To understand the critical gene expression changes induced in microvascular endothelial cells by brain tumors we sought to identify whether endothelial marker genes were similarly expressed in the vasculature of both primary and metastatic brain tumors and whether these genes were present in both high grade and low grade (PA) glial tumors. Our results demonstrate that a subset of 13 of our 20 putative GEMs do appear to be differentially regulated in bulk tumor and that this upregulation is preserved in both primary and metastatic brain tumors. These genes include AKAP13, COLA1, COL3A1, COL4A1, COL6A2, CXCR7, HSPG2, ITGA5, MMP14, PV1, PXDN, SOX4 and TEM1. While the majority of these genes have been previously identified by ourselves or others as being expressed by normal or tumor vascular structures, we believe this is one of the first reports of AKAP13 expression in normal brain and brain tumor blood vessels. Furthermore, we believe this study is the first confirmation of PXDN expression in brain tumor vessels. While none of the individual GEMs was able to perfectly discriminate between normal brain tissue and malignant brain tumor in all specimens, the panel of GEMs as a whole was significantly upregulated in tumor specimens compared to normals. The degree of upregulation was actually greater in our panel of metastatic brain tumors compared to primary glioblastoma. This may be due to the greater tumor heterogeneity of glioblastoma. The location of the tissue microarray specimen relative to the entire tumor as it existed in situ is unknown for these de-identified specimens. As such it is possible that the tissue microarray spot came from a more quiescent portion of the tumor that may not be as actively angiogenic. Because our mRNA samples for qPCR were generated from larger bulk tumor specimens, such microheterogeneity is less likely to influence our qPCR results that should be more reflective of average tumor gene expression levels. Interestingly, the degree of tumor GEMs upregulation was similar in our specimens of pilocytic astrocytoma, a benign but often highly angiogenic primary brain tumor, when compared to malignant brain tumors. This would seem to indicate that the gene expression changes are influenced by the level of the angiogenic process itself and not by the histologic grade of the tumor cells.
It is important to note that we did not perform transcriptional or translational studies to confirm whether the increased abundance of a particular microvascular GEM was due to elevated gene product per individual cell or due to an increased number of vessels within the bulk tumor. However, we did normalize all qPCR values to vWF, a known endothelial marker gene, in order to account for the increased vascularity of the tumor samples and make it more likely that higher expression levels were due to altered gene expression profiles and not simply an higher vascular index. Similarly, comparison of the IHC and ISH staining data to the staining score for vWF confirms that in many cases, GEMs expression profiles were regulated independently of simple tumor vessel numbers, suggesting a selective angiogenic response to CNS tumorigenesis.
We were unable to verify differential regulation of seven of our putative GEMs identified in our SAGE study. In most cases this was due to high baseline expression of the gene in question in normal brain tissue. We believe that while these GEMs may still be differentially regulated at the microvascular level as indicated in our SAGE study of purified endothelial cells, the basal expression of these genes in normal brain and tumor cells themselves, makes them difficult to assess in bulk tissue specimens or pathologic sections. Techniques such as microdissection of tumor specimens and antibody coupled magnetic bead isolation of tumor endothelial cells from gross specimens could clarify expression differences for these genes, but their high background expression levels likely indicates that they would not be selective enough targets for therapeutic manipulation.
Based on a combination of their biology and frequency of upregulation, several of our GEMs represent particularly attractive targets for both primary and metastatic brain tumors. For example, the SOX4 gene expression level in GBM and metastatic tumors was up to 3-fold higher than those in non-neoplastic tissue (; Supplemental Material 1
). Furthermore, in several samples, SOX4 appeared to show expression in tumor cells themselves. The SOX proteins compose a family of more than 20 transcription factors characterized by the presence of a high-mobility-group DNA binding domain [14
]. SOX4 belongs to the C group of the SOX gene family and plays a role in normal embryonic development of many tissues, including the CNS. It is over-expressed in medulloblastoma [15
] and may be predictive of outcome in medulloblastoma, although this remains controversial [17
]. One interesting recent study showed that micro-RNA 335 suppressed migration and metastasis of breast cancer cells through targeting of SOX4 and extracellular matrix component [19
]. SOX4 is also being evaluated as a potential vaccine target in lung cancer, further indicating its potential clinical application [20
]. However, relatively little is currently known concerning the role of SOX4 in tumor angiogenesis. SOX4 induction in human bladder carcinoma HU609 cells is associated with over-expression of vascular or angiogenesis related genes including IL-8, CTGF, JAG1 and NRP2 [21
]. In a mouse knockout model, lack of Sox4 led to disruption of the vascular and cardiac systems [22
]. The combined data suggest that SOX4 may contribute to brain tumor angiogenesis but the mechanism remains unclear.
