Multi-omics integration deciphers immune-metabolic heterogeneity in colorectal cancer: a prognostic model and therapeutic strategies targeting ANGPTL4/FABP4/RBP7

doi:10.12122/j.issn.1673-4254.2026.05.02

Abstract

Abstract:

Objective To explore the heterogeneity of tumor immune microenvironment (TIME) in colorectal cancer (CRC) and the resultant treatment resistance and develop a prognostic model for predicting the treatment outcomes of CRC. Methods The multi-omics data including transcriptomic, somatic mutation and single-cell RNA sequencing data of a total of 1136 CRC samples from a TCGA cohort (n=568) and a GEO cohort (n=568) were integrated. The TCGA cohort was divided into Immunity_High and Immunity_Low subtypes by single-sample gene set enrichment analysis (ssGSEA) and t-SNE algorithm. A prognostic model was constructed using univariate Cox regression combined with LASSO-Cox regression, and the functions of the core genes including ANGPTL4, FABP4 and RBP7 were verified by single cell RNA-seq, molecular docking, kinetic simulations and qPCR/IHC experiments. Results The Immunity_High subtype of CRC was enriched with CD8⁺ T cells and had significantly longer patient survival, while the Immunity_Low subtype was characterized by M0 macrophage infiltration. The prognostic model constructed based on 12 genes including ANGPTL4, FABP4 and RBP7 showed good predictive performance, with the AUC of 3-year survival rate reaching 0.765, an overall AUC of 0.76 in the TCGA cohort, and an overall AUC of 0.70 in the GEO cohort. Mechanistically, ANGPTL4 regulated macrophage polarization, FABP4 mediated fibroblast reprogramming, and RBP7 up-regulated PD-L1 expression. Molecular docking and kinetic simulations confirmed stable binding of retinoic acid and rosiglitazone to the core targets with the minimum binding energy of -8.3 kcal/mol. Conclusion The constructed prognostic model and the identified potential therapeutic targets provide new strategies to address immune therapy resistance for CRC.

Key words: colorectal cancer, multi-omics integration, immune-metabolic heterogeneity, prognostic model, ANGPTL4/FABP4/RBP7

Ning DING, Tianshu PENG, Li MO, Xiangdang HU, Zongliang YANG, You XIAO, Xiaobing FU, Jianzheng ZHAO, Leilei LI, Yingjie ZHANG. Multi-omics integration deciphers immune-metabolic heterogeneity in colorectal cancer: a prognostic model and therapeutic strategies targeting ANGPTL4/FABP4/RBP7[J]. Journal of Southern Medical University, 2026, 46(5): 977-993.

Figures/Tables 9

Tab.1 Primer sequences for qPCR amplification of the target genes

Gene	Forward primer	Reverse primer
ANGPTL4	TGCGTTGCCTCCTGAGATCG	CCAAGTGGTCCCTCTGAGTCC
FABP4	GATAAACTGGTGGTGGAATGCG	AATGCGAACTTCAGTCCAGGTC
RBP7	TCCACATCCAGCAGCAGAGC	AGTCCTCGTCGCCAAATGGG
β-actin	TGGCACCCAGCACAATGAA	CTAAGTCATAGTCCGCCTAGAAGCA

Fig.1 Flow chart of the study design.

Fig.3 Integrated analysis of immune heterogeneity and prognostic modeling in CRC. A: Volcano plot of differentially expressed genes (DEGs). B: Heatmap of DEG expression stratification. C: Co-expression Heatmap of 356 Immune-Related DEGs. D: Venn diagram of DEG-ImmPort overlap. E: HLA gene upregulation in High-Immunity Subtype. F: TF-gene regulatory network (Sankey diagram). G: PPI network of hub genes. H: LASSO coefficient trajectories. I: Cross-validation curve for LASSO regression.

Fig.4 Dual-cohort survival analysis and risk stratification in CRC. A: TCGA survival curve: the median overall survival (OS) of the high-risk group (red) was significantly shorter than that of the low-risk group (blue) (P<0.001). B: GEO validation: significant OS difference (P=0.026). C: TCGA risk stratification (red for high risk and blue for low risk). D: GEO distribution trend consistent with TCGA. E: TCGA scatter plot showing clustering of deceased patients (red dots) in high-risk areas. F: GEO validation showing significant mortality concentration in high-risk group. G: The risk score was negatively correlated with OS in TCGA cohort (R=-0.15, P=0.00064). H: Negative correlation between the risk score and OS in the GEO cohort (R=-0.11, P=0.0082).

