Gluconeogenic enzyme PCK1 deficiency promotes CHK2 O- GlcNAcylation and hepatocellular carcinoma growth upon glucose deprivation

Although cancer cells are frequently faced with nutrient- and oxygen-poor microenvironment, elevated hexosamine-biosynthesis pathway (HBP) activity and protein O-GlcNAcylation (a nutrient sensor) contribute to rapid growth of tumor and are emerging hallmarks of cancer. Inhibiting O-GlcNAcylation could be a promising anti-cancer strategy. The gluconeogenic enzymes phosphoenolpyruvate carboxykinase 1 (PCK1) was downregulated in hepatocellular carcinoma (HCC). However, little is known about the potential role of PCK1 in enhanced HBP activity and HCC carcinogenesis under glucose-limited conditions. In this study, PCK1 knockout markedly enhanced the global O-GlcNAcylation levels under low glucose condition. Mechanistically, metabolic reprogramming in PCK1-loss hepatoma cells led to oxaloacetate accumulation and increased de novo UTP synthesis contributing to uridine diphosphate-N-acetylglucosamine (UDP-GlcNAc) biosynthesis. Meanwhile, deletion of PCK1 also resulted in AMPK-GFAT1 axis inactivation promoting UDP-GlcNAc synthesis for elevated O-GlcNAcylation. Notably, lower expression of PCK1 promoted CHK2 threonine 378 O-GlcNAcylation counteracting its stability and dimer formation, increasing CHK2-dependent Rb phosphorylation and HCC cell proliferation. Moreover, aminooxyacetic acid hemihydrochloride and 6-diazo-5-oxo-L-norleucine blocked HBP-mediated O-GlcNAcylation and suppressed tumor progression in liver-specific Pck1-knockout mice. We reveal a link between PCK1 depletion and hyper-O-GlcNAcylation that underlies HCC oncogenesis and suggest therapeutic targets for HCC that act by inhibiting O-GlcNAcylation.


Introduction
Gaining insight into the fundamental role of metabolic reprogramming in cancer has contributed immensely to our understanding of tumorigenesis and cancer progression (1). Nutrient limitations (such as glucose deprivation) in solid tumors may require cancer cells to exhibit the metabolic flexibility required to sustain proliferation and survival (2,3). Increased aerobic glycolysis (the Warburg effect) is one of the main characteristics of cancer cells that supports their intensive growth and proliferation (4). Gluconeogenesis (the pathway opposite to glycolysis) operates during starvation and occurs mainly in the liver, and also plays key roles in metabolic reprogramming, cancer cell plasticity, and tumor growth (5,6). Several key gluconeogenic enzymes, such as phosphoenolpyruvate carboxykinase 1 (PCK1, also known as PEP-CK-C), fructose-1,6-bisphosphatase, and glucose-6-phosphatase, were previously found to be dysregulated in several types of cancer, including hepatocellular carcinoma (HCC) (7).
The cytoplasmic isoform of PCK1, the first rate-limiting enzyme in hepatic gluconeogenesis, catalyzes the conversion of oxaloacetate (OAA) to phosphoenolpyruvate (PEP). The oncogenic or tumor suppressor roles of PCK1 in different types of human cancers are rather controversial. PCK1 has antitumorigenic effects in gluconeogenic organs (liver and kidney), but has tumor-promoting effects in cancers arising from nongluconeogenic organs (5). In colon-derived tumor cells, PCK1 is hijacked to participate in truncated gluconeogenesis to meet biosynthetic and anabolic needs (8). However, the underlying mechanism determining its aberrant expression and altered function in multiple types of tumors remains incompletely understood.
