|Home | About | Journals | Submit | Contact Us | Français|
Summary: Enterotoxigenic Bacteroides fragilis (ETBF) strains are strains of B. fragilis that secrete a 20-kDa heat-labile zinc-dependent metalloprotease toxin termed the B. fragilis toxin (BFT). BFT is the only recognized virulence factor specific for ETBF. ETBF strains are associated with inflammatory diarrheal disease in children older than 1 year of age and in adults; limited data suggest an association of ETBF colonization with inflammatory bowel disease flare-ups and colorectal cancer. ETBF secretes one of three highly related BFT isoforms. The relationship between BFT isoform and disease expression is unknown. Although the mechanism of action of BFT is incompletely understood, available data suggest that BFT binds to a specific intestinal epithelial cell receptor, stimulating intestinal cell signal transduction pathways that result in cell morphology changes, cleavage of E-cadherin, reduced colonic barrier function, and increased epithelial cell proliferation and cytokine expression (such as the proinflammatory chemokine interleukin-8). Together, the data suggest that in some hosts, ETBF acts via secretion of BFT to induce colitis. However, the full spectrum of clinical disease related to ETBF and the impact of chronic ETBF colonization on the host remain to be defined.
Bacteroides species comprise nearly half of the fecal flora community and are host symbionts critical to host nutrition (e.g., Bacteroides thetaiotaomicron) and mucosal and systemic immunity (e.g., B. fragilis) (25, 35, 54, 55, 73, 98, 110, 127). Among Bacteroides species, B. fragilis strains are opportunistic pathogens, being the leading anaerobic isolates in clinical specimens, bloodstream infections, and abdominal abscesses despite comprising typically <1 to 2% of the cultured fecal flora (50, 62, 75, 87, 91). In 1984, while investigating the etiology of lamb diarrheal disease, Myers and colleagues provided the first evidence that certain strains of B. fragilis were epidemiologically associated with diarrheal disease (64). Studies by the same investigators revealed that both the isolates and their sterile culture supernatants stimulated intestinal secretion in lamb ligated intestinal loops (64, 69; see reference 105 for a review). The secretory responses in some ligated intestinal loops were so potent that the loops burst, a response reminiscent of cholera toxin-stimulated secretory responses. The biologically active factor was proposed to be a heat-labile, ~20-kDa protein toxin, now known to be one of a family of B. fragilis toxins (BFTs) (69, 106). B. fragilis strains eliciting intestinal secretion were named enterotoxigenic B. fragilis (ETBF) and their nonsecretory counterparts were termed nontoxigenic B. fragilis (NTBF). This review describes the progress over the subsequent nearly 25 years in defining the role of ETBF in human disease, the genetics and mechanism of action of BFT, and insights into the molecular evolution of ETBF strains.
Table Table11 summarizes the data on ETBF infections in animals.
ETBF strains were initially identified during an investigation of newborn lamb diarrheic disease (64). Subsequent reports indicated that calves, foals, and piglets were susceptible to ETBF-associated diarrheal illnesses in the field (17, 65, 66, 69, 105). In the small number of reports available, the diarrheal illnesses were largely self-limited, with little mortality, except possibly in newborn lambs (64). The limitations of these reports include variable assessment for additional enteric pathogens and a lack of epidemiology to assess ETBF carriage in asymptomatic livestock. Consistent with the native colonic habitat of B. fragilis, exfoliating colitis with an intense neutrophilic mucosal infiltrate was observed in a single piglet examined by histopathology (17).
In early studies, oral inoculation of ETBF into newborn lambs, piglets, or importantly, gnotobiotic (germfree) piglets reproduced the diarrheal illnesses observed in the field, providing further support for the proposal that select B. fragilis strains stimulated intestinal secretion and diarrheal disease (23, 64, 65). Histopathology from gnotobiotic piglets revealed that lesions were most severe in the colon, where crypt hyperplasia and neutrophilic infiltrates were observed. By scanning electron microscopy, the colonic surface epithelium had a cobblestone appearance associated with round, swollen epithelial cells and epithelial cell exfoliation (23). Similar but more variable lesions were also observed primarily in the distal half of the small intestine. No extraintestinal lesions were noted. Additional studies with infant and 2-week-old rabbits as well as adult rabbits with ligated ceca confirmed the enteropathogenicity of ETBF, but in these disease models bloody diarrhea and mortality were frequently observed (17, 65-68). However, ETBF virulence was variable in rabbit models (66, 67), consistent with subsequent observations that sterile culture supernatants of ETBF exhibited variable biologic activity on HT29/C1 cells (a human colonic epithelial tumor cell line) (13, 30, 82, 116, 119). Histopathologic abnormalities in these non-gnotobiotic animal models also occurred only in the distal ileum and colon, with disruption of the epithelial integrity and predominant neutrophilic or mixed neutrophilic and mononuclear cellular infiltrates in the lamina propria; animals colonized with NTBF strains exhibited normal colonic histopathology without inflammation by light microscopy. ETBF adherence and/or invasion of colonocytes was not observed by light or electron microscopy. Bacteremia has not been reported for these animal models (68).
Early studies indicated that mice (suckling and young adult) and hamsters do not exhibit secretory responses to ETBF (68, 69). Recently, colonization of gnotobiotic mice with ETBF, but not NTBF, has been shown to induce acute, sometimes lethal, colitis (71, 92). In contrast, conventional mice colonized with ETBF develop rapid-onset, transient diarrhea lasting 3 to 4 days. Subsequently, conventional mice colonized with ETBF exhibit persistent, asymptomatic colonization, with ongoing histopathologic colitis present for as long as 16 months (90, 92) (Fig. (Fig.1).1). Additional studies show that purified BFT, albeit at a pharmacologic dose (i.e., 10 μg), stimulates secretion and histologic enteritis in mouse ileal loops (42, 44).
