A candidate functional SNP rs7074440 in TCF7L2 alters gene expression through C‐FOS in hepatocytes

Abstract

The SNP rs7903146 at the transcription factor 7‐like 2 (TCF7L2) locus is established as the strongest known genetic marker for type 2 diabetes via genome‐wide association studies. However, the functional SNPs regulating TCF7L2 expression remain unclear. Here, we show that the SNP rs7074440 is a candidate functional SNP highly linked with rs7903146. A reporter plasmid with rs7074440 normal allele sequence exhibited 15‐fold higher luciferase activity compared with risk allele sequence in hepatocytes, demonstrating a strong enhancer activity at rs7074440. Additionally, we identified C‐FOS as an activator binding to the rs7074440 enhancer using a TFEL genome‐wide screen method. Consistently, knockdown of C‐FOS significantly reduced TCF7L2 expression in hepatocytes. Collectively, a novel enhancer regulating TCF7L2 expression was revealed through searching for functional SNPs.

Abbreviations

GWAS, genome‐wide association study

SNP, single‐nucleotide polymorphisms

TCF7L2, transcription factor 7‐like 2

TFEL, transcription factor expression library

Type 2 diabetes is a complex human disease caused by both genetic and environmental factors [1]. Approximately 10 years ago, a number of genetic variants associated with type 2 diabetes were identified by genome‐wide association studies (GWAS) [2-4]. Among these genomic variants, the TCF7L2 polymorphisms, first reported by Grant et al. [5] using microsatellite markers, have been reproducibly documented to be the most strongly associated markers of type 2 diabetes risk by numerous investigations among different ethnic groups, and the SNP rs7903146 has been established as the representative genetic marker [6-15].

Transcription factor 7‐like 2 (TCF7L2) is located on chromosome 10, and encodes a high mobility group (HMG) box‐containing transcription factor [16]. It participates in the Wnt‐signaling pathway and activates transcription from promoters with several copies of the TCF motif.

Regarding the mechanistic link between TCF7L2 polymorphisms and the etiology of type 2 diabetes, many studies have focused on the pancreatic beta cell, because early indications were that individuals with variants of TCF7L2 associated with increased risk for type 2 diabetes exhibit impaired insulin secretion [6, 7]. However, several studies have indicated that TCF7L2 polymorphisms are also associated with elevated rates of hepatic glucose production and reduced hepatic insulin sensitivity in humans in vivo [9, 17, 18]. Thus, the TCF7L2 risk SNPs may affect the metabolic functions of both pancreatic islets and the liver [19-21].

To date, all recognized type 2 diabetes‐related TCF7L2 risk SNPs are located within intronic regions of this gene [16]. Because these SNPs do not affect the coding sequence, one may assume that they affect expression of TCF7L2. However, the underlying mechanism for such transcriptional regulation currently remains unclear.

Recently, we have independently developed a useful library for pairing an enhancer and its transactivator [22]. By utilizing this method named TFEL scan, it became possible to easily determine the correspondence between an enhancer and its transactivator.

These situations prompted us to search for a functional enhancer variant within the TCF7L2 gene, to identify the corresponding transcription factor(s) for this enhancer, and to elucidate the molecular mechanisms determining the genetic variation of TCF7L2 expression. For this purpose, we searched for the functional SNPs highly linked with the GWAS lead SNP rs7903146 with enhancer activities as a clue and identified the corresponding transactivator using TFEL scan.

Materials and methods

Search for candidate functional SNPs

Several SNP annotation databases were used to identify possible SNP functions. The UCSC genome browser (https://genome.ucsc.edu/) was used as the source of genome‐wide maps of the chromatin state of the region of interest of the gene. The SNP rs7903146 has been identified to be the most strongly associated with type 2 diabetes by GWAS [4, 23]. The SNP rs7903146 is a ‘marker’ SNP and candidate functional SNPs highly linked with rs7903146 were selected with HaploReg V4 (http://compbio.mit.edu/HaploReg) [24], using a linkage disequilibrium (LD) threshold of r² ≥ 0.60 with rs7903146 at two or more areas among Europe, America, and Asia and a minor allele frequency (MAF) threshold of freq > 0 in all the areas. As a result, 21 SNPs including rs7903146 were selected as candidate functional SNPs. All the SNPs are located approximately within 67‐kb region extending from 112 989 975 to 113 057 250 on chromosome 10 (UCSC Genome Browser hg38).

