Key genes and pathway differences between serrated polyps and conventional adenomas: insights from multi-omics

Introduction

Colorectal cancer (CRC) is the third most prevalent malignancy worldwide and the second leading cause of cancer-related mortality (1,2). The American Cancer Society predicts a significant increase in both the incidence and mortality of CRC over the next decade for individuals 54 years or younger (3). Most CRC cases evolve from polyps, although ultimately, only a minority progress to cancer (4). Two principal precancerous pathways in CRC are conventional adenomas (CA) and serrated polyps (SP) (5,6). As a recently identified precursor lesion, SP have instigated changes in the paradigms of CRC prevention and treatment (7). SP and CA exhibit multiple pathological types, and the heterogeneity in these types contributes significantly to the varied pathways of CRC development (8,9). According to current research, removing polyps prior to malignant transformation is still widely recognized as an effective preventive measure (10,11). Missed SP contribute to preventable CRCs (12). However, even with such interventions, the occurrence of CRC cannot be averted entirely (13). While endoscopic examinations can prevent cancerous transformation, they are not fully effective. For instance, sessile SP (SS), which presents a flat morphology under endoscopic view, undoubtedly increases the difficulty of detection (14). Artificial intelligence-powered computer-aided detection (CADe) has demonstrated consistent improvements in adenoma detection rates (ADRs) and adenoma miss rates (AMRs); largely by detecting diminutive polyps, but it shows limited efficacy in identifying advanced adenomas or sessile serrated lesions (15).

Over the past decade, several single-omics technologies have been individually applied to CRC and its precancerous lesions. Genome-wide association studies (GWAS) have identified numerous germline susceptibility loci for CRC precancerous lesions, including variants that appear to influence the development of CA and serrated lesions. Previous GWAS-based analyses have shown that several CRC susceptibility variants can predict the risk of advanced or multiple adenomas and collectively account for a substantial proportion of the multiple-adenoma phenotype (16,17). However, most risk loci reside in non-coding regions and their effector genes and target cell types remain difficult to pinpoint (18). Additionally, Hang et al. utilized GWAS to investigate the variant differences between SP and CA, providing potential etiological explanations (19). Burnett-Hartman and colleagues identified that different types of intestinal polyps correspond to distinct CRC susceptibility single nucleotide polymorphisms (SNPs) (20). However, they could not provide further insights into the gene transcription and translation processes related to these findings. Single-cell RNA sequencing (scRNA-seq) has begun to elucidate the cellular architecture of CRC precancerous lesions at the individual-cell level. Recent scRNA-seq studies of serrated lesions have revealed distinct epithelial programs—including stem-like, absorptive and redox-imbalanced states—and identified lesion-specific signaling alterations in the serrated pathway (21). Complementarily, Chen et al. identified enriched serrated-specific cells (SSC) cell populations in SPs and adenoma-specific cells (ASC) cell populations in CAs. They proposed that SPs originate from differentiated cells, whereas CAs arise from stem cells (22), which opens avenues for exploring the cell communication processes in the polyp-to-cancer transformation.

Transcriptome-wide association studies (TWAS) is an emerging methodology that can estimate the impact of expression quantitative trait loci (eQTLs) expression on gene expression to infer their potential influence on traits (23). A study conducted in 2021 utilized individual-level eQTLs and conducted large-scale TWAS, identifying 25 new CRC risk genes, also validated through colocalization analysis (24). Emma Hazelwood and colleagues employed TWAS, integrating Mendelian randomization (MR) and colocalization, alongside sex- and anatomical subsite-specific analyses. SEMA4D, which encodes a protein targeted by an investigational CRC therapy, has been identified as a key risk gene (25). A recent large-scale transcriptomic analysis has revealed TGF-β-responsive stromal activation during the early phases of serrated carcinogenesis, providing critical insights into the multifaceted role of TGF-β in the initial stages of this pathway (26). However, no researchers have used TWAS to identify potential genetic influences transforming colorectal polyps into CRC. Also, no studies have explored the pre-carcinogenic changes in polyps using summary-level eQTLs.

In this study, we sought to move beyond single-omics analyses by systematically integrating CRC GWAS summary statistics with scRNA-seq data from colorectal polyps, tissue eQTL resources, TWAS and fine-mapping. GWAS combined with scRNA-seq analysis was utilized to explore the cellular populations associated with precancerous lesions. Utilizing the latest TWAS framework, summary-level eQTLs were applied for transcriptome-wide analyses of genes related to CRC, aiming to identify risk genes associated with the carcinogenic transformation of polyps. To substantiate our findings, we further conducted correlative validations in both scRNA-seq and RNA sequencing (RNA-seq) datasets. By integrating multiple omics layers at cell-type resolution, we sought to delineate the molecular characteristics that distinguish the two major colorectal polyp precancerous pathways, thereby overcoming the limitations of single-omics approaches and providing mechanistic insights that may inform personalized CRC prevention, screening and surveillance strategies. We present this article in accordance with the STREGA reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2138/rc).

Methods

Study design

The study flow chart is depicted in Figure 1. We employed various methods to identify critical genes and pathways, including single-cell analysis combined with GWAS, TWAS analysis, summary data-based Mendelian randomization (SMR), and gene set enrichment analysis (GSEA). This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Figure 1 Flowchart of the multi-omics analysis in this study. ABS, absorptive cells; ASC, adenoma specific-cells; CA, conventional adenomas; CRC, colorectal cancer; eQTL, expression quantitative trait loci; GSEA, gene set enrichment analysis; GWAS, genome-wide association study; SMR, summary data-based Mendelian randomization; SP, serrated polyps; SSC, serrated-specific cells.

Data acquisition

The GWAS on CRC involved a comprehensive analysis of 64,190 individuals (27). This participant cohort was compiled from 45 independent studies coordinated by three significant research consortia: the Genetic Epidemiology of Colorectal Cancer Consortium (GECCO), the Colorectal Cancer Family Registry (CCFR), and the Colon Cancer Family Registry (CORECT). The study included 34,869 CRC cases and 29,051 control subjects related to CRC. scRNA-seq data for colorectal polyps were sourced from the COLON MAP study (22), which involved the collection of polyp samples and corresponding normal biopsies from various participants. The discovery set included 65 specimens with 30 tumors, and the validation set comprised 63 specimens with 32 tumors. A total of 128 scRNA-seq datasets were generated from 62 tumor samples.

The transcriptome and clinical information of patients with colorectal polyps were downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). The GEO CRC cohorts included GSE76987 (n=86), GSE116305 (n=15) and GSE22242 (n=5) datasets. To source blood and transverse colon cis-eQTL summary data, the Genotype-Tissue Expression (GTEx) project’s Version 8 release was used as the foundational dataset (28). The GTEx V8 data, including the relevant eQTL information, can be accessed through the GTEx portal at the following link: https://www.gtexportal.org/home/datasets.

