HBx-mediated GPT2 suppression promotes liver cancer development by downregulating ADH1A

Reagents and cell culture

Antibodies to Flag tag, GAPDH, HA tag, and AKT; MG132; HBV plasmid; cycloheximide (CHX); TRIM25 expression plasmids; TRIM25 short hairpin RNA (shRNA) plasmids; vectors containing various HBV genes; and other reagents were obtained as previously described^27–30. Antibodies targeting TRIM25, p-mTOR, FASN, SCD1, Ki67, β-tubulin, LSD1, CBL, GPT2, and mTOR were purchased from ABclonal (Wuhan, China), Zenbio (Chengdu, China), Huabio (Hangzhou, China), and Proteintech (Wuhan, China). The AKT activator SC79, the mTOR pathway activator MHY1485, and its inhibitor AZD8055 were obtained from MedChemExpress (Shanghai, China). Anti-ADH1A was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Ke Lei Biological Technology (Shanghai, China) provided vectors containing shRNA sequences against GPT2 (5′-GCCCAGAAGATGCTAAGAAACTCGAGTTTCTTAGCATCTTCTGGGCTTT TTT-3′), CBL (5′-GCCGATGTGAAATTAAAGGTATTCAA GAGATACCTTTAATTTCACATCGGCTTTTTT-3′), ADH1A (5′-GGATGCATTAATAACCCATGTTTCAAGAGAACATGGGTTATTAATGCATC CTTTTTT-3′), and LSD1 (5′-GCTACATCTTACCTTAGTCATTT CAAGAGAATGACTAAGGTAAGATGTAGCTTTTTT-3′); Flag-labeled-GPT2 expression plasmid, Flag-labeled-ADH1A expression plasmids, Flag-labeled-CBL expression plasmids, HA-labeled-NEDD4 expression plasmids, and Flag-GPT2 mutant plasmids, including mutants with or without aminotransferase class I and II domains (detailed information can be found in the Pfam database at http://pfam.xfam.org/); GPT2 plasmids (GPT2-Mut) containing C259R, N271T, and V478R mutants³¹; and mTOR mutant plasmids, including C-terminal (with mTOR domain), double-terminal (with FAT domain), and N-terminal deletions of mTOR mutants (with the PI3/4 kinase catalytic domain) (detailed information can be found in the Pfam database). HepG2, HEK293T, and Huh7 cells were cultured, transfected with target plasmids, and selected as previously described^28,29.

Detection of lipids, triglycerides, and cholesterol

The lipid droplet content in the target cells was measured with Oil Red O staining, according to the manufacturer’s instructions (Beyotime, Shanghai, China). The concentrations of total cholesterol (TC) and triglyceride (TG) in LC cells were determined with total TC and TG detection kits purchased from Solarbio (Beijing, China).

Clinical samples, immunohistochemistry, and H&E staining

A total of 112 HBV-positive tumor tissues, 88 HBV-negative LC tissues, and 80 adjacent tissues were collected from Outdo Biotech Co., Ltd. (Shanghai, China), and the Affiliated Hospital of Xuzhou Medical University. The tissues came from patients with LC who were treated with surgical resection. Immunohistochemistry (IHC) analysis was performed to assess the relative expression of GPT2 and ADH1A. Briefly, after deparaffinization, rehydration, and incubation with sodium citrate, the tissues were treated with H₂O₂; blocked with 5% goat serum; incubated with primary antibodies to GPT2 (1:200) and ADH1A (1:400) and subsequently with horseradish peroxidase-conjugated secondary antibodies (1:500); and stained with 3,3′-diaminobenzidine. The tissues were counterstained with hematoxylin. Hematoxylin-eosin (H&E) staining was performed as previously described³². GPT2 and ADH1A protein expression levels were calculated as previously described^33,34. This study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Ethics Committee of the Affiliated Hospital of Xuzhou Medical University (approval No. XYFY2022-KL085-01).

Animal transplantation

Four-week-old BALB/c nude mice were obtained from Cavenslasales Company (Changzhou, China). The mice were randomly divided into groups (mice per group n = 5) and housed under specific-pathogen-free, temperature-controlled conditions. Matrigel solution (BD Biosciences, New Jersey, USA) (0.1 mL) combined with cell suspensions (0.1 mL, 2 × 10⁷/mL) resuspended in phosphate-buffered saline was injected into the shoulders of nude mice. After rearing for 2–3 weeks, the mice were euthanized, and tumor samples were collected. Tumor weights and volumes were measured as previously described³³. All animal experiments were approved by the Animal Care and Use Committee of Xuzhou Medical University (approval No. L20210226151).

