High Glucose Stimulates Tumorigenesis in Hepatocellular Carcinoma Cells Through AGER-Dependent O-GlcNAcylation of c-Jun

Diabetes has been established as one of the most important risks in progression of hepatocellular carcinoma (HCC) (1,2). As the prevalence of both diabetes and HCC is substantially increasing worldwide (3), it is urgent to understand the underlying molecular mechanisms of how diabetes promotes HCC. It is well established that liver is fundamentally important for glucose metabolism, and hyperglycemia is one of the main characteristics of diabetes. However, whether and how high glucose (HG) stimulates liver tumorigenesis remain largely unclear.

HG reacts with long-lived proteins to form advanced glycation end products. Advanced glycation end products account for many diabetes complications through their engagement to advanced glycosylation end product–specific receptor (AGER) (4–6). Inhibition of AGER is regarded as a therapeutic option to prevent diabetes complications (7). Despite an epidemiologic study that has demonstrated that AGER may be involved in liver injury and subsequent carcinogenesis (8), the exact roles of AGER in liver tumorigenesis are still unknown.

Glucose, glutamine, and acetyl-coA are required for the hexosamine biosynthetic pathway (HBP) to synthesize uridine diphosphate (UDP)–N-acetyl-d-glucosamine (UDP-GlcNAc) (9). The HBP couples metabolic flux to control cell proliferation and participates in O-GlcNAcylation of target proteins through O-linked N-acetylglucosamine (GlcNAc) transferase (OGT) (9–11). Despite the fact that OGT and O-GlcNAcylation have been implicated as very important in the promotion of tumorigenesis in various types of cancer (10,11), very few studies have focused on their function in HCC. Given that HBP and subsequent O-GlcNAcylation are critical for both glucose metabolism and tumorigenesis, we supposed that they might be involved in HG-associated HCC.

Here, we demonstrate that HG is capable of promoting liver tumorigenesis via AGER. AGER stimulated O-GlcNAcylation and activation of c-Jun via enhancing HBP and OGT. We also uncovered that c-Jun could conversely upregulate AGER. Altogether, we establish a positive feedback loop from HG, AGER, and HBP to c-Jun, which may be helpful in the treatment of diabetic HCC.

Insulin-deficient HG and control mice models were constructed by intraperitoneal injection of streptozocin (STZ) (resolved in solution of 0.1 mol/L citrate acids, pH 4.4) and saline, respectively, into 5-week Balb/c male mice (Bikai, Shanghai, China). To investigate dose-dependent effects, we injected STZ at a final concentration of 40 mg/kg i.p. (STZ1) and 200 mg/kg i.p. (STZ2), respectively, into the mice. The mice were treated with STZ1 once a day for four consecutive days or treated with STZ2 only once. The STZ2-treated mice were also injected with or not with insulin (Sigma-Aldrich, St. Louis, MO) at a final concentration of 2 units/kg i.p. each day. For xenograft mouse experiments, Bel-7402 cells (5 × 10⁶ cells) under different treatments were subcutaneously injected into STZ1-, STZ2-, combined STZ2- and insulin-, or saline-treated 8-week-old athymic nude mice (Bikai). The tumor size was measured 8 weeks after injection, and the tumor volume was calculated as 0.5 × L × W², where L is length and W is width. All mouse experiments were performed according to the institution guidelines of Shanghai Tenth People’s Hospital.

Slides of tissue microarray analysis (TMA) were purchased from U.S. Biomax (Xi’an, China). Fresh HCC and adjacent normal liver tissues were acquired at Shanghai Ruijin Hospital with institutional approval. The HCC cell lines Bel-7402, SMMC-7721, Huh7, HepG2, SK-Hep1, and Bel-7404 and hepatocyte line HL-7702 cells were cultured in DMEM. Cells were treated with AG490 (Beyotime, Haimen, China) at a final concentration of 50–150 μmol/L, glucose (Sigma-Aldrich) at a final concentration of 5.5–50 mmol/L, PUGNAc (Sigma-Aldrich) at a final concentration of 25 μmol/L, GlcNAc (Sigma-Aldrich) at a concentration of 4 mmol/L, and cycloheximide (CHX) (Sigma-Aldrich) at a final concentration of 0.1 mg/mL, respectively. The cDNA encoding human AGER was purchased from OriGene (Beijing, China) and subcloned into the pLJM-based lentiviral vector with a COOH-terminal Myc tag. The expressing plasmids of COP1, OGT-Myc, c-Jun–Myc, and OGT–short hairpin RNA (sh)4 were obtained from OriGene. AGER-sh11, AGER-sh12, OGT-sh7, c-Jun–sh2, c-Jun–sh3, and c-Jun–sh4 were obtained from GeneChem (Shanghai, China). Luciferase (LUC) reporter of c-Jun was purchased from Beyotime. Promoter regions of human AGER gene were PCR amplified from gDNA of Bel-7402 cells and cloned into pGL4.21 (Promega, Madison, WI) vectors. The wild-type (WT) and S73A c-Jun–FLAG were subcloned into pcDNA3.1 (+) plasmids. The mutants were constructed by overlapping PCR. The primers used for this study are listed in Supplementary Table 1.