In initial studies, ETBF or sterile culture supernatants were injected into ligated intestinal (predominantly jejunal) segments of lambs, calves, or rabbits (64, 69). These studies confirmed that ETBF stimulated intestinal secretion but, more importantly, provided the first evidence that ETBF secreted a heat-labile protein toxin responsible for stimulating intestinal secretion (69). Subsequently, Obiso and colleagues demonstrated that purified BFT (1-μg to 50-μg doses) stimulates dose-dependent secretion of sodium, chloride, and albumin in both ileal and colonic ligated loops in rats, rabbits, and lambs (80). Colonic secretory responses exceeded ileal responses in all species. However, species-dependent BFT potency was observed, with the greatest fluid responses in the ilea of lambs and the colons of rabbits. The histology of ileal or colonic ligated loops inoculated with 20 μg of purified BFT for each species exhibited marked epithelial cell rounding and detachment, neutrophilic inflammation, and in some sections, necrosis and hemorrhage. Time course analyses revealed that histologic changes (at 10 h) preceded detection of intestinal secretion (examined at 18 h). Consistent with the subsequent determination that BFT is a zinc-dependent metalloprotease toxin (see below), metal-chelating agents reduced (>90%) the secretory and histologic responses to purified BFT in ligated intestinal loops and zinc reconstituted them (~50%).
ETBF strains were first isolated from humans with diarrheal illnesses in an uncontrolled study in 1987 (70) examining 10 individuals with diarrhea of unknown origin (from Montana) and 34 infants (2 to 14 months of age; Navajo Area Indian Health Service, Tuba City, AZ). Overall, ETBF isolates were identified in 8 of 44 patients (two adults and six children of <5 years of age), with only one infant positive for a second enteric organism, enterotoxigenic Escherichia coli (ETEC). In this initial study, most patients had self-limited diarrhea, but intermittent diarrhea of more than 3 years duration, diarrhea persisting 4 weeks, high fever, and frank/occult fecal blood were noted in ETBF-positive patients (70). The first human study with control subjects to investigate a putative association between ETBF infection and human diarrheal disease was conducted in the pediatric outpatient clinics of the Apache Indian reservation at Whiteriver, AZ, in 1992 (95). The key results of this study were that children of <1 year of age did not develop ETBF-associated diarrhea, whereas ETBF isolates were associated with diarrheal illnesses in children between the ages of 1 and 5 years. ETBF isolates from single families evaluated by restriction enzyme analysis appeared to be genetically related.
Table Table22 summarizes studies with control subjects evaluating the role of ETBF as an etiologic agent of diarrheal disease. Table Table33 summarizes uncontrolled case series or reports of human ETBF infection. The majority of controlled studies conducted to date have evaluated ETBF as an agent of acute community-acquired diarrheal disease in children (Table (Table2).2). One Swedish study assessed ETBF in adults hospitalized with acute diarrheal disease (128) compared to isolation of ETBF from healthy outpatient control adults. Overall, study designs have been variable, and many studies are hampered by small numbers of patients and controls, a lack of thorough studies for other enteric pathogens, and a wide range of recovery rates (6 to 70%) for B. fragilis.
Of 17 studies with controls (Table (Table2),2), 5 did not show an association between ETBF and diarrheal disease in differing patient populations (4, 8, 48, 77, 84). Of these studies, a follow-up study of the same patient population by the same authors subsequently did show an association of ETBF with diarrheal disease (59a, 84), one study had a B. fragilis isolation rate of only 6% (77), and three studies were from Brazil (4, 8, 48). Excluding the one study (128) focused on adults, all but one (10) of the remaining studies confirmed the original observations of Sack et al. (95) suggesting that, similar to the case with Clostridium difficile, there is no association between ETBF isolation and diarrheal illnesses in children of <1 year old, whereas diarrheal disease due to ETBF emerges after age 1. The mechanisms accounting for the apparent lack of pathogenicity of ETBF in very young children are unknown. However, there is marked variability in the frequencies at which ETBF strains are associated with diarrhea in different geographic locales (from 3.5% in Bangladesh to 28% in Italy). Only one large study to date has investigated ETBF in adults (defined as individuals who are >15 years old), identifying ETBF in 27% of 728 patients with diarrhea and 12% of 194 healthy controls (P < 0.01) (128). In this study, 19% of adults of <30 years of age, 10.6% of adults between the ages of 30 and 60, and 3.7% of adults older than 60 years were asymptomatically colonized with ETBF. Among adults between 15 and 30 years of age, ETBF fecal carriage rates were similar for those with and without diarrhea, whereas an association of ETBF with diarrheal disease emerged in older patients. Other studies also suggest that asymptomatic colonization with ETBF is common (up to ~30%) (7) (Table (Table2;2; also see Table Table5).5). Similarly, 9.3% of 237 B. fragilis strains cultured from sewage influent in Bozeman, MT, were ETBF, suggesting moderately high endemic carriage of ETBF (109). ETBF carriage for up to 16 months with fecal ETBF isolates identical by restriction enzyme analysis has been reported (95). Together, the data suggest that ETBF strains are globally distributed enteric pathogens causing diarrhea in both children and adults. Similar to the case for other enteric pathogens, asymptomatic ETBF colonization is common. Geographic diversity in the recovery of enteric pathogens is also well known (10a), although the reasons underlying the variable isolation rates have not been elucidated. With respect to detection of ETBF in stool, one important variable is likely the sensitivity of the diagnostic approaches used (see “Diagnosis of ETBF Infection”).
ETBF clinical illnesses are typically characterized as self-limited with watery diarrhea; if noted, persistent diarrhea (>14 days) has been reported for a minority (0 to 22%) of patients. Table Table44 summarizes the range of clinical findings reported in studies of ETBF diarrheal illnesses. When assessed, ETBF diarrheal illnesses have been clinically indistinguishable from non-ETBF diarrheal illnesses in the populations studied (10, 95, 128). A recent prospective analysis involving 73 ETBF-infected patients (43 children under age 15 and 30 adults) in Bangladesh (102) correlated clinical findings with more detailed stool analyses to evaluate the pathogenesis of ETBF diarrheal illnesses. In this series, although illnesses were again self-limited (2 to 11 days), substantial abdominal pain (88%), tenesmus (66%), and nocturnal diarrhea (80%) were reported. In contrast, fever (>37.5°C) (7%) and fecal occult blood (8.5%) were infrequent. Leukocytosis was not detected. Stool analyses for leukocytes, fecal lactoferrin, and proinflammatory cytokines (tumor necrosis factor alpha and interleukin-8 [IL-8]) indicated that ETBF induced inflammatory diarrhea compared to stools from asymptomatic control individuals not infected with ETBF. Consistent with the inflammatory mucosal response to ETBF infection, serum immunoglobulin A (IgA) and IgG antibodies as well as fecal IgA to BFT were detected. In one report, two to four distinct, sequential episodes of ETBF-associated diarrhea were reported for children in Bangladesh, suggesting that acquired immunity to ETBF is incomplete (85).
