The last few years have seen development of multiple methodologies to dissect the tumor microenvironment into functionally distinct subpopulations. In contrast, the malignant compartment continues to be represented as a mutation and patient-specific continuum of states, representing an aberrant version of human developmental lineages.
Recent studies by our lab have shown that tumor cell states also fall into highly specific aberrant development-like states rather than a continuum, with critical similarity to normal lineage development but also important differences. The most critical similarity is that malignant cells can only occupy a small set of stable, discrete states that coexist — with virtually identical Master Regulator dependencies — across all tumors of a specific type; we have dubbed this phenomenon “quantum cancer biology.” For instance, we have shown that diffuse midline glioma (DMG), pancreatic cancer (PDA), and cholangiocarcinoma (CCA) comprise seven, six, and three malignant states, respectively, and have observed virtually identical state structures in sarcomas, breast cancer, and prostate cancer. Critically, the stable states identified by our network-based analyses coexist with virtually identical non-oncogene dependencies in all patient cohorts we could analyze at the single cell level. Finally, barcode-based lineage coding has shown that the distinct cell states that coexist in a tumor can spontaneously, albeit with kinetics that are sample specific, reprogram into each other, thus accounting for the overall continuum of cancer cell states which has emerged from bulk profile studies. Indeed, all individual cells not found in a stable state were undergoing plastic transition between two of the stable states. Taken together, these findings suggest that appropriate combination therapies could be designed to target each coexisting cell state, either simultaneously or sequentially.
Following up on this hypothesis, we have shown that eight of nine drugs predicted by OncoTarget/OncoTreat to target each of the seven DMG states were able to selectively target their predicted states, and only three of them were necessary to abrogate all seven states. Indeed, virtually all but one of the two drug combinations targeting complementary states produced striking synergy compared to the associated monotherapies and we are currently developing a multi-center DMG clinical trial to leverage all three drugs. The same approach has been highly effective in targeting cell states that help recruit an immunosuppressive TME or even in targeting selectively immunosuppressive stromal subpopulations, producing striking rescue of immune checkpoint inhibitor activity in otherwise cold tumors. With Biohub, we are now generating millions of perturb-seq profiles following CRISPR perturbation of TME subpopulations and single cell drug perturbation profiles, with ~1000 drugs, of PBMC from multiple donors. Taken together, this suggests that combination therapies targeting the heterogeneity of both the TME and the malignant compartments.