Only one species of Shigella, Shigella dysenteriae 1, has been demonstrated to produce Shiga toxin (Stx). Stx is closely related to the toxins produced by Shiga toxin-producing Escherichia coli (STEC). In STEC, these toxins are often encoded on lambdoid bacteriophages and are major virulence factors for these organisms. Although the bacteriophage-encoded stx genes of STEC are highly mobile, the stx genes in S. dysenteriae 1 have been believed to be chromosomally encoded and not transmissible. We have located the toxin genes of S. dysenteriae 1 to a region homologous to minute 30 of the E. coli chromosome, within a 22.4 kbp putative composite transposon bracketed by IS600 insertion sequences. This region is present in all the S. dysenteriae 1 strains examined. Tandem amplification occurs via the flanking insertion sequences, leading to increased toxin production. The global regulatory gene, fnr, is located within the stx region, allowing deletions of the toxin genes to be created by anaerobic growth on chlorate-containing medium. Deletions occur by recombination between the flanking IS600 elements. Lambdoid bacteriophage genes are found both upstream and within the region, and we demonstrate the lysogeny of Shigella species with STEC bacteriophages. These observations suggest that S. dysenteriae 1 originally carried a Stx-encoding lambdoid prophage, which became defective due to loss of bacteriophage sequences after IS element insertions and rearrangements. These insertion sequences have subsequently allowed the amplification and deletion of the stx region.
Shigellosis, a disease caused by bacteria belonging to the genus Shigella, is an illness of humans and some primates characterized by cramps, painful defecation, fever, diarrhoea and dysentery (Acheson and Keusch, 1995). All four species of Shigella, Shigella dysenteriae, Shigella flexneri, Shigella boydii and Shigella sonnei, contain large plasmids, which are required for the invasion of bacteria into epithelial cells, as well as virulence loci encoded on the chromosome. However, only one species of Shigella, S. dysenteriae type 1, has been demonstrated to produce a toxin, called Shiga toxin (Stx), which has cytotoxic, neurotoxic, and enterotoxic properties. Why this species alone among the Shigellae carries this virulence factor has not been understood.
The Stx produced by S. dysenteriae 1 is a heterodimeric protein, consisting of five B binding subunits of 69 amino acids (Strockbine et al., 1988) and one A catalytic subunit of 293 amino acids (Seidah et al., 1986). The A subunit is nicked by a proteolytic enzyme and its disulphide bond reduced to form the enzymatically active A1 fragment, which acts by cleaving a specific adenine residue in the 60S ribosomal subunit of a eukaryotic cell and thus disrupting protein synthesis (Tesh and O'Brien, 1991). The genes encoding the toxin subunits have been sequenced (Seidah et al., 1986; Kozlov et al., 1988; Strockbine et al., 1988). The genes lie in an operon with stxA upstream of stxB and separated from it by 12 non-coding nucleotides. Both subunit genes are preceded by ribosome-binding sites; an iron-regulated promoter, under the control of the Fur regulatory protein, is 5′ to the stxA gene (Calderwood and Mekalanos, 1988; Strockbine et al., 1988). Transcription of stxB occurs both from this promoter as well as from an independent B subunit promoter (Habib and Jackson, 1992). The cloning of the toxin genes has allowed the creation of non-toxigenic derivatives of S. dysenteriae 1 by in vivo marker exchange (Fontaine et al., 1988).
The exact location of stxAB on the S. dysenteriae 1 chromosome has not been known. Genetic studies have mapped stx near pyrF, a chromosomal locus at 28 min that is not linked to other known virulence determinants of Shigella spp. (Sekizaki et al., 1987). This approximate location was corroborated by studies with chlorate-resistant mutants of S. dysenteriae 1. Anaerobic growth of S. dysenteriae 1 strains on medium containing chlorate (ClO3−), which is reduced to the toxic compound, chlorite (ClO2−), by nitrate reductases when cells are grown under anaerobic conditions, was found to lead to a high percentage (10–25%) of spontaneous chlorate-resistant mutants that have lost the ability to produce Stx (Gemski et al., 1972). Subsequent investigation demonstrated that deletion of the stx genes was linked to loss of a putative chl (chlorate resistance) locus at ≈ 27 min, near the trp locus (Neill et al., 1988). However, the genetic basis for this deletion has been unknown.
Shiga toxin from S. dysenteriae 1 is but one of a family of cytotoxins. The Shiga toxin-producing Escherichia coli (STEC) differ from other pathogenic E. coli by producing one or more toxins closely related to the Stx from S. dysenteriae 1. These cytotoxins appear to be major virulence factors for STEC (Griffin, 1995). Toxin-producing bacteria have been associated with the clinical syndromes of haemorrhagic colitis, thrombotic thrombocytopenic purpura and the haemolytic uremic syndrome (HUS), consisting of haemolytic anaemia, thrombocytopenia, and acute renal failure. HUS from STEC infection is the most common cause of acute renal failure in children in the United States. In contrast to the experience in the United States, however, a significant proportion of HUS cases in developing countries are believed to be mediated by the Shiga toxin produced by S. dysenteriae 1 (Keusch and Bennish, 1989).
