ABSTRACT
-
Background/Aims
Metabolic dysfunction-associated steatotic liver disease (MASLD) is the most prevalent chronic liver disease worldwide. Aberrant DNA methylation, which is primarily maintained by DNA methyltransferase 1 (DNMT1), has been linked to metabolic dysregulation; however, its contribution to MASLD pathogenesis remains poorly defined. This study aimed to elucidate the role of DNMT1-mediated methylation in transcriptional regulation during MASLD progression and to determine whether DNMT1 inhibition can reverse disease-associated epigenetic and transcriptional alterations.
-
Methods
We conducted integrated analyses of the liver transcriptome (n=131) and DNA methylome (n=106) of patients with biopsy-proven MASLD. We evaluated the effect of DNMT1 inhibition with 5-aza-4′-thio-2′-deoxycytidine (Aza-TdC) on a diet-induced MASLD mouse model. Multiomics approaches, including DNA methylome profiling, lipidomics, RNA sequencing, and chromatin immunoprecipitation sequencing, were applied to elucidate the role of DNMT1-mediated DNA methylation in regulating pathogenic gene expression.
-
Results
DNA methylome profiling revealed increased methylation variability associated with increased DNMT1 expression in MASLD patients. DNMT1 inhibition ameliorated dysregulated lipid metabolism by reducing hepatic triacylglycerol accumulation and inflammation. Aza-TdC treatment partially reversed MASLD-related hypermethylation of hepatocyte nuclear factor 4 alpha (HNF4α)- and peroxisome proliferator-activated receptor alpha (PPARα)-regulated genes, restoring their transcriptional activity. Notably, Aza-TdC reactivated the gluconeogenic enzyme-encoding gene phosphoenolpyruvate carboxykinase 1 (PCK1), which was hypermethylated and transcriptionally repressed in MASLD. Targeted DNA methylation of the PCK1 promoter using CRISPRoff confirmed the direct epigenetic regulation of PCK1 expression.
-
Conclusions
Targeting DNMT1 may mitigate lipid dysregulation and inflammation by reversing hypermethylation and restoring HNF4α- and PPARα-dependent gene transcription, highlighting DNMT1 as a potential therapeutic target for MASLD.
-
Keywords: Non-alcoholic fatty liver disease; DNA methylation; DNA methyltransferase 1; Hepatocyte nuclear factor 4 alpha; Peroxisome proliferator-activated receptor alpha
Study Highlights
• Hepatic DNA methylome profiling revealed increased methylation variability associated with elevated DNMT1 expression in patients with MASLD.
• DNMT1 inhibition attenuated hepatic lipid accumulation and inflammatory features in vivo.
• Aza-TdC partially reversed the hypermethylation of HNF4α- and PPARα-regulated metabolic genes, including the reactivation of silenced PCK1, supporting a contributory epigenetic role of DNMT1 in MASLD progression.
Graphical Abstract
INTRODUCTION
Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly termed nonalcoholic fatty liver disease (NAFLD) [
1], has become the most prevalent chronic liver disease worldwide [
2–
4]. Its increasing incidence is closely linked to metabolic syndrome, obesity, and type 2 diabetes, which drive the progression from isolated steatosis to inflammation, fibrosis, and cirrhosis [
5]. Consequently, MASLD imposes a growing clinical and economic burden [
6]. Multiple factors, including diet, lifestyle, genetics, and gut microbiome alterations, contribute to insulin resistance, obesity, and MASLD progression.
Among these factors, the deregulation of epigenetic pathways has recently been recognized as a hallmark of MASLD. Epigenetic modifications, particularly DNA methylation, serve as key regulators of gene expression in metabolic disorders, including MASLD [
7–
11]. DNA methylation involves the covalent addition of a methyl group to the cytosine of CpG dinucleotides and is maintained through DNA replication by DNA methyltransferase 1 (DNMT1), a central enzyme responsible for preserving methylation patterns, enforcing transcriptional repression, and stabilizing chromatin architecture [
12,
13]. DNA methylation is dynamic and reversible and counterbalanced by active demethylation mediated by the ten–eleven translocation (TET) family of dioxygenases, which oxidize 5-methylcytosine to 5-hydroxymethylcytosine and related derivatives [
13]. Hepatic TET1 has been implicated in lipid metabolic control and MASLD progression [
14]. Emerging observations suggest that epigenetic regulation extends beyond the nuclear genome. DNMT1 has been reported to localize to mitochondria, where it may contribute to mitochondrial DNA methylation and influence mitochondrial function, while TET-mediated hydroxymethylation has also been proposed as a mechanism of mitochondrial epigenetic remodeling in metabolic disease [
15].
Aberrant DNA methylation has been linked to altered expression of genes involved in lipid metabolism [
16], insulin signaling [
17], inflammation, steatohepatitis, and fibrosis [
18]. Methylation signatures at fibrosis-related loci can distinguish mild from advanced cases of MASLD and alcohol-associated liver disease, although the informative CpG sites differ between the two conditions [
19,
20]. Importantly, the reversibility of DNA methylation suggests that targeting DNMT activity may offer intervention opportunities. Indeed, MASLD-associated changes in the methylation of metabolic and insulin signaling genes are partially reversed after bariatric surgery, highlighting the potential to modulate epigenetic states therapeutically [
17].
Despite this growing interest, the mechanistic role of DNMT1 in MASLD progression remains unclear. In particular, whether DNMT1-mediated methylation drives transcriptional dysregulation and whether inhibiting DNMT1 can reverse disease progression are unclear.
In this study, we investigated the role of DNMT1 in MASLD progression by examining its effects on nuclear DNA methylation and transcriptional regulation. We found that DNMT1-mediated DNA methylation suppressed key transcription factors (TFs), including hepatocyte nuclear factor 4 alpha (HNF4α) and peroxisome proliferator activated receptor alpha (PPARα), which govern essential metabolic genes. DNMT1 inhibition reversed aberrant methylation, restored HNF4α and PPARα occupancy at their target motifs, and reactivated downstream gene expression. Together, these findings identify DNMT1 as a critical epigenetic driver of MASLD and highlight its potential as a target for counteracting pathogenic epigenetic alterations.
MATERIALS AND METHODS
Patients
This study was approved by the Institutional Review Board of the Seoul Metropolitan Government Boramae Medical Center. Patients with biopsy-confirmed MASLD were enrolled according to the eligibility criteria outlined in our previous study [
21]. We established a prospective cohort from the Boramae MASLD registry (NCT 02206841), as previously described [
22]. The MASLD cohort included 106 Korean participants aged 25–80 years who sought care at Seoul Metropolitan Government Boramae Medical Center. The sample size was not based on a prespecified power calculation but represents an analysis of the established cohort.
Participants with radiological evidence of hepatic steatosis were eligible for study inclusion beginning in January 2013. The eligibility criteria for this study were as follows: (i) ≥18 years old, (ii) bright echogenic liver on an ultrasound scan, defined as increased hepatic echogenicity relative to the renal cortex with posterior beam attenuation, and (iii) unexplained high alanine aminotransferase levels above the reference range within the past 6 months. The following exclusion criteria were used: (i) hepatitis B or C virus infection; (ii) autoimmune hepatitis, primary biliary cholangitis, or primary sclerosing cholangitis; (iii) drug-induced liver injury or steatosis; (iv) Wilson disease or hemochromatosis; (v) excessive alcohol consumption (males >30 g/day, females >20 g/day); and (vi) a diagnosis of malignancy within the past year.
Among the eligible study participants, those with at least two of the following risk factors underwent liver biopsy: diabetes mellitus, central obesity (waist circumference ≥90 cm for men or ≥80 cm for women), a high level of triglycerides (≥150 mg/dL), a low level of high-density lipoprotein-cholesterol (<40 mg/dL for men or <50 mg/dL for women), and the presence of insulin resistance, hypertension, and clinically suspected metabolic dysfunction-associated steatohepatitis (MASH) or fibrosis [
23]. All participants were fully informed of the study protocol and provided written consent before enrollment. All the samples were obtained from a single hospital and processed together in one experimental batch within the same laboratory using an identical platform.
All liver specimens were collected from dissected parenchymal regions using standardized biopsy and processing procedures. Liver tissue samples were obtained from healthy control subjects who underwent liver biopsy as part of a pre-evaluation for liver donation, and normal liver parenchyma was collected from patients with benign intrahepatic tumors. In subsequent analyses, individuals without MASLD served as controls, whereas those with a liver affected by metabolic dysfunction-associated steatotic liver (MASL) or with MASH constituted the MASLD group. Clinical information, including concomitant medications and metabolic comorbidities, is provided in
Supplementary Table 1.
The EPIC BeadChip array (Illumina, San Diego, CA, USA) was used for methylation analysis [
21]. First, 1 μg of genomic DNA from human liver tissue was treated with 20 μL of sodium bisulfite solution using an EZ DNA Methylation-Gold kit (Zymo Research, Orange, CA, USA). Following bisulfite conversion, 4 μL of the bisulfite-modified DNA was amplified using an Infinium Methylation Assay kit (Illumina). The amplified DNA was then hybridized to an EPIC Bead-Chip and scanned using an Illumina iSCAN system. DNA methylation in human MASLD samples was analyzed using the SeSAMe (v1.20.0) pipeline for preprocessing. Several steps were applied, including qualityMask for filtering poor-quality probes, infiniumIChannel for inferring channels of Infinium-I probes, dyeBiasNL for correcting dye bias, pOOBAH for calculating
P-values, and noob for subtracting the background.
