Global tissue and serum metabolomics reveals altered tumor biochemical activity in molecular subtypes of hepatocellular carcinoma and cholangiocarcinoma

Objective: Cancer metabolite profiling (or cancermetabolomics), the global view of the biochemical endproducts of cellular processes, is a promising approach toidentify biomarkers and tumor subgroups, and to understandthe biological mechanisms underlying cancer developmentand progression. Metabolites are the best molecular indicatorsof cell status, since their rapid fluxes versus those of mRNAand proteins are an extremely sensitive measure of cellularphenotype. Hepatocellular carcinoma (HCC) and intrahepaticcholangiocarcinoma (ICC) represent two distinct histologicalprimary liver cancer (PLC) subtypes, which have beenextensively profiled by genomics-based studies. While theliver is a hub of metabolic processes, relatively little is knownabout the global metabolite alterations in PLC and how tumormetabolic networks contribute to aggressive disease and pooroutcome. This study determined metabolic tumor subtypes anddiagnostic/prognostic biomarkers of HCC and ICC throughglobal metabolic profiling and system integration of theirtranscriptomes within the Thailand Initiative in Genomics andExpression Research for Liver Cancer (TIGER-LC) cohort. Methods: Genome-wide profiling was performed in 398surgical specimens derived from 199 Thai liver cancerpatients. Metabolonu0027s DiscoveryHD4 platform was employed toidentify cancer metabolic profiles in tissue and paired serumsamples. These data were integrated with paired tissue andnon-tumor tissue transcriptome profiles from the AffymetrixHuman Transcriptome Array 2.0. Class comparison, principalcomponent analysis (PCA), Metabolite Set Enrichment Analysis(MSEA), Unsupervised Consensus Clustering (cCluster),Subclass Mapping (SM), and Pearson and rank correlationalgorithms were used to analyze omics data. The results werevalidated in independent liver cancer cohorts. Results: Unsupervised PCA and supervised class comparisonanalysis showed that among the 718 or 990 measurablemetabolites in tissue or serum samples, respectively, specificmetabolites can distinguish tumor from paired non-tumortissues in HCC or ICC and can identify different PLC types.MSEA analysis indicated that a larger number of metabolicpathway perturbations occur in ICC versus HCC. Integrationanalysis of metabolite abundance and metabolic genes revealed48 metabolites and 461 surrogate correlated genes that canstratify PLC patient subgroups. Pathway analysis of the 461 genes revealed enrichments in bile acid and amino acidbiosynthesis. The metabolite-defined patient subgroups weresimilar to the C1 and C2 transcriptome-defined subgroups ofPLC, previously identified by our team, which differed in bodymass index status and were associated with alterations in bileacid metabolism. These results were validated in two additionalPLC cohorts, LCI (n = 247) and TCGA (n = 314). Among the48 metabolites, 29 were tumor-specific and correlated with 120serum metabolites related to patient outcome. Conclusions: Global genomic profiling of HCC and ICC tissueand serum samples revealed metabolic tumor subgroups relatedto patient prognosis. These specific metabolites and genesurrogatescould serve as novel biomarkers for identifying,distinguishing, and classifying primary liver tumors and allowfor the development of targeted therapeutic interventions. DOI: 10.20892/j.issn.2095-3941.2018.S112