β-Catenin encoding gene CTNNB1 is known as the most frequently mutated proto-oncogene in liver cancer. We report that active β-catenin is essential in initiation and advancement of hepatocarcinogenesis. As a transcriptional activator of AKT2, β-catenin potentiates AKT2 phosphorylation of CAD, which in return stimulates de novo pyrimidine synthesis and liver cancer development. β-Catenin, AKT2, and pyrimidine synthesis inhibitors are promising therapeutics for the treatment of oncogenic β-catenin–associated cancer.
CTNNB1, encoding β-catenin protein, is the most frequently altered proto-oncogene in hepatic neoplasms. In this study, we studied the significance and pathological mechanism of CTNNB1 gain-of-function mutations in hepatocarcinogenesis. Activated β-catenin not only triggered hepatic tumorigenesis but also exacerbated Tp53 deletion or hepatitis B virus infection–mediated liver cancer development in mouse models. Using untargeted metabolomic profiling, we identified boosted de novo pyrimidine synthesis as the major metabolic aberration in β-catenin mutant cell lines and livers. Oncogenic β-catenin transcriptionally stimulated AKT2, which then phosphorylated the rate-limiting de novo pyrimidine synthesis enzyme CAD (carbamoyl-phosphate synthetase 2, aspartate transcarbamoylase, dihydroorotase) on S1406 and S1859 to potentiate nucleotide synthesis. Moreover, inhibition of β-catenin/AKT2-stimulated pyrimidine synthesis axis preferentially repressed β-catenin mutant cell proliferation and tumor formation. Therefore, β-catenin active mutations are oncogenic in various preclinical liver cancer models. Stimulation of β-catenin/AKT2/CAD signaling cascade on pyrimidine synthesis is an essential and druggable vulnerability for β-catenin mutant liver cancer.
Liver cancer is the sixth most common cancer and the third leading cause of cancer death worldwide (1). It is the second most lethal tumor after pancreatic cancer (2). Hepatic cancer is a heterogenous group of malignances ranging from hepatocellular carcinoma (HCC) (75 to 85%), intrahepatic cholangiocarcinoma (ICC) (10 to 15%), to several rare subtypes, such as hepatoblastoma (HB), the most common pediatric liver cancer (3). Its multifaceted etiologies include genetic aberrations, hepatitis B virus (HBV) or hepatitis C virus (HCV) infection, and chemical carcinogen induction. These complex pathological events are still not druggable. The knowledge of the molecular events governing tumor initiation and progression may aid the development of targeted therapies. Therefore, genetic alterations should be studied for their impacts on hepatocarcinogenesis.
Cancer is a genetic disease, which usually arises from synergistic interaction between activated proto-oncogenes and inactivated tumor suppressors. CTNNB1 (coding for β-catenin) (22% mutation rate) and TP53 (29% mutation rate) are the most frequently altered proto-oncogenes and tumor suppressor genes, respectively, detected mainly in HCC (https://cancer.sanger.ac.uk/cosmic/browse/tissue?wgs=off&sn=liver&ss=all&hn=all&sh=all&in=t&src=tissue&all_data=n) (4). Over half of HBs also harbor CTNNB1 mutations (5). In addition, CTNNB1 mutation is identified in more than 10% hepatocellular adenoma (HCA), which is a benign liver neoplasm with risk of malignant transformation (6). Mutations of exon 3 are the most common alterations of β-catenin in liver cancer (7). Phosphorylation of serine/threonine encoded by exon 3 of CTNNB1 leads to ubiquitination-mediated degradation of β-catenin. Exon 3 variations protect β-catenin protein from degradation and in return constitutively turn on β-catenin signaling cascade (8, 9). Although activated β-catenin signaling is required for hepatic APC knockout mice to develop liver cancer (10), mice did not developed liver tumors 6 m after hepatic Ctnnb1 exon 3 deletion (9). Thus, whether altered β-catenin is the cause or the consequence of liver cancer is largely unknown.
More than half of global HCC cases are HBV positive and 70 to 80% of HCC patients in HBV endemic regions are HBV positive (11). Since up to 10% Chinese are HBV carriers, half of HCC patients are in China (12). Liver cancer is the fourth most common cancer and the third leading cause of cancer death in China (13). Chronic hepatitis caused by HBV is thought to be the major cause of HCC. Given that only a small fraction of HBV-infected people develop hepatic cancer, additional genomic insults from ingestion of foods tainted with chemical carcinogens such as aflatoxins and excessive consumption of alcohol may contribute to eventual development of liver cancer in HBV carriers. For instance, 13.4 to 19% of liver cancer patients infected with HBV harbor CTNNB1 mutations in tumor lesions (14, 15), suggesting a potential collaboration of these two pathological events in liver cancer development.
