Despite the current standard of multimodal management, glioblastoma (GBM) inevitably recurs and effective therapy is not available for recurrent disease. A subset of tumor cells with stem-like properties, termed GBM stem-like cells (GSCs), are considered to play a role in tumor relapse. Although oncolytic herpes simplex virus (oHSV) is a promising therapeutic for GBM, its efficacy against recurrent GBM is incompletely characterized. Transforming growth factor beta (TGF-β) plays vital roles in maintaining GSC stemness and GBM pathogenesis. We hypothesized that oHSV and TGF-β inhibitors would synergistically exert anti-tumor effects for recurrent GBM. Here we established a panel of patient-derived recurrent tumor models from GBMs that relapsed after post-surgical radiation and chemotherapy, based on GSC-enriched tumor sphere cultures. These GSCs are resistant to the standard-of-care temozolomide but susceptible to oHSVs G47Δ and MG18L. Inhibition of TGF-β receptor kinase with selective targeted small molecules reduced clonogenic sphere formation in all tested recurrent GSCs. The combination of oHSV and TGF-βR inhibitor was synergistic in killing recurrent GSCs through, in part, an inhibitor-induced JNK-MAPK blockade and increase in oHSV replication. In vivo, systemic treatment with TGF-βR inhibitor greatly enhanced the anti-tumor effects of single intratumoral oHSV injections, resulting in cures in 60% of mice bearing orthotopic recurrent GBM. These results reveal a novel synergistic interaction of oHSV therapy and TGF-β signaling blockade, and warrant further investigations aimed at clinical translation of this combination strategy for GBM patients.

INTRODUCTION: Recent evidence suggests that glioblastoma is driven by a subset of tumor initiating (TI) cells characterized by their capacity to form tumors in xenograft models and self-renew in vitro. These TI cells share many properties of neural stem/progenitor cells, including the expression of certain cell surface markers. With serial passage, many cells lose their capacity to TI. The transition between TI-proficient and -deficient states remains poorly understood. METHODS AND RESULTS: There are two theoretic models for the maintenance of TI states. In the "elite" model, TI activity is restricted to a predetermined subpopulation of cells. The alternative "stochastic" model suggests that any tumor cell has a finite chance of acquiring TI capacity through random fluctuations in cell physiology. To address this issue, we examined the TI capacity of distinct subclones isolated from a distinct glioblastoma line as well as the TI capacity of single cells derived from each distinct clone. We found that only a subset of subclones from a single glioblastoma line displayed capacity for TI, suggestive of the elite model. However, single cells derived from any single subclone exhibited a wide range of TI capacity, suggesting a stochastic component to this process. Transcriptome profiling of the subclones of differing TI revealed a gene signature associated with TI capacity. Analysis of this signature showed enrichment for genes regulated by c-Myc. Indeed, clones with increased TI capacity tend to harbor increased c-Myc expression. Additionally, over-expression of c-myc increased the TI capacity of glioblastoma cells in xenograft models and led to the formation of intracranial tumor in an Ink4a/ARF null transgenic murine model. CONCLUSION: Our results are most consistent with a threshold model in which TI states in glioblastomas are driven by expression levels of critical factors such as c-Myc.

No abstract