Construction of luciferase reporter plasmids

The fragment of one or three tandem copies of the each SNP 25‐bp region which containing ‘risk allele’ was cloned into BamHI and SalI sites located at the 3′ end of firefly luciferase reporter gene of pGL3 promoter plasmid (Promega Corporation, Madison, WI, USA) (SV40pro‐Firefly Luc‐SNPx1 and SV40pro‐Firefly Luc‐SNPx3). Similarly, the fragments containing ‘normal allele’ were cloned into BamHI and SalI sites located at the 3′ end of renilla luciferase reporter gene of a modified pGL3 promoter plasmid whose firefly luciferase reporter gene was substituted by renilla luciferase reporter gene (SV40pro‐Renilla Luc‐SNPx3). These fragment sequences are described in Table S1. TCF7L2 promoter region was amplified by PCR with primer sets 5′‐ACGCGTGGCATATCCATCCTAGTGGGAC‐3′ and 5′‐AGATCTGAACGGAGTAGTCTGGGAGC‐3′ and the PCR product was cloned into pGEM‐T‐easy plasmid (Promega Corporation). The cloned fragment was cut out from the vector and ligated into MluI and BglII sites of pGL3 basic plasmid (Promega Corporation) (TCF7L2 pro‐Firefly Luc). The rs7074440 25 bp region of normal allele or risk allele was cloned into BamHI and SalI sites located at the 3′ end of firefly luciferase reporter gene of pGL3 promoter plasmid (TCF7L2 pro‐Firefly Luc‐SNP). Expression plasmid of mouse c‐Fos was included in the Transcription Factor Expression Library (TFEL) [22].

Transfection and luciferase assay

HepG2 and Huh7 cells were cultured in DMEM (Wako Pure Chemical Industries, Osaka, Japan) containing 25 mm glucose, 100 U·mL⁻¹ penicillin, and 100 mg·mL⁻¹ streptomycin sulfate supplemented with 10% FBS. HEK 293 and Hela cells were cultured in DMEM (Wako Pure Chemical Industries) containing 25 mm glucose, 100 U·mL⁻¹ penicillin, and 100 mg·mL⁻¹ streptomycin sulfate supplemented with 10% CCS. The indicated expression plasmids, firefly luciferase reporter plasmid and renilla luciferase reporter plasmid, were co‐transfected into cells using SuperFect Transfection Reagent (QIAGEN, Hilden, Germany) according to the manufacturer's protocol. Transfections were carried out in all cell lines in triplicate. Total amounts of transfected DNA were adjusted with empty vector. The luciferase activity in transfectant was measured on a luminometer as described previously [25].

In vivo imaging of enhancer activity

Six‐week‐old ICR male mice were purchased from SLC (Shizuoka, Japan). All animals were maintained in a temperature‐controlled environment with a 14 h‐light/10 h‐dark cycle and were given free access to standard laboratory diet and water. All animals studied were anesthetized and euthanized according to the protocol approved by the Tsukuba University Animal Care and Use Committee.

The fragments containing three tandem rs7074440 25‐bp region of normal allele or risk allele SNP sequence with SV40 promoter region linked to luciferase reporter gene cut from SV40pro‐Firefly Luc‐SNPx3 were inserted into the Gateway entry vector pENTR4 (Invitrogen, Carlsbad, CA, USA), and the adenoviral plasmid was generated by homologous recombination between the entry vector and the pAd promoterless vector (Invitrogen) (Ad‐SV40pro‐Firefly Luc‐SNPx3). After the transfection of the plasmid into 293A cells, recombinant adenoviruses were collected by CsCl gradient centrifugation as described previously [26].

In vivo imaging was performed as described previously [22]. Adenoviruses were injected intravenously into ICR male mice at the dose of 1 × 10⁸ P.F.U./body. After 3 days, mice were anesthetized and luciferin dissolved in PBS (7.5 mg·mL⁻¹) was injected into the intraperitoneal cavity. The luminescence in the liver was visualized using an In Vivo Imaging System (IVIS™, Xenogen, Alameda, CA, USA). For the luminescence quantification, tissue samples (about 50 mg) were homogenized with a Polytron in 300 μL of Reporter Lysis Buffer (Promega Corporation) and centrifuged at 25 000 g for 15 min, and luciferase activity in the supernatant was measured on luminometer. To determine hepatic transduction efficiency, the pellets obtained in the process described above were digested with proteinase K and the DNA samples were extracted with phenol/chloroform and ethanol precipitation, and then the amounts of adenoviral DNA in the samples were quantified using Q‐RT PCR method as described previously [27]. The primers used for the quantification of luciferase gene were 5′‐GTCCGGTTATGTAAACAATCC‐3′ and 5′‐ATGAAGAAGTGTTCGTCTTCG‐3′.

