Multi-omics analysis reveals that FABP2 suppresses colorectal cancer progression by regulating sphingomyelin metabolism via the PPAR signaling pathway

Abstract

Background

Colorectal cancer (CRC) is a highly prevalent and lethal digestive - tract malignancy. Its global incidence and mortality are on the rise. Understanding the underlying mechanisms of CRC development is crucial for improving clinical diagnosis and treatment.

Methods

The study employed Mendelian Randomization (MR) analysis, considering plasma proteins, metabolites, and tissue genes as exposure factors and CRC risk as the outcome. Key molecules were identified, and mediation MR explored their relationships. Bioinformatics analyses, using CRC Bulk RNA-Seq and scRNA-Seq data, validated the expression and functions of key molecules. Finally, the expression and regulatory mechanisms of the identified proteins and metabolites were validated using clinical tissue specimens and cell-mouse experiments.

Results

MR analysis showed 16 plasma proteins and 87 metabolites had causal relationships with CRC. FABP2, a plasma protein, was negatively correlated with CRC risk, while sphingomyelin was positively associated. Mediation MR indicated sphingomyelin mediated FABP2’s protective effect. Bulk RNA-seq data showed that FABP2 expression was significantly lower in CRC tissues compared to normal tissues. GSEA enrichment and PPI analyses revealed that FABP2 is closely associated with lipid metabolic processes. scRNA-seq further revealed that FABP2 is primarily expressed in epithelial cells, and its expression was markedly reduced in epithelial cells from tumor tissues relative to adjacent normal tissues. Clinical validation confirmed that FABP2 expression was significantly downregulated in tumor tissues and CRC patient plasma, whereas sphingomyelin levels were significantly elevated. Finally, vitro and vivo experiments combined with RNA-Seq demonstrated that FABP2 may regulate sphingomyelin metabolism and modulate CRC progression via the PPAR signaling pathway.

Conclusion

FABP2 and sphingomyelin are closely related to CRC development. FABP2 may inhibit CRC progression by regulating sphingomyelin metabolism through the PPAR signaling pathway, suggesting its potential as a candidate biomarker for CRC diagnosis and a potential target for therapeutic intervention.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12935-025-04159-1.

Background

Colorectal cancer (CRC) is among the most common and life-threatening gastrointestinal malignancies worldwide. According to the 2024 Global Cancer Observatory (GLOBOCAN) data, approximately 1.93 million new CRC cases were diagnosed in 2022, accounting for about 10% of all new cancer cases, and 904,000 deaths were attributed to CRC, making it the second leading cause of cancer-related mortality [1]. The median age at CRC onset is 67 years, and nearly 10% of patients are younger than 50 years [2]. CRC incidence is generally higher in developed countries, and its global incidence is rapidly increasing [3, 4]. Although advances in screening and treatment have improved early diagnosis and clinical outcomes, the overall survival of CRC patients remains unsatisfactory [1]. It is estimated that up to 50% of CRC cases may be preventable through early intervention [5]. Identifying reliable risk factors and optimizing prevention and treatment strategies remain major challenges, underscoring the need for in-depth research into early screening, precision prevention, and the mechanisms of tumor initiation and progression.

With rapid progress in molecular biology and metabolomics, the roles of plasma proteins and metabolites in CRC have attracted increasing attention. As indicators of systemic metabolic status, plasma proteins and metabolites can be used not only for early disease diagnosis but also as prognostic and therapeutic monitoring biomarkers [6–8]. Numerous studies have shown that specific plasma proteins and metabolites are closely associated with CRC occurrence and progression [9–11]. Metabolomics enables comprehensive characterization of tumor-specific metabolic reprogramming, thereby providing new opportunities for developing therapies targeting metabolic pathways. However, despite accumulating evidence implicating certain plasma proteins and metabolites in CRC, their precise causal relationships and translational potential remain to be fully elucidated.

Mendelian randomization (MR) is a causal inference approach that leverages genetic variants as instrumental variables to assess the causal effect of an exposure on an outcome [12–14]. By using germline genetic variants that are randomly allocated at conception, MR can largely circumvent confounding and reverse causality that commonly affect observational epidemiologic studies. In recent years, MR has been widely applied to investigate complex diseases [15, 16], and its application in CRC research has highlighted several novel risk factors and protective factors [17, 18]. These findings deepen our understanding of CRC etiology and provide a theoretical basis for risk prediction and personalized therapy. However, MR depends on high-quality genome-wide association study (GWAS) data, and the selection of valid instrumental variables is critical for robust results. Integrating MR with functional validation is therefore essential to derive comprehensive and reliable causal inferences.

In this study, we used a multi-omics MR framework to systematically evaluate causal relationships between plasma proteins, plasma metabolites, colon tissue gene expression, and CRC risk. We identified FABP2 as a key protective factor and sphingomyelin as a metabolite mediating the effect of FABP2 on CRC. Bioinformatics analyses using bulk RNA-seq and scRNA-seq datasets were conducted to verify the expression profiles and function of FABP2 in CRC. Furthermore, clinical tissue and plasma samples, together with in vitro and in vivo experiments and RNA-seq, were used to validate the expression patterns and dissect the mechanistic role of FABP2 and sphingomyelin in CRC. This integrated multi-omics and experimental approach supports the existence of a FABP2–PPAR–sphingomyelin axis in CRC and suggests that this pathway may provide a mechanistic basis for future preventive and therapeutic strategies.

Methods

Study design

The overall study design is summarized in (Fig. 1). Plasma proteins, plasma metabolites, and colon tissue gene expression were considered as exposures, and CRC risk was treated as the outcome. Bidirectional Mendelian randomization analyses were performed to evaluate the causal effects of these exposures on CRC risk and to exclude reverse causality. The causal associations between plasma proteins and CRC identified in the discovery GWAS cohort were subsequently validated in an independent GWAS cohort. For plasma proteins with robust causal associations with CRC, corresponding tissue expression quantitative trait loci (eQTLs) were used as exposures in additional MR analyses to clarify gene-level effects. This stepwise approach led to the identification of FABP2 as a key candidate molecule.

Fig. 1.

