Spatial architecture of head and neck squamous cell carcinoma (SCCHN): An IMMUcan/EORTC analysis

Background : The association of spatial architecture of SCCHN with clinical and molecular characteristics is poorly understood. Methods : 253 recurrent/metastatic (R/M) SCCHN, including 77 pre-immunotherapy (IO) samples, were analyzed by imaging mass cytometry to study cancer cells and 13 immune cell types (PMID 40930089) . Global cell densities (cellD, n/mm 2 ) and cellular neighborhoods (CN) were analyzed. Local cellD (degree of immune infiltration/exclusion within cancer cells) and cell interaction coefficients between pair-wise phenotypes, both independent of global cellD, were computed. Weighted correlation network analysis was used to detect modules of cell interaction coefficients. TLS and cancer cells states [hypoxic, cell cycling, EMT, beta2-microglobulin (beta2m) loss] were explored. Data were correlated with IO primary vs. secondary resistance (R), HPV status, survival, DNA and RNA profiles. Results : 8 CN were identified (pure cancer cells, internal and external (cancer/stroma) interfaces, aggregation of plasma cells, neutrophils (PMN), macrophages/mural cells, T-cells and macrophages/mural/T-cells). HPV+ tumors had decreased external interface. Refractory HNSCC had decreased interfaces and lower DC infiltration. Cancer cells were increasingly hypoxic or had decreasing beta2m expression and decreasing EMT in external interface vs. internal interface vs. pure cancer cells CNs. Overall, cancer cell states had profound impact by either increasing (cycling cancer cells) or decreasing (hypoxic cancer cells) immune cells infiltration. We identified 3 clusters of SCCHN: C1 (45%) had high global immune cellD but low local cellD; C2 (40%) was desertic; C3 (15%) was highly infiltrated (high local cellD) and had the best survival in IO-treated pts. RNA-based ‘Classical' SCCHN were almost exclusively C2. C3 was enriched in extracellular matrix organization, immune cell and migration. Specific TLS CN included B compartment, germinal center and plasma aggregation. HPV+ tumors had increased B compartment but decreased plasma aggregation. All 3 TLS CN were decreased in refractory SCCHN and in primary vs. secondary R. In IO-treated pts, TLS presence improved outcome and the closer TLS were to cancer cells, the better was the outcome. While the interaction between PMN and immune cells was associated with worse outcome, PMN infiltrating cancer cells was associated with improved survival together with T-cells interacting with cancer cells or with HLADR, MacCD163 and mural cells. Finally, mutations were generally associated with more exclusion of immune to cancer cells e.g. T-CD8 and PMN were more excluded in SCCHN mutated for RB1 and FAT4-PTPRT respectively; and BAP1 and HRAS mutations had exclusion of MacCD163-NK-mural-HLADR-plasma and B cells respectively. Conclusion: we provide a comprehensive description and clinical/molecular correlatives of the spatial architecture of SCCHN.