GRAPHICAL ABSTRACT

Highlights

• We explored the genetic factors of gestational diabetes (GDM) and their applications.

• We confirmed three loci and identified a novel locus for GDM.

• A polygenic risk score (PRS) from these loci effectively stratifies GDM risk.

• The PRS also stratifies the risk of abnormal glucose tolerance post-pregnancy.

• Genetic information may help identify high-risk women to prevent GDM.

INTRODUCTION

Gestational diabetes mellitus (GDM) occurs when insulin production or utilization is impaired during pregnancy, resulting in hyperglycemia. It typically resolves after delivery. GDM is a prevalent condition, affecting 14.0% of live births worldwide in 2021 [1]. Unmanaged GDM can lead to adverse perinatal outcomes [2] and increased long-term metabolic risks for both mothers and children. A recent meta-analysis assessing over 1.3 million individuals reported that women with GDM have a nearly 10-fold increased risk of developing type 2 diabetes mellitus (T2DM) compared with healthy controls [3]. Moreover, offspring exposed to hyperglycemia in utero are more likely to be obese and insulin resistant in childhood and early adulthood than offspring of mothers with normoglycemia [4-6]. These observations indicate that (1) GDM shares a common pathology with T2DM; and (2) GDM can provide insights into the predisposition of T2DM, with pregnancy serving as a stressor that unmask hyperglycemia during pregnancy among those who are at risk of future diabetes because of impaired β-cell function.

Given the increased lifetime risk of T2DM associated with GDM, many genetic studies have focused on the assumption that the genetic architecture of both conditions shared similarities. A number of T2DM susceptibility loci have been examined for their association with GDM in modest sample sizes [7]. In a recent meta-analysis of 23 studies, eight T2DM loci (insulin like growth factor 2 mRNA binding protein 2 [IGF2BP2], CDK5 regulatory subunit-associated protein 1-like 1 [CDKAL1], GLIS family zinc finger 3 [GLIS3], cyclin dependent kinase inhibitor 2A/2B [CDKN2A/2B], hematopoietically expressed homeobox [HHEX]/insulin degrading enzyme [IDE] transcription factor 7 like 2 [TCF7L2], melatonin receptor 1B [MTNR1B], and HNF1 homeobox A [HNF1A]) showed association with GDM after correcting for multiple comparisons [7]. Notably, these loci demonstrated a consistent direction of association with T2DM, supporting a shared genetic background between the two conditions. More recently, the GENetics of Diabetes In Pregnancy (GenDIP) Consortium conducted a multi-ancestry genomewide association studies (GWAS) meta-analysis for GDM, involving over 300K women [8]. Among the five genome-wide significant loci identified, four were previously reported to be associated with T2DM: MTNR1B, TCF7L2, CDKAL1, and CDKN2A/B. Interestingly, the novel locus at hexokinase domain containing 1 (HKDC1) did not show a strong association with T2DM, highlighting the presence of genetic determinants specific to glucose regulation in pregnancy. The FinnGen Study further explored genetic features distinct from T2DM and identified eight loci with effects on GDM that were three times stronger compared to T2DM [9].

There is growing interest in utilizing a polygenic risk score (PRS) to assess the individual’s risks of developing GDM and T2DM postpartum. Prior efforts to construct PRSs for GDM typically involved variants associated with T2DM, and suffered from a small number of variants and participants. As a result, these scores had limited success in enhancing predictive capability compared to models that included only clinical variables [10]. For instance, in a study of Caucasian women involving 458 GDM cases and 1,538 controls, a PRS consisting of 34 variants related to T2DM and/or fasting glucose was associated with increased risk of GDM. However, its utility in identifying GDM cases was limited, with only a modest improvement in GDM prediction when added to clinical factors (increase in c-statistic=0.03) [11].

This study aimed to (1) identify novel GDM susceptibility loci by performing a meta-analysis of three GWASs in Chinese women, followed by replication studies in independent Chinese, European, and multi-ethnic cohorts; and (2) explore the clinical utility of the identified variants by deriving a PRS for GDM, and evaluating its predictive value for GDM and abnormal glucose tolerance (AGT) at 7-year postpartum.

METHODS

Study design and participants

The overall design of the current study is shown in Supplementary Fig. 1. Details of the design, ascertainment methods, inclusion criteria, and phenotyping procedures for each cohort are outlined in the “cohort descriptions” (Supplementary Methods). Supplementary Tables 1 and 2 summarize the clinical characteristics of participants in the discovery and replication studies. At the time of assessment, all participants provided written informed consent for DNA collection and data analysis for research purposes. Institutional review boards of the respective institutions approved each study.

For the genome-wide scan, a meta-analysis was conducted on 1,681 (464 cases and 1,217 controls) women of Han Chinese ancestry from three independent cohorts, including: (1) The Hyperglycemia and Adverse Pregnancy Outcome-Hong Kong (HAPO-HK) Study, which consisted of 960 pregnant women (149 GDM cases vs. 811 non-GDM controls) attending a follow-up visit at the Hong Kong center [6]; (2) The Tianjin GDM Study, which involved 455 pregnant women (229 GDM cases vs. 226 non-GDM controls) participating in a two-step GDM screening program conducted by Tianjin’s Women and Children Health Center [12]; and (3) The Treated GDM Cases vs. Non-diabetes Controls (TGDM-NDM) Study, which included 86 pregnant women with GDM receiving antenatal treatment at the Hong Kong center [13], and 180 non-diabetes, non-pregnant women recruited from hospital staff and a community-based health screening program [14]. The flow diagrams of cohort selection in the HAPO-HK and Tianjin Studies are shown in Supplementary Figs. 2 and 3.

The first-stage de novo replication involved 1,136 Southern Han Chinese pregnant women (564 GDM cases and 572 non-GDM controls) from the Guangzhou Study [15]. In the second-stage in silico replication, we accessed data from two published GDM GWAS, including the FinnGen Study (12,332 cases and 131,109 controls) [9] and a multi-ancestry meta-analysis of GDM contributed by the GenDIP Consortium (5,485 GDM cases and 347,856 non-GDM controls) [8]. The HAPO-HK Study, our discovery cohort, participated in the GenDIP meta-analysis. Additionally, individual-level data from the Thai (HAPO-Thai; 260 GDM cases and 933 non-GDM controls) and Hispanic (HAPO-Hispanic; 207 GDM cases and 596 non-GDM controls) populations of the HAPO Study were utilized to validate the results of the PRS analysis [2].

