The Alleles of the bft Gene Are Distributed Differently among Enterotoxigenic Bacteroides fragilis Strains from Human Sources and Can Be Present in Double Copies

Abstract

Enterotoxigenic Bacteroides fragilis (ETBF) strains are associated with diarrheal disease in children. These strains produce a zinc metalloprotease enterotoxin, or fragilysin, that can be detected by a cytotoxicity assay with HT-29 cells. Recently, three different isoforms or variants of the enterotoxin gene, designated bft-1, bft-2, and bft-3, have been identified and sequenced. We used restriction fragment length polymorphism analysis of the PCR-amplified enterotoxin gene to detect the isoforms bft-1 and bft-2 or bft-3 borne by ETBF. By sequencing the portion of the bft gene corresponding to the mature toxin in some strains and applying allele-specific PCR for strains categorized as bft-2 or bft-3, we found in our collection two strains harboring bft-3, a variant that had been described for isolates from East Asia. Analysis of 66 ETBF strains from different sources showed that bft-1 is the most frequent allele, being present in 65% of isolates; it is largely predominant in isolates from feces of adults, while bft-2 is present in isolates from feces of children. This association is statistically significant (P, 0.0064). Sixteen strains were examined by Southern hybridization using, as probes, the bft and second metalloprotease genes, both included in a pathogenicity islet. Five strains were found to harbor double copies of both genes, suggesting that the whole islet was duplicated. Four of these strains, harboring bft-1 (three strains) or bft-2 (one strain), were found to produce a large amount of biologically active toxin, as determined by a cytotoxicity assay with HT-29 cells. The strains harboring bft-3, either in a single copy or in double copies, produced the smallest amount of toxin in our collection.

Bacteroides fragilis is the anaerobic microorganism most frequently isolated from infectious processes in humans (6) as well as a common component of the normal flora of the colon (7). Some of the strains belonging to this species are able to produce an enterotoxin and are therefore termed enterotoxigenic B. fragilis (ETBF) strains (17). ETBF strains are responsible for diarrheal diseases in young farm animals, including lambs, calves, and foals (17–19). Several studies have also shown an association between the isolation of ETBF from feces and acute diarrhea in children 1 to 5 years old (27, 28, 30). In addition, ETBF can be recovered from the normal fecal flora of healthy subjects, especially adults (25). In some of the subjects harboring ETBF, the enterotoxin is present in the feces in a biologically active form and can be detected by a cytotoxicity assay (26).

B. fragilis enterotoxin, recently termed fragilysin (20), has been characterized as a 20-kDa zinc-dependent metalloprotease belonging to the metzicins family, precisely to the subfamily that comprises the eukaryotic collagenases or matrixins (15). Recently, the target of fragilysin has been identified as the cell surface protein E-cadherin, which is the principal structural component of the zonula adherens and is responsible for cell-cell adhesion of eukaryotic cells (32).

Two groups of investigators have independently cloned and sequenced the enterotoxin gene from two ETBF strains and have identified two different allelic forms of this gene. The first published sequence of the enterotoxin gene, bftP (or bft-1), was obtained from strain VPI 13784, a lamb isolate (12). The other isoform, bft-2, was sequenced from a porcine isolate (8). Both isoforms code for a protein much larger than the mature enterotoxin, a “preprotoxin” of 45 kDa comprising a signal peptide and a protoxin from which the mature toxin is released upon cleavage. Although the two forms of the enterotoxin gene are 95% identical, they show lower homology in the region corresponding to the mature toxin moiety (8). New allelic forms of the enterotoxin gene were found in ETBF isolated from blood in Korea (5) and in fecal isolates from Japan (N. Kato, Final Program of the 2nd World Congress on Anaerobic Bacteria and Infections, abstr. 5.002, p. 74, 1998) and were termed Korea-bft and bft-3, respectively. As the corresponding nucleotide sequences are identical, we refer to the third allelic form of the enterotoxin gene as bft-3 throughout this paper. The bft-3 isoform shares high similarity with the other two forms but is more related to bft-2 (96% identity) (5). Its frequency among ETBF strains outside East Asia remains unknown.