Genome‐wide transcription factor screening

HEK 293 cells in a 48‐well plate were co‐transfected with the TFEL clones [22] and reporter plasmids of SV40pro‐Renilla Luc‐SNPx3 plus SV40pro‐Firefly Luc‐SNPx3. Cells were lysed in Reporter Lysis Buffer (Promega Corporation) 24 hr after transfection, and then firefly and renilla luciferase activities were measured.

Electrophoretic mobility shift assay

Full‐length mouse cDNAs of c‐Fos and c‐Jun were amplified with PCR. The PCR primers were added with T7 Promoter – Spacer – Kozak – HA Tag sequences. The PCR primers used were as follows: mouse c‐Fos, 5′‐GGATCCTAATACGACTCACTATAGGGAACAGCCACCATGTACCCATACGATGTTCCAGATTACGCTATGATGTCCTCGGGTTTCA‐3′, and 5′‐CCATAGAGCCCACCGCATC‐3′; mouse c‐Jun, 5′‐GGATCCTAATACGACTCACTATAGGGAACAGCCACCATGTACCCATACGATGTTCCAGATTACGCTATGACTGCAAAGATGGAAACGAC‐3′ and 5′‐CCATAGAGCCCACCGCATC‐3′. Eight microliters of TNT T7 Quick Master Mix (Promega Corporation), 0.2 μL of 1 mm methionine, and PCR‐amplified DNA (0.1 μg) described above were mixed and incubated at 37 °C for 90 min. Electrophoretic mobility shift assay (EMSA) was performed as previously described [28]. Briefly, three DNA probes were prepared by annealing two oligonucleotides. The probe sequences are listed in Table S2. The probes were labeled with [α‐³²P] dCTP by filling in the 5′‐overhangs with Klenow DNA polymerase (GE Healthcare, Chalfont St. Giles, UK), and purified on Sephadex G‐25 (GE Healthcare) columns. The labeled DNA probe was incubated with in vitro synthesized protein lysate in a buffer containing 10 mm HEPES at pH 7.8, 50 mm KCl, 1 mm EDTA, 5 mm MgCl₂, 5 mm dithiothreitol, 30 μg·mL⁻¹ poly(dI‐dC), and 0.1% Triton X‐100, for 30 min on ice. The protein complex was resolved on 4.6% polyacrylamide gels in 1xTBE buffer.

Chromatin immunoprecipitation assays

Nuclear extracts was prepared from HepG2 cell line. Chromatin immunoprecipitation (ChIP) assays were performed as described previously [28, 29]. Antibodies used for the ChIP assay were anti‐c‐Fos antibody (sc‐52‐G) and normal goat IgG (sc‐2061). DNA quantity was measured using real‐time PCR. 1% of the input genomic DNA was used for input genomic c‐DNA. Primers used to amplify TCF7L2 gene Locus containing SNP rs7074440 region were as follows: (125 bp was amplified) 5′‐GAATTCCTGGACTCAAGCAATC‐3′and 5′‐ACCTCTCTACGCCTCC AGATC‐3′. Primers used to amplify the control region containing 100 kb upstream of TCF7L2 gene were as follows: (100 bp was amplified) 5′‐ATCACGAGGTCAAGAGATCG‐3′and 5′‐TCAGCCTCTTGATAGCTGG‐3′.

