- Open Access
Lipolysis and gestational diabetes mellitus onset: a case-cohort genome-wide association study in Chinese
Journal of Translational Medicine volume 21, Article number: 47 (2023)
Genetic knowledge of gestational diabetes mellitus (GDM) in Chinese women is quite limited. This study aimed to identify the risk factors and mechanism of GDM at the genetic level in a Chinese population.
We conducted a genome-wide association study (GWAS) based on single nucleotide polymorphism (SNP) array genotyping (ASA-CHIA Bead chip, Illumina) and a case-cohort study design. Variants including SNPs, copy number variants (CNVs), and insertions-deletions (InDels) were called from genotyping data. A total of 2232 pregnant women were enrolled in their first/second trimester between February 2018 and December 2020 from Anqing Municipal Hospital in Anhui Province, China. The GWAS included 193 GDM patients and 819 subjects without a diabetes diagnosis, and risk ratios (RRs) and their 95% confidence intervals (CIs) were estimated by a regression-based method conditional on the population structure. The calling and quality control of genotyping data were performed following published guidelines. CNVs were merged into CNV regions (CNVR) to simplify analyses. To interpret the GWAS results, gene mapping and overexpression analyses (ORAs) were further performed to prioritize the candidate genes and related biological mechanisms.
We identified 14 CNVRs (false discovery rate corrected P values < 0.05) and two suggestively significant SNPs (P value < 0.00001) associated with GDM, and a total of 19 candidate genes were mapped. Ten genes were significantly enriched in gene sets related to lipase (triglyceride lipase and lipoprotein lipase) activity (LIPF, LIPK, LIPN, and LIPJ genes), oxidoreductase activity (TPH1 and TPH2 genes), and cellular components beta-catenin destruction complex (APC and GSK3B genes), Wnt signalosome (APC and GSK3B genes), and lateral element in the Gene Ontology resource (BRCA1 and SYCP2 genes) by two ORA methods (adjusted P values < 0.05).
Genes related to lipolysis, redox reaction, and proliferation of islet β-cells are associated with GDM in Chinese women. Energy metabolism, particularly lipolysis, may play an important role in GDM aetiology and pathology, which needs further molecular studies to verify.
Gestational diabetes mellitus (GDM), defined as abnormally high blood glucose levels among pregnant women without a diabetes diagnosis before pregnancy, is a common complication in pregnancy, and its incidence is increasing worldwide [1, 2]. GDM is associated with macrosomia, which is usually accompanied by dystocia, birth trauma, and caesarean section . GDM patients have higher risks of developing type 2 diabetes mellitus (T2DM), cardiovascular diseases (CVD), and metabolic syndrome (MS) after delivery [4,5,6,7,8,9,10,11], and their offspring also have increased risks for obesity, CVD, diabetes, and MS [8, 12,13,14,15].
It is suggested that approximately 80% of GDM cases are related to insulin resistance (IR) . Other hypotheses involving inflammation, oxidative stress, and adipose tissue/endothelial cell dysfunction have also been proposed . However, the pathogenesis of GDM has not been fully clarified . Many risk factors for GDM, including demography, family/fertility history behaviours in pregnancy, and genetic variants, have been identified [12, 16, 18, 19]; however, confounding bias is a problem affecting most studies because of the unclear onset time of GDM. Most genetic studies were candidate gene studies based on T2DM due to the potential common genetic background between the two conditions [20,21,22,23], and many shared genes, such as TCF7L2, GCK, MTNR1B, and PPARγ genes, have been found [24, 25]. However, dysglycaemia triggered by pregnancy could have specific mechanisms. Some GDM-associated variants, such as the HKDC1, MTNR1A, ACE, and VDR genes, have not been reported in T2DM genetic studies . To our knowledge, only two genome-wide association studies (GWASs) of GDM have been reported. One study observed 468 GDM cases and 1242 nondiabetic controls in Korea . This study’s controls were women aged ≥ 50 years with unknown GDM histories, which may introduce misclassification bias. The other was conducted in a smaller Chinese sample (103 GDM cases, 115 nondiabetic controls) , and no overlapping markers and genes were identified with the previous GWAS.
Screening for GDM based on risk factors, including obesity, first-degree relatives with diabetes, history of GDM or adverse pregnancy outcomes, and glycosuria, will miss approximately one-half of cases [12, 28]. Furthermore, there is no useful predictive model for GDM . The diagnosis of GDM is based on a time-consuming and daunting oral glucose tolerance test (OGTT). Preventive actions are still contentious and ineffective [6, 29,30,31,32,33]. The available management strategy for GDM is limited if only lifestyle changes and the use of insulin or oral antidiabetics are considered [31,32,33]. In summary, further studies of the pathogenesis of GDM are imperative to put forwards innovative precocious preventive or therapeutic interventions, and GWASs could help find clues from the whole genome level with few confounders and without a hypothesis.
The average incidence of GDM in mainland China was 14.8% in recent years, and relevant genetic knowledge for Chinese individuals is quite limited [27, 34,35,36,37]. To further identify the risk factors and mechanism of GDM in Chinese individuals at the genetic level, we conducted a GWAS based on a case-cohort design in Anqing, China.
Study subjects and phenotype diagnosis
The cohort enrolled 2232 pregnant women in the first/second trimester between February 2018 and December 2020 from Anqing Municipal Hospital in Anhui Province, China. Participants were followed up around their expected date of delivery. The demographical information, physical measurements, behavioural information before and during pregnancy, and medical histories of subjects were collected from structured questionnaires at baseline, structured questionnaires or phone interviews at follow-up, or medical records. The cohort profile is shown in (Additional file 1: Fig. S1). All subjects provided written informed consent when enrolled. Ethical approval for this study was obtained from the Medical Ethics Committee of Fudan University, School of Public Health (IRB00002408 & FWA00002399, Approval number: IRB#2017-09-0636).