No studies have been designed to specifically determine if ETBF contributes to persistent or chronic diarrhea or to diarrheal illnesses in travelers; limited observations are available on the role of ETBF in diarrhea defined as nosocomial (occurring >3 days after hospital admission) (16, 57) or antibiotic associated (39, 86). In one study, ETBF strains were identified in stools of 9% of patients with non-C. difficile nosocomial diarrhea, which was significantly greater than ETBF detection in outpatient controls (2%; P < 0.04), but hospitalized controls were not evaluated (16). C. difficile and ETBF were detected simultaneously in a small subset of fecal samples from patients with antibiotic-associated diarrhea (57, 86). In one study of children, there was no association of ETBF with antibiotic-associated diarrhea (39). Well-known extraintestinal or intestinal complications of enteric infections, such as reactive arthritis or diarrhea-predominant irritable bowel syndrome, have not yet been reported for ETBF.
Three recent but small reports raise the specter of adverse, long-term consequences of chronic ETBF colonization (Table (Table55 ). In two reports examining patients with irritable bowel disease (IBD) (7, 88), active (but not clinically quiescent) IBD was associated with ETBF infection, although the differences did not reach statistical significance in one study (7). Of note, in the latter study, ETBF isolates were identified by PCR in mucosal washes of ~30% of all individuals examined by endoscopy (7). In an additional prospective, cross-sectional epidemiologic report from Turkey (113), fecal ETBF was isolated from 38% of 56 patients with colorectal cancer but only 12% of 40 sex- and age-matched concurrent controls (P < 0.01).
Several studies, largely conducted in microbiology laboratories, have examined whether ETBF strains are isolated in excess among bloodstream isolates (Table (Table6).6). In six of eight studies, ETBF isolation from blood exceeded isolation from other extraintestinal specimens, but the results reached statistical significance in only one study (38). Only one study (26) examined whether isolation of ETBF from blood was associated with evidence of excess clinical morbidity or mortality, and it did not find an association. However, when the studies are combined, 232 of 1,325 extraintestinal B. fragilis isolates (17.5%) were identified as ETBF, with 86 of 368 blood isolates (23.4%) and 146 of 957 other extraintestinal isolates (15.3%) being ETBF (P < 0.0005; two-sided chi-square test), suggesting that ETBF may be isolated more frequently from bloodstream infections. Additional studies will be necessary to determine if ETBF strains exhibit increased virulence in extraintestinal infections, consistent with their virulence in colonic disease.
Table Table77 summarizes the approaches to diagnosis of ETBF infection. Diagnosis of ETBF infection from stool is difficult. Similar to C. difficile diagnosis, detection of the bft gene or its biologic activity is required to diagnose ETBF colonization or disease. Most studies have used one of two approaches, either B. fragilis fecal isolation with testing for the bft gene (by PCR) or for BFT (in the HT29/C1 cell assay) or direct examination for the bft gene in DNA extracted from stool. The HT29/C1 cell assay detects the biologic activity of BFT in culture supernatants of ETBF strains, demonstrating a sensitivity of 89% and a specificity of 100% compared to ETBF strain identification by the ability to stimulate secretion in the lamb intestinal loop assay (119) (Fig. (Fig.2).2). The HT29/C1 cell assay detects as little as 0.5 pM purified BFT; the half-maximal concentration of BFT detected by the HT29/C1 cell assay is 12.5 pM BFT (97). Analysis of our collection of B. fragilis strains (n = 304 strains) indicates that the presence of the bft gene correlates well with the ability to detect BFT in the HT29/C1 cell assay (29). In rare instances (n = 2 strains), deletions/mutations in the bft gene have been identified that result in a lack of synthesis and/or secretion of biologically active BFT by an ETBF strain (29).
Available data further suggest that overnight anaerobic cultivation of stool in a nutrient broth promotes enhanced recovery of B. fragilis and ETBF (102). BFT biologic activity has been detected directly in fecal supernatants from patients with diarrhea by use of the HT29/C1 cell assay (84). Enzyme immunoassays (EIAs) have been reported for detection of BFT, and in preliminary data, detection of BFT in stool samples by EIA has been reported (89, 116). A report of a combined immunomagnetic separation (to concentrate B. fragilis from stool) and PCR approach (termed IMS-PCR) to detect ETBF was encouraging (sensitivity, 50 CFU ETBF/g stool), but this method requires the use of two anti-B. fragilis antibodies for best sensitivity, and one antibody has been lost (129; A. Weintraub, Karolinska Institute, personal communication). Further commercial development of an EIA to detect fecal BFT or of IMS-PCR has not occurred. Limited comparative data suggest, not unexpectedly, that direct stool-based PCR approaches are more sensitive for the diagnosis of ETBF colonization or disease than fecal cultivation of B. fragilis, followed by detection of bft or BFT (16, 88, 107). Anaerobic stool culture of B. fragilis is affected by delays in processing of samples as well as (likely) the general difficulty of anaerobic microbiology (95, 107). Using stool culture, detection of ETBF is further affected by the fecal heterogeneity of B. fragilis strains, where both NTBF and ETBF may coexist (16), although one report suggested displacement of fecal NTBF by ETBF in diarrheal disease (95).
