Hypoxia and DNA hypermethylation are common in solid tumours, where reduced oxygen levels activate the HIF transcriptional complex ^{18, 19} , orchestrating the expression of a wide range of genes in an attempt to restore metabolic homeostasis ^{20, 21} . A recent study showed the reduced TET enzymatic activity in hypoxic tumour cells leads to the accumulation of 5-methylcytosine and cancer progression ³ .

We investigated the effect of culturing HepG2 cells in media with different glucose and glutamine concentrations on transcript levels of several HIF-target genes ( CA9, GLUT1, SCD1, BNIP3 and VEGFA) and dioxygenases ( TETs, ALKBH5 and FTO) genes under normoxic (20% oxygen) and hypoxic (1% oxygen) conditions. Under normoxic conditions, culturing cells in physiological glutamine or glucose media resulted in a significant change in gene transcript levels compared to cells propagated in DMEM ( Figure 2A). For example, we noted increased mRNA levels of GLUT1, SCD1, VEGFA, TET2, and TET3 in cells cultured in physiological glutamine or glucose compared to DMEM, suggesting that excess nutrients may dampen transcriptional activity. Comparing gene expression data from HepG2 cells cultured in low glucose or glutamine media relative to DMEM showed significant effects on gene expression and highlighted the gene-specific nature of the response ( Figure 2B).

Exposure of HepG2 cells cultured in DMEM to low oxygen increased the transcription of several HIF-target genes ( CA9, SCD1, BNIP3 and VEGFA) and the magnitude of the “hypoxic response” was dependent on the culture media in a gene-specific fashion ( Figure 3A). For example, hypoxic induction of CA9 was two-fold higher in HepG2 cells cultured in 0.5mM glutamine compared to DMEM, whilst mRNA levels for the TETs and FTO were significantly reduced in 5mM glucose but not in DMEM ( Figure 3A). This observation is consistent with a recent study reporting reduced TET activity in hypoxic tumour samples that associated with reduced 5hmC measurements in solid tumours ³ . Taken together, it is likely that excess glutamine/glucose levels perturb the metabolic cues that regulate cellular transcription and cellular responses to stress signals. The breakdown of homeostasis due to nutrient overload leading to aberrant cellular signalling is widely acknowledged to drive a myriad of pathologies such as cancer ^{6, 22} , fatty liver disease ^{23, 24} and type 2 diabetes ^{25– 27} .

Next, we determined whether the effect(s) of glucose/glutamine levels on HIF-target gene expression was explained by changes in HIF protein expression. HepG2 cells express HIF-1α under low oxygen conditions in all three media types ( Figure 3B), with an approximate 50% reduction in expression in cells cultured in 5mM glucose media ( Figure 3B). In HepG2 cells stably expressing a luciferase reporter under the control of a hypoxia responsive (HRE) promoter, we observed a significant reduction in promoter activity in hypoxic cells cultured in 5mM glucose media ( Figure 3C), consistent with reduced HIF-1α expression. These observations support an association between HIF-1α expression and HRE reporter activity, however, the impact of low glucose or glutamine on HRE activity did not correlate with other host transcriptional responses, demonstrating the simplistic nature of the reporter system. This illustrates the complex nature of HIF interactions with host gene promoter or enhancer elements, where other factors, such as the chromatin and methylation landscape determine the accessibility of any particular region to HIFs.

We predict that reduced TET expression in hypoxia would result in decreased DNA hydroxymethylation and increased DNA methylation. Since we observed a robust downregulation of TETs when HepG2 cells were cultured in 5mM glucose ( Figure 3A), we selected this media of choice for subsequent mass-spectrometry quantification of 5mC and 5hmC. Hypoxic cells showed a modest increase in DNA methylation and a significant reduction in DNA hydroxymethylation ( Figure 4). Low oxygen-dependent reduction of DNA hydroxymethylation is concomitant with a reduction in TET gene expression as they catalyse the conversion of 5mC to 5hmC ( Figure 3A). This observation is consistent with another report by Thienpont et al. ³ , suggesting that hypoxia-signalling promotes aberrant DNA methylation and can alter the epigenetic landscape of hepatocyte-specific transcription factors. Of note a recent study reported that low glucose could stabilize TET2 ²⁸ . Hence, glucose surplus in regular DMEM could perturb TET2 expression and may mask any further reduction of 5hmC in hypoxia.

Previous studies have demonstrated that oxidative derivatives of 5mC were decreased in hepatocellular carcinoma (HCC) ²⁹ . This correlated with a significant decrease in TET2 expression levels and alpha-ketoglutarate content in patients with HCC, suggesting that decrease in TET enzymatic activity resulted in altered expression of DNA methylation enzymes ^{29, 30} . Independently, using HCC mRNA expression profiles from The Cancer Genome Atlas, we found reduced TET2 transcripts in tumour compared to non-tumour samples ( Figure 5). However, TET1 and TET3 expression levels were upregulated in tumour samples, suggesting that they may compensate for reduced TET2 levels in HCC ( Figure 5). Importantly, decreased TET2 expression in HCC confirmed our earlier observation on hypoxic treated HepG2 cells cultured in physiological glucose media ( Figure 3A and Figure 4).

