Tcf7l2 in hepatocytes regulates de novo lipogenesis in diet-induced non-alcoholic fatty liver disease in mice

Human liver tissue samples in ESM Fig. 1a were collected from 32 individuals who underwent hepatectomy or cholecystectomy at the university-affiliated Severance Hospital, Yonsei University College of Medicine, Republic of Korea, between September 2014 and October 2019, as previously described [20]. Liver histology was assessed by an experienced pathologist and, according to diagnostic criteria [21], the samples were put into the following groups: (1) normal (n=13); (2) simple steatosis (NAFLD; n=10); and (3) NASH (n=9). All individuals gave informed consent and the study protocol was approved by the Institutional Review Board at Severance Hospital (IRB No 4–2014–0674).

Seven-week-old male C57BL/6N mice were purchased from ORIENT BIO (ORIENT BIO, Republic of Korea). Hepatocyte-selective Tcf7l2 ablation in C57BL/6N mice (liver-specific Tcf7l2 KO mice: Alb-Cre;Tcf7l2^f/f) was obtained by crossing mice carrying Cre recombinase driven by the albumin (Alb) promoter with mice carrying Tcf7l2 alleles floxed at exon 5. Six-week-old male Alb-Cre;Tcf7l2^f/f and their wild-type littermates (Tcf7l2^f/f mice; developed in-house from Tcf7l2^+/− mice [Tcf7l2^{tm2a(EUCOMM)Wtsi}; the EUCOMM Consortium; www.mousephenotype.org/data/genes/MGI:1202879↗, accessed 8 September 2022]) were given either an HFD (catalogue no. D12492; Research Diet, USA) or HCD (catalogue no. TD.98090; Envigo, USA) for the indicated periods. For GTTs and ITTs, mice were fasted for 16 h or 6 h, respectively, and injected intraperitoneally with 1.5 g/kg of glucose or 1 U/kg of insulin. Blood glucose levels were measured in samples taken from the tail vein (LifeScan; One-Touch, USA). V˙O2, carbon dioxide production (V˙CO2), respiratory exchange ratio (RER; V˙CO2/V˙O2) and energy expenditure were measured in mice at 10 weeks using the Oxylet Pro system (Panlab, Spain). All animal procedures were approved by the Institutional Animal Care and Use Committee of the Korea Research Institute of Bioscience & Biotechnology (KRIBB-AEC-21051) and were performed in accordance with the guidelines for the Care and Use of Laboratory Animals published by the US National Institutes of Health.

Primary hepatocytes were isolated from 8-week-old male C57BL/6N mice by collagenase perfusion [7]. The cells were maintained in Medium 199 (catalogue no. M4530; Sigma-Aldrich, UK) supplemented with 10% (vol./vol.) FBS, 1% (vol./vol.) antibiotics and 10 nmol/l dexamethasone, and were infected with adenovirus for 24 h. Subsequently, the cells were maintained in medium 199 without 10% FBS and were treated with 25 mmol/l glucose (catalogue no. G7021; Sigma-Aldrich) and/or 100 nmol/l insulin (catalogue no. I6634; Sigma-Aldrich) for 48 h, 250 μmol/l palmitic acid (catalogue no. P9767; Sigma-Aldrich) for 24 h, 300 μmol/l AICAR (catalogue no. A611700; Toronto Research Chemicals, Canada) for 1 h, or 1 μmol/l T0901317 (catalogue no. T2320; Sigma-Aldrich) for 24 h.

Total RNA from human liver tissues, mouse tissues or mouse primary hepatocytes was extracted using the easy-spin Total RNA Extraction Kit (catalogue no. 17221; iNtRON Biotechnology, Republic of Korea). Quantitative PCR (qPCR) was performed using the SYBR green PCR kit in a C1000 Touch Thermal Cycler (Bio-Rad Laboratories, USA). All data were normalised to the expression of ribosomal L32. Primer sequences are listed in ESM Table 1. miRNA (miR) expression was analysed as previously described [22] and the miR primer sequences are listed in ESM Table 2.

Western blot analysis of extracts from mouse tissues and whole cells was performed as previously described [23]. Primary and secondary antibodies were used according to manufacturer’s instructions and are listed in ESM Table 3.

Plasma insulin and IGF-1 levels were measured with a Mouse Insulin ELISA Kit (catalogue no. 80-INSMSE01; ALPCO, USA) and Mouse IGF1 ELISA Kit (catalogue no. EMIGF1; Thermo Fisher Scientific), respectively. Total lipids from mouse livers or mouse primary hepatocytes were extracted using the Folch method as described previously [7]. Hepatic TG, cellular TG, plasma TG and NEFA levels were measured with colorimetric assay kits (TG: catalogue no. 461-09092; NEFA: catalogue no. 438-91691; Wako, Japan). Hepatic glycogen was measured using an EnzyChrom Glycogen Assay Kit (catalogue no. E2GN-100; BioAssay Systems, USA). Hepatic β-hydroxybutyrate (β-OH) was measured with a fluorometric assay kit (catalogue no. 700740; Cayman Chemical, USA).

