A total of four LS patients (1 female and 3 males; median age 56.5 years) who fulfilled the Amsterdam II criteria and demonstrated loss of MMR proteins by IHC were enrolled. IHC screening revealed combined MLH1/PMS2 loss in two cases, combined MSH2/MSH6 loss in one case, and isolated MSH6 loss in the remaining case. The inclusion criteria for the study samples were as follows: (1) Patients pathologically diagnosed with colorectal adenocarcinoma and confirmed by colonoscopy to have concurrent adenoma lesions. (2) Three paired samples obtained from the same patient during surgery: normal colonic mucosal tissue more than 5 cm from the tumor edge (verified by HE staining and immunohistochemistry to be free of cancerous and inflammatory changes), traditional adenoma tissue diagnosed according to WHO criteria, and primary cancer tissue (AJCC 8th edition TNM staging of T2-3N0M0). (3) Confirmed by immunohistochemistry to have loss of expression in at least two of the four mismatch repair proteins MLH1, MSH2, MSH6, and PMS2 (consistent with the molecular characteristics of Lynch syndrome). (4) No preoperative neoadjuvant chemotherapy, radiotherapy, or immunotherapy. (5) Tissue sample volume greater than 0.5 cm³, with preprocessing completed within 30 minutes after excision, and single-cell suspension viability greater than 90% as verified by trypan blue staining. Exclusion criteria: (1) Coexistence with other hereditary cancer syndromes (such as FAP, Peutz-Jeghers syndrome). (2) Presence of multiple primary colorectal tumors or metastatic tumors. (3) Sample collection areas with active infections or ulcerative lesions. Surgically resected cancerous tissues, adenoma tissues, and normal colonic mucosa samples were collected. Clinical pathological data were simultaneously recorded for subsequent analysis.

Prepare a cell buffer containing 1% BSA using PBS (Biosharp, Hefei, Anhui, China) and bovine serum albumin (Solarbio, Beijing, China). Simultaneously, prepare a mixed digestion solution of type IV collagenase (2 mg/mL, Sigma, St. Louis, MO, USA) and DNase (1 mg/mL, Sigma) using PBS as the solvent, and pre-warm it at 37 °C for later use. Dilute the 10$\times$ red blood cell lysis buffer (Absin, Shanghai, China) with sterile deionized water at a ratio of 1:9 to make a 1$\times$ working solution and set aside. After rinsing the colon tissue sample three times with PBS to remove residual blood and impurities, transfer it to a sterile centrifuge tube and finely cut it into micro tissue fragments with a diameter of less than 1 mm using eye scissors. Add 5 mL of pre-warmed digestion solution (containing collagenase IV and DNase) and perform digestion at a constant temperature of 37 °C for 35 minutes, shaking and mixing every 10 minutes to enhance enzymatic efficiency. After digestion is terminated, filter the mixture through a 70 µm cell strainer (BD Biosciences, Franklin Lakes, NJ, USA), collect the filtrate, and rinse the filter membrane with PBS containing 1% BSA. Centrifuge (400 g, 5 minutes, 4 °C) to collect the cell pellet, then resuspend in 1$\times$ red blood cell lysis buffer and incubate at 4 °C in the dark for 5 minutes, followed by another centrifugation to remove the lysis products. If residual red blood cells remain in the pellet, repeat the lysis step until complete clearance is achieved. The obtained single-cell suspension, after verification of a viability rate $>$90% by Trypan Blue staining, can be used for subsequent single-cell sequencing analysis.

After confirming cell viability $>$85% by trypan blue staining, the single-cell transcriptome library was constructed using the Chromium Single Cell 3’ Kit (V3.1, 10$\times$ Genomics, Pleasanton, CA, USA). The single-cell suspension was mixed with the reverse transcription system and then co-injected with barcode- and UMI-labeled Gel Beads into the sample loading port of the microfluidic chip (10$\times$ Genomics). GEMs (Gel Beads-in-Emulsion) with a water-in-oil structure were generated through the Chromium Controller system. Collect GEMs for two-stage amplification: the first stage involves performing a reverse transcription reaction in a PCR instrument, with the program set as follows: preheating at 53 °C, a 45-minute reverse transcription reaction, inactivation at 85 °C for 5 minutes, and maintaining at 4 °C in the final stage; the second stage uses cDNA as a template for linear amplification to obtain enough library precursors. The amplified products are purified by magnetic beads, and the library preparation is completed through fragmentation, end repair, and adapter ligation. Qualified libraries are subjected to paired-end 150bp sequencing on the Illumina NovaSeq 6000 sequencer (Illumina, San Diego, CA, USA), with a data output of $\ge$50,000 reads/cell. The raw data, after passing the FastQC assessment, is stored for subsequent use. The data alignment and gene quantification analysis are then completed through the Cell Ranger pipeline.

