First-trimester DNA Methylome profiling identifies novel predictors for gestational diabetes mellitus in Indian women

First-trimester DNA Methylome profiling identifies novel predictors for gestational diabetes mellitus in Indian women | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article First-trimester DNA Methylome profiling identifies novel predictors for gestational diabetes mellitus in Indian women Kuppan Gokulakrishnan, Chinnasamy Thirumoorthy, Kuldeep Sharma, and 15 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5280336/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Gestational diabetes mellitus (GDM) poses significant health risks for both mothers and offspring, necessitating early detection strategies. We aimed to identify specific DNA methylation signatures in early pregnancy to enhance GDM prediction. We quantified the Epigenome-wide DNA methylation profiles in 258 individuals with normal fasting glucose at early pregnancy (< 16 weeks), of whom 43% (n=111) developed GDM. We identified a panel of five CpGs (cg10139015, cg23507676, cg11001216, cg04539775, cg04985016) that predicted GDM with an area under the curve (AUC) of 0.82 (sensitivity: 82% and specificity: 77%). Combining these CpGs with maternal risk factors (age, waist-height ratio, family history of diabetes, HbA1c, and blood pressure) achieved the highest predictive ability (AUC = 0.86). Our findings suggest that a panel of 5 CpGs exhibits robust predictive value, highlighting the feasibility of creating a CpG panel to predict GDM as early as the first trimester. This has the potential to enhance diagnostic and preventive strategies for GDM. Endocrinology & Metabolism Epigenetics & Genomics DNA Methylation biomarkers gestational diabetes mellitus pregnancy longitudinal study first trimester Asian Indians Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Gestational diabetes mellitus (GDM) is a prevalent metabolic disorder characterized by glucose intolerance with onset or first recognition during pregnancy. It is a significant public health concern globally [1], affecting up to 20% of pregnancies in some populations. GDM not only increases the risk of adverse outcomes during pregnancy but also has long-term health implications for both mother and child [2]. Women with GDM are at higher risk of developing type 2 diabetes mellitus (T2DM) [3], cardiovascular disease [4, 5], and obesity [6] later in life, while their offspring are more likely to face an elevated risk of macrosomia and other cardiometabolic disorders later in life [7, 8]. GDM is typically diagnosed in the late stages of pregnancy (24-28 weeks) [9, 10], but excess fetal adiposity may already be evident [11, 12]. Yet predictive markers for GDM remain an unmet demand in clinical practice. Recent efforts have focused on developing prediction models [13] using conventional clinical and blood-based markers such as maternal age, body mass index (BMI) [14], family history of diabetes, HbA1c, cardiometabolic [15], metabolomic [16, 17], and small non-coding circulating RNA [18] biomarkers for early prediction of GDM. However, the predictive accuracy of these models is, at best, moderate, particularly when used in the early stages of pregnancy. Our recent prospective pregnancy cohort of 2115 pregnant women from the STratification of Risk of Diabetes in Early pregnancy (STRiDE) study used composite risk score models using HbA1c showed better predictability than simple clinical risk factors [19]. However, the predictive ability of a ‘single-threshold’ was still moderate and a 2-threshold approach can reduce the burden of oral glucose tolerance test (OGTT) by at least 50% [20]. Hence, if a composite risk score using genomic or epigenetic markers could be developed, this could result in precision diagnosis and management of GDM. Recent advances in epigenetics, specifically DNA methylation, offer promising avenues for the early identification of GDM [21, 22]. DNA methylation is a form of epigenetic modification that can influence gene expression and has been implicated in various metabolic and inflammatory processes associated with GDM [23]. Recent Epigenome-Wide Association Studies (EWAS) have identified specific DNA methylation patterns associated with GDM [24]. While these findings are promising, the studies were conducted in late pregnancy, limited to candidate genes and remain inconclusive [22, 24]. In addition, none have shown any potential for prediction of GDM. Integrating the analysis of DNA methylation biomarkers with maternal risk factors is crucial for developing effective GDM prediction models. There is a lack of studies that have explored the combined evaluation of DNA methylation biomarkers alongside maternal risk factors, highlighting a significant gap. Our present work is an attempt to fill that gap in knowledge. We hypothesized that DNA methylation changes identified through EWAS in early pregnancy, combined with common clinical risk factors and HbA1c levels, could accurately predict GDM diagnosed via OGTT at 24-28 weeks. A nested cohort study sample of 258 women with fasting glucose levels below 5.1mmol/l from the prospective STRiDE study was selected for the EWAS analysis. Methods Study participants: The STRiDE study population consisted of 2115 Indian pregnant women from the STRiDE study, a prospective, longitudinal cohort, that recruited at the first antenatal visit but before 16 weeks of gestation. The study was conducted between 2016 and 2019 across seven clinical sites in three cities of South India, and the details of the cohort were published elsewhere [19]. For the present nested cohort study within the original STRiDE cohort, a total of 258 pregnant women (who had fasting glucose levels of <5.1 mmol/L) in early pregnancy were randomly selected. Participants with similar age and BMI were selected to minimise the influence of these common risk factors. At 24–28 weeks (OGTT visit) of gestation, a 75g OGTT was conducted, and women were classified into Normal Glucose Tolerance [NGT] (n = 147) and GDM (n = 111). The International Association of Diabetes and Pregnancy Study Groups (IADPSG) criteria were used to diagnose GDM [25]. DNA methylation profiling was carried out in all enrolled study participants in this nested cohort at early pregnancy, and at OGTT VISIT (NGT n=96, GDM n=93). All the samples (cases/control) were randomised before the experiment. The schema of