Together, the data suggest that similar to detection of ETEC, testing of multiple fecal B. fragilis colonies for bft/BFT is necessary to enhance the diagnostic sensitivity if fecal culture is the chosen diagnostic approach. Overall, it remains uncertain which method(s), and in particular which PCR approach, will provide the best diagnostic sensitivity and specificity for detection of ETBF disease or colonization. Table Table88 summarizes the primers utilized to date for detection of the bft gene in B. fragilis isolates or DNA extracted from stool. In some instances, reported primer use has included multiple base pair mismatches, suggesting the possibility of decreased diagnostic sensitivity and specificity (86). Nested PCR is the most sensitive PCR method to detect the bft gene reported to date, detecting 102 to 103 CFU ETBF/g stool (108). The development of rapid, accurate, and sensitive diagnostic testing for ETBF organisms will enhance assessments of the epidemiology of these bacteria and their disease associations as well as being an important prerequisite to consideration of therapeutic trials for ETBF disease.
Because diagnosis of ETBF as an etiology of diarrheal disease is presently limited to the research setting and the diagnosis is often delayed even if sought, no controlled studies have evaluated whether antibiotic therapy shortens ETBF-associated diarrheal illnesses. Reports of antibiotic sensitivity testing of clinical ETBF strains are limited to 58 strains from Bangladesh and 9 strains from Nicaragua (10, 102). Whereas all strains were resistant to ampicillin in Nicaragua, 97% of Bangladeshi ETBF strains were sensitive to ampicillin; conversely, all strains were sensitive to metronidazole in Nicaragua, but a relatively high proportion (7%) of ETBF strains were metronidazole resistant in Bangladesh. Approximately 10% of ETBF strains were clindamycin resistant in both locales, and 26% of ETBF strains were resistant to tetracycline in Bangladesh (not evaluated in Nicaragua). Because available reports suggest that ETBF disease is self-limited, adequate oral rehydration therapy is central to clinical care, consistent with the therapeutic approach to other diarrheal illnesses.
Three distinct alleles of bft have been identified, namely, bft-1, bft-2, and bft-3 (13, 31, 40, 46). The terminology bft-1, bft-2, and bft-3 was adopted to recognize the temporal sequence in which the alleles were identified. ETBF strains may possess two copies of a single bft genotype, but no ETBF strains have yet been reported to contain mixed bft alleles. The bft genes are chromosomal, with a G+C content of 39%, and are predicted to encode a 397-residue holotoxin with a calculated molecular mass of ca. 44.5 kDa. The predicted BFT structure is a preproprotein holotoxin (Fig. (Fig.3)3) (31, 46, 115, 121). The initial 18 amino acids of the BFT holotoxin make up a signal peptide (preprotein domain) predicted as important in delivery of the holotoxin to the B. fragilis membrane, where the 193-residue proprotein toxin domain is postulated (by analogy to other bacterial zinc-dependent metalloproteases ) to be instrumental in proper protein folding and secretion of mature, biologically active BFT (31). The pre- and proprotein domains of all BFTs exhibit high sequence homology (ca. 97 to 98%), consistent with conserved functions for these protein domains. Based on N-terminal sequencing of purified BFT, the proprotein domain is cleaved at an Arg-Ala site (amino acids 211 and 212) to release mature BFT from the bacterial cell (31, 115, 121). Consistent with these predictions, ca. 44-kDa and 20-kDa proteins are detected with anti-BFT antisera in lysates of ETBF strains (28, 29).
The predicted mature toxin domain of each bft allele contains an extended zinc-binding metalloprotease motif, HEXXHXXGXXH, and a perfectly superimposable methionine residue close to the metalloprotease motif. These data suggest that BFTs are members of the matrix metalloprotease subfamily (matrixins) of the metzincin superfamily of zinc-dependent metalloprotease enzymes (61, 79). Limited homology between eukaryotic matrix metalloproteases and BFT led to the hypothesis that BFT may be an ancestor of host matrix metalloproteases (53). As predicted, the proteolytic activity of BFT appears to be crucial to its biologic activity (61, 122), and BFT contains 1 g-atom of Zn2+ per toxin molecule (61). Furthermore, zinc chelation reduces BFT biologic activity by ca. 90% (47, 61, 80).
Point mutations to modify each of the conserved amino acids of the extended metalloprotease motif as well as the conserved downstream methionine also reduce or eliminate the biologic activities of BFT (28). Although BFT has been reported to be autoproteolytic (61, 116), BFT metalloprotease point mutations do not alter the intracellular processing and secretion of BFT by B. fragilis, suggesting that other intracellular B. fragilis proteases process the holotoxin to mature BFT (28). Additional mutational analysis of the C-terminal region of BFT indicated that this region is intolerant to modest amino acid deletions, suggesting that this region is also important for BFT activity (104). Truncation mutations removing only two C-terminal amino acids reduced BFT biologic activity, and removal of eight (or more) amino acids obliterated it. BFT mutants lacking eight or more C-terminal amino acids were expressed similar to wild-type toxin, but the mutant BFTs were unstable (104).
The predicted amino acid sequences of the BFT holotoxin proteins reveal highly homologous proteins, with BFT-1 and BFT-2 being 95% similar and 92% identical (31). BFT-3 is more closely related to BFT-2 (93% and 96% identical to BFT-1 and BFT-2, respectively) (13). The predicted protein domains of the toxins, however, exhibit differing degrees of identity, with only 2 to 5 amino acid changes between BFT-1, BFT-2, and BFT-3 in the preproprotein domains (13), whereas up to 25 amino acid changes occur in the mature toxin protein domains (13, 31). Consistent with the predicted protein diversity in the mature BFT protein, BFT-1 and BFT-2 purify with distinct biochemical profiles and differ in sodium dodecyl sulfate-polyacrylamide gel electrophoresis mobility and two-dimensional gel electrophoresis mobility, consistent with the predicted uniqueness of the BFT-1 and BFT-2 proteins (121). BFT-2 also exhibits modest but consistently greater specific biologic activity than BFT-1 in vitro, although the importance of this observation to disease pathogenesis is unknown (121). BFT-1 and BFT-2 are trypsin resistant and stable over a wide pH range (i.e., pHs 5 to 10) (97, 115, 121), potentially enabling these toxins to resist degradation in animal and human guts. BFT-3 exhibits a purification profile and specific activity (biologic activity/mg protein) similar to those of BFT-2, consistent with its greater homology to BFT-2 than BFT-1, but other details on the properties of BFT-3 are not available (13).
