The protocol for the animal experiments was reviewed and approved by the Ethics Committee of Medical Research, Nanjing Drum Tower Hospital, Affiliated Hospital of Nanjing University Medical School (Nanjing, China). Donors provided consent for any use of human samples for research and the study protocol was approved by the Ethics Committee of Medical Research, Nanjing Drum Tower Hospital, Affiliated Hospital of Nanjing University Medical School.

The differentially expressed genes (DEGs) in CCA were analyzed using data from The Cancer Genome Atlas (TCGA). CCA RNA-Seq data were downloaded from the TCGA database using the GDC Data Portal (https://gdc-portal.nci.nih.gov/), which consisted of 9 normal samples and 9 paired CCA samples. KEGG pathway enrichment analysis was performed to detect the potential biological functions and pathways of these genes in CCA. The heat map was drawn using the gplots package in Bioconductor software (Bioconductor, org., version 3.6).

We purchased the tissue microarray slides from Shanghai Outdo Biotech Co., Ltd. (Shanghai, China). Staining intensity was graded as follows: absence of staining, 0; weak, 1; moderate, 2 and strong, 3. The scoring approach in the assessment of staining was as follows: 0 (no positive cells), 1 (<25% positive cells), 2 (25–50% positive cells), 3 (>50-75% positive cells), and 4 (>75% positive cells). The score for each tissue was calculated by multiplication of these two grades, and the range of this calculation was 0–12 (23).

HuCCT1 (JCRB, Osaka, Japan) and RBE (The Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, China) cells were cultured in RPMI-1640 medium (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) containing 10% fetal bovine serum (Biological Industries, Cromwell, CT, USA), penicillin (100 U/ml; Invitrogen; Thermo Fisher Scientific) and streptomycin (100 U/ml; Invitrogen; Thermo Fisher Scientific, Inc.). Cells were all maintained at 37°C with 5% CO₂. Metformin was purchased from Sangon Biotech Co., Ltd. (Shanghai, China) and BG45 was purchased from MedChem Express (Monmouth Junction, NJ, USA).

HuCCT1 and RBE cells were transfected using Lipofectamine RNAiMax reagent (Invitrogen; Thermo Fisher Scientific, Inc.), following the manufacturer's protocol. HDAC3 siRNA was produced as described previously (20). Briefly, 50 pmol siRNA and 0.5 ml Opti-MEM I Medium (Invitrogen; Thermo Fisher Scientific, Inc.) were mixed, and then 5.5 µl RNAiMax reagent was added to each well of a 6-well plate and incubated for 15 min. Next, the mixed reagent was added to a cell suspension containing 25×10⁴ cells in 1.5 ml RPMI-1640 with 10% FBS.

Cells were lysed with ice-cold RIPA buffer (50 mM Tris-HCl at pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 1% NP-40 and 1 mM EDTA), mixed with a protease and phosphatase inhibitor cocktail (Roche Diagnostics GmbH, Mannheim, Germany) and phenylmethylsulfonyl fluoride (PMSF) (Biosharp, Hefei, China) for 15 min on ice. Extracted proteins were supplemented with loading buffer containing 5% 2-mercaptoethanol and then denatured at 100°C for 10 min. The protein samples were separated by 8–12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene fluoride (PVDF) membranes (EMD Millipore, Billerica, MA, USA). The membranes were blocked in Tris-buffered saline containing 0.1% Tween-20 with 5% non-fat milk for 2 h at room temperature. Subsequently, membranes were incubated with specific primary antibodies overnight at 4°C. All the primary antibodies were diluted in Tris-buffered saline containing 0.1% Tween-20 with 5% bovine serum albumin (BSA; Biosharp, Hefei, China). The dilution ratio for β-actin and GAPDH was 1:5,000, and all the other antibodies were in 1:1,000. Next, the membranes were treated with the appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies (1:3,000 dilution). The blots were visualized with ECL western blotting reagents (EMD Millipore), following the manufacturer's instructions. HRP-conjugated anti-mouse IgG (7076; Cell Signaling Technology, Inc.) and anti-rabbit antibodies (7074; Cell Signaling Technology, Inc.) were purchased from Cell Signaling Technology, Inc. (Danvers, MA, USA). The following antibodies were used: HDAC3 (ab32369; Abcam, Cambridge, UK), cleaved caspase 3 (9664s; Cell Signaling Technology, Inc.), cleaved PARP (5625s; Cell Signaling Technology, Inc.), PKM2 (4053s; Cell Signaling Technology, Inc.), PDHA1 (ab92696; Abcam), LDHA (3582s; Cell Signaling Technology, Inc.), GLUT1 (12939; Cell Signaling Technology, Inc.), GAPDH (ab128915; Abcam) and β-actin (A5441; Sigma-Aldrich; Merck KGaA, Darmstadt, Germany).

