Meta-analysis of genetic association studies on gestational diabetes mellitus


 Background

Several molecular epidemiological studies have analyzed the associations between genetic variants and the risk of gestational diabetes mellitus (GDM). However, all these studies suffer from inconsistent and conflicting results owing to relatively smaller sample sizes, fewer genetic variants included in the research, and limited statistical power. Hence, a coherent review and meta-analysis were carried out to provide a quantitative summary related to the associations of commonly studied SNPs with GDM risk.
Methods

Eligible studies were retrieved from PubMed,updated on Dec. 2019. Based on several inclusion and exclusion criteria, 71 articles with 42928 GDM patients and 77793 controls were finally considered for meta-analysis. The genotype data from 23 variants of sixteen genes were statistically analyzed using RevMan v 5.2 software. Newcastle-Ottawa Scale (NOS) was used to assess the quality of the research article. Heterogeneity among studies was tested by I2 and odds ratio with 95% confidence interval (CI) was carried out for all five genetic models.
Results

The overall combined odds ratio reveals that variants like MTNR1B (rs1083963, rs1387153), GCK (rs1799884), CANP10 (rs3792267), and GCKR (rs780094) are significantly associated with GDM in all genetic models while CANP10 (rs5030952), ADRB (rs4994) and FTO (rs8050136) are not significantly associated with GDM in any genetic models. Variants MTNR1B (rs1083963, rs1387153) and GCK (rs1799884) are associated with increased risk (OR>1, p<0.05) of GDM, and all these are related to insulin secretion. Other variants related to insulin secretion like TCF7L2 (rs7903146) and SLC30A8 (rs1326634) are also associated with increased risk (OR>1, p<0.05) of GDM. On the contrary, CANP10 (rs3792267) and GCKR (rs780094) are found associated with decreased risk (OR<1, p<0.05) of GDM. Other variants are significantly associated with the GDM in at least one or more genetic models.
Conclusion

Our study identified that most of the variants related to insulin secretions like MTNR1B (rs1083963), GCK (rs1799884), TCF7L2 (rs7903146), GCKR (rs780094), and SLC30A8 (rs1326634) are more strongly associated (p<0.005) with GDM as compared to the variants related to the insulin resistance like PPARG (rs1801282), IRS1 (rs1801278) and ADIPOQ (rs266729).


