Introduction

Gestational Diabetes Mellitus (GDM) is defined as a condition of carbohydrate intolerance of varying severity that begins or is first recognized during pregnancy and is a common obstetric complication. It is characterized by an impaired compensatory increase in insulin secretion to overcome the pregnancy-induced insulin resistance. Nevertheless, the metabolic pathways involved need to be better understood. Previous reports indicate that GDM is a strong risk factor for developing Type 2 Diabetes (T2D) later in life, and most importantly, it also influences the metabolic health of their offspring in the short and long term [1].

In Mexico, no nationwide prevalence has been reported for GDM due to the lack of both detection programs and a consensus diagnostic method. However, a recent study found a prevalence of 12.9% [2], compared to a T2D prevalence of 24.2% among individuals under 40 years old, both reported in 2012 [3]. Physiological similarities between T2D and GDM suggests that they could share a common genetic background. It has been reported that common T2D risk genetic variants confer predisposition to GDM development in Caucasian and Asian populations [4, 5], but up to date, there is scarce information in other ethnic groups. Common T2D risk genetic variants also associated to GDM risk include: TCF7L2 (rs7903146), MTNR1B (rs10830962, rs10830963), IGF2BP2 (rs4402960, rs1470579), KCNJ11 (rs5219), CDKAL1 (rs7754840, rs7756992), KCNQ1 (rs2237892, rs2237895), GCK (rs4607517), CDKN2A/2B (rs2383208, rs10811661), SRR (rs391300), HHEX (rs1111875, rs5015480, rs7923837), SLC30A8 (rs13266634), TCF2 (rs7501939).

Although T1D and GDM genetic similarities have been poorly explored, the evidence point out that autoimmunity is not likely to play a major role in GDM development. Although few reports have been published, none of them have shown a high prevalence of T1D risk genotypes in non-autoimmune GDM patients compared to the normoglycemic control group. Moreover, a low percent of GDM patients have shown to be positive to T1D autoimmune markers and pregnancy hyperglycemia are prone to develop T2D, but not T1D [6–9].

It has been shown that ethnicity is an important factor determining the risk of developing metabolic disorders. Particularly, Native American ancestry positively correlates with higher prevalence of T2D and GDM [10–12], a fact that may be related not only to lifestyle, diet and healthcare access, but also to genetic factors. Despite the fact that common T2D susceptibility variants are homogeneous within ancestry groups, their frequency and risk effects have been found to differ across populations [13].

A recent report showed that the inclusion of subjects with high Native American ancestry in a genome wide association study, allowed the identification of new T2D risk alleles, which are rare or absent in other human populations [14]. Furthermore, despite the inclusion of large sample sizes, some of the previously reported T2D associations could not be replicated in Native American derived populations, thus highlighting the importance of assessing the contribution of the currently known risk variants in subjects with different ethnic backgrounds.

In contrast to T2D, genetics of GDM has been less studied. However, a concordance of risk alleles, as well as direction of their effect has been reported. Regarding Mexican population, GDM genetic studies have been conducted for few candidate genes and including small sample sizes. Variants within TCF7L2, TNF and HNF4A loci have been associated to the risk of developing the disease in Mexican women [15–18]. However, no study has evaluated a complete set of the most associated T2D risk variants.

The aim of this work was to analyze whether common genetic variants, previously associated to T2D, obesity and other related traits, were also associated to the risk of GDM in Mexican population.

Methods

Results

Our GDM sample included 750 Mexican women (408 cases and 342 controls). Cases were younger than controls and showed a higher BMI (P<0.001). After adjusting for age, pregestational BMI and sampling gestational week, cases showed higher triglycerides levels (P = 0.001), but lower total HDL and LDL cholesterol levels (P<0.001, respectively). Although this finding may be interpreted as a good metabolic control, some authors have hypothesized that it could be due to incapacity to develop the physiological hyperlipidemia of the pregnancy, and that estrogens levels may play an important role [60–62]. Cases were also more insulin resistant, as demonstrated by their higher levels of HOMA IR, glucose, insulin and leptin levels (P = 0.002, <0.001, 0.005, 0.007, respectively) and by their lower values of Gutt index and adiponectin levels (P = 0.006, 0.002, respectively). Additionally, a higher proportion of cases reported to have family history of diabetes, as well as college education level and employment (P<0.001, 0.033, <0.001, respectively) (Table 1). As one of the objectives was to compare our findings to those reported for T2D Mexican individuals, we contrasted our data with that from the SIGMA T2D Consortium study [14] which included 3027 women (560 cases ≤48 years old and 2467 controls).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. GDM dataset description.