The bacteriophage-mediated mobility of the stx genes in E. coli led to an intensive search for possible transmissibility of the stx genes in S. dysenteriae 1, but this operon has not been demonstrated to be mobile. In one study, Southern blot analysis, using phage-specific probes, showed both H19A and 933W-related sequences in several STEC strains but not in Shigella spp. (Newland and Neill, 1988). However, using low-stringency conditions, other investigators found DNA sequences homologous to phage H19A in S. dysenteriae 1 and S. flexneri type 2A, but not in S. sonnei strains (Strockbine et al., 1988). Treatment of S. dysenteriae 1 strains with UV light and with mitomycin C does not produce bacteriophages that form plaques on E. coli (Strockbine et al., 1988). Kozlov et al. sequenced 3.2 kb of chromosomal DNA from S. dysenteriae 60R containing the stx operon (Kozlov et al., 1988), and noted the presence of an insertion element that was nearly identical to IS600 from S. sonnei (Matsutani et al., 1987) immediately 5′ to the toxin genes. No regions of phage homology were observed, and no IS element was located in the sequenced region downstream of the toxin genes. Nevertheless, these investigators hypothesized that the IS element could be responsible for amplification of the toxin genes in S. dysenteriae 1, and noted as unpublished results that over 10 copies of the toxin genes were present on the S. dysenteriae 1 chromosome. However, multiple studies by other investigators, using Southern blot analysis, have never revealed more than one copy of stx on the chromosome. As argued by Paton et al. (1993), the creation of stx deletion strains by allelic exchange, which has been easily accomplished, would be difficult if more than one copy of the toxin genes was present.
In this report, we provide the precise location of stxAB in the S. dysenteriae 1 genome, demonstrate that the stx operon can exist in multiple copies on the chromosome, describe the molecular basis for the amplification and chlorate-induced deletion of the stx genes and present evidence for bacteriophage-mediated horizontal transmission of stx between STEC and Shigella species.
Location of stx on the S. dysenteriae 1 chromosome
Cosmid pNAS2 contains the genes for Stx from S. dysenteriae type 1 3818T (Strockbine et al., 1988). Sequencing primers were designed from the known sequence of the cosmid vector pHC79 to enable sequencing from the vector into the chromosomal insert across the BamHI insertion site in both directions. This strategy rapidly allowed a determination of the exact site of the pNAS2 chromosomal insert within the S. dysenteriae 1 genome. At the ends of the chromosomal fragment are sequences identical to E. coli sapF and b1342 (lying within the yhcK–dbpA intergenic region), which are located at min 29.1–30.2 of the E. coli K-12 genome (Fig. 1). In E. coli, this region encompasses ≈ 56 kbp. This location correlates well with prior genetic mapping, which placed the stx operon near pyrF at min 28 and near the trp and putative chl loci at min 27 (Gemski et al., 1972; Sekizaki et al., 1987; Neill et al., 1988).
We proceeded to sequence both strands of pNAS2 by designing primers from the newly sequenced regions of the cosmid and from the published sequence of S. dysenteriae 1 stxAB (Kozlov et al., 1988; Strockbine et al., 1988). In order to overcome sequencing problems caused by the presence of multiple identical insertion elements, selected HindIII and EcoRI fragments of pNAS2 not containing intact toxin genes were subcloned into the plasmid cloning vector, pUC18. The inserts in these subclones were sequenced using universal forward and reverse pUC primers, and the sequencing was completed by using chromosomal walking. Polymerase chain reaction (PCR) amplification, spanning known regions, was used to assemble correctly the fragments and to confirm the map shown in Fig. 1.
Analysis of the stxAB region on the S. dysenteriae 1 chromosome
The S. dysenteriae 1 chromosomal region cloned into pNAS2 is 32 094 bp in length (Fig. 1). Eight insertion elements are located within this region. The stxAB operon is flanked by three pairs of IS elements: S. dysenteriae IS1 (GenBank J01731), S. dysenteriae iso-IS1 (GenBank J01737), which is present in three copies, and S. dysenteriae IS600, which is nearly identical to S. sonnei IS600 (GenBank J02459). A single S. dysenteriae IS911 element (GenBank X17613) is found immediately downstream of a region containing two open reading frames (ORFs), encoding predicted proteins homologous to bacteriophage lysis proteins. Two regions contain tandem insertion elements: iso-IS1/IS600 lie immediately upstream of stxAB and IS911/IS1 lie 4.1 kbp downstream of the stx operon. Several large deletions and rearrangements of the gene order, as found in the E. coli K12 chromosome, are present surrounding the various insertion elements.
Multiple regions homologous to lambdoid bacteriophage elements appear in the sequenced insert. Found 171 bp downstream of sapC is an ORF most homologous to the integrase of the lambdoid phage φ80 (51% identity over 399 aa; Swiss-Protein P06155), with only distant homology to the Int of phage 933W (24% identity over 236 aa). Int is followed by a φ80 Xis homologue. Immediately downstream of the excisionase is IS1. Downstream of IS1 is a truncated ORF, encoding a putative protein with homology to the EaC protein of bacteriophage P22 (Swiss-Protein Q03545), followed by genes highly homologous to lambda exo, bet and gam (92% nucleotide identity, GenBank J02459). The region downstream of stxB in S. dysenteriae 1 is similar to the corresponding regions of H19B and 933W. The highest homology among all three sequences occurs at the bacteriophage lysis gene S and endolysin gene R homologues. The product of S is 95% identical to 933W S and 91% identical to H19B S, whereas S. dysenteriae 1 R is 90% identical to 933W R and 88% identical to H19B R. Between the toxin and lysis genes in S. dysenteriae 1 is an ORF encoding a predicted 636 aa protein with homology to E. coli YjhS (Swiss-Protein P39370; 55% identity), which, in E. coli, lies in the intergenic region between fecI (in a region encoding iron transport proteins) and fimB (in a region encoding fimbrial biosynthesis proteins) and is of unknown function. Both H19B and 933W also have ORFs in similar positions encoding YjhS homologues. The S. dysenteriae 1 YjhS is more closely related to the predicted 933W YjhS homologue (645 aa, 74% identity) than to the H19B YjhS (282 aa, 43% identity). Downstream of the tandem IS911/IS1 elements is a 12.6 kbp region of DNA virtually identical to areas of the E. coli K-12 genome from ycjW to yhcK. In S. dysenteriae 1, this region is interrupted by two separate isoIS1 elements and deletions of the K-12 genes flanking the IS elements have occurred. The global regulator gene, fnr, is located within this region. The downstream IS600 element has inserted within yhcK; this element differs from the upstream IS600 in lacking the 5′ terminal 118 bp inverted repeat sequence.
