Two major domains have been functionally distinguished in CPE: the N-terminal half shares several structural similarities with ß-poreforming toxins [32] and mediates the cytotoxicity, whereas the C-terminal half contains the claudin-binding domain [34]. Analysis with recombinant deletion mutants of CPE revealed the region meditating the different activities of CPE in more detail. The 44 N-terminal amino acids have an inhibitory effect on the cytotoxicity [34]. The region 45–167 is involved in membrane insertion as well as “large complex”- and pore-formation. More precisely, a specific region between D45 and S59 is required for the formation of the “large complex” [2,34,35]. The region 81 to 106 is essential for pore formation [32].

The C-terminal region 290–319 is sufficient for binding to claudins, as was shown with corresponding recombinant C-terminal CPE-fragments and respective synthetic peptides [36]. Removal of these 30 C-terminal amino acids inhibited binding to claudins completely [34]. However, longer CPE-fragments interact much stronger with claudins than cCPE_290–319 [37]. Van Itallie and colleagues solved the crystal structure of cCPE_194–319, which most likely contains the complete claudin-binding domain [38]. This domain comprises the residues A205–F319 and forms a nine-strand β-sandwich. The core claudin-binding region—S290–F319—is formed by strands β8 and β9, which are located in the center of the opposing beta sheets, and the surface loop between these strands. Consequently, peptides restricted to the amino acids 290–319 are not expected to have the same conformation as the corresponding region in the full beta sandwich. This explains, at least partly, why cCPE_290–319 binds more weakly to claudins than longer cCPE-fragments [37].

The structure of cCPE194–319 shows similarities to the receptor-binding domains of other bacterial toxins [38]. This was unexpected, since no strong homology to these proteins is given at the amino acid sequence level. However, similarities in the 3D structure, e.g., to those domains within the ColG collagenase of Clostridium histolyticum [39] and the Cry4Ba toxin of Bacillus thuringiensis [40], suggest a common evolutionary origin.

To investigate the interaction between CPE and claudins different methods have been applied. These are described and compared in the following sections.

All methods listed in Chapter 5.1 are used to investigate CPE-claudin interaction. As mentioned in Chapter 3.2, it was shown that the 30 C-terminal amino acids of CPE (β8-loop-β9-region) are essential for binding to claudins. Involvement of residues within this region in binding to claudins has been demonstrated in several studies [35,59,60,63,66]. For instance, it was found that upon deletion of 16 C-terminal amino acids, cCPE loses its ability to interact with Cld4 [67]. The substitution Y310C reduced binding of CPE to brush border membrane in rabbit intestine [35]. Harada et al. revealed inhibition of binding and toxicity of cCPE-PSIF [63] in competition assays by substitutions Y306A, Y310A or Y312A [59]. Additionally, this study applied a pull-down assay to show that single mutations Y306A, Y310A or Y312A and double mutation Y310A/Y312A resulted in partial reduction of Cld4 binding. Furthermore, double mutants Y306A/Y310A and Y306A/Y312A, as well as triple mutant Y306A/Y310A/Y312A, lost the ability to bind to Cld4. Measurements of transepithelial resistance of cultured cells and in situ loop assays showed inhibition of TJ-opening by the multiple tyrosine mutations [59]. Y306K, but not Y306F substitution interfered with claudin binding [66]. Taken together, these results showed that Y306 is the pivotal residue for binding to claudins, where the aromatic ring and not the functional hydroxyl group is essential. However, all three tyrosine residues (Y306, Y310 and Y312) contribute to the interaction of cCPE with Cld4 and to the modulation of TJ barriers [59].

Systematic analysis of each of the last 16 C-terminal amino acids was performed to reveal other residues possibly involved in binding to claudins [60]. By replacing individual amino acids with alanine it was found that L315, in addition to the previously reported tyrosines, is critical for binding to Cld4, modulating TJ-barrier function and enhancing jejunal absorption. The crystal structure of CPE_194–319 revealed that Y306, Y310, Y312 and L315 form a hydrophobic pit on the surface of CPE, which is obviously involved in claudin binding [37,38].

