|Home | About | Journals | Submit | Contact Us | Français|
The 24-member claudin protein family plays a key role in maintaining the normal structure and function of epithelial tight junctions. Previous studies with fibroblast transfectants and naturally sensitive Caco-2 cells have also implicated certain claudins (e.g., Claudin-4) as receptors for Clostridium perfringens enterotoxin (CPE). The present study first provided evidence that the second extracellular loop (ECL-2) of claudins is specifically important for mediating the host cell binding and cytotoxicity of native CPE. Rat fibroblast transfectants expressing a Claudin-4 chimera, where the natural ECL-2 was replaced by ECL-2 from Claudin-2, exhibited no CPE-induced cytotoxicity. Conversely, CPE bound to, and killed, CPE-treated transfectants expressing a Claudin-2 chimera with a substituted ECL-2 from Claudin-4. Site-directed mutagenesis was then used to alter an ECL-2 residue that invariably aligns as N in claudins known to bind native CPE but as D or S in claudins that cannot bind CPE. Transfectants expressing a Claudin-4N149D mutant lost the ability to bind or respond to CPE, while transfectants expressing a Claudin-1 mutant with the corresponding ECL-2 residue changed from D to N acquired CPE binding and sensitivity. Identifying carriage of this N residue in ECL-2 as being important for native CPE binding helps to explain why only certain claudins can serve as CPE receptors. Finally, preincubating CPE with soluble recombinant Claudin-4, or Claudin-4 fragments containing ECL-2 specifically blocked the cytotoxicity on Caco-2 cells. This result opens the possibility of using receptor claudins as therapeutic decoys to ameliorate CPE-mediated intestinal disease.
Clostridium perfringens enterotoxin (CPE) causes the intestinal symptoms of C. perfringens type A food poisoning, the second most commonly identified bacterial food-borne illness (21), and also contributes to many cases of antibiotic-associated diarrhea (21). CPE applications are also emerging, including (i) use of this toxin as a potential anticancer agent, based upon cancer cells often overexpressing CPE receptors and thus exhibiting strong CPE sensitivity (16, 17, 25, 30) and (ii) use of noncytotoxic, but binding-capable, C-terminal CPE fragments to increase drug delivery (8, 11, 18, 38).
The first step in CPE action involves binding of this toxin to host cell receptors; however, CPE binding to these receptors is insufficient to trigger cytotoxicity (22, 40). Instead, the toxin-receptor complex must oligomerize on the enterocyte plasma membrane surface to form an ~450-kDa prepore named CH-1 (29). CH-1 then inserts into the plasma membrane, producing an active pore that allows a Ca2+ influx to trigger apoptotic or oncotic cell death pathways (4, 5). Dying CPE-treated cells develop morphological damage, which facilitates formation of an ~600-kDa CPE complex named CH-2 that may contribute to tight junction disruption (29, 33).
Increasing evidence implicates certain members of the claudin tight-junction protein family as functional receptors for CPE binding to host cells. First, expression cloning approaches showed that mouse fibroblasts, which are naturally CPE insensitive and produce no claudins, acquire the ability to bind and respond to native CPE when transfected to express Claudin-3 or -4 (9, 13, 14, 36). Similarly, fibroblast transfectants expressing Claudin-6, -7, -8, or -14 also exhibited CPE binding ability (9). However, not all claudins are CPE receptors, since fibroblast transfectants expressing Claudin-1, -2, or -5 did not bind CPE (9). Additional evidence for certain claudins functioning as CPE receptors was provided by coimmunoprecipitation studies showing that enterotoxin interacts with Claudin-3 and -4 prior to CH-1 formation in naturally CPE-sensitive, enterocytelike Caco-2 cells (29).
A claudin structure is not yet available, but hydropathy plots predict these 20- to 27-kDa proteins contain a short cytoplasmic N-terminal domain, four transmembrane domains, two extracellular loops (ECLs), and a cytoplasmic C-terminal tail (39). The first predicted ECL (ECL-1) of claudins consists of ~52 amino acids, while the second predicted ECL (ECL-2) is smaller, containing 16 to 35 amino acids (27). The C-terminal tail, the most variable region among different claudins, can trigger cell signaling events via its PDZ binding motif, although such signaling is not required for CPE action (29).
Fujita et al. (9) began mapping the Claudin-3 region involved in CPE binding. They showed that fibroblast transfectants acquire CPE sensitivity when expressing a chimeric claudin where the N-terminal half of Claudin-1 is fused with the C-terminal half of Claudin-3. However, CPE did not affect transfectants expressing a claudin chimera with the N-terminal half of Claudin-3 fused to the C-terminal half of Claudin-1. The presence of a CPE binding region in the C-terminal half of Claudin-3 received further support from in vitro overlay blot results showing CPE binding to an ECL-2-containing fusion protein (9).