The primary antibodies used in immunohistochemistry (IHC) were anti-AGER (no. ab54741; Abcam, Hong Kong, China), anti–cleaved caspase 3 (no. 9664; Cell Signaling Technology [CST], Boston, MA), anti-Ki67 (no. ab15580; Abcam), anti–c-Jun (no. 9165; CST), and anti–O-GlcNAc (no. ab2739; Abcam).

For immunofluorescence (IF), the primary antibodies used were anti-AGER (no. ab54741; Abcam), anti-OGT (no. ab184198; Abcam), anti-histone (no. 4499; CST), anti–p-c-Jun (no. 3270; CST), and anti–c-Jun (no. 9165; CST).

For Western blotting (WB), the primary antibodies used were anti-GAPDH (no. 5174; CST), anti-OGT (no. ab184198 or no. ab177941; Abcam), anti–Myc-tag (no. 2276 or no. 2278; CST), anti-AGER (no. ab54741 or no. ab37647; Abcam), anti-GlcNAc (no. ab2739; Abcam), anti-ubiquitin (no. 3936; CST), anti–c-Jun (no. 9165 or no. 2315; CST), anti–p-c-Jun (for p-S73, no. 3270, and for p-S63, no. 9261; CST), anti-COP1 (no. sc-166799; Santa Cruz Biotechnology, Santa Cruz, CA), anti-PDK1 (no. 3062; CST), anti-EDG2 (no. ab166903; Abcam), anti-CD44 (no. ab51037; Abcam), anti–c-Abl (no. 2862; CST), anti-HSP27 (no. ab109376; Abcam), anti-FoxA1 (no. ab170933; Abcam), anti-CD166 (no. ab109215; Abcam), anti-MCAM (no. ab134065; Abcam), anti-TRIB2 (no. H00028951-m04; Abnova, Taipei, Taiwan), anti-p70S6K (no. ab32359; Abcam), anti-p38 (no. 9218; CST), anti-ERK1/2 (no. 4695; CST), anti-eIF4E (no. ab33766; Abcam), anti-MEK1 (no. 2352; CST), anti-SRSF1 (no. ab129108; Abcam), anti–c-Kit (no. 3074; CST), anti–c-Fos (no. 2250; CST), and anti-FLAG (no. 8146 or no. 2368; CST).

All IHC, IF, and WB were performed conventionally, and the protocols are available elsewhere.

LUC reporter, cell proliferation, caspase 3/7 activity, soft agar colony formation, and quantitative RT-PCR (qPCR) assays were performed as described previously (12,13). Chromatin immunoprecipitation (ChIP) was performed using the ChIP-IT express kit from Active Motif (Carlsbad, CA). Protein-DNA complexes were incubated with 3 μg anti–c-Jun antibodies (no. 9165; CST) or IgG (no. sc-2345; Santa Cruz Biotechnology). The primers used for qPCR are listed in Supplementary Table 1. Gas chromatography/mass spectrometry was performed by Bojian Biotechnology, Ltd. (Shanghai, China). Metabolites were analyzed by corresponding kits from Sigma-Aldrich. Glucose was analyzed by a kit from Applygen (Beijing, China).

Coimmunoprecipitation was performed as previously described (12,13). The reagents used were protein A/G-Sepharose (Novex, Oslo, Norway) and Western/immunoprecipation (IP) lysis buffer (Beyotime) with 0.1% SDS (used for denatured IP for ubiquitin and O-GlcNAc). The antibodies used for IP were anti-FLAG (no. 2368; CST), anti–c-Jun (no. 9165 or no. 2315; CST), and anti-AGER (ab37647; Abcam).

The protein ligation assay was performed to identify the interaction between OGT and c-Jun using the Duolink In Situ Red Starter Kit (mouse/rabbit) (Sigma-Aldrich, Uppsala, Sweden) as previously described (14).

Reaction mixtures containing 1 μmol/L purified human c-Jun protein (no. H00003725-P01; Abnova), 0.125 μmol/L human OGT (H00008473-P01; Abnova), and 10 μmol/L UDP-GlcNAc (Sigma-Aldrich) in a buffer of 50 mmol/L Tris/HCl (pH 7.5) and 1 mmol/L dithiothreitol were incubated at 37°C for 90 min. Then coimmunoprecipitation was performed by anti–c-Jun antibodies (no. 9165; CST) and detected by WB using anti–c-Jun (no. 2315; CST) and anti-GlcNAc (no. ab2739; Abcam) antibodies.