the study design is shown in Supplementary Figure 1 . Written informed consent was obtained from all participants. The study was approved by the Institutional Ethics Committee of the National Institute of Mental Health and Neuro Sciences (NIMHANS), Bengaluru and the Madras Diabetes Research Foundation (MDRF), Chennai, India. DNA isolation: Genomic DNA was isolated from whole blood samples obtained from all the enrolled participants, employing a phenol-chloroform extraction method [26]. Subsequently, the quantification of DNA samples was performed using the Qubit® 2.0 Fluorometer with the Qubit dsDNA HS Assay Kit, along with NanoDrop One C (ThermoFisher Scientific, USA). Bisulfite conversion, DNA methylation profiling & Bioinformatics: Bisulfite conversion was performed using the EZ DNA Methylation Kit from Zymo Research (USA) and subsequently analyzed with the HumanMethylationEPIC 850k BeadChip assays (Illumina, San Diego, CA, USA), following the manufacturer's instructions on an iScan reader (Illumina). To minimize potential batch effects, both case and control samples were processed and run concurrently on the same BeadChip. CpGs with poor signal quality (detection p-value > 0.05), those located on the X and Y chromosomes, common SNPs, and cross-reactive probes were excluded from the analysis [27, 28]. This resulted in a final dataset comprising 258 samples and 597,856 CpGs for EWAS. Prior to conducting the EWAS, data normalization was performed using the quantile normalization method. The Champ.DMP function from the ChAMP (Chip Analysis Methylation Pipeline for Illumina HumanMethylation450 and EPIC) package was utilized to execute the EWAS, employing a linear regression model with CpGs as dependent variables and the grouping variable for GDM status (present/absent) [29]. The output included key parameters such as fold change, p-value, and Benjamin-Hochberg adjusted p-value as p fdr -value < 0.000001 to identify differentially methylated CpGs, resulting in a total of 808 CpGs. Ultimately, we selected 527 CpG sites that were annotated to genes for further data analysis. We used Random Forest (RF), Least Absolute Shrinkage and Selection Operator (LASSO), and Recursive Feature Elimination (RFE) for feature selection. Two classifiers - RF and Radial Kernel Support Vector Machines (SVM) - were applied to construct biomarker panels using the CpG lists derived from the above feature selection methods as input variables. The panel construction employed a forward stepwise strategy aimed at optimizing the AUC, incrementally adding significant CpGs one by one to enhance model performance. The overall framework for developing these biomarker panels is illustrated in Supplementary Table 1 . The analysis was conducted across six distinct scenarios, combining the three feature selection techniques with the two classifiers. For each scenario, ten iterations were conducted by randomizing the samples in the training and testing datasets, ensuring reproducibility through fixed seeds. Panels of varying sizes were constructed, starting with a single CpG and expanding to ten CpGs per panel, culminating in ten unique sets of panels for each scenario, all optimized for AUC in each iteration. The stability of each biomarker panel was evaluated through bootstrap validation, comprising 1000 runs. This validation process generated aggregated metrics, including AUC, sensitivity, specificity, and accuracy, along with their corresponding standard deviations (SD) from the 1000 iterations. The panel exhibiting the lowest SD was identified as the most robust model for each scenario, as a lower SD indicates enhanced model stability. Ultimately, this systematic approach resulted in a compilation of the best biomarker panel for each scenario. The final panel was selected based on achieving the highest AUC among the six panels identified. All analyses were executed using R (version 4.3.2), employing the "train" function from the caret package to implement both the RF and SVM classifiers. For the RF model, parameters such as "ntree" (number of trees) and "mtry" (number of randomly selected CpGs at each tree) were manually optimized, whereas the SVM model parameters, C (the margin between the classifier line and support vectors) and sigma (degree of non-linearity), were automatically tuned through 10-fold cross-validation with five repeats. Recent studies have shown a strong correlation between cell types in peripheral blood during inflammatory processes, particularly in pregnancy [30, 31]. Therefore, we did not perform Houseman cell-type adjustment in our analysis. This decision helps to avoid violations of the no multicollinearity assumption in EWAS, which can lead to reversed associations, especially when investigating inflammation-related phenotypes such as those observed in pregnancy. Additionally, it minimizes the risk of obtaining spurious significant results [31, 32]. Functional enrichment GO and KEGG analysis: Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis for the 527 combined differentially methylated positions (DMPs) and their annotated genes were performed using the SRplot online tool [33, 34]. For both the GO and KEGG analyses, the top 10 pathways were selected for further examination. A significant threshold of a false discovery rate (FDR) of less than 0.05 was established to identify biologically relevant pathways and processes associated with the DMRs. Statistical Analysis: Numeric variables were summarized with mean ± standard deviation (SD) for continuous variables and n (%) for categorical variables. The Kolmogorov–Smirnov test confirmed normality, allowing for parametric tests due to the normally distributed nature of the data. Continuous variable averages were compared using the independent t-test, while proportions were assessed using the Chi-square test or Fisher’s exact test as appropriate. Spearman correlation analysis was performed to explore the association between the best panel of CpGs at early pregnancy and OGTT glucose levels at OGTT visit. After obtaining the best panel of CpGs building step-by-ML approaches, further logistic regression models were fitted to have a better understanding of these CpGs and other maternal risk factors related to GDM. Each logistic regression model’s predictive performance was assessed by plotting the receiver operating characteristic curve (ROC), from which AUC with 95% CI was calculated as the primary