Recent findings emphasize the role of the hexosamine-biosynthesis pathway (HBP), a subbranch of glucose metabolism, in carcinogenesis (9,10). The HBP and glycolysis share the first 2 steps and diverge at fructose-6-phosphate (F6P). Glutamine-fructose-6-phosphate aminotransferase 1 (GFAT1), the rate-limiting enzyme of the HBP, converts F6P and glutamine to glucosamine-6-phosphate and glutamate. Uridine diphosphate N-acetylglucosamine (UDP-GlcNAc), the end products of HBP, is a donor substrate for O-linked β-N-acetylglucosamine (O-Glc-NAc) modification (also known as O-GlcNAcylation) (11). O-Glc-NAc transferase-mediated (OGT-mediated) protein O-GlcNAcylation is highly dependent on the intracellular concentration of the Although cancer cells are frequently faced with a nutrient-and oxygen-poor microenvironment, elevated hexosaminebiosynthesis pathway (HBP) activity and protein O-GlcNAcylation (a nutrient sensor) contribute to rapid growth of tumor and are emerging hallmarks of cancer. Inhibiting O-GlcNAcylation could be a promising anticancer strategy. The gluconeogenic enzyme phosphoenolpyruvate carboxykinase 1 (PCK1) is downregulated in hepatocellular carcinoma (HCC). However, little is known about the potential role of PCK1 in enhanced HBP activity and HCC carcinogenesis under glucose-limited conditions. In this study, PCK1 knockout markedly enhanced the global O-GlcNAcylation levels under low-glucose conditions. Mechanistically, metabolic reprogramming in PCK1-loss hepatoma cells led to oxaloacetate accumulation and increased de novo uridine triphosphate synthesis contributing to uridine diphosphate-N-acetylglucosamine (UDP-GlcNAc) biosynthesis. Meanwhile, deletion of PCK1 also resulted in AMPK-GFAT1 axis inactivation, promoting UDP-GlcNAc synthesis for elevated O-GlcNAcylation. Notably, lower expression of PCK1 promoted CHK2 threonine 378 O-GlcNAcylation, counteracting its stability and dimer formation, increasing CHK2-dependent Rb phosphorylation and HCC cell proliferation. Moreover, aminooxyacetic acid hemihydrochloride and 6-diazo-5-oxo-L-norleucine blocked HBP-mediated O-GlcNAcylation and suppressed tumor progression in liver-specific Pck1-knockout mice. We reveal a link between PCK1 depletion and hyper-O-GlcNAcylation that underlies HCC oncogenesis and suggest therapeutic targets for HCC that act by inhibiting O-GlcNAcylation.  Additionally, we used an N-nitrosodiethylamine (DEN) and CCl 4 -induced mouse HCC model to verify the above results in vivo. The O-GlcNAcylation levels in the hepatic tumors of liver-specific Pck1 knockout (LKO) mice were significantly higher than those in WT mice after 12 hours of fasting ( Figure 1, H and I). Furthermore, using an orthotopic HCC mouse model, we found that PCK1 overexpression decreased the O-GlcNAcylation levels and inhibited HCC growth (Supplemental Figure 3). Together, these data demonstrated that PCK1 decreases the global O-Glc-NAcylation levels in HCC under a low-glucose condition and suppresses hepatoma cell proliferation, both in vitro and in vivo. The enzymatic activity of PCK1 is indispensable for its tumor suppressor role in HCC.
PCK1 deficiency promotes UDP-GlcNAc biosynthesis via oxaloacetate accumulation. Next, a metabolomics assay was performed with AdPCK1-and AdGFP-infected SK-Hep1 cells to explore metabolic changes occurring after PCK1 overexpression (PCK1-OE) under a low glucose concentration (5 mM). Principal component analysis showed that PCK1 overexpression dramatically changed the intracellular metabolic profile of SK-Hep1 cells (Supplemental Figure 4A). Levels of several metabolites of the HBP decreased after PCK1 overexpression, including fructose 6-phosphate, N-acetyl glucosamine 1-phosphate (GlcNAc-1-P), and UDP-Glc-NAc (the HBP end product), as shown in Figure 2, A and B, and Supplemental Figure 4B. Our targeted liquid chromatography-tandem MS (LC-MS/MS) results showed that UDP-GlcNAc significantly decreased in PCK1-OE SK-Hep1 cells ( Figure 2C), but increased in PCK1-KO (PKO) cells ( Figure 2D), suggesting that PCK1 may negatively regulate UDP-GlcNAc biosynthesis for O-GlcNAcylation. To investigate how PCK1 modulates UDP-GlcNAc biosynthesis, we performed pathway-enrichment analysis of metabolite profiles and found that several metabolic pathways were significantly affected, including purine and pyrimidine metabolism, which is required for uridine triphosphate (UTP) synthesis (Supplemental Figure 4C). Since the HBP uses glucose, glutamine, acetyl-CoA, and UTP to produce the amino sugar UDP-GlcNAc ( Figure 2E), we assumed that PCK1 may regulate UDP-GlcNAc biosynthesis partially via UTP synthesis. Indeed, metabolomics data showed that levels of several metabolites involved in UTP synthesis, including Asp (aspartate, critical metabolite in de novo UTP synthesis), declined upon PCK1 overexpression ( Figure 2F), besides metabolites in glycolysis and the tricarboxylic acid (TCA) cycle ( Figure  2A and Supplemental Figure 4, D-G), suggesting that PCK1 may play role in TCA cataplerosis and UTP synthesis which contribute to UDP-GlcNAc biosynthesis.
PCK1 catalyzes the conversion of OAA (an intermediate of the TCA cycle) to PEP. Consistent with the results of a previous study (8), restoring PCK1 decreased the OAA concentration in SK-Hep1 cells ( Figure 2G). Considering that OAA is converted to Asp by the mitochondrial glutamate-oxaloacetate transaminase (GOT2, also known as aspartate aminotransferase 2; Supplemental Figure 4H), we speculated that PCK1 repression may promote UDP-GlcNAc biosynthesis through Asp conservation caused by OAA accumulation. As expected, adding OAA strengthened UDP-GlcNAc biosynthesis ( Figure 2H). To acquire a definitive proof of UDP-Glc-NAc synthesis from OAA accumulation, we used a stable isotopic tracing approach to detect dynamic metabolic flux. The relative donor substrate UDP-GlcNAc, which is proposed to be a nutrient sensor that couples metabolic and signaling pathways (12,13). Increased glucose flux through the HBP and elevated UDP-Glc-NAc contribute to hyper-O-GlcNAcylation in cancer cells (14). Previous data suggest that elevated O-GlcNAcylation may serve as a hallmark of cancer (15).
