Structural insight into tight junction disassembly by Clostridium perfringens enterotoxin

See allHide authors and affiliations

Science  13 Feb 2015:
Vol. 347, Issue 6223, pp. 775-778
DOI: 10.1126/science.1261833

How a toxin makes epithelial sheets leaky

The entire human body and its many compartments are shielded from their external environments by the barrier function of epithelial cell sheets. The paracellular barrier function of tight junctions (TJs) is critical for maintaining homeostasis in any multicellular organism, especially in the skin and internal organs and at the blood-brain barrier. One of the major components of TJs is a family of adhesive membrane proteins known as claudins. Several members of the claudin family are receptors for the bacterial toxin Clostridium perfringens enterotoxin. This toxin often causes food-borne illness both in humans and animals. Saitoh et al. crystallized a complex between the toxin and a claudin that reveals just how the toxin damages epithelial barriers (see the Perspective by Artursson and Knight).

Science, this issue p. 775; see also p. 716


The C-terminal region of Clostridium perfringens enterotoxin (C-CPE) can bind to specific claudins, resulting in the disintegration of tight junctions (TJs) and an increase in the paracellular permeability across epithelial cell sheets. Here we present the structure of mammalian claudin-19 in complex with C-CPE at 3.7 Å resolution. The structure shows that C-CPE forms extensive hydrophobic and hydrophilic interactions with the two extracellular segments of claudin-19. The claudin-19/C-CPE complex shows no density of a short extracellular helix that is critical for claudins to assemble into TJ strands. The helix displacement may thus underlie C-CPE–mediated disassembly of TJs.

Infection with Clostridium perfringens type A by eating contaminated food is a common cause of food poisoning in humans and animals. In the intestines, this bacterium produces Clostridium perfringens enterotoxins (CPEs) that trigger foodborne illness (1). Upon binding to their receptor, the complexes aggregate on the intestinal cell surface and form a large oligomer that inserts into the membrane and forms an ion pathway. The resulting Ca2+ influx triggers cell death (2, 3). The receptors for CPE, initially named CPE-R and RVP-1 (4, 5), were later recognized as claudin-4 and -3, respectively (6), based on their sequence similarity with claudin-1 and -2, known constituents of cell-to-cell tight junctional complexes (7, 8). The carboxyl-terminal half of CPE (C-CPE) mediates the interaction with specific claudins (9, 10), which reversibly modulates the paracellular permeability of tight junctions (TJs), whereas its amino-terminal half is responsible for pore formation and thus for cellular cytotoxicity (11).

Claudins have a common structural topology consisting of four transmembrane (TM) segments; a large first extracellular segment (ECS1), which contains the claudin consensus motif; and a shorter second extracellular segment (ECS2) (1214). The adhesion and polymerization properties of claudins enable them to form linear polymers, called TJ strands, which connect adjacent cells and form the structural backbone of TJs (15). TJs serve mainly as barriers that restrict the diffusion of solutes through intercellular spaces in epithelial and endothelial cell sheets (16), thus separating internal tissue compartments from external environments to maintain the homeostasis of our bodies (8).

To understand the structural basis for how C-CPE recognizes specific claudins, we expressed all mouse claudin subtypes in Sf9 insect cells and assessed their capacity to bind C-CPE by using fluorescent-detection size-exclusion chromatography (FSEC) (17). Mouse claudin-19 showed considerable affinity for C-CPE (fig. S1A) (18). When expressed in a mammalian epithelial-like cell line, mouse claudin-19 formed TJs in the plasma membranes of cell-to-cell contact regions (Fig. 1A). Although a previous study reported that a synthetic ECS2 peptide of claudin-19 had no affinity for CPE (9), a 24-hour incubation with C-CPE resulted in a significant delocalization of the claudin-19 signal away from the junctional borders (Fig. 1A), suggesting that binding of C-CPE causes claudin-19 to dissociate from TJs. The disruption of TJs by incubating cells expressing claudin-19 with C-CPE was also seen in freeze-fracture electron microscopy images (fig. S1, B and C). We used FSEC to quantify the affinity of C-CPE for claudin-19, as well as for claudin-3 and claudin-1 as positive and negative controls, respectively. Our analysis yielded the apparent dissociation constant (K0.5) values for C-CPE binding of 240 ± 18 nM and 7.9 ± 0.2 nM for claudin-19 and -3, respectively, which is consistent with previous reports (11, 19), and we found only negligible binding for claudin-1 (fig. S2B).