Since our primary interest concerns elucidating CPE-induced cytotoxicity during intestinal disease, the current study sought to investigate (i) why some, but not all, claudins can bind and mediate native CPE cytotoxicity and (ii) whether claudin-based receptor decoys might be useful as therapeutics to block CPE-induced cytotoxicity.
CPE was purified to homogeneity from C. perfringens strain NCTC8239, as described previously (23).
Full-length human Claudin-1 cDNA (Invitrogen) was PCR amplified by using Taq polymerase (New England Biolabs) and primers cldn1BamHI (GGCCGGATCCAACTCTCCGCCTTCTGCAC) and cldn1EcoRI (GGCCGAATTCTTGAGTATGATTACTCAA). The amplicon was cloned into TOPO-TA (Invitrogen) and then subcloned, using EcoRI and BamHI, into pCEP4 (Invitrogen). Full-length cDNA encoding human Claudin-2, Claudin-2/E4, and Claudin4/E2 (supplied by James Anderson and Christina Van Itallie) were amplified by using the primers Cldn-2EcoRIF (GCGGAATTCGGTCTGCCATGGCCTCTCTTGGCC) and Claudin-2BamHIRev (GCGGGAATTCTCACACATACCCAGTCAGGCTGTATGA). The product was cloned into TOPO-TA digested by using EcoRI and BamHI and subcloned into pTrcHisA (Invitrogen). Claudin-2 fragments were amplified by using the primers Claudin-2136-230F (CGCCGGATCCGTTGCTTGGAATCTTCATGGC), Claudin-2108-230F (CGCCGGATCCGTGTTCTGCCAGGATTCTCGGCTAAG), and Claudin-2BamHIRev (GCGGGAATTCTCACACATACCCAGTCAGGCTGTATGA). Claudin-4 fragments were amplified by using the primers Claudin-4108-209F (ATCGGCTAGCCTGGAGGATGAAAGCGCCAAGGCC) or Claudin-4136-209F (ACTCGCTAGCGTGTCCTGGACGGCCCACAAC) and Claudin-4REV (GCGCTTCGAATTACACGTAGTTGCTGGCAGC). These claudin fragments were subcloned, using NheI and BstBI, into pTrcHisA. Claudin-1, -2, and -4 mutations were made by using the QuikChange lightning site-directed mutagenesis kit (Stratagene), and these products were subcloned into (i) pCEP4 for transfection (Claudin-1D150N and Claudin-4N149D mutants, only) or (ii) pTrcHisA (for expression of soluble rClaudin-1D150N, rClaudin-4N149D, rClaudin-2S149N, and rClaudin-4N149S). All wild-type and mutant claudin constructs were verified by DNA sequencing.
Rabbit polyclonal anti-Claudin-2 or -4 and mouse monoclonal anti-Claudin-1 were purchased from Invitrogen. Rabbit polyclonal anti-CPE was prepared as described previously (42). Goat polyclonal anti-rabbit horseradish peroxidase conjugate and rabbit polyclonal anti-mouse horseradish peroxidase conjugate were purchased from Sigma.
Caco-2 cells and Rat1-R12 parental fibroblasts were cultured as previously described (29).
Rat1-R12 fibroblast transfectants stably expressing chimeric claudins, i.e., Claudin-2/E4 or Claudin-4/E2 (see Results), were kindly provided by James Anderson and Christina Van Itallie. These transfectants, were constructed as previously described for MDCK cells (6), except that the rat fibroblasts have the Tet-off vector and are thus resistant to Geneticin. To prepare stable Rat1-R12 transfectants expressing Claudin-1 (Claudin-1 transfectants), pCEP4 carrying a Claudin-1 cDNA was transfected into Rat1-R12 fibroblasts using Lipofectamine, as described previously (6). To prepare Rat1-R12 stable transfectants expressing Claudin-1D150 or Claudin-4N149D, pCEP4 carrying a cDNA encoding one of these two site-directed claudin mutants (which had been prepared by using the QuikChange mutagenesis kit, as described earlier) were separately transfected into Rat1-R12 fibroblasts, as described previously (6). All stable transfectants were cultured at 37°C in 5% atmospheric CO2 in Dulbecco modified Eagle medium, 10% Tet-off certified system fetal bovine serum, 100 U of penicillin/ml, 100 μg of streptomycin/ml, 100 μg of Geneticin/ml, and 50 μg of hygromycin/ml (omitted for all experiments).
