three strains were isolated from humans. In any case, these strains appear to define a group of similar CDT-producing strains that make a type III CDT and also produce CNF2 and Stx1. The other known type III strain, 3139-97, encoded Stx1, Stx2, and the enterohemolysin. This strain was not identified by the type III-specific primers of Clark et al. (3) and thus appears to be distinct from the already-mentioned group of type III strains.

The three sequence-verified type II strains encode intimin, but apparently they encode no known E. coli toxins other than CDT (Table 1). In addition, one of the nonsequenced type II/III strains was identified as type II on the basis of a type II-specific PCR. This strain, 3084-98, also possesses an eae gene but none of the other virulence genes tested for in this work. None of these eae-carrying type II CDT producers tested positive for the presence of the EAF plasmid and, of note, all were nonmotile. These four strains appear to constitute another group of strains with significant commonalities, although their pathogenic significance is presently unknown. However, one of these CDT type II strains, 3260-97, is one of five strains that were described previously by Janda et al. (15) as either unusual biotypes of E. coli or as representatives of a new Escherichia species. All five of these strains were isolated from humans with diarrheal disease, and all five produce CDT (data not shown).

In addition to the six type II or III strains for which partial cdtB sequence data were obtained, six additional type II or III strains are shown in Table 1. Two of these strains, 3084-98 and F5953-98, were identified as to their CDT type based on the type II- and III-specific PCR methods discussed above (3). The other four strains have not yet been conclusively identified as to their CDT type, other than that they belong to either type II or type III (Tables 1 and 2). Three of these strains encode Stx2, and one of these also encodes Stx1. Our results from the sequenced strains would suggest that these strains are likely type III strains. The fourth strain identified as carrying either type II or III cdt genes, 3014-99, encoded neither intimin nor any toxins other than CDT in repeated PCR tests.

RAPD analysis.In an effort to learn more about the possible relatedness of strain 3014-99 to selected representatives of the different CDT types, RAPD analysis was performed. Each strain listed in Table 1 as well as some known siblings of selected strains were tested. These experiments are continuing, but our results, some of which are shown in Fig. 4, indicated that strain 3014-99, which has a type II/III partial cdtB sequence based on type-specific PCR, had a primer 2 RAPD pattern identical to that of several type I strains, including 3061-99, 3008-99, and 3015-99 (Fig. 4, lanes 5, 6, 7, and 8). The 3014-99 patterns achieved with primers 1 and 3 also appeared to be closely similar to that of the type I group, including strain 3008-99 (data not shown).

• Open in new tab
• Download powerpoint

FIG. 4.

RAPD analysis of CDT-producing E. coli strains, with an agarose gel (1%) of DNA fragments produced by RAPD using primer 2 and template DNA from the indicated strains. Lanes 1 and 12, 123-bp ladder (Invitrogen Corp.); lane 2, 3142-98 (type I); lane 3, 3180-98 (type I); lane 4, 3179-98 (a sibling of 3180-98); lane 5, 3061-99 (type I); lane 6, 3015-99; lane 7, 3008-99 (type I); lane 8, 3014-99 (type II/III); lane 9, F5971-99 (type II/III); lane 10, F5953-98 (type III); lane 11, 3139-97 (type III). The lowest rung of the 123-bp ladder is indicated at the right.