Only limited investigations have characterized the epidemiology of the specific bft alleles of ETBF, but available data indicate that all three bft alleles are globally distributed (Table (Table9).9). Although the distribution of the bft-3 allele is not restricted to Southeast Asia (6, 21), ETBF strains possessing bft-3 have been reported predominantly from Korea, Japan, or Vietnam, suggesting that regional evolution of ETBF may have occurred (13, 40, 76). Overall, the data suggest that the bft-1 allele is most common among human ETBF strains evaluated to date.
Virulence genes of some organisms are clustered in unique chromosomal loci, termed pathogenicity islands, that possess at least two virulence genes and have a G+C content differing from that of the host organism chromosome, with the latter suggesting acquisition of the sequences from a foreign organism (33). Study of our bank of NTBF (n = 191) and ETBF (n = 113) strains, using probes derived from restriction enzyme mapping of cosmid clones containing the bft gene, revealed five distinct patterns of hybridization, three of which predominated (Fig. (Fig.4).4). Strains with these genetic patterns were termed pattern I, II, and III B. fragilis strains (29). The chromosomes of all ETBF strains (pattern I strains) possess a 6-kb DNA region not found in NTBF strains (29, 60). This 6-kb region contains the bft gene and a second putative virulence gene, termed the metalloprotease II gene (mpII). Because this 6-kb DNA region contains two potential virulence genes and its G+C content is 35%, differing significantly from the predicted ca. 43% G+C content for the B. fragilis chromosome (11, 49), this 6-kb DNA region, exclusively present in ETBF, was termed the B. fragilis pathogenicity island or islet (due to its relatively small size) (BfPAI) (29, 60). Sequence analysis of the BfPAI revealed a ca. 700-bp region upstream of bft with five putative B. fragilis promoter consensus sequences; this 700-bp promoter region is required for maximal BFT production by ETBF strains, although the precise regulatory sequences have not yet been mapped (30). mpII encodes a protein predicted to be similar in size to BFT and also predicted to be a zinc-dependent, catalytically active protein, but with only 56% similarity and 28% identity to the BFT proteins. To date, no in vitro biologic activity for MPII has been identified, likely due in part to its poor expression in vitro under growth conditions favoring expression of bft (A. A. Franco, personal communication). Using an mpII deletion mutant and other recombinant ETBF strains, the in vitro HT29/C1 cell biologic activity of BFT is not dependent on or modified by MPII expression (A. A. Franco and C. L. Sears, unpublished observations). In vivo expression of mpII or the role of MPII in ETBF disease pathogenesis has not yet been addressed.
The culture supernatants of ETBF strains grown in vitro vary significantly in HT29/C1 cell biologic activity (13, 30, 82, 116, 119). Although the mechanism(s) accounting for this variation is incompletely understood, at least two mechanisms likely contribute to the variable biologic activity, including bft copy number (21) and transcriptional regulation of bft yielding different amounts of BFT secreted by ETBF strains (30). In contrast, variation in the biologic activities of purified isoforms of BFT is only modest (13, 121). It is not known if similar differences in BFT expression occur in vivo and/or impact clinical disease expression.
Both pattern II and pattern III B. fragilis strains are NTBF (29). Using the B. fragilis genome database, the DNA element present in pattern III NTBF strains and a related element present in all ETBF strains have been identified as members of a new family of putative conjugative transposons, termed CTn9343 (from NTBF strain NCTC 9343) and CTn86 (from ETBF strain 86-5443-2-2) (27). Pattern II NTBF strains are distinguished by the absence of either of these CTns. In an analysis of 191 NTBF strains, 52% of NTBF strains were pattern II strains and 43% were pattern III strains. A small number of NTBF strains (5%) displayed other genomic patterns (29). CTn9343 and CTn86 are approximately 64 kb, with open reading frames organized as modules with various G+C contents, suggesting regions where the genes may be of Bacteroides origin and regions acquired from other genetic sources. These putative CTns have very limited sequence similarity to other described Bacteroides species CTns and are distinct in several properties, including their predicted mechanism of transposition, their lack of tetracycline regulation of CTn chromosomal excision, and the absence of tetQ (27). Although circularization of the CTns within ETBF and NTBF strains is readily identified, indicating that the CTns can excise from the B. fragilis chromosomes, transfer of CTn86 or CTn9343 to other organisms has not yet been observed (27; Franco, personal communication). However, it is postulated that interorganism transfer of these putative CTns with further acquisition of the BfPAI may be the mechanism by which ETBF strains evolved as a distinct class of B. fragilis strains (27). Additional data suggest that additional CTns, related to but distinct from CTn86 and CTn9343, are present in both ETBF and NTBF (9) (Fig. (Fig.5).5). It is unknown how or if these putative CTns modulate B. fragilis virulence. Similarly, it is unknown what biologic advantage, if any, CTn86 or the BfPAI confers upon ETBF strains. However, mobilization of a plasmid containing bft plus its promoter region into pattern III, but not pattern II, NTBF is efficient and yields high-level expression of BFT, indicating the differing genetic potentials of these two populations of NTBF strains and also suggesting that CTn9343 or another chromosomal locus unique to pattern III NTBF (and absent in pattern II NTBF) regulates BFT production (30).
By use of a variety of molecular techniques, B. fragilis strains have also been classified into two phylogenetic divisions (32, 34). Division I is characterized by the presence of the cepA gene (encoding a serine-β-lactamase of class A) and the absence of the cfiA gene (encoding a metallo-β-lactamase of class B, conferring, for example, imipenem resistance), whereas division II is characterized by the presence of cfiA and the absence of the cepA gene. All ETBF strains, to date, are division I B. fragilis, as are ca. 80% of B. fragilis strains isolated in clinical studies (9, 32). Multilocus enzyme electrophoresis and cluster analysis indicate the ETBF strains are nonclonal, consistent with the higher recombination rates ascribed to division I B. fragilis (32).