Combining single-cell analysis with GWAS to identify key cell populations and genes

scPagwas was used to integrate scRNA-seq data from colorectal polyps with GWAS data on CRC (29). This method annotates SNPs from GWAS to genes in specific pathways, employing a linear regression model to correlate SNP genetic effects with gene expression levels within these pathways. The approach calculates a regression coefficient for each cell, indicating the contribution of SNP to heritability based on pathway-based gene expression. Additionally, scPagwas transforms gene-by-cell matrices into pathway activity scores (PAS) using the first principal component, focusing on pathways within a specified gene size range from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

scRNA-seq data were processed using the Seurat R package (30). Cells were classified according to predefined cell types mentioned in the source text. In order to explore intercellular interactions among different cell types, we employed CellPhoneDB to identify significant ligand-receptor pairs between SSC, ASC, and other cell populations (31). Interaction scores were defined as the mean of average expression values for each ligand-receptor pair across the involved cell types. The expression of any complexes outputted by CellPhoneDB was calculated as the sum of the expressions of their constituent genes. To investigate the biological states or functional differences between SSC and ASC cells, we analyzed differentially expressed genes (DEGs) between these cell types, which were then used to explore KEGG sets through GSEA (32). Additionally, we utilized Monocle3 to infer cellular lineage trajectories (33). After constructing cell trajectories, the “differentialGeneTest” function in Monocle3 was used to detect genes differentially expressed along pseudotime. This comprehensive approach elucidated the dynamic changes in gene expression associated with cell differentiation and development.

Cell-cell communication and gene-pathway association analysis

We further extended cell-cell communication analysis using CellChat. After computing communication probabilities and filtering low-support interactions (min. cells =10), we quantified directional interaction strength using the aggregated communication matrix (net$weight). Ligand-receptor pairs between SSC/ASC and ABS were extracted using subsetCommunication, and interaction probabilities were compared across SS and AP datasets. Bubble plots and bar plots were generated to visualize pathway-level and pair-level differences. All analyses followed the standard CellChat workflow with default parameters unless otherwise specified.

To evaluate pathway activity at the single-cell level, we calculated module scores for selected KEGG and Hallmark pathways using the AddModuleScore function in Seurat. For MIR4435-2HG and SMAD9, cells were stratified into high and low expression groups based on the median expression within the corresponding cell type. Pathway scores were compared between high- and low-expression groups using Wilcoxon rank-sum tests and visualized with violin plots.

We assessed the association between gene expression and pathway activity using Spearman correlation. Correlation coefficients and P values were computed using the cor.test function in R, and scatter plots with linear smoothing were generated using ggplot2.

TWAS analysis and fine-mapping

Omnibus Transcriptome Test using Expression Reference Summary data (OTTERS) framework was used for a two-stage TWAS, incorporating polygenic risk score (PRS) methods (34). In the first stage, OTTERS trained genetically regulated expression (GReX) imputation models for each gene using PRS methods such as P value thresholding with linkage disequilibrium clumping (P+T), lassosum, summary statistics based dirichlet process regression (SDPR), and PRS-continuous shrinkage (PRS-CS). These methods inferred cis-eQTL weights using summary data and an external linkage disequilibrium (LD) reference panel of European ancestry. Next, GReX was imputed for each gene using the derived cis-eQTL weights, and gene-based association analysis was conducted using the test GWAS dataset. This process yielded a set of TWAS P values for each gene, one for each training method. Then, we combined these P values using the ACAT-O omnibus test, which uses a Cauchy distribution for inference.

Gene-based Integrative Fine-mapping for performing conditional TWAS analysis (GIFT) was used to perform conditional TWAS analysis, examining one genomic region at a time to identify marginal TWAS genes and finely infer causal genes more closely related to traits. GIFT considers the correlation in expression levels between different genes within the selected region and the LD relationship of cis-SNPs, accounting for the uncertainty in the constructed GReX. Herein, we combined the weights of SNP expression effects obtained from the OTTERS framework to calculate the final P values for fine-mapping. Fine-mapping P values <1e−100 were considered to have a stronger association with the trait.

Summary-data-based MR analysis

SMR methodology was employed to integrate summary-level data from GWAS of CRC with eQTL data from whole blood and transverse colon (35). The SMR approach is designed to identify genes with expression levels associated with complex traits attributable to pleiotropy. This analysis uses advanced statistical methods to examine the association between gene expression and CRC traits, carefully considering potential non-genetic confounders. The following settings were used: --maf 0.01, --thread-num 10, --diff-freq 1. Additionally, the SMR framework incorporates the HEIDI test to discern between pleiotropy and linkage, a critical step in determining the genetic architecture of complex human traits. To distinguish pleiotropy/causation from linkage, we also used the heterogeneity statistic and p_heidi >0.05 to exclude genes due to linkage. The false discovery rate (FDR) <0.05 was considered statistically significant.

Phenome-wide association studies analysis

Genes identified as significant in the SMR were further explored through a phenome-wide association study (PheWAS). This approach uses gene expression as the exposure variable and a wide array of disease outcomes based on summary statistics from the UK Biobank cohort (with a sample size of up to 408,961 individuals). To account for imbalances in case-control ratios, we specifically focused on disease GWAS data from the UK Biobank, which were analyzed using the Scalable and Accurate Implementation of Generalised Mixed Model (SAIGE V.0.29) (36).

Given statistical power considerations, we selected 783 traits (diseases) with >500 cases for our phenome-wide MR analyses. The summary statistics for disease-associated SNPs were obtained from the SAIGE GWAS database (https://www.leelabsg.org/resources) (36). With these data, we performed MR analysis using the eQTL of the identified druggable genes in blood. This analysis was executed using the inverse variance weighted (IVW) method. The causal effects were considered statistically significant if they met the threshold of an FDR <0.05.

Statistical analysis

A sensitivity analysis was performed by repeating the primary analysis on the complete-case dataset to ensure the robustness of the findings. Prior to analysis, missing data for key covariates were handled using multiple imputation. To address potential sources of bias, particularly from confounding variables, our primary analysis utilized a multivariable regression model to adjust for known potential confounders. To account for potential correlations between subjects, we utilized generalized linear mixed-effects models. All results were reported as odd ratios, with 95% confidence intervals (CIs). All statistical analysis and data visualization were performed in R software 4.2.2 (https://www.r-project.org/). All the above analyses were conducted on the Ubuntu 22.04 operating system (https://www.releases.ubuntu.com/22.04/).

