An asparagine metabolism-based classification reveals the metabolic and immune heterogeneity of hepatocellular carcinoma

Abstract

Introduction and objectives: hepatocellular carcinoma (HCC) is the major form of liver cancer with a poor prognosis. Amino acid metabolism has been found to alter in cancers and contributes to malignant progression. However, the asparagine metabolism status and relevant mechanism in HCC were barely understood.

Methods: By conducting consensus clustering and the least absolute shrinkage and selection operator regression of HCC samples from three cohorts, we classified the HCC patients into two subtypes based on asparagine metabolism level. The Gene Ontology, Kyoto Encyclopedia of Genes and Genomes analyses and Gene Set Enrichment Analysis of the differentially expressed genes between two subgroups were conducted. Immune cell infiltration was evaluated using CIBERSORT algorithm. The prognostic values of genes were analyzed by univariate and multivariate cox regression, ROC curve and Kaplan-Meier survival estimate analyses. Cell types of sing-cell RNA sequencing (scRNA-seq) data were clustered utilizing UMAP method.

Results: HCC patients with higher asparagine metabolism level have worse prognoses. Moreover, we found the distinct energy metabolism patterns, DNA damage response (DDR) pathway activating levels, drug sensitivities to DDR inhibitors, immune cell compositions in the tumor microenvironment and responses to immune therapy between two subgroups. Further, we identified a potential target gene, glutamic-oxaloacetic transaminase 2 (GOT2). GOT2 downregulation was associated with worse HCC prognosis and increased infiltration of T regulatory cells (Tregs). ScRNA-seq revealed the GOT2 downregulation in cancer stem cells compared with HCC cells.

Conclusions: Taken together, HCC subtype which is more reliant on asparagine and glutamine metabolism has a worse prognosis, and a core gene of asparagine metabolism GOT2 is a potential prognostic marker and therapeutic target of HCC. Our study promotes the precision therapy of HCC and may improve patient outcomes.

Keywords: Asparagine metabolism; DNA damage response; GOT2; Hepatocellular carcinoma; Tumor microenvironment.

Fig. 1

Identify asparagine metabolism subtypes of HCC in TCGA database and validation in ICGC database. A Consensus Cumulative Distribution Function (CDF) Plot and relative change in the area under the CDF curve (CDF Delta area). B Consensus matrices of the TCGA cohort for k = 2. C The expression of asparagine metabolism genes in two HCC subgroups. D, E OS and DFS analyses of the two HCC subgroups in the TCGA database. F A 9-gene signature was constructed for the HCC cases in the ICGC database by Lasso Cox analysis. Risk scores distribution, survival status of each patient in the ICGC database, and heatmaps of signature gene expression were plotted. G OS analysis of the two HCC subgroups in the ICGC database. H The 1-, 3- and 5-year ROC curves of the gene signature. The AUC was indicated. I The relationship between partial likelihood deviation and log (λ), and the LASSO coefficient profiles of the fractions of 9 genes were plotted

Fig. 2

Identification of differentially expressed genes (DEGs) between two subgroups in the TCGA database. A Volcano plot of the DEGs between high and low asparagine metabolism subgroups in the TCGA database. The threshold was set as |log2 Fold change|> 1 and p < 0.05. B Heatmap shows the gene expression profile in high and low asparagine metabolism subgroups. C, D Up-regulated GO and KEGG terms of the DEGs. E, F Down-regulated GO and KEGG terms of the DEGs. G GSEA analysis of the DEGs

Fig. 3

Expression comparison of metabolic-related genes between two HCC subgroups in the TCGA database. Glutamine metabolism genes (A), lipid metabolism genes (B) and TCA cycle enzyme complex genes (C) were acquired from Wikipathway, the expression of which in HCC subgroups were plotted. *p < 0.05; **p < 0.01; ***p < 0.001

Fig. 4

Signature validation using GSE84598 dataset in the GEO database and subgroup characteristics comparison. A The cumulative distribution function (CDF) curve and the delta area curve of consensus clustering. B Consistency of clustering results heatmap (k = 2). C Volcano plot of the DEGs between two HCC subgroups in the GSE84598 dataset. The threshold was set as |log2 Fold change|> 1 and p < 0.05. D Gene expression profile heatmap of two subgroups. E–H GO and KEGG analyses of DEGs

Fig. 5

DNA damage response pathway in the two subgroups and relative drug sensitivity analyses of HCC in the TCGA database. A The expressions of representative genes of DNA damage response pathway in high asparagine metabolism HCC subgroup were significantly higher than the low asparagine metabolism group. B The IC50 of two subgroups in response to sorafenib showed no difference. C The IC50 of high asparagine metabolism subgroup in response to CHEK1 inhibitor AZD7762 and PARP 1/2 inhibitor ABT-888 were lower than the low asparagine metabolism subgroup

Fig. 6

The immune landscape of high and low asparagine metabolism HCC subgroups in the TCGA database. A The immune cell infiltration in high and low asparagine metabolism subgroups. B Immune cell proportion in each HCC cases. C The expressions of most immune checkpoint molecules in high asparagine metabolism subgroup were significantly higher than the low asparagine metabolism group. D The TIDE score of the HCC cases predicted that high asparagine metabolism subgroup had better response to immune checkpoint blockers (ICBs) than low asparagine metabolism group

Fig. 7

Evaluation of the prognostic value of every single gene in the asparagine metabolism gene set. A, B The univariate and multivariate Cox regression of genes involved in asparagine metabolism in term of OS. C Nomograms predicting the 1-, 2- and 3-year OS of HCC based on the expression of ASPA, GOT2, NAALAD2 and SLC25A12. D Calibration curve for the OS nomogram model

Fig. 8

GOT2 down-expression predicts worse prognosis of HCC patients in the TCGA database. A The pan-cancer analysis of GOT2 expression. B The association between GOT2 expression and the OS, PFI, and DSS of HCC patients. C The Sankey diagram showing the distribution of age, pTNM stage, grade, GOT2 expression and survival status of HCC samples. D The ROC curve of GOT2 gene, with the AUC value indicated

Fig. 9

GOT2 expression and Treg cell infiltration. A Pan-cancer correlation analysis of the GOT2 expression and the infiltration of Treg cells. Correlation of GOT2 expression with OS (B) and RFS (C) in Treg enriched and decreased HCC subgroups

Fig. 10

GOT2 expression and core pathway differences in HCC cells and cancer stem cells. A, B The cell proportion and cell type distribution of HCC tissue. C GOT 2 expressions in HCC cells and cancer stem cells