Fig.5 Prognostic analysis of clinical variables and core targets in CRC. A: Univariate Cox regression analysis (clinical variables and risk score). B: Multivariate Cox regression analysis (adjusted for clinical variables). C-E: Survival curves by ANGPTL4 expression, corresponding to overall survival (OS), disease-free survival (DFS), and progression-free survival (PFS) respectively.F-H: Survival curves by FABP4 expression, corresponding to OS, DFS, and PFS respectively. I-K: Survival curves by RBP7 expression, corresponding to OS, DFS, and PFS respectively.

Fig.6 Immune-genomic prognostic stratification in CRC. A: Gene-immune cell correlation heatmap. B: Survival disparity by tumor mutational burden (TMB). C: Combined TMB-risk stratification. *P<0.05, **P<0.01, ***P<0.001.

Fig.7 Single-cell profiling of core gene expression across cell types. A: Analysis of ANGPTL4 expression characteristics: The left plot shows cancer cell distribution, the middle indicates high ANGPTL4 in epithelial cells, and the right shows 3.2 TPM. B: Analysis of FABP4 expression characteristics: The left plot labels cell types, the middle shows FABP4 high in myofibroblasts and endothelial cells, and the right suggests FABP4's role in stromal remodeling. C: Analysis of RBP7 expression characteristics: The left plot shows cell distribution, the middle highlights RBP7 in malignant cells and macrophages, and the bar chart displays malignant cells at 2.7 TPM, macrophages at 1.8 TPM, and plasma cells at <0.3 TPM.

Fig.9 Gene expression-guided drug sensitivity profiling in CRC. In the boxplots, green represents the low-expression group and red represents the high-expression group. A: ANGPTL4-low tumors show heightened sensitivity to cisplatin (P=3.6e-08). B: Enhanced response to epirubicin in ANGPTL4-low subgroup (P=3.4e-08). C: ANGPTL4-low tumors exhibit sensitivity to mitoxantrone (P=5.2e-09). D: ANGPTL4-low subgroup responds to niraparib (P=4e-10). E: Significant sensitivity to olaparib in ANGPTL4-low tumors (P=6.4e-12). F: FABP4-low tumors respond to dasatinib (P=1e-10). G: FABP4-high tumors favor lapatinib (P=1.6e-08). H: RBP7-low expression correlates with dasatinib efficacy (P=4.2e-09). I: RBP7-high tumors are more sensitive to selumetinib (P=4.4e-12). J: RBP7-high tumors exhibit significant response to trametinib (P=4.5e-09).

Fig.10 Multimodal prognostic model and molecular validation in CRC. A: Risk score distribution analysis shows higher scores in females than in males (P=0.002) and in advanced-stage than in early-stage tumors. B: ROC curves validate model accuracy with 3-year AUC of 0.765 and 5-year AUC of 0.718 (>0.7 threshold) (color-coded green: 1-year; purple: 3-year; red: 5-year). C: Calibration curves show high agreement between predicted and observed survival rates, and the 1-, 3- and 5-year curves align with ideal diagonal. D-F: Immunohistochemical staining showing ANGPTL4/FABP4/RBP7 protein overexpressions in tumors vs normal tissues validated by HPA (Original magnification: ×200). G-I: qPCR validation confirms ANGPTL4/FABP4/RBP7 mRNA upregulation in CRC cell lines (**P<0.01, ***P<0.001, ****P<0.0001 vs EIHC).