Clinical outcomes

GDM was diagnosed using criteria established in 2010 by the International Association of the Diabetes and Pregnancy Study Groups (IADPSG) [16]. These criteria included fasting plasma glucose levels ≥5.1 mmol/L, or 1-hour glucose levels ≥10.0 mmol/L, or 2-hour glucose levels ≥8.5 mmol/L, as determined from a 75-g oral glucose tolerance test conducted during the third trimester of gestation.

AGT was defined as the presence of impaired glucose tolerance, impaired fasting glucose, or diabetes, as per the diagnostic criteria of the American Diabetes Association [17].

Statistical analyses

Within cohort, logistic regression was used to examine the association between genetic markers (e.g., a single nucleotide polymorphism [SNP] or a PRS) and the risks of GDM and AGT after pregnancy, with the adjustments for the first four principal components (PCs), age and/or body mass index (BMI). Meta-analysis was conducted with use of the inverse-variance weighted method under a fixed-effects model. The area under the curve (AUC) of the receiver operating characteristic and continuous net reclassification improvement (NRI) index were used to evaluate the incremental predictive value of PRS in GDM and AGT after pregnancy, over the clinical risk factors and PCs.

Detailed methods for (1) clinical and laboratory measurements; (2) genotyping, quality controls, and imputation; (3) construction of PRS; and (4) statistical analysis can be found in Supplementary Methods.

Data availability

The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. The HAPO-Thai and -Hispanic data were obtained from database of Genotypes and Phenotypes (dbGAP; phs000096.v4.p1).

RESULTS

Discovery genome-wide scan analysis

We performed a meta-analysis of three GWASs for GDM in Chinese women. The results of the genome-wide scan are illustrated in Fig. 1, displaying Manhattan plots and corresponding quantile-quantile (Q-Q) plots. Among the 6,322,337 biallelic and autosomal SNPs available in all discovery cohorts, we identified two loci that reached genome-wide significance. These included a previously reported locus for GDM and T2DM susceptibility at MTNR1B (rs7945617, odds ratio [OR], 1.64; 95% confidence interval [CI], 1.38 to 1.96; P_meta=2.6×10^–8), and a novel locus near T-box brain transcription factor 1 (TBR1)-solute carrier family 4 member 10 (SLC4A10) (rs117781972, OR, 2.05; 95% CI, 1.61 to 2.62; P_meta=7.6×10^–9) (Fig. 2, Supplementary Tables 3 and 4). The results remained largely unchanged after applying genomic control correction (Supplementary Tables 3 and 4). Conditioning on TBR1-SLC4A10 rs117781972 and MTNR1B rs7945617 did not reveal any residual associations (data not shown).

The lead SNP from the two genome-wide significant loci underwent (1) de novo replication in the Guangzhou Study and (2) in silico replication using data from the FinnGen Study and the GenDIP meta-analysis. The TBR1-SLC4A10 rs117781972 variant did not replicate in the Guangzhou and FinnGen Studies (P>0.05) and it only showed nominal significance in the GenDIP meta-analysis (OR, 1.31; 95% CI, 1.07 to 1.60; P_meta=0.0164) (Fig. 2A and Supplementary Table 3). In contrast, the MTNR1B rs7945617 showed a P value ≤0.05 in all replication cohorts and demonstrated a smaller but concordant direction of effect in all replication cohorts (1.21< OR <1.34), compared to the discovery cohorts (Fig. 2B and Supplementary Table 4).

Candidate gene analysis

We utilized the meta-analysis results from this study to investigate the associations of established GDM-related loci in Chinese women (Table 1). In this analysis, we specifically chose 14 unique variants from nine loci that have shown associations with GDM at genome-wide significant levels in previous studies [8,9,18,19]. Of these, we confirmed the associations for four variants at the CDKAL1 locus (rs7754840 and rs9348441) and the MTNR1B locus (rs10830962 and rs10830963) after correcting for multiple comparisons (P value threshold: 0.05/14=3.6×10^–3). Our meta-analysis in Chinese women revealed a 1.33- to 1.58-fold (1.5×10^–6<P_meta<7.9×10^–4) increased risk of GDM per copy of the risk allele for these variants.

In our meta-analysis, we further investigated the impact of known T2DM-associated variants on the risk of GDM. We examined 338 variants that showed genome-wide significance in a multi-ethnic GWAS conducted by the Diabetes Meta-Analysis of Trans-Ethnic association studies (DIAMANTE) Consortium, which included 180,834 T2DM cases and 1,159,055 controls [20]. From these variants, we successfully obtained 286 independent variants at 216 loci in our meta-analysis (Supplementary Table 5). This analysis identified three variants, namely rs9348441 at the CDKAL1 locus (OR, 1.40; 95% CI, 1.18 to 1.65; P_meta=9.4×10^–5), rs2237897 at the insulin-insulin-like growth factor 2 (INS-IGF2)-potassium voltage-gated channel subfamily Q member 1 (KCNQ1) locus (OR, 1.48; 95% CI, 1.23 to 1.79; P_meta=3.8×10^–5), and rs10830963 at the MTNR1B locus (OR, 1.58; 95% CI, 1.31 to 1.91; P_meta=1.5×10^–6), that remained significant after correcting for multiple comparisons (P value threshold: 0.05/286=1.7×10^–4).

Previous studies consistently found a strong association between GDM and the CDKAL1 and MTNR1B loci, across diverse populations. To support the signals at the INS-IGF2-KCNQ1 locus, we conducted independent replication studies in the Guangzhou Study, the FinnGen Study and the GenDIP meta-analysis. Our analysis of INS-IGF2-KCNQ1 rs2237897 yielded P values ranging from 8.0×10^–6 to 2.7×10^–3, with consistent ORs of 1.13 to 1.33 in all these studies, aligning with the original findings (Fig. 3 and Supplementary Table 6). Moreover, a previous study conducted in the Mexican population (408 cases and 342 controls) reported a positive association between rs2237897 and GDM (OR, 1.85; 95% CI, 1.39 to 2.50; P=4.5×10^–5) (Fig. 3 and Supplementary Table 6) [21]. We further performed a meta-analysis for rs2237897 using data from the HAPO-HK Study, the Tianjin Study, the TGDM-NDM Study, the Guangzhou Study, the FinnGen Study, and the Mexican Study, comprising a total of 13,768 GDM cases and 133,240 controls. In this meta-analysis, the rs2237897 variant exhibited a genome-wide significant association with GDM risk (OR, 1.16; 95% CI, 1.10 to 1.22; P_meta=2.4×10^–9) (Fig. 3 and Supplementary Table 6).