Bioinformatics analysis

Biological processes associated with GPT2 and ADH1A were predicted with gene set enrichment analysis (GSEA) and the ARCHS⁴ database, as previously described²⁸. Biological processes associated with LSD1 were predicted with the ARCHS⁴ database³⁵. Information on GPT2 expression was extracted from The Cancer Genome Atlas (TCGA) and the HCCDB database³⁶. LC tissues were divided into low or high GPT2 and ADH1A expression groups for GSEA or univariate survival analysis according to median gene expression levels. GPT2 protein expression data in HBV-associated LC tissues were obtained from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) database³⁷. The interaction between GPT2 and mTOR was predicted with the ARCHS⁴ database. Information on potential interactions between ADH1A and LSD1 was collected from the PrePPI database³⁸. Additionally, the predicted E3 ligases ADH1A and GPT2 were extracted from the UbiBrowser online database³⁹.

Immunofluorescence analysis

The localization of GPT2, mTOR, LSD1, ADH1A, CBL, and TRIM25 was detected with immunofluorescence. Briefly, after fixation with ice-cold acetone, LC cells were blocked with 5% bovine serum albumin in phosphate-buffered saline. The cells were then treated with primary antibodies targeting GPT2 (1:200), mTOR (1:100), LSD1 (1:100), ADH1A (1:50), CBL (1:300), or TRIM25 (1:200). After being washed 3 times, the cells were incubated with Alexa Fluor 594 (1:300)- or Alexa Fluor 488 (1:300)-conjugated secondary antibodies. The cells were then incubated with DAPI. The results of the immunofluorescence assays were obtained with an Olympus microscope.

Cell proliferation and migration assays

Cell growth efficiency was assessed with Cell Counting Kit-8 (CCK-8) assays, and cell clone formation assays were performed as previously described³⁴. To examine cell migration, we conducted a wound-healing experiment combined with a Transwell assay, as previously reported³³.

Real-time polymerase chain reaction (PCR)

The primers for the detection of GPT2 were (F) 5′-GGTCCTACA GTGCTAGCCAGGG-3′ and (R) 5′-AGAAATGCCGTCACTAGCTCCC-3′, whereas those for the detection of ADH1A were (F) 5′-GTG GCTGTAGGAATCTGTGGC-3′ and (R) 5′-ACTGAGGAATAGCGAGTGGGA-3′. Real-time PCR analysis was performed with SuperReal PreMix Plus reagent (SYBR Green) (Tiangen Biotech, Beijing, China). On the basis of the manufacturer’s instructions, the optimized real-time PCR conditions were as follows: 5 min at 95°C, and 35 cycles of 10 s at 95°C, 30 s at 58°C, and 30 s at 72°C. GPT2 and ADH1A levels were normalized to GAPDH levels. The primers were as previously described⁴⁰.

Co-immunoprecipitation, Western blot, and ubiquitination assays

Cell extracts were collected as previously described and incubated with primary antibodies to AKT (1:400), mTOR (1:200), LSD1 (1:200), ADH1A (1:400), CBL (1:200), or TRIM25 (1:500), and Protein G Sepharose beads (Santa Cruz Biotechnology, Santa Cruz, CA, USA). After the immunoprecipitates were washed 4 times, the target proteins were measured with Western blot, as previously described^28,29. Antibodies targeting GPT2 (1:2,000), ADH1A (1:5,000), AKT (1:1,000), p-mTOR (1:500), LSD1 (1:2,000), mTOR (1:2,000), TRIM25 (1:2,000), FASN (1:1,000), SCD1 (1:2,000), HBsAg (1:400), β-tubulin (1:3,000), and GAPDH (1:5,000) were used for Western blot to determine protein expression levels. Protein G Sepharose beads and anti-ubiquitin antibodies (1:600) were used to detect ubiquitination. Western blot was performed to evaluate the ubiquitination of LSD1 (1:1,000) and GPT2 (1:2,000).

Statistical analysis

Differences in data values between 2 groups are expressed as means ± standard deviations (SDs) from at least 3 independent experiments, and were compared with the t-test. Differences among 3 or more groups were compared with one-way analysis of variance (ANOVA) (in which the least significant difference method was used to compare discrepancies between 2 groups). The significance of the correlations between various target proteins estimated by IHC was assessed with the chi-square test. The Kaplan–Meier (KM) method was used to examine the correlation of GPT2 expression with overall survival (OS) and disease-free survival (DFS), and differences were assessed with the log-rank test. Cox proportional hazards models were used to determine whether GPT2 expression was an independent prognostic factor. The results are presented as hazard ratios and corresponding 95% confidence intervals. In addition, the was verified with the Schoenfeld residuals test. The threshold for statistical significance was set at P < 0.05.