Mouse pancreas tissues in ESM Fig. 1e and ESM Fig. 2d were fixed with 4% (wt/vol.) paraformaldehyde and were embedded in paraffin. Sliced specimens were immunostained using anti-insulin (catalogue no. 53-9769-82; Thermo Fisher Scientific) and anti-glucagon (catalogue no. ab92517; Abcam, USA) antibodies with DAPI (catalogue no. H-1800; Vector Laboratories, USA). The fluorescence images were taken using an OLYMPUS IX73 microscope (Olympus, Japan). Insulin-positive areas (beta cells), glucagon-positive areas (alpha cells) and pancreatic areas were quantified using Image J software (v 1.50i; https://imagej.nih.gov/ij/download.html↗). Antibody information is listed in ESM Table 4.

Liver tissue sections were stained with Oil Red O (catalogue no. O0625; Sigma-Aldrich) and H&E, as previously described [24]. Liver fibrosis was detected with Sirius red staining (catalogue no. ab150681; Abcam, USA) according to the manufacturer’s instructions. NAFLD Activity Score (NAS) and fibrosis stage were calculated as the sum of the scores for steatosis (0–3), lobular inflammation (0–3), hepatocyte ballooning (0–2), as assessed by H&E staining, and fibrosis (0–4), assessed by Sirius red staining [21, 25]. Histological samples were anonymised and assigned a number before the slides were analysed by an experienced pathologist.

Mouse primary hepatocytes isolated from Tcf7l2^f/f or C57BL/6N mice were transfected with the indicated plasmid DNA vectors using Lipofectamine 3000 transfection reagent (catalogue no. L3000015; Invitrogen, USA). Fatty acid uptake in mouse primary hepatocytes was measured with the Fatty Acid Uptake Assay Kit (catalogue no. K408-100; Biovision, USA) according to the manufacturer’s instructions. Fluorescent fatty acid probes (green) were detected by a fluorescence microscope (DM IL LED FLUO; Leica Microsystems, Germany) and fluorescence intensity was quantified using Image J software (v 1.50i).

Mouse primary hepatocytes isolated from Tcf7l2^f/f mice were infected with adenoviruses expressing gfp only and Cre and treated with 1 nmol/l insulin for 30 min. 2-Deoxyglucose (2-DG) levels in mouse primary hepatocytes were measured using the 2-DG Uptake Measurement Kit (catalogue no. CSR-OKP-PMG-K01; Cosmo Bio, Japan) according to the manufacturer’s instructions.

Fatty acid oxidation in mouse primary hepatocytes was assessed by analysing oxygen consumption rate (OCR) with an XF24 extracellular flux analyser (Seahorse Bioscience, USA). Mouse primary hepatocytes were seeded on collagen-coated XF24 cell culture microplates. Cells were treated with BSA or 250 μmol/l palmitate and/or 100 μmol/l etomoxir (catalogue no. E1905; Sigma-Aldrich), and then stimulated with 2.5 μmol/l oligomycin (catalogue no. O4876; Sigma-Aldrich), 10 μmol/l fluoro-carbonylcyanide phenylhydrazone (FCCP; catalogue no. C2920; Sigma-Aldrich), and 2 μmol/l rotenone (catalogue no. R8875; Sigma-Aldrich) plus 5 μmol/l antimycin A (catalogue no. A8674; Sigma-Aldrich), according to the manufacturer’s instructions. OCR levels were normalised to the amount of protein in each sample.

For in vitro DNL analysis, mouse primary hepatocytes were treated with 37 kBq of ¹⁴C-labelled glucose (catalogue no. NEC042; PerkinElmer, USA) and 10 nmol/l insulin for 48 h. Lipids were extracted from cells with chloroform and ¹⁴C radioactivity was measured using the Tri-Carb 2910 TR liquid scintillation analyser (PerkinElmer). For in vivo DNL analysis, mice were fasted for 24 h, then refed an HCD for 12 h. Subsequently, they were intraperitoneally injected with ¹⁴C-labelled sodium acetate (555 kBq/mouse; catalogue no. NEC553; PerkinElmer), or 3% (vol./vol.) ethanol in saline (136.9 mmol/l NaCl) as a control, and euthanised 1 h post injection. Lipids were extracted using the Folch method [7] and incorporation rates of [1-¹⁴C]acetic acid into lipids were measured using the Tri-Carb 2910 TR liquid scintillation analyser.