P-values <0.05 were calculated to exclude low-quality CpG sites. Probes located on sex chromosomes were excluded from the analysis to minimize potential inconsistencies between samples. Additionally, CpGs that overlapped with single-nucleotide polymorphism loci were removed. Normalized M values were used for the differential methylation analysis. Differentially methylated positions (DMPs) between the control and MASLD samples were identified using the following cutoffs: false discovery rate (
FDR) <0.001,
P-value <0.001, and absolute delta beta (|Δβ|) ≥0.07 (7%).
Six-week-old male C57BL/6J mice were purchased from ORIENTBIO (Seongnam, Korea) and housed at the Korea Research Institute of Bioscience and Biotechnology. The mice were housed in a specific-pathogen-free facility under controlled conditions of 23°C, 40–50% humidity, and a 12-h light–dark cycle, with free access to water throughout the study. All procedures were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee.
The sample size was determined a priori using GraphPad Prism (v10.5.0) based on pilot data. A group size of eight mice provided 80% power to detect an effect size of 1.1 for hepatic steatosis (α=0.05). The primary endpoints were hepatic steatosis and inflammatory cell infiltration as assessed by H&E staining; the secondary endpoints included hepatic gene expression (Scd1 and Tnfa) as measured by qPCR and protein expression (SCD1 and PLIN2) as measured by western blotting. All outcome assessments were performed in a blinded manner.
At 8 weeks of age, the mice were randomly divided into three groups: (1) the control group, which was maintained on a standard diet (SD, Teklad Global 18% Protein Rodent Diet, ENVIGO); (2) the MASLD group, which was fed an MASLD diet (Teklad Custom Diet, ENVIGO) supplemented with high fat, high sucrose and 1.25% cholesterol for 18 weeks; and (3) a third group that received the same MASLD diet for 18 weeks but additionally received oral administration of 5-aza-4′-thio-2′-deoxycytidine (Aza-TdC, 0.2 mg/kg [0.1×] or 2 mg/kg [1×], PINOTBIO) during the final 4 weeks before sacrifice. Aza-TdC is a DNMT1-targeting nucleoside analog that becomes incorporated into DNA and induces covalent trapping and degradation of DNMT1, resulting in DNA demethylation [
24]. Because Aza-TdC is cleared more rapidly in mice than in humans, with shorter plasma exposure and enhanced CYP-mediated metabolism, the selected dosing regimen was designed to achieve sufficient systemic exposure for
in vivo DNMT1 inhibition, consistent with prior preclinical studies demonstrating effective DNMT1 depletion with minimal systemic toxicity [
25].
The health status, weight loss, and clinical signs of the mice were monitored daily. No animals were excluded from the analysis. Limited treatment-related mortality associated with weight loss occurred in a small number of mice, but no hematologic abnormalities were observed.
Illumina mouse methylation BeadChip array experiment and data processing
An Infinium Mouse Methylation BeadChip Array (Illumina) was used for methylation analyses. First, 1 μg of genomic DNA extracted from mouse liver tissue was treated with 20 μL of sodium bisulfite solution provided in the EZ DNA Methylation-Gold Kit. Bisulfite-converted DNA (4 μL) was amplified using an Infinium Methylation Assay kit (Illumina). The amplified DNA was then hybridized to the bead chip and scanned using an Illumina iSCAN system. The methylation of mouse DNA was analyzed using the MM285 array with preprocessing steps performed in the SEnsible Stepwise Analysis of Methylation data (SeSAMe, v1.20.0) package in R software (v4.4.0). Strain inference was first conducted to apply strain-specific masking based on genetic variants. Quality masking (qualityMask) was then applied to exclude nonuniquely mapped probes and to minimize the inflation of out-of-band signals. Channel inference (inferInfiniumIChannel) and dye bias correction (dyeBiasNL) were subsequently performed, with channel inference preceding dye bias correction, because the latter relies on correct channel assignments. The detection P-values (pOOBAH) were calculated before background subtraction (noob) to ensure accurate out-of-band signal assumptions. CpG sites were filtered using a P-value threshold of 0.05, and DMPs were identified with cutoff values of P<0.001, FDR<0.05, and |Δβ|≥0.05 (5%). Normalized M values were used for all DMP analyses.
RNA sequencing (RNA-seq)
RNA-seq was used for transcriptome analysis. Total RNA was isolated from mouse liver tissue using the RNeasy Plus Mini kit (QIAGEN, Hilden, Germany). RNA integrity was verified using the 2100 Bioanalyzer System and the RNA 6000 Nano kit (Agilent Technologies, Santa Clara, CA, USA). RNA-seq libraries were prepared using the TruSeq Stranded mRNA Library Prep kit (Illumina) and sequenced using the Illumina HiSeq X Ten platform to generate 150-bp paired-end reads. The quality of the FASTQ files was assessed using FastQC (v0.11.9). Adapter trimming and removal of low-quality bases (Phred score <30) were performed using TrimGalore (v3.5).
The trimmed reads were subsequently aligned to the mm10 (GRCm38) mouse reference genome using STAR (v2.7.3a). Postalignment quality control was conducted using MultiQC (v1.9) to ensure data quality across samples. Read counts were quantified using HTSeq (v2.0.8), and differentially expressed genes were analyzed using edgeR (v4.3.4).
Chromatin immunoprecipitation sequencing (ChIP-seq)
Liver tissue (150 mg) from C57BL/6J mice was processed to isolate nuclei, cross-linked with 1% formaldehyde, and lysed using nuclear lysis buffer. Each condition included three biological replicates. Chromatin was sheared to <500 bp using Covaris M220 and immunoprecipitated with anti-HNF4α (Abcam, Cambridge, UK) or anti-H3K27ac (Abcam) antibodies prebound to Protein A Dynabeads (Invitrogen, Waltham, MA, USA). Following sequential low- and high-salt washes, the cross-links were reversed, and the DNA was purified with SPRIselect Beads (Beckman, Brea, CA, USA). Libraries were generated from 10 ng of DNA using the NEBNext Ultra II DNA Library Prep Kit (NEB, Ipswich, MA, USA), quantified on an Agilent 4200 TapeStation, and sequenced on an Illumina NovaSeq X platform (~20 million paired-end reads per sample). ChIP–PCR validation was performed on Pck1 promoter regions using primers targeting −436 to −233 and −28 to +140 relative to the transcription start site (TSS). The PCR products were resolved on 1.2% agarose gels and visualized using GelRed (Biotium, Fremont, CA, USA).
ChIP-seq analysis
Paired-end ChIP-seq raw data were assessed for quality using FastQC (v0.11.9) and adapter sequences and low-quality bases (Phred score <30) were removed using TrimGalore (v3.5). Trimmed and quality-controlled reads were aligned to the mm10 (GRCm38) mouse reference genome using BWA-MEM2 (v2.2.1). BAM files were processed using SAMtools (v1.13). Mitochondrial reads were removed, and PCR duplicates were filtered using PicardTools (v3.0) to retain unique reads. Aligned reads were filtered based on FLAG scores and a MAPQ ≤30 to remove multimapped reads and reduce nonspecific signals. MACS3 (v3.0.1) was used for peak identification. Peaks were called independently for each replicate using MACS3 (q=0.01). FRiP scores were calculated as the fraction of mapped reads overlapping MACS3 peak regions, and all H3K27ac samples exhibited FRiP scores >1%, which is consistent with ENCODE quality guidelines for histone ChIP-seq [
26]. The FRiP scores are provided in
Supplementary Table 2. To ensure reproducible peak identification, we generated a consensus peak set by merging per-replicate peaks and retaining only regions detected in all three replicates within each condition. This consensus peak set was used for read counting and differential binding analysis in DiffBind (DESeq2 framework). The design matrix included condition as the primary variable, with no additional covariates, as all the samples were processed in a single batch.
For H3K27ac ChIP–seq analysis, peaks were called independently for each biological replicate using MACS3 with uniform parameters. Consensus regions for read quantification were defined automatically by DiffBind during the construction of the binding matrix, which merged all sample-level peak sets into a unified reference space. Differential binding analysis was performed using normalized read counts derived from these consensus regions. For visualization, replicate bigWig signal tracks were averaged within each condition using the WiggleTool mean values to generate representative profiles. These averaged tracks were used as inputs for deepTools (computeMatrix and plotHeatmap) to generate condition-level heatmaps and signal profiles.
Epigenome editing of PCK1 using CRISPRoff and CRISPRon
Single guide RNAs (sgRNAs) targeting the PCK1 promoter and 5′ UTR were cloned and inserted into a pgRNA backbone (Addgene #44248) at the AarI restriction site. Lentiviral particles encoding sgRNAs were generated in HEK293T cells by cotransfection with psPAX2 and pMD2. G, concentrated by ultracentrifugation, and were used to transduce HEK293T or HepG2 cells. Stable PCK1 sgRNA-expressing lines were established using puromycin selection (1 μg/mL, 5 days). For dCas9 delivery, stable cells were seeded in 6-well plates and transfected with CRISPRoff-V2.1 (Addgene #167981) or CRISPRon_TETv4 (Addgene #167983) plasmids using FuGENE HD (Promega, Madison, WI, USA). Five days after transfection, BFP+/ mCherry+ double-positive cells were isolated using FACS and expanded. The cells were harvested 14 days after transfection for downstream analyses. Genomic DNA was extracted using a DNeasy Blood & Tissue Kit (QIAGEN) and subjected to bisulfite conversion (EZ DNA Methylation-Gold Kit; Zymo, Irvine, CA, USA), and DNA methylation at six CpG sites within the PCK1 promoter/5′ UTR was quantified by pyrosequencing.
