B3galt5 deficiency attenuates hepatocellular carcinoma by suppressing mTOR/p70s6k-mediated glycolysis

Hepatocellular carcinoma (HCC) is one of the most common malignancies with high morbidity and mortality. Beta-1,3-galactosyltransferase 5 (b3galt5) plays crucial roles in protein glycosylation, but its function in HCC remains unclear. Here, we investigated the role and underlying mechanism of b3galt5 in HCC. We found that b3galt5 is highly expressed and associated with a poor prognosis in HCC patients. In vitro studies showed that b3galt5 promoted the proliferation and survival of HCC cells. We also demonstrated that b3galt5 deficiency suppressed hepatocarcinogenesis in DEN/TCPOBOP-induced HCC. Further investigation confirmed that b3galt5 promoted aerobic glycolysis in HCC. Mechanistically, b3galt5 promoted glycolysis by activating the mTOR/p70s6k pathway through O-linked glycosylation modification on mTOR. Moreover, p70s6k inhibition reduced the expression of key glycolytic enzymes and the glycolysis rate in b3galt5-overexpressing cells. Our study uncovers a novel mechanism by which b3galt5 mediates glycolysis in HCC and highlights the b3galt5-mTOR/p70s6k axis as a potential target for HCC therapy.


Introduction
Protein glycosylation is one of the most common and complex modifications and includes N-linked and O-linked glycosylation [1,2]. It has been reported that protein glycosylation is usually altered in tumor cells, which affects cell adhesion, recognition, invasion, and metastasis [3]. In addition, glycoproteins are commonly used as tumor biomarkers in the clinic, such as alpha-fetoprotein (AFP) and glypican-3 for liver cancer, cancer antigen 125 (CA125) for ovarian cancer, carcinoembryonic antigen (CEA) for colon cancer, prostate-specific antigen (PSA) for prostate cancer, and carbohydrate antigen 19-9 (CA19-9), also known as sialyl-Lewis A (SLA), for gastrointestinal cancer [4].
Glycosylation is catalyzed by glycosyltransferase through the transfer of active glycans to specific receptors with several linkages [5]. The β-1,3-galactosyltransferase family, one of the major types of glycosyltransferase families, transfers active UDP-galactose, UDP-N-acetylglucosamine or UDP-N-acetylgalactosamine to acceptor sugars in a β-1,3 linkage. At present, there are seven members in the b3galt family (namely, b3galt1-7). B3galt5 is directly responsible for the synthesis of Galβ1-3GlcNAcβ (type 1 chain) and other N-glycans with the GlcNAc β1-3 Gal β1-4 GlcNAc β1-R side chains present in CEA [6]. B3galt5 also exhibits a preference for the O-linked core3 structure (GlcNAc β1-3 GalNAc) [7]. And the core3 structure is important in the O-linked glycosylation. It has been reported that b3galt5 contributes to the progression of diverse cancers. High expression of b3galt5 is associated with poor clinical outcomes, including shorter overall survival (OS) and Xiaoling Zhang and Hao Liu have contributed equally. recurrence-free survival (RFS) in breast cancer patients, especially early-stage patients [8]. In addition, b3galt5 promotes the proliferation, migration, invasion and epithelialmesenchymal transition (EMT) of breast cancer by upregulating the expression of β-catenin and EMT activator zinc finger E-box binding homeobox 1 (ZEB1) [8]. Chuang et al. reported that knockdown of b3galt5 disrupted the focal adhesion kinase (FAK)/caveolin-1/AKT/receptor-interacting protein kinase (RIP) complex and caused the dissociation of RIP from the complex, thus inducing breast cancer apoptosis [9]. B3galt5 expression promotes the production of CA19-9, which results in rapid and severe pancreatitis and aggressive pancreatic cancer in cooperation with the Kras G12D oncogene [10]. B3galt5 also functions as a tumor marker for the diagnosis of gynecological cancers. For instance, the b3galt4/ b3galt5-positive rates in the sera from ovarian cancer and uterine cervical cancer patients were comparable with the CA125 and SCC antigen-positive rates for these cancers [11].
Primary liver cancer (hepatocellular carcinoma, HCC), ranking second in malignant tumor-associated deaths, is one of the most common malignant tumors in China with a rising incidence [12]. The molecular mechanisms of HCC are very complex. Accumulating evidence indicates that metabolic alterations are deeply involved in HCC pathogenesis [13,14]. Metabolic reprogramming from oxidative phosphorylation (OXPHOS) to glycolysis was found in very early stages of hepatocarcinogenesis and was maintained in advanced HCC, suggesting that it is an important event in the progression of HCC development [15]. Studies have shown that liver cancer cells exploit glycolysis to fulfill energetic and biosynthetic needs of rapid proliferation and survival [16]. Multiple signaling pathways, such as mTOR/p70s6k, c-Myc, and p53, are important regulators of glycolysis in cancer [17,18] and increase the expression of key glycolytic enzymes, including hexokinase-2 (HK2), pyruvate kinase M2 (PKM2), and lactate dehydrogenase A (LDHA) [19].
High expression of b3galt5 is associated with advanced Tumor-Node-Metastasis stage, metastasis, poor survival and a high recurrence rate of HCC [20]. It was reported that b3galt5 was upregulated in metastatic HCC cell lines [21]. Moreover, hepatitis B virus X protein (HBx) induced the expression of b3galt5, which was associated with higher expression of SLA, thereby promoting liver cancer metastasis by interacting with endothelial cells [22]. However, the molecular mechanisms underlying b3galt5-mediated glycosylation in HCC have not been elaborated. In this study, we demonstrated that b3galt5 was upregulated in HCC and that its deficiency inhibited liver tumorigenesis in vivo and in vitro. Mechanistically, b3galt5 activated the mTOR/ p70s6k pathway through O-linked glycosylation modification on mTOR to enhance glycolysis in HCC. Thus, our study reveals an important role and novel mechanism of b3galt5-mediated glycosylation in promoting the progression of HCC, which indicates a potential target for the diagnosis and treatment of HCC.