Fig. 1 Effect of C-CPE on TJs formed by claudin-19 and overall structure of mCldn19 in complex with C-CPE.

(A) Immunofluorescence microscopy of SF-7 epithelial-like cells stably expressing mouse claudin-19. Confluent SF-7 cells plated on filters were incubated for 24 hours without (control) or with 100 μg/ml C-CPE (+C-CPE) and then stained with antibodies to claudin-19 and ZO-1. Scale bar, 10 μm. (B) Structure of mCldn19cryst in complex with C-CPE shown in ribbon representation and viewed parallel to the membrane. The color of mCldn19cryst changes from the N terminus (blue) to the C terminus (red), and C-CPE is colored green. Dashed lines indicate disordered regions. The gray bars suggest the membrane boundaries of the outer (Ext.) and inner (Cyt.) leaflets.

For crystallization, we removed 26 residues from the C terminus of mouse claudin-19 and substituted three membrane-proximal cysteines with alanines (fig. S3). After crystallization of this construct, mCldn19cryst, in complex with C-CPE, the structure was determined at 3.7 Å resolution using the molecular replacement method (Fig. 1B, fig. S4, and table S1). Crystals of mCldn19cryst with bound C-CPE contain two 1:1 claudin-toxin complexes per asymmetric unit (fig. S5), which, despite different crystal contacts, have almost identical conformations, with an average root mean square deviation (RMSD) between their Cα atoms of 0.39 Å (fig. S5, B and C).

The overall structure of mCldn19cryst is similar to that of mCldn15 (14), a CPE-insensitive claudin, with the four TM segments forming a typical left-handed bundle (Fig. 1B and fig. S6). In addition, the binding does not induce a conformational change in C-CPE (fig. S7). However, the two extracellular β-sheet segments of mCldn19cryst, ECS1 and ECS2, are oriented differently compared with those of mCldn15 (Fig. 2, A to C, and fig. S6). In mCldn19cryst, the loop between β1 and β2 (residues 35 to 42), which is disordered in the mCldn15 structure (fig. S6), is clearly resolved and makes contact with C-CPE (Fig. 2, A and E). Also, the β5 strand in the loop connecting TM3 to TM4 is longer in mCldn19cryst than in mCldn15 and can thus form more hydrogen bonds with the adjacent β1 strand (Fig. 2B). In addition, whereas the disulfide bond between highly conserved cysteine residues at β3 and β4 (C54 and C64, respectively) is also apparent in the mCldn19cryst structure, the loop between β3 and β4 seen in mCldn15 is disordered (Fig. 2A and fig. S3). Finally, the residues in ECS1 that connect β4 to TM2 (residues 69 to 74) are disordered in mCldn19cryst, whereas the corresponding region in mCldn15 forms a short extracellular helix (ECH) that is oriented almost parallel to the membrane surface (Fig. 2C and fig. S4C).

Fig. 2 Interactions of mCldn19cryst with C-CPE.

(A) Close-up view of the extracellular domains of mCldn19cryst (ribbon representation) and C-CPE (surface representation). Colors are the same as in Fig. 1. Side chains of residues in conserved claudin sequences involved in C-CPE binding are shown in stick representation and labeled. Key C-CPE residues for claudin binding are colored yellow and labeled. (B and C) The extracellular domains of mCldn19cryst and mCldn15 are shown as ribbon diagrams in solid colors and in transparent gray, respectively. Hydrophobic residues involved in linear claudin oligomerization are shown in stick representation and labeled. The ECH region, which is disordered in the mCldn19cryst structure, is indicated by a dashed line. (D) Close-up view of the interaction of ECS2 with C-CPE. Main-chain carbonyl groups and side chains of ECS2 residues involved in interactions with C-CPE are shown in stick representation. (E) Close-up view of the interaction of ECS1 with C-CPE. Side chains in the loop region are shown in stick representation. Inter- and intramolecular hydrogen bonds and ionic bonds are shown by green, cyan, and red dashed lines, respectively. C-CPE residues that interact with ECS1, ECS2, or both are colored in light green, salmon, or gray, respectively.