Results

SSC and ASC exhibit significant differential pathways and cell communication compared to the ABS population

We clustered the single-cell transcriptome atlas of colorectal polyps from Chen et al. After filtering, 65,088 cells were obtained; 21,275 were from adenoma polyps (AP), 6,094 were from SS, and 37,719 were from other tissues. After Uniform Manifold Approximation and Projection (UMAP) clustering analysis, we annotated cell clusters using recommended marker genes, resulting in the identification of the following nine cell types: absorptive cell (ABS), crypt top colonocytes (CT), enteroendocrine cells (EE), goblet cells (GOB), stem cells (STM), transit amplifying cells (TAC), and tuft cells (TUF), as well as ASC and SSC (Figure 2A).

Figure 2 ASC and SSC cell groups exhibit distinct cell communication and key pathways. (A) UMAP dimensionality reduction clusters identify different cell groups in intestinal polyps. (B) Analysis based on scPagwas infers key cell groups relevant to CRC onset. (C) Cell communication analysis infers differential pathways in SS and AP cells. (D) GSEA analysis of the expression differences in the TGF-β pathway in SSC and ASC cells. (E,F) Directional communication strength from ABS to SSC and ASC. ABS→SSC interactions were consistently stronger than ABS→ASC in both SS and AP tissues. (G) Ligand-receptor interaction probabilities comparing SSC→ABS and ASC→ABS. Most key pairs show higher activity in SSC than in ASC. (H) TGF-β pathway module scores across ABS, SSC, and ASC populations. ASC exhibits the highest TGF-β activity, whereas ABS and SSC show similarly lower levels. Cell communication analysis displays receptor differences in SS and AP. ABS, absorptive cells; AP, adenoma polyps; ASC, adenoma-specific cells; CI, confidence interval; CRC, colorectal cancer; CT, crypt top colonocytes; EE, enteroendocrine cells; GOB, goblet cells; GSEA, gene set enrichment analysis; scPagwas, pathway-based polygenic regression method that integrates scRNA-seq and genome-wide association study (GWAS) data; SS, sessile serrated polyps; SSC, serrated-specific cells; STM, stem cells; TAC, transit amplifying cells; TUF, tuft cells; UMAP, Uniform Manifold Approximation and Projection.

Next, an association analysis was performed on the single-cell transcriptomes of colorectal polyps, excluding normal tissues, and GWAS data for CRC, utilizing the pathway-based polygenic regression method (scPagwas). This approach revealed that the ABS and SSC cell groups had a significant role in the onset and progression of CRC (Figure 2B). Under the scPagwas analysis, no significant relevance of the ASC cell group to the onset of CRC was found.

Cell communication in SS and AP tissues was analyzed separately for a more detailed examination. We identified distinct interactive pathways in both tissues; notably, the VEGF pathway was enriched in SS tissue but not in AP. Pathways exclusively enriched in AP included GDF, VISFATIN, and GALECTIN (Figure 2C).

Differential genes were sought in the SSC and ASC cell groups, and GSEA analysis was conducted using the KEGG database with pathways starting with ‘has’ for human relevance. This analysis revealed an overall downregulation of TGF-β in SSC, while an opposite trend was observed in ASC (Figure 2D), suggesting that in the SSC cell group, serrated cells may often be unable to differentiate into various cell types, potentially transforming into cancer cells under certain stimuli.

Using SSC and ASC cell groups as the origin, receptor pair displays were performed separately in the two types of tissues (Figure 2E). The GDF pathway, uniquely enriched in AP, was paired with the receptor GDF15-TGF-β, which aligned with the GSEA analysis results. The EGFR-ERBB2 receptor pair was found exclusively in SSC, suggesting a richer epithelial growth in SSC cells, potentially promoting carcinogenesis (Figure S1A,S1B). To further examine epithelial communication patterns, we quantified signaling sent from ABS cells toward SSC and ASC. ABS→SSC interactions were consistently stronger than ABS→ASC in both SS and AP tissues (Figure 2F), indicating preferential coupling between absorptive cells and SSC. We next compared ligand-receptor usage across the two epithelial populations. Most key pairs—including GUCA2A-GUCY2C, GRN-SORT1, and MDK-associated interactions—showed substantially higher probabilities in SSC→ABS than in ASC→ABS (Figure 2G), suggesting that SSC maintains a richer signaling interface with ABS. Finally, TGF-β module scoring showed that ASC exhibited the highest pathway activity, whereas SSC and ABS displayed similarly lower levels (Figure 2H), consistent with their shared signaling features observed above.

The OTTERS framework for TWAS identified susceptibility genes for CRC

The OTTERS framework established four imputation model-based cis-eQTL weights, and after ACAT-O testing, key genes combined with GWAS were identified (Figure 3A). The Eqtl from Yang Lab’s transverse colon and a reference LD panel were utilized to estimate cis-eQTL weights, identifying 112 key genes through CRC GWAS screening (OTTERS_P<0.05). A Manhattan plot was created to display the relationship between key genes and their chromosomal positions (Figure 3B). Detailed locations and OTTERS_P values for each gene are shown in Table S1.

Figure 3 Summary-level eQTLs are utilized for TWAS analysis in CRC to infer susceptibility genes. (A) The process of TWAS analysis using the OTTERS framework. (B) Manhattan plot of susceptibility genes across 22 chromosomes. (C) Manhattan plot displaying the causal genes with P values less than 1e−100 across all chromosomes after GIFT fine-mapping the TWAS genes. CRC, colorectal cancer; eQTL, expression quantitative trait loci; GIFT, Gene-based Integrative Fine-mapping for performing conditional TWAS analysis; GWAS, genome-wide association study; LD, linkage disequilibrium; OTTERS, Omnibus Transcriptome Test using Expression Reference Summary data; P+T, P value thresholding with linkage disequilibrium clumping; PRS-CS, polygenic risk score-continuous shrinkage; SDPR, summary statistics based dirichlet process regression; TWAS, transcriptome-wide association studies.

We conducted GIFT analysis to obtain the causal genes associated with traits after fine-mapping the significant TWAS genes on each chromosome. We presented the causal genes with P values <1e−100, further mitigating the influence of marginal TWAS genes (Figure 3C).