First, immunoprecipitated c-Jun with protein A/G-Sepharose (Novex, Oslo, Norway) was added into reaction buffer (20 mmol/L HEPES, pH 7.9; 50 mmol/L NaCl; 1 μmol/L PUGNAc; and 5 mmol/L MnCl₂ with protease and phosphatase inhibitors). Next, 2 μL Gal-T1Y289L (Invitrogen, Carlsbad, CA) and 2 μL 0.5 mmol/L UDP-GalNAz (Invitrogen) were added into reaction buffer to a volume of 20 μL. The reaction was performed overnight at 4°C. The beads were washed twice with reaction buffer to remove excess UDP-GalNAz. The samples were then reacted with biotin alkyne (Invitrogen) or tetramethyl-6-carboxyrhodamine (TAMRA) alkyne (Invitrogen) according to the manufacturer’s instructions. Biotin- or TAMRA-labeled samples were finally detected by WB using horseradish peroxidase (HRP)–labeled streptavidin (no. A0303; Beyotime) or antibodies against TAMRA (no. A6397; Invitrogen).

Tests to examine the differences between groups included the Student t test and χ² test. P < 0.05 was regarded as statistically significant.

Compared with the normal glucose level (5.5 mmol/L), increasing HG levels (15–50 mmol/L) resulted in a dose-dependent induced cell proliferation (Fig. 1A) and colony-formation capacity (Fig. 1B and C), whereas a reduced caspase 3/7 activity (Fig. 1D) was seen in HCC cell lines Bel-7402 and SMMC-7721, which have shown high carcinogenic properties in our previous studies (13,14).

Then, mice were intraperitoneally injected with two different concentrations of STZ to induce HG. We found that STZ could induce a dose-dependent elevation of serum glucose. By contrast, the HG that resulted from STZ-induced insulin deficiency could be reversed by intraperitoneal injection of insulin (Fig. 1E). Compared with the saline-treated control, there were more dividing cells with decreased levels of cleaved caspase 3, but increased levels of Ki67 can be detected in the liver of STZ-treated Balb/c mice (Fig. 1F), suggesting that HG promotes proliferation but inhibits apoptosis. In xenograft mice models, a dose-dependent increasing xenograft tumor volume was detected in nude mice treated with increasing concentrations of STZ (Fig. 1G). Similar to the serum glucose level (Fig. 1E), insulin could reverse STZ-accelerated xenograft growth (Fig. 1G). These results demonstrated that HG is capable of enhancing liver tumorigenesis.

Next, we investigated whether AGER is glucose inducible and important for tumorigenesis in HCC cells. AGER expression was tested in saline- and STZ-treated Balb/c mice, and it was found that STZ could induce a dose-dependent elevation of AGER in the liver but could not in the lung and colon. Interestingly, insulin was able to reverse STZ-induced elevation of AGER in the liver (Fig. 1H). These results suggested that AGER has important functions in liver when glucose is elevated. Also, AGER could be dose-dependently induced by increasing concentrations of HG in both Bel-7402 and SMMC-7721 cells (Fig. 1I). In human HCC tissues, AGER was highly upregulated compared with the paired adjacent normal liver tissues (Fig. 1J and K). Moreover, AGER had much higher expression levels in established HCC cell lines than that in the transformed hepatocyte line (Fig. 1L). To test whether AGER is important for HCC malignant function, we used two independent shRNAs against AGER with high knockdown efficiency (Fig. 1M). Contrary to the procarcinogenic roles of HG, depletion of AGER led to a significant reduction of cell proliferation (Fig. 1N) and colony-formation capacity (Fig. 1O and P) accompanied by an induction of caspase 3/7 activity in both Bel-7402 and SMMC-7721 cells (Fig. 1Q). Notably, HG-enhanced transformative phenotype was remarkably inhibited in AGER-depleted cells compared with the control (Supplementary Fig. 1A–C). In addition, knockdown of AGER in Bel-7402 cells impaired xenograft growth and inhibited a significant increase of tumor volume in STZ-treated nude mice compared with the saline-treated control (Fig. 1R). Together, these data illustrated an essential role of AGER in the HG-induced liver tumorigenesis.

