Abstract LB122: Dynamic interactions of spatially organized tumor, stromal and immune cellular niches predict outcome to immune checkpoint blockade (ICB) in recurrent or metastatic head and neck squamous cell carcinoma (RM-HNSCC)

Abstract Background: Spatial transcriptomics has revealed conserved tumor core (TC) and leading edge (LE) architectures in primary oral squamous cell carcinoma (Arora and Bose et al. Nat Commun 2023). Whether spatial organization of malignant states and tumor microenvironment (TME) interactions predict immunotherapy response in RM-HNSCC is unknown. Methods: 28 tumor biopsies (19 pre-treatment, 7 on-treatment, 2 post-treatment) from 20 ICB-naive RM-HNSCC patients treated with pembrolizumab/nivolumab in 2 investigator-initiated studies, INSPIRE (NCT02644369) and IOKIN (NCT04606940), were profiled using 10x Visium. Malignant regions were defined by reference-based deconvolution, pathological review, and TC/LE gene signatures. Malignant subclusters were functionally annotated using Ingenuity Pathway Analysis. Copy number variation (CNV) analysis used inferCNV. Cancer-associated fibroblasts (CAFs) and tumor-associated macrophages (TAMs) subsets were characterized. Cell-cell interactions were inferred using CellChat. Results: We analyzed 9, 819 malignant and 16, 269 non-malignant spots. Beyond conserved TC and LE states, distinct cycling (upregulating cell cycle checkpoints and DNA synthesis pathways) and neutrophil-inflamed (upregulating neutrophil degranulation and NETosis pathways) malignant compartments were identified. Despite spatial segregation and transcriptional divergence, all malignant niches shared a common CNV backbone (3q, 8q, 11q13 gains; 3p loss), suggesting TME-driven phenotypic plasticity underlies intratumoral heterogeneity. Cycling malignant compartments functioned as dominant signaling hubs, with strongest CAF engagement (IFN-I, WNT) and acting as major sources of immune checkpoint ligands (PD-L1/TGFβ/PVR). Responders showed myCAF depletion post-ICB, identifying CAF remodeling as a key response gatekeeper. Unsupervised clustering revealed immune-permissive (TAMs adjacent to cycling and neutrophil-inflamed regions) versus immune-constrained (CAFs near LE) states, with differential PFS (HR=0. 31; 95% CI 0. 09-1. 01; p=0. 05). Spatial TME features retained predictive value when applied to bulk transcriptomics, yielding an immune-inflamed signature validated in 1, 936 ICB-treated patients (26 cohorts) across 7 cancer types, predicting improved PFS (HR=0. 79; 95% CI 0. 63-0. 98; p=0. 03) and response (OR=0. 32; 95% CI 0. 12-0. 87; p=0. 03). Conclusions: Spatially organized malignant plasticity and tumor-stroma-immune interactions, rather than tumor genetics alone, govern immunotherapy sensitivity in RM-HNSCC and across cancer types, establishing spatial TME architecture as a clinically relevant determinant of ICB response. Citation Format: Grégoire Marret, Jinsu An, Farnoosh Abbas Aghababazadeh, Kewei Ni, Lucas Penny, Martin Hyrcza, Ben X. Wang, Enrique Sanz-Garcia, Anna Spreafico, Giselle Boukhaled, Simone Stone, Aaron R. Hansen, Troy Ketela, Benjamin Haibe-Kains, Scott V. Bratman, Pinaki Bose, Lillian L. Siu. Dynamic interactions of spatially organized tumor, stromal and immune cellular niches predict outcome to immune checkpoint blockade (ICB) in recurrent or metastatic head and neck squamous cell carcinoma (RM-HNSCC) abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts) ; 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86 (8Suppl): Abstract nr LB122.