Conditional analysis at the CDKAL1 and MTNR1B loci

Supplementary Fig. 4 summarizes the GDM variants identified in this study. Through the genome-wide scanning and candidate gene approaches, we found two variants at the CDKAL1 locus (rs7754840 and rs9348441) and three variants at the MTNR1B locus (rs7945617, rs10830962, and rs10830963). These variants showed high linkage disequilibrium (LD) within each locus in our Chinese samples (LD r²=0.81 for CDKAL1 and 0.78–0.93 for MTNR1B) (Supplementary Table 7). Conditional analysis revealed that the effect sizes of variants within the two loci were attenuated when conditioned on each other (Supplementary Table 7). As a result, we selected four variants for further analysis: CDKAL1 rs7754840, MTNR1B rs7945617, TBR1-SLC4A10 rs117781972, and INS-IGF2-KCNQ1 rs2237897.

Results for (1) sensitivity analysis and (2) associations of identified variants with glycemic and metabolic traits during pregnancy are available in the Supplementary Results.

Development and evaluation of PRS for GDM

We constructed a PRS using the four identified variants, weighted by their corresponding effect sizes for GDM risk, as estimated in our meta-analysis. This PRS demonstrated robust association with increased GDM risk among the three discovery cohorts of Chinese women (1.2×10^–13<P<1.0×10^–3) (Supplementary Table 8). Our meta-analysis revealed a 1.84-fold (95% CI, 1.62 to 2.09; P_meta=9.2×10^–21) increase in GDM risk per 1-standard deviation (SD) increase in the PRS. Each quintile increase in the PRS was associated with a 1.48-fold (95% CI, 1.35 to 1.62; P_meta=3.4×10^–17) increased odds of GDM (Fig. 4A). Women in the top quintile had a three-fold (OR, 3.07; 95% CI, 2.30 to 4.09; P_meta=3.1×10^–14) and more than five-fold (OR, 5.20; 95% CI, 3.40 to 7.96; P_meta=3.0×10^–14) increased risk of GDM, compared with the remaining individuals and those in the bottom quintile, respectively (Fig. 4A and Supplementary Table 8). The association between PRS and GDM was further validated in the Guangzhou Study (OR, 1.16; 95% CI, 1.03 to 1.31; P=0.0133) and the HAPO-Hispanic Study (OR, 1.41; 95% CI, 1.18 to 1.67; P=1.2×10^–4) (Fig. 4B, C and Supplementary Table 8). However, no significant association was seen in the HAPO-Thai Study (P=0.2695) (Fig. 4D and Supplementary Table 8).

Next, we assessed the predictive value of our PRS for GDM (Supplementary Table 9). In the model incorporating only the PRS, the AUC for predicting GDM varied between 0.63 and 0.69 in the three discovery cohorts; but it was below 0.57 in all the validation cohorts. When PRS was added to the base model consisting of PCs, age and BMI, there was a significant increase in the AUC by 5.0% to 7.8% in the HAPO-HK and Tianjin Studies (1.4×10^–4<P_increase<0.0134). Computation of NRI index indicated enhanced reclassification in all discovery and HAPO-Hispanic Studies (25.8%< NRI <57.4%; P_NRI<0.05). Similar findings were observed when additional clinical risk factors were considered in the base model of both the HAPO-HK and HAPO-Hispanic Studies.

Evaluation of PRS for AGT after pregnancy

Given the well-established association between GDM and post-pregnancy hyperglycemia, we evaluated the role of our GDM-related PRS in predicting AGT risk at 7-year postpartum in the HAPO-HK Study (129 AGT cases and 829 controls). The clinical characteristics of these participants during the follow-up study are summarized in Supplementary Table 10. With the adjustments for PCs, age and BMI, the PRS was associated with a 1.40-fold (95% CI, 1.15 to 1.71; P=6.7×10^–4) increased risk of postpartum AGT per 1-SD increase in the PRS (Supplementary Table 11). Each quintile increase in this PRS raised the odds of AGT by approximately 27% (OR, 1.27; 95% CI, 1.10 to 1.46; P=1.1×10^–3) (Fig. 5). Women with a high PRS (top quintile) had a 71% (OR, 1.71; 95% CI, 1.08 to 2.71; P=0.0220) increased risk of AGT compared to the remaining individuals (Fig. 5 and Supplementary Table 11). This risk was nearly 3-fold (OR, 2.98; 95% CI, 1.51 to 5.86; P=1.6×10^–3) higher compared to women with a low PRS (bottom quintile). However, adjusting for either a history of GDM (P=0.0571) or a PRS derived from 286 T2DM-related variants (P=0.0408) attenuated the association between PRS and AGT (Supplementary Table 11).

The AUC for predicting postpartum AGT using only the PRS was 0.60 (95% CI, 0.54 to 0.65) (Supplementary Table 12). Incorporating the PRS into the base model, which included PCs, age and BMI, significantly increased the AUC by 1.7% (P=0.0187) and improved risk reclassification for AGT, as evidenced by an NRI index of 31.7% (95% CI, 12.6% to 49.6%). When either the GDM history or the T2DM-related PRS was included in the base model, the addition of the PRS did not impact the AUC. However, a significant NRI index was still observed, suggesting an improvement in reclassification (25.5%< NRI <25.6%; P<0.05).

DISCUSSION

This study reported a meta-analysis of three GWASs for GDM in Chinese women, with subsequent replication studies in Chinese and multi-ethnic populations. We identified four susceptibility loci at TBR1-SLC4A10, MTNR1B, CDKAL1, and INSIGF2-KCNQ1, using both genome-wide scan and candidate gene approaches. Moreover, the PRS derived from these loci showed improved risk stratification for Chinese women who are at risk of developing GDM or AGT postpartum.