Moncrief et al. (16) found that the B. fragilis enterotoxin gene is located in a pathogenicity islet together with a putative second metalloprotease (MP II) gene which has a low identity (28%) with the enterotoxin gene. The pathogenicity islet of B. fragilis is a genetic unit that shares some characteristic features with the larger pathogenicity islands present in pathogenic strains of Escherichia coli, Salmonella, or Helicobacter pylori. It contains virulence genes (for the enterotoxin and the putative virulence factor MP II), has a lower G+C content than the B. fragilis chromosome, contains almost identical direct repeats in proximity to its ends, and is inserted in the same chromosomal region in different strains (16).

In a previous study, we used PCR to amplify a portion of the bft gene and showed that this gene is present in all ETBF strains but not in nonenterotoxigenic strains (24). In the present study, using PCR-restriction fragment length polymorphism (RFLP) analysis, an allele-specific PCR, and sequencing data, we have shown the frequency of the two principal isoforms, bft-1 and bft-2, in a large collection of ETBF strains of human origin. We have also found that the variant bft-3, although rare, is present in ETBF isolated in Europe. Moreover, we have found a duplication of both the bft and MP II genes in five strains that probably represents a duplication of the entire islet.

MATERIALS AND METHODS

Bacterial strains.

Sixty-eight ETBF strains from different sources were studied; these comprised 28 strains isolated from clinical (extraintestinal) samples, 23 strains isolated from the feces of children (18 with diarrhea), and 17 strains isolated from the feces of adults (10 with diarrhea). The extraintestinal isolates came from the collection of the Istituto Superiore di Sanità and included isolates obtained from Italy (18 strains), the United Kingdom (6 strains), and the United States (4 strains) and characterized in previous studies (22, 23). The fecal isolates were obtained in Italy during previous (22, 25) and ongoing studies on the prevalence of ETBF in the country and were isolated in different geographical areas over a span of 5 years. The reference strain ATCC 43858, originally isolated in the United States from the feces of an infant with diarrhea, was also included. The bacterial strains were identified as ETBF by previously described methods, including a cytotoxicity assay with HT-29 cells (22) and PCR amplification of an internal fragment of the enterotoxin gene (24). Strain VPI 13784 (a gift from T. D. Wilkins, Virginia Polytechnic Institute and State University, Blacksburg) was included in the study as the prototype of the bft-1 genotype. The nontoxigenic strain B. fragilis NCTC 9343 was used as a negative control.

Chemicals and enzymes.

Nylon transfer membranes (Hybond N) were obtained from Amersham International (Little Chalfont, Buckinghamshire, United Kingdom). Restriction endonucleases were purchased from New England Biolabs (Beverly, Mass.) and Roche Molecular Biochemicals (Milan, Italy). For PCR amplification, DynaZyme II (obtained from Finnzyme, Oy, Finland) or AmpliTaq (obtained from PE Applied Biosystems, Roche Molecular Systems, Branchburg, N.J.) DNA polymerase was used. The oligonucleotide primers were synthesized at Laboratori Genenco, M-Medical, Florence, Italy. Electrophoresis-grade agarose and low-melting-point agarose were purchased from Gibco BRL (Life Technologies Italia, San Giuliano Milanese, Milan, Italy); NuSieve was obtained from FMC BioProducts (Rockland, Maine); and agarose D-5, used for pulsed-field gel electrophoresis (PFGE), was obtained from Hispanagar (Burgos, Spain). Proteinase K and lysozyme were obtained from Roche. All other reagents and chemicals were purchased from Sigma Chemical Co. (St. Louis, Mo.). The media for bacterial growth were obtained from Oxoid (Basingstoke, United Kingdom).

PCR-RFLP analysis.