The raw sequencing data (FASTQ format) were aligned to the GRCh38 reference genome using STAR (v2.7.11b, Cold Spring Harbor Laboratory, https://github.com/alexdobin/STAR) alignment based on Cell Ranger (v4.0, 10$\times$ Genomics, Pleasanton, CA, USA), generating a gene expression matrix. After importing the data into Seurat (4.0.2) software (New York Genome Center, New York, NY, USA, https://satijalab.org/seurat/), quality control filtering was performed: first, cells with gene expression levels below 300 or above 7000 were excluded, and abnormal cells with mitochondrial gene content exceeding 10% were removed, the DoubletFinder (v2.0.3, McGinnis et al., University of California, San Francisco, CA, USA, https://github.com/chris-mcginnis-ucsf/DoubletFinder) algorithm was then used to identify and eliminate doublet interference. Subsequently, the batch effects were corrected using the FastMNN (v1.12.0, Marioni Lab, University of Cambridge, Cambridge, UK, https://github.com/MarioniLab/FurtherMNN2018) integration algorithm, and the UMI counts and mitochondrial gene ratios were normalized (scaling factor of 10,000), followed by adjustment of gene expression levels based on the log-normalization method.

The top 2000 most variable feature genes were selected using the FindVariableFeatures function (based on the vst algorithm), with the selection criteria considering both the mean expression and coefficient of variation to ensure the capture of transcripts with significant biological relevance. During the principal component analysis phase, an iterative feature selection strategy was employed, and the top 30 principal components (cumulatively explaining $>$85% of the variance) were determined based on JackStraw analysis, effectively retaining the core variability characteristics of the data. Subsequently, the Uniform Manifold Approximation and Projection (UMAP) algorithm was employed for nonlinear dimensionality reduction, with parameters set as follows: neighborhood size n_neighbors = 30, minimum distance min_dist = 0.3, balancing the preservation of local and global data structures. Cell clustering was performed using a modularity optimization algorithm based on the Shared Nearest Neighbor (SNN) graph, constructing cell neighborhood relationships with a resolution parameter of 0.5, and applying the Leiden algorithm for community detection. The identification of marker genes for each subpopulation was conducted using a two-stage strategy: initially, differentially expressed genes (adj. p-val) were screened using the FindAllMarkers function (with the Wilcoxon rank-sum test as the testing method), followed by the validation of their biological specificity through Gene Ontology enrichment analysis, ultimately determining the top 10 characteristic marker genes for each subpopulation.

Phenotypic annotation of subpopulations was conducted by integrating the CellMarker database and literature. For the target subpopulation, differentially expressed genes (top 10 high/low expression genes) were extracted, and GSEA or GSVA analyses were performed to reveal their biological characteristics. Cell-cell communication analysis was further conducted to investigate intercellular interactions.

To further dissect the heterogeneity of epithelial cells during the progression from normal intestinal mucosa to colorectal cancer (CRC) in Lynch syndrome, we performed in-depth subpopulation analysis of epithelial cells using single-cell RNA sequencing (scRNA-seq) data. Quality control and clustering analysis are shown as (Supplementary Fig. 1A,B).

