Gliosarcoma: distinct molecular pathways and genomic alterations identified by DNA copy number/SNP microarray analysis

Glioblastoma (GBM) is the most aggressive primary malignant adult central nervous system tumor [1, 2]. Gliosarcoma, a histologic variant, accounts for 2% of all GBMs. Overall survival of gliosarcoma is similar to GBM except that systemic metastasis and skull invasion have been reported more frequently in gliosarcoma [3]. However, gliosarcoma exhibits several unique morphologic, immunohistochemical and molecular characteristics. Gliosarcoma is defined histologically as having biphasic neoplastic components; displaying both mesenchymal (fibroblastic, osseous, muscle or adipose differentiation) and glial differentiation [2]. The genomic alterations implicated in the malignant transformation of astrocytes are diverse, of which, IDH, TERT, EGFR, CDKN2A, TP53, PTEN, PDGFRA and NFKB1A are the most commonly reported [3, 4]. These genes are involved in oxidative decarboxylation, maintenance of telomeres, stimulating protein tyrosine kinase, tumor suppression, cell signaling and a variety of other cellular processes. Gain of 7p in combination with 10q loss is associated with EGFR amplification and is a frequent finding in IDH-wildtype GBM. Of GBMs with EGFR protein overexpression, 70–90% demonstrate EGFR gene amplification. Other mutations such as the EGFRvIII and missense mutations involving the extracellular domain are distinct from EGFR mutations in non-glial cancers [5]. Interestingly, EGFR alterations vary amongst GBM subtypes, being rare in IDH-mutated GBM, and more prevalent in IDH-wildtype GBM [6]. Similar to IDH-wildtype GBM, gliosarcomas contain PTEN, CDKN2A and TP53 alterations, but amplification of the EGFR (epidermal growth factor receptor) are uncommon (4–8%) [7‐9]. Furthermore, the type of EGFR alterations reported in gliosarcoma are not usually seen in GBM; particularly EGFR point mutations have been detected in gliosarcomas (c.1831G > A) [10]. Amplification of EGFR is present in 35–45% of IDH-wildtype GBMs [2]. Of the detected EGFR amplifications in gliosarcoma, it is speculated that these results are derived from the glial component of the tumor rather than the sarcomatous component. Immunohistochemistry supports this, staining for EGFR is negative in the sarcomatous component and positive in the glial component [11]. Gains of chromosome 7 without EGFR amplification is frequent in gliosarcoma [9], leading one to believe the oncogenic driver for gliosarcomas may reside on chromosome 7, but not necessarily related to the EGFR pathway. The myriad of known candidate genes located on chromosome 7 (CDK6, PDGF-A, c-MET) may support this theory.

Given its poor prognosis, more research has been directed toward identifying specific mutations for targeted treatment of GBM. Several novel agents have been introduced that specifically target EGFR, but treatment with TKI’s and EGFR antibodies have not yielded successful clinical results. The hypothesis is that TKI’s work best for exon 19 and 21 mutations, which have not been detected in GBM [12]. The poor response of EGFR-targeted therapies raises the question of whether EGFR alterations truly represent key drivers in the genesis of glioblastomas. Therefore, we applied a whole-genome approach using the OncoScan^® Assay to examine DNA copy number alterations, and identify any chromosome regions known to harbor oncogenic drivers.

Overall, our data showed that DNA copy number losses were frequent and amplifications were infrequent in gliosarcoma. The majority of copy number loss occurred on chromosomes 9 and 10; localizing to regions containing CDKN2A and CDKN2B genes. The CDKN2A gene encodes for proteins p16 and p14arf, tumor suppressors that regulate the p53 and RB1 cell cycle [29]. The CDKN2B gene encodes for the p15^ink4b protein, a member of the p16^ink4 (CDKN2A) family, and a cell growth regulator that inhibits G1 progression [30]. Homozygous loss of CDKN2A is common in GBM (35–50% [2]) and loss of this locus occurred in 14 of 18 gliosarcoma specimens [31]. Other potential pathways involved in gliosarcoma include activation of the OX40L/OX40 pathway, which has been shown to induce strong immunity and antitumor effects in GBM [32]. DNA copy number loss was present in areas coding for several protein components of the OX40L/OX40 pathway (NF-kB2 (p52), IKK-alpha, NF-kB2 (p100), Calcineurin A (catalytic), PKC-theta, Perforin, NF-kB). Conversely, several regions that showed chromosome loss were areas that coded for WNT pathway proteins (Tcf (Lef), WNT, Sirtuin1, beta-TrCP, BMI-1, TCF7L2 (TCF4), DKK1). In the activated or overexpressed state, most of these WNT pathway proteins (with the exception of DKK1) promote cell proliferation and cell survival [33]. However, DKK1 (Dickkopf-1) is considered to be a negative regulator of WNT pathway, and has been implicated as a candidate gene that is epigenetically silenced in medulloblastoma [34], loss of which, may provide an avenue for WNT pathway activation with subsequent cell proliferation and survival. One study showed that DKK1 expression led to glioma cell sensitivity to chemotherapy-induced apoptosis [35]. Another protein with repetitive representation after pathway enrichment was NF-kß, a protein complex that controls DNA transcription, but can induce cell proliferation and anti-apoptosis if misregulated or constitutively activated. Aberrant activation of NF-kß in glioblastoma, leading to cell invasive capabilities, resistance to radiotherapy, and even promotion of mesenchymal phenotype has been reported [36]. Studies to evaluate the therapeutic effect of inhibition of NF-kß, based on these mechanisms have been published [37]. However, due to the multifactorial role NF-kß plays in a diverse number of biological processes (cell proliferation, survival, motility, DNA repair, inflammation), a direct pathway to neoplasia in GBM is unclear. Our data shows that there is copy number loss in the region encoding for NF-kß in gliosarcoma, implying that activation of the NF-kß pathway does not play a role in gliosarcoma genesis. However, it is possible that loss of NF-kß could lead to loss of DNA repair mechanisms, resulting in neoplasia.