References 34

[1]	Morgan E, Arnold M, Gini A, et al. Global burden of colorectal cancer in 2020 and 2040: incidence and mortality estimates from GLOBOCAN[J]. Gut, 2023, 72(2): 338-44. doi：10.1136/gutjnl-2022-327736
[2]	Wang C, Zeng Q, Gül ZM, et al. Circadian tumor infiltration and function of CD8⁺ T cells dictate immunotherapy efficacy[J]. Cell, 2024, 187(11): 2690-702. doi：10.1016/j.cell.2024.04.015
[3]	Williams HL, Frei AL, Koessler T, et al. The current landscape of spatial biomarkers for prediction of response to immune checkpoint inhibition[J]. NPJ Precis Oncol, 2024, 8(1): 178. doi：10.1038/s41698-024-00671-1
[4]	Ma Y, Li J, Zhao X, et al. Multi-omics cluster defines the subtypes of CRC with distinct prognosis and tumor microenvironment[J]. Eur J Med Res, 2024, 29: 207. doi：10.1186/s40001-024-01805-8
[5]	Chowell D, Morris LGT, Grigg CM, et al. Patient HLA class I genotype influences cancer response to checkpoint blockade immunotherapy[J]. Science, 2018, 359(6375): 582-7. doi：10.1126/science.aao4572
[6]	Hutter C, Zenklusen JC. The cancer genome atlas: creating lasting value beyond its data[J]. Cell, 2018, 173(2): 283-5. doi：10.1016/j.cell.2018.03.042
[7]	Clough E, Barrett T, Wilhite SE, et al. NCBI GEO archive for gene expression and epigenomics data sets: 23-year update[J]. Nucleic Acids Res, 2024, 52(D1): D138-44. doi：10.1093/nar/gkad965
[8]	Bhattacharya S, Dunn P, Thomas CG, et al. ImmPort, toward repurposing of open access immunological assay data for translational and clinical research[J]. Sci Data, 2018, 5: 180015. doi：10.1038/sdata.2018.15
[9]	Petrella BL, Armstrong DA, Vincenti MP. Interleukin-1 beta and transforming growth factor-beta 3 cooperate to activate matrix metalloproteinase expression and invasiveness in A549 lung adenocarcinoma cells[J]. Cancer Lett, 2012, 325(2): 220-6. doi：10.1016/j.canlet.2012.07.009
[10]	Taing L, Dandawate A, L' Yi S, et al. Cistrome Data Browser: integrated search, analysis and visualization of chromatin data[J]. Nucleic Acids Res, 2024, 52(D1): D61-6. doi：10.1093/nar/gkad1069
[11]	Szklarczyk D, Nastou K, Koutrouli M, et al. The STRING database in 2025: protein networks with directionality of regulation[J]. Nucleic Acids Res, 2025, 53(D1): D730-7. doi：10.1093/nar/gkae1113
[12]	Liu XH, Zhang WY, Zhang Q, et al. Development and validation of a machine learning-augmented algorithm for diabetes screening in community and primary care settings: a population-based study[J]. Front Endocrinol, 2022, 13: 1043919. doi：10.3389/fendo.2022.1043919
[13]	Wu XH, Lin K, Gao C, et al. miR-627-5p inhibits malignant progression of cervical cancer by targeting ANGPTL4[J]. Eur J Histochem, 2025, 69(2): 4161. doi：10.4081/ejh.2025.4161
[14]	Zuo YY, He Z, Chen Y, et al. Dual role of ANGPTL4 in inflammation[J]. Inflamm Res, 2023, 72(6): 1303-13. doi：10.1007/s00011-023-01753-9
[15]	Harjes U, Bridges E, Gharpure KM, et al. Antiangiogenic and tumour inhibitory effects of downregulating tumour endothelial FABP4[J]. Oncogene, 2017, 36(7): 912-21. doi：10.1038/onc.2016.256
[16]	Yang J, Liu XJ, Shao YQ, et al. Diagnostic, prognostic, and immunological roles of FABP4 in pancancer: a bioinformatics analysis[J]. Comput Math Methods Med, 2022, 2022: 3764914. doi：10.1155/2022/3764914
[17]	Parmar MB, Shams R, Wright JM. Genomic organization and transcription of the medaka and zebrafish cellular retinol-binding protein (rbp) genes[J]. Mar Genomics, 2013, 11: 1-10. doi：10.1016/j.margen.2013.04.001
[18]	Kim DH, Ahn J, Suh Y, et al. Retinol binding protein 7 promotes adipogenesis in vitro and regulates expression of genes involved in retinol metabolism[J]. Front Cell Dev Biol, 2022, 10: 876031. doi：10.3389/fcell.2022.876031
