Abstract
Lipid metabolism rearrangements in nonalcoholic fatty liver disease (NAFLD) contribute to disease progression. NAFLD has emerged as a major risk for hepatocellular carcinoma (HCC), where metabolic reprogramming is a hallmark. Identification of metabolic drivers might reveal therapeutic targets to improve HCC treatment. Here, we investigated the contribution of transcription factors E2F1 and E2F2 to NAFLD-related HCC and their involvement in metabolic rewiring during disease progression. In mice receiving a high-fat diet (HFD) and diethylnitrosamine (DEN) administration, E2f1 and E2f2 expressions were increased in NAFLD-related HCC. In human NAFLD, E2F1 and E2F2 levels were also increased and positively correlated. E2f1−/− and E2f2−/− mice were resistant to DEN–HFD-induced hepatocarcinogenesis and associated lipid accumulation. Administration of DEN–HFD in E2f1−/− and E2f2−/− mice enhanced fatty acid oxidation (FAO) and increased expression of Cpt2, an enzyme essential for FAO, whose downregulation is linked to NAFLD-related hepatocarcinogenesis. These results were recapitulated following E2f2 knockdown in liver, and overexpression of E2f2 elicited opposing effects. E2F2 binding to the Cpt2 promoter was enhanced in DEN–HFD-administered mouse livers compared with controls, implying a direct role for E2F2 in transcriptional repression. In human HCC, E2F1 and E2F2 expressions inversely correlated with CPT2 expression. Collectively, these results indicate that activation of the E2F1–E2F2–CPT2 axis provides a lipid-rich environment required for hepatocarcinogenesis.
These findings identify E2F1 and E2F2 transcription factors as metabolic drivers of hepatocellular carcinoma, where deletion of just one is sufficient to prevent disease.
Graphical Abstract
Introduction
Hepatocellular carcinoma (HCC) is the most common primary liver cancer (1), being the fifth most frequent malignancy in men and the seventh in women globally (2). During the last decade, nonalcoholic fatty liver disease (NAFLD), which affects approximately 80% of obese patients, has emerged as an important risk factor for cancer, including HCC (3). The therapeutic options for HCC are still very limited (4) and there is no specific pharmacological treatment for NAFLD. NAFLD-driven HCC has been associated with shorter survival time and more advanced tumor stage (5).
NAFLD begins with the simple storage of lipids within hepatocytes, which may progress to hepatocellular ballooning and cell death, as well as to inflammation, leading to the development of steatohepatitis (nonalcoholic steatohepatitis, NASH). Mechanisms involved in NAFLD progression are still not fully understood. However, lipotoxicity derived from the accumulation of certain lipids may induce inflammation and endoplasmic reticulum stress (6), both processes involved in obesity-driven HCC (7).
A shared feature of HCC and NAFLD is the reprogramming in lipid metabolism. In cancer cells, lipids are required as membrane components, signaling molecules, and energy storage, which allow their growth and proliferation. Many lipids are synthesized from acetyl-CoA and/or the derived fatty acids (FA). The relevance of FA biosynthesis for cancer cell growth and survival is well established (8). In HCC, in addition to the de novo synthesis of FAs (9), exogenous FAs also support the growth of HCC cells (10). The relevance of lipid catabolism in cancer cells is less clear. In obesity-driven HCC, fatty acid oxidation (FAO) is decreased as the result of the carnitine palmitoyltransferase 2 (CPT2) downregulation, which contributes to carcinogenesis (11). Overall, it is well established that the metabolic switch for malignant cell transformation requires the coordinated modulation of several enzymes and transporters to ensure aberrant proliferation (12). Therefore, the search for pivotal master regulators of lipid metabolism may provide new insights into the mechanisms of liver pathogenesis and novel targets for therapy.
Cell-cycle regulators play a dual role triggering proliferation and adapting metabolism in response to external stimuli (13). Some of them have also been described as modulators of metabolism in non-proliferative cells such as p53, whose deficiency is associated with hepatosteatosis (14) or E2F1 (15–17), the most studied factor of the E2F family.
Here, we investigated the contribution of the cell-cycle regulator E2F2 to NAFLD-related HCC by using in vitro and in vivo models, as well as human samples. Metabolic fluxes, chromatin immunoprecipitation (ChIP), and transcriptomic analyses were performed. Our study unveils a critical function for E2F2 in the regulation of key metabolic enzymes that mediate NAFLD-related hepatocarcinogenesis. Furthermore, we show that the regulation of this metabolic pathway is shared with E2F1, arguing for both E2F1 and E2F2 factors as novel putative therapeutic targets in NAFLD-related liver cancer.
Materials and Methods
Human samples
In this study, 78 liver samples were included: non-obese healthy human liver (NL) samples (n = 18) from donors (Marqués de Valdecilla University Hospital, Santander, Spain) and liver samples from obese patients (n = 60; Supplementary Table S1) who underwent a liver biopsy with a diagnostic purpose (Cruces University Hospital, Barakaldo, Spain). NAFLD was diagnosed according to Kleiner's criteria (18). Written consent was obtained from each patient included in the study. This research project was performed in agreement with the Declaration of Helsinki and with local and national laws. The Human Ethics Committee of each hospital and the University of Basque Country approved the study procedures and a written informed consent was obtained before inclusion in the study.
E2F1 and E2F2 expression analysis in TGCA-LIHC database
Level 3 RNA-seq and clinical datasets generated by The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC) project were downloaded from Firehose portal (Broad Institute). In total, 371 tumors and 50 surrounding normal tissues were analyzed. For correlation studies, data normality of genes expression values was assessed by the D'Agostino and Pearson normality test. Subsequently, matched expression levels were evaluated by Spearman correlation.
Animal models
Male E2f1 knockout (E2f1−/−), E2f2 knockout (E2f2−/−), E2f1/E2f2 double knockout (DKO), and their respective wild-type (WT) littermate control mice (mixed background C57BL/6J and 129/Sv) were produced at the UPV/EHU animal facility. C57BL/6J mice (The Jackson Laboratory) were used for E2f2 knockdown or overexpression in the liver. Animal procedures were approved by the Ethics Committee for Animal Welfare of the University of the Basque Country UPV/EHU, in accordance with the European Union Directives for animal experimentation.
HCC, with or without NAFLD background, was induced by an intraperitoneal injection of the hepatic carcinogen diethylnitrosamine (DEN; 25 mg/kg of mice, Sigma-Aldrich) in 14 day-old E2f1−/− or E2f2−/−, and WT mice (7). One month after weaning, mice were fed a high-fat diet (HFD; DEN–HFD) or a chow diet (CD; DEN–CD) for either 32 weeks and sacrificed at 9 months of age (HCC or NAFLD-related HCC model) or for 10 weeks and sacrificed at 3 months of age (NAFLD model). A group of mice fed with HFD or CD alone was also included.