Phylogenetic data indicate that B. fragilis strains are diverse, and functional studies and sequences of the genomes of two NTBF strains identified DNA inversion regulatory mechanisms, suggesting that these organisms are highly adaptable, with rapid and dynamic variability in surface molecule expression patterns (11, 20, 49). B. fragilis can express up to at least eight distinct capsular polysaccharides, a previously unprecedented complexity for a single organism (18). It is unknown if the enteric pathogenicity of ETBF is modulated by expression of specific surface characteristics (including capsular polysaccharides), for example, influencing adherence of ETBF to the intestinal mucosa and/or delivery of BFT in vivo.
To date, the in vitro biologic activity of BFT has been restricted to continuous epithelial cell lines capable of forming polarized monolayers (119, 126). Predominantly intestinal cell lines (HT29, HT29/C1, Caco-2, T84, SW480, and HCT116) have been studied (78, 123, 126; S. Wu and C. L. Sears, unpublished observations). More limited studies indicate that renal (MDCK) and pulmonary (Calu-3) cell lines that can polarize in vitro also develop morphological changes and thus exhibit a biologic response to BFT (78; Wu and Sears, unpublished observations). Overall, subconfluent cloned HT29/C1 cells have been studied in greatest detail due to their exquisite concentration-dependent sensitivity to BFT (63, 97). The half-maximal BFT concentration altering HT29/C1 cell morphology is ca. 12.5 pM, with the onset of activity at 0.5 pM (97). Using subconfluent HT29/C1 cells, the hallmarks of the BFT biologic response are a rapid onset (within 10 to 15 min) and temperature-dependent (maximal activity detected at 37°C) morphological changes in which cell rounding and swelling with loss of cell-to-cell contact occur (Fig. (Fig.2)2) (61, 63, 97, 119, 122). These changes are reminiscent of the morphological changes in the surface epithelial cells of intestines infected with ETBF (Fig. (Fig.1).1). Although the total F (filamentous)-actin content of cells treated with BFT is unaltered (22, 47, 96, 97), marked redistribution of F-actin occurs, with decreased stress fibers and peripheral F-actin condensation observed in unpolarized HT29 cells treated with BFT (22, 47). The mechanism(s) responsible for the cell morphology changes stimulated by BFT is unknown, although a broad-spectrum tyrosine kinase inhibitor was noted to delay the onset of BFT-induced cell morphology changes (124). Inhibitors of microtubules and endosomal or Golgi trafficking do not alter induction of cell morphology changes by BFT (22, 78, 97).
Although the bioactivity of BFT is not reversible by washing early on, subconfluent HT29/C1 cells recover a normal appearance by light microscopy within 2 to 3 days after BFT treatment, indicating that the biologic activity of BFT on HT29/C1 cells is reversible over time (97, 119, 122). The majority of available evidence indicates that BFT is a nonlethal and noncytotoxic protein, namely, BFT stimulates rather than diminishes protein synthesis (12, 47), it does not stimulate cellular lactate dehydrogenase (LDH) or 51Cr release or the cellular uptake of vital dyes (trypan blue and propidium iodide) (12, 47, 78, 120), and DNA synthesis continues normally (22). BFT treatment of polarized colonic epithelial cell (HT29 and T84) monolayers in vitro resulted in delayed (>36 h after BFT treatment) apoptosis of a minority of treated cells, although the initial response to BFT was in fact shown to be induction of an antiapoptotic protein (cellular inhibitor of apoptosis protein 2 [cIAP2]) (43; see “Molecular Mechanism of Action of BFT”). After BFT treatment of T84 cells or human colon biopsies in vitro, a delayed loss of cell viability and apoptosis of detached (and thus dying) epithelial cells were noted (99, 100). Of potential importance to human ETBF disease, heterogeneity was noted in the rate of onset (2 to 18 h) and severity of epithelial morphological changes in human colon biopsies treated with BFT, with no BFT response noted in 20% of the individuals sampled (99).
BFT stimulates a concentration- and time-dependent increase in the permeability of epithelial monolayers (T84, MDCK, HT29, HT29/C1, and Caco-2 cells) (12, 78, 105). Notably, BFT exhibits polar monolayer bioactivity, increasing permeability rapidly and at lower toxin concentrations when placed on the basolateral membranes rather than the apical membranes of epithelial monolayers or human colons studied in vitro (12, 78, 93, 105, 120; Wu and Sears, unpublished observations). In addition, basolateral but not apical BFT can rapidly but transiently increase the short-circuit current (indicative of chloride secretion) (12). BFT also stimulates predominantly basolateral release of proinflammatory CXC chemokines, consistent with a role for BFT in inducing mucosal inflammation (45, 124; see “Molecular Mechanism of Action of BFT”).
The changes in T84 cell monolayer permeability are accompanied by cellular morphological changes, with apical BFT causing only relatively slow (with onset at 6 h), focal changes on the apical epithelial cell membranes. In contrast, basolateral BFT more rapidly (by 90 min) modifies the morphology of every cell in the monolayer, with cell swelling and unraveling of the apical membrane microvilli resulting in a striking domed cellular appearance (12, 47). The disappearance of the microvilli is associated with a loss of F-actin from the apical pole of the cells, with a marked reassembly of the F-actin at the basolateral pole of the cells (12). Concomitantly, as detailed by transmission electron microscopy, BFT stimulates structural changes and even dissolution of the zonula occludens (tight junction) and zonula adherens, electron-dense structures that regulate the permeability of epithelial monolayers, whereas desmosomes remain intact (12, 105). The redistribution of F-actin and loss of the zonula occludens and zonula adherens readily explain the measured increase in monolayer permeability, although full mechanistic details are not available. BFT-induced colon permeability may further expose the submucosa to other bacterial luminal antigens and thus contribute to how ETBF fosters colonic inflammation (7, 37, 88, 90, 113). Wells and colleagues further reported that the basolateral membranes of BFT-treated HT29 cells permitted increased association and invasion of pathogenic enteric bacteria, except for Listeria monocytogenes (120). BFT-treated HT29 cells are likely resistant to L. monocytogenes invasion due to BFT-induced loss of cellular E-cadherin (see “Molecular Mechanism of Action of BFT”), one ligand for invasion of this bacterium (59). Lastly, BFT, presumably by modifying mucosal permeability, has been reported to act as a mucosal adjuvant, enhancing the systemic antibody response to an intranasal antigen challenge in mice (117).
