Chromatin occupancy of transcription factor 7-like 2 (TCF7L2) and its role in hepatic glucose metabolism
- First Online:
- 1.5k Downloads
The mechanisms by which transcription factor 7-like 2 (TCF7L2) regulates the pathways that are important in the pathogenesis of type 2 diabetes are unknown. We therefore examined the role of TCF7L2 in hepatic glucose production (HGP) in vitro and characterised the whole-genome chromatin occupancy of TCF7L2 in hepatocytes.
We investigated the effect of TCF7L2 silencing and overexpression on HGP from gluconeogenic precursors and used chromatin-immunoprecipitation (ChIP) combined with massively parallel DNA sequencing (ChIP-Seq) to investigate the DNA binding patterns of TCF7L2 across the whole genome.
Silencing of TCF7L2 induced a marked increase in basal HGP, which was accompanied by significant increases in the expression of the gluconeogenic genes Fbp1, Pck1 and G6pc. Overexpression of Tcf7l2 reversed this phenotype and significantly reduced HGP. TCF7L2 silencing did not affect the half-maximal inhibitory concentration of insulin or metformin, but HGP remained elevated in TCF7L2-silenced cells due to the increased baseline HGP. Using ChIP-Seq, we detected 2,119 binding events across the genome. Pathway analysis demonstrated that diabetes genes were significantly over-represented in the dataset. Our results indicate that TCF7L2 binds directly to multiple genes that are important in regulation of glucose metabolism in the liver, including Pck1, Fbp1, Irs1, Irs2, Akt2, Adipor1, Pdk4 and Cpt1a.
TCF7L2 is an important regulator of HGP in vitro and binds directly to genes that are important in pathways of glucose metabolism in the liver. These data highlight the possibility that TCF7L2 may affect fasting and postprandial hyperglycaemia in carriers of at-risk TCF7L2 genetic polymorphisms.
KeywordsHepatic glucose metabolism TCF7L2 Type 2 diabetes
ChIP combined with massively parallel DNA sequencing
False discovery rate
Hepatic glucose production
Hepatocyte nuclear factor 4α
Half-maximal inhibitory concentration
Ingenuity Pathway Analysis
Pyruvate dehydrogenase complex
Pyruvate dehydrogenase kinase 4
Small interfering RNA
Transcription factor 7-like 2
Transcriptional start site
Morphogenic wingless-type MMTV integration site family
The most significant type 2 diabetes candidate gene to emerge from recent genome-wide association studies is TCF7L2. Linkage analysis originally identified a type 2 diabetes susceptibility locus on chromosome 10 , which was later refined to a single nucleotide polymorphism within intron 3 of TCF7L2 . This association has since been replicated many times in several populations throughout the world . The gene product, transcription factor 7-like 2 (TCF7L2), is a high-mobility box-containing transcription factor that plays an important role in executing downstream signals of the morphogenic wingless-type MMTV integration site family (WNT) pathway. The WNT signalling cascade is a highly conserved developmental pathway, which, classically, has been associated with embryogenesis and tumorigenesis. Activation of the pathway leads to accumulation of β-catenin in the nucleus, which associates with the LEF/TCF family of transcription factors to regulate the transcription of WNT target genes, as detailed elsewhere . Following the genetic association, numerous laboratories have attempted to describe a link between WNT signalling and the pathways involved in the aetiology of type 2 diabetes. Most of these studies have focused on the pancreatic beta cell, as early indications were that TCF7L2 is important in controlling glucose- and incretin-stimulated insulin secretion in vivo and in vitro .
However, several studies have indicated that TCF7L2 polymorphisms are also associated with elevated rates of hepatic glucose production (HGP) and reduced hepatic insulin sensitivity in humans in vivo [6, 7]. In carriers of the at-risk T allele at rs7903146, rates of basal HGP were elevated and failed to be suppressed normally during a euglycaemic–hyperinsulinaemic (40 mU m−2 min−1) clamp .