Data analysis and presentation
Procedures to correct for multiple testing were applied consistently across all the omics layers. For each analysis, we report the corresponding FDR control method and threshold. The Benjamini–Hochberg (BH) procedure for FDR correction was used for differential expression analysis using RNA-seq, differential methylation analyses, and motif or ChromHMM enrichment tests. An FDR <0.05 was considered significant unless otherwise stated. All methylation–expression association tests were corrected for multiple comparisons using the BH FDR procedure, with significance supported by FDR-adjusted P-values and effect size estimates. The association models incorporated age, sex, body mass index, and T2D status to minimize confounding. Formal power calculations were not performed for each omics endpoint, as large-scale integrated methylation–expression, lipidomic, and single-cell analyses generally rely on FDR-controlled inferences rather than prespecified power thresholds.
Pearson’s correlation was used to assess pairwise associations. Correlation coefficients (R) and two-sided P-values were calculated using the cor.test function in R (v4.4.0). Visualizations, including scatter plots with fitted linear regression lines and 95% confidence intervals, were produced using the ggpubr package.
The data were analyzed using GraphPad Prism (v10.5.0). Differences between two groups were analyzed using an unpaired, Student’s two-tailed t-test. For comparisons involving more than two groups, we performed one-way ANOVA followed by the Holm–Šidák post hoc test to correct for multiple comparisons. Assumptions of normality and homoscedasticity were evaluated using the Shapiro–Wilk test and Levene’s test, respectively. Sample sizes (n) are provided in each figure legend. All the statistical tests were two-tailed, and P<0.05 was considered to indicate statistical significance. All the data are presented as the mean ± standard error of the mean (SEM) or standard deviation, as indicated. Gene expression and DNA methylation heatmaps were generated using the heatmap function of the ComplexHeatmap package (v2.18.0).
RESULTS
Increased heterogeneity of DNA methylation and DNMT1 expression in MASLD
Methylation profiles were analyzed in a cohort of 106 MASLD patients using the Illumina 850K array, which revealed 8,656 hypermethylated and 5,101 hypomethylated DMPs compared with those in healthy control subjects (
P<0.001, |Δβ|>0.07, and
FDR<0.001) (
Fig. 1A). Both hyper-and hypomethylated DMPs were enriched in open sea and intronic regions (
Supplementary Fig. 1A, 1B). Gene Ontology analysis indicated that hypermethylated DMPs were enriched in genes governing lipid metabolism, whereas hypomethylated DMPs were enriched in genes mediating immune responses (
Fig. 1B).
To identify key regulators of DNA methylation in MASLD patients, we analyzed RNA-seq data obtained from liver biopsies [
27]. Notably, DNMT family members, particularly
DNMT1 and
DNMT3A, were upregulated in MASLD patients compared with control subjects, and this upregulation correlated with MASLD progression (
Fig. 1C). Although the expression of the TET family, which functions in DNA demethylation, was slightly downregulated in MASH subjects, this difference was not statistically significant.
Consistent with the findings from the Korean MASLD cohort, increased
DNMT1 expression was also noted in a European MASLD cohort (GSE135251) [
28] compared with that in the control group (
Fig. 1D). A meta-analysis integrating multiple independent MASLD cohorts further confirmed a reproducible increase in
DNMT1 expression, as reflected by positive study-specific log
2 fold-change estimates with 95% confidence intervals and a statistically significant pooled effect (
Fig. 1E). In contrast,
DNMT3A and
DNMT3B did not show consistent upregulation across cohorts (
Supplementary Fig. 1C, 1D).
Pearson’s correlation analysis revealed that
DNMT1 expression was positively correlated with hepatocellular ballooning, lobular and portal inflammation, the nonalcoholic fatty liver disease activity score (NAS), and fibrosis, with corresponding correlation coefficients and 95% confidence intervals summarized in the forest plot (
Fig. 1F, 1G). Intriguingly,
DNMT1 expression was also positively correlated with clinical parameters associated with insulin resistance, including the circulating insulin concentration, homeostatic model assessment for insulin resistance (HOMA-IR), circulating HbA1c concentration, and insulin resistance in adipose tissue (
Supplementary Fig. 1E). However,
DNMT1 expression was not significantly correlated with clinical parameters such as body mass index or the triglyceride index (
Supplementary Fig. 1E).
In addition, a human skin stem cell-derived MASH model (GSE126484) [
29] showed increased
DNMT1 expression (
Supplementary Fig. 1F). In a mouse model, dietary manipulation leading to MASLD progression also resulted in elevated hepatic DNMT1 levels (
Supplementary Fig. 1G).
Collectively, MASLD is characterized by extensive hepatic DNA methylation remodeling involving genes linked to lipid metabolism and immune regulation. Across multiple patient cohorts and experimental models, DNMT1 is consistently upregulated and closely associated with histological severity, highlighting it as a molecular hallmark and potential mechanistic contributor to MASLD progression.
Inhibition of DNMT1 attenuates lipid accumulation and inflammation during MASLD progression
To explore the contribution of DNMT1 to MASLD, we tested the effect of Aza-TdC [
25] on a diet-induced mouse model (
Fig. 2A). Aza-TdC, a cytidine analog featuring an aza-substitution in the cytosine ring and a thioether-modified deoxyribose sugar, has recently been identified as a potent DNMT1 inhibitor with preclinical antitumor activity [
24,
25]. In our study, Aza-TdC treatment selectively depleted DNMT1 protein levels in hepatocyte cell lines (
Supplementary Fig. 2A) and attenuated MASLD progression
in vivo (
Fig. 2B), as evidenced by an improvement in hepatomegaly, a hallmark feature of diet-induced MASLD (
Supplementary Fig. 2B).
Histological analysis with BODIPY staining showed that excessive hepatic lipid accumulation induced by the MASLD diet was markedly reduced by DNMT1 inhibition (
Fig. 2C). Consistent with these findings, immunohistochemical staining showed decreased macrophage infiltration, as evidenced by reduced anti-F4/80 reactivity in the livers of Aza-TdC-treated mice (
Fig. 2C), indicating that DNMT1 inhibition also alleviated MASLD-associated inflammation. Moreover, histological analysis demonstrated marked reductions in hepatic steatosis and lobular inflammation, resulting in significantly lower NAS and fibrosis levels in the Aza-TdC-treated group (
Fig. 2D).
To determine the impact of MASLD on hepatic lipid composition and to assess whether DNMT1 inhibition could restore lipid homeostasis, we performed comprehensive lipidomic profiling in each mouse group. A total of 132 distinct lipid species were identified, and partial least squares discriminant analysis clearly separated the lipid profiles among the SD, MASLD, and MASLD plus Aza-TdC groups (
Fig. 2E). A 1,000-fold permutation test confirmed that the PLS-DA model was not overfitted (pR
2Y=0.001; pQ
2=0.001), supporting the robustness of the differences in lipidomic profiles between the groups (
Supplementary Fig. 2C). Among these lipid species, 63 were significantly altered in the MASLD group compared with the SD group (
Supplementary Fig. 2D). The MASLD diet increased in the levels of 28 lipid species, whereas DNMT1 inhibition resulted in decreased levels of five lipid species, predominantly triacylglycerols (TAGs) (
Fig. 2F). Notably, the levels of five TAG species that were markedly elevated in the MASLD group were substantially reduced by Aza-TdC treatment, particularly those containing unsaturated fatty acid chains, including 50:1–16:0/16:0/18:1, 50:2–16:0/16:0/18:2, 50:3–16:1/16:1/18:1, 52:2–16:0/18:1/18:1, and 56:4–18:1/18:2/20:1 (
Fig. 2G). In contrast, other classes of lipids, such as phosphatidylethanolamine and phosphatidylcholine, remained largely unaffected by Aza-TdC treatment, despite the overall reduction in hepatic lipid accumulation (
Supplementary Fig. 2D). These findings indicate that DNMT1 suppression selectively alleviates lipid dysregulation in a model of MASLD primarily by reducing TAG accumulation, which may contribute to the mitigation of metabolic stress and liver inflammation.
To evaluate transcriptomic changes, we performed RNAseq analysis to investigate the molecular mechanisms underlying DNMT1 inhibition in MASLD. Bulk RNA-seq revealed that DNMT1 inhibition significantly reduced the expression of genes involved in lipid metabolism and inflammation (
Supplementary Fig. 3A). qRT–PCR further confirmed that the inhibition of DNMT1 suppressed the MASLD-induced upregulation of genes related to lipid uptake (
Cd36), transport (
Fabp1,
Fabp4, and
Fabp7), and droplet dynamics (
Plin2,
Plin3, and
Cidec) (
Fig. 2H). In contrast, the expression of key regulators of
de novo lipogenesis, including
Acaca and
Fasn, remained unchanged, whereas the expression of
Scd1, encoding a critical enzyme for fatty acid desaturation, was markedly downregulated. These results suggest that DNMT1 plays an essential role in controlling lipid accumulation under MASLD conditions.
Consistent with these findings, RNA-seq and qRT–PCR demonstrated that DNMT1 inhibition reduced the expression of inflammatory markers such as
Adgre1 (which encodes the macrophage marker F4/80) and cytokine genes such as
Tnfa,
Il6, and
Tgfb1 (
Fig. 2I,
Supplementary Fig. 3A). Although the expression of collagen genes associated with fibrosis (
Col1a2,
Col3a1,
Col4a2, and
Col5a2) was significantly decreased (
Fig. 2I,
Supplementary Fig. 3A), Sirius Red staining did not reveal collagen deposition, which is consistent with an early-stage MASH phenotype. These findings suggest that DNMT1 inhibition may attenuate the transcriptional activation of fibrogenesis, likely through its anti-inflammatory effects. Protein analyses corroborated these results, showing that the elevated levels of PLIN2, CD36, and SCD1 in MASLD livers were reduced after Aza-TdC treatment (
Supplementary Fig. 3B).
