Alterations in the expression levels of tight junction proteins, especially claudins, continue to be reported in several cancers. However, an overall view of claudin gene expression in normal and cancer tissues has been lacking. In this report, we first use the large public SAGE database to investigate claudin expression of the 21 human CLDN genes we have identified in GenBank. We find that, while some CLDN genes are ubiquitously expressed, the majority of these genes exhibit a very restricted expression pattern. CLDN14,16,17,19,20, and 22, for example, are found in only a few rare libraries. Others such as CLDN3,4,5,7,11, and 12 are much more widely expressed. Our analysis allows for the identification of general expression patterns, such as the high expression of CLDN3,4 and 7 in epithelial tissues, and lower expression in other tissues, such as the brain.
Our data also reveal claudin expression patterns that were not previously known and that may have clinical implications. According to our data, gastric cells (both normal and neoplastic) express high levels of CLDN18
, while other tissues do not express this gene. Interestingly, a recent study shows that claudin-18 is highly expressed in normal gastric cells and that this high expression is retained in approximately half the gastric tumors [17
]. Because of its highly restricted pattern, claudin-18 may therefore represent a useful target for therapy of gastric cancer, especially in those tumors that maintain high levels of this gene. Claudin-18 is likely involved in TJ formation in normal gastric cells, while cancer cells, which typically do not form TJ's, may have a more available form of claudin-18. Therefore cancer cells may be more sensitive to therapy involving the targeting of this molecule. We also find that CLDN5
is not generally expressed in epithelial tissues but is expressed at high levels in all vascular endothelial cell libraries analyzed. Although also expressed in the brain, CLDN5
may represent a target for antiangiogenic therapy, especially if using compounds that cannot cross the blood-brain barrier.
Our RT-PCR experiments provide a more quantitative look at claudin gene expression in several normal and neoplastic tissues. It is important to note that these RT-PCR investigations do not represent an exhaustive study of CLDN gene expression, but rather a survey of expression in a large number of different human tissues. Follow-up studies on multiple samples for these different malignancies will be necessary to clearly establish the extent and levels of expression of claudins in these tissues. However, it is important to note that the patterns of gene expression obtained by real-time RT-PCR (Figure ) closely mirrors the in silico findings using the SAGE Genie database (Figure ). For example, the CLDN3 and CLDN4 expression patterns are consistent between the two analyses (as described above). In addition, we also observe high correspondence in the two approaches when examining CLDN5 expression, which appears to be especially high in normal brain and brain cancer. Interestingly, when clustering our RT-PCR data for gene expression patterns, we find CLDN3,4,7 are very similar in their expression, suggesting coordinate regulation. The fact that the CLDN3,4,7 cluster is present in both normal and tumors suggests that the mechanisms that lead to the coordinated expression of claudins in normal cells is conserved in tumor cells, although it may be inappropriately activated in cancer. It will certainly be interesting to elucidate the mechanisms that lead to the inappropriate activation of these genes.
Our in silico
and RT-PCR results are consistent with numerous previous reports showing that CLDN3
are overexpressed in breast [11
], ovarian [7
], and prostate tumors [9
]. In addition, our data showing overexpression of CLDN4
in pancreatic cancer is also in agreement with previous reports [8
]. The finding of expression of these claudins in other tumors, such as bladder, thyroid, fallopian tubes, stomach, colon, and uterus, is novel and warrants further investigation. CPE-based therapy, which specifically targets cells expressing claudin-3 or claudin-4 [8
], may be worth exploring in these malignancies as well. The fact that CLDN3
, and CLDN4
are expressed in several normal tissues (Figure ) certainly suggests that systemic administration of CPE may have significant toxic effects. However, the therapeutic index of this compound will depend on the level of up-regulation in the various tumors under study and the mode of administration. In ovarian cancer, for example, where both CLDN3
are highly up regulated and where intraperitoneal therapy is possible, CPE treatment is certainly an interesting possibility.
In this report we study the expression of the CLDN
genes at the mRNA level, but it will obviously be essential to validate these findings at the protein level when all the antibodies are available, as posttranslational mechanisms have been shown to regulate claudin protein levels and localization [5
]. In addition, it will be important to investigate the various claudins studied here for their potential clinical use in cancer therapy and diagnosis. With over 20 known members, many of which, as we show in this report, exhibit high tissue-specific expression and deregulation in various cancers, the claudin family of membrane proteins may represent ideal targets for cancer diagnosis and therapy.