[19]	Chen YB, Zhou XY, Xie YW, et al. Establishment of a seven-gene signature associated with CD8⁺ T cells through the utilization of both single-cell and bulk RNA-sequencing techniques in clear cell renal cell carcinoma[J]. Int J Mol Sci, 2023, 24(18): 13729. doi：10.3390/ijms241813729
[20]	Curiel TJ, Coukos G, Zou LH, et al. Specific recruitment of regulatory T cells in ovarian carcinoma fosters immune privilege and predicts reduced survival[J]. Nat Med, 2004, 10(9): 942-9. doi：10.1016/S0960-8524(03)00033-6
[21]	Gogoi M, Chandra K, Sarikhani M, et al. Salmonella escapes adaptive immune response via SIRT2 mediated modulation of innate immune response in dendritic cells[J]. PLoS Pathog, 2018, 14(11): e1007437. doi：10.1371/journal.ppat.1007437
[22]	Rassart E, Desmarais F, Najyb O, et al. Apolipoprotein D[J]. Gene, 2020, 756: 144874. doi：10.1016/j.gene.2020.144874
[23]	Zhou ZJ, Xie X, Wang X, et al. Correlations between tumor mutation burden and immunocyte infiltration and their prognostic value in colon cancer[J]. Front Genet, 2021, 12: 623424. doi：10.3389/fgene.2021.623424
[24]	Morimura S, Sugaya M, Sato S. Interaction between CX3CL1 and CX3CR1 regulates vasculitis induced by immune complex deposition[J]. Am J Pathol, 2013, 182(5): 1640-7. doi：10.1016/j.ajpath.2013.01.023
[25]	Olson ME, Harris RS, Harki DA. APOBEC enzymes as targets for virus and cancer therapy[J]. Cell Chem Biol, 2018, 25(1): 36-49. doi：10.1016/j.chembiol.2017.10.007
[26]	Xu X, Lu YD, Wu YL, et al. A signature of seven immune-related genes predicts overall survival in male gastric cancer patients[J]. Cancer Cell Int, 2021, 21(1): 117. doi：10.1186/s12935-021-01823-0
[27]	He XF, Chen HJ, Zhong XY, et al. BST2 induced macrophage M2 polarization to promote the progression of colorectal cancer[J]. Int J Biol Sci, 2023, 19(1): 331-45. doi：10.7150/ijbs.72538
[28]	Rwandamuriye FX, Wang T, Zhang HF, et al. Local therapy with combination TLR agonists stimulates systemic anti-tumor immunity and sensitizes tumors to immune checkpoint blockade[J]. Oncoimmunology, 2024, 13(1): 2395067. doi：10.1080/2162402X.2024.2395067
[29]	Sugiura D, Maruhashi T, Okazaki IM, et al. Restriction of PD-1 function by cis-PD-L1/CD80 interactions is required for optimal T cell responses[J]. Science, 2019, 364(6440): 558-66. doi：10.1126/science.aav7062
[30]	Gorría T, Sierra-Boada M, Rojas M, et al. Metabolic singularities in microsatellite-stable colorectal cancer: identifying key players in immunosuppression to improve the immunotherapy response[J]. Cancers, 2025, 17(3): 498. doi：10.3390/cancers17030498
[31]	Niu TM, Zhu XJ, Zhao DS, et al. Unveiling interaction mechanisms between myricitrin and human serum albumin: Insights from multi-spectroscopic, molecular docking and molecular dynamic simulation analyses[J]. Spectrochim Acta A Mol Biomol Spectrosc, 2023, 285: 121871. doi：10.1016/j.saa.2022.121871
[32]	Guccione C, Patel L, Tomofuji Y, et al. Incomplete human reference genomes can drive false sex biases and expose patient-identifying information in metagenomic data[J]. Res Sq, 2024: rs.3.rs-rs.4721159. doi：10.1038/s41467-025-56077-5
[33]	Xu HX, Jiao Y, Qin S, et al. Organoid technology in disease modelling, drug development, personalized treatment and regeneration medicine[J]. Exp Hematol Oncol, 2018, 7: 30. doi：10.1186/s40164-018-0122-9
[34]	Bao YC, Qiao J, Gong WJ, et al. Spatial metabolomics highlights metabolic reprogramming in acute myeloid leukemia mice through creatine pathway[J]. Acta Pharm Sin B, 2024, 14(10): 4461-77. doi：10.1016/j.apsb.2024.07.004