HCC patient specimens
Tissue microarrays (TMAs), which contained 90 dots corresponding to HCC tissues and matching normal tissues with complete clinical pathological information, were purchased from Shanghai Outdo Biotech Co., Ltd. (Shanghai, China). Written informed consent was obtained from all enrolled subjects according to the Declaration of Helsinki. Three hundred and seven liver tumor tissues and adjacent nontumorous tissues were obtained from Sir Run Run Shaw Hospital. This study was approved by the Ethics Committee of Sir Run Run Shaw Hospital, Zhejiang University (20201113-30).

Cell lines and culture
The human HCC cell lines HUH7, PLC/PRF/5, HCCLM3, and HLE and the normal hepatocyte cell line LO2 were purchased from the Shanghai Institute of Biological Science, Chinese Academy of Science (Shanghai, China). All cells were cultured in DMEM (high glucose) with 10% fetal bovine serum (FBS, Gibco).

Mice
The b3galt5 −/− mice were a generous gift from Prof. Jinhan He (Department of Pharmacy, West China Hospital of Sichuan University). C57BL6/J mice were purchased from Shanghai Laboratory Animal Center (SLAC, Shanghai, China). All mice were bred and maintained in a specific pathogen-free facility at Zhejiang University Animal Center. All animal experiments were performed in accordance with the protocols approved by the Animal Care and Use Committee of Zhejiang University.
For b3galt5 knockout mice, HCC was induced by injection with DEN (25 mg/kg, i.p.) at day 14 followed by biweekly injections of TCPOBOP (2.5 mg/kg, i.p.) 5 times starting at 4 weeks. Mice were sacrificed at 20 weeks.

siRNA transfection
The target cells were cultured at 80% confluence and transfected with human b3galt5-specific siRNA, p70s6k-specific siRNA, or a negative siRNA control using RNAimax

Glycosidase treatment
Samples were heated at 100 °C for denaturation. The denatured proteins were chilled on ice, and then centrifuged. Afterward, peptide-N-glycosidase F (PNGase F, New England Biolabs) or O-glycosidase (New England Biolabs) was added to the denatured proteins, and incubation at 37 °C for 1 h was performed according to the manufacturer's instructions.