Total RNA from mouse liver was extracted using the easy-spin Total RNA Extraction Kit (iNtRON Biotechnology). The library preparation, mRNA sequencing and analysis were performed by eBiogen (Republic of Korea). Differentially expressed genes were determined based on counts from unique and multiple alignments using coverage in Bedtools (v 2.25.0; https://bedtools.readthedocs.io/en/latest/index.html↗). The read count (RC) data were processed based on the quantile normalisation method using EdgeR (v 3.20.1; https://bioconductor.org/packages/release/bioc/html/edgeR.html↗) within R (v 3.4.4; www.r-project.org↗) using Bioconductor (v 3.6; https://bioconductor.org/install/↗). From the results obtained by setting the RC (log₂) value to 5 or more, genes with a log₂ fold change greater than 1.5 (log₂ fold-change cut-off >1.5) in KO vs wild-type comparisons were selected and analysed using Gene Set Enrichment Analysis (GSEA; www.gsea-msigdb.org/gsea/msigdb/mouse/annotate.jsp↗, accessed 1 November 2021), based on canonical pathways gene sets derived from the Reactome pathway database (www.gsea-msigdb.org/gsea/msigdb/mouse/genesets.jsp?collection=CP:REACTOME↗, accessed 1 November 2021).

Expression vectors for mouse Tcf7l2, nuclear Srebf1c and Mlxipl(α) were amplified by RT-PCR using liver RNA derived from C57BL/6N mice and inserted into pcDNA3-Flag or pcDNA3-HA expression vectors. The pGL4-6X SRE-luc (containing six copies of SRE) and pGL4-4X ChoRE-luc (containing four copies of ChoRE) constructs were a kind gift from S.-H. Koo (Korea University, Korea). Mouse Fasn (−2000/+200), Pklr (−191/+200), Srebf1c (−1195/+77), miR-33-5p (−852/+16), miR-33-5p (−1412/+16), Mlxipl(α) (−2538/+124) and Mlxipl(β) (−536/+352) promoter sequences were amplified by PCR using mouse genomic DNA and inserted into the pGL4-luciferase reporter vector. miRNA mimics (mmu-miR-33-5p, mmu-miR-132-3p, mmu-miR-212-3p and negative control) were purchased from GenePharma (Shanghai, China). Primer sequences are listed in ESM Table 5. In total, 50 μmol/l miRNA mimics were transfected into cells using Lipofectamine 3000 transfection reagent (Invitrogen), according to the manufacturer’s instructions. Adenoviruses expressing gfp only, Tcf7l2 and Cre have been described previously [7]. For animal experiments, the viruses were purified on a CsCl gradient, dialysed against PBS buffer containing 10% (vol./vol.) glycerol and stored at −80°C.

Luciferase assays were performed to determine the effects of TCF7L2 on the sterol regulatory element binding protein 1 (SREBP1)c, ChREBP/max-like protein X (MLX) transcriptional activities and the miR-33-5p promoter activity in HEPG2 cells (catalogue no. HB-8065; ATCC, USA) and Srebf1 3′UTR promoter activity in HEK293T cells (catalogue no. CRL-3216, ATCC). HEPG2 and HEK293T cells were tested for mycoplasma using the BioMycoX Mycoplasma PCR Detection Kit (catalogue no. D-50; CellSafe, Republic of Korea) and were found to be mycoplasma negative. Promoter activity was measured using a luciferase reporter assay system (catalogue no. E1910; Promega, USA) and normalised to β-galactosidase activity.

HEPG2 cells were transfected with HA-tagged TCF7L2 and Flag-tagged ChREBPα for 48 h and treated with 10 μmol/l MG132 (catalogue no. M1157; AG Scientific, USA) or DMSO (catalogue no. D2650; Sigma-Aldrich) for 3 h.

Total lysates from liver tissues were centrifuged at 16,000 × g for 10 min and the supernatant was extracted. For immunoprecipitation, 3 mg of protein lysates were incubated with carbohydrate response element binding protein (ChREBP) antibody and 30 μl of protein G plus A agarose beads (catalogue no. IP05; Millipore, USA) was added to each sample. For wheat germ agglutinin (WGA) precipitation, 3 mg of protein lysates were incubated with 30 μl of WGA agarose beads (catalogue no. AL-1023; Vector Laboratories), after which beads were eluted in 2× sample buffer without β-mercaptoethanol.

To generate Tcf7l2-KO alpha-mouse-liver-12 (AML12) cell lines (catalogue no. CRL-2254; ATCC), we developed a CRISPR/Cas9 system for gene editing using standard methods [26]. The single-guide RNA (sgRNA) sequence was cloned into pSpCas9(BB)-2A-Puro(PX459), which was a gift from F. Zhang (Broad Institute of MIT and Harvard, Cambridge, USA). The target guide RNA (gRNA) sequence was 5′-AGCAATGAACACTTCACCCC-3′ (Tcf7l2 exon 5; Genome Reference Consortium Mouse Reference 39 [GRCm39;]: 19:55,896,921–55,896,940; https://asia.ensembl.org/Mus_musculus/Info/Annotation↗). The vector expressing gRNA was transfected into AML12 cells and then cells were selected using puromycin. Cells tested negative for mycoplasma contamination. Genomic DNA was extracted from cells using the Exgene Tissue SV Kit (catalogue no. 104-152; GeneAll, Republic of Korea) Gene knockout was verified by PCR and western blot analysis. Cells were transfected with the pGL4-Pklr (−191/+200) promoter and then co-treated with 25 mmol/l glucose and 40 μmol/l OSMI-1 (catalogue no. ab235455; Abcam, UK) for 24 h.

