HDAC11 Regulates Glycolysis through the LKB1/AMPK Signaling Pathway to Maintain Hepatocellular Carcinoma Stemness

Hepatocellular carcinoma (HCC) represents 85% to 90% of all primary liver cancers, which is the fourth leading cause of cancer-related deaths worldwide (1, 2). Despite the increasing awareness of the disease's biology, a significant percentage of its drivers are nevertheless undruggable. Besides, multikinase inhibitors, such as sorafenib and lenvatinib, are currently available for first-line systemic treatment. Nevertheless, they extend survival by a few months before resistance arises. Although promising outcomes have been seen with several immunotherapies, such as nivolumab, their rate of responsiveness rate is less than 20% (3). In this case, new prognostic biomarkers or therapeutic strategies to optimize the efficacy of sorafenib are urgently required.

Histone deacetylase (HDAC) enzymes mediate various biological functions such as metabolism, inflammation, and neurologic development via regulating gene expression, making them attractive targets for intervention in a set of diseases containing certain hematologic cancers (4–6). Apart from their success in lymphoma and multiple myeloma (4), the efficacies of inhibitory compounds observed against solid tumors have been disappointing, possibly owing to the lack of specificity (7). HDAC11, the newest member of the 11 human zinc-dependent HDACs, is also the smallest protein in this family and has the least characterized biological function. Over 80% of its 347 amino acid sequence belongs to the classical deacetylase domain (8). Its unique structure, function, and phylogeny have classified into a distinct class of HDACs defined as class IV (9). HDAC11 has been implicated in diverse immune functions (10, 11), myoblast differentiation (12), metabolism, and obesity (13). The ONCOMINE platform discloses overexpression of HDAC11 in several carcinomas as compared with corresponding healthy tissues. Its depletion is sufficient to promote cell death and inhibit metabolic activity in HCT-116 colon, MCF7 breast, PC-3 prostate, and SK-OV-3 ovarian cancer cell lines (14). A role for HDAC11 in the tumors is further supported by Hodgkin lymphoma results regarding suppression of OX40L expression (15). In myeloproliferative neoplasms (MPN), HDAC11 deficiency improved splenic architecture and reduced fibrosis. It increased survival in the MPLW515L-MPN mouse model during primary and secondary transplantation (16). In terms of solid tumors, it has been reported to interfere with p53 expression in pituitary tumor cells (17), to accelerate cell-cycle progression to boost survival in neuroblastoma cells (18), and to inhibit breast cancer metastasis from lymph nodes (19). Studies have found that inhibiting HDAC11 promotes cell apoptosis in liver cancer (20). Here, we screened from four GSE databases and verified the importance of HDAC11 in HCC progression and malignancy. RNA-sequencing analysis identified differentially expressed genes (DEG) in Hdac11^−/− mice, indicating that HDAC11 was essential for maintaining the stemness of HCC.

Glycolysis is closely related to cancer stem cells (CSC). CSCs need glycolysis and lipid metabolism to provide energy and preferentially use glycolysis to maintain homeostasis (21–24). Recent studies have shown that CD44ICD maintains breast cancer stemness by regulating the glycolysis process (25). Besides, various studies have shown that glycolysis was upregulated in liver CSCs and HBx-expressing HCC cells, and the intervention of glycolysis weakens the cancer stem-like phenotype (26). Consistently, our study also displayed the implication of HDAC11 in regulating glycolysis via energy metabolomics analysis. Further analysis of DEGs by RNA-seq and chromatin immunoprecipitation (ChIP)-sequencing (ChIP-seq) revealed that HDAC11 inhibits the AMP-activated protein kinase (AMPK) signaling pathway by regulating the expression of LKB1 to promote the glycolysis process, thereby maintaining CSCs and sorafenib resistance.

The standard liver cell line MIHA and liver cancer cell lines PLC/PRF/5, BEL-7404, HepG2, Huh7, MHCC-97L, MHCC-97H, BEL-7402, H22, and HEK 293 Phoenix ampho packaging cells were purchased from Cell Bank of Type Culture Collection of Chinese Academy of Sciences, Shanghai Institute of Cell Biology, Chinese Academy of Sciences. Liver cancer cell lines were routinely cultured as described previously (27). Cell lines were maintained at 37C in an atmosphere containing 5% CO₂ in DMEM or RPMI1640 (Gibco) supplemented with 10% FBS. Cells were passed every 1–2 days to maintain logarithmic growth. All cell lines used were cultured within 35 generations and regularly tested for Mycoplasma contamination by the Plasmo TestTM Kit (InvivoGen, rep-pt1). The short tandem repeat (STR) analysis method was used to verify the identity of cell lines twice a year at the core institution.