Our scRNA-seq analysis provided a detailed single-cell atlas of the progression from normal intestinal mucosa to colorectal cancer in Lynch syndrome patients. The workflow (Fig. 1A) highlights the comprehensive approach taken to generate high-quality single-cell data. The representative images (Fig. 1B) confirm the histological integrity of the collected tissues, ensuring the reliability of the subsequent analysis. The dot plot (Fig. 1C) illustrates the expression of marker genes in different cell subpopulations, facilitating the accurate annotation of cell types. This step is crucial for understanding the cellular heterogeneity within the tissues. The UMAP clustering (Fig. 1D, Supplementary Fig. 1C) identified 12 distinct cell clusters, with eight cell types successfully annotated, including T cells, B cells, endothelial cells, epithelial cells, macrophages, plasma cells, fibroblasts, and mast cells. This comprehensive cell type annotation provides a detailed view of the cellular landscape in CRC development. The UMAP plots (Fig. 1E) show the distribution of these cell types across normal mucosa, adenoma, and cancer tissues, revealing significant shifts in cellular composition during disease progression. Notably, the box plot (Fig. 1F) quantitatively demonstrates the changes in the proportions of epithelial cells, T lymphocytes, and macrophages. The observed increase in T lymphocytes and macrophages, along with the decrease in epithelial cells, suggests a dynamic interplay between these cell types during CRC development.

Fig. 1.

Single-cell atlas in colorectal cancer (CRC) development. (A) Schematic diagram of the single-cell RNA sequencing workflow. The workflow includes tissue collection, single-cell dissociation, library preparation, sequencing, and bioinformatics analysis. (B) Representative images of normal intestinal mucosa, adenomatous polyps and colorectal cancer tissues, along with HE staining verification images (scale bar = 100 µm (magnification $\times$100) and scale bar = 50 µm (magnification $\times$200)). (C) Dot plot illustrating the expression of markers in colorectal cell subpopulations. (D) The Uniform Manifold Approximation and Projection (UMAP) clustering in single-cell RNA-sequencing data of biopsies from 12 CRC patients. (E) UMAP clustering in normal intestinal mucosa, adenomatous polyps and colorectal cancer tissues from CRC patients. (F) The box plot illustrates the changes in the proportions of epithelial cells, T lymphocytes, and macrophages during the CRC development.

These findings are consistent with previous studies that have highlighted the role of immune cells in the tumor microenvironment and the importance of epithelial cell changes in CRC progression [8]. The detailed single-cell atlas generated in this study provides valuable insights into the cellular dynamics and molecular mechanisms underlying Lynch syndrome-related CRC development. Future work may focus on further elucidating the functional roles of these cell types and their interactions, potentially leading to the identification of novel therapeutic targets and biomarkers.

To further elucidate the cellular dynamics and molecular mechanisms underlying the progression from normal intestinal mucosa to colorectal cancer (CRC) in Lynch syndrome, we conducted a detailed subpopulation analysis of epithelial cells using single-cell RNA sequencing (scRNA-seq) data. This analysis aimed to identify distinct epithelial cell subtypes and their roles in disease progression.

Our detailed subpopulation analysis of epithelial cells provides critical insights into the cellular dynamics during CRC development in Lynch syndrome (Supplementary Fig. 2). The UMAP visualization (Fig. 2A) and CNV analysis (Fig. 2B, Supplementary Fig. 3) highlight the distinct epithelial cell subtypes and their potential cancerous characteristics. The marker gene expression (Fig. 2C) and their differential expression across normal mucosa (N), adenoma (P), and tumor (T) tissues (Fig. 2D) further support the subtype-specific signatures: BEST4 expression was significantly downregulated in colorectal cancer tissues versus normal intestinal mucosa (p = 4.7 $\times$ 10^-6), while CEMIP and EGFR were dramatically upregulated in colorectal cancer tissues versus adenomatous polyps (both p 0.0001), consistent with their roles as tumor-promoting markers. The changes in cell proportions (Fig. 2E) underscore the dynamic shifts in epithelial cell populations during disease progression. The pseudotime analysis (Fig. 2F, Supplementary Fig. 4) suggests a differentiation trajectory from BEST4⁺ to CEMIP⁺ and EGFR⁺ epithelial cells, indicating a potential pathogenic pathway in CRC development. The GSVA analysis (Fig. 2G) further elucidates the activation of specific signaling pathways in CEMIP⁺ and EGFR⁺ epithelial cells, highlighting their roles in promoting tumor progression. These findings are consistent with previous studies that have identified similar pathways in CRC development [8, 9]. The immunohistochemical validation (Fig. 2H) confirms the scRNA-seq findings, providing additional evidence for the changes in epithelial cell populations during CRC progression. We further explored communication intensity network between epithelial cell subsets. BEST4⁺ and CEMIP⁺ epithelial cells show strong communication via CLDN3-CLDN3, highlighting their interactions during CRC development (Supplementary Figs. 5,6). These results collectively underscore the importance of understanding the cellular and molecular mechanisms underlying CRC development in Lynch syndrome and may inform the development of targeted therapies and early detection strategies.