https://doi.org/10.1371/journal.pone.0126408.t001

Even though T2D prevalence has been consistently correlated with a higher proportion of Native American ancestry and low socioeconomic status (SES), no statistical differences between GDM cases and controls were observed (Native American ancestry: 65.4% in controls vs. 63.4% in cases, P = 0.300; and low SES: 32.8% in controls vs. 44.4% in cases, P = 0.272) (Table 1). In contrast, women-only SIGMA T2D cases had a higher Native American ancestry proportion when compared to controls (51.8% in controls vs. 70.9% in cases, P = 1.02x10^-73) (S1 Fig). Fig 1 shows a PCA projection of GDM and women-only SIGMA T2D samples onto parental populations, as well as the centroids of eigenvectors 1 and 2 for cases and controls. A comparative plot of global ancestry proportions by group is in S1 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Principal component analysis projection of GDM and SIGMA samples.

Principal components were calculated using parental data from the Human Genome Diversity Project (HGDP) and the Mexican Genome Diversity Project (MGDP). Centroids of eigenvectors 1 and 2 for cases and controls are plotted in black triangles for GDM dataset and in black circles for women-only SIGMA dataset.

https://doi.org/10.1371/journal.pone.0126408.g001

We first examined 176 SNPs previously associated to the risk of T2D, GDM, high BMI and adverse pregnancy traits in other populations. Six variants within two loci reached statistical significance for the association to GDM risk after FDR multiple comparisons adjustment (Table 2): TCF7L2 (rs7901695, Q = 0.00314, OR = 2.14[1.47–3.11]; rs4506565, Q = 0.00314, OR = 2.06[1.45–2.94]; rs7903146; Q = 0.00324, OR = 2.07[1.44–2.98]; rs12243326, Q = 0.00061, OR [95%CI] = 2.89[1.85–4.52]) and KCNQ1 (rs2237892, Q = 0.007, OR = 0.58[0.43–0.78]; rs2237897, Q = 0.00314, OR = 0.54[0.40–0.72]). The allele effect sizes and frequencies of these SNPs are consistent with previous studies of GDM and T2D [14, 63]. When stratifying by pregestational BMI (lean vs. non-lean), association analyses did not reveal a modulation of the TCF7L2 and KCNQ1 allele effects sizes (S1 Table). In addition, eight genetic variants showed nominal statistical significance (P<0.05): KLF14 (rs972283), KCNQ1 (rs163184), FAIM2 (rs7138803), HHEX (rs5015480), GRB14 (rs3923113), DUSP9 (rs5945326), FTSJD1/CALB2 (rs6499500), PEPD (rs3786897) (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Genetic variants associated to the risk of GDM in Mexican women.

https://doi.org/10.1371/journal.pone.0126408.t002

Next, we tested the haplotypes of the TCF7L2 and KCNQ1 for association to GDM and various related metabolic quantitative traits. TCF7L2 risk haplotype CTTC (r² rs7901695 vs. rs4506565 = 0.96, r² rs7901695 vs. rs7903146 = 0.88, r² rs7901695 vs. rs12243326 = 0.55, r² rs4506565 vs. rs7903146 = 0.92, r² rs4506565 vs. rs12243326 = 0.58, r² rs7903146 vs. rs12243326 = 0.60) was associated to the risk of GDM and to higher levels of fasting glycemia, 60’ and 120’ OGTT glycemia and area under the curve. Similarly, apart from association to GDM, KCNQ1 protection haplotype TTT (r² rs2237892 vs. rs163184 = 0.41, r² rs2237892 vs. rs2237897 = 0.91, r² rs163184 vs. rs2237897 = 0.39) was also associated to lower levels of 60’ and 180’ OGTT glycemia, and area under the curve (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Haplotypes associated to the risk of GDM and related metabolic traits.