Human hepatoma HepG2 cells were cultured with regular, physiological glucose or glutamine media as follows: regular media is basic DMEM (ThermoFisher Scientific - A1443001) supplemented with 25mM glucose, 2mM glutamax, 10% fetal calf serum (FCS) and 1x penicillin-streptomycin (P/S); physiological glucose is basic DMEM with 5mM glucose, 2mM glutamax, 10% FCS and 1x P/S; physiological glutamine media is basic DMEM with 25mM glucose, 0.5mM glutamax, 10% FCS and 1X P/S. Cells were seeded at 26,000 cells/cm ² in collagen (Sigma) coated 6-well plates and incubated for 48h under 20% or 1% oxygen (Invivo2, Baker Ruskinn).

Cellular RNA samples were extracted from normoxic and hypoxic samples using the RNeasy Mini kit (Qiagen) with in-column DNAse digestion performed according to the manufacturer’s instructions. One-step qPCRs for CA9, GLUT1, BNIP3 and VEGFA were performed using a TaqMan master mix (Takyon, Eurogentec) and TaqMan MGB probes (Applied Biosystems). For detection of SCD1, TET1, TET2, TET3, FTO and ALKBH5 mRNA, cDNA syntheses from 1μg of total RNA were performed using a QuantiTect Reverse Transcription Kit (Qiagen), followed by qPCR using a SYBR green master mix (Applied Biosystems) ( Table S1). The hypoxia insensitive housekeeping gene Beta-2-Microglobulin ( B2M) was used as an endogenous control. Log2 fold changes were calculated using the 2 ^-∆∆Ct method. In order to depict gene expression levels under normoxia, ∆Ct values (Ct gene – Ct B2M) were calculated.

HepG2-NTCP cells were seeded at 26,000 cells/cm ² in collagen (Sigma) coated 6-well plates in the various media and left to adhere overnight. Cells were incubated at 20% or 1% oxygen for 48h and lysed using 8M urea lysis buffer. Approximately 20μg of protein was separated by PAGE and transferred to Polyvinylidene difluoride (PVDF) membrane and incubated with primary antibodies for HIF-1α (BD Biosciences, USA), or β-actin (Thermo Fisher Scientific, USA). Bound antibodies were detected using horseradish peroxidase (HRP) conjugated secondary antibodies (Dako, Agilent Technologies, USA) with chemiluminescent detection (SuperSignal, Thermo Fisher Scientific) using a PXi Touch Gel Imaging System (Syngene).

HepG2-NTCP cells cultured as described above were lysed and luciferase activity assessed using GloMax luminometer in accordance with the manufacturer’s instructions (Promega). Lysates were quantified for protein concentration using a BCA Protein Assay Kit (Thermo Fisher Scientific) and luciferase activity (relative light units, RLU) expressed relative to protein concentration.

DNA samples were incubated in hydrolysis solution containing 45 mM NaCl (Invitrogen), 9 mM MgCl ₂ (Ambion), 9 mM Tris-HCl (pH 7.9, Gibco), 25 U Benzonase Nuclease (Sigma-Aldrich), 5 mU Phosphodiesterase I (Sigma-Aldrich), 0.5 μg Alkaline phosphatase (Sigma-Aldrich), 9.36 ng/μL EHNA hydrochloride (Sigma-Aldrich) and 3.52 μM Deferoxamine (Sigma-Aldrich) at 37°C for 4 hours. The digested samples were filtered with Amicon Ultra-0.5 mL 10K centrifugal filters (Merck Millipore) to remove the proteins. The same volume of HPLC buffer A (H ₂O containing 10 mM ammonium acetate, pH 6.0) was added to the filtered samples, and then subjected to HPLC-MS/MS analysis.

The HPLC-MS/MS analysis was carried out with 1290 Infinity LC Systems (Agilent) coupled with a 6495B Triple Quadrupole Mass Spectrometer (Agilent). A ZORBAX Eclipse Plus C18 column (2.1 x 150mm, 1.8-Micron, Agilent) was used. The column temperature was maintained at 40°C, and the solvent system was water containing 10mM ammonium acetate (pH 6.0, solvent A) and water-acetonitrile (60/40, v/v, solvent B) with a 0.4 mL/min flow rate. The gradient was: 0–5 min; 0 solvent B; 5–8 min; 0–5.63 % solvent B; 8–9 min; 5.63% solvent B; 9–16 min; 5.63–13.66% solvent B; 16–17 min; 13.66–100% solvent B; 17–21 min; 100% solvent B; 21–24.3 min; 100–0% solvent B; 24.3–25 min; 0% solvent B. The dynamic multiple reaction monitoring mode (dMRM) of the MS was used for quantification. The source-dependent parameters were as follows: gas temperature 230°C, gas flow 14 L/min, nebulizer 40 psi, sheath gas temperature 400°C, sheath gas flow 11 L/min, capillary voltage 1500 V in the positive ion mode, nozzle voltage 0 V, high pressure RF 110 V and low-pressure RF 80 V, both in the positive ion mode. The fragmentor voltage was 380 V for all compounds, while other compound-dependent parameters were summarized in Table S2.

The liver cancer dataset (LIHC) used in this study was obtained from The Cancer Genome Atlas (TCGA) ³¹ . Gene expression profiles for LIHC were downloaded from the Broad Institute GDAC Firehose ( http://gdac.broadinstitute.org/). RSEM normalized expression profiles were converted to log ₂(x + 1) scale and separated according to TCGA barcodes to tumour and non-tumour categories. Box plots were generated using the R ggplot2.

Statistical analyses were performed using R (version 3.5.1). Details of statistical tests performed can be found in the figure legends.