predictive measure. Statistical analyses were conducted using R (Version 4.3.2), with p-values < 0.05 deemed statistically significant. Model developments: - a. Conventional risk predictors considering clinical accessibility: Overall, we considered well-documented clinically accessible maternal risk factors and blood-based biomarkers associated with the development of GDM, which are available in the original full cohort of 2115 pregnant women and included the same in this nested GDM case-control subset as potential conventional risk predictors (age, waist-height ratio, family history (FH) of diabetes, point of care (POC) HbA1c and blood pressure). These predictors are routinely assessed at relatively affordable clinical costs. First, we built logistic regression models using the maternal risk factors at early pregnancy, with the outcome being the presence or absence of GDM based on OGTT results at OGTT visit. Model 1a fitted with how age, BMI, FH of diabetes and venous HbA1c predicted GDM. Model 1b tested the association of age, BMI, FH diabetes, and POC HbA1c instead of venous HbA1c adjusted for each other. As both venous and POC HbA1c performed equally well and POC HbA1c can be used in rural settings without the need for a lab for HbA1c measurement, we retained the POC HbA1c in the final model. Additional analyses using waist-height ratio were presented instead of BMI to assess the potential role of waist circumference (Model 1c & 1d) ( Supplementary Figure 2 ). Further, we considered other variables significantly associated with GDM in our dataset (age, waist-height ratio, FH of diabetes, POC HbA1c and blood pressure) as conventional risk predictors. b. Candidate biomarkers predictors: These include EWAS-derived novel DNA methylation biomarkers, which are not yet available for clinical use and require specialized laboratories for measurement. c. Full model predicting GDM risk in the clinical context : To develop a clinically efficient comprehensive GDM prediction model, we integrated conventional risk predictors with novel epigenetic candidate biomarkers predictors. Results Early pregnancy and OGTT characteristics of the whole STRiDE cohort and the nested case-control study participants are shown in Table 1 . The mean gestational age in early pregnancy was around 10 weeks. Women who developed GDM at 24-28 weeks had marginally higher systolic blood pressure and glycated haemoglobin levels at early pregnancy. The composite risk score model of the whole STRiDE study cohort (n=2115), combining age, BMI, FH of diabetes and POC HbA1c had an AUC of 0.61 for predicting GDM. The nested EWAS cohort (n=258) using the same composite risk score had a very similar AUC of 0.63 (Figure 1A) suggesting that this is representative of the larger whole STRiDE study cohort of 2,115 individuals. As waist-height ratio may be a better metabolic predictor than BMI in South Asians [35], we replaced BMI with waist-height ratio and added systolic BP to assess the predictive ability of GDM. This prediction model, combining age, waist-height ratio, FH of diabetes, POC HbA1c, and blood pressure, had an AUC of 0.71 for GDM (Figure 1B). All these risk factors are either routinely measured or can be done at relatively affordable clinical costs. Identification of DNA Methylation signatures: After initial processing, our dataset included 597,856 DNA methylation probes from each study participant. We performed a discovery EWAS, applying a significance threshold of p fdr -value <0.000001 to identify CpG sites significantly associated with GDM. This analysis identified 527 DMPs related to GDM (Supplementary Table 2). Developing DNA Methylation signature panel as candidate biomarkers and its performance metrics for early GDM prediction: To leverage the strengths of each method and ensure a consistent selection of the most informative CpGs for GDM prediction out of the 527 available at early pregnancy, we applied three machine learning techniques: RF, RFE, and LASSO. The top 40 CpGs were chosen based on their feature-important scores using RF, which utilises a feature-importance ranking algorithm. Then, we used RFE, which systematically removes the least significant features, and found out that the model performed best with 40 CpGs. Lastly, we selected 21 CpGs as the most predictive features using LASSO, which penalises less significant features by shrinking their coefficients towards zero. All three feature selection methods were applied to the same dataset. Using these selected CpGs from each approach, we developed prediction models for GDM utilizing the classifiers – RF and SVM methodologies, incorporating bootstrap validation with 1000 repetitions. The RF method demonstrated the best performance, identifying a panel of five CpG sites at early pregnancy (cg10139015, cg23507676, cg11001216, cg04539775, cg04985016) for GDM prediction. In the test dataset, this panel achieved an AUC of 0.98, with a sensitivity of 0.98 and a specificity of 0.83. During bootstrap validation, the panel maintained robust performance, yielding an AUC of 0.89, a sensitivity of 0.83, and a specificity of 0.80. Supplementary Table 3 summarises the performance metrics for the various feature selection methods in test and validation datasets. The identified panel of five CpG sites, along with their annotated genes and the chromosome-wide distribution of 527 DMPs, is illustrated in the Manhattan plot ( Figure 2A ). Figure 2B presents the hierarchical clustering of the five CpG sites at early pregnancy showing the differential expression patterns among GDM and NGT separately. Detailed information regarding these five CpGs, including their annotated genes and corresponding chromosomes are summarized in Supplementary Table 4. Figure 3A-E shows that women who were diagnosed to have GDM at OGTT, had higher methylation levels of all these five CpGs compared to women who were NGT in early pregnancy. This higher methylation pattern of 5 CpGs were similar when measured at the time of OGTT ( Supplementary Figure 3 A-E ). Prediction of GDM combining clinical, POC HbA1c and DNA methylation signatures: We conducted an ROC analysis to illustrate the performance of the candidate biomarker panel using the 5 CpG sites in early pregnancy to predict GDM at OGTT. The 5 CpG marker panel demonstrated a robust AUC of 0.82 (95% CI: 0.77–0.87), with a sensitivity of 82% and a specificity of 77% (Figure 4A). Combining