Coimmunoprecipitation (Co-IP) assay
Total proteins were extracted from target cells by NP40 buffer (Meilunbio, Dalian, China) according to the manufacturer's instructions. Ricinus communis agglutinin (RCA)-I/II lectin (Vector laboratories, CA, USA) antibody, which is a plant-derived protein that binds specifically but nonimmunologically to galactose or N-acetylgalactosamine residues of glycoconjugates [23], protein A/G agarose beads and protein lysate were incubated together at 4 °C overnight. The beads were washed 5 times with NP40 buffer and collected by centrifugation at 15,000 rpm. SDS (2 ×) was added to the beads, and the beads were subjected to western blotting.

Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) assays
The ECAR and OCR were analyzed using a Seahorse XFe 96 extracellular flux analyzer (Agilent, CA, USA). Briefly, 1.5 × 10 4 cells were plated in a Seahorse XF 96 cell culture microplate. For ECAR, 10 mM glucose, 1 μM oligomycin and 50 mM glycolytic inhibitor 2-DG were sequentially injected into each well at the indicated time points. For OCR, 1 μM oligomycin, 1 μM p-trifluoromethoxy carbonyl cyanide phenylhydrazone (FCCP) and 0.5 μM mitochondrial complex III inhibitor rotenone/antimycin A (Rote/AA) were sequentially injected into the wells. Data were analyzed by Seahorse XF-96 wave software. The ECAR and OCR were normalized to cell number.

Lactate content measurement
Mouse serum and cell supernatants were collected at the indicated times for lactate content measurement using a lactate colorimetric assay kit (Jiancheng, Nanjing, China) following the manufacturer's protocol.

RNA extraction and quantitative RT-PCR analysis
RNA was extracted from liver tissues using TRIzol reagent (Thermo Fisher Scientific, MA, USA), and mRNA was reverse transcribed into cDNA with the HiScript III RT SuperMix for qPCR Kit (Vazyme, Beijing, China). cDNA was quantified with the Ultra SYBR One Step qPCR Kit (CWBIO, Beijing, China). B2M was used as an endogenous control. The primer sequences were as follows: mouse b3galt5 forward CAT GCC CTT CAT CCA CCT GT, reverse ATA CCC AGT TGG GGA TCG AC; mouse B2M forward TCG CGC TAC TCT CTC TTT CT, reverse TGT CGG ATG GAT GAA ACC CA.

Statistical analysis
Data are presented as the mean ± SEM of the indicated number of litters. mRNA expression, tumor volume, tumor number and tumor weight/body weight ratio were analyzed using two-tailed unpaired Student's t test. The analysis was performed with GraphPad Prism 7 software. The correlation analysis was measured by Pearson analysis. P < 0.05 was considered statistically significant.
All other methods are described in the supplementary materials.

B3galt5 is upregulated in HCC patients
We first examined the mRNA and protein expression of b3galt5 in tumor tissues and adjacent normal tissues in HCC patients using RNA-seq and western blotting. The mRNA level of b3galt5 in HCC tissues was significantly increased compared with that in paired normal tissues (Fig. S1A). Analysis of b3galt5 expression in HCC from the Gene Expression Omnibus dataset GSE1898 also confirmed a consistently higher b3galt5 level in tumor tissues (Fig. S1B). Furthermore, we determined the protein level of b3galt5 in HCC samples, which was also dramatically elevated compared with the adjacent normal tissues, indicating the upregulation of b3galt5 in HCC tissues (Fig. S1C). Importantly, Kaplan-Meier survival analysis revealed that HCC patients with high b3galt5 expression (n = 259) were associated with poor overall survival, while low expression (n = 111) prolonged overall survival (Fig. S1D). Overall, these results indicated that b3galt5 might act as a tumor promoter in HCC.