Liver-specific E2f2 knockdown or overexpression was performed in C57BL/6J mice through the injection of recombinant adeno-associated viruses, serotype 8 (AAV8-shE2F2/AAV8-shScramble for knockdown and AAV8-E2f2/AAV8-GfP for overexpression), diluted in saline buffer. The in vivo Creb downregulation was performed with antisense oligonucleotides (ASO). For this, DEN-HFD E2f2−/− and DEN-HFD WT mice received an injection of ASO CREB or ASO control (Ionis Pharmaceuticals).
Statistical analysis
Data are represented as mean ± SEM. Differences between groups were analyzed with the Student t test. Significance was defined as P < 0.05. For correlation studies in human samples, data normality was assessed by the D'Agostino and Pearson normality test. Subsequently, matched gene expression levels were evaluated by Spearman correlation. See figure legends for more details.
Data availability
The datasets and computer code produced in this study are available in the following databases:
—RNA-seq data: Gene expression Omnibus GSE117420
Additional methods are described in Supplementary Material.
Tables with the sequences of oligonucleotides (Supplementary Table S2) and the dilutions and references of antibodies (Supplementary Table S3) are included in the Supplementary Material.
Results
E2F2 promotes hepatocarcinogenesis
Expression of E2F2 in human HCC was analyzed using the TCGA-LIHC cohort (19). We found that E2F2 is upregulated in HCC tissue compared with the surrounding normal liver (SL) and that, in general, its expression increases with tumor and disease stage (Supplementary Fig. S1A), consistent with previous reports (20, 21). We next evaluated whether E2F2 is required for HCC development. For this purpose, HCC tumors were induced in WT and E2f2−/− mice through the administration of DEN followed by a long-term HFD feeding regimen (a well-established NAFLD-driven HCC model) for 32 weeks (Supplementary Fig. S1B; ref. 7). Similar to our observations in human HCC, the expression of E2f2 was found upregulated in mouse HCC tumors compared with normal liver tissue (Fig. 1A). Strikingly, E2f2−/− mice (Supplementary Fig. S1C) were protected against hepatocarcinogenesis, as a nearly complete reduction of liver tumors induced by HFD, by DEN or by DEN–HFD was observed in these mice compared with controls (Fig. 1B). Moreover, liver-to-body weight ratio, serum alanine aminotransferase (ALT) levels (Fig. 1C), and a reduced expression of several E2F target genes involved in cell-cycle progression (Supplementary Fig. S1D and S1E) confirmed the protective features of E2f2−/− mice to NAFLD-driven HCC development.
E2F2 is required for hepatocellular carcinoma development. Mice were treated for 32 weeks to induce hepatocarcinogenesis and were sacrificed at 9 months of age. A, Analysis of hepatic E2f2 mRNA levels (n = 6–8 mice per group). B, Representative photographs from WT and E2f2−/− mouse livers (left) and quantification of tumor number and size (right; n = 7–9 mice per group). C, Liver-to-body weight ratio (n = 6–9 mice per group) and ALT levels in liver homogenates (n = 6–9 mice per group). Values represent mean ± SEM. Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN–HFD and CD are denoted as #, P < 0.05; ##, P < 0.01; ###, P < 0.001.
E2F2 is required for hepatocellular carcinoma development. Mice were treated for 32 weeks to induce hepatocarcinogenesis and were sacrificed at 9 months of age. A, Analysis of hepatic E2f2 mRNA levels (n = 6–8 mice per group). B, Representative photographs from WT and E2f2−/− mouse livers (left) and quantification of tumor number and size (right; n = 7–9 mice per group). C, Liver-to-body weight ratio (n = 6–9 mice per group) and ALT levels in liver homogenates (n = 6–9 mice per group). Values represent mean ± SEM. Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN–HFD and CD are denoted as #, P < 0.05; ##, P < 0.01; ###, P < 0.001.
Metabolic rearrangements in E2f2−/− mice confer protection against hepatocarcinogenesis by preventing lipid storage
A hallmark of NAFLD-related HCC is the lipid accumulation inside tumor hepatocytes (22). Remarkably, the increased storage of triglycerides (TG) and cholesteryl esters (CE) observed in NAFLD-driven HCC samples from WT mice was totally prevented in E2f2−/− mice (Fig. 2A). Furthermore, the lipid droplets found in livers of DEN–HFD WT mice, which are mainly located within tumor hepatocytes, were absent in similarly treated E2f2−/− mice (Fig. 2B). These data indicate an important role for E2F2 in the regulation of the hepatic lipid metabolism.
Rewiring of lipid metabolism in E2f2−/− mice prevents the lipid-rich environment for NAFLD-related HCC development. Mice were treated for 32 weeks to induce hepatocarcinogenesis and were sacrificed at 9 months of age. A, Liver TG and CE quantification after 32 weeks of treatment (n = 4–9 mice per group). B, Representative liver sections stained with hematoxylin and eosin. C, RNA was extracted from E2f2−/− and WT liver homogenates and a gene expression microarray analysis was carried out (n = 4 mice per group). Differential gene expression between both genotypes was analyzed and significant probes were assigned into pathways with Reactome database. Biological pathways encompassing upregulated genes (P ≤ 0.001) in DEN–HFD E2f2−/− mice vs. DEN–HFD WT mice are shown (–log2 of 0.001 is indicated with a red line). D, Hepatic mRNA levels of genes involved in lipid oxidation (n = 7–8 mice per group). E, Fatty acid β-oxidation rate was determined in liver homogenates by measuring the amount of [14C]-CO2 (complete oxidation of [14C]-palmitate) and [14C]-ASM (incomplete oxidation of [14C]-palmitate; n = 4 mice per group). F, Lipid content analysis in HepG2 cells upon E2F2 knockdown with siRNAs. Lipids were labeled with Bodipy Green dye and cell nucleus with DAPI. Representative confocal micrographs of cells with or without oleic acid are shown (left). Right, cell TG content (n = 4–5 per group). G, Fatty acid oxidation rate was determined in HepG2 cell cultures by measuring the amount of [14C]-CO2 and [14C]-ASM released by cells (n = 4 per group). H, Cell cycle distribution analysis in HepG2 cells transfected with siE2F2 compared with siControl transfected cells (n = 6). Values represent mean ± SEM. Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN-HFD and CD are denoted as #, P < 0.05; ##, P < 0.01; ###, P < 0.001. ASM, acid soluble metabolites.