Study design of the research CRC (Colorectal cancer). The workflow illustrates how FABP2 was identified as a key molecule through Mendelian randomization (MR) analyses of plasma proteins, plasma metabolites, and colon tissue expression quantitative trait loci (eQTLs), followed by validation using bulk and single-cell RNA-seq–based bioinformatics, clinical tissue and plasma samples, and in vitro and in vivo functional and mechanistic experiments. Together, the workflow outlines a comprehensive, multi-stage strategy to identify and validate FABP2 as a causal protective factor in CRC

Mediation MR was then applied to integrate plasma FABP2 levels, plasma metabolites, and CRC risk, with the aim of identifying metabolites mediating the effect of FABP2 on CRC. Bioinformatics analyses based on bulk RNA-seq and scRNA-seq data were used to validate the expression pattern and potential biological functions of FABP2 in CRC. Finally, clinical tissue and plasma samples, together with in vitro and in vivo experiments and RNA-seq, were used to confirm the expression changes and regulatory mechanisms of FABP2 and sphingomyelin.

Mendelian randomization analysis

Our MR analyses rely on the standard instrumental variable assumptions: (1) relevance, the selected SNPs are robustly associated with the exposure; (2) independence, the SNPs are independent of confounders of the exposure–outcome relationship; and (3) exclusion restriction, the SNPs influence CRC risk only through the exposure and not via alternative pathways. These assumptions were supported by stringent instrument selection, assessment of F-statistics, and pleiotropy-robust sensitivity analyses.

Data sources for plasma proteome and genetic instrument selection

GWAS summary statistics for 1,001 plasma proteins were obtained from the genomic atlas of the human plasma proteome reported by Karsten Suhre (10.1016/j.xgen.2024.100506). This study included 54,000 individuals of European ancestry, and the corresponding GWAS summary data are available in the GWAS Catalog (https://www.ebi.ac.uk/gwas/). To obtain a sufficient number of instrumental variables for each protein while maintaining statistical power, we selected single-nucleotide polymorphisms (SNPs) significantly associated with each plasma protein at P < 5 × 10⁻⁶. Linkage disequilibrium (LD) pruning was performed using r² < 0.001 and a physical distance threshold of 10,000 kb, yielding 12,113 SNPs corresponding to 1,001 plasma proteins.

Data sources for eQTLs and genetic instrument selection

Colon tissue eQTL GWAS summary statistics were downloaded from the GTEx Portal (https://www.gtexportal.org/home/). The GTEx v8 dataset includes 838 donors and 17,382 samples across 52 tissues and two cell lines. For the gene–CRC association analysis, we used precomputed eQTL summary statistics for colon tissue. eQTL effect estimates represent the change in gene expression (in SD units) associated with each additional allele. To ensure sufficient instrument strength for mRNA expression, SNPs significantly associated with each gene at P < 1 × 10⁻⁶ were selected as instruments. LD pruning was performed using r² < 0.01 and a 100-kb window.

Data sources for plasma metabolites and genetic instrument selection

GWAS summary statistics for plasma metabolites were obtained from the study by Yiheng Chen et al. (10.1038/s41588-022-01270-1), which included 1,091 blood metabolites and 309 metabolite ratios measured in 8,299 individuals of European ancestry. GWAS summary data for 1,400 plasma metabolites are publicly available in the GWAS Catalog (https://www.ebi.ac.uk/gwas/). To maintain adequate statistical power, SNPs associated with each plasma metabolite at P < 1 × 10⁻⁶ were selected as instruments, followed by LD pruning with r² < 0.001 and a 10,000-kb distance threshold.

Data sources for CRC and genetic instrument selection

GWAS summary statistics for CRC in the discovery cohort were obtained from the Finngen R11 dataset (https://storage.googleapis.com/finngen-public-data-r11/summary_stats/finngen_R11_CRC.gz), which includes 8,801 CRC cases and 345,118 controls. For validation, we used publicly available GWAS summary statistics from the IEU OpenGWAS project (https://gwas.mrcieu.ac.uk), specifically dataset ebi-a-GCST90018808, comprising 6,581 CRC cases and 462,421 controls. For reverse MR analysis, SNPs associated with CRC at P < 5 × 10⁻⁶ in the discovery cohort were selected as instruments, with LD pruning at r² < 0.001 and 10,000 kb.

Analysis methods

Five MR methods were used to estimate the causal effects of exposures on CRC risk: inverse variance weighted (IVW) as the primary method, and MR-Egger, weighted median, simple mode, and weighted mode as complementary methods. Causal effect estimates were reported as odds ratios (ORs) with 95% confidence intervals (CIs).

Horizontal pleiotropy was assessed using MR-Egger regression and MR-PRESSO, with P > 0.05 indicating no evidence of pleiotropy. Heterogeneity was evaluated using Cochran’s Q statistic for both IVW and MR-Egger; P > 0.05 indicated no heterogeneity. Leave-one-out sensitivity analyses were performed to assess whether the overall causal estimate was driven by any single SNP.

For plasma protein–CRC associations in the discovery and validation cohorts, P values were adjusted using the Benjamini–Hochberg method, with a false discovery rate (FDR) < 0.05 considered statistically significant. For analyses of plasma metabolites and eQTL–CRC associations, P < 0.05 was regarded as statistically significant.

For mediation MR, we defined β1 as the total causal effect of FABP2 on CRC, β2 as the causal effect of FABP2 on sphingomyelin, and β3 as the causal effect of sphingomyelin on CRC. The mediating effect was calculated as β2 × β3, and the proportion mediated was given by (β2 × β3)/β1.

Bioinformatics analysis

Data sources and processing

RNA-seq data for CRC tumor tissues and matched normal controls were downloaded from The Cancer Genome Atlas (TCGA) (https://www.cancer.gov/ccg/research/genome-sequencing/tcga). This dataset includes 455 tumor samples and 41 normal tissues. scRNA-seq data for three pairs of CRC tissues and corresponding normal tissues were obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) under accession GSE231559.

Gene set enrichment analysis (GSEA)

To explore the functional implications of FABP2, we performed KEGG pathway-based GSEA using the R package “clusterProfiler.” Pathways were considered significantly enriched if P < 0.05 and q < 0.25.

Protein–protein interaction network (PPI)

GeneMANIA (https://genemania.org/) is an online platform for gene function prediction and gene association network construction. It integrates multiple data sources, including gene coexpression, protein–protein interactions, gene co-localization, gene co-evolution, gene annotations, and pathway data. We used GeneMANIA to identify the top 20 proteins interacting with FABP2 and to construct a PPI network, thereby predicting the potential functions of FABP2 and its interactors. Query is scheduled for 24 November 2025, Application version: 3.6.0.