For investigation of the function of AGER in metabolism, gas chromatography/mass spectrometry was performed. Among 366 metabolites tested, 22 metabolites were commonly changed in both Bel-7402 and SMMC-7721 cells when AGER was ectopically expressed (Fig. 2A and Supplementary Table 2). Among these metabolites, glutamine and glycogenic amino acids including isoleucine, phenylalanine, tyrosine, and threonine were elevated, while glutamic acid and lactic acid were reduced (Fig. 2A). All these changed metabolites are important for HBP (see schematic diagram of HBP [Fig. 2B]). To further verify the results gained from overexpression of AGER, we knocked down AGER in Bel-7402 and SMMC-7721 cells and reexamined metabolites involved in the HBP. As expected, lactic acid and glutamic acid were induced, while glutamine, acetyl-coA, and fructose-6-phosphate were reduced (Fig. 2C). However, glycogen, a metabolite not directly involved in HBP, was not changed (Fig. 2C). We also found that knockdown of AGER reduced both glucose consumption from culture media (Fig. 2D) and intracellular glucose levels (Fig. 2E), while overexpression of AGER led to the opposite effects (Fig. 2D and E). These results suggested that AGER may enhance metabolic flux to stimulate HBP.

To test whether AGER also regulates expression of genes involved in HBP, we performed qPCR to test mRNA levels of GLUT1, HK1 and -2, NUDT9, GUCY1A3, CANT1, GFPT1, GNPNAT1, PGM1, UAP1, OGT, SLC35A3, and PGM2 and -3 and found only OGT was upregulated by overexpression of AGER in both Bel-7402 and SMMC-7721 cells (Fig. 2F). By contrast, knockdown of AGER led to a reduction of OGT mRNA levels (Fig. 2G). Compared with the control, WB experiments also indicated an induction of OGT protein accompanied with increased levels of O-GlcNAcylation, an end event of HBP in both Bel-7402 and SMMC-7721 cells with AGER overexpressed (Fig. 2H), while a reduction of OGT protein was accompanied by decreased levels of O-GlcNAcylation in cells with AGER knocked down (Fig. 2I). These results further demonstrated that AGER may function as a regulator of HBP in HCC cells.

Because O-GlcNAcylation is the terminal effect of HBP, we then investigated the roles of O-GlcNAcylation in liver tumorigenesis. We found elevated levels of O-GlcNAcylation and OGT in the liver from STZ-treated Balb/c mice compared with the saline-treated control (Fig. 3A). Increased levels of O-GlcNAcylation and OGT were also detected in human HCC tissues (Fig. 3B) and established HCC cell lines (Fig. 3C), respectively, compared with the paired adjacent normal liver tissues and hepatocyte line. These data suggested that elevation of O-GlcNAcylation may be a common event in the liver from both HCC patients and patients with diabetes.

To test the importance of O-GlcNAcylation in liver tumorigenesis, we stimulated O-GlcNAcylation by using GlcNAc, an agonist of O-GlcNAcylation, and PUGNAc, an antagonist of O-GlcNAcase (an enzyme that has the opposite function against OGT). We found that treatment of GlcNAc and PUGNAc could significantly induce O-GlcNAcylation, and such effects were more obvious when GlcNAc and PUGNAc were simultaneously used (Fig. 3D). Because the efficacy of PUGNAc in stimulation of O-GlcNAcylation was more significant than that of GlcNAc (Fig. 3D), we used PUGNAc in combination with or without GlcNAc in the following study. We found stimulation of O-GlcNAcylation enhanced cell proliferation (Fig. 3E) and colony-formation capacity (Fig. 3F and G) while inhibiting caspase 3/7 activity (Fig. 3H). Similar to the endogenous O-GlcNAcylation levels (Fig. 3D), the transformative phenotypes stimulated by the combination of PUGNAc and GlcNAc were more obvious than that generated by PUGNAc alone (Fig. 3E–H). To exclude nonspecific effects by PUGNAc and GlcNAc, we knocked endogenous OGT down by specific shRNAs. We found that depletion of OGT could also lead to a significant reduction of O-GlcNAcylation (Fig. 3I) accompanied by significant reduction of cell proliferation (Fig. 3J) and colony-formation capacity (Fig. 3K and L) but induction of caspase 3/7 activity (Fig. 3M). By contrast, overexpression of OGT could stimulate O-GlcNAcylation and transformative phenotypes in both Bel-7402 and SMMC-7721 cells (Fig. 3N–Q). All these data suggested that OGT and O-GlcNAcylation are essential for tumorigenesis in HCC cells.

