Introduction

The tight junction (TJ) is critical for the epithelial barrier protecting our cells, organs, and body systems from pathogens and other harmful agents [23]. These structures reside in the most apical region of the lateral membranes, where they function as both gates and fences. As gates, TJs create selective barriers between tissue spaces. As fences, TJs inhibit the diffusion of lipids and membrane proteins between the apical and basolateral membrane domains, thus maintaining cell polarity. However, epithelial carcinogenesis often introduces aberrations into the proteins that form the TJs, consequently disrupting their gate and fence functions [64]. Thus, TJ proteins are attractive candidate targets for cancer therapy [48].

One group of proteins that is integral to TJs and to their gate and fence functions are the claudins (CLDNs) [19]. Each of the 27 mammalian members of the CLDN family is a tetra-transmembrane protein that contains 2 extracellular loop domains and N- and C-terminal intracellular domains [55]. The expression pattern of CLDNs differs among tissues [23]. A series of gene and histological analysis has revealed that abnormal expression patterns of CLDNs are observed in many cancers (Table 1). These findings indicate that CLDN binders might be potent therapeutic and diagnostic molecules for cancer therapy.

Table 1 Cancers characterized by changes in CLDN protein or mRNA expression
Full size table

Clostridium perfringens enterotoxin (CPE) is responsible for the diarrhea and abdominal cramps of C. perfringens type A food poisoning, which is the second most common foodborne illness in the USA [21]. The receptors for CPE were identified in 1997—before the “discovery” of CLDNs—and, in 1999, the CPE receptors were found to be identical to CLDN-3 and -4 [18, 31, 32, 58]. Thus, by being identified as the first CLDN binder, CPE became an important nexus between the concept of CLDN-targeted cancer therapy and, through its application as a model for drug discovery, the reality of this treatment modality. In this review, we overview the journey to CLDN-targeted cancer therapy with CPE technology.

CPE and its receptor-binding domain

The CPE protein comprises 319 amino acids and has two domains: the N-terminal cytotoxic domain, which is involved in oligomerization and pore formation, and the C-terminal receptor-binding domain (Fig. 1a, b) [35]. To elicit the high cytotoxic activity of the N-terminal domain requires digestion of the extreme N-terminal residues by intestinal protease [36]. The CPE oligomerization domain corresponds to amino acids 45–53, and residues 80–106 are responsible for the insertion of CPE into the membrane and for pore formation [10, 36]. The C-terminal fragment of CPE (corresponding to amino acids 184–319) is the receptor-binding domain [31, 32]. CLDN-3 and -4 were the first recognized CPE receptors, with the second extracellular loop domain considered to be the CPE receptor domain [17, 79]. CPE typically binds to CLDNs that are “free” on the cell surface and only infrequently binds to CLDNs that are incorporated into TJs [42, 94]. After interacting with CLDNs, CPE first becomes a small (∼90 kDa) complex consisting of CPE and a cell surface CLDN, followed by oligomerization of six of these small complexes into a large (∼450 kDa) complex. This large complex forms a permeable pore, causing a robust influx of Ca2+, and the subsequent disruption of the cellular osmotic equilibrium and activation of Ca2+-dependent protease cause rapid cell lysis (Fig. 2) [9]. In this context, CPE is the first CLDN-targeted cytotoxic molecule. Numerous binding studies have revealed that CPE strongly binds to CLDN-3, -4, -6, -7, and -9 and weakly binds to CLDN-8, -14, and -19 [17, 71, 94].

Fig. 1
figure 1

Structure of Clostridium perfringens enterotoxin (CPE). a Structure of CPE (PDB code, 3AM2) by X-ray crystallography. The N-terminal domain is shown in green, and the C-terminal domain is turquoise. Regions involved in oligomerization (red), pore formation (pink), and claudin binding (navy) are indicated. b Schematic diagram of the roles of the CPE domains. Amino acids shown in blue are particularly important for claudin binding

Full size image
Fig. 2
figure 2

CPE-mediated cytolysis. This schematic diagram illustrates the mechanism underlying CPE-mediated cytolysis. In brief, CPE binds to CLDNs to form a hexameric complex in the membrane. This complex creates a pore through which Ca2+ enters the cell. The rapid influx of Ca2+ triggers Ca2+-dependent protease, leading to necrosis

Full size image

Recombinant C-terminal fragments of CPE (C-CPE), corresponding to amino acids 184–319 or 194–319, bind to CLDNs [79, 89, 91]. Treatment of epithelial cell sheets with C-CPE disrupts the intercellular seal function of TJs without inducing cytotoxicity [39, 85, 89]. Functional domain-mapping studies indicate that the minimal CLDN-binding domain comprises amino acids 290–319 in the C-terminus and that Y306, Y310, Y312, and L315 are critical for the interaction of C-CPE with CLDN-4 [24, 84].