The potential link between TCF7L2 and hepatic glucose metabolism is further strengthened by increasing evidence of a key role for the WNT/β-catenin pathway in driving metabolic zonation in the liver . Hepatocytes are metabolically specialised, depending partly on their position along the porto–central axis of the liver-cell plate . Periportal hepatocytes in close proximity to the biliary duct, portal vein and hepatic artery are known to be gluconeogenic in nature and express key enzymes involved in this function. Hepatocytes with a more central location, designated as pericentral, are naturally glycolytic and display a gene expression profile more appropriate for this phenotype . Functional zonation, therefore, is directed by the transcriptional regulation of genes that are specific to the function concerned. For example, key genes of gluconeogenesis FBP1, PCK1 and G6PC are all upregulated in periportal hepatocytes. In contrast, glycolytic genes such as GCK and LDHD are overexpressed in pericentral hepatocytes . Remarkably, inactivation of WNT/β-catenin signalling in mouse liver is sufficient to transform hepatocytes from a pericentral to a periportal phenotype . This metabolic switch is associated with dramatic changes in the expression of genes involved in gluconeogenesis, including Pck1 and G6pc .
In the current study, we demonstrate that silencing of Tcf7l2 specifically in hepatocytes leads to a marked increase in HGP. We also provided evidence that this effect occurs primarily through increased expression of genes involved in gluconeogenesis. Using whole-genome chromatin-immunoprecipitation (ChIP) combined with massively parallel DNA sequencing (ChIP-Seq), we also document that TCF7L2 specifically binds to multiple genes that are important for the regulation of glucose metabolism in hepatocytes.
The rat hepatoma cell line, H4IIE, was used in the present study and has been described in detail elsewhere . H4IIE cells are highly differentiated, display a greater signal/noise ratio for HGP than other hepatic cell lines (i.e. HepG2, HuH6) and have been used extensively for the study of HGP in vitro [15, 16, 17]. Low passage cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and routinely cultured in DMEM supplemented with 4.5 mmol/l glucose and 10% (vol./vol.) FBS.
TCF7L2 silencing and glucose production assay
After treatment, H4IIE cells were washed in ice-cold PBS and cellular protein was collected by scraping the cells into 50 μl protein extraction buffer, after which western blotting was carried out as described . Membranes were probed with the following antibodies, all supplied by Cell Signaling Technology (Danvers, MA, USA) unless stated otherwise: TCF7L2 (number 2569), total AKT (number 4691), phosphorylated AKT (pAKT) (number 4060), myc-tag (number 2278) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (G9545; Sigma, St Louis, MO, USA).
Quantitative RT-PCR analysis
Cells in 24-well plates were immediately washed with ice-cold PBS following treatment and lysed with 500 μl Trizol reagent (Invitrogen). Total RNA was extracted, cDNA prepared and real-time PCR carried out as previously described . TaqMan primer and probe pre-mixes were purchased (Applied Biosystems, Carlsbad, CA, USA). For ChIP assays, primers were designed for SYBR Green quantification, the sequences of which are listed in the electronic supplementary material (ESM Table 1).
ChIP DNA was prepared using a kit (number 9003; Enzymatic SimpleChIP Magnetic Kit; Cell Signaling) and anti-TCF7L2 antibody (number 2569; Cell Signaling). This antibody was designed against a peptide sequence surrounding Leu330 of the TCF7L2 protein. This residue is at the 3′ end of exon 9 and represents a region of conservation in all Tcf7l2 mRNA transcripts [20, 21]. Sequencing libraries were prepared from 30 ng ChIP DNA using a kit (ChIP-Seq Library Preparation; Illumina, San Diego, CA, USA). The TCF7L2 library was prepared from DNA obtained from replicate ChIP reactions (n = 4).
