The participants were selected from the ongoing Tübingen Family Study, which currently includes ∼2,000 individuals at increased risk of diabetes [27]. Individuals on medication affecting glucose metabolism were excluded. Inclusion of the participants in the present study was based on availability of: (i) DNA samples for genotyping (n = 1,750); and (ii) complete OGTT data (glucose, insulin, C-peptide and proinsulin levels available for all time points during the OGTT, n = 1,502).

The anthropometric characteristics of the study population are shown in Table 1. Most (69%) of these subjects had a family history of diabetes, i.e. at least one second-degree relative with type 2 diabetes mellitus. All participants were genotyped for the following SNPs: PTBP1 rs351974, rs11085226, rs736926, and rs123698. Informed written consent was obtained from all participants and the local Ethics Committee (University of Tübingen) approved the protocol.

Using the publically available phase II data of the International HapMap Project derived from a population of Utah residents with ancestry from Northern and Western Europe (release 24, November 2008, http://www.hapmap.org, [28]), we screened in silico the complete PTBP1 gene (14.92 kb, 15 exons, 14 introns, located on human chromosome 19p13.3) and 5 kb of both the 3′ and 5′ flanking regions (Figure 1).

Twelve informative SNPs with minor allele frequency (MAF)>5% are located in this region. Their HapMap linkage disequilibrium data (D′ and r-squared values) are shown in Figure 1. Among these, we selected four tagging SNPs that cover 100% of the common genetic variation (MAF>5%) within this locus for genotyping: PTBP1 rs351974 (A/C) is located in the 5′ flanking region of the gene, rs11085226 (A/G) and rs736926 (C/T) are located within intron 2, while rs123698 (C/G) is located in intron 10.

DNA from whole blood was isolated using a commercial DNA isolation kit (NucleoSpin, Macherey& Nagel, Düren, Germany). Genotyping was done using the TaqMan assay (Applied Biosystems, Forster City, CA, USA). The TaqMan genotyping reaction was amplified on a GeneAmp PCR system 7000 and fluorescence was detected on an ABI PRISM 7000 sequence detector (Applied Biosystems). The genotyping success rate for PTBP1 rs351974 was 99.9%, for rs11085226 was 99.9%, for rs736926 was 99.8% and for rs123698 was 99.4%. Genotypes were verified in 50 randomly selected subjects by bidirectional sequencing, and both methods gave identical results.

After an overnight fast of 10 h, subjects ingested a solution containing 75 g glucose at 08:00 hours. Venous blood samples were obtained at 0, 30, 60, 90, and 120 min and plasma glucose, insulin, C-peptide, and proinsulin concentrations were determined. Insulin sensitivity during the OGTT was estimated according to Matsuda and DeFronzo [29].

In a subgroup of informed subjects (n = 320), an IVGTT was performed after a 10-h overnight fast, as described earlier [30]. After baseline samples had been collected (−10, −5, and 0 min), a glucose dose of 0.3 g/kg body weight was given at time point 0. Blood samples for the measurement of plasma glucose and insulin were obtained at 2, 4, 6, 8, and 10 minutes. In these subjects, a hyperinsulinemic-euglycemic clamp for the determination of insulin sensitivity was performed as well [31].

Blood glucose was determined using a bedside glucose analyzer (glucose oxidase method; Yellow Springs Instruments, Yellow Springs, OH, USA). Plasma insulin and proinsulin were determined by microparticle enzyme immunoassay (Abbott Laboratories, Tokyo, Japan, and IBL, Hamburg, Germany, respectively). The proinsulin assay has 0% cross-reactivity with human insulin and C-peptide. The insulin assay has 0% cross-reactivity with proinsulin.

The area under the curve (AUC) AUC_Ins0–30/AUC_Glc0–30 was calculated as (Ins0+Ins 30)/(Glc0+Glc30). Insulinogenic index (IGI) was calculated as (Ins30 – Ins0)/(Glc30 – Glc 0). Cleared insulin response (CIR) was (100×Ins30)/Glc30×(Glc30 – 3.89). AUC was calculated according to the trapezoid method as: 0.5×(C₀/2+C₃₀+C₆₀+C₉₀+C₁₂₀/2) with c = concentration.

Unless otherwise stated, data are given as mean ± SD. Data that were not normally distributed were logarithmically transformed prior to statistical analysis. A Bonferroni-corrected p-value<0.0127 was considered statistically significant according to the number of SNPs tested.

Hardy-Weinberg equilibrium was tested using χ² test.