A total of 203 GDM cases were identified using medical records, and these patients were diagnosed with GDM in the present pregnancy according to the Chinese guidelines for diagnosis and treatment of gestational diabetes mellitus (2014) by physicians. Pregnant women without diabetes before pregnancy met one or more of the following criteria: (a) blood glucose after 75 g OGTT ≥ 10.0 mmol/L in the first hour or ≥ 8.5 mmol/L but < 11.1 mmol/L in second hour and (b) 5.1 mmol/L ≤ fasting blood glucose < 7.0 mmol/L after the 23rd gestational week (only for pregnant women who have no access to an OGTT) .
A subcohort was sampled from the cohort by systematic sampling (interval = 1). Finally, 193 GDM patients (93 in the subcohort) and 819 subjects without diabetes diagnosis (controls) were selected for further analyses per the following criteria: singleton pregnancy, in the first/second trimester, aged 18 years or more, Chinese nationality, without a history of antibiotic use in the last four weeks at baseline, without systemic or organ diseases (such as diabetes, cardiovascular diseases, polycystic ovary syndrome) before pregnancy, with a gestational age of termination ≥ 30 weeks, and with genotyping data (Fig. 1).
Biosample and genotyping
A 1.5-mL peripheral blood sample was collected from each subject at baseline for DNA extraction. All samples with a call rate > 95% were genotyped using an ASA-CHIA (Infinium Asian Screening Array-China Health Industry Alliance) Bead chip (Illumina, Inc., San Diego, California, USA) according to the Illumina Infinium HTS chip standard operating procedures (SOP) . All genetic data sets used in this study were based on the human genome Hg19 assembly, and the rsIDs were mapped to dbSNP build 146 if necessary.
Preprocessing of genotyped data
Single nucleotide polymorphisms (SNPs), insertions-deletions (InDels), and copy number variants (CNVs) were obtained using GenomeStudio version 2 according to Illumina GenomeStudio genotyping QC SOP v.1.6 .
Quality control (QC) and imputation for SNPs and InDels
For SNPs and InDels, QC was conducted per the following steps: merging the overlapped loci based on the Illumina GenTrain score (those with lower scores were excluded) and removing the loci on the Y chromosome using SAS 9.4. Subsequent QC steps were done following the guideline provided by Weale et al.  using PLINK  version 1.9 and R version 4.1.1, in which variants and individuals were excluded based on the following criteria: (1) individual and variant missingness > 0.02, (2) variants with minor allele frequency (MAF) < 0.05, (3) variants deviating from Hardy–Weinberg equilibrium (HWE) with P value < 1e− 10 in cases and < 1e− 6 in controls, (4) individuals deviating ± 3 standard deviation (SD) from the samples' heterozygosity rate mean, (5) individuals with a lower call rate from sample pairs with relatedness pi-hat value > 0.2, or (6) variants sharing the same coordinate and allele code. Genotype imputation was performed with the 1000 Genomes Project phase 3 East Asian population (1000G3 EAS)  using IMPUTE2  for autosomal chromosomes and chromosome X. Only variants with info metric > 0.9, MAF < 0.01, missingness > 0.02, and unique coordinate and allele codes were kept.
Quality control and merging for CNVs
QC for CNV was performed by PennCNV  version 1.0.5 following the protocols provided by Lin et al. . Signal intensity files obtained from GenomeStudio version 2 were split into one sample per file by column.pl script. The population frequency of the B allele file was compiled by compile_pfb.pl script. CNVs were called by the detect_cnv.pl script with a GCmodel adjustment to reduce false positive calls, and the cal_gc_snp.pl script was used to generate the customized GC model file. QC for called CNVs was conducted only for CNV calls with SNP numbers ≥ 10 by running filter_cnv.pl. Called CNVs in HLA and genomic regions near the centromeres or the telomeres within one million bases (Mb) were removed. CNVs with confidence scores < 10 and lengths > 5 Mb were further filtered by the HandyCNV R package . Called CNVs from the same genomic locus can have various start and end points across individuals; thus, CNVs were merged into common CNV regions (CNVRs) by CNVRuler  version 1.2 trimming with CNV regional density < 0.1 and MAF < 0.05 to simplify the analysis.
Annotating and prioritizing variants and genes from SNP/InDel-GWAS results in FUMA
Most hits from GWAS are in noncoding or intergenic regions and typically cannot be directly translated into causal variants; therefore, prioritizing the most likely causal variants and genes based on functional and biological knowledge is necessary. The FUMA  version 1.4.1 SNP2GENE process was used to analyse computed LD structure, to annotate functions to variants, to prioritize candidate genes based on functional gene mapping, and to perform gene-level analyses using GWAS summary statistics of variants with P values < 0.01. The total sample size was 985. Parameters were set as recommended by FUMA if not explained (Additional file 1: Note S1).
Characterizing genomic loci based on GWAS
Independent significant SNPs/InDels (ind. sig. markers) were defined as markers with P values ≤ 1e− 5 (suggestive significance level) and those that were independent of each other at r2 < 0.6, among which the lead markers were identified at r2 < 0.1. All known markers in the 1000G3 EAS reference panel or with P values < 0.05 in input GWAS statistics having r2 ≥ 0.6 with one of the ind. sig. markers (candidate markers) were included for further annotation. The LD blocks of ind. sig. markers located to each other within 250 kilobases (kb) were merged into one genomic locus. The data from the 1000G3 EAS were used to gauge the LD structure. Markers and genes in the MHC region were excluded for the complicated LD structure.
Functional gene mapping
Three strategies for gene mapping were applied to GWAS summary statistics with the following settings:
Positional mapping identifies the genes associated with candidate markers. The mapping was based on ANNOVAR annotations within a 10-kb physical distance between candidate markers and genes from Ensembl genes build 92.
Expression quantitative trait loci (eQTL) mapping maps markers to genes that likely affect the expression of those genes up to 1 Mb (cis-eQTL). The mapping was performed based on combined evidence from multiple eQTL data sources provided by FUMA (Additional file 1: Note S1). Only significant marker-gene pairs with a FDR P value < 0.05 were used.
Chromatin interactions (CI) regulate gene expression by bringing distal regulatory elements, such as superenhancers, to promoters in close spatial proximity. CI mapping was performed to map markers to genes if there was a three-dimensional DNA–DNA interaction between the marker region and another gene region without a distance boundary. For mapping databases, please see Additional file 1: Note S1. Interactions were filtered by a FDR P value < 1e− 6. Genes that were 250 bp upstream and 500 bp downstream of the transcription start site and that overlapped with the significantly interacting regions were mapped to further prioritize candidate genes. Predicted enhancer and promoter regions from the Roadmap epigenomics project for 111 epigenomes were also annotated to interaction regions.