RNA interference

Full‐length mouse cDNAs of c‐Fos were amplified by PCR with primer sets 5′‐GGATCCATGATGTTCTCGGGTTTCAACG‐3′ and 5′‐CTCGAGTGACTGCTCACAGGGCCA‐3′, and the PCR product was cloned into pGEM‐T‐easy plasmid. The c‐Fos cDNA fragment was cut out from the plasmid and ligated into BamHI and XhoI sites of pcDNA3.1(+) (Invitrogen) with an N‐terminal Flag tag expression plasmid (pcDNA3.1‐Flag‐c‐Fos). Human and mouse C‐FOS‐specific shRNA expression plasmids were subcloned into pENTR/U6 entry vector (Invitrogen) using the synthesized oligo DNA fragments with the following sequences (C‐FOSi‐1 and C‐FOSi‐2, respectively): 5′‐GAGTCTGAGGAGGCCTTCATGTGAAGCCACAGATGGTGAAGGCCTCCTCAGACTC‐3′ and 5′‐CTTCTATGCAGCAGACTGGGATGTGAAGCCACAGATGGTCCCAGTCTGCTGCATAGAAG‐3′. The above two target sequences were designed from the consensus sequence of human and mouse C‐FOS coding region. To examine knockdown efficiency for c‐Fos protein, pcDNA3.1‐Flag‐c‐Fos was co‐transfected with C‐FOSi‐1 or C‐FOSi‐2 in HEK 293 cells. Twenty‐four hours after transfection, cells were lysed and then western blotting was performed using anti‐FLAG (F3165) (mouse monoclonal) antibody (Sigma‐Aldrich, Zwijndrecht, Netherlands) as described previously [22]. Both C‐FOSi‐1 and C‐FOSi‐2 effectively knocked down the c‐Fos protein expression (data not shown).

Adenovirus vectors for knockdown were generated by homologous recombination between the entry vector and the pAd promoterless vector (Ad‐C‐FOSi‐1 and Ad‐C‐FOSi‐2) using the Gateway system. LacZ‐specific shRNA expression plasmids (LacZi) and adenovirus vector (Ad‐LacZi) were described previously [26].

RNA isolation and quantitative reverse transcription PCR

Total RNA was extracted from HepG2 cells with Sepazol reagent (Nacalai Tesque, Kyoto, Japan) and transcribed into cDNA using reverse transcription reagent kit (TOYOBO, Osaka, Japan). Quantitative reverse transcription PCR was performed using ABI 7500 System (Applied Biosystems^® 7500 Fast, Applied Biosystems, Framingham, MA, USA). Q‐RT PCR was performed using SYBR Green PCR master mix. The primer sets were as follows: human C‐FOS, 5′‐ CTTCAACGCAGACTACGAGG‐3′ and 5′‐ATGAAGTTGGCACTGGAGA‐3′; human TCF7L2, 5′‐TCGCCTGGCACCGTAGGACA‐3′ and 5′‐GGATGCGGAATGCCCGTCGT‐3′; human Cyclophilin, 5′‐GCATACGGGTCCTGGCATCTTGTCC‐3′ and 5′‐ATGGTGATCTTCTTGCTGGTCTTGC‐3′.

Statistical analyses

Data are expressed as means ± SEM. Differences between two groups were assessed using an unpaired two‐tailed Student's t‐test. Data sets involving more than two groups were assessed by ANOVA using statcel3 Software (OMS Publishing Inc., Saitama, Japan). The differences were considered statistically significant at P < 0.05.

Results

Selection of potential functional SNPs

The SNP rs7903146 has been identified to be the most strongly associated marker of type 2 diabetes by GWAS [2] and is located in intron 4 of the TCF7L2 gene. In order to identify potential functional SNPs linked with rs7903146, we selected 21 candidate SNPs with a strong linkage disequilibrium to rs7903146 (r² ≥ 0.60) including itself (Fig. 1A). The details of SNPs selection criteria were described in ‘Materials and Methods’. Of 21 candidate SNPs, 9 SNPs are located within 10 kb upstream of rs7903146 locus and other 11 SNPs are located within 60 kb downstream of that. All SNPs are located within intron 4 of the TCF7L2 gene (Fig. 1B).

Figure 1

Functional assessment of TCF7L2 SNPs shows enhancer activity at rs7074440. (A) Method of screening functional SNPs highly linked with rs7903146. (B) Positions of candidate SNPs at TCF7L2 locus. rs7903146 is shown in yellow and distances from rs7903146 are indicated. Linkage disequilibrium (LD) map was captured from UCSC Genome Browser on Human Mar. 2006 (NCBI36/hg18) assembly (http://genome.ucsc.edu/index.html) using Phase 2 HapMap for the CEU Population (Northern and Western European Ancestry in Utah, US – CEPH). (C) Construction of two types of luciferase reporter plasmids. Three tandem copies of 25‐bp sequences spanning these SNPs were ligated into the SV40 promoter‐driven luciferase reporter vector to evaluate the enhancer activity of these SNPs. Each ‘Normal allele’ SNP sequence was inserted at the 3′ end of renilla luciferase reporter (SV40pro‐Renilla Luc‐SNPx3) and that of ‘Risk allele’ was inserted at the 3′ end of firefly luciferase reporter (SV40pro‐Firefly Luc‐SNPx3). (D) 21 normal SNP‐constructed renilla reporter vectors and SV40 promoter only vector (indicated as ‐) were co‐transfected, respectively, with 21 risk SNP‐constructed firefly reporter vectors into HepG2 and Huh7 human hepatoma cells. The relative luciferase activities are shown as a fold change to that of the control (n = 3 for each group). Each SNP position relative to rs7903146 on TCF7L2 gene locus is described below.