these novel DNA methylation signatures with conventional clinical risk predictors and POC HbA1c has increased the AUC of the predicted GDM risk to 0.86 (95% CI: 0.81–0.91), suggesting that the full GDM prediction model at early pregnancy improved the discrimination (Figure 4B). Association of novel candidate biomarkers with glycemic variables at OGTT: To assess whether there are any differential associations between the identified candidate biomarker panel of 5 CpGs and glucose levels at OGTT, we tested the individual CpGs with fasting, 60 min and 120 min glucose levels at the time of OGTT. Four CpGs (cg10139015, cg23507676, cg11001216, cg04985016) were positively correlated with fasting glucose levels (p-value<0.05) ( Supplementary Figure 4) . All the 5 CpGs showed a positive correlation with 60-mins (p-value <0.01) ( Supplementary Figure 5) and 120-mins glucose levels (p-value <0.01) ( Supplementary Figure 6) . Interestingly, among the 5 CpGs, cg23507676 showed a higher magnitude of positive relationship with fasting, 60- and 120-minute glucose levels during the OGTT. Functional analysis GO and KEGG: To explore the potential mechanisms involved in GDM, we conducted GO and KEGG analyses on the annotated genes of 527 DMPs. The DMP classification included Biological Processes (BP), Cellular Components (CC), and Molecular Functions (MF). The top ten GO enrichment results indicated a significant predominance in BP, highlighting processes such as nucleotide catabolism, coenzyme A biosynthesis, Ras protein signal transduction, and small GTPase-mediated signaling. In the CC analysis, notable enrichments were observed in complexes such as the GABA-A receptor, early endosome, guanyl nucleotide exchange factor, and chloride channel complexes. The MF analysis revealed enrichments in activities including molecular adaptor, GABA-A receptor, guanyl-nucleotide exchange factor, ion channel, and Ras GTPase binding (Figure 5A) . The most significant enriched pathways, including morphine addiction, endocytosis, TH17 cell differentiation and phosphatidylinositol signaling pathways, were identified through KEGG analysis for the 527 DMPs (Figure 5B). Discussion In this EWAS study which investigated 850,000 CpGs from peripheral blood samples of a prospective early pregnancy cohort, we show the following novel findings: 527 CpGs were differentially methylated and associated with the development of GDM. We identified a panel of 5 CpGs using modern machine learning techniques as candidate biomarkers with the best discriminatory power for GDM prediction. We developed a clinically efficient comprehensive early GDM prediction model comprising the clinically accessible maternal risk factors, POC HbA1c and novel DNA methylation candidate biomarkers achieving the highest predictive ability for GDM with an AUC of 0.86. Pathway analyses indicated potential involvement of pathways such as endocytosis, phosphatidylinositol signaling, Th17, Th1 and Th2 cell differentiation in GDM. Findings from this large-scale epigenomic study in an Indian pregnancy cohort highlight the potential of early pregnancy novel epigenetic biomarkers and conventional maternal risk factors in predicting GDM, and to our knowledge, this is the first comprehensive study to report this in GDM. Studies on GDM prediction models are advancing, with significant improvements seen by including early trimester conventional risk factors, blood-based inflammatory markers, metabolites related to glucose, lipid, and amino acid metabolism, genetic markers, miRNA signatures, extrachromosomal circular DNA, and cell-free DNAs [17, 36–40]. In our recent STRiDE study, we developed composite risk score models that combined HbA1c with factors such as age, BMI, and FH of diabetes during early pregnancy, showing promising results for GDM prediction [19]. We improve the predictive ability of our approach by the addition of 5 CpGs in this study in a representative STRiDE-India cohort. Unlike many previous studies, we focused on developing prediction models based on clinically accessible maternal risk factors, which are crucial for practical implementation, and extended our analysis by incorporating novel DNA methylation biomarkers identified through EWAS. Our panel consisting of five CpGs, along with these easily accessible clinical risk factors and POC HbA1c demonstrated strong predictive ability for GDM around 10 weeks of gestation with robust sensitivity and specificity. Our GDM prediction model is more accurate than previous ones [21, 22, 24]. These findings are particularly important, as early lifestyle interventions based on accurate predictions could help in reducing the burden of GDM and its complications in Indian women. However, this will require future randomized controlled trials (RCTs) based on our composite risk score model including these CpGs. The top 5 CpG sites identified in our study were annotated to the BLK, A1BG, TRIP12, CLSTN3, and PSMC3 genes, all of which were implicated in various metabolic disorders. Among the 5 CpGs identified in the present study, the CpG cg23507676 annotated to the BLK had the highest correlation with OGTT fasting, 60- and 120-min glucose levels. This CpG annotated to the BLK gene, has been shown to play a vital role in insulin synthesis and secretion in pancreatic beta-cells by upregulating transcription factors like Pdx1 and Nkx6.1 [41]. Mutations, such as p.Ala71Thr compromise this function, reduce insulin output and were linked to BLK MODY and T2DM, although not all studies support BLK as a cause for MODY [42]. Studies also have reported the association of this variant with a higher risk for T2DM and obesity [43]. The CpG cg11001216 annotated to the A1BG protein levels, were shown to be downregulated in exosomes among individuals with obese T2DM and reduced in the inflamed livers of obese mice [44, 45]. A1BG has also been linked to diabetic kidney disease [46, 47], and higher levels were noted in PCOS individuals [48], indicating a complex relationship with metabolic disorders. TRIP12 (cg04985016) plays a vital role in pancreatic cell homeostasis by stabilizing transcription factors like PTF1A, essential for insulin secretion [49]. The CLSTN3 (cg04539775), which encodes calsyntenin-3, is associated with obesity and dysfunction in white adipose tissue (WAT) [50]. The variant rs7296261 in CLSTN3 has been linked to an increased risk of obesity, highlighting its role in