The ability to predict dependencies given the molecular features of a patient’s tumor is central to cancer precision medicine. The systematic use of CRISPR/Cas9 and pharmacologic tools in established cancer models is showing great potential to discover new targets. However, existing model development approaches require long periods of culture time during which evolutionary pressures reduce heterogeneity. And, it remains challenging to create long-term models for certain tumor types and genotypes, making it challenging to use perturbational tools to experimentally map dependencies. To address these challenges, our overarching goal is to develop ‘rapid ex vivo tumor biosensors’ whereby we would be able to interrogate cancer dependencies in an immediate short-term ‘culture’ of cancer cells taken from a patient biopsy/surgery/fluid collection as a novel research-grade experimental model of cancer. In doing so, we aim to couple the timing of drug or CRISPR/Cas9 perturbation with the preservation of subcellular heterogeneity. If successful, we hypothesize that this modelling approach will more accurately recapitulate patient tumors and may ultimately serve as a stronger foundation for preclinical therapeutic studies. This work should also substantially expand the fraction of patient samples that can be interrogated. Here, we propose using gastroesophageal adenocarcinoma (GEA) as a test case for this strategy due to our experience as well as the existence of marked intra-tumor heterogeneity. However, once established, this novel modeling platform should enable a wide range of basic and translational questions (both for GEA and other tumors) that require model formats that include heterogeneous cell populations. Our goal will be achieved via two Specific Aims including (1) using patient-derived organoids created on rapid time frames for CRISPR/Cas9 editing to validate emerging GEA dependencies; and (2) developing the ability to directly visualize and perturb single cells from matching patient ascites fluid or disaggregated primary tumors ex vivo using label-free imaging methods. We will benchmark these approaches against each other using the same clinically annotated, serially collected patient samples. In following the instructions for this RFP, we focus on technology-development focused goals as opposed to deeper mechanistic studies. We focus on benchmarking predictions and assessing reproducibility, sensitivity and specificity. This work is innovative, in that it brings together expertise at the intersection of functional genomics, advanced computational approaches for image-analysis and GEA genomics. If successful, this effort could have significant impact by establishing a foundation to expand this approach to other disease (tumor and non-cancer) indications.

The proposed research is relevant to public health since long-term survival rates for patients with tumors of the stomach or esophagus remain stubbornly low despite advances in genomics. This work directly addresses this challenge based on the hypothesis that it is possible to use living patient samples ex vivo on rapid timeframes to refine and validate associations between genotype and therapeutic dependency. Since this approach will enable both basic and translational cancer research it is centrally focused on the NCI’s mission to support research to reduce the morbidity and mortality of difficult-to-treat tumors such as gastroesophageal adenocarcinomas.