Multi-Omics interaction analysis identifies key genes influencing carcinogenic transformation in SSC and ASC cell population

The integration of scRNA and GWAS data through scPagwas identified the top 2000 significant genes, six of which, PITPNC1, SMAD9, POU2AF2, LIMCH1, MIR4435-2HG, and HES6, were also recognized by the OTTERS framework (Figure 4A). Among them, MIR4435-2HG, LIMCH1, and SMAD9 had P values <1e−100 in the fine-mapping analysis. The expression of these six genes was validated in the original single-cell atlas of colorectal polyps. Specifically, PITPNC1 and MIR4435-2HG showed notably higher expression in the SSC cell group. However, no genes were distinctly overexpressed in the ASC cell group (Figure 4B). To account for the influence of other cell groups on expression levels, the expression data of SSC and ASC were extracted and analyzed separately for differences (Figure 4C,4D). Compared to SSC cells, GDF15 and TGFBR2 were significantly downregulated in ASC cells. SMAD9, a part of the TGF-β pathway, also showed a trend of downregulation. Among the evaluated genes, MIR4435-2HG and SMAD9 showed the most statistically significant differential expression between SSC and ASC cells, further supporting their prominence among the risk candidates. Similarly, these genes were reanalyzed in SS and AP tissues (Figure S2A). Pseudotime analysis was used to analyze the significant receptors and six genes. In SS tissues, MIR4435-2HG and PITPNC1 came to be activated in SSC cells over time, while GUCA2A underwent a downregulation followed by reactivation (Figure 4E). In AP tissues, an interesting observation was made with the rise in GDF15, TGFBR2, and PITPNC1 expression in ASC cells, aligning with previous findings. GUCA2A exhibited a reverse process of activation followed by inhibition (Figure 4F). The scPagwas analysis provided pathway scores for all cells. To further affirm the significance of these key genes, a Pearson correlation analysis was conducted between them and the pathway scores. Notably, in SS tissues, MIR4435-2HG was correlated with the TGF-β and ErbB pathways, which was not observed in AP tissues (Figure S2B,S2C).

Figure 4 Single-cell combined TWAS analysis identifies key differential genes in SP and CA. (A) Venn diagram showing overlap of TWAS susceptibility genes with key genes from scPagwas analysis. (B) Dot plot of key gene expression across different cell groups. (C) Expression values of key genes in the two types of polyps. (D) Volcano plot of differential genes in ASC and SSC cells. (E,F) Pseudo-time analysis to determine the expression trends of key genes over time in both types of polyps. (G) SMR analysis to validate the genetic effect of key gene POU2AF2. ABS, absorptive cells; AP, adenoma polyps; ASC, adenoma-specific cells; CA, conventional adenoma; CT, crypt top colonocytes; EE, enteroendocrine cells; GOB, goblet cells; scPagwas, pathway-based polygenic regression method that integrates scRNA-seq and genome-wide association study (GWAS) data; SMR, summary data-based Mendelian randomization; SP, serrated polyps; SS, sessile serrated polyps; SSC, serrated-specific cells; STM, stem cells; TAC, transit amplifying cells; TUF, tuft cells; TWAS, transcriptome-wide association analysis.

We used SMR to re-evaluate the aforementioned key genes. Only the eQTL associated with POU2AF2 was found to be linked to CRC, showing a reduced risk of its occurrence (b<0) (Figure 4G). POU2AF2 had been previously correlated with SMAD9, suggesting it may be a regulator of the TGF-β-SMAD pathway. Furthermore, POU5F1B, identified as one of the most significant genes in TWAS, also showed statistical significance in SMR analysis (Figure S2D). Unfortunately, significant P values were not observed for the eQTLs associated with the other five genes, which may be due to the higher discovery likelihood in summary-level screening using the OTTERS framework.

Pathway alterations in ABS cell group across SS and adenomas polyps

In this bioinformatics study, significant focus was placed on analyzing the ABS cell group within AP and SS tissues based on key cell groups identified through scPagwas. The analysis involved differentially evaluating the scPagwas pathway scores for these two tissues, identifying the top five pathways with the most pronounced upregulation and downregulation (Figure 5A). Re-extraction of the ABS cell group from the original atlas was followed by a GSEA analysis, which revealed opposing trends in oxidative phosphorylation, chemical carcinogenesis-reactive oxygen pathways, and thermogenesis between SS and AP (Figure 5B-5G). We identified genes PRKAA1, COX5B, COX14, and ATG10, recognized by TWAS, within the aforementioned three pathways. In addition to EGFR-ERBB2, receptor pairs GRN-SORT1 and GUCA2A-GUCY2C were exclusively present in ABS-SSC interactions (Figure 2F,2G). GUCA2A was found to be correlated with all three pathways in AP tissue, a phenomenon not seen in SS. A similar pattern was noted with GUCY2C and PRKAA1, even though they were related only to a subset of these pathways (Figure 5H).

Figure 5 Pathway changes of ABS cell population and relationship with TWAS key gene. (A) Differential analysis of scPagwas pathway scores identifies the top 5 upregulated and downregulated pathways in SS and AP. (B-G) GSEA analysis reveals differences in the expression of oxidative phosphorylation, chemical carcinogenesis-reactive oxygen pathways, and thermogenesis between SS and AP. (H) Correlation analysis of key gene expression and pathway scores from scPagwas. ABS, absorptive cell; AP, adenoma polyps; GSEA, gene set enrichment analysis; scPagwas, pathway-based polygenic regression method that integrates scRNA-seq and genome-wide association study (GWAS) data; SS, sessile serrated polyps; TWAS, transcriptome-wide association studies.

MR analysis establishes causal relationships between key genes and metabolic phenotypes

Mitochondrial-related metabolic changes were observed as a significant factor influencing differences between SS and AP tissues. In order to explore potential genetic correlations between genes and metabolic traits, MR analysis was conducted on key SNPs extracted from all the genes and receptors, focusing on their association with mitochondrial-related metabolic phenotypes. Notably, a negative correlation was found between SMAD9 and hydroxymethylglutaryl-CoA synthase (prot-a-1356), indicating that the pronounced expression of SMAD9 in ASC cells inhibits oxidative phosphorylation, aligning with our previous hypothesis (Figure 6). In the ABS cell group, COX5B and PRKAA1, highly expressed in AP tissue, also seemed to suppress related mitochondrial metabolic phenotypes. However, this effect was not observed in PITPNC1, which is highly expressed in SS tissue.

Figure 6 Mendelian randomization analysis of key genes and their genetic association with oxidative phosphorylation phenotypes. prot-a-641, cytochrome c oxidase subunit 8A, mitochondrial; prot-a-640, cytochrome c oxidase subunit 7A1, mitochondrial; prot-a-64, apoptosis-inducing factor 1, mitochondrial; prot-a-534, coiled-coil-helix-coiled-coil-helix domain-containing protein 10, mitochondrial; prot-3015, transmembrane protein 70, mitochondrial; prot-a-2454, tRNA pseudouridine synthase A, mitochondrial; prot-a-2025, NADH dehydrogenase [ubiquinone] iron-sulfur protein 4, mitochondrial; prot-a-1965, Peptide chain release factor 1-like, mitochondrial; prot-a-1945, ribosome-recycling factor, mitochondrial; prot-a-1356, hydroxymethylglutaryl-CoA synthase, mitochondrial; prot-a-1220, glutaredoxin-2, mitochondrial; prot-a-1055, phenylalanine-tRNA ligase, mitochondrial. nSNP, nonsynonymous single nucleotide polymorphism.