metabolic dysregulation [51]. Studies were reported that altered expression of PSMC3 in T2DM impairs pancreatic β-cell function, affecting insulin production and disease progression [52]. While these genes are linked to metabolic disorders, further studies are needed to explore their pathophysiological roles as GDM and T2DM have shared common etiological pathways [53]. Our findings provide proof of concept and would be useful for future mechanistic studies examining the potential role of these DNA Methylation signatures in the etiology of GDM in other ethnic groups. Our KEGG pathway analyses of 527 CpG-annotated genes revealed key pathways related to NK cell-mediated cytotoxicity, phosphatidylinositol signaling, Th17 cell differentiation, and Th1/Th2 differentiation. While NK cells play a critical role in immunity by eliminating pathogens; however, in diabetes, their function declines, especially under hyperglycemia, leading to reduced numbers and activity of NK cells [54]. Studies have reported that in obesity, high-fat diets activate NK cells through increased NCR1 ligands on adipocytes, promoting inflammation and insulin resistance [55, 56]. Studies have demonstrated that Th1/Th2 differentiation plays an important role in shaping the immune response in GDM. Women with GDM show lower Th1 levels and higher Th2, Tregs, and Th17 cells, indicating an immune adaptation that balances inflammation and supports fetal development [57, 58]. These findings highlight the interplay between immune regulation and metabolic disorders. These cellular pathway alterations implicated in GDM, diabetes, insulin resistance, and inflammation were driven by shared underlying biological mechanisms. Our findings suggest that the altered methylation patterns identified in the present study are related to these cellular pathways and it may have a mechanistic role in the development of GDM. However, additional confirmatory and mechanistic studies are warranted to establish their biological relevance. Our study's strength lies in the comprehensive longitudinal assessment of novel DNA methylation signatures through the EWAS approach along with a combined analysis of clinically accessible maternal risk factors during early pregnancy and in the OGTT visit within a large prospective early pregnancy cohort in India. This approach offered us to perform a comprehensive analysis of GDM prediction and how they are impacted by GDM. Our study is the first to report these findings in GDM. It is also of interest that the panel of DNA methylome biomarkers presented in our study could be tested in interventional studies to demonstrate the molecular benefits of lifestyle modification for GDM. In clinical practice, various methods are commonly used to analyze DNA methylation, including targeted methylation sequencing, DNA methylation arrays, methylation-specific PCR (MSP), pyrosequencing, and methylation-sensitive restriction enzyme (MSRE) assays as predictive, diagnostic, and prognostic tools [59, 60]. In this context, our study paves the way for developing in vitro diagnostic kits (IVDs) utilizing the five CpGs for GDM prediction and we propose to try to develop a point of care (POC) kit incorporating these 5 CpGs. However, our study also has some limitations. Firstly, due to the study design, the associations found cannot imply causality. Secondly, there is a need for additional studies across ethnically diverse cohorts to validate the clinical usefulness of our discriminating and predictive models in other pregnancy cohorts. Thus, though we used sophisticated ML approaches like RF, SVM and bootstrap validations, validating these prediction models in external cohorts is needed to confirm our findings. We propose to test these in other cohorts which we have already approached for this purpose. Finally, the sample size for this study was modest, although admittedly, most GDM studies tend to be small. Additionally, the functional implications of the identified methylation changes need to be experimentally validated to establish their biological relevance. Given the limited studies on DNA methylation in GDM prediction, our findings, despite these limitations, are significant in providing insights into the design of large-scale longitudinal studies for population-based screening for GDM prediction. In conclusion, we have identified a panel of DNA methylation signatures consisting of 5 CpGs, which has a robust predictive value for early detection of GDM especially when integrated with clinically accessible maternal risk factors in a large longitudinal pregnancy cohort of 2,115 pregnant women. These findings provide proof of concept for developing a CpG panel for GDM prediction and may form the basis for future diagnostic and preventive strategies for GDM. Declarations Acknowledgements/Funding Dr. Gokulakrishnan Kuppan is a current recipient of a DBT-Wellcome Trust India Alliance Intermediate Clinical & Public health Fellowship (Grant Number IA/CPHI/18/1/503964) and acknowledges the funding received from the DBT-Wellcome Trust India Alliance for this study. STRiDE study was funded by MRC-DBT Newton fund (MRC Newton Fund MR/N006232/1). P. Popova was financially supported by the Ministry of Science and Higher Education of the Russian Federation (Agreement No. 075-15-2022-301). Dr M. Balasubramanyam acknowledges the funding from ICMR and DHR. The authors would like to thank all the study participants and staff of STRiDE study. Conflict of Interest The authors declare no conflict of interests. Author’s contributions KG has conceptualized, written, and taken the lead in completing manuscripts through all stages of preparation and submission. KG, PS, MB, RMA, NT, HG, UR, VR, PP and VM have reviewed the project from the conceptualization stage and have contributed to each version of the manuscript. MD, RMA, and UR are members of the research team who participated in the conduct of the study. CT performed experiments. KS, YW, HA, VU, BB, and SS were involved in data analysis. PS was the principal investigator for the STRiDE study and one of the mentors for KG’s fellowship. All authors have contributed to the article critically for intellectual content and have provided final approval of the version to be published. Financial Interest All authors have no relevant financial or non-financial interests to disclose. Data availability statement The data that support the findings of this study are