The role of CPE for cancer therapy

Ovarian cancers have been studied most frequently as targets for CPE and its derivatives. A large-scale serial analysis of gene expression revealed that CLDN-3 and -4 are overexpressed (83- and 109-fold, respectively) in several cases of ovarian cancers compared with normal epithelium [29]. In addition, DNA microarray analysis has shown that CLDN-3 and -4 are two of the top five differentially expressed genes in ovarian cancers [75]. CLDN-6 and -7 are also overexpressed in ovarian cancers [13, 87]. This means that the expression of several CPE-sensitive CLDNs (that is, those that strongly bind CPE) is frequently increased in ovarian cancers [74]. Moreover, the expression levels of CLDN-3 and -4 were higher in chemotherapy-resistant or recurrent ovarian cancers than in chemotherapy-sensitive ovarian cancers [74, 95]. Intraperitoneal injection of CPE into intraperitoneal mouse xenografts models of chemotherapeutic-resistant primary ovarian cancer cells attenuated tumor growth without apparent side effects [8, 74].

In studies comparing primary breast carcinoma and normal mammary epithelial tissues, CLDN-3 and -4 were overexpressed in 62 and 26 % of breast cancers, respectively, and CLDN-7 gene expression was increased in breast cancers [38, 60]. Furthermore, CPE has been shown to be cytotoxic in freshly resected samples of breast cancer tissue [38]. However, 5–10 % of breast cancers comprise a “CLDN-low” subtype that shows decreased expression of CLDN-3, -4, and -7 [25].

Among the CLDNs, both CLDN-3 and -4 are strongly expressed in prostate cancers [51, 82]. In particular, a cDNA microarray analysis of prostate cancer cell lines and normal prostate cell lines showed that CLDN-4 mRNA has the greatest upregulation among 51 upregulated genes in prostate cancer cells, and an immunohistologic study indicated that CLDN-4 is highly expressed in primary prostate cancers and at the secondary sites of metastatic prostate cancers and is moderately expressed in benign prostatic hyperplasia [43]. Although intratumoral injection of CPE inhibited tumor growth in mice bearing PC-3 prostate cancer cells, the possibility of side effects of this treatment scheme should be considered, given that CLDN-3 is strongly expressed in normal prostate tissues [51, 52].

A Northern blot analysis demonstrated frequent high CLDN-4 expression in pancreatic cancer tissues, whereas CLDN-4 expression was low or absent in normal pancreatic tissue [54]. In addition, immunohistologic analysis using tissue samples from patients with pancreatic cancers revealed that expression of CLDN-4 was increased in primary and metastatic pancreatic cancer cells compared with normal pancreatic duct epithelial cells [63]. Intratumoral injection of CPE in mice carrying Panc-1 xenografts (pancreatic cancer cells) completely inhibited tumor growth and induced widespread tumor necrosis without causing any adverse effects [54].

One of the hurdles to the clinical application of human induced pluripotent stem cells for regenerative medicine is the risk of teratoma formation from contaminating undifferentiated cells [46]. In this context, CLDN-6 has been shown to be present in pluripotent-undifferentiated cells and germ cell tumors but not in differentiated cells [3, 90]. Although transplantation of a mixture of undifferentiated and differentiated cells into mice led to teratoma formation, treatment of the cell mixture with CPE prior to transplantation prevented this consequence [3]. Of note, CLDN-6-deficient mice showed a normal phenotype, indicating that CLDN-6 is dispensable for the self-renewal and survival of pluripotent cells [1].

The role of C-CPE as an adjunct to chemotherapy

An increasing tumor mass causes an increase in tumor interstitial fluid pressure and thus hampers the uptake and penetration of anti-tumor drugs [2]. This situation prompted the search for a strategy to increase the tissue penetration of anti-tumor drugs by opening epithelial junctions, given that intercellular junctions, including TJs, are not fully compromised in cancer cells [53]. C-CPE can be used as such an enhancer of anti-tumor drug penetration (Fig. 3a) [37]. In particular, C-CPE induced morphologic changes in ovarian cancer cells (causing them to become spheroids) and repeated combined intraperitoneal administration of paclitaxel and C-CPE that yielded significant synergic anti-tumor effects in mice implanted with subcutaneous tumors [20]. Another molecule, a categorized enhancer of anti-tumor drug penetration, increased the therapeutic efficacy of paclitaxel, protein-bound paclitaxel, doxorubicin, and a monoclonal antibody [4, 5].