Genetic variation in the stx region among S. dysenteriae 1 strains
PCR amplification was performed with S. dysenteriae 1 strains 3818T, 60R and SC503, using primer pairs specific for each locus throughout the 32.1 kbp chromosomal region in pNAS2. Amplification of these strains with the PCR primers 196 (within gam) and 197 (within stxA) produced products of 3.4 kbp from 3818T and 2.6 kbp from 60R and SC503 (Fig. 2A). Complete sequencing of these amplified products demonstrated the presence of an iso-IS1 element upstream of stxAB in strain 3818T that was not present in strains 60R and SC503 (Fig. 1). The iso-IS1 element inserted within the upstream terminal inverted repeat of the IS600 in this region, producing 8 bp flanking direct repeats (TGTTCTTT) (Fig. 2B). PCR amplification of all other areas demonstrated that S. dysenteriae 1 strains 3818T, 60R and SC503 only differed in the presence or absence of this iso-IS1 element and otherwise maintained the identical gene order and location and presence of insertion elements throughout the 32.1 kbp chromosomal region (Fig. 1).
Creation of stx deletion strains
The creation of chlorate-resistant strains of S. dysenteriae 1 that had concurrent deletion of the stx operon was performed as has been previously described (Gemski et al., 1972; Neill et al., 1988). Dilutions of S. dysenteriae 1 60R were plated onto Luria–Bertani (LB) medium with added potassium chlorate (2 g l−1) and glucose (1 g l−1), then incubated anaerobically for 48 h. Colonies were subcultured onto the chlorate-containing medium and then incubated for another 48 h. In two separate experiments, a total of 30 colonies was screened for the deletion of stxAB, using PCR with primers internal to stxAB. Three of the 30 did not produce a PCR product and were subsequently confirmed, using Southern blot analysis, to have lost stxAB (Fig. 3A). The extent of the deletion in two independently derived deletion strains (60RchlorR1 and 60RchlorR2) was determined using long-range PCR. Primers 196 and 199 were designed upstream of stxAB within lambda gam and downstream of the distal IS600 element respectively (Fig. 1). A 2.3 kbp PCR product was amplified from 60RchlorR1 and 60RchlorR2, but not from the parent strain 60R (Fig. 3B). Sequence analysis and comparison with Fig. 1 demonstrated that the deletion in these strains occurred by recombination between the two IS600 elements, leaving a single IS600 in place. PCR amplification with primer pairs throughout the 32.1 kb region further confirmed the extent of the deletion in 60RchlorR1and 60RchlorR2 (Fig. 1).
Amplification of the stx region
S. dysenteriae 1 strain SC503 contains a spectinomycin resistance cassette from the interposon omega inserted within stxA (Fontaine et al., 1988). The spectinomycin-resistant phenotype of this strain allowed a study of the amplification of the stx region in vitro. SC503 was grown overnight in LB, then plated on agar containing varying amounts of spectinomycin (50, 200, 500, 1000 and 2000 μg ml−1). Nearly the same efficiency of plating was seen on medium containing 1000 μg ml−1 spectinomycin as on medium containing 50 μg ml−1 spectinomycin. However, a marked decrease in growth was seen on media containing higher levels of antibiotic, suggesting that a single copy of the Spr allele encodes resistance to ≈ 1000 μg ml−1 spectinomycin. Single colonies of SC503, which were able to grow on LB with 2 mg ml−1 spectinomycin, were isolated; these occurred at a frequency of ≈ 1 × 10−8.
The strains resistant to high levels of spectinomycin were maintained at 2 mg ml−1 spectinomycin. A Southern blot of EcoRI-digested chromosomal DNA from one of these strains was probed with an internal stxAB fragment (amplified from strain 60R with primers 171 and 172) (Fig. 4A). The wild-type hybridizing fragment in strain SC503 was ≈ 6.2 kbp in size, as expected. However, in the highly spectinomycin-resistant strain SC503-Sp2K a larger band at ≈ 7.6 kbp was detected, in addition to a 6.2 kbp band. Moreover, the hybridization intensity of the larger band was greater than that of the 6.2 kbp band in this strain.