Metabolic aberration is a hallmark of cancer (16). Oncogenic mutations of proto-oncogenes or tumor suppressors reprogram metabolism to meet enhanced nutritional and growth requirements of unchecked cell proliferation and tumor growth (17). Liver is the major metabolic organ. The relationship between abnormal metabolism and hepatic cancer remains largely elusive. If mutated β-catenin contributes to hepatocarcinogenesis, elucidation of its downstream events such as aberrant metabolism and its governing signaling transduction pathway may provide novel targets for the treatment of liver cancer.
To establish the causative relationship between aberrant β-catenin activation and liver cancer, we not only dissected the role of CTNNB1 mutation alone in liver carcinogenesis but also demonstrated its collusion with Tp53 deletion or transgenic HBV in liver cancer propagation. Moreover, we identified oncogenic β-catenin–boosted de novo pyrimidine synthesis via stimulation of the AKT2–CAD (carbamoyl-phosphate synthetase 2, aspartate transcarbamoylase, dihydroorotase) signaling pathway in liver cancer development and its clinical significance in cancer treatment.
Constitutively Activated β-Catenin Causes Hepatic Tumors.
Because exon 3 is the most common locus of CTNNB1 mutation in cancer (7), exon 3–deleted mice are widely used for the study of β-catenin activation (9). To mimic CTNNB1 active mutations in hepatoblastoma and HCC, we first generated albumin (Alb)-Cre–mediated hepatic Ctnnb1 exon 3–deleted mice. However, these mice died within 20 to 30 d after birth with hepatomegaly (SI Appendix, Fig. S1 A–C). Since early mouse lethality due to high efficacy of β-catenin activation precluded us from using this model to study hepatocarcinogenesis, we inoculated 7-wk-old β-cateninlox(ex3)/+ mice with Cre-adenoviruses (6 × 108 pfu per mice) via tail vein injection to generate mosaic Ctnnb1 exon 3 deletion in mouse livers (Fig. 1A). A total of 20 to 40% of β-cateninlox(ex3)/+ mice died within 50 to 100 d after Cre-adenovirus incubation (SI Appendix, Fig. S1 D–F) (9). The remaining β-cateninlox(ex3)/+ mice also had a shorter lifespan albeit at a slower pace (Fig. 1B). By 13 mo old of age, 73% (19/26) of the survived β-cateninlox(ex3)/+ mice but not wild-type (WT) mice developed prominent liver tumors with activated β-catenin (Fig. 1 C–F). Among 19 tumor-bearing β-cateninlox(ex3)/+ mice, three had benign neoplasms, eight had malignant ones, and the remaining eight mice had both types of tumors according to hematoxylin-eosin (H&E), reticular fiber, and CD34 stainings (Fig. 1G and SI Appendix, Table S1). Malignant lesions had strongly diffused staining of CD34 and reticulin widening, while benign lesions showed minimal/focal positivity of CD34 and relative normal reticulin staining. Positive Heppar1 and negative CK19 stainings indicated that the malignant tumors were HCC rather than ICC (Fig. 1H). As it took a long latency for mutant β-catenin to cause liver cancer, additional pathological events may have participated in this pathological process.
β-Catenin activation causes hepatic tumors. (A) Schematic illustration of Cre-adenovirus–induced β-catenin active mutation in mouse liver. (B) Survival of WT mice (n = 23) and β-cateninlox(ex3)/+ mice (n = 27) 50 d post Cre-adenovirus tail vein injection. (C) Representative liver pictures of 13-mo-old mice. (D) Tumor incidence (at least one visible tumor nodule on the surface of liver) of 13-mo-old β-cateninlox(ex3)/+ mice (n = 26). (E) The maximum liver tumor size (Left), tumor number (Middle), and ratio of liver weight to body weight (Right). WT mice (n = 12), β-cateninlox(ex3)/+ mice (n = 26). Data are shown as mean ± SD. (F) Immunoblotting of mouse liver tissues. (G) Representative liver H&E, reticular fiber, and CD34 stainings of β-cateninlox(ex3)/+ mice. (H) Immunohistochemistry stainings of mouse liver tissues. (Scale bar: 100 μm.) ***P < 0.001.