Fig. 3
figure 3

C-CPE-based therapeutic and diagnostic approaches. a C-CPE-mediated modulation of the epithelial barrier to enhance the penetration of anti-tumor drugs. b C-CPE-mediated sensitization of cancer cells to anti-tumor drugs. c Using C-CPE derivatives to target and image cancer cells

Full size image

The biological roles of CLDNs in cancer cells are still not fully understood, but several reports suggest that the suppression of CLDNs in cancer cells increased their sensitivity to anti-tumor drugs. Knockdown of CLDN-4 in ovarian cancer cells enhanced the cellular uptake of cisplatin and thus their sensitivity to this drug [95]. Samples of ovarian cancer tissue resected from cisplatin-resistant patients consistently showed significantly higher CLDN-4 expression than did those resected from cisplatin-sensitive patients [95]. CLDNs recruit various membrane proteins to the plasma membrane and regulate others at the gene-expression level, suggesting that the loss of CLDN-4 might affect the recruitment or expression of a drug influx or efflux transporter [41, 83, 97]. Moreover, knockdown of CLDN-7 in pancreatic cancer cells increased their cisplatin sensitivity by reducing the expression of anti-apoptotic proteins [88]. Because C-CPE treatment causes the degradation of CLDN-4 proteins, C-CPE might enhance the sensitivity to anti-tumor drugs by increasing their cellular uptake (Fig. 3b) [79]. In an example of an alternative mechanism, CLDN-4 associates with the cisplatin influx transporter CTR1 and knockdown of CLDN-4 in ovarian cancer cells hampered the internalization of cisplatin by reducing CTR1 mRNA [76]. The cell type-specific biological functions of CLDNs should be considered for the utility of CLDN binders as adjunct therapeutics.

The role of C-CPE in cancer diagnosis

CLDNs are targets not only for cancer therapeutics but also for cancer diagnosis. C-CPE is a non-toxic CLDN binder, and concepts for CLDN-targeted cancer diagnosis have been proven by using C-CPE (Fig. 3c).

Intravenous injection of fluorescent dye-conjugated C-CPE into mice intraperitoneally grafted with OSPC-ARK-1 cells (intraperitoneal metastatic ovarian cancer cells) detected tumor tissues as small as 1 mm2 in area, which went undetected by conventional visual observation [12]. C-CPE-mediated detection of peritoneal micrometastases and CPE-mediated cytolysis have been useful in the treatment of uterine serous cancers, which are gynecologic cancers showing high expression of CLDN-3 and -4 [12, 73].

SPECT imaging using 111In-conjugated glutathione S-transferase–C-CPE showed significantly increased accumulation of the radioactive tracer in the tumor tissues of MDA-MB-468 xenografts (CLDN-4-positive breast cancer cells) compared with CLDN-4-negative xenografts [59]. This SPECT imaging technique also detected aplastic lesions in genetically engineered mice that spontaneously developed breast cancer (Balb/neuT mouse). As another example, the accumulation of fluorescein-conjugated C-CPE was greater in Capan-1 xenografts (CLDN-4-positive pancreatic cancer cells) than in CLDN-4 negative xenografts [62]. In addition, the C-CPE conjugate accumulated more strongly in spontaneously developed pancreatic intraepithelial neoplasms than in normal pancreatic tissues [62]. Together, these findings suggest a potential application of C-CPE in the detection of precursor lesions in some cancers.

Safety evaluation of CPE and C-CPE

Because CLDN-3 and -4 are expressed in normal tissues, including lung, thyroid, liver, kidney, and intestines, damage to these tissues is inevitable once CPE or C-CPE-conjugated drugs are applied. C-CPE was distributed in thyroid, liver, kidney, and intestines after intravenous injection [49]. For instance, a single intraperitoneal injection of CPE in excess of 0.75 mg/kg was lethal to mice [93]. Similarly, a single intravenous injection of more than 10 μg/kg of C-CPE conjugated with protein synthesis inhibitory factor, which is the exotoxin derived from Pseudomonas aeruginosa, induced hepatic injury and loss of body weight in mice [49]. The high systemic toxicity of CPE limits the administration route and dose of CPE and C-CPE-conjugated drugs. Indeed, CPE injected directly into xenografted tumors or into the peritoneal cavity of mice with peritoneal metastatic cancer inhibited tumor growth in the absence of systemic effects [8, 52, 54, 74]. To reduce the systemic cytotoxicity of CPE and its derivative, one group designed a CPE-based protoxin [69]. The protoxin comprised two domains: CPE and the CPE-binding motif derived from the second extracellular domain of CLDN-3. The two domains were connected by flexible linker that contained a cleavage site for prostate-specific antigen (PSA). PSA is an enzyme secreted from prostate cancer cells and is activated in tumor microenvironments because serum protease inhibitors bind to (and thus inactivate) PSA in the systemic circulation [15]. This CPE-based protoxin was cytotoxic to cells that expressed both CLDN-3 and -4, and PSA yet showed reduced adverse effects.