Three out of the four GDM-associated loci identified in this study have also been implicated in glycemic traits and the development of T2DM in non-gravid populations. Aligning with previous GWASs in Korean [18], Finnish [9], and multi-ethnic populations [8], as well as candidate gene studies across different populations [21,22], our findings further support the associations of CDKAL1, MTNR1B, and KCNQ1 loci with GDM risk, with comparable effect sizes estimated in Chinese pregnant women. To the best of our knowledge, this is the first study showing a genome-wide significant association between the KCNQ1 variant and GDM risk. These three genes have been found to impact insulin secretion and glucose homeostasis, contributing to the development of diabetes [23-25]. Although the role of CDKAL1 in pancreatic β-cell function remains incompletely understood, it is known to be expressed in human adult pancreatic islets [26]. In mice, CDKAL1 gene deletion has been observed to cause a modest impairment in insulin secretion during high-fat feeding [27]. Moreover, emerging evidence indicates that functional loss of CDKAL1 affects the accuracy of protein translation, resulting in abnormal insulin synthesis and subsequently triggering endoplasmic reticulum stress in β-cells [28]. These findings provide evidence linking the molecular function of CDKAL1 with GDM. MTNR1B plays a crucial role in regulating circadian rhythm. Disruptions in circadian rhythms have been observed to promote systemic metabolic dysfunction, including obesity or diabetes, in both humans and rodents [29]. Consequently, altered MTNR1B function may influence circadian rhythm regulation, potentially affecting glucose metabolism during pregnancy. The KCNQ1 gene encodes for a voltage-gated potassium channel (Kv7.1) that is expressed in multiple organs/tissues, including pancreas [30]. In the pancreas, this channel drives an electrical signal in the β-cells that facilitates glucose-stimulated insulin secretion [30]. Changes of KCNQ1 function may impact the proper functioning of pancreatic β-cells, leading to impaired insulin secretion and glucose regulation in GDM women. Overall, our findings suggest shared genetic determinants between GDM and T2DM. The lack of association observed between other known T2DM-related loci and GDM may indicate partial differences in the genetic architecture of glucose metabolism between the pregnant and non-pregnant states, and/or limited power in the current study.

While many previous studies have identified susceptibility loci associated with both GDM and T2DM, two recent large-scale GWASs involving 143K–353K participants, predominantly of European descent, have uncovered a total of nine novel loci (i.e., HKDC1, glucokinase regulator [GCKR], SPC25 component of NDC80 kinetochore complex [SPC25], adenylate cyclase 5 [ADCY5], proprotein convertase subtilisin/kexin type 1 [PCSK1], estrogen receptor 1 [ESR1], cyclin D2 [CCND2], NEDD1 gamma-tubulin ring complex targeting factor [NEDD1], c-Maf inducing protein [CMIP], and mitogen-activated protein kinase kinase kinase 15 [MAP3K15]) specifically linked to glucose regulation during pregnancy [8,9]. These findings motivate further investigation into the distinct genetic mechanisms underlying GDM compared to T2DM, particularly within non-European populations. In this study, we have discovered a novel common variant rs117781972 located between the TBR1 and SLC4A10 gene regions in Chinese population. This variant has not been reported in GWASs for T2DM or glycemic traits. We found that the GDM-associated risk allele of rs117781972 was also associated with higher postprandial glucose levels (i.e., 1-hour glucose and 2-hour glucose) during pregnancy. Interestingly, inconsistent associations have been observed among replication studies, which may be due to differences in genetic background. For instance, the allele frequency of rs117781972 is significantly higher in East Asian populations (minor allele frequency [MAF]=0.227) compared to European (MAF=0.002 to 0.014), South Asian (MAF=0.01), and African (MAF=0.003) populations (https://gnomad.broadinstitute.org/). This genetic heterogeneity can be further modified by differences in phenotypes, lifestyles, and treatments. Our study, along with other research, suggests that while metabolic changes during pregnancy may resemble those observed in the progression of T2DM, and despite the shared genetic risk factors between T2DM and GDM, the underlying mechanisms driving the development of hyperglycemia during pregnancy may have some important differences.

The product of TBR1 is vital for the development and functioning of brain regions, including cerebral cortex, hippocampus, and amygdala [31,32]. SLC4A10, a member of the sodium-coupled bicarbonate transporters (NCBTs) family, regulates neuronal pH and excitability, and the secretion of cerebrospinal fluid [33]. Both genes are primarily expressed in the brain (data source: GTEx Analysis Release V8). TBR1 protein and multiple variants at SLC4A10 (rs4500960, rs7604885, and rs7580486) have demonstrated associations with major neurological disorders such as autism spectrum disorder (ASD) [32], cognitive ability [34], schizophrenia [35], and intelligence [36]. In line with the observational data linking exposure to maternal diabetes in utero with ASD risk [37], the effects of TRB1-SLC4A10 locus on GDM may also influence the fetal brain development, potentially leading to ASD after birth. However, little is known about the GDM-associated locus at these genes. A few recent GWASs have linked common variants at the TBR1-SLC4A10 locus (rs55732192, rs62189012, rs2053865, rs77882688, and rs1449631) to cardiometabolic traits, including BMI, mean arterial pressure, and systolic blood pressure [38-41]. Our identified variant rs117781972, strongly associated with GDM, appears to be independent of these reported variants (r²<0.1 in East Asian population). Furthermore, our data did not show any association between rs117781972 and blood pressure or BMI during pregnancy (Supplementary Table 13). To pinpoint the relevant gene and functional variant(s), further fine-mapping in larger multi-ancestry cohorts may be required.