https://doi.org/10.1371/journal.pone.0126408.t003

We then examined the 176 SNPs for association to fluctuations of glycemic metabolic quantitative traits. Two variants reached statistical significance after FDR multiple comparisons adjustment: CENTD2 (60’ OGTT glycemia: rs1552224, Q = 0.02599), MTNR1B (HOMA B: rs1387153, Q = 0.03881) (Table 4 and S2 Table). Interestingly, MTNR1B rs1387153 variant was nominally associated to GDM in non-lean women subset (lean: OR = 0.888, P = 0.6337 vs. non-lean: OR = 1.761, P = 0.01297, t test P value = 0.0432) (S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Genetic variants significantly associated with metabolic trait fluctuations in pregnant Mexican women.

https://doi.org/10.1371/journal.pone.0126408.t004

We were also interested in comparing the risk effects of TCF7L2 and KCNQ1 variants between GDM and T2D. For this purpose, we compared the Beta coefficients obtained in the T2D case-control group from SIGMA T2D Consortium study [14] with those obtained in GDM. As previously reported, both loci reached genome-wide significant association to T2D in Mexican population, TCF7L2 (rs7903146, P = 2.5x10^-17, OR [95%CI] = 1.41[1.30–1.53]) and KCNQ1 (rs2237897, P = 4.9x10^-16, OR = 0.74[0.69–0.80]) [14]. The effect of TCF7L2 risk alleles was significantly different between the women-only T2D and GDM patients (t test P<0.01). In contrast, no difference was found in the risk effect of the KCNQ1 variants between the two groups (t test P>0.1) (Table 5). S3 Table shows allelic frequencies for the analyzed datasets.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Risk allele effect comparison between GDM and T2D samples.

https://doi.org/10.1371/journal.pone.0126408.t005

Despite the association with GDM observed for TCF7L2 and KCNQ1, no association was found for SLC16A11 haplotype, which was recently identified as a major contributor for T2D risk in Mexican population [14]. Even though the frequency of SLC16A11 risk variants in GDM cases was found to be similar to that of women-only SIGMA T2D cases (P>0.1), no statistical difference was found for the SLC16A11 risk variants between GDM cases and controls (P>0.1) (S3 Table). As all association analyses included a previously validated set of AIMs, lack of association was not due to a deficient population stratification control.

Finally, considering that variants near or within 59 gene loci showed nominal association to the risk of GDM or to metabolic and pregnancy related traits, we performed a Gene Set Enrichment Analysis (GSEA) using an association P value threshold <0.01. Besides evident related GO biological processes (i.e. glucose homeostasis, negative regulation of signal transduction, positive regulation of metabolic process, regulation of insulin secretion, generation of precursor metabolites, energy reserve metabolic process, feeding behavior), the analysis revealed two additional interesting processes: transcription regulation from RNA polymerase II promoter and gene-specific transcription (P<0.05). It is noteworthy that all genetic variants that reached statistical significance after FDR multiple comparisons adjustment, are located either in coding untranslated regions or within non-coding sites (S4 Table).

Discussion

Our results support a common genetic background for T2D and GDM. Not only did we find TCF7L2 and KCNQ1 variants to be associated to GDM, but GDM cases reported higher frequency of family history of diabetes than controls. In the recent SIGMA T2D GWAS in Mexicans, it was demonstrated that common variants within the TCF7L2 and KCNQ1 genes were among the strongest associated loci [14]. In the current study, associated genetic variants in these two loci were directionally consistent with T2D results previously reported for Mexicans and other populations [5, 14, 15, 64]. Also, we found that allele effect sizes of TCF7L2 variants were significantly different between GDM and T2D, suggesting a major risk effect of this locus in the development of GDM. Additionally, when analyzed variants within TCF7L2 and KCNQ1 were grouped in LD blocks, the TCF7L2 risk haplotype was significantly associated with increased glucose levels during OGTT; whereas, the KCNQ1 protection haplotype, was associated with decreased glycemia.