These results could be explained by amplification of the omega interposon alone, by tandem duplication of a region containing stxAB, or by non-tandem duplication (transposition) of the stx region. The interposon omega carries the aadA gene from Tn21 and is flanked on each side by T4 translation–transcription stop signals. The PvuII site used to construct the omega fragment is in the middle of the integrase gene, so no site-specific recombination activity is encoded (Prentki and Krisch, 1984; Prentki et al., 1991). The presence of flanking inverted repeats potentially allowed amplification of the omega interposon. To investigate this possibility, chromosomal DNA from SC503 and SC503-Sp2K was digested with EcoRI and BglII, and hybridized with an internal stxAB fragment probe. A BglII site lies upstream of stxA, 3′ to IS600; there are no BglII sites within the omega interposon. Given a single omega interposon, a 4.5 kbp EcoRI–BglII fragment in SC503 should hybridize with the stxAB probe. Duplications of omega would lead to larger hybridizing fragments. Both strains showed identical 4.5 kbp fragments (data not shown), demonstrating that duplication of the omega interposon was not the explanation for the additional, larger EcoRI fragments seen in the highly spectinomycin-resistant strain.
Two observations suggested that amplification of the stxAB region occurred by multiple tandem amplification. First, the size of the observed bands in the Southern blot (Fig. 4A) was consistent with the EcoRI sites flanking the upstream and downstream IS600 elements, as determined from the sequence of the region. Second, the hybridization intensity of the larger bands was greater than that of the wild-type bands. SC503-Sp2K was therefore amplified by PCR with the divergent primers 210 (within yhcK, upstream of the distal IS600) and 197 (within stxA, downstream of the proximal IS600), which would yield a PCR product in the event of tandem amplification occurring (Figs 1 and 4C). A single fragment of ≈ 1.7 kbp was amplified from SC503-Sp2K (Fig. 4B). The PCR product was sequenced in its entirety, confirming that the product had the expected sequence flanking a single IS600 element (Fig. 4C). Primers 210 and 197 would, however, produce the identical PCR results if the amplified DNA was on an extrachromosomal element, such as an excised circular form of the putative composite transposon produced by homologous recombination between the two IS600 elements. To examine this possibility, 2 μg each of SC503 and SC503-Sp2K chromosomal DNA and 2 μg of DNA from a plasmid preparation from SC503-Sp2K (purified with the methods used to isolate the 38.6 kbp cosmid pNAS2) were digested with EcoRI and probed with an internal stxAB fragment. Hybridizing fragments were not observed in the plasmid preparation (Fig. 4A). These results suggest that even if circularized intermediates are present and are being detected by PCR they are not in sufficient quantity to account for the intense 7.6 kbp band seen in Southern blots of EcoRI-digested SC503-Sp2K chromosomal DNA. However, it remains possible that the amplified DNA lies on a very large extrachromosomal element which co-purifies with chromosomal DNA.
The hybridization intensity of tandem fragments is known to be proportional to the degree of fragment amplification (Mekalanos, 1983; Goldberg and Mekalanos, 1986). Densitometry analysis of the radiograph shown in Fig. 4A demonstrated that the 7.6 kbp band is approximately three times as dense as the 6.2 kbp band (data not shown). As the lower band is considered to represent a single copy of the stx element at the 5′ end of the tandem array, roughly four tandem stx elements are present in the chromosome of the amplified strains.
The contribution of gene amplification to increased production of Stx was evaluated as follows. The spectinomycin resistance cassette in strain SC503 is inserted within stxA and therefore disrupts transcription of stxB from the operon promoter, but transcription persists from the stxB promoter alone. Supernatants were collected from cultures of SC503 grown in LB as well as from SC503-Sp2K grown in 2 mg ml−1 spectinomycin. The supernatants were evaluated for the presence of StxB uisng ELISA as has been previously described (Calderwood et al., 1990); all assays were performed in triplicate. The mean amount (± standard deviation) of StxB in supernatants from SC503 grown in LB alone (0.180 ± 0.02 ng ml−1 OD600−1) was approximately 10-fold less than the amount of StxB from SC503-Sp2K grown in 2 mg ml−1 spectinomycin (1.901 ± 0.13 ng ml−1 OD600−1).
To investigate whether tandem duplication of the stx region in S. dysenteriae 1 strains occurs spontaneously, primers 210 and 197 were used to amplify potential tandem junctions in the wild-type strains 60R and 3818T, the spectinomycin-resistant strain SC503, SC503-Sp2K grown in 2 mg ml−1 spectinomycin and the clinical S. dysenteriae 1 strains 109, 118 and 133 (Fig. 4B). A strong 1.7 kbp product was amplified from SC503-Sp2K, but notably, fainter 1.7 kbp products were also amplified from strains 60R, 3818T and SC503, as well as from the clinical isolates 109, 118 and 133, suggesting that low-level spontaneous duplication occurs during growth of both laboratory and clinical strains (presumably at a level too low to be detected by Southern hybridization) without the requirement of antibiotic pressure.
Lysogeny of S. dysenteriae 1 with STEC bacteriophage
To investigate the ability of STEC bacteriophages to lysogenize Shigella species, an antibiotic-marked stxAB was created on bacteriophage H19B. We employed in vivo marker exchange (Butterton et al., 1993; Butterton et al., 1995), rather than using a minitransposon to create an antibiotic-marked H19B as has been recently described (Acheson et al., 1998), in order to create a STEC bacteriophage marked with a chloramphenicol cassette inserted within stxA (see Experimental procedures and Fig. 5A). The receptors for certain Stx2-converting bacteriophages in E. coli have been recently elucidated. They are the maltose-inducible outer membrane protein, LamB, which serves as the lambda receptor, and the outer membrane protein, FadL, involved in long-chain fatty acid transport (Watarai et al., 1998). S. dysenteriae 1 strains 60R and 60RchlorR1 and S. flexneri strains 2457T and M90T were therefore grown with added maltose to induce the lambda receptor (Watarai et al., 1998), then incubated with purified H19BCmR bacteriophage (see Experimental procedures).