GPT2 expression is significantly diminished in LC

To investigate the association between GPT2 and LC, we analyzed in-house clinical specimens and data from TCGA database. GPT2 expression was lower in LC tissues than adjacent normal tissues (Figure 1A, B). This finding was corroborated by analysis of 7 LC cohorts from the HCCDB database³⁶, which also demonstrated lower GPT2 expression in tumor tissues than adjacent non-tumor tissues (Figure 1C). Further analysis in TCGA cohort revealed an association between GPT2 expression and clinical parameters: GPT2 levels were lower in patients with LC with serum alpha fetoprotein (AFP) ≥ 20 ng/mL rather than < 20 ng/mL (Figure 1D). To further evaluate the clinical significance of GPT2, we divided patients into low- and high-expression groups according to median GPT2 expression. KM analysis with the log-rank test indicated that lower GPT2 expression was significantly associated with poorer OS but not DFS (Figure 1E, F). In addition, univariate Cox proportional hazards analysis identified low GPT2 expression as an independent prognostic factor for OS but not DFS (Figure 1G, H). Schoenfeld residual tests confirmed no violation of the proportional hazards assumption, thereby supporting the validity of the Cox models (Figure 1I, J). Together, these results indicated that low GPT2 expression significantly correlated with poor OS in patients with LC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

GPT2 expression and clinical association with LC. (A) GPT2 protein expression in LC tissues and adjacent tissues, detected with IHC analysis. (B) Relative expression of GPT2 in TCGA LC cohorts. (C) Relative expression of GPT2 and associated statistical results in 7 LC cohorts from the HCCDB database. (D) Association of GPT2 with AFP levels in patients with LC in TCGA LC cohorts. (E) KM analysis of the association between GPT2 and OS in TCGA LC cohorts. (F) KM analysis of the association between GPT2 and DFS in TCGA LC cohorts. (G) Univariate analyses of GPT2 expression for correlation with OS according to the Cox proportional hazards model. (H) Univariate analyses of GPT2 expression for correlation with DFS according to the Cox proportional hazards model. (I) Schoenfeld residual test on the proportional hazards assumption for GPT2 with OS. (J) Schoenfeld residual test on the proportional hazards assumption for GPT2 with DFS. Statistical analysis was performed with the chi-square test, t-test, and log-rank test. *P < 0.05, ***P < 0.001. AFP, alpha fetoprotein; DFS, disease-free survival; GPT2, glutamic-pyruvic transaminase 2; HCCDB, hepatocellular carcinoma database; IHC, immunohistochemistry; LC, liver cancer; OS, overall survival; TCGA, The Cancer Genome Atlas.

GPT2 inhibits the growth and migration of LC cells

We next assessed the effects of GPT2 on the proliferation and migration of LC cells. A plasmid for expression of Flag-tagged GPT2 was transfected into LC cells (Figure S1A). With CCK-8 and colony formation assays, we observed that GPT2 inhibited LC cell growth in vitro (Figure S1B, C). Transwell and wound healing assays (Figure S1D, E) further revealed that GPT2 overexpression decreased the migratory ability of LC cells. Additionally, in xenograft tumor models, GPT2 effectively suppressed tumor growth in vivo, according to measured tumor weight, volume, and expression of the proliferation marker Ki67 (Figure S1F, G). To further investigate GPT2 function, we generated a specific shRNA targeting GPT2. GPT2 knockdown (Figure S1H) promoted both LC cell growth and migration in vitro (Figure S1I–L). Moreover, GPT2 inhibition in nude mice enhanced tumor proliferation in vivo (Figure S1M, N). In summary, GPT2 was found to suppress malignant phenotypes in LC cells.

GPT2 regulates ADH1A expression in LC

GSEA was performed to further assess the biological processes associated with GPT2 in LC⁴¹ in TCGA tumor cohorts⁴². A significant association was observed between GPT2 expression and fatty acid metabolism (Figure 2A). In agreement with this finding, analysis of the ARCHS⁴ database³⁵ predicted GPT2’s involvement in both fatty acid metabolism and degradation (Figure 2B). We evaluated fatty acid metabolism-related genes associated with GPT2 with GSEA (Table S1), and selected ADH1A for further investigation. A previous analysis of TCGA data by our group has indicated ADH1A downregulation in LC¹³. Herein, ADH1A was similarly associated with fatty acid metabolism and degradation (Figure 2C, D) and therefore might be an important downstream target of GPT2 in modulating lipid metabolism.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Effect of GPT2 on ADH1A expression in LC. (A) GSEA of pathways associated with GPT2, on the basis of TCGA LC cohort. (B) KEGG pathways associated with GPT2 in the ARCHS⁴ database. (C) GSEA of pathways associated with ADH1A, on the basis of TCGA LC cohort. (D) KEGG pathways associated with ADH1A in the ARCHS⁴ database. (E) Relative expression levels of ADH1A and associated statistical results in 7 LC cohorts from the HCCDB database. (F) ADH1A expression in LC tissues, detected with IHC. (G) Association of GPT2 with ADH1A in LC tissues, detected with IHC. (H) Effect of GPT2 on ADH1A gene expression, examined with real-time PCR. (I) Influence of GPT2 on ADH1A protein expression, evaluated with Western blot. Mock, LC cells transfected with control expression plasmids; GPT2, LC cells transfected with GPT2 expression plasmids. Statistical analysis was performed with chi-square test and t-test. *P < 0.05, **P < 0.01, ***P < 0.001. ADH1A, alcohol dehydrogenase 1A; GPT2, glutamic-pyruvic transaminase 2; GSEA, gene set enrichment analysis; HCCDB, hepatocellular carcinoma database; IHC, immunohistochemistry; LC, liver cancer; TCGA, The Cancer Genome Atlas.