Similar to phosphorylation, O-GlcNAcylation is a dynamic posttranslational modification that regulates protein subcellular localization, stability, protein-protein interactions, or enzymatic activity according to the nutrient demands of cells (16). OGT and O-GlcNAcase (OGA) are the only enzymes known to be responsible for adding and removing N-acetylglucosamine (GlcNAc) on serine and threonine residues of target proteins (17). Numerous oncogenic factors, including c-MYC, Snail, and β-catenin, are targets of O-GlcNAcylation (18)(19)(20). Therefore, modulating the HBP or O-GlcNAcylation (which regulate oncogenic activation in human cancers) represents a promising anticancer strategy that can potentially be used in combination with other treatments (21,22).
In this study, we explored the role of the gluconeogenic enzyme PCK1 in regulating the HBP and HCC proliferation under low-glucose conditions. We unravel a molecular mechanism responsible for enhanced UDP-GlcNAc biosynthesis and O-GlcNAcylation induced by PCK1 depletion, and delineate the functional importance of checkpoint kinase 2 (CHK2) O-GlcNAcylation in HCC tumorigenesis. Importantly, our study reveals a novel link between the gluconeogenic enzyme PCK1 and HBP-mediated O-GlcNAc modification, which suggests a therapeutic strategy for treating HCC.  Figure 1D). Interestingly, the catalytically inactive G309R mutant of PCK1 (23) was unable to reduce the O-GlcNAcylation levels in these hepatoma cells under the same culture conditions ( Figure 1C and Supplemental Figure 1, B and C), suggesting that the enzymatic activity of PCK1 may play an attenuating role in regulating cellular O-GlcNAcylation levels. In addition, this observation was confirmed via pharmacological or transcriptional inhibition of OGT and OGA (Figure 1, D-G), suggesting PCK1 inhibits OGT-mediated O-GlcNAcylation. Interestingly, PCK1 did not change the mRNA or protein expression levels of OGT, OGA, and GFAT1, the key enzymes involved in regulating O-GlcNAcylation and HBP (Supplemental Figure 1, E-N). In addition, our cell-proliferation assays indicated that PCK1 suppresses hepatoma cell proliferation, depending on its enzymatic activity and the cellular O-GlcNAcylation levels (Supplemental Figure 2). er groups showed that AMPK activation reduces O-GlcNAcylation through GFAT1 phosphorylation (which diminished GFAT1 activity) in endothelial cells and cardiac hypertrophy (25,26). We speculated that PCK1 may also inhibit UDP-GlcNAc synthesis through the AMPK-GFAT1 axis. Thus, we tested whether PCK1 can suppress the HBP through the AMPK-GFAT1 axis under low-glucose conditions. As expected, PCK1 overexpression promoted the phosphorylation of both AMPK and GFAT1 ( Figure 3A), whereas PCK1-KO downregulated p-AMPK and p-GFAT1 production ( Figure 3B). The AMPK activator metformin partially offset hyper-O-GlcNAcylation mediated by PCK1 depletion ( Figure 3C). However, shRNA-mediated knockdown of AMPK in PCK1-OE cells rescued the inhibitory effects of PCK1 on O-GlcNAcylation ( Figure 3D).

PCK1 deficiency increases global O-GlcNAcylation in hepatoma cells
Furthermore, we investigated whether the inhibition of hepatoma cell proliferation in response to PCK1 depended on the AMPK-GFAT1 axis. We found that metformin suppressed PKO cell proliferation ( OGT mediates CHK2 O-GlcNAcylation in PCK1-deficient hepatoma cells. To further explore how OGT-mediated protein O-GlcNAcylation facilitates hepatoma cell proliferation in PKO cells, we used immunoprecipitation coupled with tandem MS (IP-MS/MS) to screen for proteins that specifically interact with OGT ( Figure 4A and Supplemental Figure 6A). Flag-tagged OGT was transiently expressed in PKO cells, and subsequent IP-MS identified 618 candidate OGT-binding proteins (Supplemental Table 1). Pathway-enrichment analysis indicated that several proteins were involved in metabolic processes, apoptotic processes, and cell-cycle progression ( Figure 4B). We then focused on CHK2, which is required for checkpoint-mediated cell-cycle arrest (27). Interactions between OGT and CHK2 were confirmed by coimmunoprecipitation (co-IP) experiments in HEK293 cells (Supplemental Figure 6, B and C) and PKO cells ( Figure 4, C-E). Confocal analysis also indicated that OGT and CHK2 colocalized in the nucleus ( Figure 4F and Supplemental Figure 6D). To define the precise region(s) in CHK2 required for this interaction, we expressed full-length HA-tagged OGT in combination with different Flag-tagged fragments of CHK2 in HEK293 cells ( Figure 4G). The C-terminal region of CHK2 (amino acids 69-543) containing kinase domains showed a strong interaction, whereas the N-terminal region (amino acids 1-175) did not interact with OGT ( Figure 4H).
