Network mechanisms and bioengineering of precision immunotherapies for glioblastomas

Project Summary/Abstract Glioblastoma multiforme (GBM) is the most common and lethal malignancy of the human central nervous system (CNS). This devastating disease has a median overall survival of 3 months from time of diagnosis in untreated patients.

Despite the significant cost and morbidity with standard of care surgical resection combined with adjuvant chemotherapy and radiation, life expectancy is only extended by a few months. Oncolytic viruses (OVs) are a class of immunotherapies with an FDA-approved treatment for solid tumors.

These work by directly lysing tumor cells, which then release tumor antigens with danger signals that elicit antitumor immunity.

Adenovirus is a strong candidate OV immunotherapy for GBM because of its low neurotoxicity when delivered by intratumoral injection and its amenability to bioengineering.

Ad5-?24, an adenovirus modified to replicate in GBM but not in other CNS cells, is perhaps the most promising OV in clinical trials. Major hurdles remain, however, including innate immunoresistance.

This study will overcome these by elucidating complex tumor-OV interaction mechanisms at the systems level, designing and constructing improved OV candidates that exploit these mechanisms, and then validating the mechanisms by experimentally testing the OVs.

Preliminary studies for this proposal demonstrate that GBM patient-derived xenograft (PDX) models have distinct transcriptomes in vivo and can be cultured as neurospheres (Sp-GBMs) under serum-free conditions ex vivo. Ad5-?24 OV can lyse several different Sp-GBMs ex vivo and the level of this oncolytic activity is PDX specific.

Additionally, preliminary in silico models suggest that such OV activity levels are contextually dependent on protein-protein interaction networks in the target GBM cells.

This project will test the central hypothesis that dynamic transcriptional states of human GBM play key mechanistic roles in Ad5-?24 oncolytic efficacy; therefore, modeling the emergent system behaviors will guide the engineering of precision OV immunotherapies.

The first aim will classify human GBM PDXs using transcriptomes to predict Ad5-?24 OV responses and identify context-specific network features.

The resulting classifier and network models will be validated by assessing their ability to predict the response of Sp-GBM to Ad5-?24 OV treatment. The second aim will elucidate Ad5- ?24 OV gene dependencies and resistance mechanisms in human GBM PDXs. Predictions from both aims will both experimentally and computationally validated.

Synthetic Ad5-?24 OVs with shRNAs that perturb these gene dependencies and resistance mechanisms will be constructed and tested for efficacy.

These precisely targeted alterations result in the quantifiable, predictable phenotypic effects of modulating human GBM gene expression and oncolysis ex vivo and in vivo. Such synthetic OVs will be useful as GBM research tools.

As a future direction, the big data generated will be a valuable public resource to further interrogate antitumor immunity and inflammation states.

Successful completion of these aims will yield novel GBM diagnostic subtypes with paired engineered OVs ready for preclinical development as precision immunotherapies.