ChIP-Seq data analysis: identification of TCF7L2-enriched regions
Analysis of ChIP-Seq data was performed with CLC Genomics Workbench version 4.5.1 (Cambridge, MA, USA). Raw 40 bp reads for the input control (29,226,566 reads) and TCF7L2 ChIP reaction (30,688,073 reads) were first mapped to the repeat-masked Rat Reference (RN4) Genome (UCSC Genome Browser, http://genome.ucsc.edu/, accessed 28 September 2010). ChIP-Seq peak detection was then carried out using a window size of 250 bp and a false discovery rate (FDR) of 1%.
We searched for occurrence of the canonical (T/A–T/A–C–A–A–A–G) or evolutionarily conserved (A–C/G–T/A–T–C–A–A–A–G) TCF7L2 binding motif in each peak (±400 bp flanking DNA). To perform a bias-free search for novel and potentially unknown TCF7L2 motifs in the rat, we also used MEME-ChIP  on a small subset of the ChIP peaks. For this analysis, each peak was extracted with 25 bp of the 5′ and 3′ flanking sequences, and the analysis was run with default conditions, but with a minimum and maximum motif width of 6 and 15 respectively.
Data are presented as means±SEM. Two-tailed t tests were used to detect statistical differences in cell culture studies.
TCF7L2 and glucose production
We initially determined that TCF7L2 was highly abundant in the hepatic cell line H4IIE (Fig. 1a). We were able to silence TCF7L2 protein production by over 70% and 80% for the 78 and 58 kDa bands, respectively (Fig. 1a). Protein levels began to recover 48 h after silencing, returning to near normal levels after 96 h. Notably, we did not observe any additional protein isoforms that were indicative of alternative splicing. We also did not detect any effect of TCF7L2 silencing on cell proliferation as measured by the 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) assay (ESM Fig. 1).
In rat hepatocytes in which TCF7L2 was silenced, addition of lactate/pyruvate caused a three- to fivefold increase in glucose production compared with the scrambled siRNA control (Fig. 1a). This effect was highly reproducible and was reversed when wild-type TCF7L2 was overproduced. Overabundance of myc-tagged TCF7L2 in hepatocytes induced a 50% reduction in glucose production determined over a 6 h period (Fig. 1b).
Expression of genes involved in gluconeogenesis
The mRNA expression of a number of key gluconeogenic enzymes was also increased significantly in TCF7L2-silenced cells, when compared with the scrambled siRNA control (Fig. 1b). Pck1 (36%, p < 0.01; Fig. 1c), Fbp1 (152%, p < 0.0001; Fig. 1d) and G6pc (132%, p < 0.0001; Fig. 1e) were all significantly upregulated. The additional rate-controlling enzyme of gluconeogenesis, pyruvate carboxylase (Pc), was not affected by TCF7L2 silencing (ESM Fig. 2). Expression of Hnf4a was elevated following TCF7L2 silencing, whereas Ppargc1a mRNA was unaffected (ESM Fig. 2).
Insulin and metformin sensitivity of glucose production
ChIP assay validation
Transcriptional targets of TCF7L2 in hepatocytes
Illumina sequencing libraries were generated from input-control and TCF7L2-ChIP reactions and size-selected such that the final sequencing libraries were approximately 200 to 250 bp in length (Fig. 3c) (GEO accession: GSE28782). We generated 40-base pair sequencing reads (input 29,226,566 reads; TCF7L2 ChIP 30,688,073 reads) and mapped them to the repeat-masked Rat Reference (RN4) Genome. Approximately 46% of the input and TCF7L2 reads aligned uniquely to the repeat-masked genome (ESM Table 2).
