An amino acid metabolism-based seventeen-gene signature correlates with the clinical outcome and immune features in pancreatic cancer

Abstract

Background: Pancreatic cancer is an aggressive tumor with a low 5-year survival rate and primary resistance to most therapy. Amino acid (AA) metabolism is highly correlated with tumor growth, crucial to the aggressive biological behavior of pancreatic cancer; nevertheless, the comprehensive predictive significance of genes that regulate AA metabolism in pancreatic cancer remains unknown. Methods: The mRNA expression data downloaded from The Cancer Genome Atlas (TCGA) were derived as the training cohort, and the GSE57495 cohort from Gene Expression Omnibus (GEO) database was applied as the validation cohort. Random survival forest (RSF) and the least absolute shrinkage and selection operator (LASSO) regression analysis were employed to screen genes and construct an AA metabolism-related risk signature (AMRS). Kaplan-Meier analysis and receiver operating characteristic (ROC) curve were performed to assess the prognostic value of AMRS. We performed genomic alteration analysis and explored the difference in tumor microenvironment (TME) landscape associated with KRAS and TP53 mutation in both high- and low-AMRS groups. Subsequently, the relationships between AMRS and immunotherapy and chemotherapy sensitivity were evaluated. Results: A 17-gene AA metabolism-related risk model in the TCGA cohort was constructed according to RSF and LASSO. After stratifying patients into high- and low-AMRS groups based on the optimal cut-off value, we found that high-AMRS patients had worse overall survival (OS) in the training cohort (a median OS: 13.1 months vs. 50.1 months, p < 0.0001) and validation cohort (a median OS: 16.2 vs. 30.5 months, p = 1e-04). Genetic mutation analysis revealed that KRAS and TP53 were significantly more mutated in high-AMRS group, and patients with KRAS and TP53 alterations had significantly higher risk scores than those without. Based on the analysis of TME, low-AMRS group displayed significantly higher immune score and more enrichment of T Cell CD8⁺ cells. In addition, high-AMRS-group exhibited higher TMB and significantly lower tumor immune dysfunction and exclusion (TIDE) score and T Cells dysfunction score, which suggested a higher sensitive to immunotherapy. Moreover, high-AMRS group was also more sensitive to paclitaxel, cisplatin, and docetaxel. Conclusion: Overall, we constructed an AA-metabolism prognostic model, which provided a powerful prognostic predictor for the clinical treatment of pancreatic cancer.

Keywords: amino acid metabolism; chemosensitivity; genomic alterations; immunotherapy; pancreatic cancer; prognosis; tumor microenvironment.

FIGURE 1

The expression landscape of AA-metabolism-related genes. (A) The volcano plot showed downregulated and upregulated AA metabolism-related DEGs. (B) The heatmap displayed DEGs were divided into high- and low-expression groups. (C) KEGG analysis of DEGs. (D) Reactome analysis of DEGs. AA: amino acids, DEGs: differentially expressed genes, KEGG: Kyoto Encyclopedia of Genes and Genomes.

FIGURE 2

Construction and verification of AMRS in pancreatic cancer. (A) 42 combinations of machine learning algorithms of AMRS and the Concordance-index were calculated through TCGA and GSE57495 cohorts. (B) Identification of hazard factors and protective factors by multivariate Cox regression analysis. (C) Risk score distribution and survival status in TCGA cohort. (D) Box plot showing the expression level of 17 genes between high- and low-AMRS group. (E) Survival analysis and ROC curve for predicting OS of 12-, 24-, and 36-month in TCGA cohort. (F) Kaplan–Meier curve and ROC curve for predicting OS at 12-, 24-, and 36-month in GSE57495 cohort. (G) The distribution of Moffitt subtypes and PurIST subtypes in high- and low-AMRS group. (H) AUC value for the AMRS, Moffitt subtypes and PurIST subtypes in TCGA cohort. AMRS: amino acid metabolism-related risk score; TCGA: The Cancer Genome Atlas; ROC: receiver operating characteristic; OS: overall survival; AUC: Area Under the ROC Curve.

FIGURE 3

Construction of a Nomogram in pancreatic cancer. (A) Univariate Cox regression analysis. (B) Multivariate Cox regression analysis. (C) Nomogram for predicting 12-, 24-, and 36-month of OS in pancreatic cancer patients in TCGA cohort. (D) ROC curve of the nomogram. Calibration curves for predicting the fitness of the nomogram in 12 months, 24 months, and 36 months.E

FIGURE 4

Functional and pathway enrichment analysis of AMRS. (A) GO analysis. (B) KEGG analysis. (C) GSEA analysis. GO: Gene Ontology, GSEA: gene set enrichment analysis.

FIGURE 5

Overview of mutated genes in pancreatic cancer. Waterfall plot of somatic mutation in high AMRS group (A) and low AMRS group (B). (C) The forest plot shows the 27 genes with the highest differences between the high- and low-AMRS groups. Heat map of mutually exclusive and co-occurring genes in the high-AMRS group (D), and low AMRS groups (E).

FIGURE 6

KRAS and TP53 mutation in pancreatic cancer. The risk score of MT and WT in mutation status: KRAS (A) and TP53 (B). The proportion of MT and WT between high- and low-AMRS groups in KRAS (C) and TP53 (D). (E) Survival analysis of MT and WT of KRAS between high- and low-AMRS groups. (F) Survival analysis of MT and WT of TP53 between high- and low-AMRS groups. MT: mutation, WT: wild-type.

FIGURE 7

TME analysis of AMRS. (A) Heat map of the distribution of immune cell infiltrations in high- and low-AMRS groups. (B) Immune score of MT and WT in KRAS status between high- and low-AMRS groups. (C) Immune score of MT and WT in P53 status between high- and low-AMRS groups. (D) The expression level of 22 tumor infiltrating cells between the two groups. (E) Immune checkpoint gene expression level between the two groups. TME: tumor microenvironment.

FIGURE 8

Immune cell infiltration landscape of mutant genes KRAS and TP53. Correlation analysis between 22 tumor infiltrating cells and WT, MT of KRAS (A) and TP53 (B) in high- and low-AMRS groups. Box plots showing the 22 tumor infiltrating cells expression in MT and WT subgroups of KRAS (C) and TP53 (D).

FIGURE 9

Prediction of the immunological therapeutic benefits in high- and low-AMRS groups. (A) Comparison of TMB between high- and low-AMRS groups. (B) Kaplan–Meier survival curve in the IMvigor210 cohort. (C) The proportion of different immunotherapy responses in high- and low-AMRS groups. (D) Analysis of the proportions of four different immunotherapy responses by both TMB groups and the AMRS model. (E) Risk score distribution for four different immunotherapy outcomes. (F) The difference in AMRS between CR/PR and SD/PD. (G) Comparison of risk scores in DCB and PD. (H) Difference in risk score among patients with three types of treatment response in the GSE63557 cohort. The distribution of TIDE score (I), T Cells dysfunction score (J), and T Cells exclusion score (K). The relationship between AMRS between TIDE score (L), T Cells dysfunction score (M), and T Cells exclusion score (N). CR/PR: complete response/partial response, SD/PD: stable disease/progressive disease, DCB: durable clinical benefit, TIDE: tumor immune dysfunction and exclusion.

FIGURE 10

Relationships between AMRS and chemotherapeutic drug sensitivity. (A) IC50 comparison of four chemotherapy drugs in high- and low-AMRS groups. Pathways analysis between AMRS and gemcitabine resistance DN (B), cisplatin resistance UP (C), and cisplatin response and XPC UP (D).