HCC tumor and paracancerous tissues, which were used for qRT-PCR and Western blot analysis, were randomly collected from patients with HCC who underwent curative resection with informed consent between 2010 and 2014 at the Department of Hepatobiliary Surgery, Shandong University, Affiliated Qilu Hospital. All tissues were collected immediately upon resection of the tumors in the operation theatre, transported in liquid nitrogen, and then stored at −80°C. Another 364 samples of HCC tissue (Supplementary Table S1), which were used forIHC analysis, were randomly collected from patients with HCC who underwent curative resection with informed consent between 2006 and 2010 at the Department of Hepatobiliary Surgery, Shandong University, Affiliated Qilu Hospital. Tumor staging was based on the 6th edition of the Tumor-Node-Metastasis (TNM) classification of the International Union Against Cancer. Follow-up data were summarized at the end of March 2017, with a median observation time of 73.7 months. The research protocol complied with the ethical standards of the Declaration of Helsinki and was approved by the Ethics Committee of Shandong University Hospital, and the patient's written informed consent was obtained according to the Declaration of Helsinki.

For tumorigenesis assays, a group of six Balb/c nude mice was injected subcutaneously with infected cells into the left and right flanks. The tumor size was assessed by measuring tumor dimensions with calipers for up to 42 days. For metastasis assays, cells were resuspended in PBS, and the cell suspension was injected into the tail veins of nude mice. All animals were maintained under the guidelines of Shandong University and the evaluation and approval of the Institutional Animal Care and Use Committee (Shandong University, Jinan, China). Food and water were provided ad libitum.

Conditional knockout (KO) mice (Hdac11^loxP/loxP) were obtained from the Cyagen Biotechnology Co. Ltd. According to guidelines, all mice were maintained in filter topped cages on autoclaved food and water at Shandong University. A single dose of diethylnitrosamine (DEN; Sigma-Aldrich) at 100 mg/kg body weight was injected intraperitoneally into 4-week-old male mice, which were fed with phenobarbital in a 0.07% diet from the age of 6 weeks until sacrifice. These mice were killed 9 months after treatment with DEN. The development of HCC was compared among offspring from the same breeding.

The tumors and lungs were dissected from mice and fixed in 4% paraformaldehyde in PBS overnight and subsequently embedded in paraffin wax. Sections were cut at a thickness of 5 μm and stained with hematoxylin and eosin for histologic analysis. IHC staining was performed according to the previous study (27, 28). In brief, 5-μm consecutive sections were cut by standard methods and were stained for HDAC11 and LKB1. All paraffin-embedded sections were dewaxed in xylene and rehydrated in graded ethanol solutions. Antigens were retrieved with citrate buffer (0.01 mol/L citric acid: pH 6.0) for 15 minutes at 100°C in a microwave oven. After being treated with 3% H₂O₂ for 10 minutes to block the endogenous peroxidase, the sections were incubated with 10% FCS for 30 minutes to reduce nonspecific binding. The primary antibodies were applied to the sections at 4°C overnight. Immunoreaction visualization and negative control setup were performed according to the procedure mentioned previously. The proportion of stained cells (lower, <30% staining; higher, ≥30% staining) was semiquantitatively determined following published protocols.

Cells were labeled with EPCAM microBeads (Miltenyi Biotec) and separated on the MACS MS column (Miltenyi Biotec). Briefly, cells were incubated with the beads for 30 minutes at 4°C, and the labeled cells were collected in a magnetic separation column. Then, the column was removed from the magnetic field and washed, collecting EPCAM-positive cells. CD133/90-positive cells were obtained using a cell sorter after antibody incubation. Briefly, cells were incubated with anti-CD133/90-FITC antibody (Miltenyi Biotec) in labeling buffer (PBS pH 7.2, 0.5% BSA, and 2 mmol/L EDTA) in the dark for 10 minutes at 4°C. The cells were then washed and sorted using the MoFlo XDP Cell Sorter (Beckman Coulter).