Single-cell transcriptomic analysis

scRNA-seq data for three pairs of CRC tissues and matched normal tissues (GSE231559) were processed using the “Seurat” R package. Cells were filtered using the following criteria: (1) gene count < 200 or > 40,000; (2) mitochondrial gene proportion > 20% were excluded. Remaining cells were designated as core cells. Highly variable genes were identified using variance analysis among core cells. Gene expression data were normalized by linear regression, followed by principal component analysis (PCA). The top 20 principal components were used for visualization by uniform manifold approximation and projection (UMAP). The “FindAllMarkers” function was used to identify marker genes for each cluster.

Clinical specimen validation and in vitro and in vivo experiments

Clinical specimens

CRC tissue and plasma samples were obtained from the Digestive Disease Biobank of the First Affiliated Hospital of the Fourth Military Medical University (Xijing Hospital). All tissue samples underwent pathological examination, and histological features were confirmed by experienced pathologists. The study was approved by the Ethics Committee of the First Affiliated Hospital of the Fourth Military Medical University, and informed consent was obtained from all participants.

Immunohistochemistry

Resected tissue specimens were fixed, paraffin-embedded, and sectioned. After deparaffinization and rehydration, antigen retrieval was performed, followed by blocking with goat serum. Sections were incubated overnight at 4 °C with primary antibody against FABP2 (Affinity, Jiangsu, China, DF6508). After incubation with horseradish peroxidase (HRP)-conjugated secondary antibody, staining was developed with 3,3’-diaminobenzidine (DAB) and counterstained with hematoxylin. Slides were dehydrated, cleared, mounted, and examined under a light microscope.

Western blotting

Total protein from tissues or cells was extracted using radioimmunoprecipitation assay (RIPA) buffer, and protein concentrations were determined using the bicinchoninic acid (BCA) method. Equal amounts of protein were separated by 12% SDS–polyacrylamide gel electrophoresis and transferred onto nitrocellulose (NC) membranes. Membranes were blocked with 5% skim milk for 1 h at room temperature and incubated overnight at 4 °C with primary antibodies against FABP2 (DF6508, Affinity), PPAR-α (AF5301, Affinity), ACOX1 (10957-1-AP, Proteintech, Wuhan, China), and CPT1A (15184-1-AP, Proteintech). After washing with TBST, membranes were incubated with appropriate HRP-conjugated secondary antibodies and developed using a Bio-Rad imaging system. β-Actin served as the loading control. Band intensities were quantified using Image Lab software.

qPCR

Total RNA from tissues or cells was extracted using a commercial RNA extraction kit, and cDNA was synthesized using a Toyobo reverse transcription kit. Gene expression was quantified by SYBR Green-based qPCR. Cycling conditions were: 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s, 60 °C for 15 s, and 72 °C for 20 s. GAPDH was used as the internal control. Relative expression levels were calculated by the 2^−ΔΔCt method.

FABP2 forward: ATGGCGTTTGACAGCACTTG; reverse: TCAGTTCCGTCTGCTAGATTGTA.

PPAR-α forward: TTCGCAATCCATCGGCGAG; reverse: CCACAGGATAAGTCACCGAGG.

ACOX1 forward: ACTCGCAGCCAGCGTTATG; reverse: AGGGTCAGCGATGCCAAAC.

CPT1A forward: TCCAGTTGGCTTATCGTGGTG; reverse: TCCAGAGTCCGATTGATTTTTGC.

GAPDH forward: CTGGGCTACACTGAGCACC; reverse: AAGTGGTCGTTGAGGGCAATG.

Sphingomyelin and plasma FABP2 detection

Quantified tissue and plasma samples were collected, and sphingomyelin levels were measured using a sphingomyelin assay kit (Elabscience, Wuhan, China, E-BC-F079) according to the manufacturer’s instructions. Plasma FABP2 levels were determined using a commercial FABP2 ELISA kit (Elabscience, Wuhan, China, E-EL-H0159).

Cell culture

Human CRC cell lines HCT116 and HCT8 were obtained from the State Key Laboratory of Holistic Integrative Management of Gastrointestinal Cancers (Shaanxi, China) and stored in liquid nitrogen. Cells were cultured in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin at 37 °C in a humidified atmosphere with 5% CO₂. All cell lines were authenticated by short tandem repeat (STR) profiling within the past three years, and all experiments were performed with mycoplasma-free cells.

Cell transfection

Cells in the logarithmic growth phase were harvested and seeded to achieve approximately 30% confluence after 24 h. Overexpression plasmids were prepared according to the manufacturer’s instructions and transfected into cells. After 24 h, cells were collected for RNA and protein extraction to assess transfection efficiency by qPCR and Western blotting.

CCK-8 assay

Cell proliferation was assessed using the CCK-8 assay. Cells (1 × 10³ per well) were seeded into 96-well plates with 200 µL of culture medium. After 24 h, the medium was replaced with a mixture of CCK-8 reagent and culture medium and incubated for 2 h in the dark. Optical density (OD) at 490 nm was measured using a microplate reader.

Transwell assay

For migration and invasion assays, Transwell chambers were used. For invasion assays, Matrigel was diluted in serum-free medium and added to the upper chamber membrane, then incubated at 37 °C for 2 h to solidify. A total of 4 × 10⁴ cells in 500 µL of serum-free medium were added to the upper chamber, and 800 µL of medium containing 10% FBS was placed in the lower chamber. After 24 h of incubation, cells on the membrane were fixed with 4% paraformaldehyde for 20 min and stained with crystal violet for 15 min. Images were acquired with an inverted microscope (Olympus; 400×), and migrated or invaded cells were counted using ImageJ.

Plate clonogenicity assay

Cells (1 × 10³ per well) were seeded into 6-well plates and cultured in medium containing 10% FBS at 37 °C and 5% CO₂ for 14 days. Colonies were then fixed with methanol and stained with 0.1% crystal violet. Colony numbers were counted using ImageJ.

RNA sequencing analysis

Total RNA was extracted from cells using TRIzol reagent (MajorBio Technology). RNA-seq expression levels were quantified as transcripts per million (TPM). Differentially expressed genes (DEGs) between FABP2-overexpressing and control cells were identified using the “edgeR” package, with adjusted P < 0.05 and |log₂ fold change| ≥ 1 considered significant.

Mouse model

Five-week-old male BALB/c nude mice (Foxn1nu homozygous) were purchased from the Animal Experiment Center of the Fourth Military Medical University. Mice were randomly assigned to the control (Ctrl) or FABP2-overexpressing (OE-FABP2) groups (n = 5 per group). HCT116 cells (3 × 10⁶) in 100 µL PBS were subcutaneously inoculated into the right flank of each mouse. Tumor growth was monitored for 21 days, after which mice were euthanized and tumor specimens were collected. Tumor volume was calculated using the formula V = (length × width²)/2. Humane endpoints were defined as tumor diameter > 1.5 cm or signs of significant distress.