Functional assessment of TCF7L2 SNPs shows enhancer activity at rs7074440

To examine which of these 21 SNP candidates are functional SNPs affecting TCF7L2 gene expression in human hepatocytes, we performed a luciferase reporter assay in human hepatoma cell lines HepG2 and Huh7. To this end, we constructed reporter plasmid containing three copies of 25‐bp region surrounding each SNP from ‘normal’ allele placed at the 3′ end of renilla luciferase reporter gene and those from ‘risk’ allele placed at the 3′ end of firefly luciferase reporter gene (SV40pro‐Renilla Luc‐SNPx3 and SV40pro‐Firefly Luc‐SNPx3, respectively) to assess enhancer activity of the sequence around each SNP (Fig. 1C). Next, we co‐transfected the two plasmids for each SNP as a pair of normal and risk into HepG2 and Huh7 hepatocytes. As shown in Fig. 1D, the normal allele reporter plasmid for rs7074440 exhibited significantly higher luciferase activity compared with risk allele reporter plasmid in both HepG2 and Huh7 cell lines. In contrast, in nonhepatic cell lines HEK 293 and HeLa cells, significant allele‐specific differences for any SNP plasmids were not observed (Fig. S1). To further compare the normal and risk alleles around rs7074440 in the same reporter background, two plasmids based on firefly luciferase reporter (SV40pro‐Firefly Luc‐rs7074440x3) were constructed and transfected individually in HepG2 cells. As shown in Fig. 2A, SV40pro‐Firefly Luc‐rs7074440x3 for normal (G) allele exhibited 15‐fold higher luciferase activity compared with that for risk (A) allele, demonstrating a strong enhancer activity of normal (G) allele at rs7074440. The reporter plasmid containing ‘one’ copy of the 25‐bp region surrounding rs7074440 from normal allele also showed significantly higher luciferase activity compared with that from risk allele reporter plasmid (SV40pro‐Firefly Luc‐rs7074440x1) in HepG2 cells (Fig. S2). Consistently, a similar result was obtained about reporter plasmids driven by TCF7L2 gene promoter (TCF7L2 pro‐Firefly Luc‐rs7074440x1) in HepG2 cells (Fig. 2B). The results in Fig. 2A,B also indicate that rs7074440 risk (A) allele has no enhancer activity in HepG2 cells.

Figure 2

Enhancer activity of rs7074440. (A) Enhancer activity of rs7074440 was examined in HepG2 cells. HepG2 cells were transfected with three tandem rs7074440 sequences of normal allele (G) or risk allele (A) inserted into SV40 promoter firefly luciferase reporter vectors (SV40pro‐Luc‐rs7074440x3). Firefly luciferase activity was normalized to SV40 promoter renilla luciferase activity and shown as a fold change to that of the SV40 promoter only firefly luciferase vector (n = 3 for each group). (B) Enhancer activity of rs7074440 was examined in the presence of TCF7L2 promoter in HepG2 cells. HepG2 cells were transfected with a 1 copy rs7074440 sequences of normal allele (G) or risk allele (A) inserted into TCF7L2 promoter driven firefly luciferase reporter vectors (TCF7L2 pro‐Luc‐rs7074440). Firefly luciferase activity was normalized to SV40 promoter renilla luciferase activity and exhibited as a fold change to that of TCF7L2 promoter only firefly luciferase vector (n = 3 for each group). (C) In vivo analysis of hepatic enhancer activities of rs7074440 is shown. SV40pro‐Luc‐rs7074440x3 adenovirus (Ad‐SV40pro‐Luc‐rs7074440x3) (1.0 × 10⁸ P.F.U./body) of normal allele (G) or risk allele (A) was intravenously injected into ICR male mice. After 3 days, luciferin was injected i.p. in nonfasted mice and the luminescence from liver was captured with IVIS (left panel). The color overlay on the image represents the photons/s emitted from the animal with a range of 1.5 × 10⁵–1.5 × 10⁶ photons/s as indicated by the color scale next to the images. The amount of transduced adenoviral DNA was quantified by real‐time PCR technique on each group, and the luminescence of homogenized liver samples was normalized to transduced adenoviral DNA (right panel; n = 4–5 per group). Results are expressed as means ± SEM. *P < 0.05 and **P < 0.01. N.S., not significant.