Our study was sufficiently powered (1–β>0.8, α = 0.05) to detect effect sizes ≥0.18 for the SNP with the lowest minor allele frequency (rs11085226). Power calculations were performed using G*power software (www.psycho.uni-duesseldorf.de/aap/projects/gpower/). Pairwise linkage disequilibrium (D′, r²) was determined using the Java linkage disequilibrium plotter (http://www.genepi.com.au/projects/jlin). For all other statistical analyses, the software package JMP 7.0.1 (SAS Institute, Cary, NC, USA) was used.

GWA-data on glycaemic traits have been made publicly available by the MAGIC consortium [32] and have been downloaded from www.magicinvestigators.org. We interrogated data from the meta-analysis results for the homeostatic model assessment of β-cell function (HOMA-B).

A search for SNPs in linkage disequilibrium with SNP rs123698was performed on the Broad Institute's SNAP(SNP Annotation and Proxy Search) website (http://www.broadinstitute.org/mpg/snap/ldsearch.php) using the CEU panel from both the 1000 genomes project pilot 1 dataset as well as well as HapMap release 22.

For these analyses, the JASPAR (http://jaspar.genereg.net) database [33] “homo sapiens” was used with a relative profile score threshold of 80%. We analyzed the 50 bp around each of the two SNPs that showed associations with insulin levels during IVGTT and all linked SNPs with r²>0.8 (1000 genomes project pilot 1: SNPs rs10420407 and rs10420953 for SNP rs11085226; SNPs rs351976, rs8103323, rs8100561, and rs351995 for SNP rs351974).

The participants' clinical characteristics are presented in Table 1. In this group of relatively young subjects 66% were females and 34% males. They were slightly overweight and non-diabetic. 10% of the subjects had impaired fasting glucose, 10% had impaired glucose tolerance and 7% had both.

We analyzed the following tagging SNPs: rs351974, rs11085226, rs736926, and rs123698 in PTBP1. All SNPs were in Hardy-Weinberg equilibrium (all p≥0.3) and showed MAFs very close to those reported by the HapMap project (http://www.hapmap.org, [28], Table 2).

The minor allele of SNP rs11085226 in PTBP1 was nominally associated with lower insulin secretion measured as IGI and CIR during the OGTT in the dominant inheritance model (p≤0.04; Table 3). Whereas the third insulin-based parameter, i.e., AUC_Ins0–30/AUC_Glc0–30 revealed a trend for association, there was no association with C-peptide-based measures (Table 3). The three other tested SNPs in the PTBP1 gene did not show any association with any of the tested parameters of insulin secretion (all p>0.07; Table 3 and 4). None of the analyzed SNPs was associated with proinsulin levels or proinsulin-to-insulin ratio. In our data, we detected no significant association with HOMA-B.

The minor allele of SNP rs11085226 showed significant association with lower insulin levels within the first ten minutes following glucose injection (AUC_{0–10 min} p = 0.0103, dominant inheritance model; Figure 2 A), while glucose levels did not diverge (AUC_{0–10 min} p = 0.3; Figure 2 B).

Neither SNP rs736926, nor SNP rs123698 were associated with insulin levels during the IVGTT (AUC_{0–10 min} p = 0.4 and p = 0.7, respectively), in a model adjusted for gender, age, BMI, and clamp-derived insulin sensitivity index. Performing the same adjustment, SNP rs351974 was associated with insulin levels during the first 10 minutes of the IVGTT (AUC_{Insulin 0–10 min} p = 0.0108, dominant inheritance model). SNP rs351974 was furthermore associated with glucose levels during the same time period after adjustment for gender, age, and BMI (AUC_{Glucose 0–10 min} p = 0.0172, dominant inheritance model). Both, glucose and insulin levels were lower in carriers of the minor allele of SNP rs351974. The association of this SNP with AUC_{Insulin 0–10 min} remained present after adjustment for AUC_{Glucose 0–10 min} in addition to gender, age, BMI, and clamp-derived insulin sensitivity index (p = 0.0377).

In the publicly available data on glycaemic traits from the MAGIC consortium on 46,186 non-diabetic participants [32], SNP rs11085226 was associated with the provided measure of insulin secretion, HOMA-B (p = 0.01815) with higher insulin secretion in carriers of the A-allele. Of the other three investigated PTBP1 SNPs, rs736926 showed association with HOMA-B in MAGIC, too (p = 0.04164), and SNP rs351974 did not show such an association (p = 0.7). Since SNP rs123698 is not included in the provided MAGIC data set, we searched for linked polymorphisms (r²>0.8)in 1000 genomes pilot 1, as well as HapMap release 22 but found no such linked SNPs. Since we were unable to identify another polymorphism as a proxy for PTBP1rs123698, we could not check for associations between this SNP and HOMA-B in MAGIC data.