To the Editor: Maturity onset diabetes of the young (MODY), and permanent and transient neonatal diabetes mellitus (NDM), are the most prevalent monogenic forms of diabetes. Dominant activating mutations in the KCNJ11 gene [1, 2], which encodes the inwardly rectifying Kir6.2 subunit of the ATP-sensitive potassium (KATP) channel expressed at the surface of the pancreatic beta cell, or in the ABCC8 gene [3], which encodes the SUR1 regulatory subunit of the channel, account for more than 30% of cases of NDM in the French population. So far, genes involved in the aetiology of MODY include glucokinase (GCK) [4] (also occasionally responsible for permanent NDM) and at least five beta cell-expressed transcription factors [5]. Common DNA polymorphisms in most of these genes have also been repeatedly associated with beta cell dysfunction and with type 2 diabetes in a polygenic context [6]. Recently, intronic single nucleotide polymorphisms (SNPs) of the transcription factor 7-like 2 (TCF7L2) gene have been associated with type 2 diabetes in different ethnic groups, including French white subjects [7, 8]. The physiological consequences of carrying the rs7903146 T allele, at risk for type 2 diabetes, remain unclear but there are some arguments supporting a primary effect on insulin secretion [8–10].

These data prompted us to evaluate the putative contribution of TCF7L2 in monogenic forms of diabetes. In this study, we sought variants by direct sequencing of all potential exons and intron–exon junctions of TCF7L2 in unrelated white individuals with still unexplained NDM (n = 28) and MODY (n = 17), as described in Table 1. Two other groups were also screened for mutations: 205 non-MODY early onset type 2 diabetes (EOD; age at diagnosis <40 years), and 93 normoglycaemic (NG) subjects (Table 1). The TCF7L2 gene (formerly known as TCF4) is subject to extensive alternative splicing [11], and its exon–intron structure was obtained using the Genbank sequences AJ270770 to AJ270778. Seventeen different exons have been described, eight of them being totally (exons 4, 13, 14, 15 and 16) or partially (exons 7, 9 and 17) alternatively expressed (Fig. 1). Alternative splicing events of exons 14 and 15 change the reading frame in exon 17, leading to three different COOH-terminal ends. When exons 14 and 15 are transcribed together, it creates a short COOH-terminal end (STOP 1). When only one of these two exons is transcribed, it creates instead a long COOH-terminal end (STOP 3) and when both are lacking, there is a medium COOH-terminal end (STOP 2). Nineteen PCR products were necessary to cover all exons and intron–exon junctions (Electronic supplementary material Table 1). Each fragment was sequenced using an ABI3730xl DNA analyser. Sequence variants were confirmed by re-amplifying and re-sequencing a fresh dilution of DNA. No mutations were found in unexplained NDM and MODY subjects. In the other screened subjects, six rare SNPs, all in the heterozygous state, were identified: one in the 5′-untranslated part of the gene (c.−156G > A), four in the coding region (c.963A > G, c.1,637C > A, c.1,674C > G and c.1,675C > G) and one in the 3′-untranslated region (c.2,015C > T) (Fig. 1). No rs numbers have been assigned to these rare SNPs according to the dbSNP build 126 database. No subjects were carriers of more than one SNP. Subjects carrying one of these mutations did not show any specific features, such as a more severe diabetic phenotype or differences in glucose-lowering treatment, compared with non-carrier EOD individuals.

Table 1 Clinical characteristics of patients with unexplained NDM, MODY or EOD, and NG subjects
Full size table
Fig. 1
figure 1

TCF7L2 gene variants and their potential impact on the different possible transcripts: The TCF7L2 gene sequence was obtained from the Genbank sequences AJ270770 to AJ270778. Seventeen exons have been described [11], eight of them being totally (exons 4, 13, 14, 15 and 16) or partially (exons 7, 9 and 17) alternatively expressed, as noted by hatched squares. Nineteen PCR amplifications were needed to cover the overall gene. The SNP locations are displayed by base numbers counting from the ATG-translation initiation codon (Human Genome Variation Society recommendation) and from the p-arm telomere of chromosome 10 (http://genome.ucsc.edu/cgi-bin/hgGateway). The codon locations are not displayed due to multiple possible transcripts. For each SNP, the variant nucleotide is underlined

Full size image

Therefore, it is unlikely that TCF7L2 contributes to unexplained MODY or NDM in the study sample of subjects of French white origin. In addition, no significant linkage with MODY was observed at the TCF7L2 locus on chromosome 10q25 in European families including a set of French families [12]. By re-analysing the markers nearest to TCF7L2 in French families only, the LOD-score values were consistently below −4 for every genetic model. These observations suggest that TCF7L2 mutations are not likely to be responsible for MODY in these families. Our study has also led to the identification of six SNPs, three of them only present in EOD subjects (c.−156G > A, c.963A > G and c.2,015C > T), which may be due to the low frequency of these mutations. Recently, four exonic SNPs have been reported in the TCF7L2 gene of patients with permanent neonatal diabetes [13]. We assumed that Groves et al. used the NM_030756 mRNA sequence as reference sequence. This reference sequence does not take into account exons 4, 15 and 16, as well as the beginning of exons 9 and 17 reported in sequences AJ270770 to AJ270778 submitted by Duval et al. [11]. Two SNPs were found to be common between the two studies, c.879A > G and c.1429C > A (according to the NM_030756 sequence used by Groves et al.) corresponding to c.963A > G and c.1,637C > A, respectively (according to the AJ270770 to AJ270778 sequences used in our study). The two other reported SNPs were not found within our subjects. Further studies, with more participants, are necessary to investigate the potential impact of these SNPs on TCF7L2 function and/or the WNT signaling pathway. Interestingly, the two non-synonymous polymorphisms (c.1,637C > A, Pro → Thr or His → Gln and c.1,674C > G, Pro → Arg) may change the protein conformation in the transcripts with long or medium COOH-terminal ends. In this regard, transcripts harbouring long COOH-terminal ends have been demonstrated to mediate TCF7L2 transcriptional repression, as they contain two specific binding domains for C-terminal binding protein (CtBP), an ubiquitous transcriptional repressor [11]. Additional studies are needed to determine which transcript isoforms are expressed in beta cells, given that some of the sequence variants could differentially affect TCF7L2 isoforms. However, at this stage, there is no functional argument supporting a role of these variants in beta cell function.