Gene-based analysis by MAGMA
Gene-based analysis was performed for the input summary statistics using MAGMA version 1.08 in FUMA to increase the power of detecting genotype–phenotype associations. The Bonferroni correction was used to correct multiple testing of gene-based P values.
Annotating genes for CNVRs
The HandyCNV R package was used to identify the genes located in each CNVR, and the reference gene panel was Human Release 19 version. CNVRs with a frequency of less than 0.05 were excluded because of their extremely high RRs and small P values.
WebGestalt  and g:Profiler  were used to perform overrepresentation analysis of genes. Genes mapped by more than two mapping methods (positional mapping, eQTL mapping, CI mapping, or Bonferroni corrected P value < 0.05 in MAGMA gene analysis) and annotated genes of significantly associated CNVRs were combined in these analyses as candidate genes. For parameters for WebGestalt and g:Profiler, please see (Additional file 1: Note S2, S3). Additional file 1: Fig. S2 shows the whole analysis strategy after variant calling.
For continuous variables, the normality of distribution was tested by the one-sample Shapiro‒Wilk test (when sample size ≤ 2000) or Kolmogorov‒Smirnov test (when the sample size was > 2000). Continuous variables with a normal distribution are presented as the mean (SD), and those with a nonnormal distribution are presented as the median (interquartile range [IQR]). Means of 2 continuous normally distributed variables were compared by independent samples Student’s t-test. The Wilcoxon test and Kruskal‒Wallis test were used to compare the means of 2 and 3 or more groups of variables with nonnormal distributions, respectively. The frequencies of categorical variables were compared using Fisher’s exact test. For all statistical tests, a P value < 0.05 was considered significant . All analyses above were performed in SAS 9.4 (SAS Institute Inc., NC, USA). Associations for SNPs, InDels, and CNVRs with GDM were estimated using the regression-based method for the case-cohort study proposed by Chui et al. . The risk ratios (RRs), their 95% confidence intervals (CIs), and P values were reported. SNPs/InDels with a P value < 1e− 5 were defined as suggestively significant, and those with a P value < 1.61e− 7 (the empirical genome-wide significance threshold for SNP-GWAS in the East Asian population ) were defined as genome-wide significant. The study-wide P value threshold was defined as a false discovery rate (FDR)-corrected threshold assuming the whole false positive rate at 0.05 for CNVRs. All association analyses were performed in R 4.1.1.
According to the law of causality, the relation between genotypes and phenotypes is affected by only a few potential confounders, of which the most important is population stratification (PS). It can be estimated by the distinctions of observing genotypes among populations . For SNPs/InDels, PS was calculated using the linkage-disequilibrium-pruned (LD-pruned) (window size 1500 variant count, step size 150 variant count, and pairwise r2 < 0.2) variant set on autosomal chromosomes [53, 54]. The top 50 principal components (PCs) of the variance-standardized relationship matrix were extracted using PLINK 1.9. Then, the “twstats” program of Eigensoft  version 6.1.4 was adopted to select the statistically significant PC axes (P value < 0.01)  to be included as covariates for the association analyses. Three PCs for CNVRs were selected using CNVRuler.
Characteristics of participants and the representativeness of the subcohort
The median age of the study cohort was 28.0 years with an IQR of 25.0 to 30.0 years, and the median gestational age at baseline was 16.4 (IQR: 15.9–17.1) weeks. Approximately 99.2% (1953/1968) of the subjects in the cohort were of Han nationality. Balance tests for the subcohort group and nonselected group verified the representativeness of the sampling, in which the difference was significant only for the weight before pregnancy (P = 0.006) (Additional file 1: Table S1). Based on the case-cohort design and inclusion criteria, this study included 193 GDM patients and 912 subcohort subjects with 93 overlapping GDM cases between the two groups. The demographic characteristics of the study subjects are shown in Table 1.
SNPs/InDels associated with GDM and gene mapping
A total of 1,544,258 SNPs and 123,886 InDels (referred to as “input markers”) from 985 subjects (190 cases and 795 controls) were included in the GWAS based on the imputation of 318,716 SNPs and 470 InDels. Three PCs were selected as covariates based on the “twstats” method (Additional file 1: Table S2). The inflation factor for the distribution of P values (Additional file 1: Fig. S3) was λGC = 1.01. There was no genome-wide significant marker, but two reached the suggestive significance level: SNP rs78175392:T:C (MAF: 0.14 in GDM group, 0.07 in non-DM group) in the intronic region of SLC12A8 gene on the chromosome (Chr) 3 (RR: 2.0 [1.5–2.7], P value: 3.5e− 06) and SNP rs12253503:A:G (MAF: 0.58 in GDM group, 0.45 in non-DM group) in the intergenic region of RP11-186O14.7 gene on Chr10 (RR: 1.6 [1.3–2.0], P value: 9.0e− 06) (Fig. 2, Additional file 1: Table S3). These two were defined as independent significant markers.
A total of 39 candidate markers (markers in LD with at least one independent significant marker at r2 ≥ 0.6) in two genomic risk loci (Fig. 3) were mapped to 64 genes by three mapping methods, including three by positional mapping (protein coding genes: SLC12A8 and LIPF; pseudogene: RP11-186O14.7), nine by eQTL mapping, and 61 genes by CI mapping (Additional file 1: Table S4). The SLC12A8 gene on Chr 3 and the RNLS, LIPJ, LIPK, KRT8P38, and LIPN genes on Chr 10 were mapped by eQTL mapping and CI mapping, and the LIPF and RP11-186O14.7 genes were also mapped by CI mapping (Fig. 3).
All input markers were mapped to 1007 protein coding genes in MAGMA gene-based analysis, and twelve reached the Bonferroni corrected significance threshold (Additional file 1: Table S5).