Soon after the identification of Cld3 and Cld4 as CPE-receptors it was demonstrated that CPE does not bind to all claudin subtypes. In particular, binding to Cld3 and Cld4 but not to Cld1 and Cld2 was reported [4]. The association constants for the binding of CPE to Cld3, -4, -6, -7, -8 and -14 were determined to be 8.4 × 10⁷ M^-1, 1.1 × 10⁸ M^-1, 9.7 × 10⁷ M^-1, 8.8 × 10⁷ M^-1, 1 × 10⁶ M^-1 and 1 × 10⁶ M^-1, respectively [4,43]. Recently, detectable binding to Cld1, Cld2 [44] and Cld9 [37] has been demonstrated also. However, CPE-binding to Cld1 and -2 is much weaker than that to Cld3 and -4. For Cld5 no interaction with either full length CPE or CPE_194–319 was obtained [37]. However, binding of CPE to Cld5 was detected with GST-CPE_116–319 [37] containing more amino acids than the well described claudin-binding domain (194–319). It was suggested that the very weak affinity between cCPE and Cld5 is presumably enhanced by multivalent binding of GST-CPE_116–319-oligomers [37]. Different results concerning CPE-binding of the diverse claudin subtypes could be obtained due to differences in the CPE-constructs and the sensitivity of the assay used. However, Cld3, -4, -6, -7, -8 and -9 (controversial for Cld8) can be considered to be high affinity CPE-receptors; Cld1, -2, -5, -8 and -14 (controversial for Cld5 and Cld8) to be low affinity CPE-receptors and Cld10–13 as well as Cld15–24 apparently do not interact with CPE.

To date, it is well known that the ECL2 of CPE-sensitive claudins is involved in binding with enterotoxin. In different independent studies it was confirmed that CPE interacts with the ECL2 of Cld3 and -4, but not (strongly) with those of Cld1 and -2 [43,62]. In addition, CPE does not recognize ECL1 of Cld3 and -4 [43,44]. Recently, residues in the ECL2 of CPE-binding claudins, involved in the interaction with CPE, were identified by amino acid substitutions in ECL2 of claudins [37]. Arrays with immobilized Cld3-ECL2 peptides were incubated with GST-cCPE. Substitution of D149 and T151 in murine Cld5 by the corresponding residue of Cld3 (N148, L150) enabled the binding of cCPE to Cld5. Similarly, substitution S149N in murine Cld2 led to binding of GST-cCPE.

Furthermore, substitution mapping was performed to reveal the role of residues in Cld3-ECL2 in interaction with cCPE. Each position of Cld3-ECL2 peptide was substituted with every other amino acid except cysteine. This mapping identified the motif 148NPLVP152 as essential for binding of cCPE to Cld3-ECL2 peptides. Moreover, pull-down and cellular binding assays verified strong involvement of N148 and L150 in interaction between CPE and native Cld3 [37]. Sequence comparison revealed that all cCPE-binding subtypes of claudin contain the consensus sequence NP(L/M)(V/L/T)(P/A) in the turn region of the predicted helix-turn-helix structure of the ECL2 (Chapter 4 and [54]). Strikingly, all claudins, which are able to interact with cCPE, belong to the group of classic claudins [45]. As mentioned above (Chapter 4), the classic claudins probably share a common helix-turn-helix structure. This appears to be a precondition for CPE-binding, since non-classic claudins, which have most likely a different fold, do not bind [37]. However, differences within this helix-turn-helix structure have to determine the ability to interact with cCPE. The positions 2, 4 and 5 in the identified consensus sequence (P, V/L/T and P/A) are highly conserved among classic claudins and, according to the model, are involved in stabilization of the turn region of the loop structure. In contrast, the NH2-group of arparagine at position 1 and side chain of leucine/methionine at position 3 point away from the loop. In all CPE-sensitive claudins, but in nearly none of the CPE-insensitive claudins, asparagine is present at position 1 and leucine or methionine at position 3 of the consensus sequence. Taken together, the discussed data demonstrate that the residues corresponding to N148 and L150 in murine Cld3 are strongly involved in the CPE-claudin interaction.

Subsequently, other studies verified the importance of the asparagine in the proposed consensus sequence. The substitution N149D in full length human Cld4 abrogated the binding of CPE [62]. Conversely, the corresponding D150N substitution in human Cld1 as well as S149N in human Cld2 strongly increased binding of CPE [62]. Kimura et al. analyzed several claudin chimera of Cld5 and Cld4 including D149N substitution in human Cld5 [44]. These data are also consistent with conclusions mentioned above.

In addition, Kimura et al. reported that a Cld4/Cld1 chimera bound to cCPE as long as they contained the particular residue M160 from Cld4 at the end of ECL2, but not the corresponding F161 from Cld1 [44]. In another study, substitution mapping of Cld3-ECL2 peptides [37] suggested that this methionine, which is conserved between Cld4 and Cld3, is rather not directly involved in interaction with CPE. Alternatively, it could change the structure, especially at the transition between ECL2 and the 4th transmembrane segment (TM4) of Cld3/-4 compared to Cld1 and, therefore, play a role in CPE-binding. Results from SRP measurements using recombinant Cld4 protein as analyte are in line with this interpretation. So it was found that CPE_290–319 binds much stronger to Cld4-constructs containing the TM4 in addition to the ECL2 than the ECL2 alone [58]. Similarly, a recombinant Cld4 construct containing ECL2 and TM4 was able to compete for binding of full length CPE in cellular binding assays [62].