Statistical analysis

Data were analyzed using GraphPad Prism 10.0 and are presented as mean ± standard deviation (SD) from independent biological replicates. For comparisons between two groups, unpaired t-tests were used for normally distributed data, and Mann–Whitney U tests were used for non-normally distributed data. P < 0.05 was considered statistically significant (*P < 0.05, **P < 0.005, ***P < 0.001). All cell-based experiments (CCK-8, Transwell, colony formation, qPCR, and Western blotting) were independently repeated at least three times (n = 3). For in vivo experiments, each group contained five mice (n = 5). Details of statistical tests and exact P values are provided in the figure legends.

Results

Identification of Plasma Proteins and Tissue Genes Causally Associated with CRC through pQTL and eQTL Screening

We first evaluated the causal effects of plasma proteins on CRC risk in the discovery cohort. A total of 162 suggestive associations (P < 0.05) were identified, of which 65 remained significant after FDR correction. Among these, 34 plasma proteins were associated with decreased CRC risk, while 31 were associated with increased risk (Fig. 2A, Supplementary Table 1). These 65 plasma proteins were then subjected to validation in an independent CRC GWAS cohort. In the validation cohort, 31 proteins showed suggestive associations (P < 0.05), and 16 retained FDR-significant causal associations with CRC: 11 were associated with decreased risk and 5 with increased risk (Fig. 2B, Supplementary Table 2).

Fig. 2.

Causal effects of plasma and tissue FABP2 on CRC risk identified by Mendelian Randomization (MR). (A) Volcano plot showing the causal associations between 1,001 plasma proteins and CRC risk in the discovery cohort (FinnGen R11: 353,919 participants, 8,801 cases and 345,118 controls). Each point represents one protein. Red points indicate proteins where higher levels are associated with increased CRC risk, and blue points indicate proteins where higher levels are associated with decreased CRC risk (FDR < 0.05). (B) Volcano plot validating the causal associations of the 65 candidate proteins from (A) in the independent validation cohort (UK Biobank/IEU: 469,002 participants, 6,581 cases and 462,421 controls). Red points indicate risk-increasing proteins and blue points risk-decreasing proteins (FDR < 0.05). (C) Scatter plot of MR analysis for the causal effect of plasma FABP2 on CRC risk in the discovery cohort. Each point represents an instrumental SNP. The slope of the fitted line represents the causal effect estimated by the MR method. The downward trend indicates that higher FABP2 levels are associated with a lower risk of CRC (protective association). (D) Forest plot showing the effect of each SNP on CRC risk, calculated using the Wald ratio method. The horizontal axis represents the SNP-specific causal effect estimate; the vertical axis lists the instrumental SNPs. Black dots denote point estimates and horizontal lines represent 95% confidence intervals. (E–F) Scatter plot (E) and forest plot (F) for the causal effect of colon tissue FABP2 gene expression (GTEx eQTL) on CRC risk in the discovery cohort, analyzed similarly to (C–D) Overall, both plasma protein levels and colon tissue gene expression of FABP2 show a consistent protective causal effect against CRC

To further refine the candidate list, we examined the tissue-level gene expression effects by linking pQTL-identified proteins to corresponding eQTLs in colon tissue. Among the 16 plasma proteins with validated causal associations with CRC, only FABP2 exhibited a consistent causal association at the gene expression level (P < 0.05).

In the discovery cohort, higher plasma FABP2 concentrations were significantly associated with reduced CRC risk (OR = 0.812, 95% CI: 0.744–0.887, P = 3.74 × 10⁻⁶, FDR = 1.7 × 10⁻⁴) (Fig. 2C, D). In the validation cohort, the protective effect persisted (OR = 0.928, 95% CI: 0.883–0.974, P = 0.0026, FDR = 0.013) (Supplementary Fig. 1 A, B). Consistently, eQTL-based MR analyses showed that higher FABP2 expression in colon tissue was significantly associated with lower CRC risk (OR = 0.767, 95% CI: 0.723–0.816, P = 1.9 × 10⁻¹⁷) (Fig. 2E, F) (Table 1.).

Table 1.

Mendelian randomization analysis of plasma proteins and tissue gene with CRC

	Mendelian randomization using pQTL				Mendelian randomization using eQTL
	Discovery		Validation
	OR(95% CI)	FDR	OR(95% CI)	FDR	OR(95% CI)	P
FABP2	0.812 (0.744~0.887)	1.7E-04	0.928 (0.883-0.974)	0.013	0.767 (0.723~0.816)	1.9E-17

Although heterogeneity and potential horizontal pleiotropy were detected for plasma FABP2 in the discovery cohort, neither Cochran’s Q test nor MR-Egger intercept analysis nor MR-PRESSO identified heterogeneity or pleiotropy in the validation cohort. eQTL-based analyses did not detect heterogeneity or horizontal pleiotropy for FABP2 in relation to CRC (Supplementary Table 3). Leave-one-out sensitivity analyses confirmed that the observed causal association between FABP2 and CRC was not driven by any single SNP (Supplementary Fig. 1 C - E).

Reverse MR analysis showed no evidence of a causal effect of CRC on plasma FABP2 levels (OR = 0.999, 95% CI: 0.948–1.051, P = 0.958) (Table 2). Cochran’s Q test and MR-Egger intercept analysis also indicated no heterogeneity or horizontal pleiotropy in the reverse direction (Supplementary Table 4).

Table 2.

The causal association of CRC with plasma protein FABP2 in MR analysis

Exposure	Outcome	IVW		MR Egger		Weighted median		Simple mode		Weighted mode
		OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
CRC	FABP2	0.999 (0.948–1.051)	0.958	1.026 (0.899–1.171)	0.710	0.993 (0.925–1.066)	0.841	0.997 (0.873–1.139)	0.968	1.008 (0.905–1.122)	0.891

Causal effects of plasma metabolites on CRC

We next examined the causal relationships between plasma metabolites and CRC risk. MR analyses identified 87 plasma metabolites with significant causal associations (P < 0.05), of which 44 were associated with decreased CRC risk and 43 with increased risk (Supplementary Table 5).

