Endosialin, also referred to as TEM1, was originally discovered as a human embryonic fibroblast-specific antigen and later reported to be expressed in the endothelium. Endosialin is barely detectable in normal tissues other than its moderate expression in the smooth muscle of colon and prostate . Therefore, TEM1 was once considered as an important candidate as a vascular target [6, 9, 19]. However, recent studies have demonstrated that endosialin is expressed in pericytes of breast tumours and brain gliomas, and not selectively in tumour endothelium [10–12]. In the present study, endosialin was prominently expressed in the tumour-associated stroma, especially in fibroblasts and blood vessels, and only weakly expressed in the stroma of distant or adjacent normal mucosa in both the non-RT group and in the RT groups. When compared to the normal mucosa, endosialin expression was much higher in tumour, thus agreeing with other reports on colon and breast cancers [20, 21]. Up-regulated endosialin in the tumour-associated stroma may play a role in the tumorigenesis of rectal cancers.
In the present study, endosialin expression in the stroma was more frequently observed in advanced TNM stages, and positively related to the tumour cellular expression of Cox-2, p73, and PRL in the RT group, whereas there were no such associations in the non-RT group. The level of endosialin expression in the tumour-associated stroma was significantly higher in breast cancers with nodal involvement compared to those with negative nodes . Endosialin has also been found in glioblastoma multiforme, anaplastic astrocytomas, and metastatic carcinomas that are of highly invasive activity [12, 22]. Endosialin is also more abundant in melanoma metastases than in the primary tumours . In colorectal cancer, endosialin was up-regulated in Dukes' B compared to Dukes' A . Cox-2 is an inducible isoenzyme of cyclooxygenase that is undetectable in normal colonic mucosa but is overexpressed in 80% of colonic tumours . Cox-2 is involved in a multistep process of colorectal tumorigenesis, such as apoptosis inhibition of cellular proliferation and angiogenesis enhancement, tumour cell invasion and differentiation. One of our previous studies has shown that Cox-2 expression is higher in more advanced tumours . It is interesting to see, in the further study, whether or not endosialin and Cox-2 have interactions in the tumour development, especially in enhancing angiogenesis. We have also found that p73 expression is increased during the development of colorectal cancers and its overexpression is further associated with poor prognosis in patients . PRL, which stimulates the Rho signalling pathway to promote cell motility and invasion , is up-regulated in colorectal cancer and associated with tumour invasion and metastasis . In the same series of the patients, our previous studies have demonstrated that tumours with p53-negative expression (wild type p53), or p73-negative expression, or weak Cox-2 expression had less local recurrence after RT [16, 17, 27]. PRL expression is related to distant recurrence and poor survival after RT . In further studies of colon cancer cell lines, we found that, after radiation, the antiapoptotic ΔNp73 and mitosis factor PRL-3 increase  and the overexpression of ΔNp73β increases the viability of cell lines and cisplatin induces the degradation of ΔNp73β in a dose-dependent manner (unpublished data). All these results indicate that certain biological factors may be involved in response to therapy in rectal cancer patients. If we could further confirm the data obtained from studies of these biological factors in relation to the clinicopathological issues above as well as their pathways in therapies, we may be able to apply them to clinical practices where rectal cancer patients may receive individual therapies based on their biological profile. For example, the targeting of multiple biological factors instead of only one in certain therapies may yield greater responses to them.
Radiated stromal fibroblasts in the carcinogenic process have been shown to induce sub-lethal DNA damage. Radiation-induced alterations of the stroma have been found to produce more mammary carcinomas compared to non-radiated stroma . All of these results hinted that the patients with up-regulated expression of endosialin in a tumour-associated stroma may be linked to other biological variables that are related to aggressive characterization after radiation. However, we did not find that endosialin expression was further related to survival in either the whole group or subgroups (with or without RT) of patients. This may be partly due to the limited number of the patients and/or the hypothesis that endosialin may play a survival role in certain groups of the patients. It would be interesting to determine the survival significance of endosialin in subgroups where the endosialin is related to the clinicopathological variables, such as TNM, Cox-2, p73 and PRL. We did find that endosialin presented different results in each subgroup, but firm conclusions are difficult to draw at this point due to the limited number.
