Spatially resolved C1QC+ macrophage-CD4+ T cell niche in colorectal cancer microenvironment: implications for immunotherapy response

Abstract

Colorectal cancer (CRC), including both microsatellite instability (MSI) and microsatellite stability (MSS) subtypes, frequently exhibits intrinsic resistance to immunotherapy. However, the spatial tumor microenvironment (TME) and its role in distinguishing immunotherapy responders from non-responders remain poorly understood. In this study, spatial multiomics, including imaging mass cytometry (n = 50 in-house), spatial proteomics (n = 50 in-house), and spatial transcriptomics (n = 9 in-house), were employed to elucidate the spatial TME of metastatic CRC (mCRC) patients receiving immunotherapy. These methodologies were integrated with single-cell RNA sequencing (scRNA-seq), bulk RNA-seq, and bulk proteomics for comprehensive analysis and validation. A spatial immune atlas containing 314,774 cells was constructed. We found that C1QC⁺ resident tissue macrophages (RTMs) were more abundant in responders regardless of microsatellite status. Co-localization of C1QC⁺ RTMs with CD4⁺ T cells was observed in responders, and MHC-II expression facilitated their interaction. In contrast, cancer-associated fibroblasts inhibited this interaction in non-responders. Moreover, whole genome screening identified key genes involved in antigen presentation in C1QC⁺ RTMs. Hence, our study highlights the importance of spatial immune mapping in revealing the complex spatial topology of CRC and corresponding immunotherapy response.

Fig. 1. A subset of MSS and MSI mCRC patients responded to immunotherapy.

a Overview of the study design. b Progression-free survival (PFS) of systemic chemotherapy and immunotherapy in 16 microsatellite stable (MSS) distant mCRC patients, including 8 responded (R) patients and 8 non-responded (NR) patients. c PFS of immunotherapy in 9 microsatellite instable (MSI) CRC patients, including 5 R patients and 4 NR patients. d Representative CT images of MSS/MSI_R and NR patients.

Fig. 2. Spatial immune phenotypes of MSS/MSI_R and NR CRC patients at single-cell resolution.

a The workflow of IMC analysis. b UMAP plot visualizing the clustering of broad cell types within the IMC dataset (top left). Box plot showing the relative frequencies of each major cell type among all cells in the IMC dataset, providing an overview of their abundance in the analyzed samples (top right). UMAP plots showing the clustering of monocytes and macrophages, T cells and NK cells in the IMC dataset (bottom). c The immune cell subsets frequencies in total cells. d Spatial distribution of cells (immune, stromal, and epithelial cells, and major cell subtypes) in representative sections from the four groups, annotated based on manual classification results. e Prevalence of immunotherapy outcomes for each cell cluster, as estimated by the cell frequency-based Ro/e analysis. f Risk ratios illustrating the associations between cell frequency and immunotherapy response are shown for each cell subpopulation in the MSS (left) and MSI (right) TME, as determined by logistic regression analysis. g Workflow of cosine similarity and ResNet18-based deep learning. h Heatmap visualization of cosine similarity scores between TME images. Each cell represents the pairwise similarity score between two samples, with warmer colors indicating higher similarity (scale: 0.75–1.0). The labeled values indicate the average similarity computed within individual groups or between different groups. i Feature importance calculated by ResNet18-based deep learning.

Fig. 3. The spatial cell neighboring and cell–cell distance phenotype of MSS/MSI_R and NR CRC patients.

a Schematic diagram of cell neighbor (CN) identification. b Representative Network, Voronoi and CN diagrams of the TME in MSS/MSI_R and NR CRC samples. c Heatmap of 15 distinct CNs based on the 29 original cell clusters and their respective abundances within each CN. d Tissue prevalence of each CN cluster estimated by the Ro/e analysis. e Pairwise cell‒cell interaction heatmap. The background color represents overall interaction count intensity across all samples, with darker shades indicating higher interaction intensities. Small squares denote statistically significant interactions in comparisons between different patient groups, with their color representing the log2 fold change (log2FC) in interaction strength. f Violin plots illustrating the distances between C1QC⁺ RTM and CD4⁺ T cells across the four groups. Significance between the two groups was evaluated by the t tests with ****P < 0.0001. The one-way ANOVA test was used to compare the four groups. g Heatmap displaying the expression levels of T cell-related markers (e.g., CD38, CD57, GZMB, TNFα, PD-1) in CD4⁺ T cells proximal to C1QC⁺ RTMs vs those distal to C1QC⁺ RTMs, colored by z-score normalized expression levels. h Ro/e analysis of CD4⁺ T cells proximal to C1QC⁺ RTMs and distal to C1QC⁺ RTMs in the four groups. i The distance of immune cell subsets to fibroblasts associated with the prognosis of CRC. The Hazard Ratio and P value was calculated on data from IMC by univariate Cox analysis. (j) Schematic diagram of fibroblast barrier score calculation. The barrier score measures the degree of spatial interpositioning of C1QC⁺ RTM–adjacent fibroblasts between CD4⁺ T cells and their nearest C1QC⁺ RTM (s) in a tissue core. In the lower half of the schematic, four nearest C1QC⁺ RTMs are defined for the purple CD4⁺ T cells. C1QC⁺ RTMs–adjacent fibroblasts are found on three of these four paths from CD4⁺ T-cell to C1QC⁺ RTMs, resulting in a barrier score of 3/4. (k) Boxplot comparing the fibroblast barrier scores in the four groups. Statistical significance was determined by one-way ANOVA comparing MSI_R, MSI_NR, MSS_R, and MSS_NR groups.