Rewiring of lipid metabolism in E2f2−/− mice prevents the lipid-rich environment for NAFLD-related HCC development. Mice were treated for 32 weeks to induce hepatocarcinogenesis and were sacrificed at 9 months of age. A, Liver TG and CE quantification after 32 weeks of treatment (n = 4–9 mice per group). B, Representative liver sections stained with hematoxylin and eosin. C, RNA was extracted from E2f2−/− and WT liver homogenates and a gene expression microarray analysis was carried out (n = 4 mice per group). Differential gene expression between both genotypes was analyzed and significant probes were assigned into pathways with Reactome database. Biological pathways encompassing upregulated genes (P ≤ 0.001) in DEN–HFD E2f2−/− mice vs. DEN–HFD WT mice are shown (–log2 of 0.001 is indicated with a red line). D, Hepatic mRNA levels of genes involved in lipid oxidation (n = 7–8 mice per group). E, Fatty acid β-oxidation rate was determined in liver homogenates by measuring the amount of [14C]-CO2 (complete oxidation of [14C]-palmitate) and [14C]-ASM (incomplete oxidation of [14C]-palmitate; n = 4 mice per group). F, Lipid content analysis in HepG2 cells upon E2F2 knockdown with siRNAs. Lipids were labeled with Bodipy Green dye and cell nucleus with DAPI. Representative confocal micrographs of cells with or without oleic acid are shown (left). Right, cell TG content (n = 4–5 per group). G, Fatty acid oxidation rate was determined in HepG2 cell cultures by measuring the amount of [14C]-CO2 and [14C]-ASM released by cells (n = 4 per group). H, Cell cycle distribution analysis in HepG2 cells transfected with siE2F2 compared with siControl transfected cells (n = 6). Values represent mean ± SEM. Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN-HFD and CD are denoted as #, P < 0.05; ##, P < 0.01; ###, P < 0.001. ASM, acid soluble metabolites.
We next analyzed the transcriptome that underlies the resistance to develop HCC in DEN–HFD-treated E2f2−/− mice. The most significantly upregulated pathway within the E2f2−/− mice compared with WT was that concerning to “metabolism,” which included genes related to lipid catabolism (Fig. 2C). Among the downregulated genes, the most significantly altered pathways were “Immune system,” “Signaling by Rho GTPases,” and “Platelet activation, signaling, and aggregation” (Supplementary Fig. S2A), which are linked to HCC development, as described recently (23). For the validation of these results, we selected genes involved in lipid oxidation (Fig. 2D; Supplementary Fig. S2B) and oxidative phosphorylation (Supplementary Fig. S2B). The expression of several key genes involved in fatty acid β-oxidation (FAO), typically downregulated in NAFLD-driven HCC in WT mice, were increased in DEN–HFD E2f2−/− mice (Fig. 2D; Supplementary Fig. S2B) together with an elevated FAO rate (Fig. 2E). Similar results were obtained when we compared FAO rates and the mRNA expression levels between E2f2−/− and WT mice fed HFD alone (Supplementary Fig. S2C and S2D). In concordance, a comparison of the expression levels of genes involved in cell cycle, lipid oxidation, and oxidative phosphorylation in CD, HFD, and DEN–HFD-treated E2f2−/− and WT mice showed that E2f2 deficiency results in changes with the same trend in the DEN–HFD and the HFD groups (Supplementary Fig. S2E).
In line with these findings, E2F2 knockdown (siE2F2) in cycling HepG2 cancer cells (Supplementary Fig. S3A) reduced significantly lipid droplet accumulation (BODIPY staining) and cellular TG levels (Fig. 2F), and increased FAO rate (Fig. 2G), features that resemble those observed in the E2f2−/− mouse livers resistant to HCC. Concomitant to the extensive changes in metabolism, acute E2F2 silencing in HepG2 cells had a subtle effect in the cell cycle, as shown by the approximately 5% reduction in S/G2–M cells compared with siControl (Fig. 2H). Overexpression of E2F2 (pE2F2) in cycling HepG2 cells (Supplementary Fig. S3A) led to a decreased FAO rate (Supplementary Fig. S3B) together with a modest induction of E2F target gene expression and cell-cycle progression (Supplementary Fig. S3C). Finally, we overexpressed E2F2 in vivo in non-cycling mouse liver cells through the injection of recombinant AAV8-E2f2 to healthy mice (Supplementary Fig. S3D). Sustained liver E2f2 overexpression promoted a substantial reduction in FAO rate when compared with control AAV8-GfP mice (Supplementary Fig. S3E) without inducing significant changes in cell-cycle gene expression (Supplementary Fig. S3F). The results suggest that E2F2 is a prominent regulator of lipid metabolism in cancer and healthy liver cells. Interestingly, this novel role of E2F2 appears to be independent of its classical role in cell-cycle regulation.
E2F2 deficiency in mice prevents nonalcoholic fatty liver disease
We next investigated whether E2F2 could also play a metabolic role in the onset of NAFLD. For this, a short-term HFD-induced carcinogenic protocol was used (Supplementary Fig. S1B) whereby mice were sacrificed at an earlier time (3 vs. 9 months of age), before tumor development and to induction of cell-cycle gene expression and Ki67 (Supplementary Fig. S4A and S4B). A group of mice that received the HFD alone was also included. Upon short-term HFD or DEN–HFD exposure, E2f2−/− mice were resistant to the TG accumulation (Fig. 3A) and lipid droplet storage (Fig. 3B), and maintained lower levels of liver CE than their corresponding WT mice (Fig. 3A). The DEN–HFD-mediated increase in ALT values was totally prevented in E2f2−/− mice, in which the liver-to-body weight was decreased (Supplementary Fig. S4C). The initiation of inflammation (Supplementary Fig. S4D) and the activation of endoplasmic reticulum stress (Supplementary Fig. S4E), usually linked with NAFLD progression, were also attenuated in E2f2−/− mice compared with the corresponding WT mice, as shown by the expression of Il1-β and the phosphorylation of eIF2-α, respectively. Among all the genes involved in FAO and oxidative phosphorylation that were found upregulated in E2f2−/− mice under long-term DEN–HFD exposure, only the expression of Ppargc1a, Cpt2, and Acsl1 was increased after a short-term administration (Fig. 3C; Supplementary Fig. S5A and S5B). When mice were fed the HFD alone, mRNA levels of Cpt2 and Ppargc1a were also increased; however, those of Acsl1 (Supplementary Fig. S5C), and of some genes involved in oxidative phosphorylation were decreased (Supplementary Fig. S5D). The FAO rate was already elevated in short-term DEN–HFD E2f2−/− mice (Fig. 3D) and in mice fed the HFD alone (Supplementary Fig. S5E) to levels that were similar to those found after a long-term exposure. These changes in FAO rate were not linked to changes in TOM20 protein levels, a marker of mitochondria (24, 25), or in PGC1α, involved in mitochondrial biogenesis (Fig. 3E; ref. 26). A significant upregulation in FAO rate was also observed in primary hepatocytes isolated from DEN–HFD E2f2−/− mice compared with their WT counterparts (Fig. 3F), which was coupled to increased mitochondrial oxygen consumption rate (OCR; Fig. 3G). This was not observed in CD fed E2f2−/− hepatocytes (Supplementary Fig. S5F).