H19BCmR did not form visible plaques on any of the Shigella strains tested. Strains incubated with H19BCmR bacteriophage were grown overnight in LB medium, then plated onto LB agar with added chloramphenicol. Using these experimental conditions, chloramphenicol-resistant colonies were always isolated from each host strain. In five experiments, E. coli C600 consistently demonstrated the highest frequency of lysogeny, with 65 ± 27% (mean ± SD) of cells acquiring chloramphenicol resistance. The proportion of chloramphenicol-resistant cells recovered from strains 60R, 60RchlorR1, 2457T and M90T varied more widely over five experiments, ranging from 5.0 × 10−10 to 1 × 10−2, but for each strain reached a maximum of ≈ 1% in at least one experiment. As a negative control, strains that had not been incubated with the H19BCmR phage were similarly evaluated for the acquisition of chloramphenicol resistance; no spontaneous chloramphenicol-resistant strains were detected.
The presence of the H19BCmR prophage in chromosomal DNA from chloramphenicol-resistant colonies of all strains was demonstrated using PCR amplification of H19BCmR bacteriophage sequences that are not present in S. dysenteriae 1 (Neely and Friedman, 1998; Fig. 5B). Primer pairs 179/181, Cm1/Cm2 and Cm1/178 amplified fragments unique to H19BCmR from 60R(H19BCmR), 60RchlorR1(H19BCmR), 2457T(H19BCmR) and M90T (H19BCmR). No products were seen with these primer pairs in amplifications with the parent S. dysenteriae 1 and S. flexneri strains (data not shown). In 60R(H19CmR), primer pair 177/178 amplified the predicted 500 bp product from the parent strain 60R as well as the 1.4 kbp product from the H19BCmR prophage, whereas only the H19BCmR product was amplified from the stx deletion strain 60RchlorR1(H19BCmR) and the S. flexneri strains. In addition, all lysogens produced StxB as detected by ELISA (data not shown). However, the interesting possibility that regulation of the H19B stx operon promoter differs in the different strain backgrounds cannot be examined using this system, because stxA in H19BCmR is interrupted by the chloramphenicol resistance cassette. H19BCmR was easily re-isolated from the S. dysenteriae 1 and S. flexneri lysogens after phage induction with mitomycin C and propagation on E. coli C600 (data not shown).
The stability of lysogeny was evaluated by determining the proportion of lysogens that maintained the prophage after 24 h of growth in antibiotic-free medium. No significant loss of the H19BCmR bacteriophage was detected from the E. coli or Shigella H19BCmR lysogens. The percentage of colonies of the H19BCmR lysogens remaining chloramphenicol resistant (expressed as mean ± SD of triplicate cultures) was 99 ± 31% for C600, 100 ± 26% for 60R, 104 ± 23% for 60RchlorR1, 96 ± 26% for 2457T, and 98 ± 12% for M90T.
The location of virulence factors on mobile genetic elements as a mechanism for microbial evolution and diversity is a scenario that has been increasingly recognized over the past decade. Whether on bacteriophages, transposons or mobile pathogenicity islands (Cheetham and Katz, 1995; Waldor and Mekalanos, 1996; Hacker et al., 1997; Lindsay et al., 1998), genes encoding toxins and other virulence factors have been demonstrated to amplify, delete and undergo horizontal transmission. Several pieces of data suggested that the mobility of the stx operon in S. dysenteriae 1 should be re-examined. First, the location of the nearly identical toxin genes on bacteriophages in E. coli argued for phage-mediated horizontal transmission among the E. coli and Shigellae. Second, the high-frequency deletion of the toxin genes in strains grown on chlorate-containing medium was reminiscent of the chlorate-induced deletion of regions of the gal and bio operons in lambda prophage excision (Adhya et al., 1968). Third, the location of IS600 immediately upstream of the toxin genes suggested recombination between insertion sequences as an alternative potential mechanism for gene mobility. Finally, the undocumented observation of Kozlov et al. that stx was represented in the genome of S. dysenteriae 1 by over 10 copies (Kozlov et al., 1988), although not corroborated by other investigators, was intriguing.
We have located the Shiga toxin genes of S. dysenteriae 1 at a region homologous to minute 30 of the E. coli chromosome. The presence of both bacteriophage sequences and multiple insertion elements flanking stxAB, and the ability of STEC bacteriophages to lysogenize Shigella species, lead us to propose that S. dysenteriae 1 originally carried a Stx-encoding lambdoid prophage, which became defective owing to loss of phage sequences after IS element insertions and rearrangements. By analogy to the gene order of the 933W prophage in the STEC strain EDL933 (Plunkett III et al., 1999), we suggest that the putative defective prophage sequence within S. dysenteriae 1 begins with attL followed by int and xis, and that no additional deletions of bacteriophage genes occur upstream of this region. 933W integrates within the wrbA locus, at minute 23 of the E. coli genome, whereas the prophage in S. dysenteriae 1 apparently integrated at a separate site, adjacent to the translational start site of sapC at minute 29. Given the loss of attR from the prophage sequence in S. dysenteriae 1 and the heterogeneity of the sequences between the known stx-carrying bacteriophages, the precise att site for the bacteriophage that integrated into the S. dysenteriae 1 genome cannot be identified.