Sphingomyelin mediates the protective effect of FABP2 on CRC

To identify metabolites potentially mediating the effect of FABP2 on CRC, we used plasma FABP2 concentration as the exposure and the 87 metabolites with causal associations with CRC as outcomes in MR analyses. Seven plasma metabolites showed significant causal relationships with FABP2 (P < 0.05) (Supplementary Table 6). FABP2 was protective for six of these metabolites and acted as a risk factor for one. Specifically, the metabolites identified were: 3-(5-hydroxyphenyl)propionate (OR = 0.897, 95% CI: 0.805–0.999, P = 0.049), citraconate/glutaconate (OR = 0.890, 95% CI: 0.804–0.984, P = 0.023), sphingomyelin (d18:1/22:1, d18:2/22:0, d16:1/24:4) (OR = 0.901, 95% CI: 0.821–0.989, P = 0.028), 1-oleoyl-2-docosahexaenoyl-GPC (18:1/22:9) (OR = 1.124, 95% CI: 1.020–1.238, P = 0.018), X-12,130 (OR = 0.862, 95% CI: 0.767–0.970, P = 0.013), X-12,547 (OR = 0.895, 95% CI: 0.806–0.995, P = 0.039), and another X-12,547-related measure (OR = 0.823, 95% CI: 0.685–0.989, P = 0.038) (Fig. 3A).

Fig. 3.

MR analysis evaluates the metabolites-CRC risk and the causal relationship between FABP2 and metabolites. (A) Forest plot showing MR estimates for the causal effects of the seven plasma metabolites associated with FABP2 on CRC risk. Data are presented as odds ratios (ORs) with 95% confidence intervals from the IVW method. (B) Scatter plot of MR analysis for the causal effect of sphingomyelin (d18:1/22:1, d18:2/22:0, d16:1/24:4) levels on CRC risk. (C) Scatter plot of MR analysis for the causal effect of plasma FABP2 on sphingomyelin levels Together, these analyses indicate that higher FABP2 levels are associated with lower sphingomyelin levels, whereas higher sphingomyelin levels are linked to an increased risk of CRC, supporting sphingomyelin as a potential mediator of the FABP2–CRC relationship

Combining these findings with the earlier metabolite–CRC analyses, we observed that higher sphingomyelin levels were associated with increased CRC risk (OR = 1.133, 95% CI: 1.022–1.257, P = 0.018) (Fig. 3B, Supplementary Table 7). Although heterogeneity between sphingomyelin and CRC was detected by Cochran’s Q test, MR-Egger intercept analysis did not indicate horizontal pleiotropy (Supplementary Table 8). In contrast, higher plasma FABP2 levels were significantly associated with lower sphingomyelin levels (OR = 0.901, 95% CI: 0.821–0.989, P = 0.028) (Supplementary Table 9, Fig. 3C). For the FABP2–sphingomyelin analysis, Cochran’s Q test did not detect heterogeneity, and MR-Egger intercept analysis and MR-PRESSO indicated no horizontal pleiotropy (Supplementary Table 10). Reverse MR analyses further excluded reverse causality between FABP2 and sphingomyelin (Supplementary Tables 11 and 12).

Based on the mediation MR framework, β1 (total effect of FABP2 on CRC) was − 0.208, β2 (effect of FABP2 on sphingomyelin) was − 0.104, and β3 (effect of sphingomyelin on CRC) was 0.125. The mediation effect (β2 × β3) accounted for 6.25% of the total effect, indicating that sphingomyelin partially mediates the protective effect of FABP2 on CRC risk.

Bulk RNA - Seq reveals differential expression and function of FABP2 in CRC tissues

To investigate the expression pattern and potential function of FABP2 in CRC tissues, we analyzed TCGA RNA-seq data. FABP2 expression was significantly downregulated in tumor tissues relative to adjacent normal tissues (Fig. 4A-C). KEGG-based GSEA showed that higher FABP2 expression was positively associated with pathways related to fatty acid metabolism and pancreatic β-cell function (Fig. 4D), whereas lower FABP2 expression was associated with pathways such as allograft rejection, epithelial–mesenchymal transition, IFN-γ response, and myogenesis (Fig. 4E).

Fig. 4.

Bulk RNA-Seq reveals the expression and function of FABP2 in CRC tissue. (A–C) FABP2 mRNA expression is significantly downregulated in CRC tissues compared with adjacent normal tissues in The Cancer Genome Atlas (TCGA) cohort (n = 455 tumors, n = 41 normals). Data are shown as (A) volcano plot, (B) unpaired comparison between tumor and normal tissues, and (C) paired comparison of matched tumor–normal samples. Statistical test: paired t-test for (C), ***P < 0.001. (D–E) Gene set enrichment analysis (GSEA) of KEGG pathways shows that high FABP2 expression is positively associated with fatty acid metabolism–related pathways (D) and negatively associated with allograft rejection and epithelial–mesenchymal transition (EMT)–related pathways (E), NES>2, FDR<1. (F) Protein–protein interaction (PPI) network for FABP2 generated using GeneMANIA and functional enrichment analysis of its interactors, highlighting involvement in neutral lipid metabolic processes Overall, FABP2 is consistently downregulated in CRC tumors, and its expression pattern is strongly linked to lipid metabolic pathways

Using GeneMANIA, we constructed a PPI network for FABP2 and its top 20 interactors. Key interacting proteins included FABP1, CDX1, TERF2IP, HAPLN1, SI, and FABP6. Functional enrichment analyses indicated that these proteins are mainly involved in neutral lipid catabolic processes, acylglycerol catabolic processes, glycerolipid catabolism, triglyceride metabolism, and neutral lipid metabolism (Fig. 4F), further supporting a role for FABP2 in lipid metabolic regulation.

scRNA - Seq reveals FABP2 expression at the Single - Cell level

We next assessed FABP2 expression at single-cell resolution using scRNA-seq data from three pairs of CRC and adjacent normal tissues (Fig. 5A). A total of 24 clusters were identified (Fig. 5B). Clusters 0, 1, and 4, characterized by high CD3D and CD7 expression, were annotated as T cells; clusters 2, 3, 9, 12, 13, 14, and 17, with high KRT8 and EPCAM expression, were annotated as epithelial cells; clusters 6, 7, 8, 11, 15, and 18, expressing CD79A and MS4A1, were classified as B cells; clusters 5 and 21, expressing LYZ and CD68, were annotated as myeloid cells; clusters 10 and 19, with high COL1A1 and DCN, were classified as fibroblasts; clusters 16 and 23, with high MZB1 and JCHAIN, were identified as plasma cells; cluster 20, expressing PECAM1 and VWF, was annotated as endothelial cells; and cluster 22, expressing TPSAB1 and CPA3, was annotated as mast cells (Fig. 5C).