Introduction
Gestational diabetes mellitus (GDM), a complex metabolic disorder, is de ned as any degree of carbohydrate intolerance with onset or rst recognition during pregnancy (American Diabetes Association, 2010). Risk factors include maternal age, pre-pregnancy obesity, and previous delivery of a newborn with congenital malformations such as macrosomia, previous/prior history of GDM, cesarean section, and a family history of diabetes in rst degree relatives (Reece et al., 2009).However, intrinsic factors like environmental interaction and genetic predisposition can't remain unaddressed . Women with GDM had an increased risk of developing diabetes type 2 which accounts for 90% of cases of diabetes. GDM can progress when a genetic susceptibility to pancreatic islets cells are exposed to incremental insulin resistance during pregnancy. GDM is often associated with adverse pregnancy outcomes, including fetal macrosomia, stillbirth, neonatal metabolic disturbances, and related problems, and causes short-and long-term complications in women and their offspring (Bellamy et al., 2009;"Gestational Diabetes Mellitus," 2004;Reece et al., 2009). GDM women are at over seven-fold higher risk of developing type 2 diabetes mellitus (T2DM) later in life (Bellamy et al., 2009).
The Burden of GDM is growing at a much higher rate in developing and low-to-middle income countries than in developed countries. The prevalence of GDM varies widely from 1.8-25.1% of all pregnancies. It is higher among the Middle East and North Africa, South Asia, and Western Paci c regions while lowest in Europe (Zhu and Zhang, 2016). In the western world, the incidence of GDM is about 1-3% of all pregnancies while 5-10% in Asian pregnancies . In India, the incidence of GDM is estimated to be 10-14.3% which is much higher than in developed countries (Lowe et al., 2016). This signi cant variation in its prevalence is attributed to racial and ethnic differences a social-economic variations (Zhu and Zhang, 2016). The Asian/Paci c Islander women have a higher incidence of GDM than non-Hispanic white, Black, or Hispanic women (Chu et al., 2009;Kim et al., 2012). Due to racial and regional differences in GDM prevalence, exploring the relationship of susceptible gene polymorphism in GDM women of different racial backgrounds will be quite informative.
In this study, we systematically analyzed all the current evidence regarding the genetic associations of GDM to quantitatively summarize the effect size of replicated single nucleotide polymorphisms (SNPs) on GDM risk and identify the possible sources of heterogeneity among the eligible researchers. A total of 23 different SNPs related to 16 genes were included for meta-analysis. These sixteen genes are namely TCF7L2 (Transcription factor 7-like 2), MTNR1B (Melatonin receptor 1B), FTO (Alpha-ketoglutarate-dependent dioxygenase), PPARG (Peroxisome proliferator-activated receptor-gamma), GCK (Glucokinase), GCKR (Glucokinase Regulator), ADIPOQ (Adiponectin), TNF (Tumor necrosis factor), IRS1 (Insulin receptor substrate 1), KCNJ11 (Potassium inwardlyrectifying channel, subfamily J, member 11), IGF2BP2 (Insulin-like growth factor 2 mRNA-binding protein 2), ADRB3 (Adrenoceptor beta 3), CDKAL1 (CDK5 regulatory subunit associated protein 1-like 1), HNF1A (Hepatocyte nuclear factor 1-alpha), CANP10 (Calpain-10), and SLC30A8 (Solute carrier family 30 member 8). These genes and their respective SNPs were involved in pathways like type 1 diabetes mellitus, type1 diabetes mellitus, insulin signalling pathway, Maturity onset diabetes of the young, PPAR signalling pathway, Adipocytokine signalling pathway, glycolysis, and amino sugar metabolism. To the best of our In the present study, a total of 23 SNPs related to 16 different genes were analyzed. Characteristic features of genes and their related alleles are provided in Table 3. Of these, CANP10 and TCF7L2 have three polymorphisms; MTNR1B, FTO, and GCKR have two polymorphisms, while the rest genes have only one SNPs included in the study ( Fig. 2A). Association between all twenty-three polymorphism and GDM risk were assessed in ve genetic models, and detailed results have been shown in Table 4. Out of 23 SNPs analyzed, the results showed that only 17 SNPs were signi cantly associated with GDM risk in at least one genetic model ( Fig. 2B and D, Table 4). A total of 13 polymorphisms regarding genotypes in the dominant model, 11 polymorphisms regarding genotypes in the recessive model, 14 polymorphisms regarding genotypes in the homozygous model, ten polymorphisms regarding genotypes in the heterozygote model and ten polymorphisms regarding genotypes in the allele model were found to be signi cantly (p<0.05) associated with GDM risk (Fig. 2D). The four polymorphisms, namely rs780094 (GCKR), rs1387153 (MTNR1B), rs1799884 (GCK), rs1083963 (MTNR1B) were showing signi cant association (p<0.05) with GDM risk in all ve genetic models (Fig. 2C). On the other hand, two polymorphisms (rs3792267, rs9939609) exclusively present the dominant, recessive, homozygote, and heterozygote model each while another two polymorphisms (rs2975760, rs12255572) exclusively present in the dominant, recessive, homozygote, and allele models were found to be signi cantly (p<0.05) associated with GDM risk (Fig. 2C). Association between polymorphisms and GDM In all ve genetic models analyzed, the number of polymorphisms associated with increased risk (OR>1) of GDM is larger than those polymorphisms which have protective effects (OR<1) (Fig 3A). Analysis of all 23 SNPs in the dominant genetic model revealed signi cant heterogeneity (p<0.05) in only seven polymorphisms. Thus, a random-effect model was conducted to pool results for total OR estimation. Genotype analysis indicates that all these seven polymorphisms are signi cantly associated with increased risk (OR>1, p<0.05) of GDM (Supplementary le 1). In recessive models (Fig. 3B), signi cant heterogeneity (p<0.05) was found in 10 polymorphisms out of 23 analyzed. Random effect model showed signi cant protective association (OR<1) of eight polymorphisms while two polymorphisms namely GCK rs1799884 (OR = 1.77(1.56-2.00); I 2 = 86%, p<0.00001) and TCF7L2 rs12255572 (OR = 2.24(1.81-2.77); I 2 = 87%, p<0.00001) are signi cantly associated with increased risk (OR>1) of GDM. In the homozygote model ( Fig. 3B), signi cant heterogeneity was observed in total eleven polymorphisms, of which ten were strongly associated with increased risk (OR>1) of GDM while only one, i.e., GCKR rs780094 was showing protective association (OR = 0.52(0.38-0.70); I 2 = 83%, p<0.00001). A signi cant heterogeneity (p<0.05) was observed in six polymorphisms out of 23 in the analysis of the heterozygote model (Fig. 3B). Genetic analysis through the random effect model revealed that ve polymorphisms are signi cantly associated with increased risk (OR<1) of GDM while GCKR rs780094 showed protective association (OR = 0.72(0.53-0.97); p = 0.03). Genotypes in the allele model ( Fig. 3B) revealed a total of nine signi cant (p<0.05) heterogeneity, of which six were signi cantly associated with increased risk (OR>1) while three were showing protective association (OR<1) for GDM. The overall results revealed the signi cant heterogeneity (p<0.05) of ve polymorphisms, namely rs1326634 (SLC30A8), rs780094 (GCKR) rs1083963 (MTNR1B), rs7903146 (TCF7L2), rs9939609 (FTO) in all ve genetic models analyzed. Among these ve, rs780094 (GCKR) rs1083963 (MTNR1B) were signi cant for overall effect in all ve genetic models analyzed. Polymorphisms rs1799884 (GCK) and rs1387153 (MTNR1B) were also signi cantly associated with the disease. However, they are not signi cant at the heterogeneity level in all ve genetic models analyzed.