Then, we determined whether CHK2 can be modified via O-GlcNAc. Immunoprecipitated, Flag-tagged CHK2 exhibited a strong O-GlcNAc signal in HKE293 cells ( Figure 4I). Endogenous CHK2 O-GlcNAcylation was confirmed by affinity chroma-abundance of M+3 UDP-GlcNAc was significantly increased in PKO cells when incubated with 13 C 5 -glutamine, whereas it decreased in PCK1-OE cells ( Figure 2, I and J). The isotopomers of the metabolites involved in the TCA cycle including OAA, and UTP de novo synthesis were reduced in the PCK1-OE cells, meanwhile Asp was increased in PKO cells (Supplemental Figure 5), suggesting that PCK1 deficiency strongly boosts the entry of OAA into UDP-GlcNAc synthesis.
Accordingly, both OAA and Asp enhanced O-GlcNAcylation in SK-Hep1 cells, while high concentrations of PEP (above 0.5 mM) showed an opposite effect ( Figure 2K). Next, we tested whether GOT2 can contribute to HBP-mediated O-GlcNAcylation in PKO cells. We found that treatment with aminooxyacetic acid (AOA), a specific inhibitor of GOT2, decreased the UDP-GlcNAc levels in PKO cells (Supplemental Figure Figure 4J). Taken together, these data provide strong evidence supporting the contributions of OAA accumulation and the GOT2-mediated pathway to enhanced UDP-Glc-NAc biosynthesis and hyper-O-GlcNAcylation in PKO cells.
Restoration of PCK1 suppresses O-GlcNAcylation by activating the AMPK-GFAT1 axis. The final rate-limiting step of UDP-GlcNAc synthesis involves UTP and GlcNAc-1-P ( Figure 2E). Our metabolomics analysis showed that both UTP and GlcNAc-1-P levels were significantly decreased in PCK1-OE cells (Figure 2, B and F). Since OAA accumulation contributes to the downstream UTP increase and GFAT1 is the rate-limiting enzyme in GlcNAc-1-P synthesis, we explored whether GFAT1 activity also regulates UDP-GlcNAc production. Previously, we reported that PCK1 activates AMP-activated protein kinase (AMPK) upon glucose deprivation in HCC (24). Oth- Next, we sought to map the O-GlcNAcylation site(s) on CHK2. Flag-tagged CHK2 was purified from PKO cells and analyzed by MS. As shown in Figure 4N, threonine 378 (T378) was the main O-GlcNAcylation site on CHK2. We then generated site-specific point mutants of CHK2. Mutating T378 to Ala (T378A) largely reduced the O-GlcNAc signal compared with WT CHK2 and the T383A mutant control ( Figure 4O and Supplemental Figure 6E), indicating that T378 is the major CHK2 site carrying the O-Glc-NAc modification. The potentially O-GlcNAcylated residue T378 and the surrounding amino acids are highly conserved among vertebrates ( Figure 4P), indicating that it serves an evolutionarily conserved role regulating the CHK2 protein. Taken together, these O-GlcNAcylation on T378 stabilizes CHK2 and promotes hepatoma cell proliferation. To examine the effect of O-GlcNAcylation on CHK2 under the low-glucose conditions, Flag-tagged WT and T378A CHK2 were overexpressed with HA-tagged ubiquitin (HA-Ub) in PCK1-KO and parental PLC/PRF/5 cells. WT CHK2 ubiquitination was alleviated in PKO cells compared with parental cells, whereas the T378A mutation or OGT inhibitor ST045849 enhanced CHK2 ubiquitination ( Figure 5A). We then performed a series of cycloheximide-chase experiments to assess the half-life of these proteins. Endogenous CHK2 was more stable, with a halflife of more than 24 hours, in PLC/PRF/5 cells treated with PUG-NAc, indicating CHK2 O-GlcNAcylation may enhance its stability (Supplemental Figure 7, A and B). In comparison with the parental cells, the CHK2 half-life was prolonged in PKO cells, but the T378A mutation or ST045849 treatment reduced CHK2 half-life from 24 hours to 12 hours ( Figure 5, B-E). In addition, overexpression of WT PCK1 promoted CHK2 ubiquitination and degradation (Supplemental Figure 7, C-G). As expected, the G309R PCK1 mutation did not affect the half-life of CHK2 (Supplemental Figure 7, D-G). These results suggested that O-GlcNAc modification of T378 stabilizes CHK2 by preventing its ubiquitination and degradation in PCK1-deficient hepatoma cells.