E2F2 deficiency prevents NAFLD through the induction of fatty acid oxidation rate. Mice were treated for 10 weeks to induce NAFLD and were sacrificed at 3 months of age. A, Liver TG and CE quantification (n = 8–9 mice per group). B, Representative liver sections stained with H&E. C, Hepatic mRNA levels of genes involved in lipid oxidation (n = 7–8 mice per group). D, Fatty acid β-oxidation rate was determined in liver homogenates by measuring the amount of [14C]-CO2 and [14C]-ASM (n = 4 mice per group). E, Western blot analysis of TOM20 and PGC1α protein levels in livers (n = 6–8 mice per group). F, Fatty acid β-oxidation rate was determined as above in hepatocytes (n = 5 per group). G, Oxygen consumption rate in hepatocytes was measured using the Seahorse analyzer (n = 11–12 per group). Values represent mean ± SEM. Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN–HFD and CD are denoted as #, P < 0.05; ##, P < 0.01. ASM, acid soluble metabolites.
E2F2 deficiency prevents NAFLD through the induction of fatty acid oxidation rate. Mice were treated for 10 weeks to induce NAFLD and were sacrificed at 3 months of age. A, Liver TG and CE quantification (n = 8–9 mice per group). B, Representative liver sections stained with H&E. C, Hepatic mRNA levels of genes involved in lipid oxidation (n = 7–8 mice per group). D, Fatty acid β-oxidation rate was determined in liver homogenates by measuring the amount of [14C]-CO2 and [14C]-ASM (n = 4 mice per group). E, Western blot analysis of TOM20 and PGC1α protein levels in livers (n = 6–8 mice per group). F, Fatty acid β-oxidation rate was determined as above in hepatocytes (n = 5 per group). G, Oxygen consumption rate in hepatocytes was measured using the Seahorse analyzer (n = 11–12 per group). Values represent mean ± SEM. Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN–HFD and CD are denoted as #, P < 0.05; ##, P < 0.01. ASM, acid soluble metabolites.
E2F1, an E2F family member closely related to E2F2, plays an essential role in regulation of glycolysis in liver, a process that provides compounds required for the de novo lipogenesis (17). Given the relevance of glycolysis in NAFLD and HCC, we measured the extracellular acidification rate (ECAR) upon glucose injection into a Seahorse analyzer, as a readout of glycolysis, in primary hepatocytes isolated from CD and DEN-HFD E2f2−/− and the corresponding WT mice. Although glycolysis was decreased in E2f2−/− hepatocytes from CD fed mice when compared with the corresponding WT hepatocytes (Supplementary Fig. S6A), it remained unaltered in E2f2−/− hepatocytes from DEN–HFD mice (Supplementary Fig. S6B). The expression of glucokinase (Gck) was downregulated in CD and DEN-HFD E2f2−/− hepatocytes whereas that of piruvate kinase (Plkr) remained unaltered (Supplementary Fig. S6C).
One of the mechanisms involved in the generation of hepatosteatosis is the de novo lipogenesis. E2F1 promotes de novo fatty acid synthesis through modulation of Acaca and Fasn expression (17). Thus, we analyzed in E2f2−/− mice the expression of genes involved in de novo lipogenesis. The levels of malonyl-CoA, the product of the rate limiting reaction in the de novo lipogenesis and an inhibitor of FAO, and the metabolic flux that regulates de novo synthesis of TG from 3H-acetate were also examined. Expression of Acaca and Fasn genes was increased in untreated CD E2f2−/− mouse livers when compared with corresponding normal controls (Supplementary Fig. S6D). However, their expression (Supplementary Fig. S6D), as well as the liver concentration of malonyl-CoA (Supplementary Fig. S6E) and de novo lipogenesis of TG from acetate (Supplementary Fig. S6F), were unchanged in DEN–HFD E2f2−/−, indicating that this pathway is not involved in E2F2-mediated remodeling of lipid metabolism.
Another mechanism involved in lipid storage in NAFLD is the dysregulated VLDL secretion (27). However, the hepatic expression (mRNA) of Apob or Apoe (Supplementary Fig. S6G), required for VLDL assembly, and the in vivo hepatic TG secretion rate did not change in E2f2−/− DEN–HFD mice (Supplementary Fig. S6H). Mice fed with HFD alone showed similar results, except for an increased expression of Apob mRNA (Supplementary Fig. S7A–S7D). Altogether, these results indicate that the increased FAO rate is the main mechanism involved in the resistance that E2f2−/− mice exhibit to lipid storage upon NAFLD induction.
E2F2 represses FAO in the liver by targeting Cpt2
Among the genes involved in FAO, Cpt2 was consistently upregulated in both short-term and long-term DEN–HFD-treated E2f2−/− mice, as well as in CD E2f2−/− mice as compared with their WT counterparts (Fig. 3C). To validate our results in another setting, liver-specific knockdown of E2f2 was performed in CD WT mice with AAV8 carrying specific small shRNAs against E2f2. In agreement with our results in E2f2 knockout animals, acute silencing of E2f2 led to increased CPT2 protein levels and FAO rates (Fig. 4A and B).
E2F2 represses CPT2 gene expression in the liver. A, E2F2 and CPT2 protein levels in CD WT mice after liver-specific knockdown of E2f2 using AAV8s (n = 6 mice per group). B, Fatty acid β-oxidation rate was determined in liver, after specific knockdown of E2f2 using AAV8s, by measuring the amount of [14C]-CO2 and [14C]-ASM (n = 6 mice per group). C, Negative correlation between E2F2 and CPT2 expression in HCC tumors from human samples (data from the TCGA-LIHC project). Spearman correlation coefficient is shown. D, Schematic representation of Cpt2 gene promoter indicating the localization of a consensus E2F motif detected with Insect 2.0 at a 0.88 threshold level. ChIP-q-PCR analyses of E2F2 target gene Cpt2 in WT (n = 8–10 mice per group) and E2f2−/− mice (n = 3 mice). ChIP assays were performed using anti-E2F2 and anti-IgG (negative control) antibodies. E, CPT2 protein levels in liver (n = 4 mice per group). Values represent mean ± SEM (mouse analyses) or mean ± SD (human analyses). Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN-HFD and CD are denoted as #, P < 0.05. ASM, acid soluble metabolites; RSEM, RNA-seq by expectation maximization.