Recombinant claudins, claudin chimeras, claudin fragments and claudin mutants were expressed as His6-tagged proteins from pTrcHisA and then affinity enriched to >95% homogeneity (data not shown) by HisTrap HP affinity chromatography (GE Healthcare), as described previously (35). Aliquots containing recombinant claudin (rClaudin) species were identified by Western immunoblotting with rabbit polyclonal anti-Claudin-2 or -4 or mouse monoclonal anti-Claudin-1 (as appropriate) and quantified by a Lowry protein assay (20).
Confluent Caco-2 cells or fibroblast transfectants were treated with 1 μg of CPE/ml in Hanks balanced salt solution without calcium and magnesium (iHBSS; Mediatech) for 60 min, scraped, washed once, and lysed with 2× sodium dodecyl sulfate (SDS) buffer. Samples were electrophoresed on SDS-containing gels containing 4% polyacrylamide and then transferred onto a nitrocellulose membrane (Bio-Rad). Western immunoblotting was performed on this membrane, as described previously (29) for detecting the SDS-resistant CPE complexes.
A 100-fold excess of each affinity-enriched, soluble, recombinant claudin species was preincubated with 1 or 2.5 μg of CPE/ml in 10 ml of iHBSS, with gentle rocking, for 20 min at 37°C. Confluent Caco-2 cells were then treated with those preincubation mixes for 60 min at 37°C and processed for CPE large complex formation, morphological damage, and cytotoxicity, as described elsewhere in Materials and Methods.
Confluent monolayers of each transfectant were washed with iHBSS and then treated with or without 2.5 μg of CPE/ml in iHBSS for 30 min at 37°C. Confluent Caco-2 cells were similarly treated, except that the CPE used had been preincubated with a 100-fold excess (per weight) of the appropriate soluble, affinity-enriched, recombinant claudin species for 20 min at 37°C before application to those cells. Control or damaged cells were photographed by using a Canon Powershot G5 fitted to the Zeiss Axiovert 25 microscope, and images were processed by using Adobe Photoshop 6.0.
A Live/Dead viability/cytotoxicity assay kit (Invitrogen) for mammalian cells was used to evaluate CPE-induced cytotoxicity. The assay, as described by the supplier, was performed after treatment of Rat1-R12 transfectant or Caco-2 monolayers for 60 min at 37°C with 1 or 2.5 μg of CPE/ml. For the Caco-2 cell experiments, CPE was preincubated for 20 min at 37°C with a 100-fold excess (per weight) of the appropriate affinity-enriched, recombinant claudin species prior to addition to this preincubation mix to the monolayer, as described earlier. The results are expressed as a percentage of the total cells.
CPE (1 μg/ml) and 100 μg of affinity-enriched, soluble, rClaudin-4 or rClaudin-1/ml were mixed in 600 μl of buffer A (50 mM Tris-HCl [pH 7.4], 1% NP-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and additional protease inhibitors [Complete protease inhibitor cocktail tablets; Roche]). That mixture was incubated with gentle rotation for 30 min at 37°C before coimmunoprecipitation, as described previously (29). Western immunoblotting was then performed on the immunoprecipitate by using rabbit polyclonal anti-Claudin-1, followed by horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibody, or mouse anti-Claudin-4 antibody primary, followed by horseradish peroxidase-conjugated rabbit anti-mouse IgG secondary antibody serum.
Western blot analysis of claudin expression by transfectants was carried out as previously described (29) using mouse anti-Claudin-1, rabbit anti-Claudin-2, or rabbit anti-Claudin-4 primary antibody. Secondary antibody conjugates were used as described above for coimmunoprecipitation studies.
To establish a simple in vitro assay for assessing interactions between claudins and native CPE, the we initially evaluated whether preincubating CPE with soluble, affinity-enriched rClaudin-4 would inhibit CPE binding and action if that preincubation mixture was later applied to Caco-2 cells. This experiment was also performed with two affinity-enriched, rClaudin-4 C-terminal fragments (see Fig. S1A in the supplemental material) to confirm previous suggestions (9, 19, 41) that ECL-2-containing C-terminal Claudin-4 fragments can bind native CPE.