For holoclone assays, we plated cells at a clonal density (i.e., 100 cells/well) in a 6-well dish, counted the number of holoclones several days later, and presented the percentage of cells that established a holoclone as cloning efficiency. For clonogenic assays, we plated cells generally at 1,000 cells/well in Matrigel (MG) or methylcellulose (MC) at a 1:1 ratio in 100–200 μL and enumerated colonies 1–2 weeks after plating. Mammosphere culture was performed as described by Dontu and colleagues with slight modifications (29). Single-cell suspensions were plated in ultra-low attachment 96-well plates (Costar) at a different density of viable cells. Cells were grown in a serum-free mammary epithelial growth medium (MEGM) supplemented with 1:50 B27 (Invitrogen), 20 ng/mL EGF, 20 ng/mL essential fibroblast growth factor (bFGF; BD), and 10 μg/mL heparin (Sigma). The numbers of spheroids were counted after 7–10 days. For in vitro propagation, primary spheres were collected, dissociated into single-cell suspensions, and plated in ultra-low attachment 96-well plates. The secondary numbers of spheroids were counted after 14 days of plating.

Protein A (88845, Thermo Fisher Scientific) and G (88847, Thermo Fisher Scientific) dynabeads were mixed at a 1:1 ratio and preincubated with antibodies 3 hours before immunoprecipitation. Cells were cross-linked by 1% formaldehyde for 10 minutes, and the reaction was quenched with 125 mmol/L glycine. The antibodies used for ChIP assays were against H3K9ac (ab4441, Abcam). The beads were washed in RIPA buffer, and elution buffer (0.1 mol/L NaHCO₃, 1% SDS, and proteinase K) was used for reverse cross-linking of DNA-protein complexes at 65°C for 8–16 hours. DNA was purified by phenol-chloroform extraction and subjected to the sequence.

Data were described as the mean ± SD The association between HDAC11 and LKB1 expression in HCC tissue was assessed using Spearman rank correlation test. Comparisons between different groups were performed using Student two-tailed t test. The limit of statistical significance was P < 0.05. Statistical analysis was done with SPSS version 11.0 software (SPSS, Inc.).

For the initial screening of potential biomarkers of HCC, four gene-expression profiles (GSE76427, 64041, 62232, and 39791) were downloaded from the Gene Expression Omnibus (GEO) database, each comprising data of HCC and control samples. The network diagram depicted the interactions among 29 overlapped genes differentially expressed between HCC and the control group, highlighting the core position of HDAC11. Indeed, we found frequent amplification and upregulated expression of HDAC11 in the TCGA and four GEO datasets of HCC compared with their matched normal tissues (Supplementary Fig. S1A; Fig. 1A–C). In addition, analysis of the clinical information of patients with HCC found that HDAC11 was closely related to tumor size, microvascular invasion, tumor differentiation, and clinical staging (Supplementary Table S1).

To further validate these bioinformatical searches, we performed IHC and immunoblotting methods to analyze HDAC11 expression in a set of 364 human liver cancers and their corresponding adjacent tissues. As expected, tumor tissues showed an apparent rise in HDAC11 protein levels (Fig. 1D; Supplementary Figs. S1B–S1D and S2A and S2B). The same as in HCC cell lines (PLCPRF5, Huh7, MHCC-97L, HepG2, MHCC-97H, BEL-7402, and BEL-7404), HDAC11 expression was markedly higher than that in the regular liver cell lines (MIHA; Supplementary Fig. S2C–S2E). Moreover, high levels of HDAC11 predicted poor clinical outcomes (Fig. 1E); this was also observed in the other three types of cancers (endometrial carcinoma, oesophageal carcinoma, and thymoma; Supplementary Fig. S3A–S3I). Altogether, the results mentioned above that were derived from human datasets and tissues strongly proved that HDAC11 might act as an oncogenic driver in HCC.

To explore the role of HDAC11 in tumorigenesis in vivo, we focused on a DEN/phenobarbital-induced (DEN/PB-induced) mouse model of HCC (30). Consistent with human data, HDAC11 protein levels were dramatically higher in the liver tissues from DEN-induced mice compared with those in controls (Supplementary Fig. S4A and S4B). Meanwhile, oncogenic HDAC5 increased, whereas tumor suppressor HDAC6 reduced in the same context (31, 32).

HDAC11 expression was effectively silenced in the KO mice, as shown in Fig. 2A and Supplementary Fig. S5A and S5B. Fully developed liver cancers were developed in mice within 9 months after treatment with DEN. Hdac11^−/− mice exhibited alleviated hepatocarcinogenesis with decreased tumor masses, tumor sizes, tumor numbers, and Ki67-positive cells as markers for proliferation (Fig. 2B–D). In addition, improved survival rates were noted in HDAC11^−/− mice rather than HDAC11^WT mice (Fig. 2E). Altogether, the above observations indicated that HDAC11 might facilitate HCC progression by stimulating cell proliferation. The results also highlighted that HDAC11 was necessary for efficient DEN-induced mice liver cancer.