To further examine hepatocyte-specific responses, we performed single-nucleus RNA sequencing (snRNA-seq) of liver tissues from each dietary group and selected nuclei with a hepatocyte transcriptional profile (
Supplementary Fig. 3C). Gene Ontology analysis revealed significant downregulation of lipid metabolism-related genes in hepatocytes following DNMT1 inhibition (
Supplementary Fig. 3D), consistent with the results of bulk RNA-seq. Notably, while the expression of genes related to hepatocyte lipid metabolism was strongly affected (
Supplementary Fig. 3E), the change in the expression of genes related to hepatocyte inflammation was relatively limited (
Supplementary Fig. 3F).
Together, these findings indicate that DNMT1 inhibition mitigates MASLD progression primarily by reducing lipid accumulation and, to a lesser extent, by modulating inflammatory and fibrotic responses.
HNF4α and PPARα are essential TFs whose binding is substantially influenced by altered DNA methylation in MASLD
We next examined the DNA methylome in a mouse model using a Mouse Methylation BeadChip Array. Principal component analysis of global methylation profiles revealed clear separation among the control, MASLD, and Aza-TdC-treated groups (
Supplementary Fig. 4A), indicating marked epigenetic reprogramming in MASLD that was altered by DNMT1 inhibition. DMPs were classified into hypermethylated (C1) and hypomethylated (C2) clusters following MASLD induction, confirming the extensive remodeling of the methylation landscape (
Fig. 3A). Aza-TdC treatment induced a pronounced reversal of these methylation patterns (
Fig. 3A). Gene Ontology enrichment analysis revealed that C1 DMPs were predominantly associated with lipid metabolic pathways, whereas C2 DMPs were enriched in immune-related processes (
Fig. 3B), underscoring the pivotal role of DNMT1 in the epigenetic regulation of metabolic and immune functions in a MASLD model.
To investigate the role of DNA methylation in restricting TF binding, we analyzed TF binding motifs within hypermethylated and hypomethylated DMPs in human MASLD patients compared with those in healthy controls (
Fig. 3C,
left panel;
Supplementary Fig. 4B,
upper panel). Motif enrichment analysis revealed the significant enrichment of several hepatic TFs, including HNF4α, PPARα, HNF1α, HNF1β, FOXA1, and FOXM1, among the hypermethylated DMPs in MASLD (
Figs. 1A and
3C,
left panel).
We next assessed whether these TFs, which are consistently affected in humans with MASLD, are also associated with DNMT1-dependent changes in methylation in a mouse model. Specifically, we examined C1 DMPs that were hypermethylated in response to the MASLD diet and became hypomethylated after DNMT1 inhibition (
Fig. 3C,
right panel). Consistent with the human data, the same set of TFs, including the six mentioned above, was enriched within C1 DMPs in the MASLD mouse model. We also analyzed C2 DMPs that displayed opposite methylation patterns (
Supplementary Fig. 4B,
lower panel).
Correlation analyses further revealed that
HNF4A and
PPARA expression was inversely associated with histological indicators of disease activity, such as ballooning and lobular inflammation, in MASLD subjects (
Fig. 3D). In comparison, the expression of
HNF1A,
HNF1B,
FOXA1, and
FOXA2 showed weak or no correlations, whereas
FOXM1 expression correlated positively with ballooning and lobular inflammation (
Supplementary Fig. 4C). Moreover,
DNMT1 expression inversely correlated with
HNF4A and
PPARA expression in human samples (
Fig. 3E).
Together, these findings indicate that specific TFs, particularly HNF4α and PPARα, are significantly enriched in hypermethylated genomic regions associated with MASLD, with consistent evidence obtained from both patients and the MASLD mouse model. Importantly, the inverse correlations between TF expression and histological indicators underscore their functional relevance in MASLD, with their binding sites notably affected by changes in DNA methylation during disease progression.
DNMT1 inhibition reactivates the expression of genes regulated by HNF4α and PPARα
Given the established roles of HNF4α and PPARα in lipid metabolism and inflammation associated with MASLD [
30,
31], we examined the relationship between DNA methylation and the transcriptional activity of genes targeted by these TFs. Genes that were hypermethylated and downregulated in MASLD mice compared with controls were first identified, with a particular focus on those containing HNF4α- or PPARα-binding motifs (
Fig. 4A,
left panel). Conversely, we selected hypomethylated genes whose expression was increased in the Aza-TdC (1×) group compared with that in the MASLD model mice (
Fig. 4A,
right panel). This analysis revealed 42 overlapping genes that were altered in both the MASLD/SD and Aza-TdC/MASLD comparisons, including
solute carrier family 38 member 3 (
Slc38a3),
phosphoenolpyruvate carboxykinase 1 (
Pck1),
interleukin 6 receptor alpha (
Il6ra), and
prolactin receptor (
Prlr).
Pyrosequencing confirmed reduced CpG methylation within
Slc38a3,
Pck1,
Il6ra, and
Prlr following DNMT1 inhibition (
Fig. 4B), whereas qRT–PCR demonstrated the significant transcriptional upregulation of these genes (
Fig. 4C). Consistently, snRNA-seq analysis revealed restored expression of these genes in hepatocytes from Aza-TdC-treated mice (
Fig. 4D).
In the human MASLD cohort, the expression of these genes was inversely correlated with that of
DNMT1 and positively correlated with that of
HNF4A or
PPARA (
Fig. 4E). The progressive downregulation of these genes accompanied disease progression (
Fig. 4F), whereas their methylation levels increased (
Fig. 4G,
Supplementary Fig. 5A).
Taken together, these findings indicate that the inhibition of DNMT1 in individuals with MASLD attenuates hypermethylation at regulatory regions of key metabolic and inflammatory genes, thereby restoring HNF4α- and PPARα-mediated transcriptional activity.
Restoration of HNF4α binding and PCK1 regulation following DNMT1 inhibition
To investigate the transcriptional regulatory mechanisms associated with differential DNA methylation, we analyzed the genomic binding targets of HNF4α and PPARα using publicly available ChIP-seq datasets (GSE90533 and GSE61817). These analyses were performed in the context of chromatin states associated with hypermethylated (C1) and hypomethylated (C2) DMPs (
Fig. 3A). Chromatin state enrichment analysis revealed that both C1 and C2 DMPs were predominantly enriched in enhancer regions (Enh, Enh-Lo, and EnhPois) and quiescent regions (QuiesG) (
Fig. 5A,
left panel) [
32]. Notably, HNF4α and PPARα binding was selectively enriched within the C1 group (
Fig. 5A,
right panel).
Next, we assessed whether DNMT1 inhibition could restore Hnf4α binding. ChIP-seq analysis of liver tissue from MASLD mice showed a global reduction in Hnf4α binding at promoter and enhancer regions marked by H3K27ac (
Fig. 5B). Treatment with Aza-TdC significantly restored Hnf4α binding at these loci, whereas regions outside the promoters and enhancers were unaffected, indicating that DNMT1-mediated methylation selectively impaired Hnf4α recruitment to regulatory regions in MASLD.
Among the Hnf4α targets,
Pck1, which encodes a key gluconeogenic enzyme [
33], emerged as the top locus with dynamic Hnf4α recruitment (
Fig. 5C). Hepatic PCK1 deficiency not only promotes lipogenesis but also activates the RhoA/PI3K/AKT pathway through increased GTP levels, PDGF-AA secretion, and hepatic stellate cell (HSC) activation [
34]. A meta-analysis of multiple MASLD cohorts revealed consistent downregulation of
PCK1 (
Fig. 5D). The Aza-TdC-responsive sites in
Pck1 overlapped with the shared Hnf4α and Pparα occupancy (
Fig. 5E). Similarly, Aza-TdC restored Hnf4α binding at additional targets, including
Il6ra,
Slc38a3, and
Prlr (
Supplementary Fig. 5B). ChIP–PCR further validated the reduction in Hnf4α binding to the
Pck1 promoter in MASLD mice, which was reversed by Aza-TdC treatment (
Fig. 5F).
To determine whether promoter methylation directly regulates
PCK1 transcription, we employed CRISPRon/off–based targeted epigenome editing (
Fig. 6A) [
35]. Directed methylation of the
PCK1 promoter using CRISPRoff markedly reduced
PCK1 expression in HEK293T cells, whereas targeted demethylation with CRISPRon increased its expression (
Fig. 6B, 6C). These results demonstrate that DNA methylation is sufficient to modulate
PCK1 transcription.
To assess target specificity, we quantified CpG methylation across a 5-kb region surrounding the
PCK1 transcription start site (TSS ±2.5 kb), encompassing 109 CpG sites. Pronounced changes in methylation were confined to the CpGs proximal to the gRNA—specifically, 15 CpG sites— while methylation at other CpGs remained comparable to that of the control (
Fig. 6D). These data indicate that CRISPRon/ off editing induces localized and specific CpG modifications within the
PCK1 TSS region.
We next applied this system to HepG2 hepatocyte-derived cells. Stable
PCK1 promoter methylation and transcriptional repression were achieved with
PCK1 CRISPRoff (
Supplementary Fig. 6), resulting in significant TAG accumulation (
Fig. 6E). The expression of lipogenic genes (
SCD1,
SREBP1, and
FASN) and the lipid-droplet-associated gene
PLIN2 increased with
PCK1 inhibition (
Fig. 6F), which is consistent with the role of PCK1 in hepatic lipid metabolism. The levels of inflammatory markers were not appreciably induced, likely reflecting the limited immune component of HepG2 cells relative to the liver
in vivo.
