Abstract 7156: Spatial single-cell transcriptomics identifies tumor niches predictive of response to TKI therapy in non-small cell lung cancer

Abstract Non-small cell lung cancer (NSCLC) is the tumor with the highest mortality in the world. Tyrosine kinase inhibitors (TKI) are commonly used treatment options for NSCLC, albeit with significantly different mechanisms of action. Responses to TKIs vary greatly among patients. Biomarkers for determination of response in patients treated with these therapies are needed to improve outcomes for patients and allow for more personalized therapeutic approaches. In this study, we retrospectively collected 34 patients with stage III or IV EGFR-mutated NSCLC. These patients were treated with TKIs. Responses to TKI therapies were dichotomized to early progression (<12 month) and late progression (>12 month) based on their progression free survival data. We performed 10X Xenium spatial single-cell-resolution transcriptomics and imaging mass cytometry (IMC) on tissue microarrays (TMAs) generated from these patients, representing a total of 91 cores. In this cohort, we detected an enrichment in low tumor infiltrating lymphocyte (TIL) numbers with high epithelial proportions in early progressors using niche detection methods and neighborhood analysis algorithms. The proportion of tumor cells in these immune cold regions is significantly higher in early-progression patients than the late-progression patients. Amphiregulin, encoded by AREG, was significantly upregulated in immune cold tumor regions associated with TKI resistance. AREG has been demonstrated to inhibit TKI-induced apoptosis with multiple mechanisms. Additionally, we found increased myofibroblast-like cancer-associate fibroblasts (Myo-CAFs) surrounding the tumor region in the early-progressor group. We next unbiasedly extracted >144, 000 features, including gene expression data, cell types, spatial correlations, and gene correlations. We then used Stabl, which is a feature selection algorithm for isolating a sparse and reliable set of features in predictive models when the number of features is significantly greater than the number of patient samples. Using this platform, we identified a 6-feature program which had an AUC of 0.973 in predicting early vs. late progression in TKI-treated EGFR-mutant NSCLC patients. One of the identified features was the expression level of ARL14 in Myo-CAFs. We further found these ARL14+ Myo-CAFs with higher tendency to form spatial niches with tumor cells than other CAFs. The AREG+ immune cold tumor regions and their surrounding ARL14+ Myo-CAFs commonly form the spatial niches determining TKI responses. We identified novel spatial features leading to the TKI resistance in NSCLC using the highly multiplexed nature of spatial single-cell transcriptomics. These spatial biomarkers have a high predictive power in predicting response to TKI treatment and provide a basis for the precise treatment of lung cancer and provide potential targets to overcome treatment resistance. Citation Format: Haochen Li, Joseph Lownik, Simeon Mahov, Matthew Gayhart, Jean Lopategui, Karen Reckamp, Akil Merchant. Spatial single-cell transcriptomics identifies tumor niches predictive of response to TKI therapy in non-small cell lung cancer [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 7156.