Spatial architectures of colorectal cancer microenvironment underlying immune checkpoint inhibitor response

Colorectal cancer (CRC) remains a leading cause of cancer mortality. Immune checkpoint inhibitors (ICI) are among the most effective systemic therapies, yet they benefit only a subset of MSI-H/MMRd patients. To investigate how spatial tumor-immune-stromal organization contributes to heterogeneous treatment responses, we profiled FFPE CRC tissues from PD-1 inhibitor pembrolizumab-treated (n=10) and treatment-naïve (n=14) patients using the CosMx Spatial Molecular Imager (SMI) with a 1,000-gene single-cell panel. After image processing and cell segmentation, cell states were annotated through unsupervised Leiden clustering, gene-module scoring, and supervised InsituType prediction using a public CRC single-cell atlas, followed by spatial analyses including niche identification, neighborhood enrichment, distance-based metrics, and ligand-receptor inference. Across 24 patients (balanced sex distribution; stages I-IV; median age 66.5), ICI-treated tumors exhibited markedly higher frequencies of all CD8⁺ T cell subsets, CXCL8⁺ cancer-associated fibroblasts (CAFs), myofibroblasts, and diverse macrophage and neutrophil populations compared with treatment-naïve tumors, alongside higher proportions of CMS1-like malignant cells. De novo nonnegative matrix factorization (NMF) revealed eight tumor-intrinsic programs, with Inflammatory/MHC-II, Type I IFN/Antigen Presentation, and Innate Inflammatory programs enriched in ICI-exposed tumors, whereas Invasion/Angiogenesis, Proliferation/Stress, and CEA-high programs characterized untreated tumors. Spatial mapping uncovered two recurrent architectures: immune-infiltrated tumors enriched for tertiary lymphoid structures (TLSs) marked by T-B lymphocyte aggregates and focal LTB and CXCL13 expression, as well as fibroblast-dominated tumors demonstrating stromal encapsulation, limited immune intermixing, and preliminary enrichment of CAF-immune suppressive ligand-receptor circuits. Together, these findings delineate inflamed versus fibrotic CRC microenvironments with distinct tumor-immune communication states. Integration of spatial features with clinical response will support refined stratification for ICI-based therapy and nominate TLS density, CAF patterning, and specific ligand-receptor modules as candidate spatial biomarkers for predicting or modulating treatment responsiveness.