Given reports that hepatocyte PCK1 deficiency promotes MASLD progression via paracrine PDGF-AA secretion [
34] and evidence that PCK1 and PPARα protect against MASH, particularly during intermittent fasting [
36], we evaluated PDGFR signaling. The expression of genes in the PDGF pathway was upregulated in patients with MASLD (
Fig. 6G), whereas DNMT1 inhibition suppressed
Pdgfa,
Pdgfra,
Pdgfrb, and
Des expression in MASLD mice (
Fig. 6H), suggesting that restoring PCK1 expression through DNMT1 inhibition may counteract PDGFR signaling and thereby mitigate pathways that drive MASLD progression.
DISCUSSION
Research on MASLD has traditionally emphasized environmental factors and genetic variation [
37], yet these influences alone cannot account for its increasing global burden. Increasing evidence points to epigenetic mechanisms as critical contributors to the development and progression of MASLD. Our study identifies DNMT1 as a central driver of MASLD-associated DNA methylation, specifically at loci containing binding sites for HNF4α and PPARα. This hypermethylation diminishes the binding of these TFs, leading to reduced transcription of key genes such as
Pck1,
Slc38a3,
Prlr, and
Il6ra, which regulate lipid metabolism, inflammation, and fibrosis. Importantly, pharmacological inhibition of DNMT1 with Aza-TdC reversed these epigenetic changes, restored TF binding, and reactivated gene expression, ultimately alleviating MASLD features
in vivo.
We observed consistent upregulation of DNMT1 expression in liver tissues from patients with MASLD and in a diet-induced mouse model, in line with previous reports [
38]. In contrast, DNMT3A expression exhibited greater interpatient variability. Prior studies have suggested diet-specific roles of DNMT3A, as a high-fat diet increased DNMT1 and DNMT3A levels, worsening hepatic steatosis through hypermethylation of the
Klb promoter, whereas genetic deletion of these enzymes alleviated lipid accumulation [
38]. Other studies have implicated DNMT3A in the FGF19–SHP-mediated repression of lipogenesis [
39]. These discrepancies underscore the complexity of DNMT3A regulation and highlight the need for further mechanistic investigation.
Aza-TdC showed relative selectivity for DNMT1 and mitigated MASLD progression with minimal off-target effects in our model [
24,
40]. DNMT1 inhibition reduced hepatic TAG accumulation, which is consistent with previous findings [
41]. However, cholesterol levels and the expression of associated genes were unaffected, suggesting that the contribution of cholesterol metabolism to MASLD warrants further investigation, as lipidomic studies have proposed a link to cholesterol dysregulation [
41–
43].
DNMT1 inhibition significantly influences inflammation beyond lipid metabolism. The inhibition of DNA methylation or genetic deletion of
DNMT1 in macrophages promotes alternative activation, reduces inflammation, and protects against obesity-related insulin resistance [
44]. These findings emphasize the importance of examining the cell type-specific roles of DNMT1, particularly in macrophage-mediated inflammatory responses, to clarify its contribution to MASLD pathogenesis.
Our data suggest that DNMT1-mediated hypermethylation restricts HNF4α and PPARα activity, whereas DNMT1 inhibition restores TF binding and downstream gene expression. HNF4α protects against diet-induced MASLD through the regulation of lipolysis, p53 signaling, and bile acid pathways [
30], whereas PPARα remains a central regulator of hepatic lipid homeostasis [
31]. However, the broader epigenetic dimensions of these TFs in MASLD progression remain poorly understood.
Among the affected genes,
Slc38a3,
Il6ra, and
Prlr exhibited the strongest DNMT1-dependent repression.
Slc38a3 encodes the primary glutamine transporter in the liver, which is essential for ammonia detoxification [
45–
47]. Il6ra mediates IL-6 signaling, which is elevated in individuals with MASH and driven by myeloid cells such as macrophages and neutrophils [
48,
49]. Prlr is involved in metabolic homeostasis, as prolactin signaling through PRLR and CD36 ameliorates hepatic steatosis by increasing fatty acid uptake [
50]. These genes represent biologically plausible mediators of MASLD progression and may hold translational potential.
We also identified
PCK1 as an HNF4α-regulated gene related to promoter hypermethylation in individuals with MASLD. Beyond its canonical role in gluconeogenesis [
33], PCK1 has recently been implicated in restraining MASLD progression [
34,
36]. Hepatocyte-specific loss of PCK1 increases PDGF-AA secretion, driving HSC activation and extracellular matrix deposition via RhoA–PI3K–AKT signaling, likely through increased intracellular GTP availability [
34]. Increased DNMT1 expression may contribute to the repression of
PCK1 transcription via promoter hypermethylation, thereby impairing HNF4α binding and augmenting PDGFR signaling. To date, only a limited number of studies have reported this hepatocyte–HSC paracrine mechanism, and independent confirmation is still lacking. Thus, while our results are consistent with the existing evidence, broader validation is needed to confirm the role of this pathway in MASLD.
Our experimental model recapitulates early-stage MASLD characterized by a lipid imbalance and mild inflammation/ fibrosis. While this model provides mechanistic insights, DNMT1 inhibition was evaluated only in hepatic tissue; the systemic consequences remain uncertain. Chronic DNMT1 suppression may disrupt global epigenetic homeostasis, potentially inducing genomic instability, immune activation via endogenous retrovirus reactivation, or hematopoietic toxicity. These considerations highlight the need for liver-targeted or transient DNMT1 modulation strategies. Liver-activated prodrugs, CRISPR-based epigenetic editing, or hepatocyte-derived exosomal delivery could increase therapeutic specificity.
Although DNMT1 upregulation may, in principle, be influenced by metabolic or inflammatory stress, reverse causality cannot be completely ruled out. However, the consistent directionality observed across early-stage human samples, diet-induced mouse models, and the rescue of transcriptional repression by DNMT1 inhibition collectively argue against this alternative interpretation.
Genome-wide methylome profiling revealed that DNMT1-dependent changes in methylation predominantly occur in CpG island shores and open-sea regions, suggesting that enhancer rewiring and altered TF accessibility occur rather than classical promoter silencing. These findings highlight both the mechanistic specificity and the translational boundaries of DNMT1 targeting in patients with MASLD.
Although DNMT1 inhibition and CRISPRon/off perturbations provide compelling evidence of DNMT1 activity in regulating hepatic methylation and transcriptional programs as a cause of MASLD, genetic validation remains essential. Hepatocyte-specific Dnmt1 knockdown or conditional knockout would confirm hepatocyte-intrinsic mechanisms, exclude systemic or nonparenchymal contributions, and enable long-term safety assessments of DNMT1 suppression in vivo.
In conclusion, our findings demonstrate that DNMT1 promotes MASLD progression by epigenetically repressing HNF4α- and PPARα-dependent metabolic pathways. DNMT1 inhibition with Aza-TdC restored the transcription of metabolic and inflammatory regulators, improved lipid homeostasis, and attenuated hepatic inflammation. These results position DNMT1 as a central epigenetic regulator of hepatocellular function in MASLD and underscore the need for tissue specific, validated and safe approaches prior to translational application. Future studies should elucidate the systemic role of DNMT1 and its interactions with major transcriptional regulators to guide the development of precise, liver-targeted epigenetic interventions.
FOOTNOTES
-
Access to data and data analysis
The findings of this study are supported by the data, which are available from the corresponding authors upon reasonable request. The raw data obtained in this study were deposited in the Korea BioData Station (K-BDS, https://kbds.re.kr) under the accession numbers KAP241056 and KAP241057 and in the Gene Expression Omnibus (GEO) with the following accession numbers: GSE284446 for mouse RNA-seq, GSE283953 for the mouse DNA methylation array, GSE285014 for mouse sn-RNA-seq, and GSE284617 for mouse ChIP-seq.
-
Authors’ contribution
M.K., W.K. and K.-J.O. conceived and designed the study. W.K. and D.H.L. contributed to sample acquisition. H.A.S., H.G., T.H.A., J.M.L., H.-J.K., K.H., H.-J.J., Y.-J.S., H.J.L., and S.-H.P. performed the experiments. K.-J.O. and T.H.A. generated the mouse model. H.A.S., H.G., A.M., Y.J., M.J.S., E.-W.C., E.-S.K., M.C., W.K., and M.K. analyzed and interpreted the data. K.C.P. and E.-S.K. were responsible for project administration. Y.B., Y. Jung, and G.-S.H. conducted the lipidomic analysis. J.H.P. contributed to the analysis of the immunohistochemistry data and histological scoring. H.A.S., H.G., W.K., and M.K. drafted the manuscript with input from all the authors. M.K., W.K., and K.-J.O. supervised the research and served as guarantors of this work. All the authors read and approved the final manuscript.
-
Acknowledgements
We would like to express our gratitude to Doo Young Jung, CEO of PINOTBIO (http://pinotbio.com), for providing Aza-TdC. This work was supported by National Research Foundation of Korea (NRF) grants funded by the Korean government (2020R1A2C2102308, RS-2021-NR056442, RS-2022-NR067269, RS-2023-00223831, RS-2024-00440883, RS-2024-00449704, RS-2023-NR076422, RS-2025-00522077, RS-2025-25458964, and RS-2025-02305013), the National Research Council of Science & Technology (NST) Aging Convergence Research Center (CRC22014-400), the Korea Basic Science Institute (A423200), and the Korea Research Institute of Bioscience & Biotechnology (KRIBB) Research Initiative Program (KGM5192531 and KGM5392414).