Validation of key gene correlations and phenome-wide association study

Our gene expression findings were validated using data from GEO (GSE76987). It was observed that MIR4435-2HG and PITPNC1 remained highly expressed in SS tissue, while expression of SMAD9 was significantly lower in AP tissue (Figure 7A). Correlation analysis revealed a negative correlation between MIR4435-2HG and PITPNC1 with SMAD9 in SS tissue and a positive correlation in AP tissue. Interestingly, a high correlation between SMAD9 and TGFBR2, as well as GDF15, was observed in AP tissue, with almost no correlation in SS tissue (Figure 7B).

Figure 7 Validation of key genes in RNA-seq and PheWAS analysis. (A) Differential expression analysis of key genes in SP and CA using GSE76987 RNA-seq data. (B) Pearson correlation analysis of key genes. (C,D) Prognostic analysis of MIR4435-2HG and SMAD9. (E) PheWAS analysis for SMAD9 identifies significant associated phenotypes. *, P<0.05; **, P<0.01; ***, P<0.001; ns, not significant. AP, adenoma polyps; CA, conventional adenoma; CI, confidence interval; HR, hazard ratio; PheWAS, phenome-wide association study; SP, serrated polyps; SS, serrated-specific cells.

To investigate the impact of these genes on patient survival, we extracted survival information for all CRC patients from GTEx and TCGA databases, finding that patients with high expression of MIR4435-2HG had worse prognoses compared to those with lower expression, while those with high expression of SMAD9 had better outcomes (Figure 7C,7D). We also explored the prognoses of all other genes and plotted survival curves (Figure S3), further underscoring the importance of these key genes in CRC. We expanded our analysis to include all genes, such as SMAD9, for a comprehensive phenotype association study and to determine their possible effects on other bodily systems or different types of cancers (Figure 7E, Figure S4). Regrettably, no relevant SNPs were associated with MIR4435-2HG. Notably, the genes, including SMAD9, impacted CRC in our PheWAS analysis, indicating their broader implications beyond the initial scope of the study (Table S2).

Pathway differences associated with MIR4435-2HG and SMAD9 expression

To further characterize the functional relevance of MIR4435-2HG and SMAD9, we compared pathway activity between their high- and low-expressing cells within SSC and ASC, respectively. SSC cells with high MIR4435-2HG expression displayed significantly lower TGF-β signaling activity and modestly higher oxidative phosphorylation scores, whereas SMAD9-high ASC cells showed increased TGF-β signaling but no substantial change in oxidative phosphorylation (Figure 8A). We next assessed whether MIR4435-2HG and SMAD9 expression levels were associated with TGF-β pathway activity at the single-cell level. Both genes are statistically significant positive correlations with TGF-β signaling in their respective cell groups (SSC for MIR4435-2HG and ASC for SMAD9) (Figure 8B). These results suggest that the two risk-linked genes influence TGF-β-related signaling in distinct epithelial contexts.

Figure 8 Pathway differences associated with MIR4435-2HG and SMAD9 expression. (A) TGF-β and oxidative phosphorylation scores in MIR4435-2HG high/low SSC cells and SMAD9 high/low ASC cells. (B) Correlation between MIR4435-2HG or SMAD9 expression and TGF-β pathway activity in SSC and ASC. ASC, adenoma-specific cells; SSC, serrated specific cells.

Discussion

The debate over prognostic differences in transforming SP and traditional adenomas into CRC remains contentious. Herein, we proposed potential key genes and mechanisms that may contribute to the poorer prognosis associated with SP. In this study, we integrated single-cell and GWAS data to suggest that SPs are more likely to lead to carcinogenesis than CAs, primarily due to differences in the SSC and ABS cell populations. In this context, the TGF-β pathway exhibits opposing enrichment trends in SSC and ASC cells. Combining TWAS based on summary data and single-cell analysis identified six risk genes: PITPNC1, SMAD9, POU2AF2, LIMCH1, MIR4435-2HG, and HES6. Moreover, the pseudo-time analysis revealed that MIR4435-2HG, GDF15, and TGFBR2 exhibited divergent expression trends in the two types of tissues. Notably, SMAD9, identified by TWAS, is a part of the downstream pathway of TGFBR2. The ABS cell populations in the two types of polyps showed marked differences in the enrichment of the oxidative phosphorylation pathway. Further analysis demonstrated a causal relationship between genes like SMAD9 and mitochondrial oxidative phosphorylation. Additionally, we revalidated the expression of risk genes in RNA-seq and discovered that MIR4435-2HG and SMAD9 were associated with survival outcomes.

The carcinogenic transformation of SP, a relatively recently discovered pathway distinct from that of CA, exhibits significant differences in the location of CRC development and its molecular biological characteristics (37). Right-sided CRCs often develop from SPs, characterized by a predominance of BRAF mutations and high levels of CRC DNA methylation (38,39). In contrast, left-sided CRCs typically originate from CAs, frequently associated with KRAS mutations or TP53 inactivation (40,41). In the present study, we observed a downregulation of TGF-β in SSC cells and an opposite trend in ASC cells. Additionally, the GDF15-TGFBR2 pathway, significant in cell communication within SSCs, appeared absent, suggesting a potential inactivation of TGF-β factors during the carcinogenic process of SPs. The notion that TGF-β signaling distortions contribute to right-sided CRC development is not new. For instance, in their study, Leach and colleagues demonstrated that mice with BRAF mutations and a loss of TGF signaling were driven toward the development of right-sided CRC (42), which aligns with the observed patterns in human CRC, where TGF-β pathway aberrations are increasingly recognized as significant contributors to cancer progression. On the other hand, it is noteworthy that SPs often develop into microsatellite instability-high (MSI-H) CRCs, implying a high infiltration of immune cells in the tissue (43). These results provide evidence for the complex interplay between genetic aberrations and immune responses in the pathogenesis of CRC. TGF-β is often considered an immunosuppressive factor, suggesting that its downregulation in SSC cell populations could lead to an activated state of immune cells (44), which is consistent with the immune microenvironment of MSI-H CRCs, where heightened immune activity is a characteristic feature.

