Association of Six Single Nucleotide Polymorphisms with Gestational Diabetes Mellitus in a Chinese Population

Background To investigate whether the candidate genes that confer susceptibility to type 2 diabetes mellitus are also correlated with gestational diabetes mellitus (GDM) in pregnant Chinese women. Methodology/Principal Findings In this study, 1764 unrelated pregnant women were recruited, of which 725 women had GDM and 1039 served as controls. Six single nucleotide polymorphisms (rs7754840 in CDKAL1, rs391300 in SRR, rs2383208 in CDKN2A/2B, rs4402960 in IGF2BP2, rs10830963 in MTNR1B, rs4607517 in GCK) were genotyped using TaqMan allelic discrimination assays. The genotype and allele distributions of each SNP between the GDM cases and controls and the combined effects of alleles for the risk of developing GDM were analyzed. We found that the rs4402960, rs2383208 and rs391300 were statistically associated with GDM (OR = 1.207, 95%CI = 1.029–1.417, p = 0.021; OR = 1.242, 95%CI = 1.077–1.432, p = 0.003; OR = 1.202, 95%CI = 1.020–1.416, P = 0.028, respectively). In addition, the effect was greater under a recessive model in rs391300 (OR = 1.820, 95%CI = 1.226–2.701, p = 0.003). Meanwhile, the joint effect of these three loci indicated an additive effect of multiple alleles on the risk of developing GDM with an OR of 1.196 per allele (p = 1.08×10−4). We also found that the risk alleles of rs2383208 (b = −0.085, p = 0.003), rs4402960 (b = −0.057, p = 0.046) and rs10830963 (b = −0.096, p = 0.001) were associated with HOMA-B, while rs7754840 was associated with decrease in insulin AUC during a 100 g OGTT given at the time of GDM diagnosis (b = −0.080, p = 0.007). Conclusions/Significance Several risk alleles of type 2 diabetes were associated with GDM in pregnant Chinese women. The effects of these SNPs on GDM might be through the impairment of beta cell function and these risk loci contributed additively to the disease.


Introduction
Gestational diabetes mellitus (GDM) is defined as glucose intolerance with onset or first recognition during pregnancy [1]. It affects 5% to 10% of Asian women with an increasing trend observed in some developing countries, including China [2]. Epidemiological studies have confirmed that GDM is associated with increased feto-maternal morbidity and long-term complications in mothers and offsprings. However, the pathogenesis of GDM is still largely unknown. Given the fact that women with a history of GDM are at an increased risk of developing type 2 diabetes (T2D) later in their lives [3] and women with a family history of diabetes may be predisposed to an increased risk of GDM [4], GDM may share the same risk factors and genetic susceptibilities with T2D.
Considering the lack of sufficient evidence about the effect of candidate genes of T2D on GDM and the discrepancy among races [2,37], we aimed to determine whether the results derived from other races are applicable to the Chinese population, or whether different genetic background can cause different quantitative traits related to GDM. Moreover, the replication performed in multiple ethnicities did help to identify population specific risk variants. In the study, we selected six genes, including CDKAL1, IGF2BP2, MTNR1B, GCK, SRR and CDKN2A-CDKN2B, and tested their association with GDM in Chinese pregnant women. The results may provide additional insights to the mechanisms which underlie the genetic variants associated with the risk of GDM.

Ethnic statement
Written informed consent was obtained from each participants, and the study was approved by the Institutional Review Board of Peking Union Medical College Hospital.

Study subjects
During the period of 2006 to 2010, we recruited a total of 1,764 participants of Chinese pregnant women residing in Beijing from Peking Union Medical College Hospital. The pregnant women without a previous diagnosis of glucose intolerance were routinely screened for GDM between 24 and 28 weeks of gestation by two procedures. First of all, a 50 g glucose challenge test (GCT) was used as preliminary screening. It was considered as GCT negative (GCT2) if the plasma glucose concentration was less than 7.8 mmol/l after glucose intake 1 hour. Otherwise, diagnosed as GCT positive (GCT+). The pregnant women of GCT+ were then given a 100 g oral glucose tolerance test (OGTT). Diagnosis of GDM was based on the criteria as set by the American Diabetes Association [38]. The glucose threshold values were as follow: fasting 5.3 mmol/l, 1 h 10.0 mmol/l, 2 h 8.6 mmol/l and 3 h 7.8 mmol/l. A diagnosis of GDM was made if 2 or more of the glucose values met or exceeded the threshold value. Normal glucose tolerance (NGT) was diagnosed when all plasma glucose values were below the threshold values. Based on the above criteria, 725 subjects with GDM, 641 with NGT, and 275 GCT2 participants were recruited to the study. The NGT and GCT2 groups were taken as controls.