ETBF strains were processed for PCR amplification as previously described (24). Briefly, the bacterial cells were boiled for 10 min, centrifuged at 13,000 × g for 5 min, and stored at −20°C until used as templates (2 μl in each PCR). All PCRs were performed with a GeneAmp PCR System 9600 (Perkin-Elmer) and a reaction volume of 100 μl, containing a 0.5 μM final concentration of each oligonucleotide, 200 μM each deoxynucleoside triphosphate, and 2.5 U of DynaZyme II Taq polymerase. Each amplification was preceded by 5 min at 94°C and consisted of 35 cycles of 60 s at 94°C, 60 s at 56°C, and 120 s at 72°C, followed by a final 5 min at 72°C.

For PCR-RFLP, the primer pair used, BF5-BF6 (Table 1), was designed to amplify a 976-bp internal fragment of the three isoforms of the enterotoxin gene. The PCR amplification products were purified using a QIAquick PCR Purification Kit (QIAGEN GmbH, Hilden, Germany) and then digested with the restriction enzymes AccI and DraI according to the manufacturers' recommendations. These enzymes were chosen because they have different recognition sites in the isoforms bft-1 and bft-2 or bft-3 and therefore generate segments of different lengths. The sequence identity between bft-2 and bft-3 in these recognition sites does not allow them to be distinguished. According to the published sequences, digestion of bft-1 with DraI yields two fragments (640 and 336 bp), and digestion of bft-2 or bft-3 yields three fragments (561, 336, and 79 bp); restriction of bft-1 with AccI yields three fragments (701, 213, and 62 bp), and restriction of bft-2 or bft-3 yields two fragments (914 and 62 bp). The digestion products were separated in a 3% agarose gel (2% NuSieve, 1% agarose) containing 0.5 μg of ethidium bromide per ml.

TABLE 1.

Oligonucleotide primers used to amplify the bft and MP II genes

Primer	DNA sequence	Accession no.a
BF5	5′-GATGCTCCAGTTACAGCTTCCATTG-3′	U67735
BF6	5′-CGCCCAGTATATGACCTAGTTCGTG-3′	U67735
MP1	5′-CAGCATGTGCTGACGATCTT-3′	AF038459
MP2	5′-ATCCACATGTTCCGCTCCTA-3′	AF038459
BFTF	5′-CCTCAAGTACCTCATGGAAT-3′	U67735
BFTR	5′-ATTCCATTAATCGAACTTCG-3′	U67735
BFT2R	5′-TTGGATCATCCGCATGCCT-3′	U90931
BFT3R	5′-TTGGATCATCCGCATGGTT-3′	AF081785

^aPrimers were derived from the published sequences in the GeneBank database under the indicated accession numbers. 

Allele-specific PCR.

In order to distinguish between bft-2 and bft-3, we devised allele-specific PCR assays. For bft-2, the primers used were BF5 and BFT2R (Table 1). BFT2R was designed based on the oligonucleotide sequence specific for bft-2, mapping in a gene region divergent from both bft-1 and bft-3, according to Franco et al. (8). For bft-3, the pair used was BF5-BFT3R (Table 1); BFT3R was designed in the same position as BFT2R but with a 2-base substitution at the 3′ end, according to the sequence of bft-3 (5). The PCR conditions used were the same as those described for PCR-RFLP analysis.

Sequencing of the bft gene.

The portion of the enterotoxin gene corresponding to the whole mature moiety was amplified using the primer pair BFTF-BFTR. These primers were designed based on consensus regions for the three isoforms (Table 1).

Sequencing reactions were performed with the PCR products as templates and with a Perkin-Elmer ABI 370A DNA Sequencer and an ABI Prism Dye Terminator Cycle Sequencing Ready Reaction Kit (PE Applied Biosystems).

DNA extraction and Southern hybridization.