Given that CHK2 dimerization is essential for its activation (28), we next detected CHK2 dimerization in cells cotransfected with vectors encoding Flag-CHK2 and Myc-CHK2. Our results indicated that PKO cells displayed strengthened CHK2 dimer formation, whereas ST045849 treatment or CHK2 T378A mutant weakened this association (Supplemental Figure 7, H and I). A similar result was observed by crosslinking analysis (Supplemental Figure 7J), suggesting O-GlcNAcylation of CHK2 may promote its dimerization. Interestingly, T378, the autophosphorylation site of CHK2, is located in the dimerization interface (28). We then performed dimeric CHK2 homology modeling, followed by molecular dynamic (MD) simulation. Our model disclosed that the O-GlcNAcylated residue T378 interacts with the amino acid VSLK of another CHK2 kinase domain. The acetylglucosamine group occupies a cavity located in the edge of interaction interface and forms 3 hydrogen bonds with the backbone of VSLK motif, thus might strengthen the stability of the CHK2 dimer (Supplemental Figure 7, K-M). Since dimerization promotes CHK2 activation and phosphorylates its downstream targets, such as retinoblastoma (Rb) in HCC (29), we subsequently checked the phosphorylation of CHK2 substrates and downstream signaling in response to PCK1 expression. Overexpressing WT PCK1 decreased the p-Rb and p-CDK2 levels, but increased the p27 levels ( Figure 5F). In contrast, PCK1 KO or OAA treatment reversed the regulatory effects of these molecules ( Figure 5, G and H). Notably, the OGT inhibitor ST045849 or the GFAT1 inhibitor DON partially offset the regulatory effects mediated by PCK1 deficiency (Figure 5, G and I). These data indicated that O-GlcNAcylation promotes CHK2 dimerization and subsequently enhances downstream Rb phosphorylation.
To further test whether the loss of CHK2 O-GlcNAcylation affects its downstream signaling and hepatoma cell proliferation, we transiently overexpressed WT CHK2 or the T378A mutant in CHK2-KO cells. WT CHK2 restored the phosphorylation of Rb and CDK2 and promoted HCC proliferation, whereas the CHK2 T378A mutant failed to exert this stimulatory role on tumorigenesis ( Figure 5J and Supplemental Figure 8, A and B), indicating that T378 O-GlcNAcylation plays an essential role in CHK2 activation. The T378A point mutation (which eliminated the O-Glc-NAc modification) decreased the capacity of CHK2 to phosphorylate Rb. In agreement, the OGA inhibitor PUGNAc enhanced the ability of WT CHK2 to phosphorylate Rb and CDK2, but not that of the CHK2 T378A mutant ( Figure 5J). Accordingly, PCK1 KO or OAA treatment promoted WT CHK2 activation ( Figure 5, K and L). Finally, we explored whether PCK1 deficiency-induced malignancy may rely on CHK2 O-GlcNAcylation. As expected, CHK2 depletion suppressed hepatoma cell proliferation and G1/S transition induced by PCK1 deficiency, which was rescued by reexpressing WT CHK2, but not the T378 mutant (Supplemental Figure 8, C-H). In addition, CHK2 depletion dramatically reduced the growth of orthotopic tumor bearing PKO PLC/PRF/5 in nude mice (Supplemental Figure 9). These data suggest that CHK2 T378 O-GlcNAcylation conferred a growth advantage for the PKO cells. Collectively, these findings indicate that the O-Glc-NAcylation of residue T378 stabilizes CHK2 and activates its downstream targets such as Rb, thus promoting PCK1-deficient hepatoma cell proliferation in vitro.
Targeting HBP-meditated O-GlcNAcylation suppresses DEN/ CCl 4 -induced hepatocarcinogenesis in vivo. Next, we used the DEN/ CCl 4 -induced mouse model of liver cancer to further verify our results in vivo. Based on our in vitro data, we proposed that blocking the HBP with an inhibitor of GOT2 (AOA) or GFAT1 (DON) could suppress the growth of HCC by reducing O-GlcNAcylation. LKO mice ( Figure 6, A and B) were generated as previously described (30). The mice were treated with DEN/CCl 4 to induce hepatocarcinoma, which was followed by administering AOA or DON (twice a week) for 16 weeks ( Figure 6C and Supplemental Figure 10). The LKO mice exhibited accelerated liver tumorigenesis with increased tumor masses and nodules, and higher serum levels of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) (Figure 6, D-H). Our MS data showed that the OAA and UDP-GlcNAc levels were higher in the liver tumors of LKO mice (Figure 6, I and J), indicating that PCK1 deficiency promotes the HBP in vivo. In contrast, mice treated with AOA or DON exhibited slower tumor growth and a reduced number of tumor nodules compared with untreated LKO mice ( Figure  6, D-F). Furthermore, AOA or DON treatment also decreased UDP-GlcNAc and O-GlcNAcylation levels in LKO mice ( Figure  6, J-L). These data suggested that hyper-O-GlcNAcylation conferred a growth advantage for tumor cells in vivo. Consistent with our in vitro data, the levels of total CHK2, O-GlcNAcylated CHK2, and p-Rb were significantly enhanced in the liver tumors of LKO mice, and were partially reversed by administering AOA or DON ( Figure 6L). In summary, these findings suggested that PCK1 depletion increases susceptibility to DEN/CCl 4 -induced carcinogenesis and promotes hepatocarcinogenesis via enhanced CHK2 O-GlcNAcylation; thus, blocking HBP-meditated O-Glc-NAcylation suppresses HCC in LKO mice.