As observed with the STEC bacteriophages 933W and H19B, the S. dysenteriae 1 stx region is composed of a mosaic of different phages (Neely and Friedman, 1998; Plunkett III et al., 1999), suggesting active and complex gene rearrangements among the members of this bacteriophage family. Lambdoid bacteriophage genes with homology to φ80 int and xis, P22 eaC, and lambda exo, bet and gam are found upstream of the putative transposon; within the region, downstream of the stx operon, are sequences similar to those downstream of stxAB in STEC bacteriophages, terminating in bacteriophage lysis gene S and R homologues. Although sequence comparisons suggest that the S. dysenteriae 1 int and xis are distant from those of lambda and 933W, genes common to the recombination region (exo, bet and gam) in all three genomes have both gene order and predicted amino acid sequences highly conserved. Sequence comparisons show greater DNA homology between S. dysenteriae 1 and the Stx2-carrying 933W than between S. dysenteriae 1 and the Stx1-carrying H19B in the region from stxAB to the bacteriophage lysis genes. As more sequences of STEC bacteriophages are available, the evolutionary relationship between the Stx1- and Stx2-carrying bacteriophages should become more clearly defined.
The observation that the lysis genes are highly conserved among the S. dysenteriae 1, 933W and H19B genomes is of particular interest. In both 933W and H19B, the stx operon lies downstream of the Q gene, whose protein product acts as a transcriptional activator of late phage genes by antiterminating transcription that initiates at the late promoter pR′. Putative pR′ and tR′ (a transcription terminator) sites are found upstream of the toxin genes in 933W and H19B. Q-stimulated transcription from pR′ through the toxin genes to the lysis genes after prophage induction has been proposed to link phage induction and subsequent bacterial lysis of strains with enhanced toxin production and release (Neely and Friedman, 1998; Plunkett III et al., 1999). In S. dysenteriae 1, the Q, pR′ and tR′ regions have been deleted, but the sequences from the toxin genes to the lysis genes have been conserved, suggesting that there has been selective pressure for maintaining the linkage of the toxin to the lysis cassette (Neely and Friedman, 1998). The speculation that transcription from the stx promoter, independent of Q, also plays a role in expression of the downstream phage lysis genes in STEC bacteriophages (Neely and Friedman, 1998) would be supported if transcription of the toxin genes is similarly found to be linked to cell lysis in S. dysenteriae 1.
The in vitro instability of the stx genes in E. coli, as well as in other species of enteric bacteria, is a well-known problem for clinical microbiologists (Karch et al., 1992; Schmidt et al., 1993; Paton and Paton, 1997). Determinants of prophage stability are unknown, but are apparently strain dependent (Paton and Paton, 1997). In order to avoid the rapid loss of STEC prophages in this study, an antibiotic-marked stx1 bacteriophage was used to demonstrate the ability of such phages to lysogenize both S. dysenteriae 1 and S. flexneri. The antibiotic resistance phenotype allowed for rapid selection for strains with integrated bacteriophages and the stable maintenance of the prophage. Measurements of lysogen stability demonstrated that, once integration occurred, these Shigella species maintained the STEC prophage for at least 24 h without the requirement of antibiotic pressure, suggesting that Shigella species infected with STEC bacteriophages may have the ability to become stable toxin producers.
The IS element insertions and rearrangements that led to the loss of bacteriophage sequences have subsequently led to the ability of the stx region to undergo amplification and deletion. In all the S. dysenteriae 1 strains examined, the stx operon lies within a putative 22.4 kbp composite transposon bracketed by IS600 insertion sequences. The structure of the amplified and deleted regions, as demonstrated by Southern blot, PCR and sequence analysis, suggests that rearrangement occurs by homologous recombination between the flanking IS600 sequences. There is no evidence at present for non-tandem duplication or transposition of the putative stx-carrying composite transposon. However, if the region could transpose from the chromosome to a mobile genetic element, such as a transmissible plasmid, one could then speculate that mobility of the toxin genes between cells might not be limited to bacteriophage conversion. Tandem amplification of approximately four copies of the stx region can be selected by antibiotic pressure, leading to increased toxin production, but also occurs spontaneously at lower levels in both clinical isolates and laboratory strains. The contribution of these rearrangements to toxin production in vivo and to pathogenesis of disease remains to be determined.
Restriction fragment length variation of the Shiga toxin gene region is routinely used in molecular epidemiological studies of S. dysenteriae 1. In one study, all isolates from Mexico and Central America were shown to carry the toxin genes on a 5.1 kbp EcoRI chromosomal fragment, whereas those from Africa and Asia carried the toxin genes on a 4.3 kbp EcoRI fragment, suggesting that there has been no significant exchange of organisms between continents in recent decades (Strockbine et al., 1991). In fact, all prior studies that have reported EcoRI restriction fragment lengths encompassing stxAB have noted ≈ 5.1 or 4.3 kbp hybridizing fragments from the S. dysenteriae 1 strains studied (Fontaine et al., 1988; Kozlov et al., 1988; Neill et al., 1988; Strockbine et al., 1988). The results of the present study have demonstrated that the 800 bp size difference between strains in this region is a result of the presence or absence of an isoIS1 element within the IS600 upstream of stxAB, suggesting that all Western hemisphere strains are clonally derived from an S. dysenteriae 1 isolate that acquired the isoIS1 element at this location in the distant past. The availability of the nucleotide sequence of the stx region will allow the development of PCR techniques that should facilitate the epidemiological analysis of S. dysenteriae 1 infections.