B. fragilis cells were pelleted from 5 ml of an overnight culture in Wilkins-Chalgren broth. Chromosomal DNA was extracted as follows. Lysis was performed with 50 mM Tris–50 mM EDTA buffer (pH 8.0) containing 50 mM glucose, 5 mg of lysozyme per ml, 40 μg of proteinase K per ml, and 10 μg of RNase A per ml at 37°C for 30 min. The DNA was extracted three times with phenol-chloroform, the aqueous phase was collected, and the DNA was recovered by ethanol precipitation and resuspended in water. The DNA was digested with restriction enzymes according to the manufacturers' recommendations, run on an 0.8% agarose gel at 30 V for 20 h, and transferred to a nylon membrane by capillary blotting.

Two probes were used; both were generated by amplification of the chromosomal DNA of VPI 13784. The bft probe is the product of the primer pair BF5-BF6; the MP II probe is the product of the primer pair MP1-MP2 and corresponds to a 1,050-bp internal fragment of the MP II gene. Probe labeling and filter hybridization were performed using a nonradioactive method based on enhanced chemiluminescence (Amersham).

PFGE.

For the preparation of genomic DNA suitable for PFGE, ETBF strains were grown in 10 ml of Wilkins-Chalgren broth to late log phase (optical density at 600 nm, 0.8 to 0.9). The bacterial cells were placed in agarose plugs and lysed by standard procedures (13).

Digestion of the DNA-containing agarose plugs was performed with the restriction enzyme NotI (1) by use of a 150-μl reaction mixture containing bovine serum albumin (100 μg/ml), NotI (20 U), and the buffer provided by the manufacturer; the mixture was incubated at 37°C overnight.

The restricted chromosomal DNA was separated on 1% agarose gels using the CHEF Mapper Pulsed-Field Electrophoresis System (Bio-Rad Laboratories, Milan, Italy). Electrophoresis was performed in the two-state mode with a 120° pulse angle at 5.4 V/cm for 34 h. The switch times were increased from 1 to 40 s by the ramping factor 0.357. Following electrophoresis, the gels were stained with ethidium bromide, and the DNA bands were photographed under UV light. For Southern hybridization, DNA was transferred to nylon membranes by vacuum blotting, and hybridization was performed as described above.

RESULTS

Detection of the enterotoxin gene alleles.

With the primer pair BF5-BF6, the expected product of 976 bp was amplified from all of the ETBF strains examined. The nonenterotoxigenic strain NCTC 9343 yielded no amplification product (data not shown).

By PCR-RFLP, we were able to classify the enterotoxin genes of all the ETBF strains examined as belonging to isoform bft-1 or isoform bft-2 or bft-3, as the observed sizes of the fragments generated agreed with the predicted sizes of the fragments and no anomalous profile was observed (Fig. 1). Strain VPI 13784 was found to harbor the isoform bft-1, as expected, while strain ATCC 43858 was found to harbor the isoform bft-2; these findings were subsequently confirmed by sequencing. These two strains were therefore used as prototypes of the two isoforms throughout the study.

FIG. 1.

Agarose gel electrophoresis of the bft gene PCR products of ETBF strains and of the RFLPs obtained with DraI or AccI digestion. Lanes: 1, strain VPI 13784; 2, ATCC 43858; 3, PR 838; 4, MT 2. U, unrestricted PCR products; D, PCR products digested with DraI; A, PCR products digested with AccI. M, molecular size markers. In accordance with the fragment lengths indicated in the text, the isoforms bft-1 and bft-2 or bft-3 can be distinguished in lanes 1 and 3 and in lanes 2 and 4, respectively. The lower-molecular-size fragments are not visible. By sequence analysis and isoform-specific PCR, ATCC 43858 was shown to harbor bft-2 and MT 2 was shown to harbor bft-3.

To confirm the results of PCR-RFLP, the portion of the bft gene corresponding to the mature toxin of 14 ETBF isolates, 9 with bft-1 and 5 with bft-2 or bft-3, was analyzed by sequencing. The nucleotide sequence of the bft gene of 13 strains showed complete identity with the published sequence of either bft-1 or bft-2, and the isoforms deduced by sequencing were in accordance with the PCR-RFLP results. However, strain MT 2 showed a number of divergences from the sequences of both isoform bft-1 and isoform bft-2, while it showed 100% identity with the sequence of bft-3.