Analysis of the HCCDB database revealed diminished ADH1A expression in LC (Figure 2E) within the same cohorts (Figure 1C). IHC staining further demonstrated lower ADH1A expression in tumor tissues than adjacent normal tissues (Figure 2F). Moreover, we observed a significant positive correlation between GPT2 and ADH1A expression in LC tissues (Figure 2G). Using LC cell models, we evaluated the effect of GPT2 on ADH1A and observed that exogenous GPT2 increased ADH1A expression (Figure 2H, I). Together, these results highlighted the critical role of GPT2 in regulating ADH1A expression in LC.

We also examined whether ADH1A might be involved in GPT2-mediated regulation of LC cell behavior. Although GPT2 silencing promoted LC cell growth and migration, these effects were reversed by the addition of exogenous ADH1A (Figure S2). Therefore, ADH1A was determined to contribute to the inhibitory effects of GPT2 on LC cell growth and migration.

GPT2 suppresses the mTOR signaling pathway and consequently upregulates ADH1A

Next, we assessed the mechanism through which GPT2 modulates ADH1A expression. Given that the mTOR signaling pathway inhibits ADH1A¹⁴, we evaluated its role in regulating ADH1A expression in LC cells treated with the mTOR inhibitor AZD8055 and activator MHY1485. mTOR inhibition increased ADH1A expression, whereas mTOR activation decreased ADH1A levels (Figure 3A, B). Analysis of the ARCHS⁴ database suggested a potential interaction between GPT2 and mTOR (Figure S3A). Therefore, we examined whether GPT2 might regulate ADH1A via the mTOR pathway. Co-immunoprecipitation (co-IP) and immunofluorescence assays confirmed that GPT2 interacted and colocalized with mTOR in tumor cells (Figure S3B, C). We further generated GPT2 mutants containing or lacking aminotransferase class I/II domains (GPT2-1 and GPT2-2, respectively). The finding that GPT2-1 mutant, which retained these domains, interacted with mTOR indicated that aminotransferase domains are necessary for the GPT2–mTOR interaction (Figure S3D–F).

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Influence of GPT2 on mTOR signaling pathway activation and regulation of ADH1A expression. (A) Effects of inhibition of mTOR with AZD8055 on ADH1A expression in LC cells. (B) Effects of activation of mTOR with MHY1485 on ADH1A expression in LC cells. (C) Effects of GPT2 on mTOR activation and ADH1A expression in LC cells. (D) Effects of AKT activation mediated by the activator SC79 on mTOR sensitization and ADH1A expression in LC cells. (E) Effects of GPT2 on the activation of AKT and mTOR in LC cells. (F) Effects of GPT2 on mTOR activation and ADH1A expression mediated by AKT activation in LC cells. (G) AKT–mTOR interaction, assessed with co-IP assays. (H) Effects of GPT2 on AKT–mTOR interaction, measured with co-IP assays in LC cells. ADH1A, alcohol dehydrogenase 1A; AKT, protein kinase B; co-IP, co-immunoprecipitation; GPT2, glutamic-pyruvic transaminase 2; LC, liver cancer; mTOR, mammalian target of rapamycin.

Additionally, to determine which regions of mTOR bind GPT2, we constructed and expressed mTOR mutants with C-terminal, double-terminal, and N-terminal deletions (Figure S3G, H). Co-IP assays revealed that the double-terminal deletion mutant of mTOR interacted with GPT2, thereby suggesting that the central region of mTOR mediates its binding GPT2 (Figure S3I). Furthermore, GPT2 suppressed mTOR sensitization and promoted ADH1A upregulation (Figure 3C).

Subsequently, we investigated the mechanism through which GPT2 inhibited mTOR in regulating ADH1A expression. Given that AKT, a key kinase in the PI3-K pathway, enhances mTOR sensitization⁴³, we used the AKT activator SC97 to examine the involvement of AKT in mTOR sensitization in the regulation of ADH1A. AKT activation enhanced mTOR sensitization and suppressed ADH1A expression in LC cells (Figure 3D). We next tested whether GPT2 might influence mTOR via AKT. Although GPT2 did not alter AKT activation (Figure 3E), it impaired AKT-mediated mTOR sensitization (Figure 3F). Because GPT2 disrupted the binding between AKT and mTOR (Figure 3G, H), we concluded that decreased AKT-mTOR interaction contributed to GPT2-induced mTOR inhibition and subsequent ADH1A modulation in LC cells.