-
Conflicts of Interest
Two patent applications related to this work have been filed by the authors’ institution, listing some of the authors as inventors. The authors declare no other conflicts of interest.
SUPPLEMENTARY MATERIAL
Supplementary material is available at Clinical and Molecular Hepatology website (
http://www.e-cmh.org).
Supplementary Figure 1.
DNMT1 expression correlates with MASLD. (A, B) Stacked bar plots of genomic features (A) and CpG islands (B) of hyper- and hypomethylated CpGs compared to background CpGs. (C, D) Random-effects model estimates, weighted by inverse variance, are presented in boldface type alongside sample size–weighted estimates of DNMT3A (C) and DNMT3B (D) expression differences between Control and MASLD groups from public RNA expression data. (E) Scatter plots showing relationships between DNMT1 expression and clinical parameters. (F) Heatmap of scaled DNMT and TET family gene expression in a human skin–derived induced pluripotent stem cell-MASH model (GSE126484). (G) Immunoblot of DNMT1 in a standard diet (SD) and a high-fat and choline-deficient (HFCD) diet-induced MASLD mouse model. BMI, body mass index; DBP, diastolic blood pressure; DNMT1, DNA methyltransferase 1; HOMA-IR, homeostatic model assessment for insulin resistance; MASH, metabolic dysfunction–associated steatohepatitis; MASLD, metabolic dysfunction–associated steatotic liver disease; SBP, systolic blood pressure; TET, ten-eleven translocation; TG, triglyceride.
cmh-2025-1099-Supplementary-Fig-1.pdf
Supplementary Figure 2.
DNMT1 inhibition attenuates MASLD-related phenotypes. (A) DNMT1 inhibition in Huh-7, HepG2, Hep3B, and SNU-387 cells following 24-h treatment with Aza-TdC (4 μM). (B) Representative liver images from mice across the indicated diet groups. (C) PLS-DA performance and permutation analysis. The observed R²Y and Q² values exceeded those from 1,000 permuted models (pR²Y=0.001; pQ²=0.001), confirming model robustness (left). The PLS-DA score plot shows clear separation among SD, MASLD, and Aza-TdC groups with 95% confidence ellipses (R²X=0.707; R²Y=0.863; Q²Y=0.687; RMSEE=0.196) (right). (D) Heatmap showing normalized relative abundance of 63 significantly different lipids from lipidomic analysis (*P<0.05). Lipid classes are abbreviated as follows: free fatty acids (FFA), lysophosphatidylcholine (LysoPC), lysophosphatidylethanolamine (LysoPE), phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylserine (PS), sphingomyelin (SM), and triacylglycerol (TAG). Aza-TdC, 5-aza-4′- thio-2′-deoxycytidine; DNMT1, DNA methyltransferase 1; MASLD, metabolic dysfunction–associated steatotic liver disease; SD, standard diet.
cmh-2025-1099-Supplementary-Fig-2.pdf
Supplementary Figure 3.
(A) Heatmap showing gene expression related to lipid metabolism and inflammation/fibrosis (n=3 per group). (B) Immunoblot analysis of PLIN2, CD36, SCD1, and DNMT1. (C) UMAP visualization of hepatocytes from snRNA-seq analysis. (D) Gene Ontology analysis of differentially expressed genes in hepatocytes upon DNMT1 inhibition. (E, F) Dot plot depicting relative expression of lipid metabolism (E) and inflammation-associated genes (F) in hepatocytes from snRNA-seq analysis. Dot size represents fraction of cells expressing the gene, and the color gradient shows relative expression levels. Aza-TdC, 5-aza-4′-thio-2′-deoxycytidine; DNMT1, DNA methyltransferase 1; MASLD, metabolic dysfunction–associated steatotic liver disease; SD, standard diet; snRNA-seq, single-nucleus RNA sequencing; UMAP, Uniform Manifold Approximation and Projection.
cmh-2025-1099-Supplementary-Fig-3.pdf
Supplementary Figure 4.
Enrichment of HNF4α and PPARα motifs at hypomethylated DMPs in MASLD. (A) Principal component analysis of SD, MASLD, and Aza-TdC (0.1×, 1×) samples based on common DMPs. (B) Bubble scatter plot of TF motif enrichment for hypomethylated CpGs in MASLD patients (upper) and hypomethylated CpGs restored by Aza-TdC (1×) treatment in mice (lower). Bubble size represents the P-value, with colors indicating the TF family. (C) Pearson correlations between estimated TF expression and clinical indicators (ballooning, lobular inflammation). Blue, control; red, MASLD patients. Aza-TdC, 5-aza-4′-thio-2′-deoxycytidine; DMPs, differentially methylated positions; HNF4α, hepatocyte nuclear factor 4 alpha; MASLD, metabolic dysfunction–associated steatotic liver disease; PPARα, peroxisome proliferator-activated receptor alpha; SD, standard diet; TF, transcription factor.
cmh-2025-1099-Supplementary-Fig-4.pdf
Supplementary Figure 5.
(A) Locus of methylation probes corresponding to Figure 4G. (B) DNMT1 inhibition recovers HNF4α binding. (A) Genome tracks showing peaks for Slc38a3, Il6ra and Prlr. These tracks show publicly available ChIP-seq data for Hnf4α and Pparα in wild-type (WT) and knockout (KO) mice as well as Hnf4α ChIP-seq data for our MASLD mice. ChIP-seq, chromatin immunoprecipitation sequencing; DNMT1, DNA methyltransferase 1; HNF4α, hepatocyte nuclear factor 4 alpha; MASLD, metabolic dysfunction–associated steatotic liver disease.
cmh-2025-1099-Supplementary-Fig-5.pdf
Supplementary Figure 6.
Effect of CRISPRon/off targeting the PCK1 promoter on CpG methylation (A) and mRNA expression (B) in HepG2 cells. Student’s two-tailed t-test. (C) Schematic overview for editing PCK1 promoter methylation via CRISPRon/off system. PCK1, phosphoenolpyruvate carboxykinase 1; TF, transcription factor.
cmh-2025-1099-Supplementary-Fig-6.pdf
Figure 1Changes in DNA methylation and DNMT1 activation in MASLD. (A) Heatmap of differentially methylated positions (DMPs) (|Δβ|>0.07; ***FDR<0.001, ***P<0.001) between healthy controls (Con; gray) and patients with MASLD (MASL; yellow) or MASH (navy). DNA methylation is shown as scaled β values. Violin plots (right) display the distribution of scaled DNA methylation levels for hypermethylated and hypomethylated DMPs. (B) Gene Ontology analysis of hypermethylated and hypomethylated DMPs performed using GREAT, with filtered CpGs from the EPIC array serving as the background. (C, D) Heatmap showing scaled expression of DNMT and TET family genes in the Korean cohort (C) and European cohort (GSE135251) (D). (E) Estimates from the random effects model, weighted by inverse variance, are shown in bold along with sample size-weighted estimates of differences in DNMT1 expression between the control and MASLD groups from public RNA expression datasets. (F) Forest plot of Pearson’s correlation coefficients between DNMT1 expression and clinical parameters in the MASLD cohort. (G) Scatter plots with linear regression lines and 95% confidence intervals (gray shading) showing the associations between DNMT1 expression and clinical parameters. BMI, body mass index; BP, biological process; CC, cellular component; DNMT1, DNA methyltransferase 1; MASH, metabolic dysfunction–associated steatohepatitis; MASL, metabolic dysfunction–associated steatotic liver; MASLD, metabolic dysfunction–associated steatotic liver disease; MF, molecular function; NAS, nonalcoholic fatty liver disease activity score; TET, ten-eleven translocation.
Figure 2DNMT1 inhibition attenuates lipid accumulation and inflammation in a mouse model of MASLD. (A) Schematic of the MASLD mouse model. (B) Body weight, liver weight, and hepatic index (liver-to-body weight, %) in mice fed a standard diet (SD), a MASLD diet, or a MASLD diet plus Aza-TdC (0.1× or 1×) (n=5–8 mice per group). (C) Representative liver sections subjected to H&E staining (top panel), BODIPY staining (middle panel), and F4/80 immunostaining (bottom panel). Scale bar: 100 μm. (D) Histological scoring for steatosis, lobular inflammation, nonalcoholic fatty liver disease activity score (NAS), and fibrosis. (E–G) Differential lipidomic analysis. (E) Partial least squares discriminant analysis of the lipidomic data. (F) Bubble plot of log2 fold changes in relative lipid abundance. (G) Relative abundance of hepatic triacylglycerol (TAG) species (n=8 mice per group). (H, I) qRT–PCR analysis of genes related to lipid metabolism (H) and inflammation/fibrosis (I) (n=5–7 mice per group). The data are presented as the means±SEMs. Statistical significance was assessed using one-way ANOVA with the Holm–Sidak multiple comparison test (B, D, G, H, I). *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001; ns, not significant. Aza-TdC, 5-aza-4′-thio-2′-deoxycytidine; DNMT1, DNA methyltransferase 1; MASLD, metabolic dysfunction–associated steatotic liver disease.
