Abstract
Colorectal cancer (CRC) is an aggressive malignancy with a poor prognosis. Recent research has focused on the impact of dietary restriction (DR) and circadian rhythm (CR) on cancer; however, the roles of DR-related genes (DRRGs) and CR-related genes (CRRGs) in CRC remain poorly understood. Bulk RNA and single-cell sequencing data for CRC were retrieved from public databases, while DRRGs and CRRGs were compiled from existing literature. Prognostic genes were identified through univariate Cox regression and machine learning approaches. A prognostic risk model and a clinically applicable nomogram were developed. Differences in functional pathways, tumor microenvironment characteristics, and drug sensitivities were analyzed between the high-risk cohort (HRC) and the low-risk cohort (LRC). The expression of prognostic genes was assessed at the single-cell level and validated in clinical samples. Five prognostic genes, namely SOCS3, MMP3, SLC3A1, FABP4, and SERPINE1, were identified. Both the model and nomogram demonstrated strong predictive power. LRC was primarily associated with DNA replication pathways, whereas HRC exhibited higher ESTIMATE, immune, stromal, and TIDE scores, with significant differences across 24 immune cell types. Sensitivity to 61 drugs was observed in HRC, while 56 drugs were sensitive to LRC. Single-cell analysis highlighted fibroblasts as key cells, with dynamic expression of FABP4 and SOCS3 during differentiation. With the exception of SERPINE1, prognostic genes showed differential expression in clinical samples. The risk model effectively predicts CRC prognosis and emphasizes the prognostic relevance of DRRGs and CRRGs, offering novel insights for personalized CRC therapy.
Introduction
Colorectal cancer (CRC) is a highly heterogeneous and deadly malignancy. Emerging evidence shows that lifestyle factors, specifically dietary restriction (DR) and circadian rhythm (CR), play significant roles in cancer development by regulating metabolism, cellular repair, and gut microbiota. However, the specific genetic crosstalk between DR and CR pathways in CRC remains poorly understood. Previous multi-omics studies have struggled to connect tissue-level survival data with cell-specific mechanisms. This study bridges that gap by integrating bulk RNA sequencing with single-cell RNA sequencing (scRNA-seq) to identify specific DR- and CR-related genes that drive CRC and build a clinically actionable predictive model.
Methods
The authors utilized a hybrid bioinformatics and experimental pipeline. They began by intersecting literature-derived DR- and CR-related genes to identify common regulatory elements.
These genes were then intersected with Differentially Expressed Genes (DEGs) found in CRC vs. normal tissues using bulk RNA-seq data from the GEO (GSE39582) and TCGA-CRC datasets. Using univariate Cox and LASSO regression, the candidate genes were narrowed down to a 5-gene risk signature. The team also applied scRNA-seq clustering and pseudo-time trajectory analysis to dataset GSE161277 to identify cell-specific expression patterns. Finally, the findings were validated using RT-qPCR on clinical CRC and adjacent normal tissue samples.
Key Insights and Ideas
- The 5-Gene Signature: The study isolated SOCS3, MMP3, SLC3A1, FABP4, and SERPINE1 as core prognostic genes linking lifestyle-related metabolic pathways to CRC survival.
- Fibroblasts as Primary Drivers: scRNA-seq revealed that tumor-associated fibroblasts are the primary cells expressing these prognostic genes. Pseudo-time analysis showed that as fibroblasts differentiate, FABP4 and SOCS3 expression dynamically shifts, suggesting these genes actively shape the tumor's physical and immune environment.
- High Risk and Immune Evasion: Patients in the high-risk cohort exhibited significantly higher TIDE scores, indicating a greater capacity for immune evasion. They also showed higher stromal scores, suggesting a dense physical barrier around the tumor.
- Rewired Cellular Communication: In CRC tissues, fibroblasts were found to have reduced communication with immune cells compared to healthy tissue, contributing to an immunosuppressive environment.
Key Results
The risk model successfully stratified patients, with the high-risk cohort showing significantly poorer overall survival. Initial clustering based on identified genes also demonstrated clear survival differences between patient groups.
- Predictive Accuracy: The combined nomogram (Risk Score + M-stage) achieved solid predictive accuracy, with AUC values of 0.72 and 0.72 for 1- and 3-year survival, respectively.
- Therapeutic Targets: The study identified 61 drugs specifically effective for high-risk patients (e.g., the dual mTOR inhibitor AZD8055) and 56 for low-risk patients.
- Experimental Validation: RT-qPCR successfully validated the differential expression of 4 out of the 5 genes (FABP4, SLC3A1, MMP3, SOCS3) in clinical samples.
Limitations and Future Work
The study's experimental validation was limited to a small cohort of 5 patient sample pairs. While the bioinformatic links are strong, the model relies on public databases and requires validation in larger, independent, multi-center clinical cohorts. Future research should focus on the biological mechanisms of these 5 genes in CRC progression through in vitro and in vivo experiments.
Summary updated