To further evaluate the rs7074440 enhancer activity in hepatocytes in the in vivo setting, we performed animal experiments using the in vivo Ad‐luc analytical system originally developed by ourselves [22]. Adenoviruses containing SV40 promoter and reporter gene plus three copies of rs7074440 SNP region from normal and risk allele (Ad‐SV40pro‐Luc‐rs7074440x3) were transduced in the mouse livers. The quantification results of luciferase activities using both in vivo imaging system and liver homogenized sample demonstrated that the enhancer activity of normal (G) allele was significantly increased compared with risk (A) allele in the liver (Fig. 2C). These results further confirmed that rs7074440 normal (G) allele has a strong enhancer activity in hepatocytes.

TFEL genome‐wide screen of trans‐acting factors for rs7074440 identifies c‐Fos

To identify transcription factors that can bind to rs7074440 and potentially regulate the expression of TCF7L2, we screened 1588 genome‐wide transcription factors from the Transcription Factor Expression Library (TFEL), our original library that is composed of nearly all the transcription factors in the mouse genome [22]. As a result of TFEL scan, c‐Fos transcription factor was identified as an activating factor, because c‐Fos exhibited higher transcription activity against the normal (G) allele reporter plasmid compared with the risk (A) allele (Fig. 3A). To investigate strictly the transcriptional effect of c‐Fos on SNP rs7074440 region, we performed luciferase assays with the same reporter of normal (G) and risk (A) allele (SV40pro‐Luc‐rs7074440x3) co‐transfected with c‐Fos expression plasmid in HEK 293 cells. Consistent with the screening experiment, luciferase activity of normal (G) allele construct was significantly increased by c‐Fos but that of risk (A) allele was not changed (Fig. 3B).

Figure 3

Transcription factor expression library genome‐wide screen of trans‐acting factors for rs7074440 identifies c‐Fos. (A) TFEL genome‐wide screen of trans‐acting factors for rs7074440 from normal allele compared with risk allele in HEK293 cells. Screening with TFEL (named as TFEL scan) were performed by the co‐transfection of 9 or 10 clones with normal (renilla) and risk (firefly) SV40pro‐Luc‐rs7074440x3 plasmids per one well (1st TFEL scan). Continuously, individual clones selected in the 1st TFEL scan were examined by the co‐transfection of with SV40pro‐Luc‐rs7074440x3 plasmids (2nd TFEL scan). (B) Effect of c‐Fos on rs7074440 region was examined using the same luciferase reporter. HEK 293 cells were transfected with 3 tandem rs7074440 sequences of normal allele (G) or risk allele (A) inserted into SV40 promoter firefly luciferase reporter vectors (SV40pro‐Luc‐rs7074440x3). Firefly luciferase activity was normalized to SV40 promoter renilla luciferase activity and shown as a fold change to that of the SV40 promoter only firefly luciferase vector (n = 3 for each group). Results are expressed as means ± SEM. *P < 0.05. N.S., not significant.

C‐FOS directly binds to rs7074440

Next, to assess how the variations of the SNP alter protein‐DNA binding, we performed EMSA using a radiolabeled DNA probe of rs7074440 region and c‐Fos and c‐Jun protein synthesized by in vitro translation system. Figure 4A showed that a radiolabeled DNA probe containing the normal (G) allele of rs7074440 more strongly bound to c‐Fos and c‐Jun heterodimer (AP‐1) protein complex compared with the risk (A) allele. Furthermore, the chromatin immunoprecipitation analysis to confirm the endogenous binding of C‐FOS proteins to the rs7074440 loci in vivo revealed that C‐FOS protein actually bound to the rs7074440 within TCF7L2 gene in HepG2 cells (Fig. 4B). We also ascertained that HepG2 and Huh7 genome have rs7074440 normal allele (data not shown).