Fig. 5.

scRNA-Seq reveals the expression of FABP2 at the single-cell level in CRC. (A) Quality control metrics for the single-cell RNA-seq data from GEO dataset GSE231559, showing the distribution of detected genes, UMI counts, and mitochondrial gene percentage per cell. (B) UMAP visualization of 24 unsupervised cell clusters. (C) Bubble plot of top marker genes for each cluster in (B). (D) UMAP plot annotated with eight major cell types using canonical markers from the CellMarker database. (E) Stacked bar plot showing the proportional abundance of each cell type in normal and tumor tissues. (F) Heatmap showing FABP2 expression across the eight annotated cell types. (G) Bar plot of FABP2 expression within the epithelial compartment in normal versus tumor tissues. Statistical test: unpaired t-test, ***P < 0.001 These analyses demonstrate that FABP2 is predominantly expressed in epithelial cells and is significantly reduced in tumor epithelial cells compared with normal epithelium

Using the CellMarker database, we further grouped these clusters into eight major cell types: T cells, B cells, plasma cells, mast cells, myeloid cells, endothelial cells, epithelial cells, and fibroblasts (Fig. 5D). Comparison of cell-type proportions between tumor and normal tissues revealed higher proportions of T cells and epithelial cells and lower proportions of B cells and fibroblasts in tumor tissues (Fig. 5E). Among all cell types, FABP2 expression was predominantly restricted to epithelial cells (Fig. 5F). Importantly, FABP2 expression in epithelial cells was significantly lower in tumor tissues than in normal tissues (Fig. 5G).

Clinical tissue specimen validation of FABP2 expression and sphingomyelin metabolism levels

We next validated FABP2 expression and sphingomyelin metabolism in clinical samples. Paired CRC tumor tissues and adjacent normal tissues, as well as plasma samples from CRC patients and healthy controls, were collected (Table 3). FABP2 expression at both the mRNA and protein levels was evaluated by qPCR, Western blotting, and immunohistochemistry. Consistent with the bioinformatics analyses, FABP2 expression was significantly lower in tumor tissues than in adjacent normal tissues (Fig. 6A–C). Plasma FABP2 levels were also significantly reduced in CRC patients compared with healthy controls (Fig. 6F).

Table 3.

Clinical characteristics of patients with colorectal cancer and healthy individuals

Characteristics	Colorectal cancer	Percentage (%)	Healthy individuals	Percentage (%)
	Number (n = 20)		Number (n = 20)
Age(years)
≤ 60	8	40	13	65
> 60	12	60	7	35
Gender
Male	13	65	12	60
Female	7	35	8	40
Tumor location
Rectum	11	55
colon	9	45
Tumor size (cm)
≤ 4	10	50
> 4	10	50
Differentiation status
Well	1	5
Moderately	17	85
Poorly	2	10
T stage
T1	1	5
T2	2	10
T3	15	75
T4	2	10
N stage
N0	9	45
N1	7	35
N2	4	20
M stage
M0	20	100
M1	0	0

Fig. 6.

Clinical tissue samples validate the expression levels of FABP2 and sphingolipid metabolism levels. (A) Western blot analysis and densitometric quantification of FABP2 protein expression in 16 paired normal (N) and tumor (T) tissues. Statistical test: paired t-test, **P < 0.01. (B) Quantitative PCR (qPCR) analysis of FABP2 mRNA in 20 paired normal and tumor tissues. Statistical test: paired t-test, ***P < 0.001. (C) Representative immunohistochemistry (IHC) images and semi-quantitative IHC scores of FABP2 protein expression in 20 paired normal and tumor tissues. Statistical test: paired t-test, **P < 0.01. (D) Sphingomyelin levels measured in 10 paired normal and tumor tissues. Statistical test: paired t-test, **P < 0.01. (E) Sphingomyelin levels in plasma from 10 CRC patients and 10 healthy controls. Statistical test: unpaired t-test, ***P < 0.001. (F) Plasma FABP2 levels in 20 CRC patients and 20 healthy controls. Statistical test: unpaired t-test, ***P < 0.001 Collectively, these data confirm that FABP2 is downregulated and sphingomyelin is upregulated at both tissue and plasma levels in CRC patients

Sphingomyelin levels in tissues and plasma were quantified using sphingomyelin assay kits. Sphingomyelin metabolism was significantly elevated in tumor tissues compared with normal tissues and was also higher in the plasma of CRC patients than in healthy controls (Fig. 6D, E). These findings corroborate the MR and transcriptomic results and support a negative relationship between FABP2 and sphingomyelin metabolism in CRC.

FABP2 regulates CRC cellular functions and tumor progression

To investigate the functional role of FABP2 in CRC progression, we established FABP2-overexpressing (OE-FABP2) HCT116 and HCT8 cell lines, with successful overexpression confirmed by Western blotting and qPCR (Fig. 7A–C). CCK-8 assays and colony formation assays showed that FABP2 overexpression significantly inhibited CRC cell proliferation (Fig. 7D, E, H, I). Transwell assays further demonstrated that OE-FABP2 markedly reduced CRC cell migration and invasion (Fig. 7J, K). Intracellular sphingomyelin levels were significantly decreased upon FABP2 overexpression in both cell lines (Fig. 7F, G).

Fig. 7.

FABP2 Regulates CRC Cellular Functions and Tumor Progression. All cell experiments were performed with n = 3 independent biological replicates; data are presented as mean ± SD. Statistical test: unpaired t-test. **P < 0.01, ***P < 0.001. (A–C) Western blot (A, B) and qPCR (C) validation of FABP2 overexpression (OE-FABP2) in HCT116 and HCT8 cells compared with negative control (Ctrl). (D–E) Cell proliferation assessed by CCK-8 assay in HCT116 (D) and HCT8 (E) cells. (F–G) Intracellular sphingomyelin levels in HCT116 (F) and HCT8 (G) cells after FABP2 overexpression. (H–I) Plate clonogenicity assay for HCT116 (H) and HCT8 (I) cells after FABP2 overexpression. (J–K) Cell migration and invasion assessed by Transwell assay in HCT116 (J) and HCT8 (K) cells. (L–O) Subcutaneous xenograft model in BALB/c nude mice (n = 5 per group). (L) Representative images of subcutaneous tumors. (M) Tumor growth curves. (N) Tumor volumes. (O) Tumor weights at endpoint Overall, FABP2 functions as a tumor suppressor in CRC by inhibiting cell proliferation, migration, invasion, and tumor growth, concomitant with reduced sphingomyelin levels

In vivo, a subcutaneous xenograft model in BALB/c nude mice revealed that OE-FABP2 significantly inhibited tumor growth compared with the control group (Fig. 7L–O). Collectively, these findings indicate that FABP2 functions as a tumor suppressor in CRC by restraining malignant cell behaviors and reducing sphingomyelin metabolism.