We also assessed whether GPT2 enzymatic activity might be necessary for mTOR pathway regulation. Mutations in the aminotransferase domains (C259R, N271T, and V478R; Figure S4A)³¹ impair GPT2 activity. A mutant GPT2 plasmid (GPT2-Mut) carrying these mutations was therefore transfected into LC cells. Co-IP assays revealed that GPT2-Mut weakened the GPT2–mTOR interaction (Figure S4B) but enhanced mTOR–AKT binding (Figure S4C). Although, as previously shown, wild-type GPT2 inhibited mTOR and consequently upregulated ADH1A, mTOR activity was restored, and ADH1A expression was diminished, in cells expressing the enzymatically impaired GPT2 mutant (Figure S4D). Therefore, overall, GPT2-mediated regulation of mTOR and ADH1A might depend on its enzymatic activity.

ADH1A facilitates decreased LSD1 expression and lipid accumulation mediated by GPT2 in LC cells

Although ADH1A might be involved in fatty acid metabolism, a process essential for hepatic lipid synthesis (Figure 2C, D)⁴⁴, its specific role in LC remains unclear. Oil Red O staining and measurements of TC and TG levels indicated that ADH1A suppressed lipid accumulation in LC cells (Figure 4A–C). To elucidate the mechanism through which ADH1A modulates lipid storage, we consulted the PrePPI database³⁸, and identified a potential interaction between ADH1A and LSD1 (Figure 4D), a histone demethylase that promotes lipid synthesis by upregulating key lipogenic enzymes⁴⁵. Additionally, data from the ARCHS⁴ database associated LSD1 with steroid biosynthesis, unsaturated fatty acid biosynthesis, and cancer-related pathways (Figure 4E). LSD1 therefore might potentially contribute to hepatocarcinogenesis by regulating ADH1A-mediated lipid accumulation.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Effects of ADH1A and GPT2 on LSD1 expression and lipid accumulation in LC. (A) Effect of ADH1A on lipid droplet accumulation, detected with Oil Red O staining. (B) Influence of ADH1A on triglyceride content in LC cells. (C) Effect of ADH1A on cholesterol content in LC cells. (D) Interaction between ADH1A and LSD1, predicted with the PrePPI database. (E) Predicted KEGG pathways relevant to LSD1 in the ARCHS⁴ database. (F) Influence of LSD1 on FASN and SCD1 expression in LC cells. (G) Influence of ADH1A on LSD1 expression in LC cells. (H) Interaction of ADH1A with LSD1, evaluated with co-IP assays. (I) Colocalization between ADH1A and LSD1, detected with immunofluorescence assays. (J) LSD1 gene expression in TCGA LC cohort. (K) Relationship between ADH1A and the LSD1 gene in TCGA LC cohort. (L) Effects of LSD1 on FASN and SCD1 expression mediated by ADH1A. (M) Effect of LSD1 on lipid droplets mediated by ADH1A. (N) Effect of LSD1 on triglyceride content mediated by ADH1A. (O) Effect of LSD1 on total cholesterol content mediated by AHD1A. Mock, LC cells transfected with control expression plasmids; ADH1A, LC cells transfected with ADH1A expression plasmids; ADH1A-CON, ADH1A-positive LC cells transfected with control expression plasmids; ADH1A-LSD1, ADH1A-positive LC cells transfected with LSD1 expression plasmids. Statistical analysis was performed with t-test and one-way ANOVA. *P < 0.05, **P < 0.01, ***P < 0.001. ADH1A, alcohol dehydrogenase 1A; ANOVA, analysis of variance; GPT2, glutamic-pyruvic transaminase 2; LC, liver cancer; LSD1, lysine-specific demethylase 1; TCGA, The Cancer Genome Atlas.

Because LSD1 upregulates the expression of lipogenic genes SCD1 and FASN⁴⁵, we examined its effects on these molecules in LC cells. LSD1 markedly increased the expression of SCD1 and FASN (Figure 4F). Moreover, ADH1A suppressed LSD1 expression (Figure 4G). Co-IP and immunofluorescence assays confirmed the interaction and colocalization of LSD1 and ADH1A (Figure 4H, I). In TCGA cohorts, LSD1 expression was higher in LC tissues than adjacent tissues but was negatively correlated with ADH1A levels (Figure 4J, K). We further demonstrated that ADH1A overexpression repressed FASN and SCD1 expression in LC cells. However, transfection of exogenous LSD1 into ADH1A-overexpressing LC cells restored the expression of FASN and SCD1 (Figure 4L). In contrast, knockdown of ADH1A with a specific shRNA upregulated FASN and SCD1 expression; however, this expression was inhibited when LSD1 was silenced in ADH1A-deficient cells (Figure S5). Additionally, assessments of lipid droplets, TC, and TG revealed that ADH1A overexpression suppressed lipid accumulation in LC cells, but this effect was reversed by ectopic LSD1 treatment (Figure 4M–O).

