50 Tumor–Myeloid Multicellular Programs Reveal ECM-Associated Macrophage Polarization in ccRCC

Abstract

Abstract Background Clear cell renal cell carcinoma (ccRCC) is characterized by a highly heterogeneous tumor immune microenvironment. However, how tumor-intrinsic programs shape immune cell states remains incompletely understood. We aimed to identify multicellular programs (MCPs) linking tumor and immune compartments and to define tumor-derived signals that modulate immune phenotypes. Methods We applied DIALOGUE to identify MCPs across tumor and non-tumor compartments, including CD8+ T cells, CD4+ T cells, regulatory T cells, myeloid cells, endothelial cells, and fibroblasts. A total of 189,999 cells from 26 patients with sufficient tumor cell representation (>100 cells) were analyzed from our previously reported cohort (Kashima, ASCO 2024). MCP scores were derived from DIALOGUE and summarized at the sample level by averaging single-cell MCP scores. Cell–cell interactions were evaluated using CellChat and NicheNet. Spatial relationships were assessed using G4X spatial transcriptomic data (Yochum, AACR Kidney 2026). Spatial niches were defined using Seurat v5 based on k-nearest neighbor composition, and annotated as stromal, sarcomatoid, or clear cell according to cell-type composition and gene expression features. Results We identified two major MCPs, including tumor–myeloid (MCP1) and tumor–CD4+ T cell (MCP2) programs. The tumor component of MCP1 was enriched for mesenchymal and extracellular matrix (ECM)-related genes, including structural components and remodeling factors (COL5A2, MMP2, MRC2), as well as integrin-associated adhesion and cytoskeletal genes (ITGA3, ITGB1, VCL). Expression of HMGA2 further supported a dedifferentiated, EMT-like state. Clinically, a high MCP1 score was associated with sarcomatoid features (p = 0.04). In the myeloid compartment, MCP1 was characterized by immunoregulatory and tissue-remodeling markers (MRC1, MARCO, VSIG4, MS4A4A), chemokines (CCL13, CCL18, CCL23), and complement-related genes (C2, CFH, CR1), consistent with an M2-like macrophage phenotype. Cell–cell interaction analysis revealed increased communication between MCP1-high tumor cells and MARCO+ and CXCL9+ macrophage populations. Mechanistically, ECM-related signaling pathways, particularly collagen and laminin interactions, were predicted to activate CD44-mediated signaling in macrophages. These findings suggest an ECM–CD44 axis linking tumor mesenchymal programs to M2-like polarization. Spatial transcriptomic analysis further demonstrated preferential accumulation of macrophages in stromal-rich regions. Quantitative comparison across stromal, sarcomatoid, and clear cell regions showed increased macrophage infiltration in stromal and sarcomatoid areas compared to clear cell regions (3900 and 1920 vs 830 cells/mm², respectively). Differential expression analysis of macrophages across these regions showed that macrophages in sarcomatoid areas upregulated immunoregulatory and M2-associated markers, including CD163, MRC1, CSF1R, and TGFB1. CD44 expression was also significantly increased, supporting activation of ECM–CD44-associated signaling. Response to immune checkpoint inhibitors (ICIs) was evaluable in 13 patients. Although MCPs were not significantly associated with response, patients with higher myeloid MCP1 score (> mean), but not tumoral MCP1 score, showed a numerically higher response rate compared to those with low MCP1 score (4/4 vs 5/9 patients). Of note, all 4 responders in the myeloid MCP1-high group received combination therapy with CTLA-4 and PD-1 blockade. This suggests myeloid MCP1 may be associated with response to combination ICI therapy. Conclusions Our study identifies a tumor–immune MCP in which mesenchymal and ECM remodeling programs in tumor cells are associated with M2-like macrophage polarization. These findings highlight an ECM–CD44-associated signaling axis as a potential mechanism by which tumor cells shape an immunosuppressive microenvironment in ccRCC.