FABP2 regulates sphingomyelin metabolism and cellular functions via the PPAR signaling pathway

To elucidate the mechanism by which FABP2 modulates CRC progression, we performed RNA-seq on OE-FABP2 and control HCT116 cells. A total of 285 genes were upregulated and 707 were downregulated upon FABP2 overexpression (Fig. 8A, B). GO and KEGG enrichment analyses indicated significant enrichment of lipid-related processes, including response to lipid, regulation of lipid transport, and lipid metabolism (Fig. 8C), as well as signaling pathways such as cAMP, TNF, and PPAR (Fig. 8D). Given the central role of PPAR signaling in fatty acid metabolism, these results suggest that FABP2 may influence CRC progression through PPAR pathway modulation.

Fig. 8.

Transcriptomic profiling of FABP2-overexpressing cells. RNA sequencing was performed on HCT116 cells overexpressing FABP2 (OE-FABP2) versus control (Ctrl) (n = 3 per group). (A) Volcano plot of differentially expressed genes (DEGs), highlighting genes with |log₂ fold change| ≥ 1 and FDR-adjusted P < 0.05. Red, upregulated genes; blue, downregulated genes. (B) Heatmap of DEGs from (A), showing distinct clustering of OE-FABP2 and Ctrl samples. (C) Gene Ontology (GO) enrichment analysis of DEGs. (D) KEGG pathway enrichment analysis of DEGs, indicating significant enrichment in lipid-related biological processes and pathways, including PPAR signaling These results indicate that FABP2 overexpression induces widespread transcriptomic reprogramming, particularly in lipid metabolism–related pathways

To validate this hypothesis, we examined the expression of PPAR-α and its downstream fatty acid oxidation (FAO)–related target genes ACOX1 and CPT1A. FABP2 overexpression led to downregulation of PPAR-α, ACOX1, and CPT1A at both mRNA and protein levels compared with controls. Treatment with the PPAR-α agonist WY-14643 in OE-FABP2 cells (OE-FABP2 + WY-14643 group) restored PPAR-α, ACOX1, and CPT1A expression (Fig. 9A–C). Functionally, FABP2 overexpression reduced cell proliferation, colony formation, migration, and invasion in both HCT116 and HCT8 cells, accompanied by decreased sphingomyelin levels. WY-14643 treatment reversed these effects, restoring cell proliferation, clonogenicity, migration, invasion, and sphingomyelin metabolism (Fig. 9D–K). These data support a model in which FABP2 suppresses PPAR-α signaling activity, thereby reducing FAO-associated downstream gene expression, limiting sphingomyelin metabolism, and restraining malignant cell behaviors.

Fig. 9.

FABP2 Regulates Sphingomyelin Metabolism and Cellular Functions via the PPAR Signaling Pathway. All cell experiments were performed with n = 3 independent biological replicates; data are presented as mean ± SD. Cells were treated with the PPAR-α agonist WY-14643 where indicated. Statistical test: unpaired t-test. **P < 0.01, ***P < 0.001, ns, not significant. (A–B) Western blot analysis of FABP2, PPAR-α, ACOX1, and CPT1A protein expression in each group. (C) qPCR analysis of FABP2, PPAR-α, ACOX1, and CPT1A mRNA levels in each group. (D–E) Cell proliferation assessed by CCK-8 assay in HCT116 (D) and HCT8 (E) cells in each group. (F–G) Plate clonogenicity assay in HCT116 (F) and HCT8 (G) cells in each group. (H–I) Cell migration and invasion assessed by Transwell assay in HCT116 (H) and HCT8 (I) cells in each group. (J–K) Intracellular sphingomyelin levels in HCT116 (J) and HCT8 (K) cells in each group These data support that the tumor-suppressive effects of FABP2 are mediated, at least in part, through inhibition of PPAR-α signaling activity, thereby limiting fatty acid oxidation–linked sphingomyelin metabolism and malignant cell behaviors

Discussion

In this multi-omics study integrating MR, bulk and single-cell transcriptomics, clinical specimen analyses, and functional experiments, we identified FABP2 as a protective factor against CRC and elucidated a FABP2–PPAR–sphingomyelin axis that may underlie its tumor-suppressive role.

CRC remains a major global health challenge, with increasing incidence and high mortality despite advances in early detection and therapy [19, 20]. Plasma proteins and metabolites are attractive candidates for noninvasive biomarkers and may also participate in tumor initiation and progression [21–24]. Our MR framework systematically evaluated the causal relationships among plasma proteins, tissue gene expression, plasma metabolites, and CRC risk. Through a discovery–validation strategy, we identified FABP2 as a robust protective factor, with higher plasma FABP2 levels and higher colonic FABP2 expression both associated with reduced CRC risk. The attrition of significant associations from the discovery cohort (n = 65) to the validation cohort (n = 16) likely reflects differences in sample size and population heterogeneity; however, this two-stage design acts as a stringent filter, and the persistence of FABP2 as a significant candidate highlights its robustness and relevance for mechanistic investigation.

FABP proteins are intracellular lipid-binding proteins that play diverse roles in fatty acid uptake, transport, and metabolism [25, 26]. Dysregulated FABP expression has been implicated in multiple cancers [27]. For example, FABP1 and FABP4 are upregulated in hepatocellular carcinoma and pancreatic cancer [28–30], whereas FABP7 is highly expressed in glioma [31]. FABP2 is an intestinal-type FABP involved in dietary fat absorption and transport [32, 33], and polymorphisms in the FABP2 gene have been associated with metabolic diseases [34, 35]. However, its role in CRC has remained largely unexplored.

