Metabolic network-based stratification of hepatocellular carcinoma reveals three distinct tumor subtypes

Abstract

Hepatocellular carcinoma (HCC) is one of the most frequent forms of liver cancer, and effective treatment methods are limited due to tumor heterogeneity. There is a great need for comprehensive approaches to stratify HCC patients, gain biological insights into subtypes, and ultimately identify effective therapeutic targets. We stratified HCC patients and characterized each subtype using transcriptomics data, genome-scale metabolic networks and network topology/controllability analysis. This comprehensive systems-level analysis identified three distinct subtypes with substantial differences in metabolic and signaling pathways reflecting at genomic, transcriptomic, and proteomic levels. These subtypes showed large differences in clinical survival associated with altered kynurenine metabolism, WNT/β-catenin-associated lipid metabolism, and PI3K/AKT/mTOR signaling. Integrative analyses indicated that the three subtypes rely on alternative enzymes (e.g., ACSS1/ACSS2/ACSS3, PKM/PKLR, ALDOB/ALDOA, MTHFD1L/MTHFD2/MTHFD1) to catalyze the same reactions. Based on systems-level analysis, we identified 8 to 28 subtype-specific genes with pivotal roles in controlling the metabolic network and predicted that these genes may be targeted for development of treatment strategies for HCC subtypes by performing in silico analysis. To validate our predictions, we performed experiments using HepG2 cells under normoxic and hypoxic conditions and observed opposite expression patterns between genes expressed in high/moderate/low-survival tumor groups in response to hypoxia, reflecting activated hypoxic behavior in patients with poor survival. In conclusion, our analyses showed that the heterogeneous HCC tumors can be stratified using a metabolic network-driven approach, which may also be applied to other cancer types, and this stratification may have clinical implications to drive the development of precision medicine.

Keywords: biological networks; genome-scale metabolic models; hepatocellular carcinoma; personalized medicine; systems biology.

Fig. 1.

Network-based approaches to identify driver genes involved in progression of HCC. (A) Radar plot of median node absolute deviation for betweenness, closeness, degree, and eccentricity indicates a larger variability in the HCC vs. noncancer networks. (B) Relation of degree centrality and controllability in cancer and noncancerous samples. Groups of genes were classified based on their dispensability (indispensable, dispensable, neutral) when identified in each category in >80% of the fGGNs for HCC and noncancerous networks. Indispensable genes tend to be more central than neutral or dispensable, in both HCC and noncancerous tissues (for all six comparisons, Q < 10⁻⁷, Mann–Whitney U test). For each group (indispensable vs. dispensable vs. neutral), we observed no statistical differences in degrees of HCC vs. noncancerous for the three tested comparisons (Q > 0.2). (C) Silencing of controlling genes leads to lethality in >95% of HCCs (vs. <50% for silencing of other genes). In noncancerous samples, silencing either kills all or none of the samples, where all controlling genes lead to lethality (48% for other genes). Both comparisons show statistically significant differences (Q < 10⁻¹⁰⁰, Mann–Whitney U test). (D) Principal component analysis of cancer and noncancer for controllability of fGNNs. Ellipses indicate 95% confidence regions (one outlier is identified at this confidence level).

Fig. 2.

Network-based approaches identified three different HCC subtypes. (A) Gene set-enriched biological processes (Q < 0.05) in different HCC subtypes including iHCC1, iHCC2, and iHCC3. Arrows indicate direction of change (e.g., iHCC2 shows up-regulated heme metabolism compared with iHCC1). (B) Kaplan–Meier survival analysis shows significant differences in patient survival between the three HCC subtypes (iHCC1 > iHCC2 > iHCC3). (C) Correlation plot between tumors and mean gene expression in iHCC1 and iHCC3 (Q < 0.01) showed that iHCC2 tumors tend to be more similar to iHCC1 than iHCC3. This is reinforced by the higher Euclidean distance between iHCC3 fGGNs and other fGGNs in this or the two other subtypes, compared with distances within or between iHCC1 and iHCC2 (SI Appendix, Fig. S6). EMT, epithelial-to-mesenchymal transition; ROS, reactive oxygen species.

Fig. 3.

HCC tumors stratified based on metabolic network analysis. (A) We determined three novel HCC subtypes and show the stratifying genes in iHCC1 (green), iHCC2 (cyan), and iHCC3 (orange). (B) Expression of stratifying genes and enzymes catalyzing the same reactions in the three iHCC groups is shown.

Fig. 4.

Coexpression analysis highlights the association between stratifying and controlling genes in iHCC subtypes. Stratifying and controlling genes for iHCC1, iHCC2, and iHCC3 and their top 25 coexpressed genes are included. Coexpression between iHCCs was determined based on TCGA transcriptomics data of 369 HCC tumor samples. We additionally included AKT1 and MTOR, transcription factors involved in PI3K/AKT/mTOR signaling, and CTNNB1, which encodes for the transcription factor β-catenin in the Wnt signaling pathway. Edges indicate positive (red) or negative (blue) Pearson correlations (Q < 0.01).

Fig. 5.

iHCC subgroups rely on alternative enzymes catalyzing the same reactions and display specific synthetic lethal genes. (A) Flux balance analysis performed on iHCC-specific models shows that iHCC1 or iHCC3 displays the highest reaction fluxes, followed by iHCC2. The predominant color in each box shows the iHCC subtype that displays the highest flux. (B) Metabolic genes involved in transport, glycolysis, and the TCA colored according to expression in each subgroup. (C) Numbers of synthetic lethal genes found in iHCC subgroups are shown, highlighting five synthetic lethal genes per subgroup. No synthetic lethal genes are simultaneously identified in iHCC1 and iHCC3, but several are found between iHCC2 and the other subgroups.

Fig. 6.

Stratifying and controlling genes in iHCC3 show specific responses to hypoxia. HepG2 cell lines were grown under normoxic or hypoxic conditions (n = 6 per condition) and transcriptomics data were generated. Expression of stratifying and controlling genes in iHCC3 (A), iHCC2 (B), and iHCC1 (C), and gene association with enriched biological processes (D; Q < 0.05). All genes with the exception of CYP3A4, GLUL, XDH, KMO, and TDO2 (Q > 0.01) are differentially expressed between hypoxia and normoxia. NMP, NDP, and NTP indicate nucleoside mono, di-, and triphosphate, respectively.