Clinical and biochemical data
Clinical and biochemical data of all subjects were collected at 24-28 weeks gestation. Clinical data included age, height, weight at one year before pregnancy, systolic blood pressure and diastolic blood pressure. The family history of T2D in each subject was also recorded. Body mass index before gestation (pre-BMI) was calculated as body weight (kg) divided by the square of height (m 2 ).
Homeostatic model assessment (HOMA) data and the area under the curve (AUC) of insulin during a 100 g OGTT performed at the time of GDM diagnosis were calculated to assess insulin resistance and beta-cell function. Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated by (FPI in mU/L6FPG in mmol/l)/22.5. Homeostasis model assessment of beta-cell function (HOMA-B) was calculated by (FPI in mU/L620)/(FPG in mmol/l23.5) as previously reported [39]. Another assessment index of beta-cell function, the AUC of insulin at 3 hour was evaluated according to the trapezoid method: V1+V2+0.5*V0+0.5*V3, where V is the insulin concentration at the indicated time [40].

Statistical analysis
The quantitative variable with normal distribution (platelet count) was given as mean 6 standard deviation (SD), and quantitative variables with non-normal distribution were given as medians and interquartile range. The continuous data (HOMA-B, HOMA-IR, AUC of insulin, total cholesterol, triacylglycerol, and HDL-cholesterol) were log-transformed to approximate normal distributions. Quantitative data with normal distribution or logtransformed variables were analyzed by student's t test. Nonparametric tests were performed to analyze the other variables.
The chi-square tests were used to determine whether individual polymorphism was in Hardy-Weinberg equilibrium. Genotypes were given codes of 0, 1 and 2, and the odds ratio (OR) was expressed per difference in the number of risk alleles. A multiple logistic regression model was used to investigate the individual effect of these genes on GDM. These analyses were based on additive, recessive and dominant models, and adjusted for age and the family history of type 2 diabetes. The ORs with 95% confidence intervals (CIs) were presented. Multiple linear regression models with adjustment for age were also applied to analyze these quantitative traits, and the regression coefficients (b) were presented. A two-sided p value ,0.05 was considered statistically significant. The statistical analyses were performed using SPSS 11.0 (SPSS Inc, Chicago, IL, USA).
The following assumptions were made for the power calculation: a prevalence of GDM equal to 3%, a high-risk allele frequency of 0.20, and an effect size of 1.3. By studying a sample of 725 cases and 1039 controls, our present study had more than 80% power, under a multiplicative model, with a type I error rate of 0.05. When the predisposing allele frequency was .30%, the study had at least 80% power to detect an OR of 1.22 under a multiplicative model. Power calculations were performed using the Genetic Power Calculator, available at http://ibgwww.colorado. edu/,pshaun/gpc/.

Clinical and biochemical parameters
The clinical and biochemical parameters of the control and GDM groups were presented in Table 1. Mean age, systolic and diastolic blood pressure, FPG, FPI, glycated hemoglobin protein, serum triacylglycerol, high sensitivity C-reactive protein, white blood cell and platelet counts were significantly higher in the GDM group than controls (p,0.001), whereas pre-BMI was similar in both groups (p = 0.086). In addition, women with GDM displayed higher HOMA-IR, lower HOMA-B and higher AUC of insulin (p,0.001).
Subsequently we tested the joint effects of risk alleles of susceptible loci on GDM to investigate if these loci affected the disease additively. Here we just selected SNPs with p values less than 0.05 (rs4402960, rs2383208 and rs391300) and calculated the joint effects by summing up the number of risk alleles for each participant who had the genotyping information of all these three SNPs. We found that the proportion of women with GDM increased in the subgroups with more risk alleles, the subgroups carrying more risk alleles had a significantly higher risk for GDM, with each additional risk allele increased GDM risk by 1.196-fold (95%CI = 1.092-1.309, P = 1.08610 24 ). Moreover, the subjects who harbor 4, 5 and 6 risk alleles have a 2.008-fold (p = 0.011), 5.576-fold (p = 3.31610 24 ) and 9.717-fold (p = 0.047) increasing in the odds of developing GDM as compared to individuals without any risk alleles, respectively. All the analysis was based on the adjustment for age and the family history of T2D.