In the present study, we also observed endosialin expression in tumour cells. Positive endosialin expression in tumour cells was more frequently observed in tumours with infiltrative growth pattern compared to expansive growth pattern, regardless of RT. Endosialin was positively related to p73 expression in the non-RT group, and to PRL expression in the RT group. Why was stromal endosialin positively related to p73, PRL and Cox-2 in the RT-group, but not in the non-RT group? One speculation may be that radiation influenced biological factors by up-regulating or down-regulating expression. After radiation, "bad factors," such as Cox-2, p73, and PRL-3 increased their expression. This indicates that there may be some mechanism by which tumors try to protect themselves by increasing the expression of "bad factors" against the damage of radiation. Furthermore, why did stromal endosialin and tumour cellular endosialin have different relationships with biological factors (p73 and PRL)? For example, stromal endosialin was related to p73 expression in the RT group, whereas tumour cellular endosialin was related to p73 in the non-RT group. One possible explanation may be that the effects of radiation on stroma differ on different types of tumour cells, resulting in different expression and relationships to the biological factors investigated. In addition, the p73 gene contains two promoter regions, giving rise to a p53-like protein named TAp73, and the N-terminally truncated ΔNp73, which lacks the transactivation domain and p53 homology. ΔNp73 is thought to have a regulatory function, down-regulating both TAp73 by protein-protein interaction and p53 by competitive binding with DNA. In this autoregulatory loop, ΔNp73 protein is also up-regulated by both TAp73 and p53. The functional cooperation among these family members seems to vary depending on cell types, stimuli and p53 status . In other words, the roles of the two isoforms may depend on their locations in the stroma or tumour cells.
Recent studies have raised the concept of the coevolution of tumour cells with tumour-associated stroma. The stromal environment of tumours appears to be a leading factor, and not just a supporting one in the initiation of tumours . The tumour microenvironment and interactions between tumour and stromal cells have a reciprocal relationship in tumour development and progression. Another interesting concept is epithelial-mesenchymal transition (EMT), a process by which cells lose their polarized epithelial structures and concomitantly acquire a migratory or mesenchymal phenotype. EMT is essential for normal embryonic development and progression of non-invasive adenomas into malignant, metastatic carcinomas. Alterations in cell-cell adhesion, cell-substrate interaction, extracellular matrix degradation and cytoskeleton organization are the major events that occur during EMT . Overexpressed thymosin β4 (Tβ4) induces EMT in colorectal carcinoma by increasing integrin-linked kinase (ILK) complex formation with particularly interesting new cysteine-histidine rich protein (PINCH) . We have studied PINCH expression in colorectal carcinomas and found that PINCH overexpressed on fibroblasts in the tumour-associated stroma compared to its expression in normal mucosa , similar to endosialin expression in tumour versus normal mucosa. In the present study, endosialin was expressed in the both tumour-associated stroma and tumour cells. Furthermore, endosialin expression in the tumour-associated stroma was positively correlated with that in tumour cells, giving more information regarding the interactions of endosialin between tumour microenvironments and tumours. Changes in signal conduction related to endosialin appear to play an important role in enhancing tumour progression after radiation. There was evidence that antiangiogenic therapies targeting both endothelial cells and pericytes were more effective than single-agent therapies . In the present study, endosialin was not only obviously up-regulated in stroma but also up-regulated in tumour cells when compared to normal mucosa. If the therapies target endosialin in both stroma and tumour cells, they may provide more efficient strategies of therapy for the patients with rectal cancers.
Regarding the discrepancies of endosialin localization, in stroma and/or in tumour cells found in different studies, several factors may be responsible, including the number and clinicopathological characteristics of the patients included in such studies, as well as the methods used in the same. For example, if tissue arrays were used for staining endosialin , it is possible that a limited sample of tissue was obtained from the tumour blocks and the selected arrays may not be representative of the complete characteristics of the tumour because of tumour heterogeneity. In the present study, we used ordinary sections for staining endosialin and observed that only 34% of the cases showed positive endosialin expression in tumour cells. In fact, in some positive cases, only a few positive tumour cells were determined in the entire tumour sections. In addition, some studies which employed real-time PCR or quantitative real-time PCR methods could not determine the location of the expression [20, 21]. In the present study, we used immunohistochemical staining, which is one of the best methods of identifying the location of endosialin expression.