Control experiments first demonstrated (Fig. 1A and B) that Caco-2 cells exhibit morphological damage and cytotoxicity when challenged with CPE that had been preincubated for 20 min at 37°C in the absence of any claudin. In contrast, neither CPE-induced morphological damage nor cytotoxicity was observed (Fig. 1A and B and Table Table1)1) when Caco-2 cells were challenged with CPE that had been preincubated with (i) soluble full-length rClaudin-4, (ii) a fragment corresponding to the entire C-terminal half of Claudin-4 (rClaudin-4108-209), or (iii) a fragment lacking TM3 but containing the predicted ECL-2, TM4, and cytoplasmic tail sequences of Claudin-4 (rClaudin-4136-209).
This inhibitory effect was specific since CPE similarly preincubated with soluble, affinity-enriched, full-length rClaudin-2 (Fig. 1A and B and Table Table1)1) or rClaudin-1 (Table (Table1)1) still caused CPE-induced morphological damage and cytotoxicity in Caco-2 cells, a finding consistent with the known inability of Claudin-1 or -2 to serve as CPE receptors for fibroblast transfectants (9). Similarly, Caco-2 cells still developed both morphological damage and cytotoxicity (Fig. 1A and B and Table Table1)1) when challenged with CPE that had been preincubated with an affinity-enriched soluble fragment corresponding to the entire C-terminal half of Claudin-2 (rClaudin-2108-230) or to only the predicted ECL-2, TM4, and cytoplasmic tail of Claudin-2 (rClaudin-2136-230).
Western immunoblotting demonstrated (Fig. (Fig.1C1C and Table Table1)1) CPE binding to, and CH-1/CH-2 complex formation in, Caco-2 cells challenged with CPE that had been preincubated in the absence of any rClaudin species or preincubated with His6-tagged rCLIP-P Borrelia burgdorferi protease, as a control. However, neither CPE binding nor CH-1/CH-2 complex formation occurred (Fig. (Fig.1C1C and Table Table1)1) in Caco-2 cells challenged with CPE that had been preincubated with soluble rClaudin-4, rClaudin-4108-209, or rClaudin-4136-209. This inhibition was specific since native CPE binding and CH-1/CH-2 formation were detected when Caco-2 cells were challenged with CPE that had been preincubated with soluble rClaudin-1, rClaudin-2, rClaudin-2108-230, or rClaudin-2136-230 (Fig. (Fig.1C1C and Table Table11).
Coimmunoprecipitation analysis confirmed that the protection afforded Caco-2 cells by preincubating CPE with soluble rClaudin-4, as a representative of all rClaudin-4 species tested in the present study, resulted from CPE binding to Claudin-4 rather than from an indirect effect such as protease degradation during the preincubation. In this experiment, IgG raised against CPE (but not normal rabbit IgG) specifically pulled down rClaudin-4 from the preincubation mixture (Fig. (Fig.1D).1D). In contrast, CPE antibodies did not immunoprecipitate rClaudin-1; this inability of CPE antibodies to pull down rClaudin-1 was not due to rClaudin-1 degradation since Western blotting detected intact rClaudin-1 in supernatants of the coimmunoprecipitation reaction (Fig. (Fig.1D1D).
Besides demonstrating that soluble rClaudin-4, or its C-terminal fragments, can bind and neutralize native CPE, Fig. Fig.11 results were consistent with previous studies (19, 41) indicating that the TM3 sequences present in the C-terminal half of receptor claudins are unnecessary for binding CPE fragments or peptides. However, it has not yet been specifically shown that, within the context of an intact claudin receptor, the ECL-2 sequences mediate native CPE binding interactions.
To assess the specific importance of ECL-2 sequences for mediating binding interactions between claudin receptors and native CPE, we first used the assay developed in Fig. Fig.11 to examine whether preincubating CPE with soluble recombinant chimeric claudins could inhibit the development of CPE action when those mixtures were later applied to Caco-2 cells. These soluble chimeric claudins consisted (see Fig. S1B in the supplemental material) of either (i) rClaudin-2/E4, where the ECL-2 sequence of (CPE-receptor) Claudin-4 specifically replaced the ECL-2 sequence of (non-CPE receptor) Claudin-2, or (ii) rClaudin-4/E2, where the ECL-2 of Claudin-2 specifically replaced the Claudin-4 ECL-2.
Consistent with ECL-2 mediating binding interactions between claudin-4 and CPE, preincubating CPE with rClaudin-2/E4 prevented the subsequent development of morphological damage, cytotoxicity, binding, and CH-1 formation when that mixture was applied to Caco-2 cells (Fig. (Fig.22 and Table Table1).1). In contrast, CPE preincubated with the same amount of rClaudin-4/E2 still caused those effects when applied to Caco-2 cells.