The location of the global regulatory gene fnr within the stx region provides the identity of the putative chl (chlorate resistance) locus believed to be near stx (Neill et al., 1988) along with a mechanism for the chlorate-induced deletion of the toxin genes in S. dysenteriae 1 strains. Fnr, the fumarate and nitrate reductase regulator, is a global transcriptional regulator that mediates anaerobic induction of respiratory gene expression (Spiro and Guest, 1990; Stewart, 1993; Unden and Schirawski, 1997). As Fnr is necessary for the anaerobic induction of fumarate, nitrate and nitrite reductases, deletion of fnr would lead to the repression of nitrate reductase expression. Thus, S. dysenteriae 1 strains that lack fnr would be unable to reduce chlorate to the toxic metabolite chlorite. The linkage of stx to fnr therefore presents a genetic explanation for the observation that anaerobic growth of S. dysenteriae 1 strains on medium containing chlorate leads to a high percentage of spontaneous chlorate-resistant mutants that have lost the ability to produce Stx (Gemski et al., 1972).
The above studies have provided further evidence for the importance of horizontal transmission of virulence factors by mobile genetic elements, and offer new insights into the evolutionary relationship between Shigella spp. and the STEC. We speculate, from the findings presented here, that the STEC strains from which toxin-converting bacteriophages cannot be isolated also contain defective Stx-encoding prophages (Paton et al., 1995; Weinstein et al., 1988). Although only one species of Shigella, S. dysenteriae 1, has been demonstrated to produce Shiga toxin, the ability of this species to acquire a toxin-converting bacteriophage in the past, and perhaps to transmit a stx-carrying transposon in the future, suggests that other species of Shigella could undergo similar events that would allow them to become stable toxin producers. The ability of the Shigella to acquire these genetic elements will need to be addressed in present and future efforts to create live oral attenuated Shigella vaccines.
Bacterial strains and plasmids
S. dysenteriae 1 3818T and E. coli HB101(pNAS2) were gifts from Alison O'Brien. S. dysenteriae 1 3818T is a toxigenic, invasive strain isolated from a Central American patient with dysentery (Mata et al., 1969). Plasmid pNAS2 contains the genes for Stx on a 32.3 kbp fragment from a partial Sau3A digest of 3818T chromosomal DNA, cloned into the unique BamHI site of cosmid vector pHC79 (Strockbine et al., 1988). S. dysenteriae 1 strain SC503 was a gift from Philippe J. Sansonetti. S. dysenteriae 1 SC503 is a Stx− mutant of S. dysenteriae 1 SC502, a strain that lacks the invasion plasmid. SC503 contains a 2 kbp spectinomycin resistance cassette from the interposon omega (Prentki and Krisch, 1984; Prentki et al., 1991) inserted within the unique HpaI site within stxA by allelic exchange (Fontaine et al., 1988). S. flexneri 2a 2457T and S. flexneri 5a M90T were gifts from Marcia Goldberg. Standard reference strain S. dysenteriae 60R and the E. coli lysogen C600(H19B) were gifts from Stephen B. Calderwood. S. dysenteriae 1 strains 109, 118 and 133 were gifts from Edward T. Ryan. These strains were isolated in 1997 from patients with dysentery at the International Centre for Diarrhoeal Disease Research, Bangladesh. Standard plasmid cloning vectors pUC18 and pBR328 were from laboratory stock.
All strains were maintained at −70°C in Luria broth (LB) medium (Sambrook et al., 1989), containing 15% glycerol. LB medium or lambda broth were used for bacterial growth. Ampicillin (100 μg ml−1), streptomycin (100 μg ml−1), chloramphenicol (25 μg ml−1) or spectinomycin (25–2000 μg ml−1) was added as appropriate.
Isolation of plasmid and bacterial chromosomal DNA, restriction enzyme digests, agarose gel electrophoresis and transformation of plasmids into E. coli strains were performed according to standard molecular biological techniques (Sambrook et al., 1989). Purified cosmid DNA was prepared from overnight growth of HB101(pNAS2) with the Qiagen Plasmid Midi Kit (Qiagen), using the manufacturer's Very Low-Copy Plasmid protocol. Southern hybridizations were performed using Hybond N+ membranes with the ECL direct nucleic acid labelling and detection system (Amersham Pharmacia Biotech), according to the manufacturer's protocol. Restriction enzyme-digested chromosomal and plasmid DNA fragments were separated on 1% agarose gels; required fragments were cut from the gel under ultraviolet illumination and purified using GenElute spin columns (Supelco). The polymerase chain reaction was carried out using a standard DNA minicycler (M.J. Research) with primers listed in Table 1. The location of primers is shown schematically in Fig. 1. For long-range PCR, high GC content 33-mer oligonucleotides were designed, and amplification was performed using TaKaRa Ex Taq DNA polymerase (PanVera), using LA buffer and according to the manufacturer's amplification conditions (a first cycle at 94°C for 1 min; 30 cycles at 98°C for 20 s and 68°C for 10 min; and a final cycle at 72°C for 10 min).
Table 1. . Primers used.