Motif analysis of TCF7L2-enriched regions
A search for the canonical (T/A–T/A–C–A–A–A–G) and the evolutionarily conserved (A–C/G–T/A–T–C–A–A–A–G) TCF7L2 motifs showed that 65% of all peaks (1,388 of 2,119) had at least one copy of the canonical motif and 28% (589 of 2,119) had at least one copy of the long, evolutionarily conserved motif (ESM Table 5). To validate this search and to identify potentially unknown motifs in a subset of the TCF7L2 ChIP peaks, we performed a de novo motif search using MEME-ChIP on a subset of our data. Searching for motifs between 6 and 15 bp in length, we found a remarkable and highly consistent enrichment of the long, evolutionarily conserved TCF7L2 binding motif (Fig. 4c–f). Restricting the search to just 7 bp revealed only enrichment of the shorter canonical motif (not shown). We did not detect any other significant consensus sequences in the rat hepatocytes.
Functional analysis of TCF7L2 target genes
IPA of the ChIP-Seq data, showing the top five functional annotations for the metabolic disease category
7.44 × 10−12
Non-insulin-dependent diabetes mellitus
1.63 × 10−10
3.15 × 10−5
Insulin-dependent diabetes mellitus
4.40 × 10−5
Insulin resistance of mice
1.99 × 10−4
IPA of the ChIP-Seq data, showing the top five functional annotations for the genetic disorder category
Non-insulin-dependent diabetes mellitus
1.63 × 10−10
3.25 × 10−10
Coronary artery disease
2.16 × 10−9
Inflammatory bowel disease
1.59 × 10−7
2.45 × 10−7
Genes represented by the type 2 diabetes signalling category following IPA analysis
Entrez Gene name
Acyl-CoA synthetase long-chain family member 5
Adiponectin, C1Q and collagen domain containing
Adiponectin receptor 1
v-akt murine thymoma viral oncogene homolog 2
Ataxia telangiectasia mutated
Inhibitor of kappa light polypeptide gene enhancer in B cells, kinase beta
Insulin receptor substrate 1
Insulin receptor substrate 2
Nuclear factor of kappa light polypeptide gene enhancer in B cells inhibitor, alpha
Phosphoinositide-3-kinase, class 2, beta polypeptide
Phosphoinositide-3-kinase, regulatory subunit 4
Pyruvate kinase, muscle
Protein kinase, AMP-activated, gamma 2 non-catalytic subunit
Protein kinase C, alpha
Protein kinase C, delta
Protein kinase C, zeta
Protein kinase D3
Solute carrier family 27 (fatty acid transporter), member 3
Solute carrier family 27 (fatty acid transporter), member 5
Sphingomyelin phosphodiesterase 3, neutral membrane (neutral sphingomyelinase II)
Suppressor of cytokine signalling 1
Suppressor of cytokine signalling 3
We also observed enrichment at several other genes critical for insulin signalling and hepatic glucose metabolism. These included Irs1 and Irs2, Akt2, Adipor1, Prkcz, Gys2 and Cpt1a. Of these, we found that all except Cpt1a contain at least one copy of the canonical TCF7L2 motif and that Akt2 contains a canonical and an evolutionarily conserved TCF7L2 recognition site. We confirmed a number of these peaks independently using ChIP assays followed by quantitative RT-PCR (Fig. 5).
In the current study, we have demonstrated that silencing of TCF7L2, alone and selectively, is sufficient to markedly increase HGP from gluconeogenic precursors (lactate/pyruvate) in hepatocytes in vitro. Using a whole-genome ChIP-Seq approach, we have also shown here for the first time that TCF7L2 binds to a number of genes important for glucose metabolism and insulin signalling. Together, these data suggest that TCF7L2 plays an important role in the regulation of hepatic glucose metabolism.