FPG, HOMA-B, HOMA-IR and AUC of insulin
We analyzed the association between each SNP and quantitative traits in the research (as shown in table 3). The risk allele of rs10830963, rs2383208 and rs391300 showed association with increased FPG (p = 0.019, p = 0.034, p = 0.028, respectively). We further observed that these variants exerted combined effects on FPG, with a mean 0.087 mmol/L increase per risk allele (95%CI = 0.011-0.084, p = 0.012).
We also obtained homeostatic model assessment data for betacell function and insulin resistance (HOMA-B and HOMA-IR, respectively) in both groups. The loci, rs4402960 and rs10830963, were statistically associated with HOMA-B (p = 0.046, p = 0.001, respectively). For rs2383208, we found that carriers of genotype-AA showed lower level of HOMA-B (p = 0.002, the result was not listed in table 3). The significance remained in the recessive model using the logistic regression analysis (p = 0.003). Further analyses of the risk alleles (A-allele of rs4402960, A-allele of rs2383208 and G-allele of rs10830963) confirmed their joint effects on HOMA-B level (b = 20.108 unit per risk allele, 95%CI = 220.922,26.546, p = 1.86610 24 ). We just discovered that rs7754840 was significantly correlated with AUC of insulin (b = 20.080 mU l 21 6h per risk allele, 95%CI = 220.836,23.337, p = 0.007). For HOMA-IR, no significant association was detected.

Discussion
In the present study, we observed that some common variants conferring susceptibility to type 2 diabetes mellitus may increase the risk of GDM in pregnant Chinese women. The results also confirmed that the polymorphism in SRR was associated with GDM in the Chinese population for the first time.

IGF2BP2 rs4402960
Our results provided evidence that rs4402960 was a susceptible gene locus for GDM in Chinese pregnant women (OR = 1.207, 95% CI = 1.029-1.417, p = 0.021). This result was similar to that observed by a Korean GDM study (OR = 1.18, 95%CI = 1.01-1.38, p = 0.034) [23], but differed from that of the Danish (OR = 1.18, 95%CI = 0.97-1.42, p = 0.096) [24]. This discrepancy might be attributed to racial differences [2,37]. A recent study have showed an association between rs4402960 and T2DM in a case-control sample living in Beijing (OR = 1.19, 95%CI = 1.04-1.37, p = 0.009) [15]. The results of our study showed a similar effect which further suggested that GDM may share the similar genetic background with T2D.
It has been reported that the variants of IGF2BP2 can affect first-phase insulin secretion and the disposition index [29]. In our research we found that the subjects harboring the risk T allele of rs4402960 showed a negative association with HOMA-B (b = 20.057, p = 0.046) but not with HOMA-IR and fasting glucose level. The results indicated that common variation in IGF2BP2 mainly affected beta cell function rather than insulin sensitivity or fasting glucose level. It confirmed a previous study in women with GDM and in accordant with the dominant role of beta cell dysfunction in GDM [42,43].

CDKN2A-CDKN2B rs2383208
We found that rs2383208, a variant at the same LD block with rs10811661 [24], was a risk locus for GDM in Chinese population (OR = 1.242, 95% CI = 1.077-1.432, p = 0.003). The finding was consistent with the previous study on Korean GDM women [23], though the OR in our study was a little lower than that in their research. On the other hand, a lack of association at rs10811661 in Danish women with previous GDM was observed, partly due to a lower effect size (OR = 1.12, 95%CI = 0.87-1.45, p = 0.39) [24]. Because the relation between CDKN2A-2B and beta cell function has been widely reported in Western and Eastern populations [30,44,45], we also determined the association between rs2383208 and beta cell function. However, we only found that carriers of rs2383208 allele-A showed lower levels of HOMA-B, and lack association with AUC of insulin. The discrepancy may be largely caused by differences between the two indexes. The directly measured insulin data rather than surrogate measures HOMA-B and AUC of insulin may improve the specificity. In addition, our finding that rs2383208 was associated with FPG and HOMA-B, to some content, support the view that impaired beta cell function and hyperglycemia likely share the same underlying pathogenic mechanism [33,41]. Therefore, the risk allele of rs2383208 may be associated with an increased risk of GDM primarily by regulating the secretion of pancreatic beta cell and FPG.

SRR rs391300
SRR rs391300 was originally identified as genetic determinants of type 2 diabetes by GWA studies on Han Chinese in 2009 [16]. In our study, we detected its association with GDM in Chinese population and found a nominal role of this variant on the risk of GDM (OR = 1.202, 95%CI = 1.020-1.416, p = 0.028). In addition, we discovered that risk allele-T of rs391300 showed association with FPG (b = 0.059 mmol/L per allele, p = 0.028) but not with beta cell function or insulin resistance. All these findings indicated that SRR variant may affected the incidence of GDM by modulating the secretion of insulin and/or glucagon as reported previously [16].