Cross-linking, nuclear isolation and chromatin immunoprecipitation (ChIP) assays were performed on AML12 cell line samples and mouse primary hepatocyte samples, using previously reported methods [27]. The precipitated DNA fragments were analysed by PCR.

Statistical differences between two experimental groups were evaluated by the two-tailed unpaired Student’s t test. One-way ANOVA with Tukey’s multiple comparison test was performed using GraphPad Prism 8.0.1 (GraphPad Software, USA) when comparing three or more groups, as reported in the figure legends. Data are shown as mean±SD or mean±SEM. A p value <0.05 was considered statistically significant.

We found that Tcf7l2 expression was significantly reduced in the liver of individuals with NAFLD and NASH (ESM Fig. 1a). Therefore, to gain new insights into the role of TCF7L2 in NAFLD development, we generated liver-specific Tcf7l2 KO mice (Alb-Cre;Tcf7l2^f/f; Fig. 1a,b and ESM Fig. 1b). Essentially, Alb-Cre;Tcf7l2^f/f mice exhibited increased fasting glucose levels without changes in body weight or plasma insulin and IGF-1 levels (Fig. 1c–e and ESM Fig. 1c). They also displayed impaired glucose and insulin tolerance (Fig. 1f,h and ESM Fig. 1d). However, there were no changes in insulin concentration patterns during GTT (Fig. 1g) or in islet glucagon-positive alpha cell and insulin-positive beta cell areas (ESM Fig. 1e,f). Additionally, hepatic Tcf7l2 deficiency did not lead to changes in hepatic insulin signalling, RER, V˙O2, V˙CO2 and energy expenditure, as compared with wild-type controls (Fig. 1i,j and ESM Fig. 1g). Although there were no significant changes in liver TG content (Fig. 1k), hepatic Tcf7l2 depletion significantly enhanced hepatic expression of lipogenic genes, without altering genes involved in β-oxidation and lipolysis (Fig. 1l). Additionally, there were no changes in Igf1 and Igf1R mRNA expression (ESM Fig. 1h).

In rodents, studies have shown that feeding with an HFD or HCD for 22 weeks leads to the development of NAFLD, including non-alcoholic fatty liver (NAFL; simple steatosis) and NASH [28–30]. As with individuals with NAFLD, Tcf7l2 expression was markedly decreased to below-normal levels in the liver of C57BL/6N mice fed an HFD or HCD for 22 weeks (Fig. 2a,b). To assess the role of TCF7L2 in the development of NAFLD, Alb-Cre;Tcf7l2^f/f mice were fed an HFD or HCD for 22 weeks (Fig. 2c). Although hepatic TCF7L2 expression was decreased in the liver of wild-type Tcf7l2^f/f mice by both HFD and HCD feeding for 22 weeks, hepatic TCF7L2 expression in wild-type Tcf7l2^f/f mice was clearly higher compared with that in Alb-Cre;Tcf7l2^f/f mice (ESM Fig. 2a). Under HCD conditions, hepatic Tcf7l2 deficiency promoted lipid droplet formation and TG accumulation in the liver, resulting in elevated alanine aminotransferase (ALT) levels (Fig. 2c–e). In contrast, these differences were not observed when mice were fed an HFD (Fig. 2c–e). Consistently, genes responsible for lipogenesis and inflammation in NAFLD/NASH were found to be significantly increased in the liver of Alb-Cre;Tcf7l2^f/f mice, as compared with wild-type mice, following HCD feeding, whilst elevation of these genes was less consistently observed following HFD feeding (ESM Fig. 2b,c). However, there were no significant changes in islet size or islet glucagon-positive alpha cell and insulin-positive beta cell areas in mice fed an HFD or HCD (ESM Fig. 2d-g). Representative liver images of Alb-Cre;Tcf7l2^f/f mice fed an HCD for 22 weeks are shown in Fig. 2f. In response to 22 weeks of HCD, hepatic Tcf7l2 deficiency caused impairment of glucose and insulin tolerance (Fig. 2g,h). This impeded hepatic insulin signalling, as confirmed by the reduced level of Akt phosphorylation on Ser473 in the liver of Alb-Cre;Tcf7l2^f/f mice vs wild-type mice (Fig. 2i). However, it did not affect Akt phosphorylation levels in epididymal white adipose tissue and skeletal muscle (ESM Fig. 2h).