With the Fig. Fig.22 results supporting the importance of ECL-2 for CPE binding interactions with claudin receptors, we then directly tested whether ECL-2 mediates CPE binding and action on host cells. Stable Rat1-R12 fibroblast transfectants were prepared to express Claudin-4/E2 or Claudin-2/E4. Expression of the expected chimeric claudin by each transfectant was then confirmed by Western blotting (data not shown).
Control studies (Fig. 3A and B and Table Table1)1) first confirmed previous reports (29) that parental Rat1-R12 fibroblasts are not affected by CPE treatment. Also consistent with previous studies (29), Rat1-R12 fibroblast transfectants stably expressing Claudin-4 developed morphological damage (including cell rounding and detachment) and cytotoxicity within 30 min of CPE treatment, but these cells showed no morphological damage or cytotoxicity when transfected to stably express native Claudin-2.
Consistent with the Fig. Fig.22 preincubation results, we then demonstrated (Fig. (Fig.3A3A and Table Table1),1), for the first time, that Claudin-2/E4-expressing Rat1-R12 transfectants are susceptible to CPE-induced morphological damage and cytotoxicity. In contrast, Claudin-4/E2-expressing Rat1-R12 fibroblast transfectants showed no CPE-induced morphological damage or cytotoxicity.
Western blotting identified the mechanism determining whether a chimeric claudin-expressing transfectant exhibited CPE-induced effects. As previously shown (29), Western blot results (Fig. (Fig.3C3C and Table Table1)1) confirmed that (i) parental Rat1-R12 fibroblasts do not bind CPE or form CH-1, as expected since these cells do not naturally produce any claudins; (ii) CPE binding and CH-1 formation are not detectable in CPE-treated Rat1-R12 fibroblast transfectants expressing Claudin-2, but (iii) CPE binding and CH-1 formation do occur in CPE-treated, Claudin-4-expressing Rat1-R12 transfectants. Note that CPE-binding Rat1-R12 transfectants expressing Claudin-4 cannot form CH-2 since fibroblasts do not naturally express the occludin needed for CH-2 formation (33).
Importantly, these Western blot analyses also revealed (Fig. (Fig.3C3C and Table Table1)1) that Rat1-R12 fibroblast transfectants stably expressing Claudin-2/E4 show strong CPE binding and CH-1 formation. In contrast, no CPE binding or CH-1 formation was observed in CPE-treated Rat1-R12 transfectants expressing Claudin-4/E2.
Since the Fig. Fig.33 results provided strong evidence that native CPE binding to Claudin-4 on host cells involves ECL-2 sequences, an alignment (Fig. (Fig.4)4) was performed to compare the ECL-2 sequences between claudins known, from previous transfectant studies (9, 29), to bind native CPE versus claudins known not to bind CPE. This alignment revealed that the middle amino acid residue in ECL-2 is invariably an N in CPE-binding claudins, but the corresponding ECL-2 residue is usually a D for claudins that cannot bind CPE. The single exception to this pattern is the non-CPE-receptor Claudin-2, where S is present at the corresponding ECL-2 residue.
The Fig. Fig.44 alignments suggested that the N residue located in the middle of ECL-2 (e.g., residue 149 in Claudin-4) plays a key role in determining whether native CPE can bind to a claudin. This hypothesis was tested by using site-directed mutagenesis to change (see Fig. S1C in the supplemental material) the N149 residue in Claudin-4 to the D of claudin-1 (creating rClaudin-4N149D) or to the S of claudin-2 (creating rClaudin-4N149S). Conversely, the corresponding D residue present at residue 150 in the Claudin-1 ECL-2, or the corresponding S residue present at position 149 in the Claudin-2 ECL-2, were changed to the N of claudin-4 (creating rClaudin-1D150N and rClaudin-2S149N, respectively).
Each recombinant site-directed claudin mutant or control claudin was affinity enriched and then preincubated, in solution, with CPE to assess whether CPE-induced morphological damage and cytotoxicity still developed when that preincubation mix was applied to Caco-2 cell cultures. As shown in Fig. Fig.55 and Table Table1,1, the addition of a mixture containing CPE preincubated with rClaudin-4 produced no CPE-induced morphological damage or cytotoxicity in Caco-2 cells. In contrast, CPE preincubated with rClaudin-4N149D still caused the development of CPE-induced morphological damage and cytotoxicity when that preincubation mix was applied to Caco-2 cells.
