Abstract 160: A high-throughput in vivo platform (Perturb-map) accurately models patient immunophenotypes and identifies therapeutic targets

Abstract Targeting the right patient population with any given therapeutic agent largely remains an unsolved problem, and successes in pre-clinical models often do not translate into positive outcomes for patients. Recent advances in multi-omics spatial profiling have highlighted the diversity of tumor microenvironment compositions and architectures across patients. Our deepened understanding of human disease also further demonstrates that the handful of pre-clinical models used for a large proportion of pre-clinical studies seldom recapitulate human biology and are unreliable predictors of clinical success. To bridge the gap between pre-clinical models of disease and human biology, we developed a high throughput in vivo platform that can reproduce in mice the diversity of tumor phenotypes seen in patients. Our platform leverages Perturb-map, a spatial functional genomics technology compatible with high content multimodal phenotyping of 100s of tumor clones (and their associated genetic perturbations) in the same animal. A key feature of Perturb-map is that genetic perturbations can be identified by reading a protein barcode using multiplex imaging, a scalable read out. Our current custom imaging pipeline can analyze the impact of 100+ genetic perturbations - and the equivalent of 1, 000s of individual tumors - in a few weeks, while simultaneously profiling the tumor microenvironment. To date, we have used this platform to generate a dataset encompassing 600+ genetic perturbations (CRISPR KO) in a mouse model of NSCLC (KRASG12D Tp53-/-). For each perturbation, >50 tumors are profiled at baseline, and in the context of immune checkpoint blockade. We have found that, unlike mouse models traditionally used for discovery, our in vivo dataset reproduces a broad spectrum of patient immunotypes, including T cell deserts, highly infiltrated tumors, and more specific profiles of immune infiltration. Moreover, we have identified genetic perturbations that drive the establishment of these immunotypes. For example, loss of STK11 in the mouse tumors lead to a local increase in neutrophil infiltration, a feature that is also associated with STK11 mutations in the patient population. Among the 600+ genetic perturbations, we also identify specific genes that promote local inflammation or tumor immunoediting, and therefore constitute potential therapeutic targets. Altogether, we have demonstrated that our ability to scale up in vivo experimentation using Perturb-map addresses current limitations in drug development, as we can (i) generate at scale models that better reflect the various facets of human tumor biology and (ii) perform target evaluation at a high throughput without compromising phenotypic readouts. Citation Format: Maxime Dhainaut, Lucas Cavalcante, Hargita Kaplan, Carl Ebeling, Dexter Antonio, Rodney Collins, Francis Fernandez, Eric Siefkas, Anastasia Mavropoulos, Joy Tea, Nicole Snell, Tyler Van Hensbergen, Daniel Bear, Michela Meister, Eshed Margalit, Jake Schmidt, Yubin Xie, Yu (Phoebe) Guo, Jeremie Decalf, Rob Schiemann, Shafique Virani, Jacob Rinaldi, Lacey Padrón, Ron W. Alfa. A high-throughput in vivo platform (Perturb-map) accurately models patient immunophenotypes and identifies therapeutic targets [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 160.