E2F2 represses CPT2 gene expression in the liver. A, E2F2 and CPT2 protein levels in CD WT mice after liver-specific knockdown of E2f2 using AAV8s (n = 6 mice per group). B, Fatty acid β-oxidation rate was determined in liver, after specific knockdown of E2f2 using AAV8s, by measuring the amount of [14C]-CO2 and [14C]-ASM (n = 6 mice per group). C, Negative correlation between E2F2 and CPT2 expression in HCC tumors from human samples (data from the TCGA-LIHC project). Spearman correlation coefficient is shown. D, Schematic representation of Cpt2 gene promoter indicating the localization of a consensus E2F motif detected with Insect 2.0 at a 0.88 threshold level. ChIP-q-PCR analyses of E2F2 target gene Cpt2 in WT (n = 8–10 mice per group) and E2f2−/− mice (n = 3 mice). ChIP assays were performed using anti-E2F2 and anti-IgG (negative control) antibodies. E, CPT2 protein levels in liver (n = 4 mice per group). Values represent mean ± SEM (mouse analyses) or mean ± SD (human analyses). Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN-HFD and CD are denoted as #, P < 0.05. ASM, acid soluble metabolites; RSEM, RNA-seq by expectation maximization.
Considering the established role for Cpt2 in the promotion of NAFLD-related HCC (11), we evaluated its potential regulation by E2F2. Indeed, we found that E2F2 and CPT2 expression inversely correlate in human HCC (TCGA-LIHC cohort; Fig. 4C). Building on these results, we evaluated whether E2F2 could directly repress Cpt2 gene expression. The INSECT 2.0 tool revealed the presence of several E2F-binding sites in the Cpt2 gene, including one near the transcription initiation site. ChIP followed by amplification of this region of the Cpt2 promoter showed an enrichment of E2F2 binding in DEN–HFD WT mice livers compared with CD WT mice (Fig. 4D). In concordance, CPT2 protein levels were increased in DEN–HFD E2f2−/− mice compared with their WT controls (Fig. 4E). To assess more directly E2F2-binding activity in primary cells, we performed ChIP analyses after overexpressing E2f2 in hepatocyte cultures, derived from WT mice, by treatment with adenoviruses carrying the specific plasmid for it (Supplementary Fig. S8A). Overexpression of E2f2 induced downregulation in Cpt2 mRNA levels (Supplementary Fig. S8B), concomitant with an enrichment of E2F2 in the promoter of Cpt2 (Supplementary Fig. S8C). Similarly to what has been described for E2F1 (17), there was also an enrichment of E2F2 in the promoters of several genes involved in glycolysis (Supplementary Fig. S8D). However, unlike E2F1, we did not find an enrichment of E2F2 in the promoters of genes involved in fatty acid synthesis (Supplementary Fig. S8C) or a significant enrichment in those involved in mitochondrial activity (Supplementary Fig. S8E).
Chromatin location and gene expression data collected from DEN–HFD mice and E2F2-overexpressing hepatocytes suggest a role for E2F2 as a direct repressor of liver Cpt2. To examine further the mechanism of E2F2-mediated Cpt2 gene regulation, we assessed the contribution of pRB, given its well-established role in E2F-mediated target gene repression (16), pRB was downregulated in hepatocytes isolated from DEN–HFD to CD WT mice (Supplementary Fig. S9A). Contrary to what was expected, downregulation of Rb decreased Cpt2 expression in hepatocytes from DEN–HFD WT mice, whereas it did not induce changes in hepatocytes from CD WT mice (Supplementary Fig. S9B). Thus, these results suggest that pRB is not involved in E2F2-mediated Cpt2 gene repression.
The Cpt2 gene promoter harbors a CREB-binding site near an E2F site (Supplementary Fig. S9C). Given that E2F2 and CREB cooperate in regulating the expression of some target genes (28) and that CREB is also involved in modulation of lipid metabolism (29), we assessed the contribution of CREB to Cpt2 gene regulation. In DEN–HFD WT mice, in vivo administration of CREB1 ASO lowered liver CPT2 protein levels and FAO rate (Supplementary Fig. S9D). By contrast, in DEN–HFD E2f2−/− mice, reduced CREB1 levels had no effect in CPT2 levels or FAO rate (Supplementary Fig. S9E). These results suggest that CREB1 modulates E2F2-mediated Cpt2 repression and subsequent FAO reactions.
E2f1 deficiency leads to Cpt2 upregulation and prevents NAFLD-related hepatocarcinogenesis
E2F1 induces the expression of lipogenic genes, contributing to hepatosteatosis (17) and its overexpression promotes hepatocarcinogenesis (30). However, it is unknown whether E2F1 regulates hepatic FAO in NAFLD-related HCC.
Our data confirm that E2F1 is upregulated in human HCC tissue compared with the surrounding normal liver (SL; Supplementary Fig. S10A) within the TCGA-LIHC cohort, consistent with previous reports (20, 21). Accordingly, E2f1 levels were increased in mouse livers with NAFLD-related HCC (DEN–HFD; Fig. 5A), and E2f1−/− mice showed protection against NAFLD-related HCC development (Fig. 5B). Consistently, DEN–HFD E2f1−/− mice were also resistant to liver TG storage (Fig. 5C). Following these results, we investigated the degree of similarity between the transcriptome hallmark that underlies the protection against NAFLD-related HCC in E2f1−/− and E2f2−/− mice. The most significantly changed pathways within the upregulated genes in DEN–HFD E2f1−/− mice compared with the DEN–HFD WT controls were those related to metabolism of lipids and fatty acids (Supplementary Fig. S10B), and among the downregulated genes were those related to “immune system” and “platelet activation, signaling and aggregation” (Supplementary Fig. S10C). The analysis of genes commonly upregulated between DEN–HFD E2f1−/− and DEN–HFD E2f2−/− mice showed 704 genes (Fig. 5D), most of them related to metabolic pathways and some to fatty acid metabolism (Fig. 5E).