Aromatic amino acids in CPE (Y306, Y310 and Y312) are involved in the binding to claudins [60] and aromatic residues in the ECL2 of classic claudin are crucial for trans-interaction between each other [54]. It was discussed that the aromatic residues in claudin could also interact with the aromatic amino acids in CPE [58]. But removal of aromatic residues in the ECL2 of Cld3, Cld4 or Cld5 (binding to CPE_116-319) did not reduce the ability of CPE to bind to its receptors [37,44]. Moreover, in a preliminary model, the aromatic residues in the ECL2 do rather not sterically match the aromatic residues in CPE (Figure 1b and [37]). Taken together, the published data (summarized in Table 1) indicate that residues in the turn region of the loop, but not aromatic residues in the flanking helices, are involved in the claudin-CPE interaction. Residues in the C-terminal flank of the loop and in the following TM4 could contribute to the interaction with CPE either by direct interaction or by determining the structural conformation of the loop.

It was observed that cCPE binds preferentially to claudin molecules outside of tight junction strands [37]. This is consistent with the idea that CPE sequesters claudins in the plasma membrane, preventing their incorporation in TJ strands. Such a sequestering mechanism could explain the long incubation time with cCPE needed for the opening of TJ, despite the high affinity of cCPE to claudins [43].

Many promising drug candidates cannot be used clinically because of unacceptable pharmacokinetics. The competence to cross tissue barriers is a major determinant for the delivery of a drug to its target. Several approaches have been used to enhance transcellular drug delivery, including utilization of endogenous influx transporters, blocking of the barrier supporting activity of efflux transporters, or receptor-mediated transcytosis [68]. An alternative strategy tries to enhance paracellular permeation of drugs by transient opening of the TJ [69,70]. The strength of this approach is (i) the improvement of the delivery of structurally unrelated drugs and (ii) that the drug itself does not have to be modified. Most of the different TJ modulators, described within last years, have surfactant- or chelator-like characteristics. They are often compromised by low tissue specificity and severe side effects, e.g., exfoliation of cells, which irreversibly compromise the barrier functions [71,72]. It was suggested that fewer side effects may be obtained by more specific modulation of a molecular key component of the TJ [73]. cCPE is a promising tool to modulate TJ in a direct and tissue-specific manner: (1) It binds with high affinity to claudins that are essential for TJ formation. (2) CPE binds to a subset of claudins, only. This restricts its activity to tissues that express CPE-sensitive claudins. (3) It is able to increase the paracellular permeability in a reversible and concentration-dependent manner. (4) cCPE enhances drug absorption in rat jejunum 400-fold relative to sodium caprate, which is in clinical use [5]. In addition, cCPE is known to enable mucosal absorption of a biologically active peptide [65]. Cld1 and Cld5 are potential targets for transepidermal and brain drug delivery, respectively [74,75]. However, it has been reported that these claudins do not strongly interact with CPE [43]. Alternatively, Cld3 could be considered to improve the brain uptake. Moreover, modification of cCPE could shift its claudin-subtype specificity to enable drug delivery to the brain.

Several studies have suggested the use of CPE for the chemotherapy of tumors overexpressing claudins [76,77,78]. The CPE-receptors Cld3 and Cld4 are strongly upregulated in many pancreatic, ovarian, breast and uterine cancers [76,79]. The use of other surface proteins for targeting cancer cells has been already demonstrated. For instance, ligands for cytokine- and growth factor receptors have been fused with fragments of bacterial toxins, such as diphtheria toxin and Pseudomonas exotoxin [80,81]. Similarly, fusion proteins of diphtheria toxin fragment A and cCPE were generated to target Cld4-overexpressing tumor cells [82]. CPE-fragments fused to tumor necrosis factor effected human ovarian cancer cells [83]. Injections of full length and hence cytotoxic CPE into pancreatic tumors in mice resulted in tumor necrosis and reduction of tumor growth [84]. Human ovarian cancer cells transplanted in mice could be also eliminated by injection of CPE [85]. Intracranial CPE administration reduced growth of breast cancer brain metastasis and increased survival in murine models [76]. Recently, cCPE fusion proteins were successfully used for nasal vaccination, a strategy that could be used for different therapeutic approaches including tumor treatment [82]. These studies demonstrated the therapeutic potential of CPE-constructs.