Polymorphism rs11085226 in the Gene Encoding Polypyrimidine Tract-Binding Protein 1 Negatively Affects Glucose-Stimulated Insulin Secretion

Objective Polypyrimidine tract-binding protein 1 (PTBP1) promotes stability and translation of mRNAs coding for insulin secretion granule proteins and thereby plays a role in β-cells function. We studied whether common genetic variations within the PTBP1 locus influence insulin secretion, and/or proinsulin conversion. Methods We genotyped 1,502 healthy German subjects for four tagging single nucleotide polymorphisms (SNPs) within the PTBP1 locus (rs351974, rs11085226, rs736926, and rs123698) covering 100% of genetic variation with an r2≥0.8. The subjects were metabolically characterized by an oral glucose tolerance test with insulin, proinsulin, and C-peptide measurements. A subgroup of 320 subjects also underwent an IVGTT. Results PTBP1 SNP rs11085226 was nominally associated with lower insulinogenic index and lower cleared insulin response in the OGTT (p≤0.04). The other tested SNPs did not show any association with the analyzed OGTT-derived secretion parameters. In the IVGTT subgroup, SNP rs11085226 was accordingly associated with lower insulin levels within the first ten minutes following glucose injection (p = 0.0103). Furthermore, SNP rs351974 was associated with insulin levels in the IVGTT (p = 0.0108). Upon interrogation of MAGIC HOMA-B data, our rs11085226 result was replicated (MAGIC p = 0.018), but the rs351974 was not. Conclusions We conclude that common genetic variation in PTBP1 influences glucose-stimulated insulin secretion. This underlines the importance of PTBP1 for beta cell function in vivo.

Polypyrimidine tract-binding protein 1 (PTBP1), formerly known as heterogeneous nuclear ribonucleoprotein I (hnRNP I), enhances mRNA stability and translation, in addition to regulating other processes [13]. For instance, PTBP1 mediates splicing repression in a long list of pre-mRNAs and is thereby important for the regulation of alternative splicing in a tissue-specific manner [14,15]. In pancreatic b-cells and chromaffin cells, protein kinase A-induced phosphorylation of PTBP1 promotes its rapid shift from the nucleus into the cytoplasm [13,16,17]. Glucose stimulation of b-cells induces a similar nucleocytoplasmic translocation of PTBP1 [4,13,16]. By interacting with binding sites within the 39-and 59-UTRs (untranslated regions), cytosolic PTBP1, in turn, regulates mRNA localization, stability and the initiation of translation [14,[18][19][20]. Specifically, PTBP1 stabilizes the mRNA coding for insulin and other major insulin secretory granule components such as PCSK1, PCSK2, chromogranin A and ICA512 [4,13,16]. Accordingly, knockdown of PTBP1 in insulinoma INS-1 cells leads to the depletion of insulin secretory granule stores [4].
Impaired insulin secretion together with insulin resistance is a key factor in the pathogenesis of type 2 diabetes mellitus. Various mechanisms could be responsible for impairment of insulin secretion, including a decrease in the number of pancreatic bcells due to apoptosis as well as impaired exocytosis or biogenesis of insulin secretory granules [21,22]. Besides insulin transcription itself, the conversion of proinsulin into insulin and the cleavage of its C-terminal residues by carboxypeptidase E (CPE) are critical for insulin maturation [5,7,23,24]. An increased proinsulin/insulin ratio is a characteristic finding at the onset of type 2 diabetes [25,26]. Notably, PTBP1 promotes the translation of PCSK1 and PCSK2, but not those of CPE, whose translation is not regulated by glucose [4,8,13].
In view of these considerations, genetic variation in PTBP1 might affect insulin secretion or maturation in humans. To test this hypothesis, we examined the association of common genetic variants (minor allele frequency [MAF].5%) in PTBP1 with both insulin secretion and proinsulin-to-insulin conversion.

Participants
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. Twelve informative SNPs with minor allele frequency (MAF).5% are located in this region. Their HapMap linkage disequilibrium data (D9 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 59 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.

Selection of tagging SNPs and genotyping
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.

Oral glucose tolerance test (OGTT)
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].

Intravenous glucose tolerance test (IVGTT) and hyperinsulinemic-euglycemic clamp
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 (210, 25, 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].
Unless otherwise stated, data are given as mean 6 SD. Data that were not normally distributed were logarithmically transformed prior to statistical analysis. A Bonferroni-corrected pvalue,0.0127 was considered statistically significant according to the number of SNPs tested.
Hardy-Weinberg equilibrium was tested using x 2 test.
MAGIC genome-wide association (GWA) data 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 b-cell function (HOMA-B).

Analysis of linkage between polymorphisms
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.