Similarly, CPE preincubated with rClaudin-1 or rClaudin-2, which cannot bind CPE in their native forms (9), induced CPE-induced morphological damage (not shown) and cytotoxicity (Fig. (Fig.5B5B and Table Table1)1) in Caco-2 cells. However, CPE preincubated with rClaudin-1D150N or rClaudin-2S149N failed to cause CPE-induced morphological damage or cytotoxicity in Caco-2 cells (Fig. 5A and B and Table Table11).
Western blot analyses (Fig. (Fig.5C)5C) showed that CPE preincubated with soluble rClaudin-4, rClaudin-2S149N, or rClaudin-1D150N did not bind to Caco-2 cells or form CH-1 and CH-2, explaining why no cytotoxicity was observed using those three preincubation mixes. In contrast, CPE preincubated with rClaudin-1, rClaudin-2, or rClaudin-4N149D still bound to, and formed CH-1 or CH-2, in Caco-2 cells, explaining why CPE-induced cytotoxicity was observed using those three preincubation mixes.
The results presented in Fig. Fig.55 and Table Table11 showed that changing (i) the D150 ECL-2 residue in rClaudin-1 to N, (ii) the corresponding S149 ECL-2 residue in rClaudin-2 to N, or (iii) the corresponding D149 ECL-2 residue in rClaudin-4 to either D or S (see Fig. S1C in the supplemental material), reverses (relative to their parent rClaudin) the ability of those rClaudin mutants to protect Caco-2 cells against CPE in the preincubation challenge model. Therefore, experiments were performed to evaluate directly whether Rat1-R12 transfectants expressing claudins with these ECL-2 site-directed mutations show altered CPE binding ability and sensitivity.
For this experiment, stable Rat1-R12 fibroblast transfectants expressing either Claudin-4N149D or Claudin-1D150N were prepared. Western blotting confirmed that each mutated claudin transfectant expressed the appropriate claudin (data not shown). After CPE treatment, morphological damage and cytotoxicity were observed in cultures of the Claudin-1D150N-expressing transfectant, but not in Claudin-4N149D-expressing transfectant cultures (Fig. 6A and B and Table Table1).1). Western blot analyses demonstrated that the Rat1-R12 transfectant expressing Claudin-1D150N exhibits similar binding CPE and CH-1 formation as Rat1-R12 transfectants expressing Claudin-4, but contrasting with the inability of Rat1-R12 transfectants expressing wild-type Claudin-1 to bind CPE or form the CH-1 complex (Fig. (Fig.6C6C and Table Table1).1). Those Western blot analyses also showed that mutating the N149 residue in the Claudin-4 ECL-2 to D abrogated the CPE binding ability and CH-1 formation mediated by wild-type Claudin-4 (Fig. (Fig.6C6C and Table Table1).1). Collectively, these results supported the N residue in the middle of ECL-2 as being a major determinant of whether native CPE binds to a claudin on the host cell surface.
Claudins, which play a major role in maintaining the normal structural and functional properties of the epithelial or endothelial TJ (1, 9), are also emerging as important targets for pathogenic bacteria, viruses, and toxins. Some pathogens, e.g., Helicobacter pylori, disrupt normal TJ permeability properties by reducing claudin expression (2). Alternatively, other pathogens or their toxins directly exploit claudins as receptors or coreceptors. For example, Claudin-1, -6, and -9 are considered coreceptors for hepatitis C virus, although (to our knowledge) direct hepatitis C virus binding to these claudins has not been demonstrated (7, 24, 37). A second example includes the role of some, but not all, claudins as CPE receptors (9, 29, 36).