B3galt5 promotes proliferation and inhibits apoptosis in HCC cells
To explore the biological functions of b3galt5 in HCC cells, we first screened two cell lines, HLE and HCCLM3, with high b3galt5 expression profiles from a series of HCC cell lines (Fig. S2A). Then, we silenced b3galt5 in HLE and HCCLM3 cells using specific siRNA. Intriguingly, the proliferation of these two cell lines was significantly inhibited after b3galt5 knockdown, as determined by the CCK-8 assay (Fig. S2B). The apoptotic rates of HLE and HCCLM3 cells with b3galt5 silencing were 12.83% ± 0.06% and 16.87% ± 0.22%, respectively, compared with rates of 7.60% ± 0.10% and 12.27% ± 1.83% in siRNA control cells, respectively ( Fig. S2C and D). Moreover, b3galt5 knockdown disrupted cell cycle progression and induced G0/ G1-phase arrest ( Fig. S2E and F). Taken together, these results indicated that b3galt5 promoted proliferation and inhibited apoptosis in HCC cells.

B3galt5 deficiency attenuates DEN/ TCPOBOP-induced hepatocarcinogenesis
To further investigate the role of b3galt5 in liver tumorigenesis, we attempted to silence b3galt5 in a DEN/TCPOBOPinduced mouse HCC model. C57BL/6 male mice were treated with DEN at day 14 followed by biweekly injections of TCPOBOP for 8 times starting at 4 weeks postpartum. Then, mice were given tail vein injections of shNC AAV8 or shb3 AAV8 at 18 weeks old (Fig. S3A). As expected, b3galt5 expression was markedly decreased in shb3 AAV8treated mice ( Fig. S3B and C), and the incidence of liver cancer in shb3 AAV8 mice was lower than that in shNC AAV8 mice (Fig. S3D). Histopathological analysis of livers showed that there were more fat deposits and more severe dysplasia in shNC AAV8 mice than in shb3 AAV8 mice (Fig. S3E). Quantitative analysis showed that shb3 AAV8 mice had an ~ 60% decrease in tumor number (Fig. S3F). In addition, although the difference was not statistically significant, the tumor volume showed a decreasing trend in shb3 AAV8 mice (Fig. S3G). The liver weight/body weight ratio was similar in both groups of mice (Fig. S3H). We further assessed the IHC scores of PCNA and cleaved caspase-3 in shNC AAV8 and shb3 AAV8 mice, which were 5.14 ± 1.10 vs 1.43 ± 0.78 and 2 ± 0 vs 7.83 ± 1.93, respectively, indicating massive remission of proliferation-positive cells and induction of apoptosis in the livers after b3galt5 silencing (Fig. S3I). In addition, hepatic fibrosis was markedly alleviated in shb3 AAV8 mice with b3galt5 deficiency, as determined by α-smooth muscle actin (α-SMA) staining (Fig.  S3I).
Furthermore, we investigated liver tumorigenesis in b3galt5 −/− (KO) and wild-type (WT) mice under DEN/ TCPOBOP challenge (Fig. 1A). Mice were sacrificed, and whole livers were collected and analyzed 18 weeks after DEN injection. As expected, the protein and mRNA level of b3galt5 were effectively decreased in KO mice (Fig. 1B-D). Notably, KO mice developed fewer tumors in the liver than WT mice (Fig. 1E). The KO livers exhibited smaller neoplastic lesions and attenuated cirrhosis compared with control littermates, as shown by H&E staining (Fig. 1F). Quantitative analysis revealed that the number of tumor nodules in KO mice was only 3.07 ± 0.50 per liver, which was significantly lower than that in WT mice (12.14 ± 2.83 per liver, Fig. 1G). The tumor volume of KO mice was also markedly reduced compared to that of WT mice (30.33 ± 8.65 vs 131.9 ± 43.58 mm 3 , respectively) (Fig. 1H), accompanied by an ~ 20% reduction in the liver/body weight ratio (Fig. 1I). In addition, histopathological analysis demonstrated markedly inhibited proliferation, a significant induction of apoptosis, and attenuated hepatic fibrosis in the livers of KO mice compared to WT mice (Fig. 1J).
Collectively, these results revealed that b3galt5 deficiency attenuated DEN/TCPOBOP-induced HCC in mice.

