Abstract
Glioblastoma is the most common primary cancer of the brain in adults, and its characteristic heterogeneity and phenotypic plasticity contribute to its universal lethality. The cytokine-rich and hypoxic tumor microenvironment (TME) contributes to the abysmal prognosis of this disease by altering cell signaling programs driving phenotypic transitions that modulate cancer cell responses to chemotherapy. Here, we aimed to characterize the roles of the TME in driving glioblastoma-specific phenotypic shifts and the acquisition of chemotherapeutic resistance, and to identify novel pharmacologic and gene therapy combination approaches to treatment. Analysis of publicly available human tumor data demonstrated correlations between proneural-mesenchymal transition (PMT), a phenotypic shift that results in increased glioblastoma aggressiveness and therapeutic resistance, and the common TME factors TGFβ, TNFα, and hypoxia. Accordingly, we examined the roles of these factors, as well as the front-line chemotherapeutic temozolomide, for their abilities to promote PMT in vitro. In glioma-initiating cells, TGFβ, TNFα, and temozolomide promoted expression of a circumscribed set of commonly studied mesenchymal transcripts but failed to promote a clear PMT when considering all markers used to define the proneural and mesenchymal molecular subtypes. Thus, the PMT that can occur in glioblastoma tumors is unlikely to result from the acute response to a limited number of purported TME drivers. Nonetheless, the TME features we studied did produce substantial changes in glioblastoma cell death response to temozolomide and the second-line chemotherapeutic carboplatin, with magnitudes and qualitative effects that varied with TME condition and DNA-damaging agent. To determine if a conserved set of signaling pathways explains the phenotypic effects and to identify potential drug targets for promoting chemoresponse, we generated a partial least squares regression (PLSR) model using more than 4000 measurements of signaling nodes and paired cell death measurements from glioblastoma cells pre-treated with TGFβ, TNFα, or hypoxia prior to exposure to temozolomide or carboplatin. The PLSR model surprisingly nominated the tumor suppressor PTEN and the apoptosis-related protein FADD as key drivers of chemoresistance, with some TME context dependence. These predictions were experimentally validated, and the downstream effectors responsible for the roles of PTEN and FADD were identified using small molecule inhibitors or RNA interference. As an alternative approach to glioblastoma combination therapy, we engineered cancer suicide gene products stabilized through the stress-activated protein phosphorylation events that accompany chemotherapy-induced DNA damage. For the four stress-stabilized suicide gene designs, the specificity of regulation by the p38 or JNK kinases was tested. The collective results of this work reveal limitations in our understanding of the paths glioblastoma cells traverse to undergo PMT but identify new opportunities for combination therapy approaches that leverage an understanding of TME regulation of signaling and novel designs of cancer gene therapies that may cooperate with DNA damage signaling.