available from the corresponding author, [KG], upon reasonable request. References Buchanan TA, Xiang AH, Page KA (2012) Gestational Diabetes Mellitus: Risks and Management during and after Pregnancy. Nat Rev Endocrinol 8(11):639–649. https://doi.org/10.1038/nrendo.2012.96 Saravanan P, Diabetes in Pregnancy Working Group, Maternal Medicine Clinical Study Group, Royal College of Obstetricians and Gynaecologists, UK (2020) Gestational diabetes: opportunities for improving maternal and child health. Lancet Diabetes Endocrinol 8(9):793–800. https://doi.org/10.1016/S2213-8587(20)30161-3 Diaz-Santana MV, O’Brien KM, Park Y-MM, Sandler DP, Weinberg CR (2022) Persistence of Risk for Type 2 Diabetes After Gestational Diabetes Mellitus. Diabetes Care 45(4):864–870. https://doi.org/10.2337/dc21-1430 Yu Y, Soohoo M, Sørensen HT, Li J, Arah OA (2022) Gestational Diabetes Mellitus and the Risks of Overall and Type-Specific Cardiovascular Diseases: A Population- and Sibling-Matched Cohort Study. Diabetes Care 45(1):151–159. https://doi.org/10.2337/dc21-1018 Kramer CK, Campbell S, Retnakaran R (2019) Gestational diabetes and the risk of cardiovascular disease in women: a systematic review and meta-analysis. Diabetologia 62(6):905–914. https://doi.org/10.1007/s00125-019-4840-2 Chu SY, Callaghan WM, Kim SY, et al (2007) Maternal Obesity and Risk of Gestational Diabetes Mellitus. Diabetes Care 30(8):2070–2076. https://doi.org/10.2337/dc06-2559a Xiong X, Saunders LD, Wang FL, Demianczuk NN (2001) Gestational diabetes mellitus: prevalence, risk factors, maternal and infant outcomes. Int J Gynaecol Obstet Off Organ Int Fed Gynaecol Obstet 75(3):221–228. https://doi.org/10.1016/s0020-7292(01)00496-9 Lowe WL, Scholtens DM, Lowe LP, et al (2018) Association of Gestational Diabetes With Maternal Disorders of Glucose Metabolism and Childhood Adiposity. JAMA 320(10):1005–1016. https://doi.org/10.1001/jama.2018.11628 (2013) Diagnostic Criteria and Classification of Hyperglycaemia First Detected in Pregnancy. World Health Organization, Geneva American Diabetes Association Professional Practice Committee (2024) 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes-2024. Diabetes Care 47(Suppl 1):S20–S42. https://doi.org/10.2337/dc24-S002 Venkataraman H, Ram U, Craik S, Arungunasekaran A, Seshadri S, Saravanan P (2017) Increased fetal adiposity prior to diagnosis of gestational diabetes in South Asians: more evidence for the “thin-fat” baby. Diabetologia 60(3):399–405. https://doi.org/10.1007/s00125-016-4166-2 U S, Hr M, Gc S (2016) Accelerated Fetal Growth Prior to Diagnosis of Gestational Diabetes Mellitus: A Prospective Cohort Study of Nulliparous Women. Diabetes Care 39(6). https://doi.org/10.2337/dc16-0160 Mennickent D, Rodríguez A, Farías-Jofré M, Araya J, Guzmán-Gutiérrez E (2022) Machine learning-based models for gestational diabetes mellitus prediction before 24-28 weeks of pregnancy: A review. Artif Intell Med 132:102378. https://doi.org/10.1016/j.artmed.2022.102378 Popova PV, Grineva EN, Gerasimov AS, Kravchuk EN, Ryazantseva EM, Shelepova ES (2015) The new combination of risk factors determining a high risk of gestational diabetes mellitus. Minerva Endocrinol 40(4):239–247 Parkhi D, Sampathkumar S, Weldeselassie Y, Sukumar N, Saravanan P (2023) Systematic Review of risk score prediction models using maternal characteristics with and without biomarkers for the prediction of GDM. 2023.10.23.23297401 Yu J, Ren J, Ren Y, et al (2024) Using metabolomics and proteomics to identify the potential urine biomarkers for prediction and diagnosis of gestational diabetes. eBioMedicine 101. https://doi.org/10.1016/j.ebiom.2024.105008 Razo-Azamar M, Nambo-Venegas R, Meraz-Cruz N, et al (2023) An early prediction model for gestational diabetes mellitus based on metabolomic biomarkers. Diabetol Metab Syndr 15:116. https://doi.org/10.1186/s13098-023-01098-7 Dinesen S, El-Faitarouni A, Frisk NLS, Sørensen AE, Dalgaard LT (2023) Circulating microRNA as Biomarkers for Gestational Diabetes Mellitus—A Systematic Review and Meta-Analysis. Int J Mol Sci 24(7):6186. https://doi.org/10.3390/ijms24076186 Saravanan P, Deepa M, Ahmed Z, et al (2024) Early pregnancy HbA1c as the first screening test for gestational diabetes: results from three prospective cohorts. Lancet Diabetes Endocrinol 0(0). https://doi.org/10.1016/S2213-8587(24)00151-7 Benhalima K, Van Crombrugge P, Moyson C, et al (2018) A Modified Two-Step Screening Strategy for Gestational Diabetes Mellitus Based on the 2013 WHO Criteria by Combining the Glucose Challenge Test and Clinical Risk Factors. J Clin Med 7(10):351. https://doi.org/10.3390/jcm7100351 Li N, Liu H-Y, Liu S-M (2024) Deciphering DNA Methylation in Gestational Diabetes Mellitus: Epigenetic Regulation and Potential Clinical Applications. Int J Mol Sci 25(17):9361. https://doi.org/10.3390/ijms25179361 Linares-Pineda T, Peña-Montero N, Fragoso-Bargas N, et al (2023) Epigenetic marks associated with gestational diabetes mellitus across two time points during pregnancy. Clin Epigenetics 15(1):110. https://doi.org/10.1186/s13148-023-01523-8 Dłuski DF, Wolińska E, Skrzypczak M (2021) Epigenetic Changes in Gestational Diabetes Mellitus. Int J Mol Sci 22(14):7649. https://doi.org/10.3390/ijms22147649 Wang X, Huang J, Zheng Y, et al (2021) Study on the relationship between DNA methylation of target CpG sites in peripheral blood and gestational diabetes during early pregnancy. Sci Rep 11(1):20455. https://doi.org/10.1038/s41598-021-99836-2 International Association of Diabetes and Pregnancy Study Groups Consensus Panel, Metzger BE, Gabbe SG, et al (2010) International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care 33(3):676–682. https://doi.org/10.2337/dc09-1848 Thangaraj K, Joshi MB, Reddy AG, Gupta NJ, Chakravarty B, Singh L (2002) CAG repeat expansion in the androgen receptor gene is not associated with male infertility in Indian populations. J Androl 23(6):815–818 Pidsley R, Zotenko E, Peters TJ, et al (2016) Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling. Genome Biol 17(1):208. https://doi.org/10.1186/s13059-016-1066-1 Wu MC, Kuan P-F (2018) A Guide to Illumina BeadChip Data Analysis. Methods Mol Biol Clifton NJ 1708:303–330. https://doi.org/10.1007/978-1-4939-7481-8_16 Tian Y, Morris TJ, Webster AP, et al (2017) ChAMP: updated methylation analysis pipeline for Illumina BeadChips. Bioinforma Oxf Engl 33(24):3982–3984. https://doi.org/10.1093/bioinformatics/btx513 