Figure 4

C‐FOS directly binds to rs7074440. (A) Electromobility shift assay with ³²P‐radiolabeled probe for rs7074440 (normal and risk alleles) and AP‐1 consensus sequence. In vitro translated c‐Fos and c‐Jun proteins were incubated with radiolabeled probe. The normal allele (G) has the same allele‐specific binding sequence as AP‐1, indicated by arrow. (B) ChIP analysis of rs7074440 in HepG2 cells. Chromatin fragments immunoprecipitated with an anti‐C‐FOS antibody were amplified by PCR with primers covering rs7074440. Immunoprecipitation with normal IgG were used as controls. Primers at 100 kb away from TCF7L2 gene were used as a nonspecific control. Quantification results of anti‐C‐FOS relative to IgG were obtained using real‐time PCR and the fold change was shown (n = 3 for each group). Results are expressed as means ± SEM. *P < 0.05. N.S., not significant.

To obtain further evidence that C‐FOS binds to rs7074440, we searched using the ENCODE data on the UCSC Genome Browser. As shown in Fig. S3, the transcription factor ChIP‐seq data from ENCODE project (UCSC Genome Browser on Human Feb. 2009 (GRCh37/hg19) Assembly) demonstrates that rs7074440 is overlapped with the peak indicating FOS transcription factor binding site.

These lines of evidence indicate that C‐FOS directly binds to rs7074440 normal (G) allele and that the risk (A) allele mutation attenuates its binding.

Knockdown of C‐FOS decreases TCF7L2 gene expression

To clarify the role of C‐FOS in the regulation of TCF7L2 gene expression in hepatic cells, we examined the effect of C‐FOS knockdown using adenoviruses in HepG2 and Huh7, whose genomes we conformed had rs7074440 normal (G) allele (data not shown). As shown in Fig. 5A,B, the shRNA adenoviruses targeting C‐FOS (Ad‐C‐FOSi‐1 and Ad‐C‐FOSi‐2) significantly decreased TCF7L2 mRNA levels. These data suggest that C‐FOS can regulate TCF7L2 expression via its binding to rs7074440 normal (G) allele in hepatic cell lines.

Figure 5

Knockdown effects of C‐FOS. (A,B) Infection of HepG2 (A) and Huh7 (B) cells with adenovirus expressing shRNA against the C‐FOS coding sequences (Ad‐C‐FOSi‐1 and Ad‐C‐FOSi‐2) decreased TCF7L2 mRNA expression. Target sequences of Ad‐C‐FOSi‐1 and Ad‐C‐FOSi‐2 are different. Cells were infected at 60 m.o.i. with Ad‐FOSi or control adenovirus expressing LacZi. RNA was prepared from the cells 96 h after the infection, and the indicated mRNAs were quantitated by real‐time PCR (n = 3 for each group). Results are expressed as means ± SEM. *P < 0.05 and **P < 0.01.

Discussion

In the present study, we demonstrated that the SNP rs7074440 is a candidate functional SNP altering TCF7L2 gene expression through interaction with a transcription factor C‐FOS in hepatocytes.

We searched for the functional SNPs highly linked with the GWAS lead SNP rs7903146 known as a susceptibility marker of type 2 diabetes in the liver because the data pointing at the liver are scarce compared with pancreatic beta cells, and successfully identified rs7074440 as a potential candidate of causal SNP as well as a novel enhancer controlling TCF7L2 gene expression in hepatocytes. The SNP rs7074440 is one of the several dozen SNPs within the intron 4 of TCF7L2 gene (Fig. 1B), located at the position 113 025 665, approximately 27‐kb downstream of rs7903146, on chromosome 10 (UCSC Genome Browser hg38). The minor allele frequencies (MAFs) of rs7074440 are 0.32, 0.22, 016, and 0.01 for EUR, AMR, AFR, and ASN, respectively, while those of the lead SNP rs7903146 are 0.31, 0.25, 0.25, and 0.03 for EUR, AMR, AFR, and ASN, respectively (data from HaploReg V4), showing a very similar pattern. In fact, these two SNPs are in strong linkage disequilibrium (r² = 0.85 in EUR) with each other (data from HaploReg V4).