Cellular apoptosis was detected with an Annexin V-fluorescein isothiocyanate (FITC) apoptosis detection kit (556547; BD Biosciences, Franklin Lakes, NJ, USA), following the manufacturer's instructions. Briefly, HuCCT1 and RBE cells were seeded in 6-well plates for 24 h. Then cells were cultured in normal medium or treated with 10 mM metformin, 10 µM BG45, or the combination of both drugs. Next, the control and treated cells were collected and resuspended in 100 µl Annexin V binding buffer, and incubated with 5 µl FITC-conjugated Annexin V and 5 µl propidium iodide (PI) for 15 min in the dark. Annexin V binding buffer (400 µl) was then added to each tube. Samples were analyzed using a BD FACSCanto II flow cytometer (BD Biosciences) with CellQuest (BD Biosciences) version 3.3 software within 1 h.

Cell viability was measured using a Cell Counting Kit-8 assay (CCK-8; Dojindo Molecular Technologies, Inc., Kumamoto, Japan) in 96-well plates (2×10³ cells/well). Cells were seeded in 96-well plates. At 24 h, drugs were added and cells were cultured for 24, 48 or 72 h at 37°C with 5% CO₂. After treatment, the medium was removed and 100 µl CCK-8 reagent was added to each well, according to the manufacturer's instructions. Then, the cells were incubated for 90 min at 37°C with 5% CO₂. The absorbance of the samples at 450 nm was detected using a spectrophotometer (BioTek Instruments, Inc., Winooski, VT, USA). Relative cell viability (%) = (absorbance at 450 nm of the treated group-absorbance at 450 nm of the blank)/(absorbance at 450 nm of the control group-absorbance at 450 nm of the blank) × 100%.

HuCCT1 and RBE cells were seeded in 6-well plates at 2.5×10⁵ cells/well. At confluence, the cells were treated with or without 10 mM metformin for 24 h. Then the media and cells were collected separately for lactate production detection. The lactate production was detected with a lactate assay kit (K627; BioVision, Inc., Milpitas, CA, USA), according to the manufacturer's instructions. The absorbance values were measured at 450 nm with a spectrophotometer (BioTek Instruments, Inc.). The values were normalized to the cell number.

The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were detected in real time with an XF96 Extracellular Flux Analyzer from Seahorse Bioscience, Inc. (North Billerica, MA, USA), following the manufacturer's instructions. HuCCT1 and RBE cells were plated in 96-well XF cell culture microplates at 1×10⁴ cells/well and incubated for 24 h at 37°C. Then, 1 mM metformin was added into each well of the plates. Before measurement, the medium was replaced with 175 µl/well XF-96 running medium (supplemented with RPMI-1640 without serum) and pre-incubated at 37°C for 20 min in the absence of CO₂. For each analysis, different compounds that modulate mitochondrial respiration were injected into each well, according to standard protocols: for OCR, oligomycin, carbonylcyanide p-trifluoromethoxy-phenylhydrazone, rotenone and antimycin A; for ECAR, glucose, oligomycin and 2-deoxy-D-glucose. The cell number was used for data normalization. OCR is expressed as pmol/min (picomoles/minute). ECAR is expressed as mpH/min [milli-pH units (mpH) per minute].

Nude mice were purchased from the Department of Laboratory Animal Science, Nanjing Drum Tower Hospital. HuCCT1 cells (3×10⁶) in serum-free RPMI-1640 medium were subcutaneously injected into the right flank of the mice. Once palpable tumors were exhibited, mice were randomly assigned into 4 groups: control (100 µl natural saline, NS), metformin (200 mg/kg diluted in 100 µl NS), BG45 (20 mg/kg diluted in 100 µl NS), and a combination of both drugs (metformin at 200 mg/kg and BG45 at 20 mg/kg in 100 µl NS). Each group received an intraperitoneal injection 3 times/week for 4 weeks. Tumor volume (V) was calculated using the following formula: V = length × width²/2. All the experiments involving animals were reviewed and approved by the Animal Welfare Committee of Nanjing Drum Tower Hospital.

Data are expressed as the mean ± standard error of the mean (SEM). A one-way analysis of variance (ANOVA) was performed with Dunnett's multiple comparisons test (SPSS 17.0; SPSS, Inc., Chicago, IL, USA). The Chi-squared test and unpaired Student's t-test were performed for comparisons between two groups. Overall survival time was calculated by the Kaplan-Meier analysis. Staining scores were analyzed by the log-rank test. P<0.05 was considered to indicate a statistically significant difference. In figures, *P<0.05 and **P<0.01.

Reprogramming of cellular metabolism is common in tumor cells, and is regulated by the expression of multiple genes, thus accelerating the malignant behavior of tumor cells (24). Therefore, we analyzed the differentially expressed genes (DEGs) in CCA using data from The Cancer Genome Atlas (TCGA). We confirmed the DEGs and detected the potential biological functions and pathways of these genes in CCA through KEGG pathway analysis (Fig. 1A). We observed that the most relevant pathways in CCA were metabolic and tumor growth pathways (Fig. 1B). Together, these results indicate that metabolic reprogramming is an important feature of CCA; moreover, these results underscore the complex and unclear regulatory mechanisms.