Association analyses between CNVs and GDM and gene mapping
A total of 51,688 CNVs were detected and filtered by PennCNV. These CNVs were merged into 220 CNVRs that were included in the association analyses with three PC covariates. Fourteen CNVRs were defined as significantly associated with GDM through FDR correction (Fig. 4) and mapped to 12 genes (Table 2, Additional file 1: Table S6) (Fig. 5).
Overrepresentation analyses (ORA) of mapped genes
Seven genes mapped by at least two mapping methods in the SNP/InDel part (with RP11-186O14.7 gene removed because of a lack of information in all available resources) and 12 genes mapped in the CNV part were included in the ORA as candidate genes of GDM. The 19 genes included 18 protein-coding genes and 1 pseudogene, and their primary molecular functions are shown in Table 2. Details of the genes can be found in the NCBI and GeneCards databases (Additional file 1: Table S7). In WebGestalt analysis, the 19 candidate genes were mapped to 19 unique Entrezgene IDs and were classified into different biological process, cellular component, and molecular categories (Fig. 5a–c). For biological process categories, 15 genes were related to metabolic processes, and 14 were related to biological regulation. Most of these 19 genes were associated mainly with molecular binding processes, such as ion binding (N = 9), protein binding (N = 7), and nucleic acid binding (N = 6). The ORA of genes identified 7 gene sets in WebGestalt (Table 3) and 16 gene sets (Table 3, Additional file 1: Table S8) in g:Profiler. Gene sets enriched by WebGestalt were related to lipase activity (including the LIPF, LIPN, LIPK, and LIPJ genes; Table 3), oxidoreductase activity (including the TPH1 and TPH2 genes; Table 3) and molecular function (Fig. 5d) and were connected with 3 cellular components, namely, the beta-catenin destruction complex, Wnt signalosome, and lateral element (Fig. 6). The descriptions and definitions of these 7 gene sets are shown in Table 3. The g:Profiler also identified these 7 gene sets, in addition to one related to tryptophan 5-monooxygenase activity and eight related to biological processes, including the serotonin biosynthetic process, indole-containing compound biosynthetic process, primary amino compound biosynthetic process, positive regulation of protein localization to centrosome, cornification, regulation of type B pancreatic cell development, circadian rhythm, and regulation of protein localization to centrosome (Additional file 1: Table S8).
In this case-cohort GWAS based on SNP array genotyping and imputing data, we aimed to explore the genetic background of GDM in a Chinese population and identified 2 SNPs associated with GDM at a suggestive significance level and 14 CNVRs associated with GDM. A total of 19 genes that may be candidate genetic markers for GDM were mapped through functional mapping, in which twelve (APC, BRCA1, CLOCK, GRIN3B, GSK3B, NR3C1, PRDM16, SALL3, SYCP2, TMEM259, TPH1, and TPH2) were directly mapped by variants genotyped in our study. These 19 candidate genes were enriched mainly in gene sets related to the triglyceride lipase (TGL) activity and lipoprotein lipase (LPL) activity, oxidoreductase activity, and the cellular components beta-catenin destruction complex, Wnt signalosome, and lateral element based on the GO resource.
TGL is a family of lipases that catalyse the first stage of lipolysis by hydrolysing triglyceride (also termed “triacylglycerol” in recent literature, TG) to diacylglycerol (DAG) and fatty acid (FA), while LPL is responsible for hydrolysing TG in blood lipoproteins to DAG and FA . Convincing evidence indicates an association between lipolysis and IR. Kim et al. found that transgenic mice with tissue-specific overexpression of LPL have increased TG and IR in a specific tissue, and there is a causative relationship between the accumulation of intracellular TG and IR . Inhibition of TGL in adipose tissue improves whole-body insulin sensitivity [56, 58,59,60]. DAG acts as a direct activator of many kinases that impair insulin receptor substrates and insulin signalling . Girousse et al. revealed that a high lipolytic rate was associated with low insulin sensitivity in human and partial genetic and pharmacologic inhibition of hormone-sensitive lipase (an lipase intracellular lipolysis) resulted in the improvement of insulin sensitivity in mice , which also supported that lipolysis could affect the insulin sensitivity. Since insulin is the main factor inhibiting lipolysis, decreasing insulin sensitivity and increasing lipolysis in later pregnancy could result in a vicious cycle [58, 63]. Oxidoreductase is a class of enzymes catalysing oxidoreduction reactions (redox), which play an important role in human metabolism . Excessive FA, DAG, and other lipid metabolites could cause an increasing level of free FA (FFA), redox imbalance, a decrease in oxidative capacity in adipose tissue and then excessive ectopic lipid deposits in the muscle, liver, and pancreas, which further aggravates IR and hyperglycaemia . T2DM has been recently postulated to be a redox disease, while reliable evidence is absent . The Wnt signalosome is formed to increase the local concentration and signalling activity of Wnt signals . Beta-catenin is a functional protein that can regulate the expression of Wnt target genes and is the key nuclear effector of the Wnt signalling pathway [67, 68]. Wnt signalling and β-catenin are both necessary and sufficient for the proliferation of islet β-cells [69, 70]. Fair evidence indicates that Wnt/β-catenin signalling favours improved insulin/glucose and lipid homeostasis and that antagonism of this pathway by oxidative stress may contribute to IR and hyperlipidaemia . Conventional dendritic cells with constitutively activated β-catenin induce islet expansion by increasing b-cell proliferation in a mouse model of diet-induced obesity . Beta-catenin signalling could also regulate preadipocyte differentiation, and given that obesity is one of the main risk factors for diabetes, it is conceivable that changes in components of the Wnt signalling pathway could contribute to an increased risk of diabetes by impacting adipogenesis . Taken together, these findings suggest that the onset of GDM may be the result of asynchrony of metabolic changes in pregnancy, in which the insulin signalling pathway and glycometabolism are affected based on the genetic background related to lipometabolism and redox.