Polymorphisms within TCF7L2 have consistently been associated with an elevated risk of type 2 diabetes . Numerous studies have suggested a strong link between TCF7L2 and pancreatic beta cell function in vivo and in vitro [5, 7, 26]. However, there is also evidence to suggest that variants in TCF7L2 may augment basal HGP and impair the ability of insulin to inhibit hepatic gluconeogenesis. Pilgaard and colleagues  showed that basal HGP was elevated and failed to be suppressed normally during a hyperinsulinaemic–euglycaemic clamp in T allele carriers, a finding confirmed in larger studies . Elevated hepatic insulin resistance in T allele carriers has also been documented in a Danish twin cohort  and in patients with non-alcoholic fatty liver disease . Our findings support a role for TCF7L2 in regulating hepatic gluconeogenesis and suggest that this regulation may occur at the transcriptional level. However, the ability of insulin to inhibit HGP from gluconeogenesis (i.e. IC50) was preserved when TCF7L2 was silenced. Because of the marked increase in basal HGP, the absolute rate of HGP for any given insulin concentration remained elevated in hepatocytes treated with TcfF7l2 siRNA compared with those treated with scrambled siRNA.
Although we used an in vitro model, our results, if extrapolated to an in vivo system, suggest that the T allele at rs7903146 may play a role in fasting as well as postprandial hyperglycaemia, which would be consistent with results from studies employing the euglycaemic clamp technique . The precise functional impact of the T allele in vitro or in vivo is uncertain, but it is suspected to alter the expression and/or splicing of TCF7L2. Some studies have shown reduced expression of TCF7L2 in diabetic islets , whereas others have noted a marked increase in TCF7L2 mRNA in T allele carriers , or no change at all . The current study did not examine the relationship between rs7903146 and HGP, but our results suggest that TCF7L2 suppresses the expression of genes involved in gluconeogenesis and that reduced TCF7L2 protein levels lead to enhanced HGP. It is not known whether rs7903146 affects the expression of TCF7L2 in the liver in vivo, but our results may support the hypothesis that the T allele reduces abundance of the protein. In humans, the T allele is associated with elevated HGP . Further studies examining the impact of the T allele on TCF7L2 expression and function in the liver are needed.
Glucose production from hepatocytes is controlled by the expression and/or activity of several rate-limiting enzymes . The genes encoding three key gluconeogenic enzymes, Pck1, Fbp1 and G6pc, were all significantly upregulated by TCF7L2 silencing in the present study. We believe that the increased expression of these three genes played a major role in the increase of HGP observed by us. Moreover, it is possible that TCF7L2 directly regulates the expression of Pck1 and Fbp1, as we detected significant binding events at both of these genes in our ChIP-Seq dataset. Using IPA analysis, we also showed that TCF7L2 binds to genes known to be important for hepatic glucose metabolism. These genes included Irs1, Adipor1, Gys2, Cpt1a and Pdk4. Several of these genes may have a direct role in hepatic glucose metabolism in vivo. Glycogen synthase-2, which is produced exclusively in the liver, catalyses the rate-limiting step in the synthesis of glycogen, and defects in this protein lead to glycogen storage disease (type 0), a condition associated with elevated postprandial hyperglycaemia and fasting hypoglycaemia . Pyruvate dehydrogenase kinase 4 (PDK4) is an inhibitor of the pyruvate dehydrogenase complex (PDC), which catalyses the oxidative decarboxylation of pyruvate to acetyl-CoA . PDK4 enhances hepatic gluconeogenesis through inhibition of the PDC complex and thus preservation of pyruvate for gluconeogenesis, and is upregulated during starvation and in insulin-resistant states . CPT1A, again expressed exclusively in the liver, plays a key role in β-oxidation of fatty acids in the mitochondria; its selective inhibition reduces gluconeogenesis and improves glucose homeostasis in vivo . Adiponectin suppresses HGP independently of insulin in primary hepatocytes [34, 35] and adiponectin receptors have been shown to be reduced in mouse models of diabetes . Finally, signals from IRS1 (and IRS2) are critical for the maintenance and control of HGP; silencing of both together leads to severe and persistent hyperglycaemia . Clearly, the transcriptional activity of TCF7L2 could have significant implications for the regulation of HGP, insulin resistance and type 2 diabetes.