Oncogenic β-Catenin Collaborates with Distinct Carcinogenic Factors in the Promotion of Hepatocarcinogenesis.
CTNNB1 and TP53 are the most commonly mutated proto-oncogenes and tumor suppressor genes in liver cancer. Guichard et al. reported that CTNNB1 and TP53 mutations were mutually exclusive, as only 2.4% (3/125) of HCC had co-occurrence of CTNNB1 and TP53 mutations, 11.5% (3/26) of TP53 mutant HCC had CTNNB1 mutations, and 7.3% (3/41) of CTNNB1 mutant HCC had TP53 mutation in that study (14). However, we found concomitant CTNNB1 and TP53 mutations in a bigger subset of liver cancers by analyzing the HCC sequencing data of 471 HCC cases compiled from several reports (15, 18–20). While only 7.5 to 11.3% of enrolled HCC had concurrent CTNNB1 and TP53 mutations and 10.2 to 20.2% of TP53 mutant HCC harbored CTNNB1 mutations, half (46 to 53%) of CTNNB1 mutant HCC carried TP53 mutations (SI Appendix, Table S2) (15, 18–20), suggesting the critical role of TP53 mutations in mutant CTNNB1-associated liver cancer, at least for HCC patients in China. To simulate this subtype of liver cancer, we crossed Tp53l/l mice with β-cateninlox(ex3)/+ mice to generate Tp53l/l; β-cateninlox(ex3)/+ mice. Cre-adenoviruses were then injected to mutate Ctnnb1 and/or delete Tp53 in liver accordingly (Fig. 2A). Tp53l/l; β-cateninlox(ex3)/+ mice had shorter lifespans than Tp53l/l mice (Fig. 2B). Tp53l/l; β-cateninlox(ex3)/+ mice but not Tp53l/l mice developed obvious liver tumors when they were 11.5 mo old (Fig. 2 C and D). These lesions were typical HCC with activated β-catenin (Fig. 2 E and F). We also generated heterozygous Ctnnb1 exon 3–deleted mouse embryonic fibroblasts (MEFs) in Tp53 null background from Tp53l/l; β-cateninlox(ex3)/+ mouse embryos with addition of Cre-adenoviruses in cell culture (SI Appendix, Fig. S2A). Activated β-catenin and deleted TP53 MEFs exhibited stronger tumorigenic potential than TP53 null MEFs did in immunodeficient nude mice (SI Appendix, Fig. S2 B–E).
Oncogenic β-catenin colludes with distinct carcinogenic factors in furtherance of hepatocarcinogenesis. (A–F) β-Catenin activation and Tp53 deletion in mice. (A) Tp53l/l and Tp53l/l; β-cateninlox(ex3)/+ mice were injected with Cre-adenovirus via tail vein 7 wk after birth. (B) Survival of Tp53l/l mice (n = 27) and Tp53l/l; β-cateninlox(ex3)/+ mice (n = 20). (C–F) Representative pictures (C), the maximum liver tumor size (Upper), tumor number (Lower Left), ratio of liver weight to body weight (Lower Right) (D), immunoblotting (E), and representative H&E, Heppar1 and CK19 stainings (F) of liver tissues from 11.5-mo-old mice. (G–L) Mutant β-catenin and transgenic HBV in mice. (G) HBV and HBV; β-cateninlox(ex3)/+ mice were injected with Cre-adenovirus to generate HBV and HBV plus hepatic β-catenin mutant mice. (H) Survival of HBV mice (n = 21) and HBV; β-cateninlox(ex3)/+ mice (n = 39). (I–L) Representative pictures (I), the maximum liver tumor size (Upper), tumor number (Lower Left), ratio of liver weight to body weight (Lower Right) (J), immunoblotting (K), and representative H&E, Heppar1 and CK19 stainings (L) of liver tissues from 10.5-mo-old mice. (M–O) Pri-724 treatment. (M) HBV; β-cateninlox(ex3)/+ mice were injected with Cre-adenovirus by 7 wk, treated with vehicle (DMSO, dimethyl sulfoxide) or Pri-724 by 5 mo and killed by 8 mo after birth. (N) Representative liver pictures of 8-mo-old mice. (O) The maximum liver tumor size (Left), tumor number (Middle), and ratio of liver weight to body weight (Right) were analyzed. Data are shown as mean ± SD *P < 0.05, **P < 0.01, ***P < 0.001.