Compared with an HCD, an HFD did not have as marked an effect on NAFLD development in hepatic Tcf7l2-deficient mice. To investigate whether this could have been owing to saturation of liver lipid content under an HFD diet, we used Alb-Cre;Tcf7l2^f/f mice fed an HFD for 4 or 12 weeks. Hepatic Tcf7l2 depletion exacerbated HFD-induced glucose intolerance and insulin intolerance without causing changes in body weight (ESM Fig. 3a–c). However, it did not alter hepatic TG content or plasma TG and NEFA levels (ESM Fig. 3d–g). To specifically address why hepatic Tcf7l2 did not affect HFD-induced TG content, we analysed the expression of genes involved in various lipid-related metabolic pathways (ESM Fig. 3h). Notably, hepatic Tcf7l2 deficiency increased the expression of several lipogenic genes (Srebf1c [encoded by the Srebf1 gene], Fasn and Acaca), whereas it decreased the expression of several fatty acid transporter genes (Cd36, Slc27a1, Slc27a4 and Fabp3; ESM Fig. 3h). In contrast, adenovirus-mediated expression of hepatic Tcf7l2 inversely regulated many of these genes (Srebf1c, Fasn, Cd36, Slc27a4 and Fabp3) without altering HFD-induced hepatic TG content (ESM Fig. 3i,j). Further, in a cell-autonomous manner, Tcf7l2 KO via Cre deletion of floxed sequences in primary hepatocytes impeded insulin-induced fatty acid uptake, as confirmed by use of a fluorescent fatty acid probe (ESM Fig. 4a), whereas forced expression of Tcf7l2 in C57BL/6N primary mouse hepatocytes promoted it (ESM Fig. 4h). It is well known that HFD suppresses DNL [31]. Accordingly, fatty acids can inhibit insulin-stimulated glucose uptake [32, 33]. Therefore, we hypothesised that hepatic Tcf7l2 deficiency did not affect HFD-induced liver lipid content, despite there being a decrease in fatty acid uptake, because lipogenesis is balanced by a relative increase in glucose-sensing in response to reduced fatty acid uptake. Indeed, Tcf7l2 KO in primary hepatocytes impaired fatty acid-induced inhibition of glucose uptake (ESM Fig. 4b). Expression of L-type pyruvate kinase (L-PK; encoded by the pklr gene) and Pklr, a glucose-sensing marker, increased in the liver of Alb-Cre;Tcf7l2^f/f mice vs wild-type mice, even under HFD conditions (ESM Fig. 4c,d).

Fatty acid β-oxidation, another lipid-associated metabolic pathway, was also altered by hepatic Tcf7l2 modulation under HFD conditions (ESM Fig. 3h,j). The AMP-activated protein kinase (AMPK)/acetyl-CoA carboxylase (ACC)/carnitine palmitoyltransferase 1α (CPT1α) axis is a key pathway for regulating fatty acid oxidation [34]. However, neither KO nor overexpression of Tcf7l2 functionally affected fatty acid β-oxidation, as determined by changes in OCR following treatment with fatty acid and/or etomoxir, an irreversible inhibitor of CPT1α (ESM Fig. 4e,i). Further, Tcf7l2 KO or overexpression did not affect AMPK phosphorylation levels following treatment with 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR), a selective activator of AMPK, under HFD conditions (ESM Fig. 4f,g,j,k).

HFD provides excessive exogenous fatty acids, whereas HCD and refeeding mechanistically stimulate endogenous fatty acid synthesis from glucose by postprandial blood glucose and glucose-responsive insulin action, which are increased following carbohydrates consumption. Therefore, we hypothesised that the difference in effects of TCF7L2 on HCD- and HFD-induced fatty liver might originate from differential responses to exogenous fatty acid and glucose/insulin pools. Hepatic TCF7L2 expression decreased in primary hepatocytes treated with palmitic acid, the most common saturated fatty acid in the human body, and in the liver of HFD-fed mice (Fig. 3a). Conversely, hepatic TCF7L2 expression significantly increased following in vitro treatment with glucose/insulin and following in vivo refeeding (Fig. 3b and ESM Fig. 5a–c). To further investigate the relationship between TCF7L2 expression and carbohydrate sensing, we measured hepatic TCF7L2 expression at the same time as measuring the expression of lipogenic factors over varying HCD exposure time. Hepatic TCF7L2 expression increased after 2 and 4 weeks of HCD feeding vs baseline but was downregulated after 8 weeks of HCD feeding (Fig. 3c and ESM Fig. 5d,e). However, expression of the lipogenic factors ACC (encoded by Acaca) and fatty acid synthase (FAS; encoded by Fasn), and hepatic TG levels increased with HCD feeding (Fig. 3c,d and ESM Fig. 5d,e). Eventually, as shown in Fig. 2a, hepatic TCF7L2 expression decreased to below-normal levels after 22 weeks of HCD feeding, at which point excessive intrahepatic fatty acid was expected.