As a first step toward uncovering the underlying mechanism of HCC development triggered by HDAC11, we performed Affymetrix HU U133 plus 2.0 arrays with liver cells from HDAC11 KO mice. A total of 977 genes were found to be positively associated with HDAC11 deletion, while 653 genes were negatively correlated with HDAC11 deletion (Fig. 2F). Gene Ontology (GO) and GSEA analyses revealed the enriched signature related to endogenous HDAC11-dependent transcription in HCC, namely, CSCs (Fig. 2G and H). Collectively, these findings suggest a potential regulatory role for HDAC11 in the maintenance of cancer stemness.

Studies have shown that liver CSCs detected in liver cancer are characterized by expressing the following cell surface markers: EPCAM (33, 34), CD133 (35), and CD90 (36). To assess this possibility, we first investigated the expression of HDAC11 and CSC markers (Nanog, Oct4, and Sox2) in the CSC-enriched populations, including sphere formation cells, side population (SP) cells, EPCAM⁺ cells, CD133⁺ cells, and CD90⁺ cells (Fig. 3A–F; Supplementary Fig. S6A–S6E). Compared with the adherent Huh7 cells, HepG2 and BEL-7404 cells, HDAC11, and CSC markers were highly expressed in the sphere cells. Similar results were observed in the SP and EPCAM⁺ cells, instead of the non-SP and EPCAM⁻ cells. The above findings show the role of HDAC11 in maintaining stem-like populations.

To further assess the contribution of HDAC11 in the retention of cancer stemness, we next generated Huh7 and BEL-7402 cells overexpressing HDAC11 and BEL-7404 MHCC-97H cells knocking down endogenous HDAC11 via lentiviral shRNA (Supplementary Fig. S6F–SI). Hoechst SP assay revealed that forced expression of HDAC11 caused a significant increase of SP cells. In contrast, HDAC11 ablation led to a noticeable reduction of the SP cells (Fig. 3G and H; Supplementary Fig. S7A and S7B). Simultaneously, sphere formation efficiency was evaluated, demonstrating identical results showing that overexpression of HDAC11 stimulated the formation of primary and secondary spheres. At the same time, HDAC11 knockdown inhibited sphere formation (Fig. 3I and J; Supplementary Fig. S7C). We also found that the Nanog levels in the primary and secondary spheres in Fig. 3I were positively correlated with HDAC11 expression (Fig. 3K). Also, treatment with a specific HDAC11 inhibitor (SIS17) had the same effect on the stemness of HCC cells as knocking out HDAC11 (Supplementary Fig. S7D). Meanwhile, we found that overexpression of HDAC11 had a maintenance effect on hepatic stem cells (Supplementary Fig. S7E).

To test whether ectopic expression of HDAC11 could succeed in altering the tumor-initiating frequency, we injected transformed Huh7 cells with serial dilutions into nude mice. More than half of the mice injected with 10² HDAC11-expressing Huh7 cells formed tumors (5 of 10 injected hosts), whereas no tumors arose when 10² cells expressing a control vector (pBabe) were injected into the mice. Correspondingly, at least 10⁴ of the control cells were required to initiate tumor formation. Then, tumor formation was extremely inefficient (3 of 10; Fig. 3L; Supplementary Fig. S8A). Thus, excessive expression of HDAC11 considerably enhanced (by approximately two orders of magnitude) the number of tumor-initiating cells.

Meanwhile, tumor-forming ability was also measured in the HDAC11-deficient cells compared with its corresponding control. Only 1 of 10 mice formed tumors with the injection of 10⁴ HDAC11-deficient BEL-7404 cells, while tumor formation efficiency (6 of 10 mice) was notably higher than that of controls (Supplementary Fig. S8B–S8D). The above results illustrate that HDAC11 is necessary to prime and sustain the stemness of HCC cells.

As stemness is a major cause of tumor metastasis and recurrence, Huh7 and BEL-7402 cells with excessive expression of HDAC11 and BEL-7404 and MHCC-97H cells with HDAC11 depletion were subjected to assessments of cell migration and invasion (Supplementary Fig. S9A and S9B). Overexpression of HDAC11 significantly increased cell invasion and migration in Huh7 and BEL-7402 cells, whereas knocking out HDAC11 caused the opposite effect. Overexpression of HDAC11 significantly increased the ability of metastasis in malignant lung cells (Supplementary Fig. S9C and S9D).