Next, we tried to figure out a target protein that is essential for HG and AGER/OGT/O-GlcNAcylation–dependent liver tumorigenesis. Firstly, we did a screen for proteins including SRSF1, MEK1, eIF4E, c-Jun, c-Fos, ERK1/2, p38/MAPK14, p70S6K, TRIB2, MCAM, CD166, FoxA1, HSP27, c-Abl, CD44, EDG2, PDK1, and c-Kit. These proteins are involved in tumorigenesis. We found that only proto-oncoprotein c-Jun was a candidate that could be upregulated by AGER overexpression in both Bel-7402 and SMMC-7721 cells (Fig. 4A and Supplementary Fig. 2). By contrast, c-Jun could be downregulated by AGER knockdown (Fig. 4B). The phosphorylation levels at Ser63 and Ser73, two major phosphorylation sites of c-Jun (15), were tested simultaneously. Opposite the levels of c-Jun, the levels of phosphorylation at Ser73 (hereafter p-c-Jun) were decreased by overexpression of AGER while being increased by knockdown of AGER (Fig. 4A and B). However, the levels of phosphorylation at Ser63 were changed in the same direction as c-Jun (Supplementary Fig. 3A–C), suggesting the changes of phosphorylation at Ser63 by AGER were due to the changes of c-Jun. These data also suggested that only phosphorylation at Ser73 and not at Ser63 is controlled by AGER. LUC reporter assays also demonstrated that AGER had a positive influence on the c-Jun activity (Fig. 4C). By adding increasing concentrations of glucose, we found c-Jun was dose-dependently increased, while p-c-Jun was dose-dependently decreased (Fig. 4D). Also, c-Jun activity was increased as a result of HG (Fig. 4E). Because c-Jun exerts its function mainly in the nucleus, we then detected nuclear localization of c-Jun using a microscope; we found that nuclear c-Jun was dose-dependently accumulated by glucose (Fig. 4F). Similar to the levels of serum glucose and AGER expression (Fig. 1E and H), a dose-dependent elevation of c-Jun and OGT accompanied by a dose-dependent downregulation of p-c-Jun was observed only in the liver from STZ-treated Balb/c mice compared with the saline-treated control (Fig. 4G), further demonstrating that the HG/AGER/OGT/c-Jun signaling axis is liver specific. Moreover, treatment of insulin was able to partially reverse such effects (Fig. 4G). Similar effects of STZ and insulin on the expressions of c-Jun and p-c-Jun were also detected in the mouse liver using IF experiments (Fig. 4H). These results suggested that c-Jun may be a downstream effector of AGER controlled by glucose metabolism.

Then we tested whether c-Jun is also regulated by OGT and O-GlcNAcylation. We found that treating Bel-7402 and SMMC-7721 cells with combined PUGNAc and GlcNAc had more obvious effects than treating PUGNAc alone on the induction of c-Jun but reduction of p-c-Jun (Fig. 4I). Moreover, c-Jun activity was increased after stimulation of O-GlcNAcylation (Fig. 4J). By gain and loss of function of OGT, we found that c-Jun was upregulated, while p-c-Jun was downregulated by OGT overexpression in both Bel-7402 and SMMC-7721 cells (Fig. 4K). By contrast, OGT knockdown had the opposite effects (Fig. 4L). Furthermore, c-Jun activity was also positively correlated with the OGT expression levels (Fig. 4M). These data suggested that OGT and O-GlcNAcylation are also critical in the regulation of c-Jun.

Next, we further investigated how AGER and OGT regulate c-Jun. We first excluded the possibility that AGER and OGT regulate mRNA levels of c-Jun because qPCR data indicated no changes of c-Jun mRNA before and after overexpression/knockdown of AGER or OGT (data not shown). However, we found that overexpression of OGT in Bel-7402 cells resulted in a prolonged half-life time of c-Jun protein as indicated by a CHX chase analysis (Fig. 5A and B). By contrast, knockdown of OGT accelerated degradation of c-Jun (Fig. 5C and D). Similarly, overexpression of AGER also extended the half-life time of c-Jun compared with the control (Fig. 5E and F), supporting a critical role of AGER and its downstream effector OGT in the regulation of protein stability of c-Jun.

Increased stabilization was usually accompanied by decreased ubiquitination (16,17); thereby, we hypothesized that OGT decreases ubiquitination of c-Jun. To address this, we overexpressed COP1, a known ubiquitin E3 ligase of c-Jun (18), in the absence or presence of increasing concentration of OGT in Bel-7402 cells and found that OGT was able to reverse ubiquitination and followed degradation of c-Jun by COP1 (Fig. 5G). By contrast, knocking OGT down resulted in an accumulated ubiquitination of c-Jun in both Bel-7402 and SMMC-7721 cells (Fig. 5H). These results further supported a positive role of OGT in the maintenance of c-Jun stability.

Then we performed function tests to further investigate the relationship between OGT and c-Jun. Despite the fact that overexpression of c-Jun alone could induce a significant elevation of c-Jun (Fig. 5I), transformative phenotypes were not enhanced as significantly as expected (Fig. 5J–L), suggesting that c-Jun had already overloaded in HCC cells. However, simultaneous overexpression of c-Jun was able to reverse impaired c-Jun expression (Fig. 5I), cell proliferation (Fig. 5J), colony-formation capacity (Fig. 5K), and increased caspase 3/7 activity (Fig. 5L) that was induced by knockdown of OGT, suggesting that OGT-stimulated tumorigenesis in HCC cells may be c-Jun dependent.