Fig. 2.

Distinct cellular subtypes and distribution characteristics of epithelial cells during the multistage development of colorectal cancer. (A) UMAP visualization depicting the primary clusters of epithelial cells. Each epithelial cell is represented in color according to its specific condition. (B) InferCNV evaluates the genomic variation and cancerous characteristics of epithelial cells throughout the development of CRC. (C) Dot plot illustrating the expression of markers in epithelial c cell subpopulations. (D) The violin plot illustrates the expression levels of BEST4, CEMIP and EGFR in the N, P, and T tissue types. (E) The box plot illustrates the changes in the proportions of BEST4⁺, CEMIP⁺ and EGFR⁺ epithelial cells during the CRC development. (F) Trajectories analysis of BEST4⁺, CEMIP⁺ and EGFR⁺ epithelial cells using Monocle 2. (G) GSVA analysis indicating the enriched pathways in CEMIP⁺ and EGFR⁺ epithelial cell populations. (H) Immunohistochemical staining images showing the localization and expression levels of BEST4, CEMIP, and EGFR in normal mucosa (N), adenoma (P), and cancer tissues (T) (magnification $\times$400, scale bar = 50 µm).

To further dissect the heterogeneity and roles of CD4+ regulatory T cells (Tregs) during colorectal cancer (CRC) development in Lynch syndrome, we performed a detailed subpopulation analysis of CD4+ Tregs using single-cell RNA sequencing (scRNA-seq) data. This analysis aimed to identify distinct CD4+ Treg subtypes and their dynamic changes during disease progression.

Our detailed analysis of CD4+ Treg subtypes provides critical insights into the immune dynamics during CRC development in Lynch syndrome. The UMAP visualization (Fig. 5A) and dot plot (Fig. 5B) highlight the distinct CD4+ Treg subtypes and their molecular signatures. The increase in CCR8+ CD4+ Tregs (Fig. 5C) underscores the shift in the immune microenvironment towards a more immunosuppressive state during CRC progression.

Fig. 5.

Characteristics of CD4+ Reg T cell subtypes. (A) UMAP plot depicting the subtype clusters of CD4+ Reg T cells. (B) Dot plot illustrating the expression of markers in CD4+ Reg T cell subpopulations. (C) The box plot illustrates the changes in the proportions of CCR8+ CD4+ Reg T cells during the CRC development. (D) The communication intensity network between CCR8+ CD4+ Reg T cell and other CD4+ Reg T cell subsets. (E) GSVA analysis indicating the enriched pathways in CCR8+ CD4+ Reg T cell during CRC development. (F) Trajectories analysis of CD4+ Naïve and CCR8+ CD4+ Reg T cell using Monocle 2. (G) The communication intensity network between CCR8+ CD4+ Reg T cell and other T cell subsets.

The GSVA analysis (Fig. 5E) reveals the activation of key pathways in CCR8+ CD4+ Tregs, suggesting their involvement in promoting tumor progression. These findings are consistent with previous studies that have identified similar pathways in CRC development [8]. The pseudotime analysis (Fig. 5F) provides insights into the potential differentiation trajectory of CCR8+ CD4+ Tregs from CD4+ Naïve T cells, highlighting the dynamic nature of Treg populations during disease progression.

The communication intensity network (Fig. 5D,G) highlights the significant interactions between CCR8+ CD4+ Tregs and other T cell subsets, emphasizing their role in modulating the immune microenvironment. CCR8+ CD4+ Reg T cells show strong communication with CD8+ effector and CD8+ cytotoxic T cells via HLA-A-CD8A (Supplementary Figs. 8–11). This analysis provides a comprehensive view of the Treg landscape during CRC progression and underscores the importance of understanding the immune dynamics in Lynch syndrome-related CRC.