PCK1 deficiency strengthens CHK2 O-GlcNAcylation in primary human HCC. Finally, we investigated PCK1 expression and glob-are generally lower than those in nontransformed tissues (31). The mechanisms underlying tumor growth during periods of metabolic stress through enhanced HBP activity and O-GlcNAcylation have not been fully elucidated. It remains unknown whether gluconeogenesis contributes to maintaining HBP-mediated O-GlcNAcylation in cancer cells under low nutrient conditions (5). Here, we present the first evidence that deficiency of the gluconeogenic enzyme PCK1 promotes cellular O-GlcNAcylation and tumorigenesis in HCC ( Figure 7J). Moreover, we identify that CHK2 O-GlcNAcylation at T378 maintains its stability and oncogenic activity in hepatoma cells. Therefore, our study provides a link between PCK1 repression and hyper-O-GlcNAcylation underlying HCC oncogenesis.
Since the liver is the major site of gluconeogenesis during fasting, the role of gluconeogenesis in HCC has begun to draw more attention recently (32). During glucose starvation, cancer cells redistribute gluconeogenic intermediates to downstream pathways to facilitate their proliferation (5). PCK1, the rate-limiting enzyme of gluconeogenesis, is downregulated in HCC (24,33,34). The mammalian target of rapamycin complex 2 (mTORC2) and hepatitis B X-interacting protein (HBXIP) abrogate the expression of PCK1 by inhibiting the nuclear translocation of forkhead box protein O1 (FOXO1), a key transcriptional activator of the PCK1 gene (35,36). In addition, posttranslational modifications such as sumoylation and acetylation of PCK1 are enhanced in HCC cells, which promotes its ubiquitination and subsequent degradation (37,38). PCK1 mediates not only gluconeogenesis, but also glyceroneogenesis and TCA cataplerosis (39). In the present study, we found that PCK1 silencing promoted UDP-GlcNAc biosynthesis, thus enhancing cellular O-GlcNAcylation under a low-glucose condition (5 mM glucose). Previous findings showed that the maximum rate of gluconeogenesis is approached at glucose concentrations under 5 mM, whereas high glucose levels inhibit gluconeogenesis (40,41). Consistent with this observation, we did not detect any significant change in O-GlcNAcylation levels under high-glucose conditions (25 mM glucose) upon PCK1 depletion or overexpression, indicating that PCK1 regulated HBP flux and protein O-GlcNAcylation, depending on glucose availability. Moreover, the global O-GlcNAcylation levels of HCC tissues were obviously increased in Pck1-knockout mice, as compared with those in the WT mice after fasting. However, cell culture under 5 mM glucose for 12 hours could not completely simulate the low-glucose conditions of tumor microenvironment. Given these limitations, further in vitro studies under different concentrations of glucose are needed to confirm that PCK1 modulates protein O-GlcNAcylation.
O-GlcNAcylation depends on OGT/OGA levels and the donor substrate UDP-GlcNAc. As a nutrient sensor, UDP-GlcNAc levels are dependent on glucose, amino acid, fatty acid, and nucleotide availability (42). Here, we revealed a dual role for PCK1 in regulating UDP-GlcNAc biosynthesis through OAA accumulation and the AMPK-GFAT1 axis under low-glucose conditions. On the one hand, the OAA levels decreased upon PCK1 overexpression in hepatoma cells, but accumulated in the liver tumors of LKO mice. LKO mice are unable to remove oxaloacetate from the TCA cycle (43). OAA is converted to Asp by GOT2, a key enzyme that plays a role in the TCA cycle and amino acid metabolism (44). Accordingly, amino acids synthesized from the TCA cycle (including Asp) were elevated in the blood of PCK1 KO mice (45). Pharmacologi-  Table 2). In addition, downregulated PCK1 expression was significantly associated with poor tumor differentiation and prognosis (Supplemental Table  2). Moreover, the global O-GlcNAcylation was significantly higher in HCC tissues than in adjacent normal tissues (Figure 7, B and E and Supplemental Figure 11). Consistently, the p-AMPK and p-GFAT1 levels were reduced in HCC tissues ( Figure 7B). We also observed a negative correlation between PCK1 protein-expression levels and O-GlcNAcylation levels in HCC ( Figure 7F, r = -0.3565, P = 0.0240). In addition, the sWGA pull-down assay showed that enhanced CHK2 O-GlcNAcylation was associated with PCK1 downregulation (Figure 7, G and H and Supplemental Figure 11). Consistent with our in vitro data, we observed a strong negative correlation between p-Rb levels and PCK1 expression ( Figure 7I, r = -0.3852, P = 0.0168). In conclusion, this clinical validation supports the finding that PCK1 repression strengthens CHK2 O-Glc-NAcylation and promotes tumor growth in human primary HCC.