Our combined analysis of TWAS and scRNA-seq provided further biological interpretation for the six candidate risk genes identified in both SP and CA. MIR4435-2HG has been previously reported as a pan-cancer biomarker associated with poor prognosis in CRC (45), which aligns with our findings, further emphasizing its potential role in CRC pathogenesis and as a prognostic indicator. Other investigations into MIR4435-2HG suggest that it promotes the elevation of YAP1, thus increasing cell invasion and migration (46). Furthermore, its knockdown has been associated with induced sensitivity to cisplatin and enhanced cell apoptosis (47). Another study demonstrates that MIR4435-2HG is significantly overexpressed in CRC tissues, correlating with poor prognosis and resistance to oxaliplatin (L-OHP) based chemotherapy (48). In the present research, we integrated single-cell analysis, GWAS, and TWAS for the first time to elucidate the significance of MIR4435-2HG in the carcinogenic transformation of colorectal polyps into cancer, which may only activate in SSC cells within SP, a distinct difference from the lack of activation in ASC cells in CA. This suggests that this lncRNA might be used to assess the carcinogenic risk of different polyps, a conjecture that requires further experimental validation. Moreover, MIR4435-2HG has been reported to be associated with pathways involved in developing various tumors, including the TGF-β pathway. For instance, the MIR4435-2HG/miR-1224-5p/TGFBR2 axis has been observed in gliomas (49). Notably, some researchers consider MIR4435-2HG as an anti-cancer gene in mice whose absence is linked to immune suppression (41). Hence, the connection between its high expression in SP and MSI-H CRC is an area that requires further investigation by the research community. Our research also focused on SMAD9, another gene identified through multi-omics analysis. SMAD9, a receptor-regulated SMAD protein, is involved in the cooperative signal transduction of TGF-β type I and II receptors (50). There is limited reporting on the role of SMAD9/TGF-β receptors in CRC or intestinal polyps, though mutations in SMAD2, SMAD3, and SMAD4 have been documented in CRC (51). A rare report suggests a possible link between germline SMAD9 mutation and juvenile polyposis (52). A recent study on biomarkers and related molecular mechanisms shared by CRC and lung adenocarcinoma suggests that SMAD9 may contribute to the progression of both CRC and LUAD through its protein-binding functions (53). In our study, the TGF-β/SMAD9 signaling was significantly stronger in CA than in SP, potentially contributing to the differences observed between the two types of polyps. Among these risk genes, MIR4435-2HG and SMAD9 stood out because they were independently supported by multiple analytical layers—including scPagwas, TWAS and fine-mapping—and exhibited strong and opposite specificity for SSC and ASC cells, respectively. Their pathway associations with metabolic remodeling and TGF-β signaling further highlight their biological relevance to serrated and conventional adenoma progression.

Among the six genes identified in TWAS combined with sc-RNA analysis, only POU2AF2 was validated in SMR analysis, suggesting that the influence of its SNPs on phenotypes is more convincingly mediated by gene expression, particularly for POU2AF2. Notably, POU2AF2 is primarily expressed in TUF cell groups. These clustered cells, known for sensing intestinal contents, can signal the immune system, highlighting a possible functional role in intestinal health and disease mechanisms (54). POU2AF2, reported to modulate tuft cell differentiation and innate immunity in mice, may be linked to the carcinogenic transformation in intestinal polyps (55). Further research by Rajasekaran’s team confirmed that POU2AF2 serves as the primary transcriptional activator of tuft cells and exhibits tumor-suppressive effects in mouse models, suggesting that tuft cells play a critical tumor-protective role in the colonic epithelium (56). Regarding PITPNC1, LIMCH1, and HES6, we found no direct evidence of their impact on precancerous lesions in CRC, which warrants further investigation. Overall, the candidate genes identified in our study could provide new insights into the differing etiologies of the two types of precancerous polyps. Genetic regulatory variations are of essential importance in targeting gene expression, subsequently altering protein levels.

The management and monitoring of SP and CA are crucial for preventing CRC. Current guidelines support the same surveillance interval after endoscopic resection for both SP and CA (57,58). In a prospective study, Georgios Polychronidis and colleagues demonstrated through multivariate Cox proportional hazards regression that patients with high-risk polyp findings were at significantly higher risk of CRC and high-risk polyps, and may therefore benefit from early surveillance within 3 years (59). However, there remains a lack of lesion-specific preoperative surveillance strategies tailored to these distinct precancerous conditions (60). By integrating genetic association signals with cell-type-resolved single-cell transcriptomics, our multi-omics framework overcomes the limitations of single-omics approaches by linking germline risk variants to specific epithelial states and functional pathways, thereby providing biologically grounded markers for lesion-specific preoperative risk assessment. In addition to delineating the biological divergence between serrated and conventional pathways, our findings provide specific directions for potential clinical translation. The gene and cell-state signatures associated with SSC- and ASC-like populations offer lesion-type-specific molecular markers that may, after further validation, aid in the identification of high-risk serrated lesions—an area where routine colonoscopy has well-recognized limitations due to the subtle morphology and frequent miss rates of these lesions. These molecular features also represent promising candidates for developing adjunct stool- or blood-based assays, which could complement current screening tools by improving preoperative risk stratification and reducing misclassification based solely on histology. Although our findings are not yet ready for direct clinical application, the multi-omics framework provides a mechanistic basis and a practical roadmap for enhancing existing diagnostic and surveillance strategies through cell-type-resolved molecular information (61).

There are three key limitations in the present study. Primarily, TWAS concentrates on cis-eQTLs with significant cis-genetic influence, inadvertently neglecting the impactful distant regulatory roles of trans-eQTLs. Secondly, our reliance on reference data predominantly from European populations may constrain the broader applicability of our results across diverse ethnicities. Lastly, our focus was limited to SP and CA polyps, excluding inflammatory or hyperplastic variants, a choice dictated by data availability. More extensive research and refined methodologies are warranted to address these gaps.

Our study employed a multi-omics approach to thoroughly analyze the differences in CRC risk between SP and CA, identifying key genes and relevant cells. These findings enhance our understanding of the carcinogenic transformation in intestinal polyps, laying a foundation for further research into the specific functions of these genes. Moreover, these findings are important for postoperative clinical management and surveillance.