Given the implication of glucose metabolism in maintaining cancer stemness (37), we next investigated whether HDAC11 maintains cancer stemness via regulating glycolysis. Targeted metabolomics was thus used to detect glucose metabolism pathways in Hdac11^WT and Hdac11^−/− primary liver cells, showing that, compared with Hdac11^WT, glucose metabolites in Hdac11^−/− (lactate, citrate, pyruvate, etc.) primary liver cells were significantly reduced (Fig. 4A). To verify the mass spectrometry results, 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-d-glucose (2NBDG) levels were detected in Hdac11^wt, and Hdac11^−/− primary liver cells, as well as the HDAC11 knockdown cells, and the results showed that HDAC11 deficiency significantly inhibited glucose uptake (Fig. 4B). Contrarily, HDAC11 overexpression crucially increased glucose uptake in Huh7 and BEL-7402 cells (Fig. 4B). Measurements of ATP levels and lactate production showed that knocking out HDAC11 in primary hepatocytes and HCC cells significantly inhibited both ATP levels and lactate production (Fig. 4C and D). Simultaneously, overexpression of HDAC11 significantly increased ATP levels and the lactic acid production of HCC cells (Fig. 4C and D). Further analysis of the effect of HDAC11 on Glut1 expression levels showed that HDAC11 deficiency in primary hepatocytes and HCC cells significantly inhibited Glut1 expression levels (Fig. 4E and F), while HDAC11 overexpression significantly increased Glut1 expression levels (Fig. 4E and F). We also noted a dose-dependent effect of glucose on the growth ability of Huh7 cells overexpressing HDAC11 instead of control cells (Fig. 4G). Besides, treatment with 2-deoxy-d-glucose (2-DG), which competitively inhibits glycolysis, reversed the effect of HDAC11 on HCC cell growth (Fig. 4A). These results collectively demonstrated that HDAC11 might regulate glucose metabolism to promote the growth of HCC cells.

To further explore the impact of HDAC-elicited glycolysis on HCC stemness, we also used 2-DG to treat transformed Huh7 and BEL-7404 cells. The results showed that treatment with 2-DG reversed the promoting effect of HDAC11 on the formation of primary and secondary spheres (Supplementary Fig. S10A). Also, the treatment with 2-DG abolished the effect of HDAC11 on the SP grouping (Supplementary Fig. S10B). Altogether, these results indicated that HDAC11 could regulate glycolysis to affect the stem-like features of HCC cells.

To determine the regulatory mechanisms of HDAC11 in liver cancer, we first figured out the potential target of HDAC11 as acetyl-histone H3 at lys9 (H3K9ac), instead of other acetyl-histones, by Western blotting (Fig. 5A). Next, we assessed genome-wide H3K9ac binding by ChIP-seq analysis in liver cells from Hdac11^WT and Hdac11^−/− mice. As expected, analysis of enriched loci (peaks) in Hdac11^WT and Hdac11^−/− cells indicated that HDAC11 diminished H3K9ac binding in the liver (Fig. 5B and C). Analysis of the DEGs by RNA-seq and ChIP-seq of H3K9ac revealed 17 common DEGs (Fig. 5D). Among them, LKB1 was selected according to fold change. Also, we found that H3K9ac in the LKB1 promoter region was significantly increased in HDAC^−/− cells (Fig. 5E), suggesting histone acetylation is associated with activation of this promoter. To figure out whether this was the case, three possible binding sites for H3K9ac were then predicted (Fig. 5F). Transformed BEL-7404, Huh7, MHCC-97H, and BEL-7402 cells as described above were chromatin immunoprecipitated for IgG, RNA polymerase II, and H3K9ac, followed by qPCR, which validated the ChIP-seq results (Fig. 5G and H; Supplementary Fig. S11A and S11B) and demonstrated high enrichment for H3K9ac-bound sequences at the #1 and #3 sites rather than the #2 site. Furthermore, ectopic expression of mutated HDAC 11 (H142/143A) ultimately rescued the decreased binding of H3K9ac regulated by the HDAC11 wild-type (Fig. 5I; Supplementary Fig. S11A).

These results suggest that HDAC11 probably regulates LKB1 transcription by mediating H3K9 acetylation in the promoter region of LKB1. To further clarify the effect of HDAC on the expression level of LKB1, we analyzed the DEGs in liver cells from Hdac11^WT and Hdac11^−/− mice by RNA-seq, which showed that the expression level of LKB1 in Hdac^−/− primary liver cells was significantly increased with the highest fold change (Supplementary Fig. S12A and S12B). Consistently, the level of LKB1 protein in the primary liver cells of Hdac^−/− mice also increased significantly (Supplementary Fig. S12C and S12D). Besides, HDAC and LKB1 protein expression levels were negatively correlated in HCC tissues (Supplementary Fig. S12E). HDAC11 overexpression significantly inhibited LKB1 expression levels in Huh7 cells, and ectopic expression of mutated HDAC11 (H142/143A) reversed the inhibitory effect of HDAC on LKB1 expression levels (Fig. 5J). In summary, HDAC negatively regulated the expression level of LKB1 by inhibiting H3K9ac in the LKB1 promoter.