MTNR1B rs10830963
We observed that rs10830963 was not associated with GDM in Chinese women. However, we found that this variant showed moderate association with HOMA-B (b = 20.096, p = 0.001) and FPG (b = 0.062, p = 0.019). Previous studies indicated that the MTNR1B variants were significantly associated with increased fraction of glycated hemoglobin and reduced beta-cell function (HOMA-B), and not related to fasting insulin level or insulin sensitivity [46,47]. One possible explanation was that MTNR1B may down regulate GCK expression and glucose-stimulated insulin secretion by lowering intracellular cAMP level [33,48,49]. Another study also demonstrated that rs10830963 was associated with GDM by affecting islet beta cell function and fasting glucose level [25]. In our study, carriers of the risk allele G of rs10830963 showed a lower value of HOMA-B and higher level of FPG. Our finding in the subpopulation further emphasized the importance of rs10830963 for beta-cell function and FPG.

GCK rs4607517
A study in Scandinavian women showed rs1799884 in GCK was a candidate locus for GDM [22]. We did not find the similar association between rs4607517 and GDM in our study though the rs4607517 and rs1799884 exhibited strong linkage disequilibrium [41]. Furthermore, we did not find the correlation between rs4607517 and quantitative traits (FPG and HOMA-B) as reported previously [33,41,47]. One possible explanation for this contrasting result may be attributed to how much GCK rs4607517 affected these traits. In our study, the effect size of rs4607517 on FPG (0.04 mmol/l) was a little smaller than that of the other three SNPs (0.055-0.062 mmol/l FPG per allele) and than that of rs4607517 in previous study (0.06 mmol/l per allele) [47].

CDKAL1 rs7754840
CDKAL1 was originally recognized as a candidate gene for T2DM by several GWAS [6,7,10,11,13]. The rs7754840 in Table 3. Associations between risk alleles and FPG, insulin beta cell function and insulin resistance.  CDKAL1 was associated with T2DM in the Chinese population (OR = 1.127, 95%CI = 1.027-1.238, P = 0.0119) [17]. Recent studies have indicated that the variation in CDKAL1 involved in the pathogenesis of GDM with an OR range from 1.22 to 1.55 [23,24]. Unfortunately, we did not find the relation between rs7754840 and GDM in our study. A relatively lower effect (OR = 1.117, 95%CI = 0.969-1.289, p = 0.127) might result in this inconsistent conclusion. Previous study indicated that CDKAL1 probably plays a role in the regulation of insulin secretion, even under glucotoxic conditions [24,35,36]. In the study, we found that rs7754840 showed significant association with insulin AUC (b = 20.080 mU l 21 6h, p = 0.007), which was consistent with a Korean study [23] and further indicated the role of CDKAL1 variants on beta cell function.

Combined genetic risk of GDM
Individuals carrying more risk alleles had a higher risk of type 2 diabetes [17,34]. This additive effect of the variants on GDM with an OR of 1.18 per risk allele (95% CI = 1.10-1.27, P = 3.2610 26 ) was also observed by Lauenborg [24]. In our study, subjects who harbor more than 4 risk alleles have at least a 2.008-fold increase for developing GDM as compared with individuals who did not carry any risk alleles. Similarly, the combined effects of the SNPs on HOMA-B and FPG were much obvious in contrast to the effects of single SNP as described in our research. These results support the finding of an additive effect of the type 2 diabetes risk alleles on the risk for GDM.
There are some limitations in the present study. First, although the study included 725 women with GDM and 1039 controls, the statistical power of the sample was not large enough to detect a weak effect size (OR,1.2). As a result, some associations may have been overlooked. Second, it was not confirmed whether all of the subjects in the control group had experienced pregnancy without GDM. In our study population, there were 271 GCT (2) women who were not given the 100 g OGTT after a 50 g glucose challenge test. However, the effects of this overlook on our interpretation of the results should be minimal because the prevalence of GDM in GCT (2) pregnant women was estimated to be very low [50].
Our study demonstrated that several previously proven type 2 diabetes risk alleles were associated with GDM in pregnant Chinese women. The study also provided evidence of the strong genetic background for the development of GDM in a multigenetic manner. Compared to women who did not harbor any risk allele, women carrying at least five or more risk alleles had a higher risk of developing GDM. The effects of these SNPs on GDM may be through the impairment of beta cell function. Further studies are required to assess the relationship between these polymorphisms and GDM in other ethnicities.