We cannot exclude the possibility that the robust changes in expression of genes involved in gluconeogenesis were part of a more global phenotypic shift in the hepatocytes, consistent with a role for TCF7L2 in metabolic zonation. The WNT/β-catenin pathway is a major determinant of the metabolic fate of hepatocytes [9, 38, 39, 40]. In this regard, several other pathways, including those responsible for amino acid, ammonia and drug metabolism, might also have been altered in our experimental model. Notably, we did observe significant ChIP-Seq peaks at several genes involved in the regulation of these pathways, including Oat (glutamine synthesis), Slc1a2 (glutamine synthesis), Cps1 (urea formation) and Rhbg (ammonia transport). Interestingly, these genes, in addition to Pck1, Fbp1 and G6pc, have also been identified in β-catenin-activated hepatocytes [13, 14, 41]. The effect of TCF7L2 silencing on Hnf4a expression may indeed suggest that TCF7L2 silencing induced a phenotypic switch in the hepatocytes. Hepatocyte nuclear factor 4α (HNF4α) is an important determinant of the periportal (gluconeogenic) transcriptional program. Hepatic HNF4α deficiency induces periportal production of glutamine synthetase and several other pericentral enzymes , acting through cross-talk with the WNT pathway . The mechanisms underlying the interaction between WNT and HNF4α are poorly understood, but may converge on the WNT transcription factor lymphoid enhancer binding factor 1 (LEF1) .
Over 60% of the ChIP-Seq peaks detected in our analysis had at least one copy of the short consensus sequence, either inside the peak or within proximal 5′ or 3′ flanking regions. Almost one third of the peaks had the long, evolutionarily conserved motif within the flanking regions. These data are consistent with several studies documenting chromatin occupancy of TCF7L2 in colon cancer cells [23, 24, 25]. An unbiased de novo search for TCF7L2 binding motifs did not reveal any novel TCF7L2 recognition sequences. It is possible that motifs for other transcription factors (nuclear factor-1, HNF4, peroxisome proliferator-activated receptor γ) were located within or surrounding our ChIP-Seq peaks, as has been shown by others , but we did not specifically test this hypothesis in the current study. Peaks not containing a recognisable motif may represent protein × protein interactions between TCF7L2 and other factors directly recruited to these areas.
We have shown that TCF7L2 is preferentially located inside gene boundaries (intragenic) and, indeed, 65% of all peaks were within 50 kb of the TSS of a gene or within the gene enhancer region, findings that again are consistent with previously published data . Hatzis et al.  commented that TCF7L2 had a tendency to cluster (5′, 3′, internal) around target genes. While we did not formally search for this type of binding pattern, several TCF7L2 target genes (e.g. Lef1 and Pck1) had multiple ChIP peaks across the gene and flanking regions. These data are consistent with the concept, suggested by others , that this binding pattern supports an open chromatin domain or provides a more accurate β-catenin sensor mechanism.
In summary, we have provided the first evidence for the role of TCF7L2 in the regulation of HGP via the gluconeogenesis pathway in vitro. Although more work is needed to define the precise role of this protein in hepatic glucose metabolism in vivo, our data provide a possible link between genetic association studies of TCF7L2 with type 2 diabetes and a physiological function that is highly relevant and disturbed in this common metabolic disease that affects 8% of the US population .
This work was supported by a San Antonio Area Foundation research grant to L. Norton. We would like to thank W. Wakeland, C. Arana and C. Liang at the Department of Immunology, UT Southwestern Medical Center (Dallas, TX, USA) for Illumina sequencing services.
LN conceived and designed the experiments, researched data, wrote the manuscript, contributed to discussion, and reviewed and edited the manuscript. MF, FFM and DW were involved in the analysis and interpretation of the data and in critically revising the article. MAG, CPJ and RAD conceived and designed the experiments, contributed to discussion, and critically reviewed and edited the manuscript. All authors approved the final version of this manuscript.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.