Figure 3Hypermethylated differentially methylated positions (DMPs) in MASLD mice and patients are enriched for HNF4α and PPARα binding sites. (A) Heatmap of DMPs that were hypermethylated in response to the MASLD diet and reversed by Aza-TdC (1×) treatment (C1; 1,665 CpGs) and hypomethylated in response to the MASLD diet and reversed by Aza-TdC (1×) treatment (C2; 668 CpGs). (B) Gene Ontology analysis of C1 and C2 DMPs using GREAT. (C) Bubble plot of transcription factor (TF) motif enrichment for hypermethylated DMPs in MASLD patients (left panel) and C1 DMPs in mice (right panel). The bubble size reflects the P-value; the color denotes the TF family. (D) Scatter plots showing associations between estimated TF expression and clinical parameters. (E) Pearson’s correlation coefficients between DNMT1 expression and HNF4A or PPARA expression in the MASLD cohort. Blue, control subjects; red, MASLD patients. Aza-TdC, 5-aza-4′-thio-2′-deoxycytidine; BP, biological process; CC, cellular component; DNMT1, DNA methyltransferase 1; HNF4α, hepatocyte nuclear factor 4 alpha; MASLD, metabolic dysfunction–associated steatotic liver disease; MF, molecular function; PPARα, peroxisome proliferator-activated receptor alpha; SD, standard diet.
Figure 4Reactivation of HNF4α- and PPARα-regulated genes by Aza-TdC treatment. (A) Scatter plots showing changes in the expression and DNA methylation of HNF4α (top panel) and PPARα (bottom panel) target genes in mice fed a MASLD diet versus a standard diet (SD, left panel) and Aza-TdC treatment versus a MASLD diet (right panel). (B) Pyrosequencing analysis of Slc38a3, Pck1, Il6ra, and Prlr expression (n=8–13 mice per group). (C) qRT–PCR analysis of Slc38a3, Pck1, Il6ra, and Prlr expression (n=8–13 mice per group). (D) snRNA-seq analysis showing the expression of Pck1, Il6ra, Prlr, and Slc38a3 in hepatocytes from mice fed an SD, an MASLD, or an MASLD with Aza-TdC. Student’s two-tailed t-test. ****P<0.0001. (E) Pearson’s correlation coefficients for the human MASLD cohort: DNMT1 expression with SLC38A3, PCK1, IL6R, and PRLR (top panel); HNF4A expression with the same genes (middle panel); and PPARA expression with the same genes (bottom panel). Gray, control subjects; black, MASLD patients. (F) Heatmap of the scaled expression of the four genes in the human MASLD cohort. (G) Heatmap of the normalized methylation levels of DMPs overlapping these genes. The data are shown as the means±SEMs. Statistical analysis was performed using one-way ANOVA with the Holm–Sidak multiple comparisons test (B, C). *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001; ns, not significant. Aza-TdC, 5-aza-4′-thio-2′-deoxycytidine; BMI, body mass index; DMPs, differentially methylated positions; DNMT1, DNA methyltransferase 1; HNF4α, hepatocyte nuclear factor 4 alpha; MASH, metabolic dysfunction–associated steatohepatitis; MASL, metabolic dysfunction–associated steatotic liver; MASLD, metabolic dysfunction–associated steatotic liver disease; NAS, nonalcoholic fatty liver disease activity score; PPARα, peroxisome proliferatoractivated receptor alpha; snRNA-seq, single-nucleus RNA sequencing.
Figure 5Aza-TdC treatment restores HNF4α binding to regulatory regions and PCK1 signaling. (A) Chromatin-state enrichment analysis showing the enrichment of chromatin states (left panel) and TF ChIP-seq peaks (right panel) in C1 and C2 DMPs. The ChIP-seq data were sourced from the ReMap database. Chromatin states include the following: promoters (Tss and TssFlnk), transcription-related regions (Tx and TxWk), enhancers (Enh, EnhLo, EnhPois, EnhPr, and EnhG), bivalent TSSs (TssBiv), repressive domains (ReprPC and ReprPCWk, enriched in H3K27me3), heterochromatin (Het, enriched in H3K9me3), and quiescent states (QuiesG and Quies1–4). (B) Heatmap of total ChIP-seq peaks (n=6,995) detected in the mouse liver. (C) Dot plot of common differential Hnf4α binding peaks (*FDR<0.05). (D) Random effects meta-analysis of PCK1 expression in publicly available transcriptomic datasets. (E) Genome browser tracks of Pck1 showing ChIP-seq peaks for Pparα (cyan) and Hnf4α (red) in wild-type (WT) and knockout (KO) mice (Pparα: GSE61817; Hnf4α: GSE90533) and Hnf4α binding in the livers of MASLD diet-fed mice versus Aza-TdC-treated mice (blue). The x-axis indicates the genomic coordinates of Pck1; the y-axis indicates the ChIP-seq signal intensity. (F) ChIP–PCR validation of Hnf4α binding at the Pck1 promoter in mice fed a standard diet (SD), a MASLD diet, or a MASLD diet plus Aza-TdC (1×). Region 1 corresponds to –436 to –233 bp upstream of the Pck1 TSS (left panel), and Region 2 spans –28 to +140 bp relative to the Pck1 TSS (right panel). Aza-TdC, 5-aza-4′-thio-2′-deoxycytidine; ChIP-seq, chromatin immunoprecipitation sequencing; DMPs, differentially methylated positions; HNF4α, hepatocyte nuclear factor 4 alpha; MASLD, metabolic dysfunction–associated steatotic liver disease; PCK1, phosphoenolpyruvate carboxykinase 1; PPARα, peroxisome proliferator-activated receptor alpha; TF, transcription factor; TSS, transcription start site.
Figure 6Effects of CRISPRon/off targeting the PCK1 promoter. (A) Schematic illustration of the sgRNA used to selectively modify methylation marks at the PCK1 promoter. (B) Pyrosequencing analysis of CRISPRon/off-edited HEK293T cells. (C) qRT–PCR analysis of PCK1 expression following targeted epigenome editing in HEK293T cells. Student’s two-tailed t-test. **P<0.01; ****P<0.0001. (D) Targeted bisulfite amplicon sequencing of CRISPRon/off-edited HEK293T cells. The X-axis represents 109 individual CpG dinucleotides spanning the PCK1 locus (TSS±2.5 kb). The Y-axis indicates the methylation percentage (0–100%) at each CpG. (E) Quantification of intracellular triacylglycerol (TAG) levels (mg/g soluble protein) in HepG2 cells after CRISPRoff-induced PCK1 silencing. (F) qRT–PCR analysis of lipid metabolism-related genes in HepG2 cell s following CRISPRoff-mediated repression of PCK1. The data are presented as the means±SDs. Statistical significance in (E, F) was assessed using Student’s two-tailed t-test. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. (G) Heatmap of the scaled expression of PCK1 signaling-related genes in MASLD patients (n=131). (H) qRT–PCR analysis of PCK1 signaling-related genes in the livers of MASLD mice. The data are presented as the means±SEMs. One-way ANOVA with the Holm–Sidak multiple comparisons test was used to analyze the data. *P<0.05; **P<0.01; ***P<0.001. MASH, metabolic dysfunction–associated steatohepatitis; MASL, metabolic dysfunction–associated steatotic liver; MASLD, metabolic dysfunction–associated steatotic liver disease; NAS, nonalcoholic fatty liver disease activity score; PCK1, phosphoenolpyruvate carboxykinase 1; sgRNA, single guide RNA.
Abbreviations
5-aza-4′-thio-2′-deoxycytidine
chromatin immunoprecipitation-sequencing
differentially methylated position
hepatocyte nuclear factor 4 alpha
homeostatic model assessment for insulin resistance
interleukin 6 receptor, alpha
metabolic dysfunction–associated steatohepatitis
metabolic dysfunction–associated steatotic liver
metabolic dysfunction–associated steatotic liver disease
phosphoenolpyruvate carboxykinase 1
peroxisome proliferator-activated receptor alpha
standard error of the mean
solute carrier family 38 member 3
single-nucleus RNA sequencing
REFERENCES
- 1. Rinella ME, Lazarus JV, Ratziu V, Francque SM, Sanyal AJ, Kanwal F, et al. A multisociety Delphi consensus statement on new fatty liver disease nomenclature. Hepatology 2023;78:1966-1986.
- 2. Loomba R, Friedman SL, Shulman GI. Mechanisms and disease consequences of nonalcoholic fatty liver disease. Cell 2021;184:2537-2564.
- 3. Younossi ZM, Koenig AB, Abdelatif D, Fazel Y, Henry L, Wymer M. Global epidemiology of nonalcoholic fatty liver disease-Meta-analytic assessment of prevalence, incidence, and outcomes. Hepatology 2016;64:73-84.
- 4. Younossi ZM, Kalligeros M, Henry L. Epidemiology of metabolic dysfunction-associated steatotic liver disease. Clin Mol Hepatol 2025;31:S32-S50.
- 5. Hutchison AL, Tavaglione F, Romeo S, Charlton M. Endocrine aspects of metabolic dysfunction-associated steatotic liver disease (MASLD): beyond insulin resistance. J Hepatol 2023;79:1524-1541.
- 6. Tincopa MA, Anstee QM, Loomba R. New and emerging treatments for metabolic dysfunction-associated steatohepatitis. Cell Metab 2024;36:1430.
- 7. Ganesan A, Arimondo PB, Rots MG, Jeronimo C, Berdasco M. The timeline of epigenetic drug discovery: from reality to dreams. Clin Epigenetics 2019;11:174.
- 8. Vilar-Gomez E, Martinez-Perez Y, Calzadilla-Bertot L, Torres-Gonzalez A, Gra-Oramas B, Gonzalez-Fabian L, et al. Weight loss through lifestyle modification significantly reduces features of nonalcoholic steatohepatitis. Gastroenterology 2015;149:367-378e5 quiz e14–e15.
- 9. Brouwers MC, van Greevenbroek MM, Cantor RM. Heritability of nonalcoholic fatty liver disease. Gastroenterology 2009;137:1536.
- 10. Lai Z, Chen J, Ding C, Wong K, Chen X, Pu L, et al. Association of hepatic global DNA methylation and serum onecarbon metabolites with histological severity in patients with NAFLD. Obesity (Silver Spring) 2020;28:197-205.