The goal of the present study was to better identify the specific regions and amino acid residues of claudin receptors that mediate host cell CPE binding and cytotoxicity. In particular, we sought to assess why only certain claudins can bind native CPE. Claudins are thought to possess two extracellular loops, ECL-1 and ECL-2 (1, 26, 31, 39). Previous studies had shown binding of native CPE or CPE fragments to ECL-2-containing fusion proteins or to ECL-2 peptides (27, 41). We used here two approaches to demonstrate, within the context of a full-length claudin structure, that ECL-2 sequences are important for mediating CPE binding. First, swapping ECL-2 sequences between receptor versus nonreceptor claudins was shown to directly affect CPE binding and cytotoxicity in transfectants. The CPE sensitivity differences observed between transfectants expressing the chimeric claudins and their parent wild-type claudins are unlikely to involve differences in claudin expression levels based upon results from a second experiment, where soluble rClaudin-2/E4 and rClaudin-4, but not the same amount of soluble rClaudin-4/E2 or rClaudin-2, inhibited CPE action on Caco-2 cells in the preincubation challenge assay. Similarly, several observations make it unlikely that the CPE binding/sensitivity differences observed between transfectants expressing chimeric claudins with swapped ECL-2 sequences could be attributable to those ECL-2 substitutions inducing gross conformational changes in claudin structure. First, replacing the ECL-2 of Claudin-2 with the Claudin-4 ECL-2 caused transfectants to gain CPE binding and sensitivity. Given the exquisite binding specificity of CPE shown in this and previous studies (9, 13, 14, 36), this gain in CPE binding ability and/or sensitivity would be unlikely to involve nonspecific gross conformational changes. Second, as discussed below, CPE binding and action in transfectants were altered by introducing into ECL-2 fairly conservative single amino acid changes, which rarely induce gross conformational changes. Finally, both chimeric claudins exhibited (data not shown) similar trypsin digestion patterns as their wild-type parent claudins, which supports the absence of gross conformational changes (15, 34), In total, these new chimeric claudin results, which are consistent with earlier results showing CPE binding to isolated ECL-2 fragments or peptides (19, 41), clearly implicate ECL-2 in native CPE binding to host cells. This finding further establishes that both claudin ECLs can be exploited by pathogens or toxins, since ECL-1 of Claudin-1, -6, or -9 is known to mediate hepatitis C virus entry into host cells (24).
A recent study (41) reported that a GST-CPE116-319 fusion protein only bound to those mouse claudin ECL-2 peptides containing the pentapeptide sequence NPL(V/L)(P/A). On this basis, it was proposed that this ECL-2 pentapeptide forms a helix-turn-helix structure that enables GST-CPE116-319 binding to ECL-2 peptides, with the hydrophobic amino acids L and V/L in the pentapeptide representing a hydrophobic core that interacts with a hydrophobic pit in the CPE116-319 binding domain. However, as acknowledged by that previous study (41), the pentapeptide model cannot adequately explain GST-CPE116-319 binding to all full-length claudins, e.g., while full-length claudin-4 lacks the NPL(V/L)(P/A) ECL-2 pentapeptide, it still bound GST-CPE116-319. Besides these observed differences between GST-CPE116-319 binding to ECL-2 peptides versus full-length claudins, the claudin-binding properties of GST-CPE116-319 versus native CPE also apparently vary, for still-unknown reasons. For example, GST-CPE116-319 reportedly bound to transfectants expressing native Claudin-5 (41), although native CPE binding to Claudin-5 was not observed in earlier transfectant studies (9).
Given those apparent binding discrepancies when using full-length claudins versus claudin ECL-2 peptides, or native CPE versus GST-CPE fragment fusion proteins, we sought here to explore which ECL-2 residues determine whether a full-length claudin can bind native CPE. Table Table22 incorporates experimental data from the present study and previous studies (9, 29) to compare ECL-2 pentapeptide sequences present in all human or mouse claudins characterized to date for their ability to bind native CPE. This comparison indicates that the only consistent ECL-2 pentapeptide residues among all CPE-binding capable claudins are N and (usually) its adjacent P. Table Table11 also shows that the corresponding ECL-2 residue is never an N in claudins unable to bind native CPE and that the P residue adjacent to the ECL-2 N residue of binding-capable claudins is also present in nonreceptor claudins.
Therefore, the alignments shown in Table Table22 are consistent with the N residue present in the ECL-2 region playing a major role in determining whether a claudin can bind native CPE and, by extension, transmit CPE-induced cytotoxicity. Our experimental results here directly support this hypothesis. First, mutating the D naturally present at residue 150 in the ECL-2 pentapeptide of human Claudin-1 to N, or changing the corresponding S149 residue in the ECL-2 of Claudin-2 to N, caused soluble forms of those two mutant claudins to acquire CPE binding ability. Of more direct physiologic relevance, transfectants expressing Claudin-1D150N, but not transfectants expressing native Claudin-1, exhibited CPE binding and sensitivity. Conversely, while soluble Claudin-4 bound CPE in vitro, mutating the natural N present at residue 149 in the putative ECL-2 pentapeptide motif of Claudin-4 to D or S eliminated the ability of soluble forms of these two mutant claudins to bind CPE. Consistent with that finding, transfectants expressing the Claudin-4N149D mutant did not bind CPE or show CPE sensitivity.