Primer walking-directed sequencing was performed at the DNA Sequencing Core Facility at Tufts Medical School. Each PCR sequencing reaction contained 1.5 μg of template DNA, 50 pmoles of primer and 16 μl of terminator sequencing reaction mix (ABI PRISM) Big Dye Terminator Cycle Sequencing Ready Reaction Kit with Amplitaq DNA Polymerase, FS (Perkin Elmer), for a total of 40 μl. Oligonucleotides were synthesized, deprotected, dried and resuspended in distilled water to 50 μM. Cycling was performed in a GeneAmp 9600 thermal cycler (Perkin Elmer), according to the instructions in Perkin Elmer's protocol P/N 402078, with an additional initial denaturation step. The thermal cycler was heated to 96°C before inserting the tubes. These were kept at 96°C for 4 min, followed by 25 cycles of 10 s at 96°C, 6 s at 50°C and 4 min at 60°C. After cycling, the reactions were purified by Sephadex G-50 spin columns, dried and run on a 373 DNA Sequencer, Stretch (Applied Biosystems) using 48 cm wells to read plates (FMC Long-Ranger, 5% acrylamide gel).
Nucleotide and derived amino acid sequences were assembled and analysed using dna strider software version 1.0 (Christian Marck, Commissariat a l'Energie Atomique, France) and sequencher software (Gene Codes). Database comparisons were performed with the basic local alignment search tool (blast) program (Altschul et al., 1990), using the server at the National Center for Biotechnology Information.
These sequence data have been submitted to the GenBank database under accession number AF153317.
Densitometry analysis was performed using scanning radiographs and performing analysis with a Power Macintosh G3 computer using the public domain nih image program (v. 1.61) (developed at the U.S. National Institutes of Health and available on the Internet at http://rsb.info.nih.gov/nih-image/).
Construction of a chloramphenicol-marked stxAB on bacteriophage H19B
A 945 bp internal stxA′B fragment with SacI and Sph I ends was amplified, using PCR with primers 171 and 172, from the E. coli lysogen C600(H19B). The PCR product was ligated into the SacI–SphI sites of plasmid pCVD442 to give plasmid pVLH5. Plasmid pCVD442 is a suicide vector, containing the pir-dependent R6K replicon, ampicillin resistance and the sacB gene from Bacillus subtilis, thus allowing positive selection for allelic exchange (Donnenberg and Kaper, 1991). A 960 bp chloramphenicol resistance cassette with PstI ends (amplified using PCR with primers JRBCm1 and JRBCm2 from plasmid pBR328) was inserted into the compatible Nsi I site within stxA′B, to create plasmid pVLH7. pVLH7 was transformed into strain SM10λpir, and crossed with a streptomycin-resistant strain of C600(H19B). Streptomycin-, ampicillin- and chloramphenicol-resistant merodiploids were subsequently grown overnight in LB media without antibiotic selection, then plated on LB media with chloramphenicol and 10% sucrose but without NaCl and grown at 30°C for 30 h, thereby selecting for clones that had deleted the integrated sacB gene (Blomfield et al., 1991). Sucrose-resistant colonies that were ampicillin-susceptible and chloramphenicol-resistant were confirmed, using Southern hybridization and PCR analysis, to have undergone marker exchange with insertion of the chloramphenicol resistance cassette within stxA (data not shown).
Isolation, quantification and propagation of bacteriophages
Bacteriophages were induced from lysogens by overnight growth of bacterial strains in LB medium with chloramphenicol and 500 ng ml−1 mitomycin C. Bacteriophage stocks were prepared by centrifuging 1 ml of bacterial growth to remove cellular debris, adding 2 drops of chloroform to the supernatant, vortexing and filter sterilizing. To determine phage titre, dilutions of bacteriophage stock were added to 100 μl of overnight growth of E. coli C600 and incubated at 37°C for 20 min. The mixture was added to 3 ml of lambda top agar at 51°C, poured onto lambda plates, and incubated at 37°C overnight to allow the formation of plaques.
To propagate bacteriophages and determine the proportion of colonies lysogenized, host strains were grown overnight in lambda broth with 0.2% maltose and 10 mM MgSO4. One hundred microlitres of overnight growth was incubated with 100 μl of a bacteriophage stock containing 109 bacteriophages per ml at a multiplicity of infection of 1:1. The mixture was added to 5 ml of LB broth and grown overnight. Serial dilutions were plated onto both LB agar and LB agar with chloramphenicol. The proportion of colonies lysogenized was calculated by dividing the number of chloramphenicol-resistant cfu ml−1 by the total cfu ml−1 for each strain.
To determine the stability of the lysogens, lysogens were grown overnight in LB with chloramphenicol. One millilitre of overnight growth was centrifuged and the cell pellet was resuspended in 1 ml of fresh LB, then diluted 1:1000 into LB medium. After an additional 24 h of growth, serial dilutions of the cultures were plated onto LB agar with and without added chloramphenicol. The stability of the lysogens was calculated by dividing the number of chloramphenicol-resistant cfu ml−1 after overnight growth in antibiotic-free medium by the total number of cfu ml−1.
We thank Alison O'Brien, Philippe J. Sansonetti, Marcia Goldberg, Stephen B. Calderwood and Edward T. Ryan for their gifts of plasmids and bacterial strains; Veronica L. Hackethal and Elizabeth Chao for their technical help; and Ka Ly and Michael Berne of the DNA Sequencing Core Facility at Tufts Medical School for outstanding advice and sequence editing. We are grateful to Yvette M. Murley, Jaideep Behari and Stephen B. Calderwood for their critical reading of the manuscript. This work was supported in part by Public Health Service Grant AI01386 (J.R.B.) and the Claflin Distinguished Scholar Award from the Massachusetts General Hospital (J.R.B.).