To prove the involvement of LKB1 in HDAC11-induced glycolysis and malignancy, we established two stable HCC cell lines (BEL-7404 and MHCC-97H) expressing two independent shRNAs targeting HDAC11 and LKB1. As anticipated, knockdown of LKB1 reversed the inhibitory effect of HDAC11 deficiency on glucose uptake, ATP levels, and lactate production in BEL-7404 and MHCC-97H cells (Fig. 6A–D). These results showed that HDAC11 regulated cell glycolysis the efficiency of through LKB1.

To further explore the effect of LKB1 on the malignant behavior caused by HDAC11 in HCC, cell viability and colony formation assays were performed with the HDAC11/LKB1 double knockdown BEL-7404 and MHCC-97H cells. HDAC11 knockdown significantly inhibited cell activity and cloning formation ability, whereas LKB1 reduction reversed the inhibitory effect of HDAC11 knockdown on cell viability and colony-forming ability (Fig. 6E and F). Besides, nude mice were injected subcutaneously with BEL-7404 cells that knocked out HDAC11 and LKB1 alone or together. The KO of HDAC11 and LKB1 together reversed the inhibitory effect on heterologous tumor growth caused by the KO of HDAC11 alone (Fig. 6G). Similar results were also observed in the weight of heterogeneous tumors (Fig. 6H and I). We also found that overexpression of HDAC11 and LKB1 in Huh7 cells significantly inhibited cell viability, colony-forming ability, and SP cells, as well as reversed the malignant features caused by overexpression of HDAC11 alone (Supplementary Fig. S13A–S13C). These results indicate that LKB1 mediates the HDAC11-regulated malignant behavior of HCC cells.

LKB1 serves as an upstream kinase for AMPK under energy stress (38). LKB1 activates AMPK by phosphorylating 172 threonines of AMPK, and activated AMPK further phosphorylates downstream pathways, thereby affecting both fatty acid synthesis and glycolysis protein synthesis (39). Therefore, we explored whether HDAC11 affected glycolysis and malignant behavior through the LKB1/AMPK signaling pathway. In Huh7 and BEL-7402 cells overexpressing HDAC11, we found that the protein levels of p-AMPK and LKB1 were significantly reduced. However, overexpression of an HDAC11 mutant (H142/143A) reversed the decreased p-AMPK and LKB1 levels caused by HDAC11 overexpression (Supplementary Fig. S14A and S14B). Also, in HDAC11-deficient BEL-7404 and MHCC-97H cells, the protein levels of LKB1, p-AMPK, and downstream p-ACC were significantly increased. However, by knocking down HDAC11 and LKB1 together, the increased p-AMPK and p-ACC levels were offset (Supplementary Fig. S14C and S14D). We also found that overexpression of HDAC11 combined with AMPK knockdown significantly increased cell proliferation and glycolysis in Huh7 cells, which significantly improved the effect of HDAC11 overexpression alone on cell proliferation and glycolysis (Supplementary Fig. S14E and S14F). To further verify that HDAC11 affects the malignant behavior and glycolysis of HCC through the LKB1/AMPK signaling pathway, we used the AMPK activator metformin (Met) to treat Huh7 cells with or without HDAC11 overexpression. The results showed that the treatment with Met resisted the inhibitory effects of HDAC11 overexpression on p-AMPK and p-ACC expression levels (Supplementary Fig. S15A). Furthermore, we examined the effect of Met on HDAC11-mediated glycolysis. The results showed that treatment with Met reversed the HDAC11-induced increase of glucose uptake, ATP, and lactate levels in Huh7 cells (Supplementary Fig. S15B–S15D). We also measured the oxygen consumption rate (OCR) of Huh7 cells overexpressing HDAC11 in response to treatment with Met, which showed that overexpression of HDAC11 significantly inhibited the OCR of Huh7 cells. Treatment with Met significantly increased OCR in Huh7 cells and eliminated the inhibitory effect of HDAC11 overexpression on OCR (Supplementary Fig. S15E). These results suggest that the LKB1/AMPK signaling pathway mediates the positive regulation of HDAC11 on glycolysis.