Because CPE is a bacterial protein, it induces an immune response. This immunogenicity limits the clinical application of native CPE and its derivatives because the resulting antibody response may complicate the prediction of their pharmacokinetic–pharmacodynamic profiles and because the immunogenic compounds might induce complement- or IgE-mediated hypersensitivity [44]. A previous domain-mapping analysis of CPE revealed amino acids 16–65, 91–170, and 286–305 as antigenic epitopes, but the minimal CLDN-binding domain of CPE (amino acids 290–319) is considered to have low antigenicity [80]. Subsequent conjugation of this 30 amino acid peptide with tumor necrosis factor alpha increased its CLDN-targeting ability and its specific cytotoxicity against cells that express CLDN-3 and -4 [96]. In another study, screening of a phage display library yielded two CLDN-4-binding 12 residue peptides whose amino-acid motifs were similar to the minimal binding domain of C-CPE [50]. Future efforts focused on minimizing the CLDN-binding domain of CPE might overcome its antigenicity.

The role of C-CPE as a template for creating new CLDN binders

Native CPE binds to a broad range of CLDNs, leading to a risk of systemic toxicity. Consequently, a CLDN-4-specific CPE derivative might lack hepatotoxic effects, because CLDN-4 is not expressed in liver [51]. Similarly a CLDN-6-specific CPE mutant might accumulate preferentially in tumor tissues because normal differentiated cells do not express CLDN-6 [3]. To generate new C-CPE derivatives with narrow CLDN-specificity, several groups have applied a mutagenesis approach; the resulting CPE mutants N309A/S313A, L254A/S256A/I258A/D284A, L223A/D225A/R227A, and Y306W/S313H showed high affinity for CLDN-4 and high specificity for CLDN-4, -3, and -5, respectively [66, 86, 92]. In addition, recent findings regarding the structure of mouse CLDN-15 (Protein Data Bank [PDB] code, 4P79) and of the mouse CLDN-19–C-CPE complex (PDB code, 3X29) indicate that C-CPE actually binds to both the first and second extracellular domains of CLDNs and not solely to the second extracellular domain of CLDNs, as had been thought initially (Fig. 4a) [17, 71, 81, 94]. Currently the N-P-(V/L)-(V/L/T)-(P/A) motif in the second extracellular domain of CLDN is considered to determine the sensitivity of CLDNs for CPE, whereas the (A/N/S)-I-(I/L/V)-(T/V) motif in the first extracellular domain of CLDN is thought to support the interaction between CLDNs and CPE by enlarging the hydrophobic contact area [71]. In this regard, mutations within the (A/N/S)-I-(I/L/V)-(T/V) motif of mouse CLDN-19 clearly attenuated its interaction with C-CPE to the same extent as mutation in the N-P-(V/L)-(V/L/T)-(P/A) motif [71]. Further knowledge regarding the structures of other CLDNs and CLDNs–C-CPE complexes will enable the generation mono-CLDN-specific C-CPEs and their derivatives.

Fig. 4
figure 4

Structure of the C-CPE:CLDN-19 complex. a The structure of the C-CPE:CLDN-19 complex (PDB code, 3X29) by X-ray crystallography. The regions of CLDN-19 that are involved in binding CPE (red) and those of C-CPE that are involved in binding CLDN-19 (blue) are shown. b A schematic diagram of CLDN structure. Regions involved in the interaction with C-CPE are circled

Full size image

Conclusion

Since its discovery as a CLDN binder, CPE has connected the dots leading from the conceptualization of CLDN-targeted cancer therapy to its potential realization as a novel therapeutic and diagnostic modality. Many studies have shown that CPE-mediated cytolysis can be an effective anti-tumor treatment for CPE-sensitive CLDN-overexpressing cancers, regardless of whether the lesions are chemotherapy-sensitive or -resistant. In particular, CLDN-3, -4, -6, and -7 are CPE-sensitive and frequently overexpressed in various cancers. In addition, C-CPE is a useful adjunct cancer therapy because it enhances the permeability of anti-tumor drugs into tumor tissues and increases the sensitivity to such drugs at the single-cell level. Alternatively, probe-conjugated C-CPE efficiently detects micrometastatic cancers and hyperplastic lesions. However, despite these diverse potential applications, neither CPE itself nor any CPE-based derivative is currently under study in a clinical trial. To advance CPE-based cancer diagnosis and therapy, new CLDN binders with high druggability need to be developed.