A novel anoikis-related risk model predicts prognosis in patients with colorectal cancer and responses to different immunotherapy strategies

Abstract

Purpose: We aimed to study the role of anoikis-related genes (ARGs) in colorectal cancer (CRC) using bioinformatics.

Methods: GSE39582 and GSE39084, which collectively contain 363 CRC samples, were downloaded from the NCBI Gene Expression Omnibus (GEO) database as a test set. TCGA-COADREAD, with 376 CRC samples, was downloaded from the UCSC database as a validation set. Univariate Cox regression analysis was used to screen for ARGs that were significantly associated with prognosis. The top 10 ARGs were used to classify the samples into different subtypes based on unsupervised cluster analysis. The immune environments of the different subtypes were analyzed. ARGs that were significantly associated with CRC prognosis were used to construct a risk model. Univariate and multivariate Cox regression analyses were used to screen independent prognostic factors and construct a nomogram.

Results: Four anoikis-related subtypes (ARSs) with differential prognoses and immune microenvironments were identified. KRAS and epithelial-mesenchymal transition pathways were enriched in subtype B, which had the worst prognosis. Three ARGs (DLG1, AKT3, and LPAR1) were used to construct the risk model. Both the test and validation sets showed worse outcomes for patients in the high-risk group than those in the low-risk group. Risk score was found to be an independent prognostic factor for CRC. Moreover, there was a difference in drug sensitivity between the high- and low-risk groups.

Conclusion: The identified ARGs and risk scores were associated with CRC prognosis and could predict the responses of patients with CRC to immunotherapy strategies.

Keywords: Anoikis; Colorectal cancer; Epithelial-mesenchymal transition; Immune microenvironment; Risk model.

Fig. 1

Top 10 anoikis-related genes associated with prognosis of colorectal cancer. A Top 10 anoikis-related genes significantly associated with prognosis of colorectal cancer were identified by univariate Cox regression analysis. B Correlation analysis between top 10 anoikis-related genes with prognostic value. C Protein–protein interaction analysis between 10 anoikis-related genes with prognostic value

Fig. 2

Identification of anoikis-related subtypes. A Consensus clustering cumulative distribution function. B Consensus clustering matrix with k = 4. C The proportion of ambiguous clustering analysis. D Kaplan–Meier survival curve of patients with colorectal cancer in different anoikis-related subtypes. E GSVA score analysis of patients with colorectal cancer in different anoikis-related subtypes. F Proportion of the four anoikis-related subtypes in different stages of colorectal cancer. G Proportion of the four anoikis-related subtypes in patients with colorectal cancer with respect to p53 mutation status. H Proportion of the four anoikis-related subtypes in patients with colorectal cancer with respect to BRAF mutation status

Fig. 3

Immune analysis. A Proportion of immune cell infiltration in the anoikis-related subtypes. B Expression patterns of immune checkpoint genes in patients with colorectal cancer belonging to different anoikis-related subtypes. C Expression patterns of human leukocyte antigen genes in patients with colorectal cancer belonging to different anoikis-related subtypes. *, **, ***, and **** represent P < 0.05, 0.01, 0.001, and 0.0001, respectively

Fig. 4

Gene set enrichment analysis. A The top 10 enriched pathways for up-regulated and down-regulated gene sets in the order of significance between subtype B and other subtypes (A, C, D). B Differential enrichment pathways in the four anoikis-related subtypes

Fig. 5

Construction of a risk model. A Univariate Cox regression analysis screened genes significantly associated with the prognosis of colorectal cancer. B Multivariate Cox regression analysis screened genes significantly associated with the prognosis of colorectal cancer. C Kaplan–Meier survival curve of patients with colorectal cancer in the high- and low-risk groups in the GEO test dataset. D Kaplan–Meier survival curve of patients with colorectal cancer in the high- and low-risk groups in the TCGA validation dataset

Fig. 6

Construction of a nomogram. A Univariate Cox regression analysis screened clinical features significantly associated with the prognosis of colorectal cancer. B Multivariate Cox regression analysis screened independently prognostic factors of colorectal cancer. C A nomogram was constructed using the independent prognostic factors, age, clinical M stage, and risk score. D Receiver operating characteristic curves of nomogram for 1-, 3-, and 5-year prognosis of patients with colorectal cancer. E Calibration curves of 1-year, 3-year, and 5-year prognosis of patients with colorectal cancer as predicted by the constructed nomogram and actual data

Fig. 7

Treatment sensitivity analysis. A Comparison of susceptibility to patients with colorectal cancer patients in the high- and low-risk groups to eight chemotherapeutic agents. B Comparison of TIDE scores in patients with colorectal cancer in the high- and low-risk groups. C Comparison of CD8A expression in patients with colorectal cancer in the high- and low-risk groups. D Comparison of cytolytic activity scores (CYT) in patients with colorectal cancer in the high- and low-risk groups. E Comparison of tertiary lymphoid structure (TLS) scores in patients with colorectal cancer in the high- and low-risk groups. F Comparison of CD274 expression in patients with colorectal cancer in the high- and low-risk groups. G Correlation analysis between genes used for model construction and immune-related indicators

Fig. 8

The distribution of anoikis-related subtypes in different risk groups by the Sanky plots