In order to verify whether in our series other strains harbored bft-3, using an assay simpler than sequencing, we devised PCR assays specific for bft-2 and bft-3. In the bft-2-specific PCR, strain ATCC 43858 yielded the expected amplification product, while neither VPI 13784 (bft-1) nor MT 2 (bft-3) yielded any amplification product. With the bft-3-specific PCR, MT 2 yielded the expected amplification product (Fig. 2). All the strains classified as bft-2 or bft-3 by PCR-RFLP were submitted to the bft-2-specific PCR. Besides MT 2, another strain, UK 5312, was negative in the amplification reaction for bft-2 but yielded an amplification product in the reaction for bft-3. Sequencing analysis confirmed these findings.

FIG. 2.

Agarose gel electrophoresis of the amplification products obtained with allele-specific PCR for the isoforms bft-2 and bft-3. Lanes: 1, VPI 13784; 2, ATCC 43858; 3, B 115; 4, MT 2; 5, UK 5312. M, molecular size markers.

Distribution of the enterotoxin gene isoforms.

Among the ETBF isolates examined, bft-1 was found to be more common than bft-2, as it was detected in two-thirds of the strains. The bft-1 isoform was found in almost all isolates from the feces of adults, including individuals with and without diarrhea; only 1 strain out of a total of 17 was found to harbor the bft-2 isoform. However, the bft-2 isoform was as common as the bft-1 isoform in strains isolated from the feces of children. The association between isoform bft-2 and strains from children is statistically significant compared to that between bft-2 and adult strains (P, 0.0064) (Table 2). When strains isolated from children with diarrhea (nine strains bearing bft-1 and nine bearing bft-2) are compared to strains isolated from adults with diarrhea (nine strains bearing bft-1 and one bearing bft-2), the association is statistically significant (P determined by the Fisher exact test, 0.04). We found that the bft-3 isoform, originally described for isolates from Korea and Japan, also was present in our collection of strains originating from Europe and the United States, although much less frequently than the other two isoforms. Strain MT 2 was isolated from the stools of a healthy 4-month-old Italian baby in 1994; strain UK 5312 was originally isolated from a blood culture in the United Kingdom.

TABLE 2.

Distribution of enterotoxin gene isoforms (bft-1 or bft-2) according to the origin of the ETBF isolates

Source	No. of isolates tested	No. of isolates with the following enterotoxin gene allele:
		bft-1	bft-2a
Extraintestinal site	27	17	10
Feces
Children	22	12	10*
Adults	17	16	1**
Total	66	45	21

^a* versus **, chi-square test value, 7.42; P, 0.0064. 

Southern analysis of the bft and MP II genes.

We analyzed 16 ETBF strains by Southern hybridization: 10 harboring bft-1, 4 harboring bft-2, and 2 harboring bft-3 (Table 3). The intent was to hybridize the chromosomal DNA with probes for the bft and MP II genes to define whether these two genes were consistently associated in all strains, irrespective of the isoform borne. Prior to the experiments, controls were run to verify that the two probes were specific for the respective genes and that there was no cross-hybridization.

TABLE 3.

Characteristics of 16 ETBF strains from different sources, examined by a cytotoxicity assay, sequencing of the bft gene, and Southern analysis