Experiments with ECS2-derived peptides suggested that the ECS2 region of claudins contains a toxin-binding motif, NP(V/L)(V/L)(P/A), that is responsible for CPE binding (3, 9). This toxin-binding motif is probably located in a hydrophobic cavity formed by three tyrosine residues in C-CPE, Y306CPE, Y310CPE, and Y312CPE ( throughout the text, CPE residues are indicated with superscripted CPE) (Fig. 2, A and D), thought to be a claudin-binding pocket (3, 9, 20). The ECS2 loop starting with a conserved asparagine residue, N150, stabilizes the complex with CPE by forming hydrogen bonds with the main chain carbonyl of P311CPE, the side chain of the positively charged R227CPE residue, and so on (Fig. 2D and fig. S4B). Recently, a structure was determined of full-length CPE bound to a short peptide with a modified sequence of the claudin-2 ECS2 (21). The ECS2 peptide was observed in the same hydrophobic cavity on C-CPE despite the differences in ECS2 sequences of claudin-2 and -19 (fig. S8).

Because previous biochemical studies suggested that only ECS2 is involved in CPE binding (10, 19), the observed interaction of ECS1 with C-CPE is unexpected. Although the sequences connecting the β1 and β2 strands in ECS1 (residues 32 to 48) differ among claudin subtypes, the residues in the middle of this region (residues 39 to 42) are more conserved (fig. S3). The motif (A/N/S)-I-(I/L/V)-(T/V) in the claudin-19 structure clearly interacts with the surface of C-CPE, whereas the corresponding region is disordered in the mCldn15 structure (Fig. 2E and fig. S6). To test the binding site observed in the structure of the complex, we introduced point mutations in both ECS1 and ECS2 of mCldn19 in the region where they make contact with C-CPE (Fig. 3). The affinity of the mutants for C-CPE in detergent solution was assessed using FSEC (18) (table S2 and figs. S9 and S10). The binding affinity for C-CPE was affected most by substitutions of residues in the putative binding motif in ECS2 (N150-P154; Fig. 2, A and D) and of the three successive residues forming the conserved motif in ECS1 (A39-I41; Fig. 2E). Mapping of these mutagenesis results onto the C-CPE surface (Fig. 3C) shows almost perfect agreement with the mapping of the sequence conservation in CPE-sensitive claudins (Fig. 3B). Thus, both ECS1 and ECS2 are involved in the physiological interaction with C-CPE.

Fig. 3 The interface between mCldn19cryst and C-CPE.

(A) Surface representation of C-CPE with bound mCldn19cryst shown in ribbon representation. C-CPE residues that interact with ECS1, ECS2, or both are labeled and colored in light green, salmon, or gray, respectively. (B and C) The surface of C-CPE that interacts with claudins viewed as in (A). (B) The C-CPE residues involved in the interaction with mCldn19cryst are conserved and can thus mediate binding to all CPE-sensitive claudins. (C) Mapping of claudin-19 mutations that affect the binding affinity onto the surface of the interacting C-CPE residues. Violet indicates a significant reduction in binding affinity, and lime indicates a small reduction.

The surface of C-CPE that interacts with the extracellular claudin domains is mainly composed of hydrophobic residues (fig. S11, A and B) but also contains residues that can form hydrogen or ionic bonds, especially with ECS2 (Fig. 2 and fig. S11C). The side chains of the residues forming the conserved motif in ECS2 (150N-P-S-T-P154 in mCldn19) extend toward C-CPE and fit snugly into the hydrophobic cavity formed by the tyrosine triplet of C-CPE (Y306CPE, Y310CPE, and Y312CPE) (fig. S4B). Mutation of these residues to alanine reduced the binding of CPE to claudin-4 (22), and substitutions of S152 in ECS2 with hydrophobic residues resulted in mCldn19 variants with a higher affinity for C-CPE (table S2). Thus, the presence of a hydrophobic residue in ECS2 that can fit into the hydrophobic cavity may determine whether a claudin is sensitive to CPE or not. The size of the side chain of N150 and the hydrogen bonds it could form were also critical for CPE binding, and a negative charge at this position inhibited CPE binding even if the pI of the ECS2 region remains unchanged (for details, see the materials and methods and table S2). These results are consistent with the finding that it is predominantly hydrophobic and uncharged residues that form the claudin-toxin interface around the CPE tyrosine triplet (indicated by the dashed line in fig. S11). On the other hand, the residues of the conserved motif in ECS1 (39A-I-I41 in mCldn19) among all claudins were sensitive to the size of their side chains (table S2). Therefore, it seems to be predominantly ECS2 that determines the sensitivity of a claudin for CPE, whereas ECS1 contributes to the interaction simply by enlarging the hydrophobic contact area.