E2F1 is required for NAFLD-related HCC and it regulates lipid oxidation through a mechanism shared with E2F2. Mice were treated for 32 weeks to induce hepatocarcinogenesis and were sacrificed at 9 months of age. A, Analysis of hepatic E2f1 mRNA levels (n = 6–8 mice per group). B, Representative photographs from WT and E2f1−/− mouse livers. C, Liver TG and CE quantification (n = 4–6 mice per group). D, Venn diagram showing the degree of overlap between the number of differentially upregulated genes in DEN-HFD E2f1−/− and DEN-HFD E2f2−/− mice. E, RNA was extracted from E2f1−/− and WT liver homogenates and a gene expression microarray analysis was carried out (n = 4 mice per group). Differential gene expression between E2f1−/− or E2f2−/− and WT was analyzed and significant probes were assigned into pathways with Reactome database. Biological pathways, including common upregulated genes (P ≤ 0.001), are shown (–log2 of 0.001 is indicated with a red line). F, Hepatic mRNA levels of genes involved in lipid oxidation (n = 6–8 mice per group). G, Fatty acid β-oxidation rate was determined in liver homogenates by measuring the amount of [14C]-CO2 and [14C]-ASM (n = 5 mice per group). H, Negative correlation between E2F1 and CPT2 expression in HCC tumors from human samples (data from the TCGA-LIHC project). I, Hepatic mRNA levels of Cpt2 in WT, E2f2−/−, E2f1−/−, and E2f1/E2f2 double knockout (DKO) mice (n = 4–5 mice per group). J, Positive correlation between E2F1 and E2F2 expression in HCC tumors from human samples (data from the TCGA-LIHC project). Values represent mean ± SEM (mouse analyses) or mean ± SD (human analyses). Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN–HFD and CD are denoted as #, P < 0.05; ##, P < 0.01. Spearman correlation coefficient is shown. ASM, acid soluble metabolites; RSEM, RNA-seq by expectation maximization.
E2F1 is required for NAFLD-related HCC and it regulates lipid oxidation through a mechanism shared with E2F2. Mice were treated for 32 weeks to induce hepatocarcinogenesis and were sacrificed at 9 months of age. A, Analysis of hepatic E2f1 mRNA levels (n = 6–8 mice per group). B, Representative photographs from WT and E2f1−/− mouse livers. C, Liver TG and CE quantification (n = 4–6 mice per group). D, Venn diagram showing the degree of overlap between the number of differentially upregulated genes in DEN-HFD E2f1−/− and DEN-HFD E2f2−/− mice. E, RNA was extracted from E2f1−/− and WT liver homogenates and a gene expression microarray analysis was carried out (n = 4 mice per group). Differential gene expression between E2f1−/− or E2f2−/− and WT was analyzed and significant probes were assigned into pathways with Reactome database. Biological pathways, including common upregulated genes (P ≤ 0.001), are shown (–log2 of 0.001 is indicated with a red line). F, Hepatic mRNA levels of genes involved in lipid oxidation (n = 6–8 mice per group). G, Fatty acid β-oxidation rate was determined in liver homogenates by measuring the amount of [14C]-CO2 and [14C]-ASM (n = 5 mice per group). H, Negative correlation between E2F1 and CPT2 expression in HCC tumors from human samples (data from the TCGA-LIHC project). I, Hepatic mRNA levels of Cpt2 in WT, E2f2−/−, E2f1−/−, and E2f1/E2f2 double knockout (DKO) mice (n = 4–5 mice per group). J, Positive correlation between E2F1 and E2F2 expression in HCC tumors from human samples (data from the TCGA-LIHC project). Values represent mean ± SEM (mouse analyses) or mean ± SD (human analyses). Statistical analysis was determined by Student two-tailed t test. Significant differences between E2f2−/− and WT mice are denoted as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and differences between DEN–HFD and CD are denoted as #, P < 0.05; ##, P < 0.01. Spearman correlation coefficient is shown. ASM, acid soluble metabolites; RSEM, RNA-seq by expectation maximization.
The genes related to FAO and oxidative phosphorylation that were upregulated in DEN–HFD E2f2−/− mice (Fig. 2C; Supplementary Fig. S2B), were also found upregulated in DEN–HFD E2f1−/− mice (Fig. 5F; Supplementary Fig. S10D and S10E), including Cpt2 (Fig. 5F) together with the FAO rate (Fig. 5G). Moreover, E2F1 and CPT2 correlated negatively in human HCC (TCGA-LIHC cohort; Fig. 5H) and Cpt2 was upregulated in 3-month-old CD fed E2f1−/− mice (Fig. 5I). This effect was not the result of a cross-regulation between E2f1 and E2f2, because livers of E2f1/E2f2 double knockout (DKO) mice fed a CD exhibited levels of Cpt2 expression that were practically the sum of those found in single knockout mice (Fig. 5I). Moreover, during disease progression, expression of E2f1 was decreased in the E2f2−/− mice as compared with their WT mice, implying that E2f1 does not compensate for the loss of E2f2 (Supplementary Fig. S10F and S10G). Taken together, these results suggest that both E2F1 and E2F2 contribute independently to Cpt2 gene repression in the liver.
E2F1 and E2F2 protein levels are increased in human NAFLD
E2F1 and E2F2 expression positively correlated in human HCC from the TCGA-LIHC cohort (Fig. 5J). Given our evidences that E2F1 and E2F2 play a major role as regulators of the FAO rate involved in the development of NAFLD and NAFLD-related hepatocarcinogenesis, we measured E2F1 and E2F2 protein levels in livers from a cohort of obese patients with and without NAFLD (Supplementary Table S1). A group of non-obese NL patients was also included. E2F2 levels were found increased in the liver of patients with NAFLD compared with NL (Fig. 6A). The same profile was observed for E2F1 protein levels (Fig. 6B). E2F1 and E2F2 levels were higher in NL from obese patients than in NL from non-obese subjects. Correlation between both proteins was observed in obese patients' livers (Fig. 6C).
E2F1 and E2F2 are increased in human NAFLD. A, Immunohistochemical analysis of E2F2 in patients (n = 78) with NAFLD or NL. B, Immunohistochemical analysis of E2F1 in patients (n = 78) with NAFLD or NL. C, Correlation analyses between E2F1 and E2F2 expression in a human cohort of obese patients with and without NAFLD. A group of non-obese NL patients was also included. Values represent the mean. Differences between NAFLD (obese) or NL (non-obese or obese) are denoted by *, P < 0.05; **, P < 0.01; ***, P < 0.001 (Student t test). Spearman correlation coefficient is shown.
E2F1 and E2F2 are increased in human NAFLD. A, Immunohistochemical analysis of E2F2 in patients (n = 78) with NAFLD or NL. B, Immunohistochemical analysis of E2F1 in patients (n = 78) with NAFLD or NL. C, Correlation analyses between E2F1 and E2F2 expression in a human cohort of obese patients with and without NAFLD. A group of non-obese NL patients was also included. Values represent the mean. Differences between NAFLD (obese) or NL (non-obese or obese) are denoted by *, P < 0.05; **, P < 0.01; ***, P < 0.001 (Student t test). Spearman correlation coefficient is shown.