Our present study also confirmed upregulation of multiple cell surface proteins in the microvasculature of primary glial and metastatic brain tumors. One was PXDN, also known as MG50, one of the few melanoma-associated antigens that is not a differentiation antigen or a mutated protein. Recent genomic evidence revealed that MG50 was identical to PXDN, the human homologue of the Drosophila
gene peroxidasin, an extracellular matrix-associated peroxidase. Its expression is relatively restricted to tumors such as melanoma, breast cancer, ovarian cancer as well as the glioblastoma cell line U138MG, and is absent from many normal tissues [23
]. However, its function has yet to be fully defined. Mitchell et al. [24
] reported that PXDN encodes an interleukin 1 receptor antagonist containing six epitopes recognized by human cytolytic T lymphocytes. Another study identified PXDN as being induced during p53-dependent apoptosis, suggesting a link between PXDN expression and reactive oxygen production in apoptosis [25
]. Previously we showed that PXDN (MG50) expression was 17-fold higher in glioma endothelial cells than non-neoplastic brain endothelium [10
]. In the present study, we confirmed that PXDN mRNA was up-regulated in an independent set of brain tumor samples () and expression localizes to microvascular endothelial cells (). Similarly, Castronovo et al. [26
] reported that PXDN localizes in vasculature structures in renal cell carcinoma, thus PXDN may function in tumor angiogenesis in multiple cancers. Since PXDN is a surface protein accessible to extracellular pharmaceutical compounds, it may prove to be a novel target for anti-angiogenesis therapy.
Extracellular matrix (ECM) remodeling is critical for the endothelial cell activation and migration that contribute to angiogenesis [27
]. By using laser capture microdissection microscopy and microarray techniques, Pen et al. [28
] found IGFBP7 and SPARC, both of which encode proteins that are secreted into the extracellular space, were upregulated in GBM vessels. In the present study, we found several ECM components were upregulated in brain tumor samples including four distinct members of the collagen superfamily. One of these, COL4A1, was upregulated in brain tumor samples in a manner consistent with previous reports [29
]. COL4 is believed to play an important role in angiogenesis and tumor progression [30
]. Both COL4 and COL5 are expressed along tumor vessels in glioma [31
]. COL4 colocalizes with TEM1, another GEM, and MMP2 during fetal brain angiogenesis [32
]. Small-molecule inhibitors of COL4 biosynthesis can prevent endothelial tube formation and tumor growth [33
]. Structural abnormalities or cryptic sites in COL4 have been detected in the tumor microenvironment and may be therapeutically exploitable [34
]. Similarly, when treated with brivanib alaninate, an anti-angiogenic agent targeting vascular endothelial growth factor receptor 2 (VEGFR-2) and fibroblast growth factor receptor 1, athymic mice bearing L2987 human tumor xenografts showed significant reduction of COL4A1 at both mRNA and protein levels [36
]. Thus, COL4A1 may be an important biomarker of angiogenesis in cancer as well as a viable therapeutic target.
We similarly confirmed MMP14 (membrane type 1 metalloprotease, MT1-MMP) induction in primary and metastatic brain tumor samples by both real time PCR and immunohistochemistry (; ). This upregulation is supported by additional reports [37
]. MMP14 is a transmembrane metalloprotease that plays a key role in the angiogenesis response through multiple steps including degradation of ECM, endothelial invasion/migration formation of capillary tubes and recruitment of accessory cells [40
]. Multiple studies demonstrate the important role of MMP14 in tumor angiogenesis. In human melanoma cells, MMP14 overexpression is associated with increased in vivo tumor growth and vascularization [41
]. MMP14 overexpression results in up-regulation of VEGF and promotes angiogenesis in human GBM and breast cancer xenograft models [42
]. In human GBM, VEGF and MMP14 often show co-expression by the same tumor cells [38
]. The combined data suggest that the conserved expression pattern of MMP14 in primary and metastatic brain tumors make this gene an attractive target for CNS neoplasms.
In conclusion, the principle goal of the present investigation was to compare gene expression patterns of microvascular marker genes in primary brain tumors versus metastatic brain tumors. Primary brain tumors grow invasively as single cells along blood vessel walls, although they do not invade the vessel wall itself. Conversely, systemic metastatic tumors tend to invade the brain as small groups of cells, which show little tendency to grow along blood vessel tracks, but do invade the vessel walls [44
]. Certain investigators have reported significantly different gene expression profiles in primary lung cancer specimens compared with matched metastatic brain tumors from the same patient [45
]. It is certainly possible that further SAGE or gene array based studies of human metastatic brain tumor endothelium or animal models of “classic” versus “cooptional” tumor angiogenesis might reveal new, metastasis-specific angiogenic genes. However, we hypothesized that brain angiogenesis would be more likely to progress along a common preserved pathway regardless of the identity and growth characteristics of the inciting tumor and thereby provide molecular targets shared by a majority of brain tumors. We demonstrate here that despite their original sites, primary and metastatic brain tumors induce similar gene expression changes in several novel microvascular protein markers. Development of new drugs targeting these genes may benefit both primary and metastatic brain tumor patients and the treatment of metastatic brain tumors using anti-angiogenic drugs currently being tested for malignant glioma is likely to have a molecular justification as well.