Because O-GlcNAcylation interplays with ubiquitination and followed degradation of target proteins (19), we investigated whether and how c-Jun was O-GlcNAcylated in HCC cells. We found that O-GlcNAcylation of c-Jun was detectable in both Bel-7402 and SMMC-7721 cells when overexpression of OGT was induced (Fig. 6A), while knockdown of OGT reduced O-GlcNAcylation of c-Jun (Fig. 6B). Stimulation of O-GlcNAcylation by PUGNAc and GlcNAc also resulted in an increased level of O-GlcNAcylation of c-Jun protein (Fig. 6C). In addition, by adding increasing concentrations of glucose, O-GlcNAcylation of c-Jun was dose-dependently induced (Fig. 6D). Furthermore, overexpression of AGER was capable of inducing O-GlcNAcylation of c-Jun (Fig. 6E). These results suggested that c-Jun can be O-GlcNAcylated in HCC cells in a glucose/AGER/OGT-dependent manner.

Confocal microscopic analysis indicated that c-Jun and OGT were colocalized (Fig. 6F). Moreover, the direct interactions between c-Jun and OGT were further confirmed by performing PLA analysis using a Duolink kit (Fig. 6G), suggesting that O-GlcNAcylation of c-Jun may occur through its interaction with OGT. Enzymatic labeling of O-GlcNAc site by using anti-TAMRA antibodies and HRP-labeled streptavidin, respectively, in Bel-7402 and SMMC-7721 cells provided evidence that c-Jun can be O-GlcNAcylated (Fig. 6H and I). In vitro O-GlcNAcylation experiments by mixing purified c-Jun and OGT proteins with UDP-GlcNAc also resulted in an O-GlcNAcylation of c-Jun (Fig. 6J), further supporting that c-Jun can be O-GlcNAcylated by OGT.

We have described a negative association between c-Jun and p-c-Jun, and stimulation of O-GlcNAcylation elevates c-Jun expression (Fig. 4); therefore, we hypothesized that O-GlcNAc and phosphate may competitively occupy Ser73. To address this, we replaced this serine by an alanine in the c-Jun protein and generated the S73A mutant. We found that O-GlcNAcylation of S73A mutant was much reduced compared with the WT one as indicated by enzymatic labeling of O-GlcNAc sites using anti-TAMRA antibodies and HRP-labeled streptavidin, respectively (Fig. 6K). We also found that stimulation of O-GlcNAcylation by PUGNAc with or without GlcNAc could result in a dose-dependent increase in O-GlcNAcylation of WT c-Jun; however, such effects were much reduced for the S73A mutant (Fig. 6L). These data revealed that O-GlcNAcylation of c-Jun is largely due to the O-GlcNAcylation at Ser73. Then we investigated whether O-GlcNAcylation of c-Jun is accompanied by dephosphorylation at Ser73. It was observed that O-GlcNAcylation levels of c-Jun were dose-dependently induced, while phosphorylation levels at Ser73 were dose-dependently reduced in the IPs pulled down by anti–c-Jun antibodies under the treatment of PUGNAc with or without GlcNAc (Fig. 6M). Therefore, we concluded that O-GlcNAcylation at Ser73 might play a large part in suppressing phosphorylation of c-Jun at the same site.

By performing TMA using IHC, we also revealed a significant positive correlation between O-GlcNAcylation and c-Jun in HCCs (Fig. 6N and O), further supporting the importance of the relationship between O-GlcNAcylation and c-Jun in clinical samples.

Then we tested whether AGER has direct impacts on the phosphorylation of c-Jun. We found that overexpression of AGER led to a significant decreased level of phosphorylation at Ser73 in WT c-Jun. However, due to the fact that O-GlcNAc and phosphate can occupy at the same site, there were no signals of phosphorylation at Ser73 in the S73A mutant c-Jun before or after overexpression of AGER (Supplementary Fig. 4). These data suggested that AGER may directly inhibit phosphorylation of c-Jun at Ser73.

We investigated whether c-Jun has a feedback on AGER. We found knockdown of c-Jun to be reduced (Fig. 7A), while overexpression of c-Jun dose-dependently induced protein levels of AGER (Fig. 7B). qPCR experiments indicated that AGER mRNA could also be reduced by knockdown of c-Jun (Fig. 7C) while induced by overexpression of c-Jun (Fig. 7D), suggesting that c-Jun may regulate AGER largely through a transcription-dependent mechanism.