It was notable that introducing relatively conservative amino acid substitutions into the middle ECL-2 residue profoundly affected CPE binding. The N-to-D change in rClaudin-4N149D introduced a negative charge into the mutant protein. However, charge alterations at this ECL-2 residue are not necessary to affect claudin CPE binding properties since changing this critical ECL-2 residue from S to N (or vice versa), which does not involve a charged amino acid substitution, also altered claudin binding properties for CPE. Understanding why the ECL-2 N residue is so important for native CPE binding would profit from the availability of structural information regarding receptor-bound CPE.
Concluding the conserved N residue in ECL-2 of claudin-4, and likely other receptor claudins, is important for native CPE binding partially agrees with an earlier study (41) of GST-CPE116-319 binding to claudins or ECL-2 peptides. Specifically, that previous study (41) showed Claudin-2 acquires GST-CPE116-319 binding ability when the ECL-2 S150 residue is changed to N (41). Changing the ECL-2 N148 residue to D in Claudin-3, a natural CPE receptor claudin, had more ambiguous effects on GST-CPE116-319 binding in that study (41). In some assays, this mutation abolished GST-CPE116-319 binding to Claudin-3. However, in other assays, an N-to-D ECL-2 mutation at residue 148 had limited effect on GST-CPE116-319 binding to Claudin-3; GST-CPE116-319 binding was only blocked by also introducing a second mutation, changing claudin-3 residue 150 from L to A.
Combining our experimental results with the Table Table22 alignments further suggests the ECL-2 pentapeptide motif model proposed for GST-CPE116-319 binding to claudin ECL-2 peptides (41) cannot adequately explain native CPE binding to receptor claudins. Among claudins capable of binding native CPE, considerable amino acid variation exists in the ECL-2 pentapeptide sequence; notably, for the Claudin-1 ECL-2 pentapeptide, these variations include the presence of the neutral amino acids M and T, rather than the hydrophobic amino acids L and V/L considered important for GST-CPE116-319 binding to claudin-3 ECL-2 peptides (41). Given our results showing that Claudin-1D150N expression renders transfectants CPE-sensitive, the presence of neutral amino acids (rather than L and V/L) in the ECL-2 pentapeptide is clearly permissible for native CPE binding/cytotoxicity. Table Table22 alignments also indicate that the ECL-2 helix-turn-helix region important for GST-CPE116-319 binding to ECL-2 peptides is not required for native CPE binding to full-length claudins, i.e., the second P residue in the pentapeptide is missing from this ECL-2 region in CPE-binding claudin-4, -6, and -8 (9).
Our results identifying an important role for the N149 ECL-2 residue of Claudin-4 in mediating native CPE binding and action may help to explain the widespread CPE sensitivity of mammalian species. Claudin-4 has the highest CPE binding affinity of any known claudin and is expressed in the intestines, which is relevant for CPE-mediated disease (12, 28). The N149 residue now implicated in native CPE binding to Claudin-4 is conserved (at minimum) in humans, mice, rats, horses, cows, chimpanzees, rhesus monkeys, sheep, and dogs (GenBank). Interestingly, this N149 ECL-2 residue of Claudin-4 is also conserved in frogs and some fish, suggesting that CPE may act beyond mammalian species. Furthermore, as shown in Table Table2,2, the presence of a corresponding ECL-2 N residue in other known CPE-binding claudin receptors could also be physiologically significant, since many of those receptor claudins are also found in the intestines where they could contribute to CPE-mediated disease (39).
Finally, receptor decoys can interfere with the actions of several bacterial toxins (3, 10, 32). The present study has demonstrated that a 100-fold excess (per weight, equivalent to a 160-fold molar excess) of soluble rClaudin-4, or rClaudin-4 fragments containing ECL-2, can inhibit CPE action on enterocyte-like Caco-2 cells. This suggests possible therapeutic applications for claudin receptors or their fragments, i.e., CPE-induced intestinal symptoms might be ameliorated by delivering claudin receptors or receptor fragments into the small intestine so they could act as decoys to bind and neutralize CPE. In vivo studies addressing this possibility are currently under way in our laboratory.
This study was generously supported by a grant from the National Institute of Allergy and Infectious Disease (R37-AI019844-27 [B.A.M.]).
We thank Christina Van Itallie and James Anderson, Department of Cell and Molecular Physics, University of North Carolina-Chapel Hill, for their help and input into the project, as well as for supplying rat fibroblast transfectants and chimeric claudins. We also thank James Carroll, Department of Microbiology and Molecular Genetics, University of Pittsburgh, for providing the soluble, His6-tagged recombinant bacterial protein, rCLIP-P protease of B. burgdorferi.
Editor: B. A. McCormick
Published ahead of print on 2 November 2009.
†Supplemental material for this article may be found at http://iai.asm.org/.