Gene-interaction, functional and pathway enrichment
The protein-protein interaction network reveals a high degree of interaction among these sixteen genes (Fig. 9A). Gene ontology enrichment analysis highlights the role of these sixteen genes in signi cant biological processes like regulation of glucose metabolism and insulin secretion, cellular response to insulin stimulus, and fatty acid oxidation (Fig. 9B). Signi cant pathways being regulated by these genes are the PPAR signalling pathway, mTOR signalling pathway, Maturity onset of diabetes in young (MODI), and type II diabetes mellitus pathway (Fig. 9C).

Sensitivity analysis
For maternal genotype sensitivity analysis, the overall OR after exclusion of any individual study showed no change, ranging from 0.71 to 0.81, including the robust at-risk effect of the rs1326634, rs7754840 rs1083963, rs9939609 genotype against gestational diabetes.

Publication bias analysis
Egger's test was performed to assess the publication bias. No statistically signi cant evidence of publication bias was observed for studies included in the analyses.

Discussion
In the present meta-analysis, all studied genetic variants related to type1 and type2 diabetes was investigated for their association with the GDM risk. Several previous studies have included only a few variants or have missed some variants (Ref.). Moreover, the pathophysiology of GDM shares similarities with type1 (insu cient insulin secretion) and type2 diabetes (insulin resistance). Both insulin insu ciency and insulin resistance play a signi cant role in the development and progression of GDM. Besides, these two pathways, glucose and lipid metabolism pathway, also play an essential role in the pathophysiology of GDM. Hence in this meta-analysis, we have used rigorous statistical analysis to rule out the most signi cantly associated variants with increased GDM risk and related pathways. Further, we also tried to identify those pathways whose genetic variants are strongly associated with the increased GDM risk. Thus, our replication study provides a more comprehensive and concise summary of the currently available evidence regarding GDM genetic variants.
Pregnancy is accompanied by a number of changes in metabolic activity, which helps in dwelling the interaction between mothers and growing fetuses to meet their energy needs. There is a slight enhancement of insulin sensitivity seen during early gestation; however, this insulin sensitivity declines during 12-14 weeks. Moreover, in the third trimester, this insulin sensitivity increases, and these values, as reported in some cases, approach the values of T2DM. This condition is termed GDM; evidence has reported that GDM develops when a genetic predisposition of pancreatic islet B-cells impairment is unmasked by the increased insulin resistance during pregnancy. A GWAS study con rmed the association of various SNPs with impaired b-cell function (MTNR1), insulin resistance, and abnormal utilization of glucose (GCK, CANP10). We came across many studies with heterogeneous results, and this variation is liable to ethnicity, study design, and tissue is taken. Overall, we observed that variants related to insulin secretions pathways like MTNR1B (rs1083963), GCK (rs1799884), TCF7L2 (rs7903146), GCKR (rs780094), and SLC30A8 (rs1326634) are more strongly associated (p<0.005) with increased GDM compared to the variants related to the insulin resistance like PPARG (rs1801282), IRS1 (rs1801278) and ADIPOQ (rs266729).