Strain	Sourcea	Cytotoxicity titerb	bft gene
			Allele	No. of copies
VPI 13784	Intestinal, lamb	2.6 ± 0.3	bft-1	1
PR 7	Extraintestinal	2.7 ± 0.2	bft-1	2
PR 254	Extraintestinal	2.2 ± 0.4	bft-1	1
B 69	Extraintestinal	2.7 ± 0.2	bft-1	2
CM 14	Extraintestinal	2.4 ± 0.3	bft-1	1
PR 823	Intestinal, adult	2.3 ± 0.2	bft-1	1
PR 838	Intestinal, adult	2.6 ± 0.4	bft-1	1
PR 850	Intestinal, adult	2.2 ± 0.2	bft-1	1
MS 1	Intestinal, infant	3.0 ± 0.2	bft-1	1
LZ 4	Intestinal, infant	3.1 ± 0	bft-1	2
ATCC 43858	Intestinal, infant	2.0 ± 0.2	bft-2	1
AN 209	Intestinal, infant	2.7 ± 0.2	bft-2	1
B 45	Extraintestinal	2.3 ± 0.2	bft-2	1
B 115	Extraintestinal	2.7 ± 0.2	bft-2	2
MT 2	Intestinal, infant	2.0 ± 0.2	bft-3	2
UK 5312	Extraintestinal	1.6 ± 0	bft-3	1

^aSources were human, unless otherwise indicated. 

^bExpressed as the logarithm (log₁₀) of the reciprocal of the highest dilution of culture supernatant giving a 50% cytopathic effect. The geometric mean ± standard deviation of four separate experiments is reported. 

The chromosomal DNA was digested with the restriction enzyme EcoRI. Since this enzyme has no recognition site in the bft gene, in the majority of the strains the bft probe hybridized with a single band, as expected. This band had a different molecular size in each strain, ranging from approximately 9 to 25 kb. However, in five strains, the bft probe hybridized with two bands (Fig. 3A). As the presence of an additional EcoRI site internal to the bft gene was ruled out by analysis of the bft sequences, this finding was suggestive of the presence of double copies of the gene. In the same strains, two hybridizing bands also were obtained when the chromosomal DNA was digested with BamHI and SalI, which have no restriction sites the bft gene (data not shown).

FIG. 3.

Southern hybridization analysis of B. fragilis chromosomal DNA digested with EcoRI. DNA was hybridized with the bft probe (A) or with the MP II probe (B). Lanes: 1, VPI 13784; 2, ATCC 43858; 3, MT 2; 4, AN 209; 5, PR 838; 6, B 45; 7, PR 823; 8, LZ 4; 9, NCTC 9343 (negative control, nontoxigenic strain). The positions of the molecular size standards are indicated.

According to the published sequence of the fragilysin islet, EcoRI has a single recognition site in the MP II gene and a second site between the MP II gene and the left end of the islet (16). As expected, in the majority of the strains, the MP II probe hybridized with two bands. One band, with a high molecular size, corresponded to the band recognized by the bft probe and therefore comprised part of the MP II and bft genes. The smaller reactive band had the same size in all strains (approximately 800 bp, in accordance with the published sequence) and represented the EcoRI fragment internal to the fragilysin islet. In the five strains where the bft probe hybridized with two bands, the MP II probe recognized three bands: two high-molecular-size bands (the same as the bands recognized by the bft probe) and the 800-bp band (Fig. 3B).

To further confirm that some strains harbored two copies of both the bft and the MP II genes and to investigate whether these two copies were closely associated, Southern hybridization was performed following macrorestriction of the chromosomal DNA with an infrequently cutting enzyme (NotI) and separation by PFGE. In the 16 ETBF strains examined, extensive heterogeneity of the NotI restriction patterns was observed, suggesting that ETBF strains are not clonally related (Fig. 4A). The bft gene was found to be located in NotI fragments with different molecular sizes, larger than 500 kb in most strains. In strains with two hybridizing EcoRI bands, we also detected two hybridizing NotI bands with high molecular sizes, confirming the presence of double copies of the bft gene (Fig. 4B). The hybridization patterns obtained with the MP II probe were identical to those obtained with the bft probe (data not shown).

FIG. 4.

PFGE restriction patterns generated by NotI digestion (A) and Southern hybridization with the bft probe (B) of ETBF chromosomal DNA. Lanes: 1, VPI 13784; 2, ATCC 43858; 3, MT 2; 4, AN 209; 5, PR 838; 6, B115; 7, PR 823; 8, B 69; 9, LZ 4. Lanes M₁ and M₂, molecular size markers. The hybridization pattern obtained with the MP II probe was identical to that obtained with the bft probe.