Fig. 4. scRNA-seq analysis indicates C1QC⁺ RTM-mediated immune activation via MHC class II signaling.

a UMAP plot of broad cell types from the Li et al. cohort. b UMAP plot of the macrophages showing transcriptionally distinct clusters. c UMAP plots showing the expression of selected marker genes in macrophages. d–f The Resident tumor macrophage (RTM) score, M2 score and innate anti-PD-1 resistance (IPRES) score in distinct cell clusters. The one-way ANOVA test was adopted to evaluate the statistical significance. g Abundance of each macrophage cluster in the tissue of pCR and non-pCR groups was estimated via Ro/e analysis. h Cell cluster frequency shown as a fraction of total macrophages in pCR and non-pCR group. i C1QC⁺ RTM score calculated by single sample gene set enrichment analysis (ssGSEA) method on the basis of bulk transcriptome from pCR and non-pCR group. The Wilcoxon Rank-Sum Test was adopted to evaluate the statistical significance. j GO analysis of C1QC⁺ RTMs and the other three macrophage clusters. k Violin plot showing the expression levels of MHC-II molecules in C1QC⁺ RTMs and the other three macrophage clusters. l The outgoing and ingoing interaction strength of immune cells in pCR and non-pCR group. The x-axis and y-axis scales differ between the pCR and non-pCR groups. m The number of pair–ligand interactions between T cells and C1QC⁺ RTM in the pCR and non-pCR groups. n Differences in the MHC-II pathway interaction of various cell types. The thicker the line, the stronger the connection. o Up-regulated and down-regulated receptor–ligand pairs that differ significantly between pCR and non-pCR based on C1QC⁺ RTM and other cell clusters. Dot size indicates the P value, colored according to the communication probability of pathways.

Fig. 5. Paired spatial resolved proteomics confirm the antigen presenting role of C1QC⁺ RTMs in immunotherapy.

a The potential biological functions and relevant signaling pathways of MSS/MSI_R and NR CRC patients were evaluated by the GO analyses. The hypergeometric test for over-representation was adopted to evaluate the statistical significance with multiple tests corrections. b Gene set enrichment analysis (GSEA) enrichment for APC and C1QC⁺ RTMs in MSS_NR and MSS_R samples. NES, normalize enrichment score. The Kolmogorov-Smirnov test was adopted to evaluate the statistical significance with multiple tests corrections. c Fast gene set enrichment analysis (FGSEA) enrichment for top 10 upregulated pathways and top 10 downregulated pathways comparing MSS_NR and MSS_R according to the hallmark gene sets. d Heatmap showing different immunotherapy-related pathways enriched in the integrated MSS/MSI_R and NR groups by gene set variation analysis (GSVA) analysis, colored by z-score transformed mean GSVA scores. e Boxplot illustrating the C1QC⁺ RTM scores, calculated using the C1QC⁺ RTM signature, in spatial proteomics analyses. The one-way ANOVA test was adopted to evaluate the statistical significance. f Scatter plots showing Pearson’s correlation between C1QC⁺ RTM and MHC-I score, calculated using the C1QC⁺ RTM signature and MHC-I signature, in spatial proteomics analyses. g, h Scatter plots depicting the Pearson’s correlation between C1QC⁺ RTM and MHC-II scores, as determined through spatial proteomics analyses and further validated through bulk proteomics. The Pearson Coefficient Test was adopted to evaluate the statistical significance.

Fig. 6. Whole genome screening detecting the potential related pathways involved in regulating APC in C1QC⁺ RTMs.

a Schematic representation of the workflow for the genome-wide CRISPR/Cas9 screen. b Cumulative distribution function (CDF) of biological replicates of 5 representative samples of M0, MHC^low, and MHC^high cells. CPM, counts per million. n = 5 in each group. c Top hit genes of MHC^low and MHC^high cells. The P value was corrected by Benjamini-Hochberg test. d Gene ontology (GO) term analysis of MHC^low and MHC^high cells. e, f KEGG gene interaction network of the hit genes in MHC^low and MHC^high cells. Subnetworks (Neighborhoods) are colored and annotated with enriched functional categories. Gray lines, connections within a neighborhood; red lines, connections between neighborhoods. g, h Representative flow cytometric plot and quantification of MHC-II expression levels in ESRRA-KO THP-1 cells during macrophage polarization. Vector control and THP-1-Cas9 were used as control for comparison. Vector control indicates THP-1 cells infected with lentivirus carrying an empty plasmid lacking gRNA. gMFI, geometric mean fluorescence intensity. P values in h were determined using one-way ANOVA.

Fig. 7. Differences in the co-localization of C1QC⁺ RTMs and CD4⁺ T cells between R and NR CRC patients revealed by spatial transcriptomics.

a Schematic diagram of niche computation, H&E staining of each sample, and the spatial map of niches. The white dashed line in the H&E section separated the para tumor from the tumor, and the black dashed line separated the tumor from the necrotic region. b Co-localization analysis using the MISTy algorithm. Median importance of cell-type abundance in predicting the abundances of other cell types within a spot for MSI_R (top left), MSI_NR (top right), MSS_R (bottom left), and MSS_NR (bottom right). c Spatial feature plots showing signature scores for MSI_R, MSI_NR, MSS_R, and MSS_NR samples. d Violin plots illustrating the gene set scoring results for each niche in the MSI_R_1 sample.

Fig. 8. Heterogeneous TME of immunotherapy responder and non-responder CRC patients and the working model for ResNet18-based deep learning.

Immunotherapy-sensitive CRC presents higher infiltration of C1QC⁺ RTM and CD4⁺ T cell pair and lower fibroblasts than immunotherapy-resistant CRC. ResNet18-based deep learning further effectively dissects the detailed spatial topology of the CRC TME and highlights the vital role of C1QC.