Construction and validation of a prognostic model based on cuproptosis-related genes in patients with multiple myeloma

To explore the potential cuproptosis-related genes (CRGs) in patients with multiple myeloma (MM) and develop a prognostic model for improving prognosis and revealing features of the MM immune microenvironment.

① Transcriptome sequencing data and clinical information were retrieved from the GSE4581 dataset in the Gene Expression Omnibus (GEO) database and the Cancer Genome Atlas-Multiple Myeloma Research Foundation (TCGA-MMRF) database. The 859 patients from the TCGA-MMRF database were assigned into a training set, and the other 414 ones from the GSE4581 dataset into a validation set. LASSO-Cox and multivariate Cox regression analyses were used to construct prognostic models and calculate risk scores. Based on the median risk score, they were categorized into high- and low-risk cohorts. Time-dependent receiver operating characteristic (ROC) and calibration curves were plotted to assess the predictive performance and accuracy of the model. The differences between the high- and low-risk cohorts were explored using Kaplan-Meier survival curve analysis and immune microenvironment correlation analysis. ② RT-qPCR and Western blotting were used to verify the expression of prognostic model genes in MM cell lines and normal bone marrow single-nucleated cells, and CCK-8 assay, flow cytometry, and Western blotting were applied to verify the biological function of UBE2D1 in MM cells.

① LASSO-Cox and multivariate Cox regression analyses revealed that the model consisted of 4 genes, CDKN2A [HR=1.60 (95%CI: 1.24~2.05), P=2.5e-4], PDE3B [HR=1.33 (95%CI: 1.09~1.62), P=4.2e-3], UBE2D1 [HR=1.65 (95%CI: 1.20~2.26), P=2.1e-3] and COA6 [HR=1.35 (95%CI: 1.07~1.71), P=0.01]. In the training set, the time-dependent ROC curves predicted that the area under curve (AUC) value of 1-, 3-, and 5-year survival rate was 0.63, 0.71, and 0.78, respectively, and in the validation set, the AUC value was 0.656, 0.657, and 0.797, respectively. Calibration curve analysis showed excellent agreement in predicting 1-, 3-, and 5-year prognosis. In the training set, Kaplan-Meier curves showed that patients in the high-risk cohort had a significantly shorter overall survival (OS) than the low-risk cohort [HR=2.18 (95%CI: 1.58~3.02), P<0.001], and in the validation set, the high-risk cohort still had a shorter OS than the low-risk cohort [HR=2.45 (95%CI: 1.49~4.05), P<0.001]. Immune correlation analysis revealed that the ratios of immune cells, such as plasma cells and CD4 T cells were significantly lower in the high-risk cohort (P<0.05), and the risk scores were positively correlated with the expression of immune checkpoint CTLA-4, tumor-targeted therapeutic sites TNFSF4 and ENTPD1, and microenvironmental chemokines CXCL16, CCL8, and CCL16 (P<0.05). ② Remarkable differences were observed in the expression of all 4 prognostic model genes between the MM cell lines and normal bone marrow single-nucleated cells (P<0.05), and knockdown of UBE2D1 notably inhibited the proliferation of MM cells (P<0.05).

Our prognostic models based on CDKN2A, PDE3B, UBE2D1, and COA6 genes can predict the prognosis of MM patients. The risk scores of the genes are significantly correlated with immune infiltration in the tumor microenvironment, which providing new molecular markers for individualized therapy.