TCF7L2 gene encodes for the TCF4 transcription factor, involved in the Wnt signaling pathway. Risk variants are located in an intronic 92 kb interval, which has been reported to confer the strongest effect on T2D risk among the more than 90 T2D associated loci to date. Despite the mechanisms through which TCF7L2 influences disease development have not been fully understood, it has been proposed to be related to a reduced insulin secretion. One study showed that the rs7903146 risk allele T is located within a region of open chromatin and increased enhancer activity [65], thus increasing gene expression in pancreatic islets and presumably other tissues [66]. Given that others have reported downregulation of TCF7L2 gene expression, risk variants have been proposed to alter TCF7L2 splicing patterns, suggesting that functionally distinct mRNA isoforms, rather than levels of expression, define their phenotypic consequences [67]. In the case of GDM, expression of TCF7L2 was decreased in visceral adipose tissue from women who developed hyperglycemia, although after adjusting for BMI, the difference was not significant [68].

KCNQ1 gene encodes the Kv7.1 voltage-gated potassium channel. Even though it is known to play an important role in shaping the cardiac action potential, as well as in controlling the water and salt homeostasis in epithelial tissues [69], its role in diabetes development is not yet clear. KCNQ1 is located at the imprinted 11p15.4 region and its expression is monoallelic during early embryogenesis but turns biallelic during fetal heart development [70]. Two independent signals within introns 11 and 15 have been associated to T2D development, but only when risk alleles are maternally inherited [71]. KCNQ1 region is regulated by differential methylation at the promoter of KCNQ1OT1, a paternally expressed non-coding antisense RNA, which results in maternal expression of neighbouring genes. Risk allele of rs2237895 located within intron 15 was associated with impaired exocytosis and a reduced amount of docked granules of insulin, especially among non-obese allele carriers [72]. The same risk allele was reported to be associated with hypermethylation of the DMR and CTCF binding region near KCNQ1OT1 in fetal pancreas, but no relationship with risk allele number and gene expression was found [71].

Apart from TCF7L2 and KCNQ1, seven additional loci previously related to T2D (KLF14, HHEX, GRB14, DUSP9 and PEPD), increased BMI (FAIM2) or GDM (FTSJD1/CALB2), were nominally associated to the risk of developing GDM. In contrast to T2D genetic studies, those performed for GDM have yielded scarce results. This is due in part, to the inclusion of smaller sample sizes, sample heterogeneity as a result of different inclusion criteria and diagnostic methods. Although most of GDM genetic studies have been preferentially conducted for candidate genes, meta-analyses have consistently revealed variants near previously T2D associated genes, i.e. TCF7L2, MTNR1B, IGF2BP2, KCNJ11, CDKAL1, KCNQ1, GCK and IRS1 [63, 64]. In the only published GWAS for GDM performed in Korean women, the authors found association to common T2D variants and identified novel signals with marginal association [5]. It is of interest that one of them, FTSJD1/CALB2, was also nominally associated in our sample.

Of particular relevance is the finding of a differential role of SLC16A11 gene locus in the susceptibility to T2D and GDM. It was recently reported that SLC16A11 has a major contribution to T2D risk in Native American derived populations, as risk associated variants are found at a substantially higher frequency in individuals with Native American ancestry [14]. In the present study, no statistical difference was observed in Native American ancestry proportions between GDM cases and controls and consequently similar allele frequencies for SLC16A11 risk associated variants between both groups were observed. This finding suggest that apart from T2D common gene variants, other genetic and possibly environmental risk factors may play a role in the development of GDM. The association of FTSJD1/CALB2 locus with GDM risk in two independent studies supports the fact that other GDM susceptibility loci are yet to be identified.