More importantly, we examined whether GPT2 might regulate LSD1, FASN, and SCD1 expression, as well as lipid accumulation in LC cells via ADH1A. In agreement with our prediction, GPT2 silencing elevated LSD1, FASN, and SCD1 levels, and increased lipid accumulation in LC cells, on the basis of lipid droplets, TG, and TC (Figure S6). However, when ectopic ADH1A was transfected, the GPT2 deficiency-induced upregulation of LSD1, FASN, and SCD1, and the enhanced lipid accumulation were suppressed.

CBL is crucial for ADH1A-mediated inhibition of LSD1 stabilization

Although ADH1A inhibits LSD1 expression through direct interactions, the underlying mechanisms remain unclear. Given that protein interactions can regulate target protein stability via the ubiquitin–proteasome system⁴⁶, we investigated whether ADH1A might affect LSD1 stabilization. In agreement with our hypothesis, treatment of ADH1A overexpressing LC cells with the protein synthesis inhibitor CHX, or the proteasome inhibitor MG-132 significantly decreased the half-life and stability of LSD1 (Figure 5A, B). Furthermore, ADH1A enhanced LSD1 ubiquitination in LC (Figure 5C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Inhibition of LSD1 stabilization mediated by ADH1A via CBL. (A) Effect of ADH1A on LSD1 stabilization after treatment of LC cells with CHX (200 μg/mL) to inhibit protein synthesis. *P < 0.05, mock vs. ADH1A group in HepG2 cells; ^#P < 0.05, mock vs. ADH1A group in Huh7 cells. (B) Influence of ADH1A on LSD1 stabilization after treatment of LC cells with the proteasome inhibitor MG-132 (100 nM). *P < 0.05, mock vs. ADH1A group in HepG2 cells; ^#P < 0.05, mock vs. ADH1A group in Huh7 cells. (C) Influence of ADH1A on LSD1 ubiquitination in LC cells. (D) Effect of CBL on LSD1 stabilization after treatment of LC cells with CHX (200 μg/mL) to inhibit protein synthesis. *P < 0.05, mock vs. CBL group in HepG2 cells; ^#P < 0.05, mock vs. CBL group in Huh7 cells. (E) Influence of CBL on LSD1 stabilization after treatment of LC cells with the proteasome inhibitor MG-132 (100 nM). *P < 0.05, mock vs. CBL group in HepG2 cells; ^#P < 0.05, mock vs. CBL group in Huh7 cells. (F) Influence of CBL on LSD1 ubiquitination in LC cells. (G) Inhibition of CBL protein expression with a specific shRNA. (H) Effect of CBL on LSD1 stabilization mediated by ADH1A after treatment of LC cells with CHX (200 μg/mL) to inhibit protein synthesis. *P < 0.05, mock vs. ADH1A-shCON group in HepG2 cells; ^#P < 0.05, mock vs. ADH1A-shCBL group in Huh7 cells. (I) Influence of CBL on LSD1 stabilization mediated by ADH1A after treatment of LC cells with the proteasome inhibitor MG-132 (100 nM). *P < 0.05, mock vs. ADH1A-shCON group in HepG2 cells; ^#P < 0.05, mock vs. ADH1A-shCBL group in Huh7 cells. (J) Effect of CBL on LSD1 ubiquitination mediated by ADH1A in LC cells. ADH1A, LC cells transfected with ADH1A expression plasmids; CBL, LC cells transfected with CBL expression plasmids; mock, LC cells transfected with control plasmids; shCON, LC cells transfected with shRNA control plasmids; shCBL, LC cells transfected with shRNA-targeting CBL plasmids. Statistical analysis was performed with t-test. ADH1A, alcohol dehydrogenase 1A; CBL, casitas B lineage lymphoma; CHX, cycloheximide; LC, liver cancer; LSD1, lysine-specific demethylase 1.

ADH1A decreased LSD1 stability in a ubiquitin–proteasome-dependent manner. However, ADH1A is not an E3 ubiquitin ligase capable of directly ubiquitinating target proteins. Analysis of the UbiBrowser database³⁹ predicted CBL as the highest-scoring potential E3 ligase for LSD1 (Figure S7A). In agreement with this finding, CBL suppressed LSD1 expression in LC (Figure S7B). Co-IP and immunofluorescence assays confirmed the physical interaction and colocalization of CBL with LSD1 (Figure S7C, D). Moreover, CBL significantly shortened the half-life and decreased the stability of LSD1 (Figure 5D, E), whereas ectopic expression of CBL in LC cells enhanced LSD1 ubiquitination (Figure 5F). Collectively, these results established CBL as a key E3 ubiquitin ligase for LSD1 in LC cells.