B3galt5 promotes glycolysis in HCC
To investigate the underlying molecular mechanism of the tumorigenic effect of b3galt5, we performed metabolomics analysis of liver cancer cells from shb3 AAV8 and shNC AAV8 treated mice. The results revealed that the metabolites were significantly changed when b3galt5 was knocked down (Fig. S4A). Kyoto Encyclopedia of Gene and Genomes (KEGG) enrichment analysis identified that multiple carbohydrate metabolism signaling pathways were enriched, including pyruvate metabolism, pentose and glucuronate interconversion, amino sugar and nucleotide sugar metabolism and OXPHOS (Fig. S4B). Next, we performed proteomic sequencing of liver cancer cells from shb3 AAV8and shNC AAV8-treated mice. A number of key proteins closely associated with OXPHOS, including Cox7a2l [25], Ndufaf3 [26], Cmc1 [27] and Sdhaf4 [28], were significantly upregulated upon b3galt5 knockdown (Fig. S4C). To further explore the function of the b3galt5-regulated proteins, we conducted KEGG enrichment analysis. Our results showed that proteins involved in amino nucleic acid sugar metabolism, glycolysis and gluconeogenesis, and OXPHOS were markedly enriched in b3galt5 knockdown mice (Fig.  S4D), suggesting the potential roles of b3galt5 in glucose metabolism.
Inspired by metabolomics analysis and proteomic sequencing, we then examined whether b3galt5 regulated glycolysis in HCC. Compared with control mice, knockout of b3galt5 resulted in a significant decrease in key glycolytic enzymes, including HK2, 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 3 (PFKFB3) and LDHA ( Fig. 2A). Immunohistochemical staining also confirmed that the expression of LDHA, HK2, and PFKFB3 was decreased in the livers of KO mice (Fig. S5A-C). Additionally, ablation of b3galt5 led to a significant reduction in lactate production and decreased lactate dehydrogenase, hexokinase, and phosphofructokinase activity in the serum of KO mice (Fig. 2B-E). It should be noted that the absence of b3galt5 did not lead to s obvious changes in total pyruvate kinase activity (Fig. 2F), which might be because pyruvate kinase has a variety of isoenzymes, while pyruvate kinase M2 is the main form expressed in liver tissue [29].
In HCCLM3 and HLE cells, silencing b3galt5 likewise decreased the expression of several key glycolytic enzymes, such as HK2, PFKFB3, LDHA and pyruvate dehydrogenase kinase 1 (PDK1) (Fig. 2G). Conversely, overexpression of b3galt5 in LO2 and HUH7 cells increased the expression of these glycolytic enzymes (Fig. 2H). Moreover, b3galt5 knockdown significantly reduced lactate production, while b3galt5 overexpression had the opposite effect (Fig. 2I). Accordingly, we observed an obviously increased rate of ECAR, as indicated by the increased glycolysis rate and glycolysis capacity in b3galt5-overexpressing cells (Fig. 2J,  K). Knockdown of b3galt5 in HCCLM3 and HLE cells led to opposing effects (Fig. 2L, M). In addition, we also detected the OCR in HCC cells. As expected, knockdown of b3galt5 in HCCLM3 and HLE cells led to a significant increase in the OCR. In particular, the maximal respiration in HLE cells increased from 42.1 ± 3.8 to 49.1 ± 6.2 pmol/ min, and in HCCLM3 cells, it increased from 31.0 ± 3.9 to 64.7 ± 5.5 pmol/min (Fig. 3A-C). In contrast, a significant reduction in maximal respiration in LO2 and PLC/PRF5 cells was observed after b3galt5 overexpression (Fig. 3D-F). All these compelling results suggested that b3galt5 promoted glycolysis in HCC.