To date, the only cellular protein demonstrated to be rapidly cleaved after treatment of colonic epithelial cells with BFT is the zonula adherens protein E-cadherin (122, 125). In subconfluent HT29/C1 cells, the onset of E-cadherin cleavage is rapid, as it is detectable by 1 min and typically complete within 1 to 2 h. BFT induces E-cadherin cleavage via two steps, the initial release of the E-cadherin ectodomain, which is dependent on biologically active BFT, and subsequent processing of the intracellular E-cadherin fragment by host cell presenilin-1/γ-secretase (a member of the intramembrane cleavage protease [iCLips] family) (125). Only E-cadherin presented on an intact, living cell is cleaved in response to BFT (i.e., in vitro cleavage of E-cadherin cannot be demonstrated), and this E-cadherin cleavage requires cellular ATP, suggesting that protein conformation and/or other cellular properties contribute to the proteolytic event (122). Cellular recovery after BFT treatment correlates with resynthesis of E-cadherin. Cleavage of E-cadherin also occurs in vivo in a murine model of ETBF disease and correlates with the onset of colonic inflammation and disruption of the epithelial barrier by histopathology (92).
Besides its critical role in intercellular adhesion at the zonula adherens in intestinal tissue, E-cadherin is tethered at its intracellular domain to β-catenin, a nuclear signaling protein involved in both normal and dysregulated cellular growth (74). Proteolysis of E-cadherin in response to BFT induces β-catenin nuclear localization, upregulation of c-myc (a β-catenin-regulated oncogene) transcription and translation, and cellular proliferation of HT29/C1 cells (123). HT29/C1 cell proliferation is stimulated by as little as 0.5 nM BFT (123). Further studies suggest that β-catenin signaling accounts for only ca. 30 to 40% of BFT-induced cellular proliferation, indicating that other, as yet unidentified mechanisms contribute to BFT-initiated cell proliferation (123). BFT-treated HT29/C1 cells are highly mobile, consistent with observations indicating that diminished E-cadherin on tumor cells enhances metastatic potential (Wu and Sears, unpublished observations). In parallel, BFT also stimulates induction of the antiapoptotic protein cIAP2, mediated by p38 mitogen-activated protein kinase (p38 MAPK) and cyclooxygenase-2 (COX-2) signaling with production of prostaglandin E2 (PGE2) (43). BFT-induced colonocyte proliferation and mobility, combined with resistance to apoptosis, may contribute to the putative oncogenic potential of BFT and support an initial report of an association between ETBF colonization and colorectal carcinoma (113). An alternative hypothesis is that antiapoptotic signaling by BFT increases the colonocyte life span, permitting further generation of proinflammatory signaling to the colonic submucosa (43).
Reported in vitro proteolytic substrates of BFT include G (monomeric)-actin, gelatin, azocoll, tropomyosin, collagen IV, human complement C3, and fibrinogen. However, no biologic significance of these substrates has yet been identified with regard to the cellular or intestinal mechanism of action of BFT (61, 79, 96). Furthermore, BFT induces shedding of a broad array of HT29/C1 membrane proteins over time (125). However, the identities of these shed proteins and whether they are cleaved by BFT or specifically due to BFT-induced cell signaling are unknown. Of other proteins specifically tested, BFT does not cleave the cell surface proteins occludin or claudin 1 and 2 (both present in the zonula occludens) and β1-integrin (a basal intestinal epithelial cell protein) (122; Wu and Sears, unpublished observations) or the intracellular proteins β-catenin, α-catenin, zonula occludens protein-1 (ZO-1), and actin (96, 122). Although these proteins are not BFT substrates, ZO-1, β-catenin, and occludin, for example, redistribute in cells by 1 hour after treatment with BFT (78, 122).
ETBF induces intestinal inflammation in animals and humans (80, 102, 105). In addition to increased colonic permeability that may augment submucosal inflammation through exposure to luminal bacterial antigens, ETBF likely induces intestinal inflammation via BFT-stimulated intestinal epithelial cell release of proinflammatory chemokines, including IL-8 (also called CXCL8 or [the murine equivalent] macrophage inflammatory protein 2 [MIP-2]), epithelial-neutrophil activating protein 78 (ENA-78), CCL2, monocyte chemotactic protein 1 (MCP-1), and growth-related oncogene alpha (GRO-α; also called CXCL1) (41, 42, 45, 100, 124). The time course of chemokine release varies, with IL-8 appearing first (at 2 to 4 h) and GRO-α and ENA-78 appearing later (45, 124). Similarly, BFT also stimulates increased synthesis of these chemokines from isolated human colonocytes (45). BFT also induces increased expression of the biologically inactive form of transforming growth factor beta (TGF-β) by intestinal epithelial cells in vitro (100). It is postulated that inactive TGF-β is subsequently processed by proteases in the intestinal mucosa to active TGF-β and contributes to mucosal repair in ETBF disease. In contrast, no data have yet identified direct effects of BFT on immune cells or myofibroblasts (92, 100).
NF-κB appears to act as a central regulator of chemokine expression in BFT-stimulated colonic epithelial cells in vitro (41, 45, 124), and NF-κB activation in BFT-treated isolated human colonocytes has been demonstrated (41). BFT induces an unusual discrete supranuclear localization of NF-κB, a finding also previously reported in response to IL-1 (36, 124). The regulation of NF-κB activation stimulated by BFT is complex, involving receptor and nonreceptor tyrosine kinases, MAPKs (p38, extracellular signal-related kinase, and c-Jun N-terminal kinase), Ras, and AP-1 (42, 124). NF-κB is also reported to mediate BFT-induced colonic epithelial cell expression (including in human colonocytes) of COX-2 but not COX-1, resulting in increased levels of cellular PGE2 and cyclic AMP (cAMP).