We further investigated whether CBL might be required for ADH1A-mediated inhibition of LSD1. In LC cells, ADH1A increased CBL protein levels and interacted with CBL (Figure S8A, B). More importantly, ADH1A enhanced the interaction between CBL and LSD1 (Figure S8C); therefore, we reasoned that CBL might be involved in ADH1A-induced suppression of LSD1. To test this hypothesis, we constructed an shRNA plasmid targeting CBL (Figure 5G). Silencing CBL in ADH1A-overexpressing LC cells significantly extended the half-life and increased the stability of LSD1 (Figure 5H, I), while decreasing its ubiquitination level (Figure 5J). Overall, ADH1A promotes the ubiquitination of LSD1 via CBL, thereby inhibiting its stabilization.

HBx decreases GPT2 expression, thus leading to lipid accumulation in LC cells

HBV infection is a significant risk factor for LC. On the basis of the proteomic data reported by Zuo et al.⁴⁷, we identified lower GPT2 protein levels in HBV-transgenic mice (Figure 6A). Therefore, we detected whether HBV might similarly suppress GPT2 expression in human LC. IHC analysis revealed lower GPT2 expression in HBV-associated LC tissues than adjacent non-tumor tissues (Figure 6B). This finding was further confirmed in HBV-positive tumor tissue data from the CPTAC database³⁷ (Figure 6C). Analysis of the HCCDB18 cohort indicated lower GPT2 expression in LC tissues associated with HBV, hepatitis C virus (HCV), and non-HBV/non-HCV etiologies than in adjacent tissues; therefore, multiple pathogenic factors appeared to downregulate GPT2 in tumors. Among these, GPT2 expression was lowest in HBV-associated LC and was significantly lower in HBV-associated LC than HCV-positive LC (Figure 6D). Overall, HBV might play a prominent role in suppressing GPT2 expression in LC. Next, using LC cell models, we demonstrated that HBV inhibited GPT2 protein expression (Figure 6E, F). Among viral genes, HBx, whether transiently or stably expressed, was responsible for this suppression (Figure 6G, H).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Effects of HBx on GPT2 expression in LC cells. (A) Effect of HBV on GPT2 protein expression in HBV-Tg mice. (B) GPT2 expression in HBV-positive (HBV+) adjacent tissues and HBV+ tumor tissues, detected with IHC. (C) GPT2 expression in HBV-positive LC tissues and adjacent tissues from the CPTAC database. (D) GPT2 expression in HBV-positive LC tissues, hepatitis C virus (HCV)-associated LC, non-HBV-HCV-associated LC tissues, and adjacent tissues from the HCCDB18 cohort. (E) Effect of HBV on GPT2 gene expression in LC cells, assessed with real-time PCR. (F) Role of HBV in GPT2 protein expression in LC cells, examined with western blotting. (G) Influence of transient transfection of various HBV genes on GPT2 protein expression. (H) Effect of stable HBx transfection on GPT2 protein expression. (I) Effects of HBx on ADH1A, FASN, SCD1, and LSD1 expression. (J) Effects of HBV on ADH1A, FASN, SCD1, and LSD1 expression. (K) Effects of GPT2 on ADH1A, FASN, SCD1, and LSD1 expression mediated by HBx. (L) Effect of GPT2 on lipid droplets mediated by HBx. (M) Effect of GPT2 on triglyceride content mediated by HBx. (N) Effect of GPT2 on cholesterol content mediated by HBx. WT, wild-type mice; HBV Tg, HBV transgenic mice; mock, LC cells transfected with control plasmids; HBV, LC cells transfected with HBV expression plasmids; HBx, LC cells transfected with HBx expression plasmids; HBx-CON, HBx-positive LC cells transfected with control expression plasmids; HBx-GPT2, HBx-positive LC cells transfected with GPT2 expression plasmids. Statistical analysis was performed with t-test and one-way ANOVA. *P < 0.05, **P < 0.01, ***P < 0.001. ADH1A, alcohol dehydrogenase 1A; ANOVA, analysis of variance; FASN, fatty acid synthase; GPT2, glutamic-pyruvic transaminase 2; HBV, hepatitis B virus; HBX, hepatitis B virus X protein; HCV, hepatitis C virus; LC, liver cancer; LSD1, lysine-specific demethylase 1; PCR, polymerase chain reaction; SCD1, stearoyl-CoA decarboxylase 1; TC, total cholesterol; TG, triglyceride.

In LC cells, we also observed that HBx downregulated ADH1A, and upregulated LSD1, FASN, and SCD1 (Figure 6I). Similar effects were observed in the full HBV infection group (Figure 6J). GPT2 repression in HBx-positive cells decreased ADH1A expression, and increased LSD1, FASN, and SCD1 expression (Figure 6K). Furthermore, we confirmed that HBx promoted lipid accumulation in LC cells by inhibiting GPT2 (Figure 6L–N).

