Abstract Solid tumors are complex ecosystem with inherent genetic and phenotypic variability, and interactions among cancer cells with the tumor microenvironment that shape the development, progression and the response or resistance to targeted- and immunotherapies. This heterogeneity is poorly captured with profiling of bulk tumor tissues that provides an average of these signals without recapitulating cellular interactions and phenotypes. Single-cell genomics technologies, such as single-cell RNA-sequencing (scRNA-seq), represent powerful tools to resolve tumor composition and dissect interactions within the ecosystem that may determine drug resistance and reveal vulnerabilities that can be therapeutically targeted. Using scRNA-seq we profiled thousands of malignant and non-malignant cells from patients with metastatic melanoma. We found that all tumors, irrespective of prior drug exposures, contained cells expressing a cell state characterized by high expression of AXL and low expression of MITF, the master regulator of the melanocytic lineage. The AXL-high/MITF-low cell state conferred intrinsic resistance to RAF/MEK inhibitor therapy in BRAF-mutated melanoma models, and led to the emergence of a small number of pre-existing AXL-high/MITF-low cells with treatment. Interrogation of tumor-infiltrating lymphocytes (TILs) provided insights into their diversity and expression of various immune checkpoints, such as PD-1, CTLA-4, LAG3, TIGIT and others, indicating their co-expression on CD8 and CD4 cells. While these markers indicated dysfunction of the T cells, we found concurrent expression of activating markers, including key cytokines such as interferon gamma. We dissected the relationship between activation, dysfunction, and clonality in these cells, and identified markers that saliently characterized the dysfunction state across different patients. Several complement factors, including C3, were predicted to modulate the infiltration and exclusion of T cells from the TME, and we validated the relationship of these in independent melanoma patient tissue sections. We performed scRNA-seq on additional melanoma patients to assemble a total of 31 samples, of which 15 were isolated from patients with resistance to immune checkpoint inhibitors (ICI) (anti-PD-1 or anti-CTLA4 therapy or a combination of both). We performed a genome-scale inference of cancer cell-mediated drivers of T cell exclusion, a major mechanisms of ICI resistance. Next, we directly measured cell-intrinsic expression of pathways associated with immune evasion. Strikingly, the same cell state mediated both, T cell exclusion and immune evasion, indicating that one coherently regulated cancer cell program could be responsible for ICI resistance in patients. We validated the spatial impact of this resistance program by multiplexed immunofluorescence of immune infiltrates for predicted drivers of T cell exclusion in matching formalin-fixed, paraffin-embedded (FFPE) patient specimens. The signature was prognostic for survival in the melanoma TCGA cohort, indicating that the program can be intrinsically expressed. To test the predictive value of this signature, we compiled two validation cohorts, including one with 112 melanoma patients undergoing pre-treatment biopsies followed by anti-PD-1 therapy, and 26 additional patients of which some had sequential biopsies. RNA-sequencing of these cohorts revealed that the resistance program was predictive of progression-free survival, discriminated objective responders (OR), including complete response vs. partial response from patients progressive disease, and the duration of OR, supporting the possibility that the program captures both intrinsic and acquired resistance to ICI. Using an in silico prediction across >600 cell lines screened against 131 drugs, we found that cells with high expression of the resistance program were selectively susceptible to inhibition by CDK4/6 inhibitors, such as palbociclib or abemaciclib. We experimentally validated that abemaciclib reversed the resistance program in vitro, enhanced T cell responses in an ex vivo model of patient-derived melanoma cells and autologous T cells, and when combined with ICI, induced a high rate of complete responses in an otherwise ICI resistant syngeneic in vivo model of melanoma. Along with other studies in the field, our study suggest a potentially clinically active synergistic effect of combining CDK4/6 inhibitors with ICI. Together, these studies highlight the possibility of developing rationale therapeutic strategies to overcome drug resistance by systematic assessment of patient tumor ecosystems. Citation Format: Benjamin Izar, Livnat Jerby-Arnon, Itay Tirosh, Parin Shah, Michael C. Cuoco, Christopher Rodman, Mei-Ju Su, Johannes C. Melms, Rachel Leeson, Abhay Kanodia, Jia-Ren Lin, Genevieve M. Boland, Gao Zhang, F. Stephen Hodi, Peter K. Sorger, Kai W. Wucherpfennig, Eli M. Van Allen, Dirk Schadendorf, Orit Rozenblatt-Rosen, Levi A. Garraway, Asaf Rotem, Charles H. Yoon, Aviv Regev. Development of therapeutic strategies by resolving the tumor ecosystem [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr SY45-04.