Conclusions

By integrating scRNA-seq with GWAS, TWAS, fine-mapping and MR, we identified SSC and ASC as two transcriptionally and genetically distinct epithelial states in SP. SSC cells showed stronger enrichment of CRC genetic risk and displayed unique pathway features compared with ASC, particularly in TGF-β and metabolic programs. Key genes MIR4435-2HG and SMAD9 further exhibited cell type-specific associations with these pathways, highlighting divergent regulatory roles in SSC and ASC. These findings link genetic susceptibility to gene function and pathway activity at single-cell resolution. Overall, the study provides a clearer mechanistic framework for how distinct epithelial trajectories contribute to SP carcinogenesis.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STREGA reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2138/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2138/prf

Funding: This work was supported by the Health Commission of Guangdong Province and the Directive Project of the Guangdong Medical Research Foundation (C2024080).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-2138/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. The global, regional, and national burden of colorectal cancer and its attributable risk factors in 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet Gastroenterol Hepatol 2019;4:913-33. [Crossref] [PubMed]
2. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
3. Kotula AE, Korde YV, Oyler HJ, et al. Colorectal polyps: pathophysiology, malignant potential, and advancements in therapeutic strategies. Med Oncol 2025;42:287. [Crossref] [PubMed]
4. Nguyen LH, Goel A, Chung DC. Pathways of Colorectal Carcinogenesis. Gastroenterology 2020;158:291-302. [Crossref] [PubMed]
5. Li J, You L, Xu Z, et al. Healthy lifestyle and the risk of conventional adenomas and serrated polyps: Findings from a large colonoscopy screening population. Int J Cancer 2022;151:67-76. [Crossref] [PubMed]
6. Dunne PD, Arends MJ. Molecular pathological classification of colorectal cancer-an update. Virchows Arch 2024;484:273-85. [Crossref] [PubMed]
7. van Toledo DEFWM, IJspeert JEG, Dekker E. Current Approaches in Managing Colonic Serrated Polyps and Serrated Polyposis. Annu Rev Med 2022;73:293-306. [Crossref] [PubMed]
8. Sweetser S, Smyrk TC, Sinicrope FA. Serrated colon polyps as precursors to colorectal cancer. Clin Gastroenterol Hepatol 2013;11:760-7; quiz e54-5. [Crossref] [PubMed]
9. Meester RGS, Anderson JC. Long-Term Risk for Colorectal Cancer in Patients With Index Serrated Polyps. Gastroenterology 2022;162:2108-10. [Crossref] [PubMed]
10. Edelstein DL, Axilbund JE, Hylind LM, et al. Serrated polyposis: rapid and relentless development of colorectal neoplasia. Gut 2013;62:404-8. [Crossref] [PubMed]
11. Helsingen LM, Kalager M. Colorectal Cancer Screening - Approach, Evidence, and Future Directions. NEJM Evid 2022;1:EVIDra2100035.
12. Hagen R, Srivastava A, Anderson JC. The serrated pathway and colorectal cancer: what the gastroenterologist should know. Expert Rev Gastroenterol Hepatol 2025;19:593-606. [Crossref] [PubMed]
13. Song M, Emilsson L, Bozorg SR, et al. Risk of colorectal cancer incidence and mortality after polypectomy: a Swedish record-linkage study. Lancet Gastroenterol Hepatol 2020;5:537-47. [Crossref] [PubMed]
14. Crockett SD. Sessile Serrated Polyps and Colorectal Cancer. JAMA 2017;317:975-6. [Crossref] [PubMed]
15. Spadaccini M, Menini M, Massimi D, et al. AI and Polyp Detection During Colonoscopy. Cancers (Basel) 2025;17:797. [Crossref] [PubMed]
16. Cheng TH, Gorman M, Martin L, et al. Common colorectal cancer risk alleles contribute to the multiple colorectal adenoma phenotype, but do not influence colonic polyposis in FAP. Eur J Hum Genet 2015;23:260-3. [Crossref] [PubMed]
17. Abulí A, Castells A, Bujanda L, et al. Genetic Variants Associated with Colorectal Adenoma Susceptibility. PLoS One 2016;11:e0153084. [Crossref] [PubMed]
18. Edwards SL, Beesley J, French JD, et al. Beyond GWASs: illuminating the dark road from association to function. Am J Hum Genet 2013;93:779-97. [Crossref] [PubMed]
19. Hang D, Joshi AD, He X, et al. Colorectal cancer susceptibility variants and risk of conventional adenomas and serrated polyps: results from three cohort studies. Int J Epidemiol 2020;49:259-69. [Crossref] [PubMed]
20. Burnett-Hartman AN, Newcomb PA, Hutter CM, et al. Variation in the association between colorectal cancer susceptibility loci and colorectal polyps by polyp type. Am J Epidemiol 2014;180:223-32. [Crossref] [PubMed]
21. Wang J, Zhang Y, Chen X, et al. Single-Cell Transcriptomics Reveals Cellular Heterogeneity and Drivers in Serrated Pathway-Driven Colorectal Cancer Progression. Int J Mol Sci 2024;25:10944. [Crossref] [PubMed]
22. Chen B, Scurrah CR, McKinley ET, et al. Differential pre-malignant programs and microenvironment chart distinct paths to malignancy in human colorectal polyps. Cell 2021;184:6262-6280.e26. [Crossref] [PubMed]
23. Wainberg M, Sinnott-Armstrong N, Mancuso N, et al. Opportunities and challenges for transcriptome-wide association studies. Nat Genet 2019;51:592-9. [Crossref] [PubMed]
24. Guo X, Lin W, Wen W, et al. Identifying Novel Susceptibility Genes for Colorectal Cancer Risk From a Transcriptome-Wide Association Study of 125,478 Subjects. Gastroenterology 2021;160:1164-1178.e6. [Crossref] [PubMed]
25. Hazelwood E, Canson DM, Deslandes B, et al. Multi-tissue expression and splicing data prioritise anatomical subsite- and sex-specific colorectal cancer susceptibility genes. Nat Commun 2025;16:5043. [Crossref] [PubMed]
26. Tsumuraya H, Okayama H, Katagata M, et al. TGFβ-Responsive Stromal Activation Occurs Early in Serrated Colorectal Carcinogenesis. Int J Mol Sci 2024;25:4626. [Crossref] [PubMed]
27. Huyghe JR, Bien SA, Harrison TA, et al. Discovery of common and rare genetic risk variants for colorectal cancer. Nat Genet 2019;51:76-87. [Crossref] [PubMed]
28. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 2020;369:1318-30. [Crossref] [PubMed]
29. Ma Y, Deng C, Zhou Y, et al. Polygenic regression uncovers trait-relevant cellular contexts through pathway activation transformation of single-cell RNA sequencing data. Cell Genom 2023;3:100383. [Crossref] [PubMed]