A total of 13.4 to 19% of liver cancer patients infected with HBV harbor CTNNB1 mutations in tumor lesions (14, 15). To imitate this subset of HCC, we crossed HBV transgenic mice with β-cateninlox(ex3)/+ mice to generate HBV mice and HBV; β-cateninlox(ex3)/+ ones. These mice were then injected with Cre-adenoviruses to delete exon 3 of Ctnnb1 in hepatocytes of HBV; β-cateninlox(ex3)/+ mice (Fig. 2G). HBV; β-cateninlox(ex3)/+ mice had a shorter lifespan compared to HBV transgenic mice (Fig. 2H) and developed HCC with activated β-catenin at age of 10.5 mo (Fig. 2 I–L).
Comparison of the survivals of various mouse models we generated revealed that β-cateninlox(ex3)/+ mice appear to have shortened survival than that of HBV mice (P = 0.0645) or Tp53l/l mice (P = 0.0538) albeit without reaching statistical significance within the window of observation (SI Appendix, Fig. S3). Therefore, the relative potency of tumorigenic potential among β-catenin activation, transgenic HBV, and Tp53 deletion in mice is inconclusive. Combination of either transgenic HBV and β-catenin activation or Tp53 deletion and β-catenin activation was more potent than either transgenic HBV, β-catenin activation, or Tp53 deletion alone in malignant transformation (SI Appendix, Fig. S3). Pri-724, a β-catenin inhibitor, blunted the tumor growth in Cre-adenovirus–inoculated HBV; β-cateninlox(ex3)/+ mice without changes in mouse body weight (Fig. 2 M–O and SI Appendix, Fig. S4). Taken together, under various preclinical settings, oncogenic β-catenin promotes mouse hepatocarcinogenesis, which can be ameliorated by β-catenin inhibitor.
Oncogenic β-Catenin Stimulates De Novo Pyrimidine Synthesis.
Liver is the major metabolic organ. Gain-of-function mutation of specific proto-oncogene often introduces metabolic vulnerabilities, which may be amenable to therapeutic intervention (16). To check oncogenic β-catenin–mediated cellular metabolic reprogramming without potential interference of nonactivated cells, we employed tamoxifen to induce expression of Alb-Cre recombinases to globally delete exon 3 of β-catenin in hepatocytes of ERT2-Alb-Cre; β-cateninlox(ex3)/+ mice (Fig. 3 A and B and SI Appendix, Fig. S5A). Untargeted metabolomic profiling was performed on both wildtype and β-cateninΔ(ex3)/+ livers. Of 201 metabolites identified by liquid-chromatography (LC) mass spectrometry (MS) (Q-Exactive-plus), the abundances of 30 metabolites were different (P < 0.05) between wild-type and β-cateninΔ(ex3)/+ livers (Fig. 3C). Because MEFs are widely used cell models for study of cell metabolism and signaling transduction (21, 22), we also carried out untargeted metabolomics for wild-type and β-catenin exon 3–deleted MEFs (SI Appendix, Fig. S5 C and D). Comparison of the metabolites enriched in β-cateninΔ(ex3)/+ mouse livers and MEFs identified three most significantly elevated metabolites: carbamoyl-asp, dihydroorotate, and orotate, which are starting molecules for de novo pyrimidine biosynthesis pathway (Fig. 3C and SI Appendix, Fig. S5 B, D, and E).
Oncogenic β-catenin stimulates de novo pyrimidine synthesis. (A) Generation of tamoxifen-induced hepatic β-catenin exon 3–deleted mice. Seven-week-old mice were injected with tamoxifen for 6 d and killed by day 12 for metabolic analysis. (B) Immunoblotting of WT and β-cateninΔ(ex3)/+ livers. (C) Steady-state metabolite heatmaps of WT and β-cateninΔ(ex3)/+ mouse livers. (D) 15N-labeled glutamine was used to trace pyrimidine synthesis in vivo. Abundance of 15N-labeled mouse liver metabolites 45 min after 15N-glutamine injection was analyzed. (E) The flux kinetics of 15N-labeled metabolites in MEFs at different time points after addition of 15N-glutamine. (F and G) Abundance of 15N-labeled metabolites. (F) β-cateninΔ(ex3)/+ MEFs were treated with DMSO or Pri-724 (20 μM) for 24 h and a 12-min pulse labeling of 15N-glutamine (2 mM). (G) Pri-724–treated HepG2 cells were provided with excessive glutamine (6 mM) and DMSO-treated cells were provided with the standard dose of glutamine (2 mM). *P < 0.05, **P < 0.01, ***P < 0.001. Analysis was performed using t test. Data are shown as mean ± SD.