Holbrook JD, Huang R-C, Barton SJ, Saffery R, Lillycrop KA (2017) Is cellular heterogeneity merely a confounder to be removed from epigenome-wide association studies? Epigenomics 9(8):1143–1150. https://doi.org/10.2217/epi-2017-0032 Das J, Wadhwa N, Natchu UC, et al (2023) Genome-wide temporal landscaping of DNA methylation in pregnant women delivering at term: a GARBH-InI study. Epigenomics 15(9):543–556. https://doi.org/10.2217/epi-2023-0145 Barton SJ, Melton PE, Titcombe P, et al (2019) In Epigenomic Studies, Including Cell-Type Adjustments in Regression Models Can Introduce Multicollinearity, Resulting in Apparent Reversal of Direction of Association. Front Genet 10:816. https://doi.org/10.3389/fgene.2019.00816 Tang D, Chen M, Huang X, et al (2023) SRplot: A free online platform for data visualization and graphing. PloS One 18(11):e0294236. https://doi.org/10.1371/journal.pone.0294236 Yu G, Wang L-G, Han Y, He Q-Y (2012) clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters. OMICS J Integr Biol 16(5):284–287. https://doi.org/10.1089/omi.2011.0118 Sharma A, Lee-Ødegård S, Qvigstad E, et al (2022) β-Cell Function, Hepatic Insulin Clearance, and Insulin Sensitivity in South Asian and Nordic Women After Gestational Diabetes Mellitus. Diabetes 71(12):2530–2538. https://doi.org/10.2337/db22-0622 Omazić J, Viljetić B, Ivić V, et al (2021) Early markers of gestational diabetes mellitus: what we know and which way forward? Biochem Medica 31(3):030502. https://doi.org/10.11613/BM.2021.030502 Wang J, Huang P, Hou F, Hao D, Li W, Jin H (2024) Predicting gestational diabetes mellitus risk at 11–13 weeks’ gestation: the role of extrachromosomal circular DNA. Cardiovasc Diabetol 23(1):289. https://doi.org/10.1186/s12933-024-02381-1 Juchnicka I, Kuźmicki M, Niemira M, et al (2022) miRNAs as Predictive Factors in Early Diagnosis of Gestational Diabetes Mellitus. Front Endocrinol 13:839344. https://doi.org/10.3389/fendo.2022.839344 Popova PV, Klyushina AA, Vasilyeva LB, et al (2021) Association of Common Genetic Risk Variants With Gestational Diabetes Mellitus and Their Role in GDM Prediction. Front Endocrinol 12:628582. https://doi.org/10.3389/fendo.2021.628582 Tang Z, Wang S, Li X, et al (2024) Longitudinal integrative cell-free DNA analysis in gestational diabetes mellitus. Cell Rep Med 5(8):101660. https://doi.org/10.1016/j.xcrm.2024.101660 Borowiec M, Liew CW, Thompson R, et al (2009) Mutations at the BLK locus linked to maturity onset diabetes of the young and beta-cell dysfunction. Proc Natl Acad Sci U S A 106(34):14460–14465. https://doi.org/10.1073/pnas.0906474106 Bonnefond A, Yengo L, Philippe J, et al (2013) Reassessment of the putative role of BLK-p.A71T loss-of-function mutation in MODY and type 2 diabetes. Diabetologia 56(3):492–496. https://doi.org/10.1007/s00125-012-2794-8 Secolin R, Gonsales MC, Rocha CS, et al (2021) Exploring a Region on Chromosome 8p23.1 Displaying Positive Selection Signals in Brazilian Admixed Populations: Additional Insights Into Predisposition to Obesity and Related Disorders. Front Genet 12:636542. https://doi.org/10.3389/fgene.2021.636542 van Bilsen JHM, van den Brink W, van den Hoek AM, et al (2021) Mechanism-Based Biomarker Prediction for Low-Grade Inflammation in Liver and Adipose Tissue. Front Physiol 12:703370. https://doi.org/10.3389/fphys.2021.703370 Wang Y, Wu Y, Yang S, Chen Y (2023) Comparison of Plasma Exosome Proteomes Between Obese and Non-Obese Patients with Type 2 Diabetes Mellitus. Diabetes Metab Syndr Obes Targets Ther 16:629–642. https://doi.org/10.2147/DMSO.S396239 Altman J, Bai S, Purohit S, et al (2024) A candidate panel of eight urinary proteins shows potential of early diagnosis and risk assessment for diabetic kidney disease in type 1 diabetes. J Proteomics 300:105167. https://doi.org/10.1016/j.jprot.2024.105167 Ahluwalia TS, Rönkkö TKE, Eickhoff MK, et al (2024) Randomized Trial of SGLT2 Inhibitor Identifies Target Proteins in Diabetic Kidney Disease. Kidney Int Rep 9(2):334–346. https://doi.org/10.1016/j.ekir.2023.11.020 Kim Y-S, Gu B-H, Choi B-C, et al (2013) Apolipoprotein A-IV as a novel gene associated with polycystic ovary syndrome. Int J Mol Med 31(3):707–716. https://doi.org/10.3892/ijmm.2013.1250 Hanoun N, Fritsch S, Gayet O, et al (2014) The E3 ubiquitin ligase thyroid hormone receptor-interacting protein 12 targets pancreas transcription factor 1a for proteasomal degradation. J Biol Chem 289(51):35593–35604. https://doi.org/10.1074/jbc.M114.620104 Bai N, Lu X, Jin L, et al (2022) CLSTN3 gene variant associates with obesity risk and contributes to dysfunction in white adipose tissue. Mol Metab 63:101531. https://doi.org/10.1016/j.molmet.2022.101531 Li W, Jiang Q, Chen S, Liu J (2024) Adipose-specific CLSTN3B gene associates with human obesity. Metab Open 22:100269. https://doi.org/10.1016/j.metop.2024.100269 Marques ES, Formato E, Liang W, Leonard E, Timme-Laragy AR (2022) Relationships between type 2 diabetes, cell dysfunction, and redox signaling: A meta-analysis of single-cell gene expression of human pancreatic α- and β-cells. J Diabetes 14(1):34–51. https://doi.org/10.1111/1753-0407.13236 Linares-Pineda TM, Fragoso-Bargas N, Picón MJ, et al (2024) DNA methylation risk score for type 2 diabetes is associated with gestational diabetes. Cardiovasc Diabetol 23(1):68. https://doi.org/10.1186/s12933-024-02151-z Kim JH, Park K, Lee SB, et al (2019) Relationship between natural killer cell activity and glucose control in patients with type 2 diabetes and prediabetes. J Diabetes Investig 10(5):1223–1228. https://doi.org/10.1111/jdi.13002 Spielmann J, Naujoks W, Emde M, et al (2020) High-Fat Diet and Feeding Regime Impairs Number, Phenotype, and Cytotoxicity of Natural Killer Cells in C57BL/6 Mice. Front Nutr 7:585693. https://doi.org/10.3389/fnut.2020.585693 Li Y, Wang F, Imani S, Tao L, Deng Y, Cai Y (2021) Natural Killer Cells: Friend or Foe in Metabolic Diseases? Front Immunol 12:614429. https://doi.org/10.3389/fimmu.2021.614429 SIFNAIOS E, MASTORAKOS G, PSARRA K, et al (2019) Gestational Diabetes and T-cell (Th1/Th2/Th17/Treg) Immune Profile. In Vivo 33(1):31–40. https://doi.org/10.21873/invivo.11435 Atègbo J-M, Grissa O, Yessoufou A, et al (2006) Modulation of adipokines and cytokines in gestational diabetes and macrosomia. J Clin Endocrinol Metab 91(10):4137–4143. https://doi.org/10.1210/jc.2006-0980 Wagner W (2022) How to Translate