To further explore the effect of the LKB1/AMPK signaling pathway on the HDAC11-mediated malignant behavior of HCC, MTT, colony formation, and sphere formation assays were performed. These data showed that overexpression of HDAC11 significantly increased cell viability and colony-forming ability. However, the activation of AMPK by Met counteracted the promoting effect of HDAC11 overexpression on cell viability, colony formation ability, and cancer stemness (Fig. 7A–C). The results of subcutaneous tumor formation in nude mice also showed that overexpression of HDAC11 significantly increased the growth and proliferation ability of heterogeneous tumors. The activation of AMPK by Met offsets the growth=promoting effects of HDAC11 on heterogeneous tumors (Fig. 7D–F). These results suggest that the LKB1/AMPK signaling pathway mediates the effect of HDAC11 on the malignant behavior of HCC.

Of note, low levels of HDAC11 were associated with an increase in the overall survival of patients treated with sorafenib therapy compared with those without sorafenib treatment after surgery. Simultaneously, there is no such distinction in patients with high HDAC11 expression (Supplementary Fig. S16A and S16B). We then investigated the cellular basis for this effect. We found the consistent trend that the overexpression of HDAC11 undermined the sensitivity of HCC cells to sorafenib, and HDAC11 reduction increased the sensitivity of HCC cells to sorafenib (Supplementary Fig. S16C and S16D). We also found that overexpression of HDAC11 and LKB1 in Huh7 cells significantly increased the sensitivity of cells to sorafenib and reversed the malignant features caused by HDAC11 overexpression (Supplementary Fig. S16E). These results indicated that HDAC11 played an active role in protecting HCC cells from sorafenib-elicited cytotoxicity. Similar results were obtained in the xenograft mice model (Supplementary Fig. S16F–S16H). Accordingly, our data revealed that the expression of HDAC11 correlated with the HCC clinical outcomes. The response to sorafenib was impaired in patients with high levels of HDAC11, suggesting that HDAC11 may be considered as a new predictive biomarker for the response to treatment with sorafenib.

The results of this study certify the previously unknown role of HDAC11 as a positive regulator of glycolysis in HCC via control of LKB1 transcription, which correlates with cancer stemness and response to sorafenib (Fig. 7G). We exemplify that HDAC11 expression is upregulated in human HCC tissues and cell lines and predicts poor prognosis, implying the tumorigenic role of this protein. This was further supported by the subsequent in vivo experiments on KO mice. Thus, our finding, for the first time, proves HDAC11 to be one of the oncogenic lysine deacetylases aberrantly expressed in HCC.

Glycolysis provides preferential energy to tumor stem cells and the homeostasis of the cells. Besides, the rate of glycolysis was found to increase significantly in tumor stem cells (37). The level of genes related to glycolysis also increased significantly, while inhibiting the process of glycolysis reduced the number of cancer cells with stem properties (40). Cancer stemness, which is closely related to glycolysis (41), is a pivotal event in the progression, recurrence, and chemoresistance of HCC (42). Beyond chromatin remodeling via histone acetylation, HDAC isoforms (HDAC1–9) also prompt key signaling pathways pertinent to the maintenance of CSCs, especially a bundle of EMT-inducing transcription factors, including HIF-1, Notch1, Stat3, β-catenin, c-Jun, and NF-κB, each of which possesses a critical function in regulating CSCs (43). However, whether HDAC11 can maintain the stemness of liver cancer cells through glycolysis is unclear. Our work demonstrates enrichment of a novel isoform, HDAC11, in stem-like HCC populations and attests to its contribution in retaining stemness in liver cancer via regulating glycolysis. This may explain at least one piece of the jigsaw puzzle regarding the function of HDAC11 in solid tumors.

The serine/threonine kinase LKB1 is ubiquitously expressed and highly conserved throughout the evolutionary process. LKB1 acts as a kinase and may activate various downstream kinases, such as TGFβ signaling or the mTOR pathway (44). LKB1 is generally considered a tumor suppressor, and functional mutations or deletions in its genes caused an increased risk of tumor development (44–46). It has been reported that LKB1 was highly expressed in late HCC (47). Moreover, in transplanted tumors of mice with severe combined immunodeficiency (SCID), inhibition of LKB1 reduced tumor growth (48). However, other studies have shown that LKB1 is lowly expressed in patients with HCC and closely related to poor prognosis (49). Therefore, the exact role of LKB1 in HCC is still controversial. Here, we screened and verified LKB1 as the critical target of HDAC11 in HCC. Relevant mechanism and function experiments showed that HDAC11 increased the stemness and the proliferative ability of HCCs by inhibiting the level of H3K9ac in the LKB1 promoter region and suppressing the expression level of LKB1. These findings elucidated the role of LKB1 in HCC tumorigenesis and clarified its upstream regulatory mechanism. LKB1 is an upstream activator of AMPK, which is involved in phosphorylation to activate AMPK (50). The lack of LKB1 is usually accompanied by AMPK inactivation and cell growth disorders, and activated LKB1 inhibits the proliferation of many different tumor cells (51, 52). Besides, LKB1 deficiency causes the failure of AMPK activation and stimulates the rate of glycolysis (53). Our work found that LKB1/AMPK is essential for HDAC11to regulate glycolysis, which affects HCC stemness.

