Introduction

Of all common risk variants identified for type 2 diabetes, variants in the TCF7L2 gene, which encodes a transcription factor in the morphogenic wingless-type MMTV integration site family (WNT) signalling pathway, confer the highest risk of developing the disease (reviewed in [1]). The exact role of TCF7L2 in the development of diabetes has not been determined, but several links between WNT signalling and insulin secretion and proliferation of human beta cells have been established (reviewed in [2]).

The TCF7L2 gene consists of 17 identified exons (Fig. 1) and is known to display a complex pattern of splice variants with several alternative exons and splice sites [3]. A perturbed splicing pattern of TCF7L2 has previously been demonstrated in renal cell carcinomas and linked to TCF7L2 target gene regulation [4]. Little is known about the splicing pattern of TCF7L2 in human islets, nor whether splicing is influenced by risk genotypes. Previous studies on correlation of TCF7L2 expression levels and risk genotypes have not taken into account the splicing pattern. Furthermore, most of these studies have investigated adipose and/or skeletal muscle tissue [5–9]. While experiments in vitro have shown that a reduction of TCF7L2 expression leads to impaired beta cell function [10, 11] we have previously observed increased expression of TCF7L2 in islets of Langerhans from diabetic patients [12]. However, in these previous studies no attempts were made to discriminate between different splice variants of TCF7L2, and differential expression of splice variants could be an explanation for these conflicting observations. Here, we present a detailed quantitative description of the TCF7L2 splicing pattern in primary human lymphocytes, skeletal muscle and subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT), as well as human islets of Langerhans, with particular focus on islets, and relate this to the rs7903146 genotype in the TCF7L2 gene.

Fig. 1
figure 1

Gene structure of TCF7L2 (a); the location of the risk SNP, rs7903146, in intron 4 is indicated. Exon structure (b) of TCF7L2; known binding sites of the transcription factor are indicated by shading. The exact location of the binding sites of HBP1 and SMAD4 is not known. c Schematic overview of the open reading frames of the major TCF7L2 splice variants in the five tissues investigated. The 3′-ends and the 5′-ends are shown separately (intervening sequences are invariant)

Full size image

Methods

Tissues

All tissue samples were from non-diabetic individuals. Islets were obtained from the Human Tissue Laboratory at Lund University Diabetes Centre from deceased donors (six female, 11 male), BMI 17.6–29.0 kg/m2, aged 26–73 years. Purity varied from 13% to 90%. The islets were culture in CMRL 1066 (ICN Biomedicals, Costa Mesa, CA, USA) supplemented with 10 mmol/l HEPES, 2 mmol/l l-glutamine, 50 μg/ml gentamicin, 0.25 μg/ml Fungizone (GIBCO, BRL, Gaithersburg, MD, USA), 20 μg/ml ciprofloxacin (Bayer Healthcare, Leverkusen, Germany), and 10 mmol/l nicotinamide at 37°C (5% CO2) for 1–9 days prior to RNA preparation. HbA1c levels were available for nine islet donors. SAT and VAT were obtained from bariatric surgery of obese individuals (19 female, two male), BMI 32.6–55.5 kg/m2, aged 20–61 years. Muscle biopsies and blood samples were collected from 18 individuals (all male), BMI 22.6–34.0 kg/m2, aged 30–46 years. Informed consent was obtained from all study participants. All islet donors had given consent to donate organs for medical research. All procedures were approved by the ethical committees at Uppsala and Lund Universities.

RNA isolation

RNA was isolated from islets using the AllPrep DNA/RNA Mini Kit (Qiagen, Valencia, CA, USA), from muscle using the RNeasy Fibrous Tissue Kit (Qiagen), from fat using the RNeasy Mini Kit (Qiagen), and from blood using the Tempus 12-Port RNA Isolation Kit and an ABI Prism 6100 Nucleic Acid PrepStation (Applied Biosystems, Foster City, CA, USA). Concentration and purity was measured using a NanoDrop ND-1000 spectrophotometer (A 260/A 280 > 1.8 and A 260/A 230 > 1.0) (NanoDrop Technologies, Wilmington, DE, USA). No sign of degradation was observed using agarose gel electrophoresis and Experion DNA 1K gel chips (Bio-Rad, Hercules, CA, USA).

Analysis of expression and splicing of the TCF7L2 gene

A detailed description of the procedure can be found in the Electronic supplementary material (ESM). Briefly, reverse transcription was performed using 1 µmol/l dT18 oligomer and 3 µmol/l random hexamer primers. Quantitative real-time PCR was performed using TaqMan chemistry according to the manufacturer’s recommendation (Applied Biosystems) on an ABI 7900HT sequence detection system (see ESM Table 1). All samples were analysed in triplicates (maximum accepted variation in C t value: 0.1 cycles). The absolute quantity was calculated using a dilution standard curve of an oligonucleotide template of known concentration [13] (see ESM Table 2).

Statistics

Data are expressed as means ± SD. Differences between genotypes and tissues were analysed using non-parametric Kruskal–Wallis and non-parametric Mann–Whitney U tests. Correlations were analysed using non-parametric Spearman’s tests. In all tests p < 0.05 was considered statistically significant. Statistical tests were performed with SPSS 16.0 software (SPSS, Chicago, IL, USA).

Results

Large differences in splicing patterns between different tissues

cDNA was prepared from human pancreatic islets, blood lymphocytes, skeletal muscle, SAT and VAT. Sequencing, RT-PCR and restriction cleavage analysis of TCF7L2 cDNA species reveals clear tissue-dependent differences in the splicing pattern (see ESM Figs 1 and 2, ESM Table 3).

