Tumor-Selective Gene Circuits Enable Highly Specific Localized Cancer Immunotherapy

The treatment of certain cancers has been revolutionized by immunotherapy, but not all patients benefit due to a limited mechanism of action and tumor-mediated immune suppression. More potent and broadly acting immunotherapies have received considerable interest, but their narrow therapeutic window upon systemic administration creates the need for spatiotemporal regulation, which has yet to be realized in a clinical setting. To address these challenges, we developed a cancer-specific gene circuit platform that enables targeted, multifactorial immunomodulation of tumors in a highly specific and localized manner to drive efficacy through the therapeutic window. First, we constructed a library of 6,107 synthetic promoters built from tandem arrays of all human transcription factor DNA binding sites. We used next generation sequencing and machine learning to screen the library for promoters that expressed specifically and robustly in a particular cancer cell type. When applied to ovarian cancer, breast cancer, and glioblastoma, this approach yielded promoters with activities >100-fold higher in each respective cancer cell type than in corresponding non-cancerous primary cells. Second, we further enhanced expression specificity by using these promoters to build AND logic-gated gene circuits in which expression of therapeutic payloads requires two independent cues - the activity of two distinct transcription factors, or, alternatively, a transcription factor and a microRNA. When we armed an ovarian cancer-specific gene circuit with multiple immunomodulatory payloads (a T cell engager that enables antigen-independent T cell responses, the cytokine IL-12, the chemokine CCL21, and an anti-PD1 checkpoint inhibitor antibody) and delivered it to cells, the gene circuit triggered ovarian-cancer-specific immune responses both in vitro and in vivo. This approach resulted in statistically significant reduction in in vivo tumor burden (>6-fold reduction in tumor burden at day 39 post tumor implantation; p < 0.005) and prolonged mouse survival in a humanized ovarian cancer model (all control mice died by day 55 post tumor implantation, whereas 80% of mice survived to the end of the study in the treatment group; p < 0.005). These results suggest that these AND logic-gated gene circuits can be readily reconfigured for different cancer types through modular replacement of the promoters and/or miRNA-binding sites. Thus, this technology has the potential to enable highly targeted and effective treatment of cancer via precise immunological programming.