Hence, we hypothesised that although excessive increases in intrahepatic fatty acid as a result of long-term HCD feeding can eventually reduce hepatic TCF7L2 expression and activity, hepatic TCF7L2 restrains endogenous fatty acid synthesis to maintain lipid homeostasis (Fig. 3e). Therefore, we investigated whether TCF7L2 could regulate hepatic DNL in a cell-autonomous fashion. Adenovirus-mediated Tcf7l2 expression suppressed the expression of glucose-/insulin-stimulated DNL genes involved in glycolysis (Gck and Pklr) and lipogenesis (Acaca and Fasn) and inhibited DNL in primary hepatocytes, as compared with control cells infected with Ad-gfp (Fig. 3f,g). Conversely, Tcf7l2^f/f primary hepatocytes infected with adenovirus expressing Cre (Ad-Cre) exhibited increased DNL genes and promoted DNL, as compared with cells infected with Ad-gfp (Fig. 3h,i). Consistent with previous reports, Tcf7l2 deficiency did not significantly affect hepatic insulin signalling pathway in a cell-autonomous manner (ESM Fig. 5f). Further, to investigate Tcf7l2 functional knockdown whilst avoiding potential redundancy of other TCF protein members, we generated a TCF7L2 dominant-negative (DN) mutant (TCF7L2DN) that lacks the N-terminal β-catenin interaction domain. TCF7L2DN inhibited the suppressive effects of TCF7L2 on DNL-associated genes and DNL in primary hepatocytes (Fig. 3j,k).

To further assess the physiological function of hepatic Tcf7l2 deficiency in hepatic DNL, Alb-Cre;Tcf7l2^f/f mice were fasted and then refed. Compared with wild-type mice, this promoted the expression of DNL genes involved in glycolysis (Slc2a2, Gck and Pklr) and lipogenesis (Acly, Acaca and Fasn) following refeeding in a time-dependent manner (Fig. 4a). These genes are shown (in red) in Fig. 4b, which depicts the DNL process. However, there were no changes in the expression of genes involved in lipolysis and β-oxidation upon refeeding (ESM Fig. 6). Importantly, loss of hepatic Tcf7l2 led to increased liver TG content after 24 h of refeeding, but not levels of hepatic glycogen and β-OH, which are glucose products that are dispersed into and metabolised by other pathways during the DNL process (Fig. 4c–e). Additionally, there were no changes in plasma TG and NEFA levels (Fig. 4f,g). These data suggest that the activation of the glycolytic/lipogenic pathway for DNL is the primary factor that affects fat deposition in the liver of Alb-Cre;Tcf7l2^f/f mice.

Dietary carbohydrates are the primary stimulus for hepatic DNL. To further elucidate the carbohydrate dependency of Tcf7l2 deficiency-induced hepatic TG accumulation, we compared normal chow refeeding with high-carbohydrate refeeding (HCR) in Alb-Cre;Tcf7l2^f/f mice. Hepatic Tcf7l2 deficiency promoted expression of glycolytic and lipogenic factors and hepatic TG accumulation, which was more evident under HCR conditions compared with normal chow refeeding conditions (Fig. 4h–j). A high-carbohydrate intake markedly enhanced hepatic DNL induced by hepatic Tcf7l2 depletion (Fig. 4k). These data suggest that hepatic Tcf7l2 deficiency can aggravate liver steatosis and that the extent to which this occurs is dependent on carbohydrate intake, owing to increased hepatic DNL with carbohydrate loading.

To investigate the major pathways and key instigators upregulated in the liver of Alb-Cre;Tcf7l2^f/f mice, we analysed mRNA-sequencing data from liver samples from refed mice. In total, 526 genes were upregulated (log₂ fold-change cut-off >1.5) and were analysed using GSEA (www.gsea-msigdb.org/gsea/msigdb/mouse/annotate.jsp↗, accessed on 1 November 2021), based on canonical pathways gene sets derived from the Reactome pathway database (www.gsea-msigdb.org/gsea/msigdb/mouse/genesets.jsp?collection=CP:REACTOME↗, accessed on 1 November 2021). The lipid metabolism-related gene set was most significantly upregulated (its significance can be more intuitively and integratedly presented as a radial line graph; see Fig. 5a and ESM Table 6). Specifically, the SREBP- and ChREBP-related pathways were identified as major signalling pathways (Fig. 5a). Careful investigation of mRNA-sequencing data showed that specific target genes and commonly shared lipogenic target genes of SREBP1c and ChREBP increased in the liver of Alb-Cre;Tcf7l2^f/f mice (Fig. 5b and ESM Table 7). However, despite a reduction in Srebf2 mRNA expression in the liver of Alb-Cre;Tcf7l2^f/f vs wild-type mice, the expression of its target cholesterogenic genes was not altered (ESM Fig. 7a,b and ESM Table 8). SREBP1c and ChREBP are the main transcription factors for DNL [35] that regulate target gene transcription by binding to the sterol regulatory element (SRE) and carbohydrate response element (ChoRE) within their target gene promoters, respectively [36]. In the HEPG2 cell line, transfection with TCF7L2 inhibited ChREBP/MLX-induced activation of Pklr promoters containing ChoRE as well as ChREBP/MLX-induced activation of the ChoRE promoter itself (Fig. 5c). Furthermore, TCF7L2 inhibited SREBP1c-induced activation of an Fasn promoter region containing the SRE only as well as SREBP1c-induced activation of the SRE promoter itself (Fig. 5d). Liver X receptors (LXRs) are nuclear receptors that can directly regulate both SREBP1c and ChREBP [35]. However, neither KO nor overexpression of Tcf7l2 in primary hepatocytes affected lipogenic gene expression induced by the LXR agonist T0901317 (ESM Fig. 7c,d). These data suggest that TCF7L2 regulates ChREBP and SREBP1c transcriptional activities in an LXR-independent manner.