Similar to T2DM, the risk of GDM may arise from variants in multiple loci, each conferring modest individual effects. Therefore, we computed a PRS using variants identified in this study to evaluate their cumulative impact on predicting an individual’s risk of developing hyperglycemia during and after pregnancy. Previous studies in Caucasian, Chinese and Indian populations have examined the association between PRS and GDM risk, using PRSs involving six to 14,971,357 variants [11,42-45]. However, there is limited data available for evaluating the performance of PRS in risk prediction [11,42,46]. Most of these studies derived PRSs using common variants associated with T2DM or glycemic traits, and tested associations in small sample sizes (n=296 to 8,722) [10]. Consistent with previous observations [11,42,46], our PRS, consisting of four variants, showed a significant association with GDM in Chinese and Hispanic populations. Our analysis further demonstrated that this PRS modestly contributed to GDM prediction when combined with well-recognized risk factors, although its overall value in identifying GDM cases at the population level was limited. It is worth noting that this PRS effectively stratified the risk of GDM in Chinese women, independent of clinical risk factors such as age and BMI. We found that women with a high PRS (top 20% of the distribution) had over a three-fold increased GDM risk compared to their counterparts with a lower PRS (remaining 80% of the distribution). More importantly, our PRS improved risk prediction for post-pregnancy hyperglycemia, showing a 71% higher risk among women with a high PRS compared to those with a lower PRS. This relationship was attenuated when adjusting for GDM history, further supporting the causal link between GDM and future T2DM progression. Overall, our findings underscore the importance of the polygenic contributions to GDM, and demonstrated the potential utility of genetic information in clinical practice, such as early genomic screening and personalized risk assessment for pregnant women.

This study has several limitations. First, the sample size for our genome-wide scan analysis was relatively modest, limiting our ability to discover new loci and detect variants with small genetic effects. Additionally, the HAPO Study have excluded women with severe GDM, further reducing the study’s power. Nevertheless, we were able to identify a novel locus associated with GDM within our cohorts. Second, our discovery study was exclusively conducted among Han Chinese women only, with limited generalizability. Third, the glucose tolerance status of controls in the TGDM-NDM Study was not assessed during pregnancy, potentially leading to the inclusion of some women with GDM as controls. Fourth, the predictive value of our PRS may be significantly diminished due to the inclusion of only a small number of genetic variants. Fifth, there is a possibility of over-fitting when evaluating the PRS in the discovery cohort used in the GWAS; however, it has been validated in independent cohorts. Furthermore, the impact of genetic variants may vary between GDM and AGT, and certain loci may be unique to each disease. We anticipate that constructing our PRS using weights from larger GWASs will enhance its predictive accuracy.

In summary, this study identified associations of four genetic loci with GDM among Chinese women. The MTNR1B, CDKAL1, and KCNQ1 loci, known for their role in hyperglycemia in non-pregnant populations, also have a significant impact during pregnancy, suggesting shared underlying pathology between T2DM and GDM. The TBR1-SLC4A10 locus, on the other hand, showed no association with hyperglycemia in non-pregnant populations, highlighting its specific importance in glucose metabolism during pregnancy. Finally, incorporating GDM-related PRS with clinical risk factors enhanced the prediction of GDM and post-pregnancy AGT. This demonstrates the clinical potential of integrating genetic information into risk assessment tools for identifying pregnant women at risk of developing these diseases, thereby providing opportunities to delaying disease onset.

SUPPLEMENTARY MATERIALS

NOTES

CONFLICTS OF INTEREST

Cadmon King-poo Lim, Juliana Chung-ngor Chan, and Ronald Ching-wan Ma are co-founders of GemVCare, a technology start-up initiated with support from the Hong Kong Government Innovation and Technology Commission and its Technology Start-up Support Scheme for Universities (TSSSU). The other authors declare that there is no duality of interest associated with this manuscript. Ronald Ching-wan Ma is a member of the international editorial board of Diabetes & Metabolism Journal.

AUTHOR CONTRIBUTIONS

Conception or design: C.H.T., Y.W., C.C.W., J.C.C., W.H.T., X.Y., R.C.M.

Acquisition, analysis, or interpretation of data: all authors.

Drafting the work or revising: C.H.T., C.C.W., X.Y., R.C.M.

Final approval of the manuscript: all authors.

FUNDING

This work was funded by the RGC Theme-based Research Scheme (T12-402/13 N); the Research Impact Fund (R4012-18), and the University Grants Committee Research Grants Matching Scheme. The follow-up of the HAPO Study at Hong Kong f ield center was supported by the General Research Fund of the Research Grants Council of the Hong Kong SAR, China (CUHK 473408, 471713, 14118718, 14102719). The HAPO Study was funded by the National Institute of Child Health and Human Development (grant no. R01-HD34242) and the National Institute of Diabetes and Digestive and Kidney Diseases (grant no. R01-HD34243). The funding sources do not have any role in the design, interpretation of the study, or the decision to publish the results.

Acknowledgements

Ronald Ching-wan Ma and Wing Hung Tam are the principal investigators of the HAPO-HK Study and TGDM-NDM Study; Ying Wang and Xilin Yang are the principal investigators of the Tianjin Study.

We thank the HAPO Study steering committee for initiating and conducting the original study, and for their kind help and support. We are also grateful to all the study participants for their contribution. Special thanks are extended to all medical and nursing staff at all participating centers for their dedication and professionalism. We thank all team members for their kind assistance, and their efforts on the recruitment of patients and data collection. We acknowledge the participants and investigators of the GenDIP Consortium for making their data available for analysis.

Fig. 1.

Results for meta-analysis of genome-wide association study for gestational diabetes. (A) Manhattan plot. The y-axis represents the −log₁₀ P value (adjusted for principal components and age), and the x-axis represents the 6,322,337 analyzed biallelic single nucleotide polymorphisms. The dashed red horizontal line corresponds to the genome-wide significance threshold for P<5×10^–8. There are 4 points with P<5×10^–8, and the arrow and labels localize the susceptibility loci to gestational diabetes mellitus (GDM) discovered in the present study. (B) Quantile-quantile (Q-Q) plot. The dotted line corresponds to the null hypothesis. TBR1, T-box brain transcription factor 1; SLC4A10, solute carrier family 4 member 10; CDKAL1, CDK5 regulatory subunit-associated protein 1-like 1; MTNR1B, melatonin receptor 1B; INS-IGF2, insulin-insulin-like growth factor 2; KCNQ1, potassium voltage-gated channel subfamily Q member 1.

Fig. 2.