Metformin is a widely accepted first-line drug for the treatment of type 2 diabetes (5). Many studies have shown that metformin could serve as a potential cancer therapy, although little is known concerning the mechanism of its antitumor functionality (25,26). It has been reported that the antitumor effects of metformin are partially caused by altering cellular metabolism (14). We then investigated the effects of metformin on the cellular metabolic status. After employing the Seahorse bio-energy analyzer, we found that OCR was notably increased, while ECAR was markedly decreased after the use of metformin (Fig. 2A-C). This indicated the metabolic shift from glycolysis to oxidative phosphorylation in CCA cells following treatment with metformin.

To further validate the anti-Warburg effect properties of metformin, we measured the level of lactate production after metformin. We found a significant decrease in the lactate production of both CCA cell lines after the treatment, which confirmed the effect of metformin on reversing the Warburg effect (Fig. 2D).

To clarify the mechanisms underlying the fact that metformin suppressed the Warburg effect, we detected a number of candidate proteins that may be involved in the Warburg effect. We found that the expression of LDHA was notably decreased following the use of metformin, although this result was not observed with the other possible proteins PKM2, PDHA1, GLUT1 and HDAC3 (Fig. 2E). Collectively, these data suggested that metformin reversed the Warburg effect by decreasing LDHA in CCA cells.

To explore the function of metformin in CCA, we detected the proliferation and apoptosis of CCA cells. Notably, we observed that a low concentration of metformin did not inhibit cell proliferation or induce cellular apoptosis, although the Warburg effect was suppressed (Fig. 3A-D). This suggested that metformin alone may only slightly affect the proliferation or apoptosis of CCA cells, although it suppresses the Warburg effect.

Several studies have reported that metabolic abnormalities can accelerate malignant behaviors, increase chemoresistance and inhibit tumor cell apoptosis (12,27). As we mentioned above, the present study demonstrated that metformin could regulate the energy utilization of CCA by decreasing the expression of LDHA. It is possible that reversing the Warburg effect in CCA cells with metformin could then increase tumor cell fragility and render them more susceptible to chemotherapy (14,28). Recently, we found that HDAC3 inhibitors are promising chemotherapeutic agents in the treatment of CCA (20). BG45, a novel HDAC3 selective inhibitor, has been validated as a therapeutic agent in multiple myeloma (16). However, our data showed that in CCA, the antitumor properties of BG45 were not significantly effective (Fig. 4A). As a result, we tested the combination treatment of metformin and BG45, and found that this combination inhibited cell viability (Fig. 4A and B). We further explored the mechanism of this combination-induced cell viability inhibition by performing flow cytometry. The results revealed that combined metformin and BG45 led to a significant increase in the apoptosis of CCA cells, compared with single drug use alone (Fig. 4C and D). Consistent with the flow cytometry results, higher levels of cleaved caspase 3 and cleaved PARP were found in CCA cells treated with the combined treatment, compared to single drug treatment (Fig. 4E). We further inhibited HDAC3 using siRNA (Fig. 4F), and found that the expression levels of cleaved caspase 3 and cleaved PARP were significantly increased. However, the levels of these two apoptotic markers were not promoted following treatment with the combination of both drugs after HDAC3 inhibition (Fig. 4G).

Collectively, these results demonstrated that combined treatment using metformin and BG45 could significantly inhibit the proliferation of CCA cells by inducing apoptosis. Although metformin alone hardly induced cellular apoptosis at low concentrations, it did nonetheless facilitate HDAC3 inhibitor BG45-induced apoptosis.

To further validate our findings in vitro, we evaluated the antitumor effect of the combined treatment in vivo using a CCA cell tumor xenograft model. We observed that the combined treatment group significantly inhibited tumor growth compared to the monotherapy groups (Fig. 5A and B). The weight loss of the mice was not found to be significant, which indicated that the combination therapy was safe in vivo (Fig. 5B). Altogether, these data revealed that the combination of metformin and BG45 could significantly induce cellular apoptosis and inhibit proliferation in vivo.

By evaluating the expression of LDHA on the tissue microarrays from Shanghai Outdo Biotech Co., Ltd., we found that LDHA was significantly upregulated in tumor tissues compared to that noted in adjacent tissues (Fig. 5C and D). Furthermore, we evaluated the clinical data of the tissue microarrays and found that LDHA protein was overexpressed in 68/127 cases (54.5%), and was associated with tumor size (Table I). Employing the 33 follow-up cases, we found that high LDHA protein expression in CCA reduced patient survival (P<0.001, log-rank test) (Fig. 5E). Next, we assessed the expression of HDAC3 and LDHA in fresh tissues, and found that HDAC3 and LDHA were markedly upregulated in tumor tissues compared to levels noted in normal tissues (Fig. 5F). Collectively, our data suggest that LDHA is overexpressed in CCA tissues and is associated with a worse prognosis.