Our MR and transcriptomic analyses consistently demonstrated that FABP2 expression is decreased in CRC tissues and plasma compared with normal controls, suggesting a potential tumor-suppressive role. Bulk RNA-seq and PPI analyses further showed that FABP2 and its interactors are enriched in pathways related to neutral lipid and triglyceride metabolism, highlighting its central role in lipid homeostasis. scRNA-seq analysis revealed that FABP2 is predominantly expressed in epithelial cells and is markedly reduced in tumor epithelium, indicating an epithelial-intrinsic role in the colorectal mucosa. Clinical validation experiments confirmed that FABP2 expression is indeed lower in tumor tissues and that plasma FABP2 levels are reduced in CRC patients, supporting its potential as both a mechanistic factor and a biomarker candidate.

Lipid metabolic reprogramming is a hallmark of cancer [36, 37]. Tumor cells upregulate lipid uptake, synthesis, and catabolism to meet the energy and biosynthetic demands of rapid proliferation. Key pathways involving cholesterol, fatty acids, and phospholipids have been shown to influence tumor growth, invasion, and therapeutic resistance [37–39]. Sphingolipids represent a critical class of phospholipids whose metabolites, including sphingomyelin, ceramide, and sphingosine-1-phosphate, play pivotal roles in cell proliferation, apoptosis, and differentiation [40, 41]. Dysregulated sphingolipid metabolism and aberrant sphingomyelin levels have been reported in various cancers and have been linked to tumor progression and treatment response [42–47].

In our mediation MR analysis, sphingomyelin emerged as a metabolite that partially mediates the protective effect of FABP2 on CRC. Higher plasma sphingomyelin levels were associated with increased CRC risk, whereas higher FABP2 levels were associated with lower sphingomyelin levels. Clinical sample analyses confirmed that sphingomyelin metabolism is elevated in CRC tissues and in the plasma of CRC patients. These findings, together with our functional assays showing that FABP2 overexpression reduces sphingomyelin levels in CRC cells, support the notion that FABP2 restricts CRC progression at least in part by suppressing sphingomyelin metabolism.

PPAR-α is a nuclear receptor that regulates fatty acid uptake, transport, and oxidation, and its functional roles are highly tissue- and context-dependent [48]. In metabolically reprogrammed CRC cells, PPAR-α activation can preferentially enhance FAO, increase ATP production, and support tumor cell growth and migration. In our model, FABP2 overexpression repressed PPAR-α and its downstream FAO-related target genes CPT1A and ACOX1, while the PPAR-α agonist WY-14643 restored their expression and reversed the tumor-suppressive phenotypes induced by FABP2. This suggests that FABP2 suppresses CRC progression by attenuating PPAR-α–driven FAO and associated sphingomyelin metabolism.

Furthermore, several studies have described ligand-dependent transcriptional induction of nuclear receptors, including PPAR-α, whereby receptor activation enhances its own promoter activity and related regulatory networks, resulting in increased receptor mRNA and protein expression [49]. In intestinal epithelial and colon-derived models, WY-14643 has been shown to induce metabolic gene expression via PPAR-α–dependent mechanisms, consistent with a positive feedback loop that amplifies FAO-related transcriptional programs [50]. In our experiments, WY-14643 increased PPAR-α expression alongside CPT1A and ACOX1, supporting a similar feedback mechanism in CRC cells. Taken together, our data and prior literature support the concept that excessive activation of PPAR-α can promote CRC progression through FAO-associated metabolic reprogramming, reflecting the context-dependent nature of PPAR signaling.

Our study also has limitations. First, although MR provides strong evidence for a causal role of FABP2 and sphingomyelin in CRC, the findings require replication in multiple large-scale, independent, and preferably prospective cohorts to validate their generalizability and clinical utility. Second, while we demonstrated that FABP2 modulates PPAR-α signaling and sphingomyelin metabolism, the precise molecular interactions linking FABP2 to PPAR-α regulation remain to be fully clarified. Third, the in vivo xenograft experiments, although showing significant inhibition of tumor growth by FABP2 overexpression, were exploratory in nature and conducted with a relatively small sample size (n = 5 per group) without formal power calculation. Larger, well-powered animal studies and additional in vivo models will be important to corroborate our findings. Finally, our work focused on epithelial-intrinsic effects of FABP2; potential systemic and microenvironmental contributions of plasma FABP2 merit further exploration.

In summary, our integrated genetic, transcriptomic, clinical, and functional analyses support a model in which FABP2 acts as a tumor-suppressive regulator of lipid metabolism in CRC, operating at least in part through attenuation of PPAR-α–driven fatty acid oxidation and sphingomyelin metabolism. These findings broaden our understanding of metabolic reprogramming in CRC and nominate the FABP2–PPAR-α–sphingomyelin axis as a mechanistic pathway that may be exploitable in future diagnostic and therapeutic strategies. However, whether FABP2 or sphingomyelin can ultimately serve as clinically useful biomarkers or drug targets will require confirmation in larger, independent and ideally prospective cohorts, as well as rigorous preclinical modeling and pharmacologic intervention studies.

Conclusion

In conclusion, our study uncovers the important roles of FABP2 and sphingomyelin in CRC progression and demonstrates that FABP2 may exert a protective effect by by inhibiting sphingomyelin metabolism via the PPAR signaling pathway. This finding offers new insights into the molecular mechanisms of CRC and highlighting FABP2 and sphingomyelin metabolism as promising candidates for future therapeutic strategies.

Abbreviations

CRC

Colorectal cancer

MR

Mendelian randomization

scRNA-Seq

single-cell RNA sequencing

GWAS

Genome-wide association studies

IVW

Inverse Variance Weighted

TCGA

The Cancer Genome Atlas

GEO

Gene Expression Omnibus

PPI

Protein-protein Interactions

GSEA

Gene Set Enrichment Analysis

RNA-Seq

RNA-Sequencing

Author contributions

Jianyong Zheng and Fan Feng designed the experiments. Pengfei Wang, Kunli Du and Zhipeng Zhang performed most of the experiments. Pengfei Wang and Kunli Du wrote the manuscript. Zhipeng Zhang, Kunli Du and Jian Zhang supervised the study and revised the manuscript. Ye Tian, Xinyu Qiao, Gaozan Zheng, Juan Kang assisted with data analysis and performed experiments. Yumao Yang and Wenfang He collected the samples. All authors contributed to the writing and editing of the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82072655). Fund for Scientific and Technological Innovation Team of Shaanxi Innovation Capability Support Plan (2023-CX-TD-67). Key Research and Development Project of Shaanxi Province (2024SF-ZDCYL-02-1).

Data availability

Data Available StatementAll GWAS summary statistics are publicly available and can be accessed freely. The detailed information has been described in detail in the methodology section.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the First Affiliated Hospital of the Fourth Military Medical University.

Competing interests

The authors declare no competing interests.