Discovery of differentially expressed proteins for CAR-T therapy of ovarian cancers with a bioinformatics analysis

Abstract

Target antigens are crucial for developing chimeric antigen receptor (CAR)-T cells, but their application to ovarian cancers is limited. This study aimed to identify potential genes as CAR-T-cell antigen candidates for ovarian cancers. A differential gene expression analysis was performed on ovarian cancer samples from four datasets obtained from the GEO datasets. Functional annotation, pathway analysis, protein localization, and gene expression analysis were conducted using various datasets and tools. An oncogenicity analysis and network analysis were also performed. In total, 153 differentially expressed genes were identified in ovarian cancer samples, with 60 differentially expressed genes expressing plasma membrane proteins suitable for CAR-T-cell antigens. Among them, 21 plasma membrane proteins were predicted to be oncogenes in ovarian cancers, with nine proteins playing crucial roles in the network. Key genes identified in the oncogenic pathways of ovarian cancers included MUC1, CXCR4, EPCAM, RACGAP1, UBE2C, PRAME, SORT1, JUP, and CLDN3, suggesting them as recommended antigens for CAR-T-cell therapy for ovarian cancers. This study sheds light on potential targets for immunotherapy in ovarian cancers.

Keywords: CAR-T-cell antigen; chimeric antigen receptor (CAR)-T cell; differentially expressed gene (DEG); ovarian cancer; protein-protein interaction (PPI) network.

Figure 1

Determination of gene set enrichments (GSEs) and differentially expressed genes (DEGs). (A) Volcano plot of gene distributions in control and ovarian cancer samples. Gray dots represent genes that are not significantly expressed between normal and ovarian cancer cell samples. (B) Venn diagram of overlapping gene between four GSEs from which we obtained 153 DEGs. Also, the highest unique genes are GSE36668, GSE14407, and GSE27651. (C) List of DEGs. (D) Pathway terms related to DEGs, colored by –log10(p-values). (E) Gene ontology terms related to DEGs, colored by –log10 (p-values). (F) Plasma membrane-related genes from six GO terms (p > 0.05) with localization confidence scores. A higher score means a greater probability that the protein will be situated there.

Figure 2

Oncogenicity analysis. (A) Twenty-one genes were significantly expressed in ovarian cancer among plasma membrane-related genes. (B) Plasma membrane-related genes expressions in ovarian cancer. (C) Overlap between gene lists: high expression gene list (Predicted) and Control gene list, (D) including the shared term level, where blue curves connect genes that share an enriched ontology term. Gene lists are represented by the inner circle, with hits arranged along the arc. Genes that appear on multiple lists are depicted in dark orange, while genes that only appear on one list are displayed in light orange. (E) Heatmap of enriched terms across input gene lists, colored by p-values. (F) Heatmap of biological processes across input gene lists, colored by p-values. (G) Pattern genes related to predicted and control genes. (H) DisGeNET is a discovery platform containing one of the largest publicly available collections of genes and variants associated with human diseases. The heatmap is colored by p-values. Dark orange indicates a greater probability that bioactivity will occur.

Figure 3

Protein-protein interaction (PPI) network analysis. (A) PPI network of 37 enriched genes (20 control and 17 predicted genes) with 100 genes, which were analyzed with NetworkAnalyzer and CytoHubba for node and edge scoring. Nine genes with the highest betweenness-centrality scores were designated as TAAr. (B) Clustered network with the Molecular Complex Detection (MCODE) algorithm. (C) Gene ontology (GO) terms related to control and predicted genes. (C) GO term related to control and predicted genes. (D) Interaction network of the TAAr gene: yellow genes are TAAr, and blue genes are interactor proteins. (STRING database, high confidence: 0.7). (E) GeneMania network of TAAr. The nine biggest nodes with shading are TAAr genes. Edge width is labeled for confidence scores.

Figure 4

Correlation, co-expression, and pathway analysis of TAAr. (A) Co-expressed genes are colored by a dot; a darker dot means a higher co-expression score. Co-expression score: the higher the score, the higher the probability that co-expression will occur. (B) Spearman’s correlation scores for gene expressions in ovarian cancer (OC), which are comprised of 27 proteins consisting of nine TAAr and 18 control genes. (C) The complete pathway of the Reactome database related to TAAr. This pathway was constructed with the Voronoi tessellation method termed ReacFoam, which provides user-friendly access and visualization. (D) Pathway related to reproduction where the specification of primordial germ cells pathway is located. (E) Rho GTPase-related and neighbor pathways. (F) The expression data of TAAr in brain, retina, and liver tissue, provided by GeneCard, were used to analyze the possible toxic effects of targeting TAAr.

Figure 5

Gene expression levels are based on single-cell RNA sequencing based on a human sample based on the TISCH2 database. (A) Sample from GSE154600 based on 5 non-treatment patients, 42583 cells, analyzed with the 10x genomics platform. (B) UMAP plot of nine TAAr on GSE154600, where EPCAM, MUC1, RACGAP1, CLDN3, and JUP are highly expressed in malignant cell samples. (C) Sample from GSE118828 based on 9 non-treatment patients, 1909 cells, analyzed with the Smart-Seq2 platform. (D) UMAP plot of nine TAAr on GSE118828, where CXCR4, MUC1, EPCAM, UBE2C, CLDN3, JUP, PRAME, and SORT1 are highly expressed in malignant cell samples. (E) Violin diagram of TAAr expression in each sample cell in GSE118828. (F, G) Diversification and comparison of the number of samples analyzed in GSE118828. (H) Violin diagram of TAAr expression in each sample cell on GSE154600. (I, J) Diversification and comparison of the number of samples analyzed on GSE154600. (K) Transcription factor induced by TAAr in GSE118828: the higher the induction, the redder it is. (L) Number of cell-cell interactions in sample GSE118828. (M) Transcription factor induced by TAAr in GSE154600: the higher the induction, the redder it is. (N) Number of cell-cell interactions in the GSE154600 sample. (O, P) The genetical hallmarks that are up-downregulated based on the TAAr gene in GSE118828. (Q, R) The genetical hallmarks that are up-downregulated based on the TAAr gene in GSE154600.

Figure 6

Research stages and workflow. The four research red lines are differential expression analysis, differentially expressed genes (DEGs), protein localization analysis, oncogenicity analysis, and pathway and gene ontology analyses. Therefore, in this study, we carried out multilevel screening to reduce the potential for errors or discrepancies later. What we mean by multilevel screening is first looking at the significance of the expression of a gene in normal and cancer samples from the dataset, then checking the significance of the significant gene again in a different database to ensure that the gene is significantly expressed only in cancer cells.