Results for the two genome-wide significant loci for gestational diabetes. (A) Forest plot for the association between T-box brain transcription factor 1 (TBR1)-solute carrier family 4 member 10 (SLC4A10) rs117781972 and gestational diabetes mellitus (GDM) in all discovery and replication cohorts. Odds ratio (OR) and 95% confidence interval (CI) were reported according to the A-allele of rs117781972 (i.e., the GDM-associated risk allele). (B) Forest plot for the association between melatonin receptor 1B (MTNR1B) rs7945617 and GDM in all discovery and replication cohorts. ORs and 95% CIs were reported according to the C-allele of rs7945617 (i.e., the GDM-associated risk allele). A total of three studies (i.e., the Hyperglycemia and Adverse Pregnancy Outcome-Hong Kong [HAPO-HK] Study, the Tianjin Study, and the Treated GDM Cases vs. Non-diabetes Controls [TGDM-NDM] Study) were included in the “meta-analysis of discovery cohorts.” For the GENetics of Diabetes In Pregnancy (GenDIP) meta-analysis, the P values of the associations were obtained from the meta-regression implemented in Meta-Regression of Multi-AncEstry Genetic Association (MR-MEGA) and the combined OR and 95% CI was estimated by meta-analysis under a fixed effect model. (C) Regional plot of the TBR1-SLC4A10 locus. (D) Regional plot of the MTNR1B locus. The purple diamonds represent the sentinel single nucleotide polymorphisms (SNPs) rs117781972 and rs7945617 identified from the meta-analysis of genome-wide association studies. Other SNPs are colored according to their level of linkage disequilibrium, which is measured by r², with the sentinel SNPs. The recombination rates estimated from the 1000 Genomes project Asian data are shown. The genes in the interval are indicated in the bottom panel. TANK, TRAF family member associated NFKB activator; PSMD14, proteasome 26S subunit, non-ATPase 14; AHCTF1P1, AT-hook containing transcription factor 1 pseudogene 1; FAT3, FAT atypical cadherin 3; CCDC67, coiled-coil domain containing 87.

Fig. 3.

Forest plot for the association between potassium voltage-gated channel subfamily Q member 1 (KCNQ1) rs2237897 and gestational diabetes mellitus in all discovery and replication cohorts. Odds ratio (OR) and 95% confidence interval (CI) were reported according to the C-allele of rs2237897 (i.e., the type 2 diabetes mellitus-associated risk allele). For the GENetics of Diabetes In Pregnancy (GenDIP) meta-analysis, the P value of the association was obtained from the meta-regression implemented in Meta-Regression of Multi-AncEstry Genetic Association (MR-MEGA) and the combined OR and 95% CI was estimated by meta-analysis under a fixed effect model. A total of three studies (i.e., the Hyperglycemia and Adverse Pregnancy Outcome-Hong Kong [HAPO-HK] Study, the Tianjin Study, and the Tianjin Study and the Treated GDM Cases vs. Non-diabetes Controls [TGDM-NDM] Study) were included in the “meta-analysis of discovery cohorts.” A total of six studies (i.e., the HAPO-HK Study, the Tianjin Study, the TGDM-NDM Study, the Guangzhou Study, the FinnGen Study, and the Mexican Study) were included in the “overall meta-analysis.” We did not include the GenDIP samples in the overall meta-analysis because they overlapped with both the HAPO-HK and FinnGen Studies.

Fig. 4.

Association between quintiles of polygenic risk score (PRS) and gestational diabetes. (A) Meta-analysis of three discovery cohorts of Chinese women (the Hyperglycemia and Adverse Pregnancy Outcome-Hong Kong [HAPO-HK] Study, Tianjin Study, and Tianjin Study and the Treated GDM Cases vs. Non-diabetes Controls [TGDM-NDM] Study). (B) Guangzhou Study. (C) HAPO-Thai Study. (D) HAPO-Hispanic Study. P_linear is the P value testing for a linear trend across the quintile categories of PRS. P_top is the P value testing for the association of a high PRS with gestational diabetes mellitus (GDM) by comparing the top 20% with the remaining 80% of the PRS distribution. Odds ratio (OR) and 95% confidence interval (CI) of GDM were stratified by quintile categories of PRS. Within each individual cohort, P values were obtained from logistic regression with the adjustment of principal components, age and body mass index, except for the Guangzhou Study which did not adjust for any covariates. Results from the three discovery cohorts were then meta-analyzed using a fixed-effects model.

Fig. 5.

Association between quintiles of polygenic risk score (PRS) and abnormal glucose tolerance (AGT) at 7-year postpartum in the Hyperglycemia and Adverse Pregnancy Outcome-Hong Kong (HAPO-HK) Study. (A) PRS derived based on four gestational diabetes mellitus (GDM)-related variants. (B) PRS derived based on 286 type 2 diabetes mellitus (T2DM)-related variants. The T2DM-related PRS was derived based on 286 T2DM-related variants reported by the Diabetes Meta-Analysis of Trans-Ethnic association studies (DIAMANTE) consortium [20]. P_linear is the P value testing for a linear trend across the quintile categories of PRS. P_top is the P value testing for the association of a high PRS with AGT after pregnancy by comparing the top 20% with the remaining 80% of the PRS distribution. Odds ratio (OR) and 95% confidence interval (CI) of GDM were stratified by quintile categories of PRS. P values were obtained from logistic regression with the adjustment of principal components, age and body mass index.