Discussion
Emerging evidence has demonstrated that protein O-GlcNAcylation plays a key role in tumorigenesis. Increased glucose flux through the HBP elevates UDP-GlcNAc, which enhances cellular O-GlcNAcylation. However, cancer cells are frequently faced with limited nutrients due to an insufficient and inappropriate vascular supply and rapid nutrient consumption (2). Previous metabolomics data demonstrated that the glucose concentrations in tumor tissues   and breast cancer (63). Here, we showed that both DON and AOA inhibited the growth of HCC in vitro and in vivo, largely by blocking HBP-mediated O-GlcNAcylation. These data suggest that DON and AOA inhibit cell growth through what we believe is a novel mechanism and provide a strong rationale for further clinical drug development, particularly for PCK1-deficient HCC.
In summary, we uncovered a link between gluconeogenesis disruption and O-GlcNAcylation upon glucose deprivation in HCC. We demonstrated that PCK1 deficiency can promote HBP-mediated UDP-GlcNAc biosynthesis through OAA accumulation and the AMPK-GFAT1 axis. Moreover, the OGT-mediated O-GlcNAcylation of CHK2 on Thr378 stabilizes CHK2 and promotes its oncogenic activity in HCC. The results of this study expand our understanding of PCK1 in hepatic carcinogenesis and indicate the potential of targeting HBP-mediated O-GlcNAcylation for HCC therapy.
Clinical specimens. HCC samples and paired, adjacent normal liver tissues were obtained from the Second Affiliated Hospital of Chongqing Medical University between 2015 and 2018, with approval from the IRB of Chongqing Medical University.
Adenovirus production. The full-length cDNA fragment of PCK1 (NM_002591) or G309R (PCK1 mutation 925G>A) was inserted into the pAdTrack-TO4 vector (gift from Tong-Chuan He, University of On the other hand, our previous work showed that enforced PCK1 expression leads to energy reduction and activates AMPK upon glucose deprivation in HCC (24). GFAT1 is an AMPK substrate, so we hypothesized that PCK1 may suppress HBP-mediated O-GlcNAcylation through the AMPK-GFAT1 axis. In support of this hypothesis, we found that PCK1 regulated GFAT1 phosphorylation and O-GlcNAcylation levels in an AMPK-dependent manner under glucose restrictions. As a rate-limiting enzyme of HBP, GFAT1 phosphorylation at Ser243 by AMPK diminishes its enzymatic activity (46), which decreases UDP-GlcNAc biosynthesis and O-GlcNAcylation levels. The AMPK-GFAT1 axis was also reported to regulate protein O-GlcNAcylation in angiogenesis and cardiac hypertrophy, indicating that this signal axis plays multiple physiological roles (25,26). In addition, AMPK directly phosphorylated residue Thr444 of OGT, the enzyme responsible for O-GlcNAcylation (47). Therefore, PCK1 may regulate the HBP and O-GlcNAcylation through different mechanisms.
It is known that O-GlcNAcylation is crucial for cell-cycle regulation and DNA-damage responses (48). Several proteins that regulate the growth and proliferation of tumor cells can be O-GlcNAcylated. For example, the G1/S checkpoint protein Rb is heavily O-GlcNAcylated during the G1 phase (49). By characterizing the role of O-GlcNAcylation upon PCK1 deficiency, we uncovered CHK2 as novel target of OGT. CHK2, a cell-cycle checkpoint kinase, plays key roles in DNA-damage responses and cell-cycle progression. CHK2 can directly interact with Rb or activate the p-CDK2/Cyclin A axis, thereby inducing the phosphorylation of Rb (50)(51)(52). The CHK2-modified p-Rb promotes chromosomal instability and tumor progression in HCC (29). In addition, CHK2 activates p-CDK2/Cyclin A and cell-cycle progression, resulting in hepatocyte growth (53). Rb is hyperphosphorylated in liver cancer (54,55). Hyperphosphorylated Rb can no longer bind to E2F, thereby effectively allowing E2F transcription factors to activate transcription, and generating a cellular environment that is permissive for cell proliferation (56,57). CHK2 is expressed in the nucleus in a subset of HCC and correlates with HCC progression (29). Several posttranslational modifications, including phosphorylation, ubiquitination, and acetylation, have been reported to be critical for CHK2 function (58)(59)(60). Herein, we identified Thr378 as a key O-GlcNAcylation site on CHK2 using LC-MS/MS. The loss of O-GlcNAcylation by the T378A mutation increased CHK2 ubiquitination, thus promoting its degradation. Moreover, we found that O-GlcNAcylation promoted CHK2 dimerization and activation, therefore enhancing Rb phosphorylation in PCK1-deficient hepatoma cells. Further structural analysis of O-GlcNAcylated CHK2 may help us understand how O-GlcNAcylation contributes to CHK2 activation.
