TEM1/endosialin/CD248 encodes a 165 kD single-pass transmembrane glycoprotein. It is a C-lectin type receptor with homology to thrombomodulin. Several studies identified TEM1/endosialin as a marker of tumor endothelium and isolated activated fibroblasts within the tumor stroma [7, 9]. Little to no expression of TEM1/endosialin was noted in normal adult tissues. Alternate studies have argued that TEM1/endosialin in the tumor vasculature is expressed by vessel-associated pericytes [8, 10, 19]. Two additional reports identified TEM1/endosialin in populations of VEGFR2+/CD31+/CD45-/VE-cadherin+ endothelial precursor cells derived from CD133+/CD34+ cells or in CD45+/VE-cadherin+ "vascular leukocytes" [20, 22]. Both cell types have been proposed to contribute to tumor vascularization, although the degree to which either one incorporates into tumor vessels remains controversial. Human umbilical vein endothelial cells immortalized with SV40 large/small T-antigen, a model of proliferating neovascular endothelial cells, also express TEM1/endosialin . We argue that TEM1/endosialin is expressed to varying degrees by tumor endothelial cells, pericytes and stromal fibroblasts. In normal blood vessels, the pericytes bind tightly to endothelial cells and influence migration, stabilization and maturation of the endothelium. In the cerebrovasculature, pericytes are additionally critical for maintenance of the differentiated blood-brain barrier (BBB) and failure of the BBB in brain tumors is associated with disruption and loss of pericytes lining endothelial cells. Pericytes are of emerging interest as therapeutic targets as vessels may be more sensitive to chemotherapy or radiotherapy when denuded of pericytes. When Bergers, et al., targeted both the endothelial cell and pericyte compartments with combinatorial receptor tyrosine kinase inhibitory (RTKI) therapies, they achieved more favorable responses than with either therapy alone . Cao, et al., demonstrated that systemic overexpression of Ang2 blocked pericyte recruitment to tumor vessels and led to massive vascular destabilization and tumor suppression . The collective data support the hypothesis that TEM1/endosialin has a favorable expression profile for targeting both the endothelial and pericytic portions of the tumor vasculature.
Expression of TEM1/endosialin by a subset of tumor fibroblasts may also influence tumor growth and angiogenesis. Activated tumor fibroblasts contribute to the deposition of fibronectin, collagen I and collagen IV, all of which may bind TEM1/endosialin and guide vessel invasion, migration or differentiation. SDF-1 expression by activated tumor fibroblasts has been shown to drive tumor progression and angiogenesis in breast carcinoma models and has been reported to recruit endothelial progenitor cells to the microvasculature of murine gliomas, where differentiation into both pericytes and endothelium was detected [26, 27]. A recent genetic study reported that TGFβ1 treatment of proliferating endothelial cells led to an endothelial-mesenchymal transition, wherein endothelial cells downregulated CD31 and upregulated fibroblast-specific protein-1 (FSP-1). They confirmed the presence of FSP1+/CD31+ intermediate cells in tumor models . Further work is necessary to fully determine the complete expression profile of TEM1/endosialin and its relationship to these intricately related stromal cell populations.
In the US, primary brain tumors are diagnosed in approximately 40,000 new patients a year. In addition, brain metastases occur in 20-40% of cancer patients, leading to 200,000-400,000 new cases of brain metastases a year. Despite differences in the molecular fingerprints of the tumor cells themselves, it may be possible to identify common changes in the vasculature of these disparate tumor types. Initial studies by ourselves and others reported TEM1/endosialin upregulation in human brain tumors [1, 5, 7]. We wanted to expand these findings in order to better establish the full extent of TEM1/endosialin expression in brain tumors. This study confirms expression of TEM1/endosialin at the RNA and protein level in astrocytic and metastatic brain tumors, with little to no expression in normal control brain. Expression was primarily limited to the tumor vasculature but occasional stromal cells also stained for TEM1/endosialin. Greater than 30% of GBMs displayed robust TEM1/endosialin expression, while lesser percentages of lower grade tumors expressed TEM1/endosialin. Unfortunately, we did not know whether the individual specimens analyzed were from the tumor center or closer to the tumor margin. Preliminary mouse xenograft data suggested that Tem1/endosialin staining was increased along the tumor margin, but this needs to be further investigated in both clinical and mouse tumors. It will be especially useful to assess Tem1/endosialin expression in future experiments using xenografted primary glioma spheroids or glioma stem cells, which grow as a more highly invasive intracranial tumor than the U87 model, although with less pronounced angiogenesis. Interestingly, one pilocytic astrocytoma displayed robust expression of TEM1/endosialin. Although pilocytic astrocytomas are considered low-grade tumors, they are highly angiogenic and often enhance on MRI, which is indicative of vascular leakiness, disruption of the blood-brain barrier and often more aggressive disease. A recent study in breast cancer identified an association between TEM1/endosialin expression and progressive and/or recurrent disease as well as nodal involvement, all hallmarks of a more aggressive disease state . Thus, TEM1/endosialin may have prognostic value for breast cancer. We did detect a statistically significant (p = 0.004) inverse correlation between TEM1/endosialin expression in astrocytomas and patient age, although the implications of this remain unclear. While diagnosis at a younger age can be a predictor of long-term survival for GBM , our initial analyses reveal no difference in survival between TEM1/endosialin expressing and non-expressing tumors.