The overall results indicated that five strains possess double copies not only of the bft gene but also of the MP II gene and probably of the entire islet. The duplicated islets are not closely associated in the chromosome, being located in different large NotI fragments. In the strains harboring double copies of the genes, the bft allele appears to be the same in the two copies, as deduced from PCR-RFLP and sequencing results; in our series, three strains bear the allele bft-1, and one strain each bears the alleles bft-2 and bft-3 (Table 3).

Cytotoxin production in an HT-29 cell assay.

In an attempt to find an association between the cytotoxicity titer and the bft allele or the copy number of the bft gene, repeated determinations of cytotoxicity titers in HT-29 cells were obtained for 16 ETBF strains. In strains bearing either bft-1 or bft-2, the titers varied from 2 to 3 log₁₀ units. Both strains bearing bft-3 showed low toxin titers (1.6 and 2 log₁₀ units). Four out of five strains harboring a duplication of the bft gene (three strains bearing bft-1 and one strain bearing bft-2) were found to have cytotoxicity titers in the high range of the series. The only exception was the strain bearing double copies of bft-3, which had a low toxin titer (Table 3). Due to the limited number of strains examined, it was not possible to demonstrate statistically significant differences.

DISCUSSION

Although ETBF strains have been associated with diarrhea in children under 5 years of age (27, 28, 30), the pathophysiological mechanisms linking enterotoxin production to diarrhea are not completely understood. B. fragilis toxin, or fragilysin, cannot be considered a classical enterotoxin, being a metalloprotease similar to eukaryotic collagenases (15). In vitro and in vivo studies have shown that this toxin is able to damage the intestinal mucosa (both ileal and colonic) of various animal species, including humans (21, 26, 29), and to elicit fluid accumulation through increased permeability (26) and active chloride secretion (4). Recently, the zonula adherens protein E-cadherin has been recognized as the substrate for B. fragilis enterotoxin. It has been hypothesized that cleavage of the extracellular domain of E-cadherin can lead to alteration of the cytoskeletal structures and to increased intestinal permeability. This hypothesis represents a novel mechanism of action for a bacterial enterotoxin (32).

The presence of ETBF in the gut does not necessarily indicate disease, as the organism has been frequently isolated from the feces of healthy individuals. Asymptomatic carriage is particularly high in adults, but children also can harbor both the microorganism and the toxin without intestinal disturbances (25). This finding suggests that other factors that can be related to either the host or the microorganism are necessary for the development of diarrhea. A recent study has shown unresponsiveness of the colon mucosa of some subjects to B. fragilis toxin (29).

ETBF can produce different isoforms of the enterotoxin, the most common being those encoded by the bft-1 and bft-2 genes (8, 12). Although these isoforms display the same biological activities, Wu and coworkers have suggested that their potencies might be different (32). We examined several ETBF strains isolated from different sources (extraintestinal infections and feces of adults and children) by PCR-RFLP analysis to distinguish between the isoforms bft-1 and bft-2 or bft-3 and subsequently by an isoform-specific PCR to distinguish between bft-2 and bft-3. We found that the majority (65%) of the strains investigated harbor the bft-1 isoform. This allele is largely predominant in strains from the feces of adults (with or without intestinal symptoms), while strains isolated from children harbor either bft-1 or bft-2. The rarity of the isoform bft-2 in adults indicates that strains bearing this isoform are more apt to colonize (and consequently induce diarrhea in) children than adults. The factors responsible for this association are unknown, but they might consist either of different properties of the bft-2 toxin itself or of characteristics of the strains which allow better proliferation in the colon of children than in that of adults.

By sequence analysis and an allele-specific PCR, we found that two strains in our collection harbor the isoform bft-3, originally described for blood isolates from Korea (5) and for fecal isolates from Japan (Kato, Final Program of the 2nd World Congress on Anaerobic Bacteria and Infections). The two strains in our collection were isolated in the United Kingdom and in Italy; this finding indicates that the bft-3 allele, although rare, is present in geographical areas outside East Asia.