Histopathology has revealed increased O-GlcNAcylation levels in HCC tumor tissues (61). Our findings not only provide an underlying mechanism whereby disrupted gluconeogenesis may activate the HBP and increase the availability of UDP-GlcNAc for O-GlcNAcylation under nutrient limitations, but also provide potential therapeutic targets for HCC. Preclinical evaluation of DON and AOA through the inhibition of glutamine metabolism has provided promising results for acute myeloid leukemia (21), high MYC-expressing atypical teratoid/rhabdoid tumors (62), Myc, T378A-Flag/T378A-Myc, or CHK2-Flag/T378A-Myc vectors. Precleared cell lysates were incubated overnight with an anti-Flag or anti-HA antibody, and coupled with 40 μL protein A/G agarose beads. Immunoprecipitated complexes were eluted and subjected to immunoblotting using the indicated antibodies.
sWGA pull-down assay. Hepatic cells and liver tissues were lysed in a buffer containing 125 mM NaCl, 50 mM Tris (pH 7.4), 5 mM EDTA, and 0.1% NP-40. The lysate was denatured in glycoprotein-denaturing buffer and digested with PNGase (NEB P0704S) to remove N-linked glycoproteins. Precleared lysates were incubated overnight with sWGA-conjugated agarose beads (Vector Laboratories). Precipitated complexes were eluted and immunoblotted with the indicated antibodies.
CHK2 O-GlcNAcylation site mapping. MS was performed to identify CHK2 O-GlcNAcylation sites, as previously described (67). Briefly, CHK2 was immunoprecipitated from PKO cells transfected with CHK2-Flag and subjected to 10% SDS-PAGE for Coomassie blue staining. The band corresponding to CHK2 was excised from the gel and digested overnight with trypsin. Online LC-MS/MS system consisting of an Easy-nLC system and an Orbitrap Fusion Lumos Tribrid MS instrument (Thermo Fisher Scientific) equipped with a nanoelectrospray ion source were performed. The data from the raw MS files were analyzed against the UniProt database with MaxQuant software (version 1.5.2.8). The first-search peptide tolerance was 20 ppm and the main-search peptide tolerance was 6 ppm. The MS/MS tolerance was 0.02 Da. The peptide-spectrum match and the false-discovery level was set to 1%. Matches between runs were determined and the minimum score for modified peptides was set to 40.
Quantitative analysis. Integral optical density (IOD) values were measured using Image-Pro Plus software (version 6.0, Media Cybernetics) to determine the intensity of protein expression. The IOD was calculated as the mean density times the area. The relative protein expression levels were normalized to an internal reference control, such as β-actin.
Statistics. All statistical analyses were performed using GraphPad Prism 7 (GraphPad Software Inc.). Data are represented as mean ± SD. Unpaired or paired 2-tailed Student's t test was used to compare 2 groups. One-way ANOVA followed by Tukey's test was used to compare more than 2 groups. Pearson's correlation coefficient was used to test linear correlations. Statistical significance was defined as P less than 0.05. *P < 0.05, **P < 0.01, ***P < 0.001.
Study approval. All animal procedures were performed according to protocols approved by the IACUC at the Laboratory Animal Center of Chongqing Medical University. All procedures were also approved by the Research Ethics Committee of Chongqing Medical University (reference number 2017012). Clinical samples were collected from patients after obtaining informed consent in accordance with a protocol approved by the Second Affiliated Hospital of Chongqing Medical University.
For further details regarding the materials used, please refer to the online supplement (Supplemental Methods and Supplemental Tables 1-6).

Author contributions
NT, AH, and KW conceived and designed the study. J Xiang, CC, RL, and DG performed most experiments and analyzed the data. LC and WZ performed CHK2 O-GlcNAcylation site mapping. HD and LT helped with data analysis. LL generated CHK2 mutants. QG, XP, and J Xia assisted with mice experiments. LH performed molec-Chicago, Chicago, Illinois, USA). Recombinant adenovirals AdPCK1 and AdG309R were generated using the AdEasy system as described previously (24). The adenoviral AdGFP, expressing only GFP, was used as a control.
CRISPR/Cas9-mediated knockout cells. PCK1-or CHEK2-knockout cells (PKO or CKO cells) were established using the CRISPR-Cas9 system (gift from Ding Xue, the School of Life Sciences, Tsinghua University, Beijing, China), as previously described (24). Single-cell HCC clones stably expressing single guide RNA (sgRNA) sequences were propagated and validated by immunoblotting and DNA sequencing. The sequences of all oligonucleotides used to generate the knockout cell lines are listed in Supplemental Table 3.

13
C 5 -Glutamine labeling assay. PKO cells were incubated in DMEM (no glucose and no glutamine, Gibco, A1443001) supplemented with dialyzed fetal bovine serum containing 0.3 mg/mL [U-13C]glutamine (Cambridge Isotope Laboratories) and 1 mg/mL unlabeled glucose for 12 hours. PCK1-overexpressing Hep1-SK cells were preformed using the same method as above. The cells were washed with ice-cold saline, quenched with 80% methanol in -80°C for 5 minutes, vortexed and centrifuged at 17,000g for 15 minutes at 4°C. The supernatant was collected and evaporated to dryness using a SpeedVac concentrator. Metabolites were vortexed and resuspended in 0.03% formic acid in analytical-grade water, and detected for use in liquid chromatography/triple quadrupole mass spectrometer (AB Sciex QTRAP 6500+) as previously described (66).