Among the two ChREBP isoforms (ChREBPα and ChREBPβ; both encoded by the Mlxipl gene) [37], the 100 kDa protein band corresponding to ChREBPα [35] was found to be significantly increased in the refed state and after 2, 4 and 8 weeks of HCD, but decreased after 22 weeks of chronic HCD (Fig. 6a,b). However, hepatic Tcf7l2 deficiency resulted in increased protein expression of ChREBPα and its target L-PK (encoded by the pklr gene), not only under acute HCR conditions but also following 22 weeks of chronic HCD (Fig. 6c,d). However, there were no changes in mRNA levels and promoter activity of Mlxipl(α) (Fig. 6e,f). Therefore, we investigated whether TCF7L2 regulates ChREBPα protein levels to modulate ChREBPα transcriptional activity. As shown in Fig. 6g, TCF7L2 inhibited ChREBPα/MLX-induced Mlxipl(β) promoter activity. In a set of experiments using the same concentration ratio of Flag-tagged ChREBPα:HA-tagged TCF7L2 used in the promoter activity assay, ChREBPα protein levels appeared to gradually decrease in a TCF7L2 expression-level-dependent manner, without changes in Mlxipl(α) mRNA expression (Fig. 6h). Further, the TCF7L2-mediated reduction in ChREBPα protein expression was restored by treatment with the proteasome inhibitor MG132 (Fig. 6i). Conversely, ChREBP ubiquitination was found to be significantly decreased in the liver of hepatic Tcf7l2-deficient mice under HCR conditions (Fig. 6j and ESM Fig. 8a). O-GlcNAcylation of ChREBP is induced by high glucose levels, blocking its ubiquitination to increase glucose utilisation and lipogenic gene transcription [38]. Loss of hepatic Tcf7l2 significantly increased O-GlcNAcylation of ChREBP under HCR conditions, whereas hepatic Tcf7l2 restoration reduced it to normal levels (Fig. 6k,l and ESM Fig. 8b,c). Consistently, CRISPR/Cas9-mediated Tcf7l2 KO in the AML12 cell line significantly promoted ChREBPα protein expression and the expression and promoter activity of its target gene Pklr in a cell-autonomous manner, under high-glucose conditions (ESM Fig. 9). Indeed, the high-glucose-stimulated promoter activity of the ChREBP target Pklr in Tcf7l2 KO cell lines was suppressed by the O-GlcNAc transferase (OGT) inhibitor OSMI-1 (Fig. 6m). These data suggest that TCF7L2 regulates ChREBP transcriptional activity via O-GlcNAcylation (Fig. 6n).

Under refeeding conditions, hepatic Tcf7l2 deficiency significantly upregulated Srebf1c mRNA levels (Fig. 7a). However, TCF7L2 did not affect Srebf1c promoter activity induced by insulin, SREBP1c (autoregulation) or the LXR agonist T0901317 (Fig. 7b). Therefore, we investigated whether TCF7L2 regulates miRs that can trigger Srebf1c mRNA decay during the development of hepatic steatosis [39–41]. We found that hepatic Tcf7l2 deficiency decreased the hepatic expression of miR-33-5p, miR-132-3p and miR-212-3p under refeeding conditions (Fig. 7c). We noted that miR-33-5p showed apparently similar expression patterns to Tcf7l2 during periods of HCD feeding and was markedly decreased in the liver and primary hepatocytes of Alb-Cre;Tcf7l2^f/f mice fed a chronic HCD (Fig. 7d and ESM Fig. 10 [comparative HCD feeding data for TCF7L2 are shown in Fig. 3c and ESM Fig. 5d,e]). miR-33-5p also had a potent inhibitory effect on the activity of the murine Srebf1 3′UTR, as determined by luciferase reporter assay (Fig. 7e). Similarly, compared with the other miRs investigated, miR-33-5p significantly reduced basal Fasn and Acaca expression and Tcf7l2 deficiency-induced expression of Srebf1c, Fasn and Acaca (Fig. 7f). Consistent with this, miR-33-5p significantly reduced Tcf7l2 deficiency-induced DNL and hepatic TG content (Fig. 7g,h). To further elucidate the relationship between TCF7L2 and miR-33-5p, we tested whether TCF7L2 could transcriptionally regulate miR-33-5p. Careful investigation of promoter sequences showed that the putative TCF-binding element (TBE) site was localised at −1249 to −1243 from the transcriptional start site of the miR-33-5p promoter (Fig. 7i). Indeed, expression of hepatic TCF7L2 enhanced activity of the miR-33-5p promoter, that included a putative TBE site (−1412 to +16), but did not affect the activity of the miR-33-5p promoter without a putative TBE site (−852 to +16; Fig. 7i). TCF7L2 specifically occupied the miR-33-5p promoter region containing the putative TBE site (−1249 to −1243; Fig. 7j).