As previous reports investigating the relationship between SNPs and their enhancer activities at the TCF7L2 locus, the lead SNP rs7903146 itself has been shown to affect gene expression in a mouse pancreatic MIN6 cell line [30, 31]. However, in this study, we did not found that the SNP rs7903146 has any enhancer activity in hepatic cell lines (Fig. 1D), suggesting that the involvement of rs7903146 is limited to pancreatic beta cells. Another SNP rs4132670 has also been reported to exhibit 1.3‐fold higher levels of enhancer activity in the Huh7 cell line [32], although this tendency was not observed in our study.

There are opposing views on the role of TCF7L2 in the regulation of glucose metabolism in the liver and in the pathogenesis of type 2 diabetes. In some literatures, it is considered that lower expression of TCF7L2 is preferable. For example, Boj et al. [19] reported that liver‐specific Tcf7 l2 knockout mice showed reduced hepatic glucose production during fasting and displayed improved glucose homeostasis when maintained on high‐fat diet. In contrast, Norton et al. reported that silencing of TCF7L2 induced a marked increase in hepatic glucose output, which was accompanied by significant increases in the expression of the gluconeogenic genes, while the overexpression of TCF7L2 reversed this phenotype and significantly reduced hepatic glucose output [33]. This tendency was reproduced by Oh et al. [20], where they showed that knockdown of hepatic TCF7L2 promoted increased blood glucose levels and glucose intolerance with increased gluconeogenic gene expression. Conversely, overexpression of a nuclear isoform of TCF7L2 in high‐fat diet‐fed mice ameliorated hyperglycemia with improved glucose tolerance. Ip et al. [21] also demonstrated that TCF7L2 serves a beneficial role in suppressing hepatic gluconeogenesis using a liver‐specific dominant‐negative TCF7L2 transgenic mouse model. These reports are consistent with our results that the risk allele mutation lowers TCF7L2 expression in hepatocytes, although we do not have any evidence for the functional role of the SNP.

After we identified rs7074440 as a novel enhancer controlling TCF7L2 gene expression, we found the corresponding transactivator using TFEL scan. TFEL is a plasmid library we independently developed in order to pair an enhancer and its transactivator [22], which enabled us to easily determine the correspondence between an enhancer and its transactivator. In the present study, we applied this method to find correspondence between SNPs and transcription factors and succeeded in identifying C‐FOS transcription factor, establishing a new method of ‘TFEL SNP scan’.

C‐FOS is a proto‐oncogene, which forms heterodimer with JUN to form the activation protein‐1 (AP‐1) complex and binds to DNA at AP‐1‐specific sites to affect the transcription of other genes [34]. C‐FOS is a member of Fos transcription factor family consisting of C‐FOS, FOSB, FOSL1, and FOSL2. In general, Fos family plays a role in cell differentiation and proliferation and are placed under the control of various growth factors including insulin [35]. In fact, C‐FOS and JUN can negatively regulate gluconeogenic gene expression in hepatocytes [36]. Given that TCF7L2 can suppress hepatic gluconeogenesis [21], our finding of C‐FOS‐TCF7L2 connection might be involved in this regulatory pathway, although further investigation is needed to clarify this point.

In conclusion, rs7074440 was identified as a candidate functional SNP linked with GWAS‐lead SNP for type 2 diabetes, which directly binds to C‐FOS to alter TCF7L2 gene expression. Through this approach, we found a novel enhancer regulating TCF7L2 expression.

Acknowledgements

This work was supported by MEXT/JSPS KAKENHI Grant Numbers 23116006 (Grant‐in‐Aid for Scientific Research on Innovative Areas: Crosstalk of transcriptional control and energy pathways by hub metabolites), 15H03092 (Grant‐in‐Aid for Scientific Research (B)), 21591123 and 18590979 (Grant‐in‐Aid for Scientific Research (C)), 26560392 and 16K13040 (Grant‐in‐Aid for Challenging Exploratory Research), and 03J10558 (Grant‐in‐Aid for JSPS Fellows) (to NY). It was also supported by research grants from the Uehara Memorial Foundation, Nakatani Foundation, ONO Medical Research Foundation, Takeda Science Foundation, Suzuken Memorial Foundation, Japan Heart Foundation, Kanae Foundation for the Promotion of Medical Science, Senri Life Science Foundation, Japan Foundation for Applied Enzymology, and Okinaka Memorial Institute for Medical Research (to NY).

Author contributions

NY conceived the experiments. XP performed the experiments under the guidance of YT and analyzed the data together with NY. XP, YT, and NY co‐wrote the paper. All authors discussed the results and commented on the manuscript.