30. Hao Y, Stuart T, Kowalski MH, et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol 2024;42:293-304. [Crossref] [PubMed]
31. Jin S, Plikus MV, Nie Q. CellChat for systematic analysis of cell-cell communication from single-cell transcriptomics. Nat Protoc 2025;20:180-219. [Crossref] [PubMed]
32. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 2005;102:15545-50. [Crossref] [PubMed]
33. Trapnell C, Cacchiarelli D, Grimsby J, et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol 2014;32:381-6. [Crossref] [PubMed]
34. Dai Q, Zhou G, Zhao H, et al. OTTERS: a powerful TWAS framework leveraging summary-level reference data. Nat Commun 2023;14:1271. [Crossref] [PubMed]
35. Zhu Z, Zhang F, Hu H, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat Genet 2016;48:481-7. [Crossref] [PubMed]
36. Zhou W, Nielsen JB, Fritsche LG, et al. Efficiently controlling for case-control imbalance and sample relatedness in large-scale genetic association studies. Nat Genet 2018;50:1335-41. [Crossref] [PubMed]
37. Zhou YJ, Lu XF, Chen H, et al. Single-cell Transcriptomics Reveals Early Molecular and Immune Alterations Underlying the Serrated Neoplasia Pathway Toward Colorectal Cancer. Cell Mol Gastroenterol Hepatol 2023;15:393-424. [Crossref] [PubMed]
38. Yang JF, Tang SJ, Lash RH, et al. Anatomic distribution of sessile serrated adenoma/polyp with and without cytologic dysplasia. Arch Pathol Lab Med 2015;139:388-93. [Crossref] [PubMed]
39. Kambara T, Simms LA, Whitehall VL, et al. BRAF mutation is associated with DNA methylation in serrated polyps and cancers of the colorectum. Gut 2004;53:1137-44. [Crossref] [PubMed]
40. Hazewinkel Y, de Wijkerslooth TR, Stoop EM, et al. Prevalence of serrated polyps and association with synchronous advanced neoplasia in screening colonoscopy. Endoscopy 2014;46:219-24. [Crossref] [PubMed]
41. Yu H, Chen C, Han F, et al. Long Noncoding RNA MIR4435-2HG Suppresses Colorectal Cancer Initiation and Progression By Reprogramming Neutrophils. Cancer Immunol Res 2022;10:1095-110. [Crossref] [PubMed]
42. Leach JDG, Vlahov N, Tsantoulis P, et al. Oncogenic BRAF, unrestrained by TGFβ-receptor signalling, drives right-sided colonic tumorigenesis. Nat Commun 2021;12:3464. [Crossref] [PubMed]
43. Marisa L, de Reyniès A, Duval A, et al. Gene expression classification of colon cancer into molecular subtypes: characterization, validation, and prognostic value. PLoS Med 2013;10:e1001453. [Crossref] [PubMed]
44. Nakanishi Y, Duran A, L'Hermitte A, et al. Simultaneous Loss of Both Atypical Protein Kinase C Genes in the Intestinal Epithelium Drives Serrated Intestinal Cancer by Impairing Immunosurveillance. Immunity 2018;49:1132-1147.e7. [Crossref] [PubMed]
45. Zhong C, Xie Z, Zeng LH, et al. MIR4435-2HG Is a Potential Pan-Cancer Biomarker for Diagnosis and Prognosis. Front Immunol 2022;13:855078. [Crossref] [PubMed]
46. Zanconato F, Cordenonsi M, Piccolo S. YAP/TAZ at the Roots of Cancer. Cancer Cell 2016;29:783-803. [Crossref] [PubMed]
47. Barrera G, Cucci MA, Grattarola M, et al. Control of Oxidative Stress in Cancer Chemoresistance: Spotlight on Nrf2 Role. Antioxidants (Basel) 2021;10:510. [Crossref] [PubMed]
48. Zhuang X, Jin J, Cheng J, et al. P65-Driven MIR4435-2HG Enhances Prognostic Value and Mediates Oxaliplatin Resistance via the miR-378G/ABCB9 Axis in Colorectal Cancer. Macromol Biosci 2025;25:e00663. [Crossref] [PubMed]
49. Xu H, Zhang B, Yang Y, et al. LncRNA MIR4435-2HG potentiates the proliferation and invasion of glioblastoma cells via modulating miR-1224-5p/TGFBR2 axis. J Cell Mol Med 2020;24:6362-72. [Crossref] [PubMed]
50. Fagerberg L, Hallström BM, Oksvold P, et al. Analysis of the human tissue-specific expression by genome-wide integration of transcriptomics and antibody-based proteomics. Mol Cell Proteomics 2014;13:397-406. [Crossref] [PubMed]
51. Fleming NI, Jorissen RN, Mouradov D, et al. SMAD2, SMAD3 and SMAD4 mutations in colorectal cancer. Cancer Res 2013;73:725-35. [Crossref] [PubMed]
52. Ngeow J, Yu W, Yehia L, et al. Exome Sequencing Reveals Germline SMAD9 Mutation That Reduces Phosphatase and Tensin Homolog Expression and Is Associated With Hamartomatous Polyposis and Gastrointestinal Ganglioneuromas. Gastroenterology 2015;149:886-9.e5. [Crossref] [PubMed]
53. Gao K, Yu L, Wang L, et al. Investigation of biomarkers and associated molecular mechanism shared between colorectal cancer and lung adenocarcinoma. Discov Oncol 2025;16:1540. [Crossref] [PubMed]
54. Burclaff J, Bliton RJ, Breau KA, et al. A Proximal-to-Distal Survey of Healthy Adult Human Small Intestine and Colon Epithelium by Single-Cell Transcriptomics. Cell Mol Gastroenterol Hepatol 2022;13:1554-89. [Crossref] [PubMed]
55. Wang YX, Zhou CP, Wang DT, et al. Unraveling the causal role of immune cells in gastrointestinal tract cancers: insights from a Mendelian randomization study. Front Immunol 2024;15:1343512. [Crossref] [PubMed]
56. Rajasekaran V, Harris BT, Osborn RT, et al. Genetic variation at 11q23.1 confers colorectal cancer risk by dysregulation of colonic tuft cell transcriptional activator POU2AF2. Gut 2025;74:787-803. [Crossref] [PubMed]
57. Gupta S, Lieberman D, Anderson JC, et al. Recommendations for Follow-Up After Colonoscopy and Polypectomy: A Consensus Update by the US Multi-Society Task Force on Colorectal Cancer. Am J Gastroenterol 2020;115:415-34. [Crossref] [PubMed]
58. Rutter MD, East J, Rees CJ, et al. British Society of Gastroenterology/Association of Coloproctology of Great Britain and Ireland/Public Health England post-polypectomy and post-colorectal cancer resection surveillance guidelines. Gut 2020;69:201-23. [Crossref] [PubMed]
59. Polychronidis G, He MM, Vithayathil M, et al. Risk of colorectal neoplasia after removal of conventional adenomas and serrated polyps: a comprehensive evaluation of risk factors and surveillance use. Gut 2024;73:1675-83. [Crossref] [PubMed]
60. Jung YS, Park JH, Park CH. Serrated Polyps and the Risk of Metachronous Colorectal Advanced Neoplasia: A Systematic Review and Meta-Analysis. Clin Gastroenterol Hepatol 2022;20:31-43.e1. [Crossref] [PubMed]
61. Khalaf N, Samadder NJ. Utility of Genetic Testing in Persons With Multiple Colorectal Polyps. Clin Gastroenterol Hepatol 2019;17:1942-4. [Crossref] [PubMed]