Then, we constructed LUC reporters containing truncated versions of AGER promoter from –923 to 120 nt relative to the transcription start site (TSS, +1) of human AGER gene (gene identification no. 117). We found overexpression of c-Jun was unable to induce AGER promoter activity when the –334 to –257 region was lost because −256 LUC reporter had no response to c-Jun compared with the −334 LUC reporter in Bel-7402 cells (Fig. 7E). Hereafter, we named the –334 to –257 region as a c-Jun response element (JRE) in the following studies. We also found that overexpression of OGT had a similar stimulation of the AGER promoter activity as seen with overexpression of c-Jun (Fig. 7E), supporting the idea that OGT has a positive impact on c-Jun, as described above. Data from SMMC-7721 cells were also compromised (Fig. 7F). We then knocked down c-Jun and OGT, respectively, and found that promoter activities from the −334 LUC reporter were remarkably reduced, whereas no changes of promoter activities were detected from the −256 LUC reporter (Fig. 7G). Potential transcription factor binding sites, as indicated in Fig. 7H, were then predicated by PROMO online software (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3) within the JRE. Despite the fact that no c-Jun binding site was predicated, deletion of JRE in the −923 LUC reporter resulted in no response to either overexpression or knockdown of c-Jun/OGT compared with the WT one (Fig. 7I), suggesting that c-Jun may indirectly bind with JRE within the AGER promoter. Furthermore, physical binding of c-Jun around JRE was confirmed by ChIP; however, parallel ChIP analysis on the −2k and 2k regions, either upstream or downstream of the human AGER gene, showed no occupancy of c-Jun (Fig. 7J). These results demonstrated that c-Jun is capable of inducing AGER expression through activation of its promoter.

AGER may be regarded as a promising therapeutic target in HCC cells because AGER was detected mainly at the cell membrane or cytoplasm (Fig. 8A). Therefore, we tried to figure out a possible inhibitor by testing a series of commonly used chemical inhibitors. As shown in Fig. 8B, only treatment of AG490, a Janus kinase inhibitor, was found to significantly reduce AGER expression in Bel-7402 cells. In SMMC-7721 cells, AGER could also be dose-dependently reduced when increasing concentrations of AG490 were used (Fig. 8C). qPCR and CHX chase analysis demonstrated that AG490 was only able to reduce protein half-life time and not mRNA levels of AGER (data not shown and Fig. 8D and E). We further found that using AG490 could induce a significant ubiquitination of AGER (Fig. 8F). These results suggested that AG490 affects AGER expression maybe largely through inducing ubiquitination and reducing protein stability of AGER.

Then, we tested whether AG490 affects the function of AGER in HCC cells. O-GlcNAcylation could be dose-dependently reduced by increasing concentrations of AG490 in both Bel-7402 and SMMC-7721 cells (Fig. 8G). In addition, the levels of p-c-Jun were induced, while the levels of c-Jun were reduced (Fig. 8H), accompanied by reduced c-Jun activity (Fig. 8I) after addition of AG490 into the cultured Bel-7402 and SMMC-7721 cells. Furthermore, de-O-GlcNAcylation of c-Jun was also observed after treatment of cells with AG490 (Fig. 8J). By testing malignant phenotypes in Bel-7402 and SMMC-7721 cells, we found that cell proliferation (Fig. 8K) and colony-formation capacity (Fig. 8L) were reduced, while caspase 3/7 activities (Fig. 8M) were induced after AG490 was used. Interestingly, the inhibitory effects of AG490 on the malignant phenotypes were much reduced in cells with AGER ectopically expressed compared with the control (Fig. 8K–M), further demonstrating that AG490 maybe served as a potential inhibitor to AGER.

HCC has a strong relationship with diabetes (3). Despite the fact that the influences of some risk factors of diabetes (such as insulin, IGF-I, and chronic inflammation) on cancer initiation and progression have been extensively studied, the mechanism of how hyperglycemia promotes tumorigenesis in cancer, especially in HCC, has received less attention. Here, we only focused on the potential roles of HG in mice with STZ-induced diabetes and HCC cell lines. The risk factor of insulin was excluded because STZ is used to destroy pancreatic β-cells in mice and insulin was not added into the cell culture media. We found that HG supports cell proliferation, colony-formation capacity, and in vivo tumor growth while inhibiting apoptosis in HCC cells, which is in agreement with several large cohort and case-control studies that indicate that hyperglycemia stimulates carcinogenesis in other cancer types (20–22). HG not only can provide a high-energy source (23) but also can enhance expression of genes that are critical in maintaining transformative phenotypes of cancer cells (24). We identified that AGER, which has been demonstrated to account for diabetes complications (4), is also glucose inducible and critical for liver tumorigenesis. Therefore, we suggest that AGER and its downstream signaling may be responsible for diabetes-induced HCC. Interestingly, a recent study has shown that the expression of AGER is significantly higher in HCC tissues than in adjacent normal liver tissues and is closely associated with pathological staging and lymph-vascular space invasion (25), which further supports the importance of AGER in liver tumorigenesis.