The MTNR1B polymorphisms
The MTNR is reported to modulate pancreatic islet B-cells function. Our study coincides with the result of Kim et al., who rst reported a signi cant association of GDM with MTNR1B rs1387153. A study conducted by Zheng et al. in 2013 observed the T allele of rs1387153 associated with increased risk of GDM, Vlassi et al. (2012) studies supported. However, the study on Chinese women conducted by Wang et al. disregarded the ndings.

The GCK polymorphisms
The Glucokinase (GCK) with the rs179884 has been widely studied in different ethnic populations with con icting results. Chiu et al. and Zaidi et al. reported no association between GDM and rs179884. However, Shaat et al., Freathy et al., Santas et al. found signi cant association when they conducted research on a relatively larger population. Our replication study also aligned with their ndings and the meta-analysis also presented the signi cant association with no signi cant heterogeneity.

The CANP10 polymorphism
The CAPN10 gene belongs to the calpain family and is a Ca2+ dependent intracellular cysteine protease. CAPN10 is found to be involved in glucose homeostasis as it regulates the activity of pancreatic B islet-cells, liver, skeletal muscle, and adipocytes. However, the direct association between GDM and rs3792267 remains elusive due to reported contradictory results (Heinz et al., Luo et al., N Shaat et al., Thomas et al.) Conclusion The superiority of this study was that multiple databases were included to search the literature as thoroughly as possible. The subgroup and random effect analysis were utilized to decline heterogeneity, and the comprehensive assessment of publication bias was done, which identi ed the results of our study effectively and reliably. These ndings suggest insulin resistance or defects in insulin secretion have major implications in aetiology in GDM however ethnicity plays a crucial role in it.     Venn diagram and Bar graph: Bar graph is showing the genes and their associated variants which have been included in the study. TCF7L2, MTNR1B, FTO, GCKR and CANP10 have more than one variant. Venn diagram shows the number of variants which have been found to be signi cantly associated with GDM risk in all ve genetic models.Out of 23 variants analyzed, only four variants namely rs780094, rs1387153, rs1799884, and rs1083963 have been found to be signi cantly associated with increased risk of GDM.

Figure 3
Odds Ratio distribution: A. Most of the SNPs have odds ratio grater than one in all genetic models analyzed (except recessive; OR<1) and hence are associated with increased risk of GDM. B. Value of ORs signi cantly associated with GDM among different genetic models have been plotted.

Figure 4
Forest plot: of association between TCF7L2 rs7903146 polymorphism and risk of gestational diabetes (all genetic model). The shadowed squares and their lateral tips indicate the ORs and the corresponding 95% CIs in individual studies, with the sizes of squares proportional to weights used in the meta-analyses.
The central lines and lateral tips of the diamonds indicate the pooled ORs and the corresponding 95% CIs. The solid vertical lines indicate no effect.

Figure 6
Forest plot: The risk of GDM in association of genetic variants GCKR rs780904. The shadowed squares and their lateral tips indicate the ORs and the corresponding 95% CIs in individual studies, with the sizes of squares proportional to weights used in the meta-analyses. The central lines and lateral tips of the diamonds indicate the pooled ORs and the corresponding 95% CIs. The solid vertical lines indicate no effect.

Figure 7
Forest plot: The risk of GDM in association of genetic variants of MTNR1B A. rs1083963 B. rs1387153. The shadowed squares and their lateral tips indicate the ORs and the corresponding 95% CIs in individual studies, with the sizes of squares proportional to weights used in the meta-analyses. The central lines and lateral tips of the diamonds indicate the pooled ORs and the corresponding 95% CIs. The solid vertical lines indicate no effect.

Figure 9
Networking, GO and KEGG pathway enrichment. A. String network of all 16 genes shows the high interconnection among these genes. B. Gene Ontology (GO) enrichment analysis revealed major function; processes and cellular components related to these genes and have been plotted against log of p-value. C. Signi cant pathways associated with these have been plotted against log of p-value.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.