B3galt5 promotes glycolysis in HCC via the mTOR pathway
The mTOR pathway is one of the main signaling pathways that regulates glycolysis. To clarify its role in the promotion of glycolysis by b3galt5, we detected the total and phosphorylated protein levels of mTOR and its downstream target, p70s6k. Immunoblot analysis showed that both phosphorylated and total protein levels of mTOR, as well as p70s6k, were downregulated in b3galt5-KO liver cells (Fig. 4A). Consistent results could also be observed in HLE and HCCLM3 cells (Fig. 4B). Conversely, in b3galt5overexpressing LO2 and HUH7 cells, total and phosphorylated mTOR and p70s6k levels were accordingly upregulated (Fig. 4C). As expected, these changes were in line with the expression of key glycolytic enzymes. To further verify the clinical correlation between b3galt5 and p-p70s6k, we performed immunohistochemistry staining in a 90-dot tissue microarray (Fig. S6A). Encouragingly, b3galt5 expression was positively correlated with p-p70s6k in HCC samples (r = 0.32, p = 0.0021) (Fig. S6B). Then, we examined whether activation of mTOR mediated the effects of b3galt5 on glycolysis. For this purpose, we treated HCC cells with a p70s6k activator or inhibitor   (Fig. 5A). Conversely, the inhibition of p70s6k by PF-4708671 reduced the levels of these key glycolytic enzymes in b3galt5-overexpressing LO2 and PLC/PRF5 cells (Fig. 5B, C). To further confirm the role of p70s6k in b3galt5-mediated glycolysis, we carried out live monitoring using a Seahorse XF Extracellular-Flux Analyzer for ECAR. Knockdown of p70s6k markedly inhibited the glycolysis rate and glycolysis capacity (Fig. 5D-F). Overexpression of p70s6k in b3galt5 knockdown cells rescued the glycolysis rate and glycolysis capacity (Fig. 5G, H).
Collectively, these results illustrated that b3galt5 promoted glycolysis by activating the mTOR/p70s6k pathway.

B3galt5 regulates the O-linked glycosylation of mTOR
It has been reported that glycosylation mediates protein stability, function and transport. We therefore investigated whether the activation of mTOR is regulated by glycosylation. RCA-lectin is a reagent that is generally used to identify the galactose or N-acetylgalactosamine residues of glycoconjugates [30]. Thus, we traced the galactose transferring mediated by b3galt5 in the elongation of glycoconjugates. Immunoblot analysis revealed that b3galt5 knockdown significantly blocked the glycosylation modification of p-mTOR, while overexpression of b3galt5 enhanced their glycosylation modification (Fig. 6A). In the immunofluorescence assay, we confirmed a high fluorescence intensity of p-mTOR after b3galt5 overexpression, accompanied by the high fluorescence intensity of RCA-lectin, indicating the presence of galactose in p-mTOR (Fig. 6B). In contrast, b3galt5 knockdown correspondingly suppressed the fluorescence intensity (Fig. 6C).
B3galt5 is involved in both N-linked and O-linked glycosylation. PNGase F is an enzyme that hydrolyses the N-linked oligosaccharides [31]. O-glycosidase is an enzyme that catalyzes the removal of core 1 and core 3 O-linked disaccharides from glycoproteins, which is also known as Endo-α-N-acetylgalactosaminidase [32]. To further confirm the glycosylation of mTOR, we treated LO2 and PLC/PRF5 cells with PNGase F and O-glycosidase. Interestingly, the molecular weight of mTOR was significantly smaller upon O-glycosidase treatment, which increased its electrophoretic mobility (Fig. 7A). However, no changes in molecular weight or electrophoretic mobility of mTOR were observed after PNGase F treatment (Fig. 7B). Then, we examined whether b3galt5 mediated the glycosylation of mTOR. For this purpose, we treated HCC cells with O-glycosidase. The results indicated that the ratio of glycosylated mTOR decreased from 0.53 ± 0.28 to 0.30 ± 0.15 after b3galt5 silence in HCCLM3 cells, while the ratio increased from 0.64 ± 0.17 to 1.39 ± 0.20 in PLC/PRF5 cells after b3galt5 overexpressing (Fig. 7C, D). So the change of glycosylated mTOR was in line with the total mTOR. In addition, benzylα-GalNAc, an O-linked glycosylation inhibitor, led to the decreased expression of phosphorylated mTOR in HCCLM3 and LO2 cells (Fig. 7E, F).
Above all, these findings suggested that b3galt5 O-glycosylated and activated the mTOR pathway in HCC cells.

