Tumor heterogeneity is a defining feature of cancer, yet it remains difficult to model accurately in mice. To address this, we combined in vivo CRISPR screening with Perturb-Map—a powerful approach that links tumor cells to their specific genotypes (i.e., CRISPR-induced mutations) at single-cell resolution, enabling spatially resolved functional genomics. Using this strategy to model ovarian cancer, we uncovered that tumors form distinct clonal immune microenvironments, which critically influence clonal fitness through immune surveillance and response to immunotherapy.

Remarkably, we found that cancer cells actively shape these local immune landscapes by secreting cytokines—immune-modulatory signaling molecules—to promote their survival. Among these, IL-4 emerged as a key regulator of response to PD-1 immune checkpoint blockade. IL-4 reprograms the tumor microenvironment primarily through cancer cell–macrophage interactions. Notably, blocking IL-4 signaling significantly enhanced immunotherapy efficacy in mouse models.

These findings reveal that localized communication between cancer and immune cells within tumor clones drives functional heterogeneity and underlies resistance to treatment. Targeting these cancer-immune cell signaling networks offers a promising strategy to overcome therapeutic resistance and improve patient outcomes.

Lung cancer remains the leading cause of cancer-related mortality worldwide. Non-small-cell lung cancer (NSCLC), the most prevalent type of lung cancer, comprises two primary subtypes: adenocarcinoma and squamous cell carcinoma. These subtypes are characterized by distinct driver mutations and immune microenvironments. While it has long been recognized that both adenocarcinoma and squamous tumors can coexist within a single patient’s biopsy, emerging evidence suggests that targeted therapies (e.g., EGFR and KRAS inhibitors) can select resistant squamous clones in patients. Through the development of multiple genetically engineered mouse models (GEMMs) for NSCLC, we uncovered a competitive relationship between SOX2 and NKX2-1, two transcription factors essential for lineage commitment during lung development. We demonstrated that SOX2 and NKX2-1 drive squamous and adenocarcinoma lineages, respectively. Genetic modulation of these factors led to heterogeneity in lung tumors, including the coexistence of adenocarcinoma and squamous components, as well as adenocarcinoma-to-squamous transdifferentiation. In addition to elucidating lineage plasticity, we discovered that SOX2 and NKX2-1 inversely regulate neutrophil recruitment to tumors. Furthermore, neutrophils were found to promote adenocarcinoma-to-squamous transdifferentiation. This work established a unique GEMM for studying intratumoral heterogeneity (ITH) in NSCLC, identified the critical roles of lineage-specifying transcription factors in lung cancer heterogeneity, and provided the first evidence of neutrophils influencing clonal evolution in lung cancer.

Small-cell lung cancer (SCLC) is one of the most aggressive tumor types, with an average survival rate of only 10 months. Intratumoral heterogeneity (ITH) drives the rapid development of resistance to treatments, including chemotherapy. We discovered that MYC, a frequently altered oncogene in SCLC, promotes the development of aggressive and highly metastatic tumors. Notably, MYC-driven SCLC initially presents as neuroendocrine "classic SCLC" but undergoes transdifferentiation into non-neuroendocrine "variant SCLC." The molecular subtypes of SCLC are governed by a network of transcription factors, including ASCL1 and NEUROD1. These subtypes co-exist as distinct clones within tumors, exhibiting not only differing histological and molecular characteristics but also unique pharmacological vulnerabilities. This study established the first genetically engineered mouse model for non-neuroendocrine SCLC, uncovered a genetic mechanism regulating SCLC heterogeneity, and laid a critical foundation for advancing our understanding of SCLC biology.