Spatially resolved multi-region proteomics for prediction of clinical outcome in deficient mismatch repair metastatic colorectal cancer treated with PD-1 blockade.

3520 Background: In patients with metastatic colorectal cancer (CRC) and deficient DNA mismatch repair (dMMR), we previously found transcriptomic signatures reflecting stromal compartment composition and proliferative capacity that can predict the efficacy of immune checkpoint blockade (ICB). To identify proteomic markers of response and/or resistance to ICB, we performed spatially resolved proteomic analysis of the tumor microenvironment. Methods: Patients with surgically resected, primary dMMR metastatic CRCs received pembrolizumab (Mayo Clinic, N=30). In FFPE tissues, Digital Spatial Profiling of 71 proteins was performed using panels covering immune cells, immune activation, cell death, PI3/AKT signaling, MAPK signaling, and pan-tumor markers (GeoMx nCounter). In ten regions of interest/slide (~600 µm 2 ) at the tumor invasive margin, protein expression was determined in four cellular compartments: epithelium (panCK + ), immune cells (CD45 + ; leucocyte common antigen), stroma (panCK - ), and nonimmune stroma (panCK - /CD45 - ). After normalization, protein abundance per compartment was analyzed by response (RECIST 1.1) OmicVerse Python. Differentially expressed proteins (binary) were examined by progression-free survival (PFS) using multivariable Cox regression with covariate adjustment (Lifelines Python). Results: Pre-treatment panels including 71 proteins were analyzed by treatment response and PFS after PD-1 inhibition. Proteins found to be significantly associated with response (p< 0.01) and PFS are described. Across all compartments, responders showed upregulation of NK cells and T cell subsets (CD3 + , CD4 + , CD8 + ), and ICOS; strongest association was found for CD4 + : HR adj : 0.16; 95%CI: 0.08-0.31; p< 0.001. Nonresponders showed higher phospho-JNK (HR adj : 2.09; 95%CI: 1.28-3.42; p=0.003) and SMA. In the epithelial compartment, upregulation of the Treg marker CD25 and reduced caspase-9 were observed in responders. Among responders, the stromal compartment was highly enriched in HLA-DR, dendritic cells (CD11c), PD-L1, immune activation markers (CD40, CD44, CD80, CD27), CD95/Fas, and GZMB. In the stromal compartment, downregulation of BCL6 and phospho-GSK3B was observed in responders. The immune compartment of responders showed increased beta2-microglobulin and Ki-67, while the non-immune stromal compartment showed increase in the macrophage marker CD68. Importantly, proteins found to be upregulated in responders were significantly and independently associated with longer PFS whereas the opposite was true for nonresponders. Conclusions: In metastatic dMMR CRCs, spatially defined protein markers were significantly associated with anti-PD-1 response, and independently associated with patient PFS (strongest predictive utility for CD4 in immune stroma and phospho-JNK in epithelium).