To further understand the role of macrophages in the progression of colorectal cancer (CRC) in Lynch syndrome, we performed a detailed subpopulation analysis of macrophages using single-cell RNA sequencing (scRNA-seq) data. This analysis aimed to identify distinct macrophage subtypes and their dynamic changes during disease progression.

Our detailed analysis of macrophage subtypes provides critical insights into the immune dynamics during CRC development in Lynch syndrome. The UMAP visualization (Fig. 6A) and box plot (Fig. 6B) highlight the distinct macrophage subtypes and their dynamic changes during disease progression. The increase in C1QC⁺, IDO1⁺, and IL1B⁺ macrophages underscores the shift in the immune microenvironment towards a more pro-tumorigenic state during CRC progression.

Fig. 6.

Characteristics of macrophages subtypes. (A) UMAP plot depicting the subtype clusters of macrophages. (B) The box plot illustrates the changes in the proportions of C1QC⁺, IDO1⁺ and IL1B⁺ macrophages cells during the CRC development. (C) Dot plot illustrating the expression of markers in macrophages subpopulations. (D) The violin plot illustrates the expression levels of SPP1, FN1 and S100A6 in the N, P, and T tissue types. (E) Trajectories analysis of C1QC⁺ and IL1B⁺ macrophages using Monocle 2. (F) GSVA analysis indicating the enriched pathways in C1QC⁺ macrophages during CRC development. (G) The communication intensity network between macrophages cell subsets.

The marker gene expression (Fig. 6C) and changes in expression levels (Fig. 6D) further confirm the activation and accumulation of specific macrophage subtypes during CRC development. The GSVA analysis (Fig. 6F) reveals the activation of key pathways in C1QC⁺ macrophages, suggesting their involvement in promoting tumor progression. These findings are consistent with previous studies that have identified similar pathways in CRC development [8].

The pseudo time analysis (Fig. 6E) provides insights into the potential differentiation trajectory of IL1B⁺ macrophages from C1QC⁺ macrophages, highlighting the dynamic nature of macrophage populations during disease progression. The communication intensity network (Fig. 6G) highlights the significant interactions between C1QC⁺ macrophages and other macrophage subsets, emphasizing their role in modulating the immune microenvironment. To elucidate the specific ligand-receptor interactions between macrophages and epithelial cell subsets, we performed a detailed ligand-receptor interaction analysis. This analysis revealed that CEMIP⁺ epithelial cells interact with IL1B⁺ macrophages through MIF (CD74+CD44) and APP-CD74 pathways. Additionally, CEMIP+ epithelial cells communicate with IDO1⁺ and C1QC⁺ macrophages via the APP-CD74 pathway. These interactions highlight the complex communication network between epithelial cells and macrophages, which may contribute to the pro-tumorigenic microenvironment in CRC (Supplementary Figs. 12,13).

Despite the significant insights provided by our study, several limitations must be acknowledged. First, our sample size, while sufficient for initial characterization, remains relatively small for comprehensive analysis of all LS subtypes. Future studies with larger cohorts are needed to validate our findings and explore potential differences between LS cases with different MMR gene mutations. Second, our study focused primarily on the epithelial and immune compartments, but the stromal compartment (including fibroblasts and endothelial cells) also plays crucial roles in CRC development. Future work should extend our findings to these additional cell types to provide a more comprehensive view of the LS tumor microenvironment. Third, while we performed functional validation of DMBT1’s role using organoid models, additional in vivo studies would be valuable to confirm these findings in a more physiological context. The development of LS-specific mouse models incorporating DMBT1 manipulation could be particularly informative in this regard.

(1) Multi-omic integration: Combining scRNA-seq with epigenomic, proteomic, and spatial data could provide additional layers of understanding about LS-related carcinogenesis.

(2) Intervention studies: Testing therapeutic approaches targeting the DMBT1-WNT axis or immunosuppressive mechanisms we identified could lead to new prevention strategies for high-risk LS patients.

(3) Longitudinal tracking: Following LS patients over time with serial scRNA-seq analyses could reveal dynamic changes in the cellular landscape during cancer development and progression.

(4) Expansion to other LS-related cancers: Applying similar approaches to endometrial and other LS-related cancers could identify shared and unique mechanisms across tumor types.