To further investigate the physiological role of hepatic Tcf7l2 in high-carbohydrate-induced fatty liver, we restored hepatic Tcf7l2 expression in Alb-Cre;Tcf7l2^f/f mice under acute and chronic HCD conditions. The restoration of hepatic Tcf7l2 expression in mice subjected to HCR reduced the expression of glycolytic and lipogenic factors to normal levels (Fig. 8a,b). This resulted in the restoration of hepatic DNL and TG levels (Fig. 8c,d). Under chronic HCD conditions, hepatic Tcf7l2 restoration did not cause changes in body weight, adiposity or body composition, and it did not affect the weights of the liver, epididymal white adipose tissue, inguinal white adipose tissue or brown adipose tissues (Fig. 8e–i). Plasma TG and NEFA levels also remained unchanged (ESM Fig. 11a–c). However, similar to the model of acute HCD, hepatic Tcf7l2 restoration significantly recovered protein levels of key lipogenic enzymes (ACC and FAS) and mRNA levels of glycolytic Gck and lipogenic Fasn, and restored chronic HCD-induced hepatic lipid droplet formation and hepatic TG levels as shown by Oil Red O staining of liver sections (Fig. 8j–l and ESM Fig. 11d,e). Indeed, adenovirus-mediated expression of hepatic Tcf7l2 in wild-type C57BL/6N mice fed an HCD for 22 weeks decreased chronic HCD-induced hepatic lipid droplets and TG content, concomitant with reductions in DNL-associated genes, without changes in body weight (ESM Fig. 11f–j).

To further emphasise the pathological role of hepatic Tcf7l2 deficiency in the liver, we explored whether the absence of hepatic Tcf7l2 exacerbated NAFLD into more severe forms. Mice lacking hepatic Tcf7l2 displayed simple liver steatosis at 16 weeks of HCD feeding and progressive fatty liver at 34 weeks of a long-term HCD (ESM Fig. 12a). In detail, Alb-Cre;Tcf7l2^f/f mice fed an HCD for 34 weeks exhibited steatosis and lobular inflammation (F4/80-positive area) in the liver (ESM Fig. 12b,c). Although the data are not shown, we observed some hepatocellular ballooning with Mallory–Denk bodies in the liver. Alb-Cre;Tcf7l2^f/f mice had elevated NAS, which is the sum of scores for steatosis, lobular inflammation and ballooning (ESM Fig. 12d). Additionally, liver fibrosis, as determined by the proportion of the Sirius red-stained area, was significantly increased upon hepatic Tcf7l2 deficiency (ESM Fig. 12e). Inflammation and fibrosis in the liver of Alb-Cre;Tcf7l2^f/f mice was further confirmed by analysis of mRNA and protein levels (ESM Fig. 12f,g).

Although this study did not model the Tcf7l2 SNP associated with the risk of type 2 diabetes, based on its potential to utilise the excess glucose pool more efficiently, we suggest that TCF7L2 may be a promising regulator of NAFLD associated with dietary carbohydrates and diabetes. Pathologically, hepatic Tcf7l2 deficiency-induced fatty liver progressed to NASH, a severe form of NAFLD, demonstrating the potent effect of TCF7L2 on hepatic DNL. Therefore, Alb-Cre;Tcf7l2^f/f mice could be a useful mouse model to investigate NAFLD development and progression under HCD conditions.

The authors declare that there are no relationships or activities that might bias, or be perceived to bias, their work.

K-JO, K-HB and HailK supervised the experiments. K-JO, K-HB and DSL designed the experiments. DSL, THA and HyunmiK contributed to the experimental preparation and data analysis. K-JO generated liver-specific KO mice. S-HK and C-HL helped generate mice. JKS, JWP, GK and SYO contributed to the evaluation of liver histology and body composition monitoring. S-HK, EJ, T-SH and JSK helped to generate the plasmid construct. HJC, DHH and Y-hL contributed to the preparation and analysis of human liver samples. D-SK and JJ helped with bioinformatics analysis. WKK, E-WL, B-SH and SCL contributed to data interpretation. KH and Y-HL interpreted the physiological meaning of the study. S-HK provided critical reagents, samples and comments. K-JO organised the data and wrote the manuscript. All authors made substantial contributions to editing and revising the manuscript and have approved the final version of the manuscript. K-JO is the guarantor of this work and is responsible for the integrity of the work as a whole.

RNA-sequencing data have been deposited into the NCBI GEO under the accession number GSE162449↗ (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE162449↗).

RNA-sequencing data have been deposited into the NCBI GEO under the accession number GSE162449↗ (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE162449↗). Data that support the findings of this study are available from the lead contact upon reasonable request.