Chromosome gains were the second most common copy number abnormality. Gains in regions coding for signaling molecules known to promote cell proliferation (F-actin cytoskeleton, actin cytoskeletal) and cytoskeleton remodeling (G-protein alpha-12 family, F-actin cytoskeleton) were present [38]. Interestingly, the Rho GTPase family, including rac1has been implicated in the modulation of glioma cell migration through cytoskeletal rearrangement [39]. Homeobox (HOX) genes are responsible for regulation of transcription factors during embryonic development, the differential expression of which, have been tied to many different cancers, including glioblastoma [40]. Specifically, high expression of HOXA9 and HOXA10 have been reported in human glioma cell lines [40]. The expression of HOXA13 has been proposed as an activator of WNT and TGF-ß-induced epithelial to mesenchymal transition in glioma progression [41]. Resistance to radiation and chemotherapy through the activation of HOXA, thereby inducing increased proliferation and decreased apoptosis in cultured glioblastoma cells has also been reported [42]. Our microarray data showed gains in the regions encoding HOXA1, HOXA3, HOXA7, HOXA11, and HOXA13. Interestingly, several studies have demonstrated overexpression of HOX genes in Ewing sarcoma [43, 44] and undifferentiated small round blue cell sarcoma [45]; perhaps raising the question of whether HOX genes play a role in mesenchymal phenotype in gliosarcoma.

Copy neutral loss of heterozygosity (LOH) events were less frequent. OncoScan™ detected loss of the region encoding p53 (17p13.1), and while p53 alterations have been well documented in gliomas, some studies suggests that LOH of a single cell cycle regulator, such as p53 may be insufficient for development of gliomas [46]. In the context of this study, concurrent loss of p16 with LOH of p53 may play a large role in the genesis of gliosarcoma. Another pathway that was well represented across LOH, gains and losses after MetaCore™ enrichment was chemotaxis lysophosphatidic acid signaling. Lysophosphatidic acid (LPA) is a phospholipid that binds to G protein-coupled receptors (GPCRs) leading to chemotaxis, cell proliferation, cell growth and cell survival through a complex network of signaling cascades [38]. LPA was shown to be increased in GBM stem cells from the subventricular zone via LPA/Rho signaling cascades and proposed as a mechanism for GBM invasion and angiogenesis, a possible therapeutic target using LPAR antagonist and LPA synthesis inhibitors [47, 48]. Another region with copy neutral LOH encoded for the network object Ephrin-B, a protein ligand known to promote angiogenesis, as well as in developmental processes such as axon guidance, cell migration, and maturation of cortical dendrites. The Eph receptor tyrosine kinases and ephrin ligands have been implicated in both the inhibition and promotion of neoplasia and may play a key role in glioma genesis [49].

Epidermal growth factor receptor (EGFR) is mutated or amplified in 35–45% of IDH-wildtype glioblastomas [2] and implicated as a key driver [50]. Gliosarcoma, however, does not exhibit amplification or mutations of EGFR at the same frequency, suggesting there may be additional/alternate mechanisms propelling tumorigenesis and potentially mesenchymal transition into a sarcoma phenotype. Our data supports other studies that show a very low prevalence of EGFR amplification in gliosarcoma, but did show frequent gain of chromosome 7 (72%) containing EGFR locus. It is important to note that amplification or mutation of EGFR is not necessarily required for EGFR activation. For example, copy number gain of SYNJ2 in breast cancer leads to EGFR activation by altered trafficking pathways [50]. Therefore, even though EGFR amplification is not common in gliosarcoma, EGFR pathway activation may still be present. Gain as oppose to amplification of EGFR may be sufficient for EGFR pathway activation in of itself. In contrast, some believe copy number gain (as oppose to amplification) of EGFR reflects chromosomal instability in cancer cells and has no biological significance [51]. Thus, it is unclear whether EGFR pathway activation is present in gliosarcoma; perhaps it is present through indirect mechanisms and not necessarily through overexpression of EGFR due to gene amplification. Our study confirms that EGFR amplification is uncommon in gliosarcoma and provides whole-genome evidence of possible driver pathways from a DNA copy number perspective. Further analysis to pinpoint specific loci within the altered copy number regions is needed.

The strengths of this study are that it comprises the largest cohort of gliosarcoma cases with DNA copy number analysis by OncoScan™. However, there were a few limitations. We intended this to represent an initial overview of copy number changes in a large cohort of gliosarcoma cases, and did not include OncoScan™ data from GBM. In the future, it would be ideal to compare our gliosarcoma cohort to regional-matched GBM cases. Also, no general effort was made to separate sarcomatous from glioblastoma components for OncoScan™ analysis. An interesting next step would be to separate the glioblastoma from sarcoma components microscopically, and re-analyze the separate components.

Gliosarcoma, much like GBM, is a fatal diagnosis with no cure. We present a comprehensive whole-genome copy number analysis of gliosarcoma performed in an effort to identify chromosome regions that may represent plausible drivers for gliosarcoma genesis. This study may provide a starting point to direct more targeted oncologic analysis and discover genetic alterations that lead to pathways with therapeutic potential.