To our knowledge, only two GWASs of GDM have been reported. Kwak et al. conducted a two-stage study in Korean women, including 468 GDM cases and 1242 nondiabetic controls using 2.19 million genotyped or imputed markers in the first stage and 931 GDM cases and 783 nondiabetic controls with 11 genotyping loci from the first stage in further study . Nine independent significant SNPs (P value < 2e− 5 and LD r2 < 0.5) were identified in the GWAS, among which rs7754840 in the intron regions of the CDKAL1 gene and rs10830962 located upstream of the MTNR1B gene showed the strongest association with GDM. The controls were selected from women without a T2DM history ≥ 50 years in the first stage and ≥ 50 years in the second stage, which may introduce misclassification bias because of an unknown GDM history. The second GWAS included 103 GDM cases and 115 controls from a population of Chinese Han women and identified 23 SNPs mapping to four genes (CTIF, CDH18, PTGIS, and SYNPR) . No overlapping SNP markers or mapped genes exist among the two and the present study. There are many differences between the three GWASs, including different populations, GDM diagnosis strategies, sample sizes, genotyping chips, imputing methods, and analysis strategies, details are shown in Additional file 1: Table S8; therefore, it is difficult to compare the results from these GWASs, so more GWASs of GDM are needed. Most genetic studies for GDM in humans are candidate gene studies based on other types of diabetes, particularly T2DM. Some genes directly associated with GDM in our study, such as CLOCK , NR3C1 , and PRDM16 , mainly in the European population, have also been reported to be associated with T2DM. However, gene variants associated with a specific phenotype could vary among ethnicities, districts, genetic markers selected, and even sample sizes. It is difficult to define the causality of a genetic variant before more work is done to verify its association with a phenotype in a specific population, so we focused more on explanations of the combined biological functions of associated genes.
GDM has been a public health problem with both short-term and long-term metabolic influences on mothers and children. Converging lines of evidence have shaped GDM as a complex trait caused by genetic and environmental risk factors. However, the pathophysiology of GDM remains unclear and controversial, and its genetics have not been fully explored. Our study suggests that GDM may be the result of alterations in energy metabolism under a specific genetic background, including gene variants related to lipolysis and redox in addition to the Wnt signalling pathway, which affect glucose metabolism mediated by the insulin signalling pathway. This finding adds evidence for the hypothesis that GDM may be a herald of T2DM expressed under the specific metabolic conditions of pregnancy or an excessive manifestation of metabolic alterations occurring in pregnancy  and explains why obesity is a risk factor for GDM at the gene level. With changes in maternal homeostasis to contain the growing foetus, pregnancy represents a unique metabolic state and imposes many metabolic challenges on the mother, making it easier to develop metabolic diseases such as diabetes. This study could provide more insights into the prevention and treatment of GDM by highlighting the importance of lipid storage and metabolism in pregnancy, such as using lipid metabolites as an early screening biomarker for GDM or intervention target to manage GDM. Additionally, instrumental variables could be built based on the genetic variants associated with GDM to estimate the associations between GDM and related outcomes, which could help control the confounding bias in epidemiologic studies [77, 78].
The strengths of the present study are as follows. (1) This study was based on a case-cohort study design, which is cost-effective for GWAS and could produce RR estimates without the rare-disease assumption . (2) Genome-wide SNP array genotyping data imputed with 1000G3 are also cost-effective in exploring the genetic risk factors for a specific phenotype from the whole genome level. (3) Compared to other GWASs of GDM, we analysed the association between genetic variants and GDM at the SNP, InDel, and CNV levels to fully clarify our research questions. (4) To obtain reliable results, we conducted QC of genotyping data according to published guidelines and performed functional analyses using as many parallel methods as we could. However, our study has some inevitable limitations. (1) The majority of subjects enrolled were Han in Anqing, China, and GDM diagnosis is usually different among districts. Thus, the generalizability of our results may be limited because of the population’s genetic specificity and the nonuniform phenotype definition. (2) GWAS could detect only associations and not causality, and the biological knowledge of hits and loci in LD with them for functional analyses was obtained mainly from the Caucasian population. (3) Furthermore, gene expression is complicated and influenced by the environment. Thus, the results from our study cannot determine the causal genetic variants of GDM. (4) While SNP-based GWAS has standardized processes, there are few conventions for CNV analysis. We performed ORA rather than gene set enrichment analysis for the unavailable ranked scores for genes annotated from CNVRs. (5) No genome-wide significant variant was identified in our study, and we might have missed genetic variants specific to GDM but with a small effect with a small sample size for a GWAS. In summary, our results need verification from more GWASs and other mechanistic studies (such as epigenetic, proteomic, metabolomic, pathway, molecular, and animal model studies) of GDM in the Chinese population.
This GWAS of GDM in a sample of Chinese women showed evidence that GDM is a metabolic disease and that lipid metabolism, particularly lipolysis, may play an important role in the onset and development of GDM. The results provide a new direction for the prevention, early diagnosis, and treatment of GDM, focusing more on metabolites other than blood glucose. However, further studies, such as GWASs in other populations or studies using other omics, animal models, or GDM-associated genes from this study, should be conducted to verify and detail the actual mechanism of these hypotheses.
Availability of data and materials
The datasets analysed during the current study are not publicly available due to the Regulation of the People's Republic of China on the Administration of Human Genetic Resources, but the summary statistics of datasets are available from the corresponding author on reasonable request.
Copy number variant
Copy number variants region
Expression quantitative trait loci
False discovery rate
Free fatty acid
Gestational diabetes mellitus
Genome-wide association study
Minor allele frequency
Single nucleotide polymorphism
Standard operating procedure
Type 2 diabetes mellitus
IDF Diabetes Atlas. 10th edition. https://diabetesatlas.org/idfawp/resource-files/2021/07/IDF_Atlas_10th_Edition_2021.pdf. Accessed 6 May 2022.
IDF Diabetes Atlas. 9th edition. https://diabetesatlas.org/atlas/ninth-edition/. Accessed 6 May 2022.
Stotland NE, Caughey AB, Breed EM, Escobar GJ. Risk factors and obstetric complications associated with macrosomia. Int J Gynecol Obstet. 2005;90:88–88.
Bellamy L, Casas JP, Hingorani AD, Williams D. Type 2 diabetes mellitus after gestational diabetes: a systematic review and meta-analysis. Lancet. 2009;373:1773–9.
Moon JH, Jang HC. Gestational diabetes mellitus: diagnostic approaches and maternal-offspring complications. Diabetes Metab J. 2022;46:3–14.