Discussion
Here, we reported that b3galt5 might be a driving factor in the development of HCC by potentiating glycolysis. We first identified that b3galt5 expression was highly elevated in HCC and positively associated with a poor prognosis. Then, we demonstrated that b3galt5 deficiency inhibited cell proliferation and hepatocarcinogenesis in vitro and in vivo. Mechanistically, we confirmed that b3galt5 played a tumorigenic role by promoting glycolysis in an mTOR O-glycosylation-dependent manner. Moreover, b3galt5 deficiency led to an impaired glycolytic phenotype, as indicated by decreased expression of key glycolytic enzymes and the glycolysis rate, further supporting the critical role of b3galt5 in HCC glucose metabolism.
The mTOR and HIF-α pathways are canonical pathways that regulate glycolysis [19]. mTOR inhibition was also reported to retarded the DEN-induced liver cancer model by down-regulating glycolysis capacity [33]. For example, farnesyl-diphosphate farnesyltransferase 1 downregulates glycolysis in colon cancer by inactivating the mTOR pathway [17]. Active mTORC1 directly increases glucose flux when the pentose phosphate pathway backs up into glycolysis, thereby circumventing a glycolysis block and ensuring adequate ATP and biomass production [34]. In our present study, we found that b3galt5 promotes aerobic glycolysis in hepatocellular carcinoma. This observation prompted us to further explore whether b3galt5 promoted aerobic glycolysis by regulating the mTOR pathway. We observed that depletion or silencing of b3galt5 in cells inhibited mTOR activation (Fig. 4). Surprisingly, total mTOR and p70s6k levels were also decreased. However, we could not rule out that the mTOR and p70s6k transcript levels were changed. Additional studies are needed to investigate whether b3galt5 regulates mTOR translation or affects their protein stability. Consistently, the impaired glycolytic phenotype caused by b3galt5 knockdown was partially rescued by activation or overexpression of p70s6k (Fig. 5G, H), while the activated glycolytic phenotype induced by b3galt5 overexpression could be partially reversed by inhibition or knockdown of p70s6k ( Fig. 5D-F), which supported our hypothesis. Therefore, promotion of glycolysis through mTOR activation by b3galt5 might represent one mechanism for hepatocellular carcinoma to facilitate cell survival.
Altered glycosylation modification can markedly affect the functions of proteins, including interactions and signal transduction [35]. For instance, glycosylation of PD-L1 stabilizes its structure and thus participates in the interaction between PD-L1 and PD-1 [36]. Also the FUT8 mediated glycosylation on B7-H3 also affected its stability [37]. Considering that b3galt5 was involved in both N-linked and O-linked glycosylation, we further explored the underlying mechanisms of mTOR activation. To this end, we used RCAlectin to confirm that b3galt5 participated in the glycosylation modification of mTOR. Then, we examined the expression of mTOR using the PNGase F and O-glycosidase. Our data demonstrated that O-glycosidase rather than PNGase F obviously reduced the molecular weight of mTOR. The ratio of glycosylated mTOR was increased when b3galt5 was overexpressed. Interestingly, benzy-α-GalNAc, an inhibitor of O-glycosylation [38], also significantly downregulated the phosphorylated mTOR expression. Based on these results, we hypothesized that O-linked glycosylation of mTOR influenced its activation and stability. To date, little is known about the glycosylation and functions of mTOR. Only a few articles have confirmed that O-GlcNAcylation of mTOR activated the mTOR [39]. O-GlcNAcylation is the covalent linkage of GlcNAc glycans to