The total amount of TCF7L2 mRNA was highest in islets, followed by fat, blood and muscle (Fig. 2a; see ESM Table 4).

Fig. 2
figure 2

The total amount (a) of TCF7L2 in each tissue and relative distribution of TCF7L2 splice variants determined by real-time PCR. b The 5′-end: real-time PCR over exon boundaries 3–4 (dark grey bars) and 3–5 (white bars). c The 3′-end: exon boundaries 12–17 (white bars), 12–13 (dark grey bars), 12–14 (light grey bars) and 12–15 (hatched bars). The main splice variants identified by each assay are listed in ESM Table 1. Data are presented as mean±SD. d Relative incorporation of TCF7L2 exon 4 in islets plotted against plasma HbA1c levels (r = 0.758; p = 0.018); white circles represent the rs7903146 CC genotype; grey circles represent the heterozygous (CT) genotype

Full size image

Human pancreatic islets show primarily splice isoforms containing exons 4 and 15

Pancreatic islets display a high incorporation of exon 4 (Fig. 2b and see ESM Table 4), which is located immediately upstream of the diabetes-associated polymorphism rs7903146 in the TCF7L2 gene (Fig. 1a, b). In islets, exon 4 is present in 62.3 ± 4.8% of the transcripts compared with 26.9 ± 2.7 to 33.2 ± 2.0% in the other tissues (p = 2.6 × 10−9). In the 3′-end, exons 14 and 15 tend to be mutually exclusive and tissue-specific, so that exon 15 predominates in islets and blood, whereas exon 14 is more common in fat and muscle tissues (Fig. 2c and see ESM Table 5). An overview of the major splice variants in each tissue is given in Fig. 1c.

Exon 4 incorporation in islets correlates with HbA1c levels

In islets, neither BMI (r = −0.109; p = 0.676) nor age (r = −0.173; p = 0.506) was significantly correlated to incorporation of exon 4. However, exon 4 incorporation in human pancreatic islets was significantly and positively correlated with HbA1c levels in blood (r = 0.758; p = 0.018; Fig. 2d) and with total amount of TCF7L2 (r = 0.583; p = 0.014). Neither BMI (r = 0.359; p = 0.157) nor age (r = 0.097; p = 0.711) was correlated with the total amount of TCF7L2 mRNA.

No significant differences in splicing pattern or total amount of TCF7L2 in islets were observed between carriers of different rs7903146 genotypes (CC vs CT; see ESM Fig. 3, Tables 4 and 5).

Discussion

The incomplete knowledge of the complex and tissue-specific TCF7L2 splicing pattern has hampered endeavours to detail the biological role of the transcription factor in glucose homeostasis and type 2 diabetes. We show that the tissues examined exhibit distinct variations in splicing patterns and that only a few of the theoretically possible variants are represented in each tissue. In the 5′-end there is a tissue-dependent variation in the presence or absence of exon 4. In the 3′-end, exon 14 and 15 largely appear to be mutually exclusive in the transcripts. It is noteworthy that these two exons are of the same length and share 63% identity at the mRNA level and 67% identity at the protein level, and the presence of both introduces a stop codon in exon 15. This leaves the impression that the two exons can replace one another. Also, they tend to be tissue-specific, exon 15 being preferred in blood and islets and exon 14 in fat and muscle, which also display an enhanced incorporation of exon 13. A characteristic of the islets is the high incorporation of exon 15 and, most notably, of exon 4, which is retained in ~62% of the cDNA transcripts, compared with ~33% or less in the other tissues studied.

TCF7L2 interacts with numerous binding partners and several binding sites have been mapped to the sequence (reviewed in [14]; Fig. 1b). None of the known sites overlap with parts corresponding to alternative exons, but variants without either exon 14 or 15 are predicted to produce protein isoforms without the C-terminal binding protein 1 (CtBP1) site encoded by exon 17 (Fig. 1c). As CtBP1 is a co-repressor, these shorter forms will be expected to possess reduced repression capacities. Since exon 4 comprises 69 nucleotides its presence does not affect the reading frame or, consequently, the primary structure of the rest of the protein. However, a difference of 23 amino acids can influence the three-dimensional structure and, thereby, the interaction of TCF7L2 with its binding partners.

The key issue addressed was whether the intronic single-nucleotide polymorphism (SNP) rs7903146 influences the splicing pattern. Although the splicing pattern differed between islets and other tissues with exon 4 being more predominant in islets, the splicing pattern was not significantly influenced by the rs7903146 genotype (ESM Fig. 3). The variation observed in islet tissue may reflect a number of factors, i.e. purity of the preparations, material from unmatched donors and differences in cultivation times. Significant correlations between TCF7L2 expression levels and age and BMI were not found, and differences in purity of the islet preparations did not seem to have a major influence on TCF7L2 splicing or expression levels (ESM). It is premature, though, to exclude an effect of genotype on splicing pattern as the power was limited to detect such a presumably subtle effect, and because no individuals were homozygous for the T allele. Also, the positive correlation between HbA1c levels and retention of exon 4 in islets makes it likely that the genotype effect is revealed only after plasma glucose levels are taken into account. Although this correlation does not prove causality it suggests a link between TCF7L2 splicing and plasma glucose levels.

We conclude that islets of Langerhans display a unique splicing pattern predominantly consisting of isoforms containing exons 4 and 15, the former being located immediately upstream of the rs7903146 polymorphism. However, we did not observe a significant effect of genotype on splicing pattern. Interestingly, the amount of exon 4 in islets correlated with HbA1c levels, which suggests a possible link between glucose and TCF7L2 expression.