De novo pyrimidine biosynthesis is a synthetic pathway that fuses nitrogen/carbon from glutamine, bicarbonate (HCO3–), and aspartate with ribose-phosphate to form heterocyclic nucleotides (SI Appendix, Fig. S5F). To elucidate active β-catenin–mediated regulation of intermediates’ biosynthesis, we traced the metabolic flux in vivo by injecting WT and hepatic β-cateninΔ(ex3)/+ mice with 15N-amide glutamine, which is incorporated into the pyrimidine ring. Incorporations of 15N into multiple intermediates of pyrimidine synthesis were enhanced in β-catenin mutant livers over their wild-type counterparts (Fig. 3D). Consistently, the dynamic relative flux with 15N-labeled carbamoyl-asp, dihydroorotate, and orotate was dramatically increased in β-cateninΔ(ex3)/+ MEFs (Fig. 3E) and the enhancement was compromised by Pri-724 treatment (Fig. 3F). Neither glutamine uptake nor expression of glutamine transporters was obviously altered by β-catenin in livers and MEFs (SI Appendix, Fig. S5 A and C), suggesting that β-catenin–regulated flux is not due to the change of glutamine availability. Furthermore, Pri-724 blocked pyrimidine synthesis in CTNNB1 exon 3–deleted human HB cells (HepG2) and mouse HCC cells (Hepa 1-6) as well as human HCC cells (M97h and Huh7) (SI Appendix, Fig. S6).
It was reported that β-catenin transcriptionally activated GS (glutamine synthetase) to produce glutamine, which in turn activated mTOR in liver (23). S6K (P70S6K), the downstream effector of mTOR, potentiates pyrimidine synthesis (21). Therefore, we predicted that β-catenin stimulated pyrimidine synthesis through the potential GS-S6K connection. However, MEFs had no glutamate-derived glutamine and therefore no GS activity (SI Appendix, Fig. S5G), indicating that β-catenin might promote pyrimidine synthesis without GS involvement. Of note, the regulation of β-catenin on pyrimidine synthesis still existed after S6K depletion (SI Appendix, Fig. S6E). Moreover, the relative flux of pyrimidine synthesis was largely reduced by Pri-724 even in the presence of excessive glutamine (Fig. 3G). Therefore, oncogenic β-catenin stimulates pyrimidine synthesis without participation of the GS-S6K axis.
β-Catenin Potentiates CAD Phosphorylation and Transactivates AKT2 Expression.
To identify the mechanism by which oncogenic β-catenin stimulates pyrimidine synthesis, we checked the state of CAD, DHODH (dihydroorotate dehydrogenase), and UMPS (uridine monophosphate synthetase), the critical enzymes in this pathway. However, the abundances of their transcripts and proteins were not altered by β-catenin (SI Appendix, Fig. S7 A and B and Fig. 4 A and B). Carbamoyl-asp and dihydroorotate are catalyzed by CAD, whose activity strongly depends on the phosphorylation levels of its distinct sites (24). Therefore, we compared phosphoproteome between wild-type and β-catenin mutant livers as well as wild-type and β-catenin mutant MEFs to identify potential β-catenin–regulated phosphorylation sites of CAD. Among the critical phosphorylation sites for CAD enzymatic activity, S1859 and S1406 were highly phosphorylated in β-catenin mutant livers and MEFs (SI Appendix, Fig. S7 C and D and Table S3). Furthermore, hyperphosphorylation of S1859 and S1406 was abolished by Pri-724 (SI Appendix, Table S3). Immunoblotting demonstrated β-catenin stimulation of S1859 and S1406 phosphorylation on CAD in livers, MEFs, HepG2, Hepa 1-6, M97h and Huh7 cells (Fig. 4 A and B and SI Appendix, Fig. S7E). Of note, Pri-724 abolished phosphorylation of S1859 and S1406 on CAD even in the absence of S6K (SI Appendix, Fig. S7 F and G). Taken into consideration of S6K phosphorylation of S1859 but not S1406 on CAD (21), our data suggest the existence of other kinases responsible for β-catenin–promoted CAD phosphorylation.