Discussion
Altered lipid metabolism is a common feature of most cancers (8). In malignant cells, rewiring of lipid metabolism provides lipids for the synthesis of membranes, activation of signaling cascades, and energy storage (31). In the case of liver carcinogenesis this setting is even more complex. The liver controls metabolic homeostasis of the whole body, but when the uptake of dietary lipids is overloaded, such as in obesity, the increased liver lipid input is not compensated by pathways involved in lipid output, thus leading to NAFLD (32). NAFLD, with its rising worldwide prevalence linked with obesity, represents a major risk for HCC development. There are no pharmacological therapies for NAFLD and effective treatments for HCC are required. Therefore, the prevention and/or restoration of the metabolic rewiring associated to NAFLD and HCC represent novel attractive strategies. There is growing evidence of a regulatory axis between cell-cycle regulators and metabolic fluxes. Here, we investigated the etiopathogenic role of E2F1 and E2F2 transcription factors in NAFLD and NAFLD-related HCC, and their potential value as targets for therapy in these health conditions.
We show that E2F2 is upregulated in NAFLD and NAFLD-related HCC, and that loss of E2F2 prevents the development of NAFLD and NAFLD-related HCC, implying an oncogenic role for E2F2 in liver hepatocarcinoma, as has been described in other cancer types (33). The requirement of E2F2 for liver regeneration after partial hepatectomy (34) supports its proliferative role in liver. Biological pathways typically upregulated in HCC, such as platelet activation/signaling and cell cycle, were downregulated in E2f2−/− mice administered with DEN–HFD, consistent with the protection against HCC development in these mice. Importantly, we have uncovered a novel E2F2-dependent mechanism for liver disease, by demonstrating that E2F2 is a key regulator of lipid metabolism in liver and a driver of liver cancer in a manner that appears to be independent of its cell-cycle regulator function. The reprogramming in lipid metabolism is of high relevance in cancer, but the specific metabolic pathways that sustain the supply of energy and/or other sources for cell survival are a feature to be determined for each cancer type. It will be important to evaluate to what extent other cancer types rely on E2F2-regulated metabolic pathways to support cancer growth.
E2f2 knockout mouse livers and E2F2 knockdown HCC cell lines displayed similar metabolic profiles, characterized by decreased TG concentration and increased FAO rate. The decreased lipid storage was associated to a transcriptome profile characterized by the upregulation of genes involved in FAO. The relevance of FAO in cancer cells is still not fully understood. Some cancers benefit from FAO (35) whereas others such as NAFLD-related HCC exhibit the opposite response (11). The rate limiting step that controls FAO is the carnitine–acylcarnitine shuttle, where FAs are converted to acylcarnitines through CPT1. In turn, CPT2 will release them as acyl-CoAs into the mitochondrial matrix (36). It has been reported that in NAFLD-related HCC, CPT2 levels are decreased, leading to decreased FAO rate and to the accumulation of acylcarnitines (11). The decrease in FAO rate promotes lipid storage whereas the accumulation of acylcarnitines promotes the activation of STAT3 and proliferation (11). Increased lipid storage leads to lipotoxicity-induced hepatocyte cell death in some patients promoting progression of NAFLD to NASH and even to HCC (37, 38). However, HCC cells need to survive in this lipid-rich environment, and decreased CPT2 expression is associated with resistance of HCC cells to chronic exposure to palmitic acid (11). Thus, the upregulation of CPT2 and FAO, such as what we observe in the absence of E2F2, would prevent lipid accumulation, its associated lipotoxicity and activation of an ER stress response, thereby hindering disease development and progression. So far, the only known regulator of CPT2 is PPARα (39). Here, we demonstrate by using ChIP assays and transcriptomic analyses that E2F2 binds to Cpt2 promoter and represses its expression, which is consistent with the fact that E2F2 inversely correlates with CPT2 in human HCC. The mechanisms by which E2F factors may activate genes involved in cell cycle whereas repress others involved in metabolism are still unresolved. It has been shown that E2F family can function as direct repressors of transcription through their well-known interaction with pRB (40) or independently via recruitment of corepressors such as BIN1 or KAP1 (41, 42). Regarding metabolism genes, it has been reported that pRB is involved in E2F1-mediated mitochondrial gene repression (16). Conversely, our study shows that E2F2-mediated Cpt2 gene repression in DEN–HFD treated or CD fed mice is independent of pRB. The identification of putative E2F2 co-repressors involved in Cpt2 repression remains to be addressed in future studies.
E2F2 shares with E2F1 multiple target genes and cellular functions, but there is also target specificity (43). In the liver, increased E2F1 and E2F2 levels positively correlate in human HCC and in human NAFLD, suggesting that regulation of these two factors in liver disease might respond to the same effectors. A role for E2F1 in hepatocarcinogenesis has been previously reported, associated with an upregulated expression of E2F target genes involved mainly in cell cycle and DNA repair (30). We now identify a novel regulatory pathway by which both E2F1 and E2F2 contribute to NAFLD-related HCC development in mice, through a mechanism that involves Cpt2 gene repression, FAO and mitochondrial activity. Of relevance, the upregulated levels of Cpt2 in E2f1/E2f2 DKO mice were higher that the levels found in E2f1−/− and E2f2−/− mice, revealing a new overlapping role of E2F1 and E2F2 in the repression of Cpt2 in NAFLD-related HCC. However, the lack of either E2f1 or E2f2 is enough to protect from NAFLD and NAFLD-related hepatocarcinogenesis. This is an important issue given that the concomitant lack of both E2F1 and E2F2, and subsequent overexpression of CPT2, would induce FAO to a level that may generate ROS without resolving liver disease (44). However, supporting the fact that E2F1 and E2F2 also regulate different target genes, our results here show that Acaca and Fasn expression and de novo lipogenesis of TG from acetate are not regulated by E2F2, in contrast with what has been described for E2F1 (17).
In sum, our data demonstrate that both E2F1 and E2F2 repress CPT2 expression, leading to the generation of a lipid-rich environment required for hepatocarcinogenesis (Fig. 7). However, of note, the lack of just one of them is enough to prevent liver disease. These data identify E2F1 and E2F2 as new metabolic drivers that regulate the rewiring of metabolism in NAFLD-related HCC, arising as novel targets for therapy.
Proposed model for the role of E2F2 and E2F1 in the development of NAFLD and progression to HCC. Increased E2F1 and E2F2 activity in liver inhibits CPT2 expression, which would hamper fatty acid oxidation rate from increasing in response to an elevated lipid flux. This metabolic dysregulation would boost a lipid-rich environment with increased acylcarnitine levels, thereby promoting the progression of liver disease, including Kupffer cell activation and the recruitment of platelets. LSEC, liver sinusoidal endothelial cells; L, lymphocyte; P, platelet; N, neutrophil; KC, Kupffer cells; HSC, hepatic stellate cells; DS, Disse space.