REFERENCES

• 1. Wang H, Li N, Chivese T, Werfalli M, Sun H, Yuen L, et al. IDF Diabetes Atlas: estimation of global and regional gestational diabetes mellitus prevalence for 2021 by International Association of Diabetes in Pregnancy Study Group’s criteria. Diabetes Res Clin Pract 2022;183:109050.ArticlePubMed
• 2. HAPO Study Cooperative Research Group, Metzger BE, Lowe LP, Dyer AR, Trimble ER, Chaovarindr U, et al. Hyperglycemia and adverse pregnancy outcomes. N Engl J Med 2008;358:1991-2002.ArticlePubMed
• 3. Vounzoulaki E, Khunti K, Abner SC, Tan BK, Davies MJ, Gillies CL. Progression to type 2 diabetes in women with a known history of gestational diabetes: systematic review and meta-analysis. BMJ 2020;369:m1361.ArticlePubMedPMC
• 4. Lowe WL Jr, Lowe LP, Kuang A, Catalano PM, Nodzenski M, Talbot O, et al. Maternal glucose levels during pregnancy and childhood adiposity in the Hyperglycemia and Adverse Pregnancy Outcome Follow-up Study. Diabetologia 2019;62:598-610.ArticlePubMedPMCPDF
• 5. Lowe WL Jr, Scholtens DM, Kuang A, Linder B, Lawrence JM, Lebenthal Y, et al. Hyperglycemia and Adverse Pregnancy Outcome Follow-up Study (HAPO FUS): maternal gestational diabetes mellitus and childhood glucose metabolism. Diabetes Care 2019;42:372-80.PubMedPMC
• 6. Tam WH, Ma RC, Ozaki R, Li AM, Chan MH, Yuen LY, et al. In utero exposure to maternal hyperglycemia increases childhood cardiometabolic risk in offspring. Diabetes Care 2017;40:679-86.ArticlePubMedPMCPDF
• 7. Powe CE, Kwak SH. Genetic studies of gestational diabetes and glucose metabolism in pregnancy. Curr Diab Rep 2020;20:69.ArticlePubMedPMCPDF
• 8. Pervjakova N, Moen GH, Borges MC, Ferreira T, Cook JP, Allard C, et al. Multi-ancestry genome-wide association study of gestational diabetes mellitus highlights genetic links with type 2 diabetes. Hum Mol Genet 2022;31:3377-91.PubMedPMC
• 9. Elliott A, Walters RK, Pirinen M, Kurki M, Junna N, Goldstein JI, et al. Distinct and shared genetic architectures of gestational diabetes mellitus and type 2 diabetes. Nat Genet 2024;56:377-82.ArticlePubMedPMCPDF
• 10. Tian Y, Li P. Genetic risk score to improve prediction and treatment in gestational diabetes mellitus. Front Endocrinol (Lausanne) 2022;13:955821.ArticlePubMedPMC
• 11. Kawai VK, Levinson RT, Adefurin A, Kurnik D, Collier SP, Conway D, et al. A genetic risk score that includes common type 2 diabetes risk variants is associated with gestational diabetes. Clin Endocrinol (Oxf) 2017;87:149-55.ArticlePubMedPMCPDF
• 12. Leng J, Shao P, Zhang C, Tian H, Zhang F, Zhang S, et al. Prevalence of gestational diabetes mellitus and its risk factors in Chinese pregnant women: a prospective population-based study in Tianjin, China. PLoS One 2015;10:e0121029.ArticlePubMedPMC
• 13. Tam WH, Ma RCW, Ozaki R, Li AM, Chan MHM, Yuen LY, et al. Antenatal treatment of gestational diabetes and offspring’s future cardiometabolic risk. In: The 9th International Symposium on Diabetes, Hypertension and Metabolic Syndrome and in Pregnancy; 2017 Mar 8-12; Barcelona, Spain.
• 14. Ko GT, Chan JC, Chan AW, Wong PT, Hui SS, Tong SD, et al. Association between sleeping hours, working hours and obesity in Hong Kong Chinese: the ‘better health for better Hong Kong’ health promotion campaign. Int J Obes (Lond) 2007;31:254-60.ArticlePubMedPDF
• 15. Wu L, Song Y, Zhang Y, Liang B, Deng Y, Tang T, et al. Novel genetic variants of PPARγ2 promoter in gestational diabetes mellitus and its molecular regulation in adipogenesis. Front Endocrinol (Lausanne) 2021;11:499788.ArticlePubMedPMC
• 16. International Association of Diabetes and Pregnancy Study Groups Consensus Panel, Metzger BE, Gabbe SG, Persson B, Buchanan TA, Catalano PA, et al. International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care 2010;33:676-82.ArticlePubMedPMCPDF
• 17. ElSayed NA, Aleppo G, Aroda VR, Bannuru RR, Brown FM, Bruemmer D, et al. 2. Classification and diagnosis of diabetes: standards of care in diabetes-2023. Diabetes Care 2023;46(Suppl 1):S19-40.
• 18. Kwak SH, Kim SH, Cho YM, Go MJ, Cho YS, Choi SH, et al. A genome-wide association study of gestational diabetes mellitus in Korean women. Diabetes 2012;61:531-41.ArticlePubMedPMCPDF
• 19. Changalidis AI, Maksiutenko EM, Barbitoff YA, Tkachenko AA, Vashukova ES, Pachuliia OV, et al. Aggregation of genomewide association data from FinnGen and UK Biobank replicates multiple risk loci for pregnancy complications. Genes (Basel) 2022;13:2255.ArticlePubMedPMC
• 20. Mahajan A, Spracklen CN, Zhang W, Ng MC, Petty LE, Kitajima H, et al. Multi-ancestry genetic study of type 2 diabetes highlights the power of diverse populations for discovery and translation. Nat Genet 2022;54:560-72.PubMedPMC
• 21. Huerta-Chagoya A, Vazquez-Cardenas P, Moreno-Macias H, Tapia-Maruri L, Rodriguez-Guillen R, Lopez-Vite E, et al. Genetic determinants for gestational diabetes mellitus and related metabolic traits in Mexican women. PLoS One 2015;10:e0126408.PubMedPMC
• 22. Ao D, Wang HJ, Wang LF, Song JY, Yang HX, Wang Y. The rs22 37892 polymorphism in KCNQ1 influences gestational diabetes mellitus and glucose levels: a case-control study and meta-analysis. PLoS One 2015;10:e0128901.ArticlePubMedPMC
• 23. Yasuda K, Miyake K, Horikawa Y, Hara K, Osawa H, Furuta H, et al. Variants in KCNQ1 are associated with susceptibility to type 2 diabetes mellitus. Nat Genet 2008;40:1092-7.ArticlePubMedPDF