Tomkowicz, et al. recently identified collagen types I and IV and fibronectin as potential ligands for TEM1/endosialin . They showed that CHO cells engineered to overexpress TEM1/endosialin (CHO-TEM1) exhibited greater adhesion to fibronectin than control CHO cells (CHO-K1). They suggested a model by which TEM1/endosialin is induced in cells associated with endothelium and stromal fibroblasts. The subsequent interactions between TEM1/endosialin expressing cells and fibronectin, collagen I or IV present in the extracellular matrix drive cellular migration required for tumor invasion and angiogenesis. In an independent study, TEM1/endosialin and collagen IV were shown to colocalize in finger-like processes during angiogenic development in the human fetal telencephalon . We had identified collagen I and collagen IV as putative glioma endothelial markers which exhibited a minimum 4-fold induction as compared to normal endothelial cell samples and were induced in both glioma and colon carcinoma endothelium . Therefore, we were curious to determine whether we could detect co-localization of fibronectin with TEM1/endosialin in our primary and secondary tumor samples. We did detect upregulation of both TEM1/endosialin and fibronectin within the same tumor specimens, although only a subset of the samples showed co-localization along the vasculature, supporting a physiologic interaction of these proteins. In addition, Tomkowicz and colleagues showed that the CHO-TEM1 cells, but not CHO-K1 control cells, grew with web-like morphology and formed large clusters when cultured in matrigel. The CHO-TEM1 cells also exhibited enhanced migration through matrigel that was blocked by incubation with a humanized antibody targeted against TEM1/endosialin . Others have shown that incubation with anti-TEM1/endosialin antibodies can block tube formation by endothelial progenitor cells and mesenchymal stem cells [17, 20]. Our in vitro data showed that human endothelial cells grown in matrigel induce TEM1/endosialin themselves in response to the culture environment. Together, the biochemical and localization data support a functional role for TEM1/endosialin in promoting cell-cell interactions and migration during tumor angiogenesis.
Since the expression profile of TEM1/endosialin and existent xenograft data suggested TEM1/endosialin might be a viable therapeutic target for brain tumors, we wished to ascertain whether TEM1/endosialin induction was conserved in an orthotopic model of GBM. Intracranial U87MG xenografts demonstrated robust upregulation of Tem1/endosialin in the tumor vasculature with little to no expression in non-neoplastic brain. Again, staining was predominantly present in the vessels, although we could detect staining in occasional tumor stromal cells. Interestingly, when we implanted U87MG cells engineered to stably overexpress VEGF, a small percentage of tumor cells stained positive for human TEM1/endosialin, although there was no significant difference in Tem1/endosialin staining in the mouse vessels (BNW, unpublished data).
When we implanted intracranial GBM xenografts into nude Tem1/endosialin KO and WT mice, we detected no difference in tumor take or survival. This differs from the colon carcinoma data and suggests that the impact of Tem1/endosialin in the cerebral microenvironment may differ from its role in intestinal tumorigenesis. Nanda, et al., reported a significant decrease in tumor take and volume when they implanted colon carcinoma xenografts into the large intestine of nude Tem1/endosialin KO mice . However, they did not detect a difference in tumor growth when identical cells were implanted subcutaneously into KO and WT animals. They argued that this emphasized the importance of an anatomically relevant microenvironment for tumor growth and study. This is a similar argument to that proposed by Camphausen, et al., who compared the expression profiles of U87MG and U251MG GBM cell lines grown in vitro to those from cells grown as subcutaneous or intracranial xenografts . They noted that the expression profiles in tissue culture were significantly different from the subcutaneous profiles, which were, in turn, distinct from the intracranial expression profiles. The disparity between GBM xenograft expression profiles suggested that the orthotopic implantation site can dictate gene expression, presumably due to signaling between the tumor cells and the surrounding stroma. In the current scenario, it would again appear that TEM1/endosialin expression levels within distinct anatomical compartments differentially influence tumor growth and pathology. Importantly, the data suggest that inhibition of TEM1/endosialin function alone in tumors is insufficient for suppression of brain tumor growth.
Despite unchanged tumor growth and equivalent animal survival between Tem1/endosialin WT and KO animals, we did detect an increase in the number of microvessels present in the KO GBM tumors. This strengthens a role for TEM1/endosialin in the maturation of the tumor neo-vasculature . TEM1/endosialin was recently reported to be upregulated by hypoxia via a HIF-2 pathway in a limited number of cell lines, although we could detect no induction in primary human endothelial cells . However, the brain is a highly vascularized site, and it may be that the level of oxygenation is sufficient to promote xenograft growth in the absence of Tem1/endosialin induction. It is more likely TEM1/endosialin is directly or indirectly required for differentiation or maturation of the vasculature, as opposed to strict proliferation, and that lack of TEM1/endosialin protein leads to the overabundance of vessels demonstrated in the knockout mouse tumors. Further studies are needed to determine if the increased vascularity of the Tem1/endosialin KO tumors could render them more sensitive to radiation or chemotherapy, by enhancing the delivery of radiosensitizing oxygen or therapeutic drugs.