Initial experiments implicate BFT-initiated NF-κB activation in colonic epithelial cells as the orchestrator of inflammation and secretion in ETBF disease. First, inhibition of NF-κB activation diminished BFT-induced chemokine release and polymorphonuclear leukocyte transepithelial migration in colonic epithelial monolayers in vitro, suggesting that NF-κB signaling was directly linked to polymorphonuclear leukocyte mucosal influx in ETBF disease (41). Second, both p38 MAPK and COX-2 inhibition significantly decreased secretion in mouse ileal loops treated with BFT, implicating NF-κB activation by BFT as a key cellular coordinator of secretion in ETBF disease. Inhibition of p38 MAPK activation also reversed BFT-induced ileal inflammation in the mouse (42). The effect of COX-2 inhibition on murine ileal histology was not reported, although COX-2 inhibition did not diminish MIP-2 production, suggesting at least some differential regulation of secretion and inflammation in response to ETBF infection (44).
Figure Figure66 proposes a model for the pathogenesis of ETBF disease. Available data suggest that among Bacteroides species, B. fragilis seeks a mucosal niche aided by its decoration with fucosylated molecules mimicking host proteins (19, 73). Although ETBF adherence in vivo or in vitro has not yet been examined, the pathogenesis of ETBF intestinal disease is expected to be initiated by adherence of the organism to the colonic mucosa, with local delivery of BFT. The histology of ETBF disease in animals has not identified adherent organisms, but available results are limited by formalin fixation, which can interfere with detection of mucosal adherence in the colon (111). Given the in vitro potency of BFT in cell assays (97) and the limited quantities of BFT secreted into cultures in vitro (121), it seems probable that small amounts of BFT delivered by adherent ETBF cells to the colonic epithelium may be sufficient to modify colonic epithelial cell structure and function. Consistent with this postulate, recent data indicate that bft expression is necessary and sufficient to induce colitis in murine models (92). No quantitative data have yet correlated ETBF colonization or disease with levels of BFT expressed by ETBF strains. It is predicted that BFT promotes disease by binding to apical membrane receptors on colonic epithelial cells (CECs), initiating a burst of complex signal transduction resulting in rapid E-cadherin cleavage (122, 125, 126). E-cadherin cleavage releases β-catenin associated with the cytoplasmic domain of E-cadherin. β-Catenin nuclear translocation along with activation of tyrosine kinases, MAPKs, and NF-κB results in nuclear signaling with new gene transcription (41, 42, 100, 124). E-cadherin cleavage, along with cellular F-actin rearrangement, further promotes colonic permeability (93) and access of innate mucosal immune cells to luminal bacterial antigens. This likely promotes mucosal inflammatory and secretory responses that are augmented by BFT-induced CEC chemokine expression and PGE2 synthesis. Induction of c-Myc synthesis further induces CEC proliferation (123). No data have yet identified a role for BFT in pathogenesis beyond its direct CEC actions. The precise timing and order of the signaling cascades initiated by BFT remain to be deciphered, as do the details of how or which specific signal transduction mechanisms contribute to the cell morphology changes, E-cadherin cleavage, new gene expression, and CEC proliferation stimulated by BFT. Together, the data support the concept that ETBF strains, through secretion of BFT, are proinflammatory and oncogenic bacteria, at least in some hosts (99). Limited clinical observations are consistent with this concept (7, 88, 102, 113).
ETBF organisms are common human colonic symbiotes whose potential to cause human disease is incompletely understood. In murine models, ETBF induces acute, self-limited, symptomatic colitis that transitions to long-term carriage where the murine host constrains but does not eliminate ETBF colitis (92). Well-designed human investigations are needed to assess whether this paradigm occurs in humans and to determine the impact of long-term carriage of ETBF on colonic structure, function, and disease incidence. Based on available clinical and experimental data, conditions where ETBF may contribute to disease include IBD, colorectal cancer, and potentially other enigmatic conditions where colonic inflammation may be pathogenic, such as necrotizing enterocolitis in neonates, postinfectious irritable bowel syndrome, antibiotic-associated diarrhea, and even perhaps malnutrition in children in the developing world. Possible mechanisms regulating and defining the outcome of the host-ETBF interaction include differences in ETBF strain virulence, genetically determined host differences in adhesion, immune, or inflammatory responses to ETBF, and/or modulation of ETBF virulence by the other host colonic flora. Studies to detect ETBF in patient populations should employ both sensitive molecular microbiologic fecal diagnosis and serology to detect anti-BFT antibodies that develop, at least for patients with acute ETBF diarrheal disease (102). Whether anti-BFT antibodies are useful to further evaluate the epidemiology of ETBF human infection or colonization is as yet unknown. There is surging interest in the impact of symbiont bacteria, particularly the colonic flora, on normal host physiology, immunology, and disease. Among Bacteroides species, the molecularly diverse B. fragilis strains are distinguished by being critical symbionts and important human pathogens. Future studies of the colonic flora, including ETBF, hold promise for illuminating the mechanisms governing the essential yet sometimes pathogenic relationship between flora and host.
I thank R. Bradley Sack, Shaoguang Wu, and Augusto Franco-Mora for reviewing the manuscript and providing their insights. I also thank A. Franco-Mora for providing Table Table88 in the manuscript and all the contributors over time to studies from my laboratory.
I have no conflicts of interest to report.
This work was supported by grants RO1 DK45496 and RO1 DK080817 and by a Senior Investigator Award from the Crohn's and Colitis Foundation.
Cynthia L. Sears (M.D.) received her medical degree from Thomas Jefferson Medical College, followed by training in Internal Medicine at The New York Hospital in New York City. She trained in Infectious Diseases at the Memorial Sloan Kettering Cancer Institute and the University of Virginia, where she was a member of the infectious diseases faculty until 1988. Subsequently, Dr. Sears joined the faculty at Johns Hopkins University School of Medicine, where she is now a Professor of Medicine in the Divisions of Infectious Diseases and Gastroenterology, Department of Medicine. As a result of international experiences, she developed a clinical interest in food-borne and enteric infections. She has conducted laboratory as well as clinical research on enterotoxigenic Bacteroides fragilis (ETBF) over the past 15 years. Her present work focuses on studies to define how ETBF induces colitis and the link between chronic colonic inflammation induced by bacteria and colonic tumorigenesis.