- 11. Sodum N, Kumar G, Bojja SL, Kumar N, Rao CM. Epigenetics in NAFLD/NASH: targets and therapy. Pharmacol Res 2021;167:105484.
- 12. Smith ZD, Hetzel S, Meissner A. DNA methylation in mammalian development and disease. Nat Rev Genet 2025;26:7-30.
- 13. Greenberg MVC, Bourc’his D. The diverse roles of DNA methylation in mammalian development and disease. Nat Rev Mol Cell Biol 2019;20:590-607.
- 14. Chen H, Nisar MA, Mulla J, Li X, Cao K, Lu S, et al. Liver TET1 promotes metabolic dysfunction-associated steatotic liver disease. EMBO Mol Med 2025;17:1101-1117.
- 15. Yin MT, Guo L. Mitochondrial DNA methylation: state-of-theart in molecular mechanisms and disease implications. J Adv Res 2025;S2090-1232(25)00644-7.
- 16. Gomez-Alonso MDC, Kretschmer A, Wilson R, Pfeiffer L, Karhunen V, Seppälä I, et al. DNA methylation and lipid metabolism: an EWAS of 226 metabolic measures. Clin Epigenetics 2021;13:7.
- 17. Ahrens M, Ammerpohl O, von Schönfels W, Kolarova J, Bens S, Itzel T, et al. DNA methylation analysis in nonalcoholic fatty liver disease suggests distinct disease-specific and remodeling signatures after bariatric surgery. Cell Metab 2013;18:296-302.
- 18. Murphy SK, Yang H, Moylan CA, Pang H, Dellinger A, Abdelmalek MF, et al. Relationship between methylome and transcriptome in patients with nonalcoholic fatty liver disease. Gastroenterology 2013;145:1076-1087.
- 19. Moran-Salvador E, Garcia-Macia M, Sivaharan A, Sabater L, Zaki MYW, Oakley F, et al. Fibrogenic activity of MECP2 is regulated by phosphorylation in hepatic stellate cells. Gastroenterology 2019;157:1398-1412e9.
- 20. Barcena-Varela M, Paish H, Alvarez L, Uriarte I, Latasa MU, Santamaria E, et al. Epigenetic mechanisms and metabolic reprogramming in fibrogenesis: dual targeting of G9a and DNMT1 for the inhibition of liver fibrosis. Gut 2021;70:388-400.
- 21. Magdy A, Kim HJ, Go H, Lee JM, Sohn HA, Haam K, et al. DNA methylome analysis reveals epigenetic alteration of complement genes in advanced metabolic dysfunction-associated steatotic liver disease. Clin Mol Hepatol 2024;30:824-844.
- 22. Koo BK, Joo SK, Kim D, Bae JM, Park JH, Kim JH, et al. Additive effects of PNPLA3 and TM6SF2 on the histological severity of non-alcoholic fatty liver disease. J Gastroenterol Hepatol 2018;33:1277-1285.
- 23. Koo BK, Kim D, Joo SK, Kim JH, Chang MS, Kim BG, et al. Sarcopenia is an independent risk factor for non-alcoholic steatohepatitis and significant fibrosis. J Hepatol 2017;66:123-131.
- 24. Parker WB, Thottassery JV. 5-Aza-4′-thio-2′-deoxycytidine, a new orally bioavailable nontoxic “Best-in-Class”: DNA methyltransferase 1-depleting agent in clinical development. J Pharmacol Exp Ther 2021;379:211-222.
- 25. Thottassery JV, Sambandam V, Allan PW, Maddry JA, Maxuitenko YY, Tiwari K, et al. Novel DNA methyltransferase-1 (DNMT1) depleting anticancer nucleosides, 4′-thio-2′-deoxycytidine and 5-aza-4′-thio-2′-deoxycytidine. Cancer Chemother Pharmacol 2014;74:291-302.
- 26. Landt SG, Marinov GK, Kundaje A, Kheradpour P, Pauli F, Batzoglou S, et al. ChIP-seq guidelines and practices of the ENCODE and modENCODE consortia. Genome Res 2012;22:1813-1831.
- 27. Yoo T, Joo SK, Kim HJ, Kim HY, Sim H, Lee J, et al. Disease-specific eQTL screening reveals an anti-fibrotic effect of AGXT2 in non-alcoholic fatty liver disease. J Hepatol 2021;75:514-523.
- 28. Govaere O, Cockell S, Tiniakos D, Queen R, Younes R, Vacca M, et al. Transcriptomic profiling across the nonalcoholic fatty liver disease spectrum reveals gene signatures for steatohepatitis and fibrosis. Sci Transl Med 2020;12:eaba4448.
- 29. Boeckmans J, Buyl K, Natale A, Vandenbempt V, Branson S, De Boe V, et al. Transcriptomics data of a human in vitro model of non-alcoholic steatohepatitis exposed to elafibranor. Data Brief 2019;25:104093.
- 30. Xu Y, Zhu Y, Hu S, Xu Y, Stroup D, Pan X, et al. Hepatocyte nuclear factor 4α prevents the steatosis-to-NASH progression by regulating p53 and bile acid signaling (in mice). Hepatology 2021;73:2251-2265.
- 31. Montagner A, Polizzi A, Fouché E, Ducheix S, Lippi Y, Lasserre F, et al. Liver PPARα is crucial for whole-body fatty acid homeostasis and is protective against NAFLD. Gut 2016;65:1202-1214.
- 32. Ernst J, Kellis M. ChromHMM: automating chromatin-state discovery and characterization. Nat Methods 2012;9:215-216.
- 33. Wang Z, Dong C. Gluconeogenesis in cancer: function and regulation of PEPCK, FBPase, and G6Pase. Trends Cancer 2019;5:30-45.
- 34. Ye Q, Liu Y, Zhang G, Deng H, Wang X, Tuo L, et al. Deficiency of gluconeogenic enzyme PCK1 promotes metabolic-associated fatty liver disease through PI3K/AKT/PDGF axis activation in male mice. Nat Commun 2023;14:1402.
- 35. Nuñez JK, Chen J, Pommier GC, Cogan JZ, Replogle JM, Adriaens C, et al. Genome-wide programmable transcriptional memory by CRISPR-based epigenome editing. Cell 2021;184:2503-2519e17.
- 36. Gallage S, Ali A, Barragan Avila JE, Seymen N, Ramadori P, Joerke V, et al. A 5:2 intermittent fasting regimen ameliorates NASH and fibrosis and blunts HCC development via hepatic PPARα and PCK1. Cell Metab 2024;36:1371-1393e7.
- 37. Qin Y, Grimm SA, Roberts JD, Chrysovergis K, Wade PA. Alterations in promoter interaction landscape and transcriptional network underlying metabolic adaptation to diet. Nat Commun 2020;11:962.
- 38. Wang S, Zha L, Cui X, Yeh YT, Liu R, Jing J, et al. Epigenetic regulation of hepatic lipid metabolism by DNA methylation. Adv Sci (Weinh) 2023;10:e2206068.
- 39. Kim YC, Seok S, Zhang Y, Ma J, Kong B, Guo G, et al. Intestinal FGF15/19 physiologically repress hepatic lipogenesis in the late fed-state by activating SHP and DNMT3A. Nat Commun 2020;11:5969.
- 40. Laranjeira ABA, Hollingshead MG, Nguyen D, Kinders RJ, Doroshow JH, Yang SX, et al. DNA damage, demethylation and anticancer activity of DNA methyltransferase (DNMT) inhibitors. Sci Rep 2023;13:5964.
- 41. Puri P, Wiest MM, Cheung O, Mirshahi F, Sargeant C, Min HK, et al. The plasma lipidomic signature of nonalcoholic steatohepatitis. Hepatology 2009;50:1827-1838.
- 42. Ioannou GN. The role of cholesterol in the pathogenesis of NASH. Trends Endocrinol Metab 2016;27:84-95.
- 43. Musso G, Saba F, Cassader M, Gambino R. Lipidomics in pathogenesis, progression and treatment of nonalcoholic steatohepatitis (NASH): recent advances. Prog Lipid Res 2023;91:101238.
- 44. Wang X, Cao Q, Yu L, Shi H, Xue B, Shi H. Epigenetic regulation of macrophage polarization and inflammation by DNA methylation in obesity. JCI Insight 2016;1:e87748.
- 45. Tambay V, Raymond VA, Voisin L, Meloche S, Bilodeau M. Reprogramming of glutamine amino acid transporters expression and prognostic significance in hepatocellular carcinoma. Int J Mol Sci 2024;25:7558.
- 46. Rubio-Aliaga I, Wagner CA. Regulation and function of the SLC38A3/SNAT3 glutamine transporter. Channels (Austin) 2016;10:440-452.
- 47. Gu S, Roderick HL, Camacho P, Jiang JX. Identification and characterization of an amino acid transporter expressed differentially in liver. Proc Natl Acad Sci U S A 2000;97:3230-3235.
- 48. Dogru T, Ercin CN, Erdem G, Sonmez A, Tapan S, Tasci I. Increased hepatic and circulating interleukin-6 levels in human nonalcoholic steatohepatitis. Am J Gastroenterol 2008;103:3217-3218.
- 49. Hunter CA, Jones SA. IL-6 as a keystone cytokine in health and disease. Nat Immunol 2015;16:448-457.
- 50. Zhang P, Ge Z, Wang H, Feng W, Sun X, Chu X, et al. Prolactin improves hepatic steatosis via CD36 pathway. J Hepatol 2018;68:1247-1255.
, Won Kim15