Sweeting A, Wong J, Murphy HR, Ross GP. A clinical update on gestational diabetes mellitus. Endocr Rev. 2022. https://doi.org/10.1210/endrev/bnac003.
Li JW, He SY, Liu P, Luo L, Zhao L, Xiao YB. Association of gestational diabetes mellitus (GDM) with subclinical atherosclerosis: a systemic review and meta-analysis. BMC Cardiovasc Disord. 2014;14:132.
Pathirana MM, Lassi ZS, Ali A, Arstall MA, Roberts CT, Andraweera PH. Association between metabolic syndrome and gestational diabetes mellitus in women and their children: a systematic review and meta-analysis. Endocrine. 2021;71:310–20.
Tranidou A, Dagklis T, Tsakiridis I, Siargkas A, Apostolopoulou A, Mamopoulos A, Goulis DG, Chourdakis M. Risk of developing metabolic syndrome after gestational diabetes mellitus—a systematic review and meta-analysis. J Endocrinol Invest. 2021;44:1139–49.
Fu J, Retnakaran R. The life course perspective of gestational diabetes: an opportunity for the prevention of diabetes and heart disease in women. EClinicalMedicine. 2022;45:101294.
Flachs Madsen LR, Gerdoe-Kristensen S, Lauenborg J, Damm P, Kesmodel US, Lynge E. Long-term follow-up on morbidity among women With a history of gestational diabetes mellitus: a systematic review. J Clin Endocrinol Metab. 2022;107:2411–23.
Bulletins-Obstetrics C. ACOG practice bulletin no. 190: gestational diabetes mellitus. Obstet Gynecol. 2018;131(2):e49-64.
Catalano PM, Hauguel-De Mouzon S. Is it time to revisit the Pedersen hypothesis in the face of the obesity epidemic? Am J Obstet Gynecol. 2011;204:479–87.
Friedman JE. Obesity and gestational diabetes mellitus pathways for programming in mouse, monkey, and man-where do we go next? Diabetes Care. 2015;38:1402–11.
Tocantins C, Diniz MS, Grilo LF, Pereira SP. The birth of cardiac disease: mechanisms linking gestational diabetes mellitus and early onset of cardiovascular disease in offspring. WIREs Mech Dis. 2022;14:e1555.
Plows JF, Stanley JL, Baker PN, Reynolds CM, Vickers MH. The pathophysiology of gestational diabetes mellitus. Int J Mol Sci. 2018. https://doi.org/10.3390/ijms19113342.
de Mendonca E, Fragoso MBT, de Oliveira JM, Xavier JA, Goulart MOF, de Oliveira ACM. Gestational diabetes mellitus: the crosslink among inflammation, nitroxidative stress, intestinal microbiota and alternative therapies. Antioxidant. 2022. https://doi.org/10.3390/antiox11010129.
Hod M, Kapur A, Sacks DA, Hadar E, Agarwal M, Di Renzo GC, Cabero Roura L, McIntyre HD, Morris JL, Divakar H. The International Federation of Gynecology and Obstetrics (FIGO) Initiative on gestational diabetes mellitus: a pragmatic guide for diagnosis, management, and care. Int J Gynaecol Obstet. 2015;131(Suppl 3):S173-211.
Petry CJ. Gestational diabetes: risk factors and recent advances in its genetics and treatment. Br J Nutr. 2010;104:775–87.
Powe CE, Kwak SH. Genetic studies of gestational diabetes and glucose metabolism in pregnancy. Curr Diab Rep. 2020;20:69.
Abu Samra N, Jelinek HF, Alsafar H, Asghar F, Seoud M, Hussein SM, Mubarak HM, Anwar S, Memon M, Afify N, et al. Genomics and epigenomics of gestational diabetes mellitus: understanding the molecular pathways of the disease pathogenesis. Int J Mol Sci. 2022. https://doi.org/10.3390/ijms23073514.
Martin AO, Simpson JL, Ober C, Freinkel N. Frequency of diabetes mellitus in mothers of probands with gestational diabetes: possible maternal influence on the predisposition to gestational diabetes. Am J Obstet Gynecol. 1985;151:471–5.
Williams MA, Qiu C, Dempsey JC, Luthy DA. Familial aggregation of type 2 diabetes and chronic hypertension in women with gestational diabetes mellitus. J Reprod Med. 2003;48:955–62.
Liu S, Liu Y, Liao S. Heterogeneous impact of type 2 diabetes mellitus-related genetic variants on gestational glycemic traits: review and future research needs. Mol Genet Genomics. 2019;294:811–47.
Ortega-Contreras B, Armella A, Appel J, Mennickent D, Araya J, Gonzalez M, Castro E, Obregon AM, Lamperti L, Gutierrez J, Guzman-Gutierrez E. Pathophysiological role of genetic factors associated with gestational diabetes mellitus. Front Physiol. 2022;13:769924.
Kwak SH, Kim SH, Cho YM, Go MJ, Cho YS, Choi SH, Moon MK, Jung HS, Shin HD, Kang HM, et al. A genome-wide association study of gestational diabetes mellitus in Korean women. Diabetes. 2012;61:531–41.
Wu NN, Zhao D, Ma W, Lang JN, Liu SM, Fu Y, Wang X, Wang ZW, Li Q. A genome-wide association study of gestational diabetes mellitus in Chinese women. J Matern Fetal Neonatal Med. 2021;34:1557–64.
Chinese Society of Obstetrics. Chinese medical association: diagnosis and therapy guideline of pregnancy with diabetes mellitus. Zhonghua Fu Chan Ke Za Zhi. 2014;49:561–9.
Lamain-de Ruiter M, Kwee A, Naaktgeboren CA, Franx A, Moons KGM, Koster MPH. Prediction models for the risk of gestational diabetes: a systematic review. Diagn Progn Res. 2017;1:3.
Griffith RJ, Alsweiler J, Moore AE, Brown S, Middleton P, Shepherd E, Crowther CA. Interventions to prevent women from developing gestational diabetes mellitus: an overview of Cochrane Reviews. Cochrane Database Syst Rev. 2020. https://doi.org/10.1002/14651858.CD012394.pub3.