Low SES has also been associated to the risk of developing T2D [12], yet similar proportions of GDM cases and controls were classified in the low SES category. It is also noteworthy that a major proportion of cases reported to be college graduates and to be employed. These findings suggest that environmental risk factors, rather than Native American ancestry are important determinants of GDM susceptibility. Reported environmental factors associated with the risk of developing GDM include excess high energy and saturated fat intakes [73], as well as dietary deficiencies of vitamins B12 and 25OHD [74, 75].

Besides the relationship between the common T2D gene risk variants with GDM, we also found association of two other T2D related loci with increased 60’ OGTT glucose levels (CENTD2) and decreased HOMA B (MTNR1B). The CENTD2 LD region comprises at least three other genes (PDE2A, QTG16L2, FCHSD2). Because none of them have been implicated in insulin processing or secretion, CENTD2, also known as ARAP1, is proposed to be the responsible gene within this locus. It encodes a protein which activates Arf and Rho GTPases to regulate membrane trafficking and actin cytoskeleton reorganization. Apart from T2D, this risk locus has been strongly associated to fasting proinsulin and 32-33-split proinsulin. Recently, it was proposed that rs11603334, which is in perfect LD with rs1552224 (analyzed in this study), is a functional variant regulating ARAP1 expression. The variant rs11603334 is located at the P1 promoter and overlaps transcription factor binding sites within a region of open chromatin that is marked by DNA hypersensitivity, H3K4me3 and H3K9ac in pancreatic islets. Risk allele C disrupts binding of PAX4/6 transcriptional factors and increases transcriptional activity by 2-fold, thus leading to higher gene expression [76]. PAX4/6 are known regulators of both glucagon and insulin promoters. Coding variants within them have been previously implicated in glucose homeostasis [77].

As to the role of MTNR1B, genetic variants within this locus have been associated to an increase in fasting glucose levels, reduced beta cell function, as well as T2D and GDM risk [35]. Circadian rhythms have been recognized as critical for glucose homeostasis maintenance and melatonin is among their major regulators. Melatonin is a diurnal hormone produced by the pineal gland, which binds to melatonin receptor 2, encoded by MTNR1B gene, to exert its functions. Even though the causal variant is yet to be identified, intronic variant rs10830963 is the strongest associated signal. No evidence has been reported regarding its molecular mechanism, but it does not seem to disrupt consensus transcription factor binding or cryptic alternative splice sites. Nevertheless, risk allele G was associated to MTNR1B increased gene expression in pancreatic islets [78], thus impairing insulin secretion. Of particular importance is the fact that during pregnancy, plasma melatonin levels are elevated and it crosses the placenta and fetal blood-brain barrier, playing a key role not only in the prevention of pregnancy loss, but also in the development of fetal organs and its adaptation to extra-uterine life [79].

GWAS have allowed an accelerated identification of T2D risk allele variants in a short period of time. Nevertheless, we are still missing the mechanisms by which they confer risk to the disease. It is remarkable that all associated variants are located within non-coding regions and their nearest corresponding gene belong to GO terms involved in the regulation of gene transcription and further transcript abundance, with a higher frequency that may occur by chance. Two interesting enriched GO terms were regulation of transcription from RNA polymerase II promoter and gene-specific transcription, which emphasize the putative regulatory function of involved genes. More studies are needed to clarify the mechanisms involved on disease susceptibility. Apart from T2D, our results pointed at other implicated metabolic pathways and transcriptional regulatory mechanisms, thus highlighting the complex pathophysiology of GDM. Genome-wide studies with larger sample size are still needed to identify genes with smaller allele effects and possibly GDM specific risk variants. Additional analyses, such as global expression studies in tissues relevant to disease physiology, will also be required to elucidate the role of different gene risk variants in disease susceptibility.

Supporting Information

Acknowledgments

We would like to thank Fernández-López, JC. for his valuable assistance with global ancestry estimation; Rodríguez-Torres, M., Ramírez-Jiménez, S. and Ordoñez-Sánchez, ML. for their technical assistance. We also appreciate the SIGMA Consortium participants collaboration. The research team laments the death of co-author Dr. Carlos Vargas, a respected colleague who made important contributions to women´s health in Mexico. We dedicate this work to his memory.