Proposed model for the role of E2F2 and E2F1 in the development of NAFLD and progression to HCC. Increased E2F1 and E2F2 activity in liver inhibits CPT2 expression, which would hamper fatty acid oxidation rate from increasing in response to an elevated lipid flux. This metabolic dysregulation would boost a lipid-rich environment with increased acylcarnitine levels, thereby promoting the progression of liver disease, including Kupffer cell activation and the recruitment of platelets. LSEC, liver sinusoidal endothelial cells; L, lymphocyte; P, platelet; N, neutrophil; KC, Kupffer cells; HSC, hepatic stellate cells; DS, Disse space.
Authors' Disclosures
F. González-Romero reports grants from Basque government during the conduct of the study. I. Aurrekoetxea reports grants from Basque Country goverment during the conduct of the study. A. Woodhoo reports grants from European Research Council Consolidator grant 865157, MCIU/AEI/FEDER, UE (RTI2018- 097503-B-I00, SEV-2016-0644), MINECO-FEDER SAF2015-62588-ERC, and Leonardo Grant for Researchers and Cultural Creators BBVA Foundation during the conduct of the study. M. Tamayo-Caro reports grants from Ministerio de Ciencia e Innovación during the conduct of the study. B. Gómez-Santos reports grants from Basque government during the conduct of the study. D. Sáenz de Urturi reports grants from Basque government during the conduct of the study. R. Lee reports personal fees from Ionis Pharmaceuticals outside the submitted work. S. Bhanot reports personal fees from Ionis Pharmaceuticals outside the submitted work. J. Crespo reports grants from Gilead, Abbvie, and Intercept outside the submitted work. P. Aspichueta reports grants from Basque government and Spanish Ministry of University and Science during the conduct of the study. No disclosures were reported by the other authors.
Authors' Contributions
F. González-Romero: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. D. Mestre: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. I. Aurrekoetxea: Conceptualization, resources, formal analysis, validation, investigation, methodology. C.J. O'Rourke: Formal analysis, investigation, methodology, writing–review and editing. J.B. Andersen: Formal analysis, investigation, methodology, writing–review and editing. A. Woodhoo: Conceptualization, formal analysis, investigation, methodology, writing–review and editing. M. Tamayo-Caro: Formal analysis, investigation. M. Varela-Rey: Conceptualization, formal analysis, investigation, methodology, writing–review and editing. M. Palomo-Irigoyen: Formal analysis, investigation. B. Gómez-Santos: Formal analysis, investigation, methodology. D. Sáenz de Urturi: Formal analysis, investigation, methodology. M. Núñez-García: Formal analysis, investigation, methodology. J.L. García-Rodríguez: Formal analysis, investigation. L. Fernández-Ares: Formal analysis, investigation. X. Buqué: Formal analysis, validation, visualization, methodology, writing–review and editing. A. Iglesias-Ara: Formal analysis, investigation. I. Bernales: Formal analysis, investigation. V.G. De Juan: Formal analysis, investigation, methodology. T.C. Delgado: Formal analysis, investigation. N. Goikoetxea-Usandizaga: Formal analysis, investigation. R. Lee: Formal analysis, Investigation, methodology. S. Bhanot: Formal analysis, investigation, methodology. I. Delgado: Formal analysis, investigation, methodology. M.J. Perugorria: Formal analysis, investigation. G. Errazti: Formal analysis, investigation. L. Mosteiro: Formal analysis, investigation. S. Gaztambide: Formal analysis, investigation. I. Martinez de la Piscina: Formal analysis, investigation. P. Iruzubieta: Formal analysis, investigation. J. Crespo: Formal analysis, investigation. J.M. Banales: Conceptualization, formal analysis, funding acquisition, investigation, writing–review and editing. M.L. Martínez-Chantar: Formal analysis, funding acquisition, investigation, writing–review and editing. L. Castaño: Formal analysis, investigation. A.M. Zubiaga: Conceptualization, resources, formal analysis, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing. P. Aspichueta: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.
Acknowledgments
This work was supported by “Ayudas para apoyar grupos de investigación del sistema Universitario Vasco” (IT971–16 to P. Aspichueta and IT1257–19 to A.M. Zubiaga), MCIU/AEI/FEDER, UE (SAF2015–64352-R and RTI2018–095134-B-100 to P. Aspichueta), MCIU/AEI/FEDER, UE (SAF2015–67562-R and RTI2018–097497-B-100 to A.M. Zubiaga), MINECO-FEDER SAF2017–87301-R (to M.L. Martínez-Chantar), MINECO-FEDER EU (“Subprograma Ramon y Cajal” RYC2010–06901), MINECO-FEDER SAF2015-65360-R, MINECO-FEDER SAF2015–72416-EXP, MINECO-FEDER SAF2015–62588-ERC, MCIU/AEI/FEDER, UE (RTI2018- 097503-B-I00, SEV-2016–0644), and European Research Council (Consolidator grant under the EU's Horizon 2020 research and innovation program; grant agreement 865157 to A. Woodhoo), Gilead Sciences Research Scholars Programs (to M. Varela Rey), Leonardo Grant for Researchers and Cultural Creators BBVA Foundation (to M. Varela Rey and A. Woodhoo) and “Ciberehd Emergente” (to M. Varela Rey), BIOEF (Basque Foundation for Innovation and Health Research); EITB Maratoia (BIO15/CA/014 to M.L. Martínez-Chantar and BIO15/CA/016 to J.M. Banales), La Caixa Fundation and “AYUDAS FUNDACIÓN BBVA A EQUIPOS DE INVESTIGACIÓN CIENTÍFICA” 2018 (HR17–00601 to M.L. Martínez-Chantar and J.M. Banales). AECC (“Cáncer Infantil”; JP Vizcaya; to A. Woodhoo). J.M. Banales was funded by the Spanish Carlos III Health Institute (ISCIII; FIS PI15/01132, PI18/01075 and Miguel Servet Program CON14/00129 and CPII19/00008) cofinanced by “Fondo Europeo de Desarrollo Regional” (FEDER), AMMF-The Cholangiocarcinoma Charity, Euskadi RIS3 (2019222054), and “Fundación Científica de la Asociación Española Contra el Cáncer” (AECC Scientific Foundation; “Rare cancers grant 2017” to J.M. Banales and M.L. Martínez-Chantar). Ciberehd_ISCIII_MINECO is funded by ISCIII. The results published here are in part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. CIC-BioGUNE thanks MINECO for the Severo Ochoa Excellence Accreditation (SEV-2016 0644). The authors thank Jose Antonio Lopez from the Department of Physiology, Faculty of Medicine and Nursing UPV/EHU and SGIKER from UPV/EHU.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.