Deciphering the cellular tumor microenvironment landscape in salivary gland carcinomas using multiplexed imaging mass cytometry.

Background: To spatially characterize the single-cell tumor microenvironment (TME) of salivary gland carcinomas (SGC) and identify prognostic biomarkers.

Methods: SGC, including salivary duct carcinomas (SDC), acinic cell, mucoepidermoid, and secretory carcinomas, were analyzed using a 13-marker imaging mass cytometry panel. Multichannel image data from 54 primary cases and nodal metastases were processed to generate single-cell datasets. Cell phenotypes (tumor cells, cancer-associated fibroblasts (CAFs), endothelia, immune cells) were classified using a validated CAF algorithm, followed by spatial analysis and clinicopathological correlation. Clinicopathological results were validated using a previously published independent bulk RNA-sequencing (RNA-seq) cohort of n = 67 SDC cases. Spatial transcriptomics data of three SDC cases was leveraged to understand the molecular mechanisms of spatial interactions.

Results: Among 509,364 cells, SDC exhibited the highest fractions of Collagen- and matrix-CAFs (mCAFs). Acinic cell carcinomas (ACC) showed enriched CD4⁺/CD8⁺ T cells and antigen-presenting CAFs (apCAFs), indicating strong immune infiltration. A spatially defined cellular neighborhood (CN8) of mCAFs and endothelia was elevated in SDC, with higher CAF infiltration in androgen receptor (AR)_high versus AR_low SDC. Elevated mCAF frequency and CN8 were significantly associated with reduced recurrence-free probability (RFP) and distant control rates (DCR) in SDC. Additionally, higher mCAF frequencies were an independent prognostic factor for decreased RFP and DCR in Cox regression analysis. The association between mCAFs and a decreased RFP as well as recurrence-free survival was confirmed with previously published RNA-seq data using an mCAF signature. Spatial transcriptomics confirmed enrichment of metastasis-associated gene signatures as well as platelet-derived growth factor and insulin-like growth factor pathways within mCAF^highendothelia^high tumor niches indicating that the latter may be involved in mCAF-endothelia crosstalk.

Conclusion: SDC are characterized by Collagen-/mCAF-rich microenvironments and mCAF-endothelial spatial interactions that are linked to metastasis. ACC display pronounced immune infiltration, suggesting its potential for immunotherapy. mCAFs in SDC emerge as prognostic biomarkers and therapeutic targets, highlighting the importance of targeting CAF-driven metastasis in future treatments. This study provides insights into the biology of SGC and identifies novel prognostic markers.