Interestingly, the sequences of the bft genes examined were 100% identical to the published nucleotide sequences of the three alleles, without a single base substitution. Although this observation is limited to the portion of the genes coding for the mature toxin, this lack of variation suggests that bft is a recent acquisition of B. fragilis, as already proposed by Smith and Callihan (31), and that each isoform is conserved because of an evolutionary advantage. The amino acid substitutions among the three deduced proteins are not abundant: bft-2 toxin diverges from bft-1 toxin in 25 amino acids out of 186, and bft-3 (which is more similar to bft-2 than to bft-1) diverges from bft-1 in 20 amino acids and from bft-2 in 8 amino acids. However, these substitutions cluster in two regions adjacent to the active site of the metalloprotease and the zinc-binding motif (8). Therefore, the three variants could exhibit subtly different receptor-binding preferences that could result in differences in host range and/or pathogenic potential. Similar differences have been described for other allelic proteins. For instance, the alleles speA1 and speA3 of Streptococcus pyogenes code for toxins which differ only in one amino acid; however, the product of speA3 has significantly greater mitogenic activity and affinity for the class II major histocompatibility complex and is associated with clinical cases of streptococcal toxic shock syndrome (11). The papG alleles of Escherichia coli, which code for variant forms of the P adhesin, are associated with different clinical syndromes, such as pyelonephritis and cystitis (10).

In our collection, the two strains carrying bft-3 were found to produce low levels of biologically active toxin in the HT-29 cell cytotoxicity assay. One explanation is that a smaller amount of the protein is produced. An alternative possibility is that the bft-3 toxin is less active on the cell system used, although Chung et al. have demonstrated that the purified bft-3 toxin from Korean isolates cleaves E-cadherin at a concentration similar to that observed with the other purified enterotoxins (5).

We cannot rule out the possibility that other enterotoxin variants with substitutions in areas not explored by the RFLP and PCR assays used in this study exist. Sequencing of more strains from different sources and geographical regions is necessary for a comprehensive analysis of the frequency of the different enterotoxin alleles.

Following the recent discovery of the pathogenicity islet of B. fragilis, which includes the fragilysin gene, B. fragilis has been added to the growing list of microorganisms carrying pathogenicity islands. This islet is much smaller than classical pathogenicity islands. It comprises only the enterotoxin gene and the MP II gene and lacks the genes coding for secretion systems necessary for the delivery of the toxin directly to target cells (14). However, the B. fragilis islet shares with the larger structures the property of transforming a typical commensal organism into a virulent organism (9).

An unexpected finding was that some ETBF strains possess double copies of the enterotoxin gene, associated with double copies of the MP II gene and possibly of the entire islet. Bacteria harboring more than one pathogenicity island have been described before: for instance, in the same strain of uropathogenic E. coli serotype O4, PAI I and PAI II are present together. However, the virulence genes carried by the two islands are different (2). In some Helicobacter pylori type I strains, the cag pathogenicity island appears as two regions separated by a long stretch of chromosomal DNA, containing different genes and probably derived from rearrangement driven by insertion sequences (3).

For B. fragilis, the whole islet, including the enterotoxin and MP II genes, appears to have undergone duplication in some strains, as deduced from Southern analysis; consequently, the two islets carry the same enterotoxin gene isoform, as confirmed by PCR-RFLP and sequencing. The islet duplication appears to be quite stable, as the Southern hybridization patterns were reproducible when the strains were examined after storage and repeated subculturing. Interestingly, with the exception of the bft-3 strain, the cytotoxicity titers obtained from supernatants of strains with double copies of the gene were among the highest in our series; however, the limited number of strains examined does not allow conclusions to be drawn.

Our observations indicate that there are complex microbiological properties of ETBF that need to be elucidated for a better understanding of the pathogenic potential of this microorganism.