It is known that metabolic disorders lead to HCC (26,27); however, it is still uncertain how AGER regulates metabolism in HCC cells. Evidence from our study suggests that AGER enhances glucose uptake, a well-established malignant phenotype of cancer cells (28), and acts as a stimulator of HBP and O-GlcNAcylation of target proteins in HCC cells. Recent work indicates that increased O-GlcNAcylation is a general feature of cancer and contributes to transformed phenotypes (29). In the current study, we concluded that there is a similar importance of O-GlcNAcylation in the maintenance of transformative phenotype in HCC cells. We further revealed that O-GlcNAcylation of c-Jun is a critical downstream event of AGER signaling and found that OGT-dependent induction of liver tumorigenesis is indispensible for c-Jun. Analogous to phosphorylation, O-GlcNAcylation plays crucial regulatory roles in cellular signaling. Emerging evidence has demonstrated at least up to four types of interplay between O-GlcNAcylation and phosphorylation on target proteins. These modifications can competitively and alternatively occupy either the same site or different sites within a protein. They can also simultaneously occupy different sites. At some occasions, competitive and simultaneous occupancies of O-GlcNAc and phosphate may occur within the same protein (30). Experimental analysis of c-Jun indicated that O-GlcNAc and phosphate competitively occupy Ser73. However, selective enrichment of low-abundance O-GlcNAcylated species, followed by high-technology mass spectrometry methods, were not involved in the current study, and mapping and quantification of multiple O-GlcNAc sites in c-Jun protein are needed in our future studies.

We also have revealed that HG can stimulate O-GlcNAcylation of c-Jun. Furthermore, the c-Jun protein can be stabilized by both AGER and OGT. As the procarcinogenic role of c-Jun has been implicated in liver tumorigenesis (31,32), we suppose that hyperglycemia stimulates liver tumorigenesis maybe via O-GlcNAcylation and stabilization of c-Jun in an AGER/OGT-dependent manner. Increased AGER signaling has been demonstrated to promote signaling nodes including p38, ERK1/2, and nuclear factor-κB (33–35). However, numerous studies also have suggested that AGER signaling is ligand- and cell type–dependent (4). This can explain why only c-Jun was found obviously upregulated under the stimulation of AGER in HCC cells. Interestingly, c-Jun can conversely increase AGER expression; therefore, a novel positive autoregulatory feedback was established in HCC cells. It is also not difficult to conclude that sustained activation of AGER, HBP, and c-Jun may lead to hyperglycemia-induced liver tumorigenesis. Notably, a follow-up study has demonstrated that inadequate maintenance of blood glucose in patients with diabetes is a significant risk factor for recurrence of HCC (36), supporting our finding that HG stimulates HCC.

In the current study, we also found that AG490 is a potential inhibitor of AGER because it can block AGER expression and its function. AG490 is a Janus kinase inhibitor that has been shown to inhibit cell growth in HCC cells (37). However, no studies have linked AG490 to glucose metabolic disorders in HCC cells. To the best of our knowledge, we are the first to report that AG490 can suppress the malignant property of HCC cells through blocking AGER-dependent O-GlcNAcylation of c-Jun. Therefore, AG490 may serve as a potential therapeutic agent for treatment of HCC in patients with diabetes.

Taken together, we report a potential mechanism underlying how diabetes boosts HCC. A signaling cascade from AGER to HBP to proto-oncoprotein c-Jun has been revealed that plays an important protumorigenic role in HG-induced liver tumorigenesis (Fig. 8N). Blocking this cascade may be helpful in the treatment of diabetic HCC.

Funding. This work was supported by the National Natural Science Foundation of China (81301689, 81201913, and 81202958), the Yangfan Project of Shanghai Municipal Science and Technology Commission (14YF1412300), the Shanghai Municipal Health and Family Planning Commission (20114y009), the Shanghai Jiao Tong University School of Medicine Outstanding Young Teacher Grant (to Y.Q.), the China central colleges and universities basic research–specific cross discipline grant (1501219096 [to J.W.]), the Outstanding Youth Training Program of Tongji University (1501219080), the Shanghai Tenth People’s Hospital Climbing Training Program (04.01.13024), and the Young College Teachers’ Training Scheme of Shanghai (ZZjdyx13007).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.Q. designed the study and researched and analyzed data. X.Z., Y.Z., and Y.W. researched data. Y.X. contributed to discussion. X.L. researched and analyzed data. F.S. designed the study. J.W. designed the study, researched data, and wrote the manuscript. J.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.