An AAV gene therapy computes over multiple cellular inputs to enable precise targeting of multifocal hepatocellular carcinoma in mice
Bartolomeo Angelici, Linling Shen, Joerg Schreiber, Anthony Abraham, Yaakov Benenson
Abstract
Clinical translation of multi-input biomolecular computing systems holds potential to lead to disease-tailored, data-driven rational design of next-generation therapeutic modalities. However, practical demonstrations of this potential are lacking. Here, we developed a clinically translatable approach for the design and implementation of therapeutic agents comprising biomolecular multi-input logic modules for precision cell targeting, compatible with adeno-associated virus (AAV) vectors. We used this approach to engineer an AAV-encoded gene therapy prototype that, when delivered systemically, successfully treated hepatocellular carcinoma in an orthotopic mouse tumor model. The therapy performed a molecular-scale computation over multiple transcriptional and microRNA inputs based on the differential molecular profiles of tumor and nontumor cells, to guide the activation of a herpes simplex virus thymidine kinase (HSV-TK) effector gene. Multi-input computation in individual cells was necessary and sufficient to drive in vivo and in situ tumor-specific expression of HSV-TK with minimal concomitant expression in nontumor liver and other organs. Intravenous vector injection in combination with ganciclovir resulted in marked reduction in tumor burden in treated mice compared with controls, without negative effects on general well-being or weight. The therapeutic approach has the capacity to perform logical integration of diseased and healthy cell–specific molecular inputs to precisely regulate therapeutic effector gene expression and is a promising avenue for the next generation of cancer therapies. Moreover, our systematic data-driven workflow illustrates how gene expression data can shape the molecular composition of future therapeutic candidates.