Structures of claudins with and without bound C-CPE could potentially provide insights into the disruption of TJ strands. The structure of mCldn15, which does not bind CPE, enables us to build a homology model for mCldn19 without bound C-CPE and thus to infer the conformational changes that may result from C-CPE binding (Fig. 4 and fig. S12). The claudin-CPE interaction would cause a disordering of the ECH region between β4 and TM2 (fig. S4C and S6B). Because the ECH could form hydrophobic interactions with TM3 of the neighboring claudin protomer (14), it is conceivable that the movement of TM3 and the disordering of ECH induced by CPE binding destabilize the linear claudin polymer in the TJ strand. Furthermore, CPE binding would also interfere with the head-to-head interaction of claudin protomers in TJ strands in adjoining cells. Several studies support our insights (9, 23), but further experiments are needed to fully understand how CPE binding causes the disassembly of TJs.

Fig. 4 Structural insights into the disassembly of a TJ strand induced by C-CPE binding.

The linear arrangement of claudin-19 in a TJ strand is modeled based on the mCldn15 crystal structure. The hydrophobic residues involved in the linear interactions are shown in stick representation and indicated by arrowheads (left panel). Binding of C-CPE disrupts the hydrophobic interactions between neighboring claudin protomers (arrowheads) and also creates steric clashes (double arrowhead) that prevent interactions with neighboring claudin protomers as well as with claudin protomers from an adjoining cell.

Our structure and mutational analyses of the interaction of C-CPE with claudin-19 may help in the design of drugs or biomarkers that selectively target specific claudin subtypes, regardless of their sensitivity to CPE (24, 25). The molecular information on C-CPE–induced TJ disassembly may also be useful to design strategies aimed at increasing the permeability of drugs across TJs in the blood-brain barrier (26).

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Tables S1 and S2

References (2745)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We thank M. Uji, Y. Yamazaki, and Y. Ito for technical support; K. Abe for calculation of affinity values; the beamline staff members at BL32XU, BL38B1, and BL41XU of SPring-8 (Hyogo, Japan) for technical help during data collection; and T. Walz for critical reading of the manuscript. The synchrotron radiation experiments were performed with the approval of the Japan Synchrotron Radiation Research Institute (proposal nos. 2013A1112, 2013B1178, 2013B1342, and 2014A1501). This research was supported by Grants-in-Aid for Scientific Research (S) (to Y.F.), (A) (to S.T.), and (C) (to K.T.); Grants-in-Aid for Scientific Research in Innovative Areas (to S.T.); and the Platform for Drug Discovery, Information, and Structural Life Science from the Ministry of Education, Culture, Sports, Science and Technology of Japan. This research was also supported by the Japan Science and Technology–Core Research for Evolutionary Science and Technology (to S.T.), the Japan New Energy and Industrial Technology Development Organization, and the National Institute of Biomedical Innovation (to Y.F.). Y.S., H.S., and A.T. screened claudin genes. Y.S. performed protein expression, purification, crystallization, and binding assays. K.I. assisted in crystallization. Y.S., H.S., and K.I. collected x-ray data. H.S. and K.T. processed diffraction data and solved and refined the structure. K.N. took electron microscopy images. Y.O. and A.T. collected fluorescence microscopy images. Y.S., H.S., K.T., S.T., and Y.F. wrote the manuscript, and all authors commented on the paper. S.T. and Y.F. supervised the research. The authors declare no competing financial interests. Coordinates and structure factors were deposited in the Protein Data Bank under accession number 3X29.
View Abstract

Stay Connected to Science

Navigate This Article