Landon MB, Spong CY, Thom E, Carpenter MW, Ramin SM, Casey B, Wapner RJ, Varner MW, Rouse DJ, Thorp JM Jr, et al. A multicenter, randomized trial of treatment for mild gestational diabetes. N Engl J Med. 2009;361:1339–48.
Casey BM, Rice MM, Landon MB, Varner MW, Reddy UM, Wapner RJ, Rouse DJ, Biggio JR Jr, Thorp JM Jr, Chien EK, et al. Effect of treatment of mild gestational diabetes on long-term maternal outcomes. Am J Perinatol. 2020;37:475–82.
Behboudi-Gandevani S, Bidhendi-Yarandi R, Panahi MH, Vaismoradi M. The effect of mild gestational diabetes mellitus treatment on adverse pregnancy outcomes: a systemic review and meta-analysis. Front Endocrinol. 2021;12:640004.
Lin PC, Chou PL, Wung SF. Geographic diversity in genotype frequencies and meta-analysis of the association between rs1801282 polymorphisms and gestational diabetes mellitus. Diabetes Res Clin Pract. 2018;143:15–23.
Lu XL, Yao XY, Liu XL, Xin Y, Zhao LL, Wang Z, Cui MM, Wu LH, Shangguan SF, Chang SY, et al. Melatonin receptor 1B gene polymorphism rs10830963 and gestational diabetes mellitus among a Chinese population–a meta-analysis of association studies. Endokrynol Pol. 2017;68:550–60.
Ao D, Wang HJ, Wang LF, Song JY, Yang HX, Wang Y. The rs2237892 polymorphism in KCNQ1 influences gestational diabetes mellitus and glucose evels: a case-control study and meta-analysis. PLoS ONE. 2015;10:e0128901.
Han X, Cui H, Chen X, Xie W, Chang Y. Association of the glucokinase gene promoter polymorphism -30G > A (rs1799884) with gestational diabetes mellitus susceptibility: a case-control study and meta-analysis. Arch Gynecol Obstet. 2015;292:291–8.
Illumina Infinium HTS Assay Reference Guide. https://support.illumina.com/content/dam/illumina-support/documents/documentation/chemistry_documentation/infinium_assays/infinium-hts/infinium-hts-assay-reference-guide-15045738-04.pdf. Accessed 2 Mar 2018.
GenomeStudio Genotyping QC SOP v.1.6. https://khp-informatics.github.io/COPILOT/GenomeStudio_genotyping_SOP.html. Accessed 15 Mar 2022.
Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, Maller J, Sklar P, de Bakker PIW, Daly MJ, Sham PC. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81:559–75.
1000 Genomes haplotypes -- Phase 3 integrated variant set release in NCBI build 37 (hg19) coordinates. https://mathgen.stats.ox.ac.uk/impute/1000GP_Phase3.html. Accessed 20 June 2022.
Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet. 2009;5:e1000529.
Wang K, Li M, Hadley D, Liu R, Glessner J, Grant SF, Hakonarson H, Bucan M. PennCNV: an integrated hidden Markov model designed for high-resolution copy number variation detection in whole-genome SNP genotyping data. Genome Res. 2007;17:1665–74.
Lin C-F, Naj AC, Wang L-S. Analyzing copy number variation using SNP array data: protocols for calling CNV and association tests. Curr Protoc Hum Genet. 2013. https://doi.org/10.1002/0471142905.hg0127s79.
Zhou J, Liu L, Lopdell TJ, Garrick DJ, Shi Y. HandyCNV: standardized summary, annotation, comparison, and visualization of copy number variant, copy number variation region, and runs of homozygosity. Front Genet. 2021;12:731355.
Kim JH, Hu HJ, Yim SH, Bae JS, Kim SY, Chung YJ. CNVRuler: a copy number variation-based case-control association analysis tool. Bioinformatics. 2012;28:1790–2.
Watanabe K, Taskesen E, van Bochoven A, Posthuma D. Functional mapping and annotation of genetic associations with FUMA. Nat Commun. 1826;2017:8.
Liao Y, Wang J, Jaehnig EJ, Shi Z, Zhang B. WebGestalt 2019: gene set analysis toolkit with revamped UIs and APIs. Nucleic Acids Res. 2019;47:W199–205.
Raudvere U, Kolberg L, Kuzmin I, Arak T, Adler P, Peterson H, Vilo J. g:Profiler: a web server for functional enrichment analysis and conversions of gene lists (2019 update). Nucleic Acids Res. 2019;47:W191–8.
Habibzadeh F. Statistical data editing in scientific articles. J Korean Med Sci. 2017;32:1072–6.
Chui TT, Lee WC. A regression-based method for estimating risks and relative risks in case-base studies. PLoS ONE. 2013;8:e83275.
Kanai M, Tanaka T, Okada Y. Empirical estimation of genome-wide significance thresholds based on the 1000 genomes project data set. J Hum Genet. 2016;61:861–6.
Weale ME. Quality control for genome-wide association studies. Methods Mol Biol. 2010;628:341–72.
Hellwege JN, Keaton JM, Giri A, Gao X, Velez Edwards DR, Edwards TL. Population stratification in genetic association studies. Curr Protoc Hum Genet. 2017. https://doi.org/10.1002/cphg.48.
Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet. 2006;38:904–9.
Zechner R, Zimmermann R, Eichmann TO, Kohlwein SD, Haemmerle G, Lass A, Madeo F. FAT SIGNALS–lipases and lipolysis in lipid metabolism and signaling. Cell Metab. 2012;15:279–91.
Kim JK, Fillmore JJ, Chen Y, Yu C, Moore IK, Pypaert M, Lutz EP, Kako Y, Velez-Carrasco W, Goldberg IJ, et al. Tissue-specific overexpression of lipoprotein lipase causes tissue-specific insulin resistance. Proc Natl Acad Sci USA. 2001;98:7522–7.
Morigny P, Houssier M, Mouisel E, Langin D. Adipocyte lipolysis and insulin resistance. Biochimie. 2016;125:259–66.