HBx downregulates GPT2, thus facilitating LC cell growth and migration

We next examined the effects of GPT2 on HBx-mediated cell proliferation and migration. HBx enhanced the growth of LC cells. However, after transfection of exogenous GPT2 into HBx-positive cells, this proliferative effect was suppressed (Figure S9A, B). Similarly, HBx promoted LC cell migration, whereas this effect was reversed by ectopic expression of GPT2 (Figure S9C, D). In a subcutaneous xenograft model, HBx facilitated tumor formation. In vivo, exogenous GPT2 treatment of HBx-positive LC cells abrogated the HBx-induced increase in tumor growth, on the basis of tumor weight, volume, and Ki67 expression in tumor tissues (Figure S9E, F).

TRIM25 is involved in the suppression of GPT2 stabilization mediated by HBx

HBx plays a critical role in modulating the stabilization of target proteins via ubiquitination, thus facilitating LC development⁴⁸. To determine whether HBx might affect GPT2 stabilization, we treated LC cells with CHX and MG-132. HBx significantly decreased the half-life and stability of GPT2 (Figure S10A, B) and increased its ubiquitination levels in tumor cells (Figure S10C).

Given that HBx regulates target protein stabilization through specific E3 ligases⁴⁸, we sought to identify the E3 ligases involved in HBx-mediated GPT2 destabilization. Using the UbiBrowser database³⁹, we predicted potential E3 ligases for GPT2, among which CBL, NEDD4, and TRIM25 ranked highest (Figure 7A). We then evaluated the effects of these E3 ligases on GPT2 and found that TRIM25 markedly suppressed GPT2 expression (Figure 7B). Immunofluorescence and co-IP assays confirmed that TRIM25 colocalized and interacted with GPT2 (Figure 7C, D). Further experiments demonstrated that TRIM25 decreased the half-life and stability of GPT2 by promoting its ubiquitination (Figure 7E–G).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Regulation of GPT2 stabilization mediated by HBx via TRIM25. (A) Predicted E3 ligases of GPT2 in the UbiBrowser online database. (B) Influence of CBL, NEDD4, and TRIM25 on GPT2 expression. (C) Colocalization of TRIM25 with GPT2, examined with immunofluorescence. (D) Interaction of TRIM25 with GPT2, evaluated with Co-IP assays. (E) Effect of TRIM25 on GPT2 stabilization after treatment of LC cells with CHX (200 μg/mL) to inhibit protein synthesis. *P < 0.05, mock vs. TRIM25 group in HepG2 cells; ^#P < 0.05, mock vs. TRIM25 group in Huh7 cells. (F) Effect of TRIM25 on GPT2 stabilization after treatment of LC cells with the proteasome inhibitor MG-132 (100 nM). (G) Effect of TRIM25 on GPT2 ubiquitination in LC cells. (H) Effect of TRIM25 on GPT2 stabilization mediated by HBx after treatment of LC cells with CHX (200 μg/mL) to inhibit protein synthesis. *P < 0.05, shCON vs. shTRIM25 group in HepG2-HBx cells; ^#P < 0.05, shCON vs. shTRIM25 group in Huh7-HBx cells. (I) Effect of TRIM25 on GPT2 stabilization mediated by HBx after treatment of LC cells with the proteasome inhibitor MG-132 (100 nM). *P < 0.05, shCON vs. shTRIM25 group in HepG2-HBx cells; ^#P < 0.05, shCON vs. shTRIM25 group in Huh7-HBx cells. (J) Influence of TRIM25 on GPT2 ubiquitination mediated by HBx in LC cells. CBL, LC cells transfected with CBL expression plasmids; NEDD4, LC cells transfected with NEDD4 expression plasmids; shCON, LC cells transfected with shRNA control plasmids; shTRIM25, LC cells transfected with TRIM25 shRNA plasmids; TRIM25, LC cells transfected with TRIM25 expression plasmids. Statistical analysis was performed with t-test. CHX, cycloheximide; co-IP, co-immunoprecipitation; GPT2, glutamic-pyruvic transaminase 2; HBx, hepatitis B virus X protein; LC, liver cancer; TRIM25, tripartite motif containing 25.

In line with our previous report²⁸, HBx did not substantially affect TRIM25 expression but was found to bind TRIM25 (Figure S11A, B). Co-IP additionally verified the interaction between HBx and GPT2 in LC (Figure S11C). Using previously constructed HBx deletion mutants (HBx1: C-terminal deletion; HBx2: double-terminal deletion; and HBx3: N-terminal deletion)⁴⁹ (Figure S11D), we observed that HBx interacted with TRIM25 and GPT2 via distinct regions (Figure S11E). Moreover, HBx functioned as an adaptor strengthening the association between GPT2 and TRIM25 (Figure S11F). Therefore, HBx promoted GPT2 ubiquitination and destabilization in a TRIM25-dependent manner. In agreement with this finding, TRIM25 knockdown in HBx-expressing LC cells abolished the HBx-mediated decrease in GPT2 stability and the increase in GPT2 ubiquitination (Figure 7H–J).