Studies have shown that HDAC11 inhibits Sox2 expression to regulate lung adenocarcinoma stem cell renewal and overcome drug resistance (54). Our clinical investigation revealed that patients with HCC with low HDAC11 expression exhibited a superior response to treatment with sorafenib following surgical resection. However, patients with high HDAC11 levels showed no significant response. This observation matches the facilitative role of HDAC11 in liver CSC expansion and glycolysis, each of which is known as a factor affecting the efficacy of sorafenib. Emerging research discloses the enrichment of CSCs after treatment with sorafenib. Long-term exposure to sorafenib shifted HCC cells toward a more mesenchymal phenotype and consequently lower sensitization to sorafenib (55, 56). Evidently, in response to sorafenib, HCC engenders stem-like properties to escape the cytotoxicity elicited by the drug. Combined treatment with drugs targeting CSCs might be a promising strategy to overcome resistance to sorafenib. An HDAC11 inhibitor, in this way, could be an ideal option considering the critical function of this protein in maintaining stem-like and mesenchymal traits.

CSCs promote tumor resistance in many different ways. CSCs positively express the ATP-binding cassette (ABC) transporter in various cancers. Many active ABC transporters quickly efflux chemotherapeutic drugs, contributing to multidrug resistance (57, 58). Also, the SP of CSCs is a subset of stem cells and has high drug efflux ability (59). CSCs can withstand high replication pressure levels, resist DNA damaging agents, and affect various DNA damage response. Conventional DNA damage pathways and novel mediators have been shown to induce CSC-mediated treatment resistance in various cancers (60). The slow circulation or fluctuating circulation characteristics of tumor cells and miRNAs regulate essential functions by directly controlling cell death mechanisms, leading to resistance to treatment (61). Our study found that the expression level of HDAC11 in HCC CSCs is very high, and ectopic overexpression significantly increases the activity of HCC CSCs. We have reason to speculate that the resistance of HCC to sorafenib caused by HDAC11 may be due to the mechanism by which HDAC11 confers characteristics to HCC stem cells. We will also explore whether there is a regulatory relationship between HDAC11 and multiple tumor resistance mechanisms (ABC transporter, ALDH, miRNA, etc.).

Altogether, by identifying the suppression of LKB1/AMPK downstream of HDAC11 to promote glycolysis in HCC and defining its regulation and contribution to clinical outcomes, we provide a framework for testing nascent therapeutic approaches targeting HDAC11 in this disease. Further efforts are still needed to elucidate the interactions between HDAC family members, for instance, HDAC6 (62), and explore selective chemical probes targeting HDAC11 to translate it into the clinic.

No disclosures were reported.

L. Bi: Conceptualization, resources, data curation. Y. Ren: Resources, data curation. M. Feng: resources, data curation. P. Meng: Resources, data curation. Q. Wang: Resources, data curation. W. Chen: Resources, data curation. Q. Jiao: Resources, data curation. Y. Wang: Resources, data curation. L. Du: Resources, data curation. F. Zhou: Resources, data curation. Y. Jiang: Resources, data curation. F. Chen: Resources, data curation. C. Wang: Resources, data curation. B. Tang: Resources, data curation, software, funding acquisition. Y. Wang: Resources, data curation, software, funding acquisition.

This research was supported in part by The National Natural Science Foundation of China (nos. 81874040, 81873219, 81774180, 81560393), Guangxi Science Fund for Distinguished Young Scholars Program (2016GXNSFFA380003), Natural Science Foundation of Guangxi (2015jjDA40010), Natural Science Foundation of Jiangsu Province (BK20181425), Young Elite Scientists Sponsorship Program by CACM (CACM-2018-QNRC2-B07), A Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), Shandong Key Research and Development Program (2018YFJH0505, 2019GSF108218, 2019GSF108139), Taishan Scholar Program of Shandong Province, Natural Science Foundation of Guangxi (ZR2020MH007).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.