Analysis of putative transcription factor binding sites
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 r 2 .0.8 (1000 genomes project pilot 1: SNPs rs10420407 and rs10420953 for SNP rs11085226; SNPs rs351976, rs8103323, rs8100561, and rs351995 for SNP rs351974).

Subject characteristics
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 nondiabetic. 10% of the subjects had impaired fasting glucose, 10% had impaired glucose tolerance and 7% had both.

Insulin secretion during the OGTT
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.

Insulin secretion during the IVGTT
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  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).

Replication of the association of SNP rs11085226 with insulin secretion in MAGIC GWA data
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

Discussion
In the present study, we found that the common variant rs11085226 in the PTBP1 gene is associated with significantly lower glucose-stimulated insulin secretion, but not with altered proinsulin-to-insulin ratio. This finding could be replicated in the huge data set from the MAGIC consortium. We furthermore found SNP rs351974 to be associated with insulin levels during the  IVGTT, but not with any of the OGTT-derived parameters in our study, or in the analyzed MAGIC data. The influence of PTBP1 rs11085226 in the IVGTT especially on the insulin levels within few minutes after glucose injection with diminished influence at later time points might explain why we could only detect effects on OGTT-derived parameters of insulin secretion that cover the first 30 minutes. The large sample size of MAGIC [32] was necessary to detect also effects of this SNP on fasting insulin secretion (HOMA-B), most likely because glucosestimulated insulin secretion plays only a minor role in the fasting state. Notably, the effect was not detected in C-peptide-based parameters (C-pep30, AUC C-pep /AUC Glc ). This discrepancy could be due to effects of the SNP on insulin clearance. So, we examined the association of PTBP1 rs11085226 with two estimates of insulin clearance (Cpep0/Ins0 and AUC Cpep /AUC Ins ). Both parameters were not associated with this SNP (p$0.3).
The association of PTBP1 SNP rs351974 with insulin levels (independently of its concomitant association with glucose) was only detectable in the IVGTT data but neither in OGTT data nor in MAGIC. Similar discrepancy between insulin levels after oral and i.v. glucose administration was noted for a polymorphism in SLC30A8 [34]. The underlying mechanism for this observation still remains open but one possible explanation might be that the incretin effect that is present only after oral glucose uptake compensates for the SNP effects and, thus, making them undetectable during an OGTT.
SNP PTBP1 rs11085226 is in linkage with two other SNPs, one in intron 3 (rs10420407 [A/G]) and one in exon 5 (rs10420953) (figure 1). The exonic variation is a silent mutation (N108N), thus does not change the amino acid sequence. Therefore, these variations are suggested to influence the expression of PTBP1 at the transcriptional level rather than altering the protein functions. For instance, they could influence the binding of transcription factors or repressors, however this is unlikely since the JASPAR database [33] contains no putative transcription factor binding sites around the SNPs or linked SNPs. Furthermore, the silent mutation in exon 5 could also affect RNA processing or protein translation rate. SNP rs11085226 could possibly affect alternative splicing, since it is located in an intron next to an alternative splicing site [14,35].
Since PTBP1 influences the expression of proinsulin convertases, we expected that genetic variations could affect proinsulin/ insulin ratios. However, such effects were not found. Interestingly, PTBP1 SNPs rs11085226 and rs351974 reduced insulin secretion very early after glucose administration but not at later time points. One might speculate therefore that the variation influences the number of insulin secretory granules within the readily releasable pool, which are mainly responsible for the first phase of insulin secretion [36]. Of interest, PTBP1 seems to be involved in impairment of this mechanism in type 2 diabetes [37]. The hypothesis of alterations in the pool of readily releasable granules is supported by the fact that PTBP1 supports the biogenesis of insulin secretory granules, while data in rodent b-cells indicate that newly-synthesized insulin secretory granules undergo preferential exocytosis ( [38,39]; Ivanova and Solimena, unpublished data). Moreover, PTBP1 may regulate the expression of proteins that are directly involved in the machinery for exocytosis of insulin secretory granules.
Our study could only detect an effect size $18% and was therefore underpowered to detect smaller effect sizes, e.g. on proinsulin conversion, and thus larger cohorts should be studied to assess such possible smaller effects. Further on, we focused on common variants with a MAF.5%. Therefore we might have overlooked effects of rare variants or mutations that could more dramatically influence PTBP1 levels or functions. Studying such rare variants will need much larger cohorts and direct sequencing strategies. Further clarification about the influence of PTBP1 SNPs on insulin secretory granule exocytosis will also require additional studies at the molecular level.
Taken together, our data suggest that the common genetic variation in PTBP1 affects very early glucose-stimulated insulin secretion. Thus, the effects of PTBP1 on insulin secretion that were discovered in model systems are detectable and relevant in humans in vivo, too.