serine and threonine residues of protein substrates, which is catalyzed by O-linked N-acetylglucosamine transferase (OGT) [40]. Thus, our study is the first attempt to explore b3galt5-mediated mTOR O-glycosylation and its effect on mTOR activation. More evidence is needed to further confirm how b3galt5 modifies the O-glycosylation of mTOR. Further experiments like isotope labeled glycan and glycoprotein spectrum analysis may help to identify the specific chains of mTOR. O-linked glycosylation was reported to influence the function of protein, for instances, intracellular O-linked glycosylation directly regulates cardiomyocyte L-type Ca2 + channel activity [41]. O-linked glycosylation on CaMKII enhances CaMKII activity autonomously [42], O-linked glycosylation of A20 also enhances the inhibition of NF-κb activation [43]. Moreover, how O-glycosylation activates mTOR needs to be further explored.
Emerging evidence suggests that glycosylation plays an important role in cancer stemness. Cancer stem cell markers are mainly glycoproteins, such as CD44, CD133, and CD24 [44]. α-1,3-Mannosyltransferase, which is involved in early N-glycan synthesis, promotes breast cancer stemness by inducing glycosylation of TGF-β receptor II [45]. β1,4-N-Acetylgalactosaminyltransferase III regulates stemness by inducing the expression of the colon cancer stem cell markers OCT4 and NANOG [46]. B3galt5 has been reported to promote breast cancer progression by enhancing cancer stemness [9]. In this study, we found that the regulatory role of b3galt5 in HCC involved another mechanism, that is, promoting mTOR glycosylation to enhance glycolysis. It should be noted that we cannot exclude whether b3galt5 may also contribute to the regulation of liver cancer stemness. Studies have indicated that metabolic reprogramming can regulate cancer cell stemness [47]. For instance, the glycolysis gatekeeper PDK1 reprogrammed breast cancer cell stemness under hypoxic conditions [48]. In addition, the mTOR pathway was reported to be closely related to the maintenance of cancer stem cells [49,50]. Inhibition of the mTOR pathway in breast cancer stem-like cells significantly suppressed not only colony-formation ability in vitro but also tumorigenicity in vivo [51]. Positive expression of p-mTOR was positively associated with positive expression of the liver cancer stem cell markers CD133, CD90, and EpCAM [52]. In addition, mTOR direct down-stream p70s6k also regulates the stemness and metastatic properties of ovarian cancer cells, knocking down p70s6k leading to a marked reduction in cancer stem cell proliferation and expansion [53]. P19 stem cells pluripotency was also relied on PI3K-AKT-mTOR-p70s6k pathway activation [54]. Therefore, it is necessary to further illustrate the role of b3galt5 in liver cancer stemness.

Conclusions
Our findings unveiled a novel function of b3galt5 in regulating glycolysis in hepatocellular carcinoma. B3galt5 promoted O-linked glycosylation modification of mTOR and thus activated its downstream signaling, resulting in metabolic reprogramming from OXPHOS to glycolysis (Fig. 8).
This study provides a new biomarker and potential target for the diagnosis and treatment of liver cancer. Fig. 8 Schematic diagram of the b3galt5-mediated effects on hepatocellular carcinoma. Elevated expression of b3galt5 in HCC cells promoted mTOR glycosylation, which contributed to enhanced aerobic glycolysis. Activation of mTOR promotes HCC cell proliferation and survival by diverting glucose metabolism from OXPHOS to glycolysis. On the other hand, depletion of b3galt5 eliminated glycosylation to inhibit mTOR activation, which promoted glucose metabolism into OXPHOS, thus inhibiting HCC cell survival