Grabner GF, Xie H, Schweiger M, Zechner R. Lipolysis: cellular mechanisms for lipid mobilization from fat stores. Nat Metab. 2021;3:1445–65.
Engin AB. What Is Lipotoxicity? In: Engin AB, Engin A, editors. Obesity and Lipotoxicity. Berlin: Springer; 2017. p. 197–220.
Korac B, Kalezic A, Pekovic-Vaughan V, Korac A, Jankovic A. Redox changes in obesity, metabolic syndrome, and diabetes. Redox Biol. 2021;42:101887.
Girousse A, Tavernier G, Valle C, Moro C, Mejhert N, Dinel A-L, Houssier M, Roussel B, Besse-Patin A, Combes M, et al. Partial inhibition of adipose tissue lipolysis improves glucose metabolism and insulin sensitivity without alteration of fat mass. PLoS Biol. 2013;11:e1001485.
Kim C, Ferrara A. 2010 Gestational Diabetes During and After Pregnancy.In: Catherine Kim, Assiamira Ferrara (eds), Springer, Berlin.
Hussein MK. Oxidoreductases: significance for humans and microorganism. In: Mahmoud Ahmed M, editor. Oxidoreductase. London: IntechOpen; 2020.
Watson JD. Type 2 diabetes as a redox disease. Lancet. 2014;383:841–3.
Colozza G, Koo BK. Wnt/beta-catenin signaling: structure, assembly and endocytosis of the signalosome. Dev Growth Differ. 2021;63:199–218.
Valenta T, Hausmann G, Basler K. The many faces and functions of β-catenin. Embo j. 2012;31:2714–36.
Manolagas SC, Almeida M. Gone with the Wnts: β-Catenin, T-Cell Factor, Forkhead Box O, and oxidative stress in age-dependent diseases of bone, lipid, and glucose metabolism. Mol Endocrinol. 2007;21:2605–14.
Wilson C. Diabetes: human beta-cell proliferation by promoting Wnt signalling. Nat Rev Endocrinol. 2013;9:502.
Jin T. The WNT signalling pathway and diabetes mellitus. Diabetologia. 2008;51:1771–80.
Macdougall CE, Wood EG, Solomou A, Scagliotti V, Taketo MM, Gaston-Massuet C, Marelli-Berg FM, Charalambous M, Longhi MP. Constitutive activation of beta-catenin in conventional dendritic cells increases the insulin reserve to ameliorate the development of type 2 diabetes in mice. Diabetes. 2019;68:1473–84.
Welters HJ, Kulkarni RN. Wnt signaling: relevance to beta-cell biology and diabetes. Trends Endocrinol Metab. 2008;19:349–55.
Marcheva B, Ramsey KM, Buhr ED, Kobayashi Y, Su H, Ko CH, Ivanova G, Omura C, Mo S, Vitaterna MH, et al. Disruption of the clock components CLOCK and BMAL1 leads to hypoinsulinaemia and diabetes. Nature. 2010;466:627–31.
Amin M, Syed S, Wu R, Postolache TT, Gragnoli C. Familial linkage and association of the NR3C1 gene with Type 2 diabetes and depression comorbidity. Int J Mol Sci. 2022;23:11951.
Zhang H, Guan Q, Wang R, Yang S, Yu X, Cui D, Su Z. Novel association of SNP rs2297828 in PRDM16 gene with predisposition to type 2 diabetes. Gene. 2023;849:146916.
Butte NF. Carbohydrate and lipid metabolism in pregnancy: normal compared with gestational diabetes mellitus. Am J Clin Nutr. 2000;71:1256S-1261S.
Burgess S, Labrecque JA. Mendelian randomization with a binary exposure variable: interpretation and presentation of causal estimates. Eur J Epidemiol. 2018;33:947–52.
Palmer TM, Thompson JR, Tobin MD, Sheehan NA, Burton PR. Adjusting for bias and unmeasured confounding in Mendelian randomization studies with binary responses. Int J Epidemiol. 2008;37:1161–8.
Possik E, Al-Mass A, Peyot ML, Ahmad R, Al-Mulla F, Madiraju SRM, Prentki M. New mammalian glycerol-3-phosphate phosphatase: role in beta-cell. Front Endocrinol (Lausanne). 2021;12:706607.
We thank all colleagues who gave us kind help at any part of this work. We thank all subjects for their kind cooperation in the study. We also appreciate the server support from the Department of Environmental Health, School of Public Health, Fudan University.
This study was funded by the National Natural Science Foundation of China [Grant No.82173582, No.81773490, No.81373065], and the China National Key R&D Program during the 14th Five-year Plan Period [Grant No. 2021YFC2701800].
Ethics approval and consent to participate
This study was approved by the Medical Ethics Committee of Fudan University, School of Public Health (IRB00002408 & FWA00002399). Written informed consent was obtained from each participant at enrollment.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Additional file 1.
Note S1. Parameters for FUMAGWAS analyses. Note S2. Parameters for WebGestalt analyses. Note S3. Parameters for g:profiler analyses. Table S1. Characteristics of selected or not sub-cohort groups in cohort. Table S2. PCs with P value < 0.01 for SNPs/InDels association analyses. Table S3. Markers in linkage disequilibrium with any independent significant markers (r2 ≥ 0.6). Table S4. Genes mapped by positional mapping, eQTL mapping, and chromatin interaction mapping. Table S5. Protein coding genes from MAGMA gene-based analyses with PBon < 0.05. Table S6. CNVRs associated with GDM FDR with p value < 0.05. Table S7. Candidate genes for gene over-representation analyses. Table S8. Gene sets enriched in the Gene Ontology resource by g: Profiler. Table S9. Genome-wide association studies of GDM reported in literatures and the present study. Figure S1. Profile of the study cohort. Figure S2. Strategies of the analyses. Figure S3. The quantile–quantile plot of P values of SNPs/InDels GWAS (λ=1.01).
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Zhang, M., Li, Q., Wang, KL. et al. Lipolysis and gestational diabetes mellitus onset: a case-cohort genome-wide association study in Chinese. J Transl Med 21, 47 (2023). https://doi.org/10.1186/s12967-023-03902-4
- Gestational diabetes mellitus
- Genome-wide association study