• 24. Pascoe L, Tura A, Patel SK, Ibrahim IM, Ferrannini E, Zeggini E, et al. Common variants of the novel type 2 diabetes genes CDKAL1 and HHEX/IDE are associated with decreased pancreatic beta-cell function. Diabetes 2007;56:3101-4.PubMed
• 25. Lyssenko V, Nagorny CL, Erdos MR, Wierup N, Jonsson A, Spegel P, et al. Common variant in MTNR1B associated with increased risk of type 2 diabetes and impaired early insulin secretion. Nat Genet 2009;41:82-8.PubMed
• 26. Zeggini E, Weedon MN, Lindgren CM, Frayling TM, Elliott KS, Lango H, et al. Replication of genome-wide association signals in UK samples reveals risk loci for type 2 diabetes. Science 2007;316:1336-41.PubMedPMC
• 27. Okamura T, Yanobu-Takanashi R, Takeuchi F, Isono M, Akiyama K, Shimizu Y, et al. Deletion of CDKAL1 affects high-fat diet-induced fat accumulation and glucose-stimulated insulin secretion in mice, indicating relevance to diabetes. PLoS One 2012;7:e49055.PMC
• 28. Wei FY, Suzuki T, Watanabe S, Kimura S, Kaitsuka T, Fujimura A, et al. Deficit of tRNA(Lys) modification by Cdkal1 causes the development of type 2 diabetes in mice. J Clin Invest 2011;121:3598-608.PMC
• 29. Shimizu I, Yoshida Y, Minamino T. A role for circadian clock in metabolic disease. Hypertens Res 2016;39:483-91.PubMed
• 30. Yamagata K, Senokuchi T, Lu M, Takemoto M, Fazlul Karim M, Go C, et al. Voltage-gated K+ channel KCNQ1 regulates insulin secretion in MIN6 β-cell line. Biochem Biophys Res Commun 2011;407:620-5.PubMed
• 31. Deriziotis P, O’Roak BJ, Graham SA, Estruch SB, Dimitropoulou D, Bernier RA, et al. De novo TBR1 mutations in sporadic autism disrupt protein functions. Nat Commun 2014;5:4954.
• 32. den Hoed J, Sollis E, Venselaar H, Estruch SB, Deriziotis P, Fisher SE. Functional characterization of TBR1 variants in neurodevelopmental disorder. Sci Rep 2018;8:14279.PMC
• 33. Jacobs S, Ruusuvuori E, Sipila ST, Haapanen A, Damkier HH, Kurth I, et al. Mice with targeted Slc4a10 gene disruption have small brain ventricles and show reduced neuronal excitability. Proc Natl Acad Sci U S A 2008;105:311-6.PubMed
• 34. Davies G, Lam M, Harris SE, Trampush JW, Luciano M, Hill WD, et al. Study of 300,486 individuals identifies 148 independent genetic loci influencing general cognitive function. Nat Commun 2018;9:2098.PubMedPMC
• 35. Trubetskoy V, Pardinas AF, Qi T, Panagiotaropoulou G, Awasthi S, Bigdeli TB, et al. Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature 2022;604:502-8.PubMedPMC
• 36. Williams CM, Labouret G, Wolfram T, Peyre H, Ramus F. A general cognitive ability factor for the UK Biobank. Behav Genet 2023;53:85-100.PubMed
• 37. Xiang AH, Wang X, Martinez MP, Walthall JC, Curry ES, Page K, et al. Association of maternal diabetes with autism in offspring. JAMA 2015;313:1425-34.PubMed
• 38. Evangelou E, Warren HR, Mosen-Ansorena D, Mifsud B, Pazoki R, Gao H, et al. Genetic analysis of over 1 million people identifies 535 new loci associated with blood pressure traits. Nat Genet 2018;50:1412-25.PubMedPMC
• 39. Sakaue S, Kanai M, Tanigawa Y, Karjalainen J, Kurki M, Koshiba S, et al. A cross-population atlas of genetic associations for 220 human phenotypes. Nat Genet 2021;53:1415-24.PubMed
• 40. Huang J, Huffman JE, Huang Y, Do Valle I, Assimes TL, Raghavan S, et al. Genomics and phenomics of body mass index reveals a complex disease network. Nat Commun 2022;13:7973.PubMedPMC
• 41. Zhu Z, Guo Y, Shi H, Liu CL, Panganiban RA, Chung W, et al. Shared genetic and experimental links between obesity-related traits and asthma subtypes in UK Biobank. J Allergy Clin Immunol 2020;145:537-49.PubMed
• 42. Popova PV, Klyushina AA, Vasilyeva LB, Tkachuk AS, Vasukova EA, Anopova AD, et al. Association of common genetic risk variants with gestational diabetes mellitus and their role in GDM prediction. Front Endocrinol (Lausanne) 2021;12:628582.PubMedPMC
• 43. Powe CE, Nodzenski M, Talbot O, Allard C, Briggs C, Leya MV, et al. Genetic determinants of glycemic traits and the risk of gestational diabetes mellitus. Diabetes 2018;67:2703-9.PubMedPMC
• 44. Shen Y, Jia Y, Li Y, Gu X, Wan G, Zhang P, et al. Genetic determinants of gestational diabetes mellitus: a case-control study in two independent populations. Acta Diabetol 2020;57:843-52.ArticlePubMedPDF
• 45. Ding M, Chavarro J, Olsen S, Lin Y, Ley SH, Bao W, et al. Genetic variants of gestational diabetes mellitus: a study of 112 SNPs among 8722 women in two independent populations. Diabetologia 2018;61:1758-68.ArticlePubMedPMCPDF
• 46. Lamri A, Mao S, Desai D, Gupta M, Pare G, Anand SS. Fine-tuning of genome-wide polygenic risk scores and prediction of gestational diabetes in South Asian Women. Sci Rep 2020;10:8941.ArticlePubMedPMCPDF

Figure & Data

References

Citations

Citations to this article as recorded by  

About this article

Tam CHt, Wang Y, Wang CC, Yuen LY, Lim CKp, Leng J, Wu L, Ng ACw, Hou Y, Tsoi KY, Wang H, Ozaki R, Li AM, Wang Q, Chan JCn, Ye YC, Tam WH, Yang X, Ma RCw. Identification and Potential Clinical Utility of Common Genetic Variants in Gestational Diabetes among Chinese Pregnant Women. Diabetes Metab J. 2024 Sep 20. doi: 10.4093/dmj.2024.0139. Epub ahead of print.

Received: Mar 20, 2024; Accepted: Jun 17, 2024

DOI: https://doi.org/10.4093/dmj.2024.0139.

Download citation ↓