DNA Methylation Biomarkers Into Clinical Practice. Front Cell Dev Biol 10:854797. https://doi.org/10.3389/fcell.2022.854797 Locke WJ, Guanzon D, Ma C, et al (2019) DNA Methylation Cancer Biomarkers: Translation to the Clinic. Front Genet 10:1150. https://doi.org/10.3389/fgene.2019.01150 Tables Table 1: Baseline characteristics at different visits of pregnancy Variables All STRiDE participants (n=2115) All (Nested case -control) participants (n=258) NGT (n=147) GDM (n=111) Early Pregnancy (V1) Age (years) 27 ± 4.1 27 ± 3.7 26 ± 3.5 27 ± 3.9 Gestational Age (weeks) at inclusion 10.5 ± 3 10.4 ± 3 10.2 ± 3 10.6 ± 3 Pre-pregnancy weight (kg) 59.2 ± 12.6 60 ± 11.7 60 ± 10.1 60 ± 13.6 Body mass index (kg/m 2 ) 24.2 ± 4.6 24.5 ± 4.1 24.3 ± 3.8 24.7 ± 4.6 Waist Circumference (cm) 86 ± 10 85.9 ± 8.6 85.0 ± 8.3 86.9 ± 8.8 Systolic Blood Pressure (mm Hg) 102 ± 10 102 ± 10 101 ± 10 103 ± 10* Diastolic Blood Pressure (mm Hg) 69 ± 9 68 ± 7 67 ± 8 69 ± 7 Fasting plasma glucose (mg/dl) 83 ± 5.2 84 ± 4 83 ± 4 84 ± 4 HbA 1c (Venous, %) 5.08 ± 0·3 5.09 ± 0.3 5.05 ± 0.3 5.14 ± 0.3* HbA 1c (Point of care, %) 5.15 ± 0·5 5.15 ± 0·5 5.13 ± 0·4 5.18 ± 0·5 Family history of type 2 diabetes n [%] 860 (40.7) 148 (57.4) 76 (51.7) 67 (60.4) Socio Economic Status (SES) n (%) Lower Class 499 (23.6) 21 (8.1) 7 (4.8) 14 (12.6) * Middle Class 1177 (55.7) 155 (60.1) 87 (59.2) 68 (61.3) * Upper Class 439 (20.8) 82 (31.8) 53 (36.0) 29 (26.1) * OGTT Visit (V2) Body mass index (kg/m 2 ) 24.2 ± 4.7 26.4 ± 3.9 26.0 ± 3.5 26.9 ± 4.4 Systolic Blood Pressure (mm Hg) 102 ± 10.5 104 ± 11 102 ± 9 105 ± 12* Diastolic Blood Pressure (mm Hg) 68.8 ± 9.4 68 ± 8 67 ± 8 69 ± 8 OGTT Fasting plasma glucose (mg/dl) 81.5 ± 8.1 83 ± 7 80 ± 5 86 ± 7** OGTT 1hr Venous plasma glucose (mg/dl) 137 ± 30.8 150 ± 33 129 ± 21 177 ± 27** OGTT 2hr Venous plasma glucose (mg/dl) 118 ± 26.1 131 ± 30 113 ± 18 154 ± 27** HbA 1c (Venous, %) 5.08 ± 0.3 4.86 ± 0.4 4.78 ± 0.3 4.97 ± 0.4** Graphical Abstract Graphical Abstract is not available with this version. Additional Declarations The authors declare no competing interests. Supplementary Files SupplementaryFigure1.tif Supplementary Figure 1: Schema of the study design - recruitment of study participants. SupplementaryFigure2.tif Supplementary Figure 2: Model development using conventional risk factors. · Model 1a: Age + BMI + FH of diabetes + venous HbA1c · Model 1b: Age + BMI + FH of diabetes + POC HbA1c · Model 1c: Age + waist-height ratio + FH of diabetes + venous HbA1c · Model 1d: Age + waist-height ratio + FH of diabetes + POC HbA1c SupplementaryFigure3.tif Supplementary Figure 3: DNA methylation levels of a panel of 5 CpGs during the OGTT visit (A-E). Bar plot depicts Mean ± SD. *p-value <0.01, **p-value <0.001, *** p-value <0.0001 compared to NGT. SupplementaryFigure4.tif Supplementary Figure 4: Correlation analysis of each CpGs at early pregnancy with glucose levels at OGTT. Spearman correlation matrix showing correlation coefficients between the 5 CpGs at early pregnancy and the OGTT fasting glucose levels at OGTT visit. *p<0.05, **p<0.01. SupplementaryFigure5.tif Supplementary Figure 5: Spearman correlation matrix showing correlation coefficients between the 5 CpGs at early pregnancy and the OGTT 60-minute glucose levels at OGTT visit. *p<0.05, **p<0.01. SupplementaryFigure6.tif Supplementary Figure 6: Spearman correlation matrix showing correlation coefficients between the 5 CpGs at early pregnancy and the OGTT 120-minute glucose levels at OGTT visit. *p<0.05, **p<0.01. SupplementaryTable1.docx Supplementary Table 1: Machine learning data analysis framework SupplementaryTable2.xlsx SupplementaryTable 2: List of 527 significant CpGs from EWAS analysis. SupplementaryTable3.docx SupplementaryTable 3: Performance of the CpG panels from different. Random Forest (RF), Least Absolute Shrinkage and Selection Operator (LASSO), Recursive Feature Elimination (RFE), Support Vector Machine (SVM). SupplementaryTable4.docx SupplementaryTable 4: Discovery of best-performed biomarker panel (5 CpGs) for GDM prediction using RF Model. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5280336","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":367093736,"identity":"d05c5590-dfac-4479-8de8-3659bc2c5bed","order_by":0,"name":"Kuppan Gokulakrishnan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzElEQVRIiWNgGAWjYFACNgaGBxUMPAwSMAFmYrQknCFZS2IbkJYgqBIK5N3b0iQS592R0Z3d/OzhDwY7eQZ23gN4tRieOXZMInHbMx6zO8fMjXkYkg0bmPkS8GuZkd4G1HKYx+xGgpk00CMJDMw8BkRomQPSkv5N8gdDPWEt8hJpQIc1gLTkmEnwMBwmrMWA51iyRcIxsJYyaR6D44ZtBG1pbzO88aHmsD3QYdskf1RUy/PznyFgywFULjie8AP5BkIqRsEoGAWjYBQAAHatPQ1gERqLAAAAAElFTkSuQmCC","orcid":"","institution":"National Institute of Mental Health \u0026 Neuro Sciences","correspondingAuthor":true,"prefix":"","firstName":"Kuppan","middleName":"","lastName":"Gokulakrishnan","suffix":""},{"id":367093737,"identity":"7843fd75-f298-4fd0-abf2-f30ad348a56b","order_by":1,"name":"Chinnasamy Thirumoorthy","email":"","orcid":"","institution":"National Institute of Mental Health \u0026 Neuro Sciences","correspondingAuthor":false,"prefix":"","firstName":"Chinnasamy","middleName":"","lastName":"Thirumoorthy","suffix":""},{"id":367093738,"identity":"0dd8c69d-2972-4cad-b264-887ad0ad53b9","order_by":2,"name":"Kuldeep Sharma","email":"","orcid":"","institution":"National Institute of Mental Health \u0026 Neuro Sciences","correspondingAuthor":false,"prefix":"","firstName":"Kuldeep","middleName":"","lastName":"Sharma","suffix":""},{"id":367099308,"identity":"187e5c1b-f2a8-417e-810d-ccacd4ac5778","order_by":3,"name":"Uma Ram","email":"","orcid":"","institution":"Seethapathy Clinic \u0026 Hospital","correspondingAuthor":false,"prefix":"","firstName":"Uma","middleName":"","lastName":"Ram","suffix":""},{"id":367099309,"identity":"6fcb28da-f85c-4841-b9b7-6de1a6d70e24","order_by":4,"name":"Mohan Deepa","email":"","orcid":"","institution":"Madras Diabetes Research Foundation","correspondingAuthor":false,"prefix":"","firstName":"Mohan","middleName":"","lastName":"Deepa","suffix":""},{"id":367099310,"identity":"e849bb8a-2429-42d4-921b-a723400fff1c","order_by":5,"name":"Yonas Weldeselassie","email":"","orcid":"","institution":"The Open University, UK","correspondingAuthor":false,"prefix":"","firstName":"Yonas","middleName":"","lastName":"Weldeselassie","suffix":""},{"id":367099311,"identity":"06a9a30a-43e4-4ae2-9256-aee891a030f7","order_by":6,"name":"Hailemariam Asres","email":"","orcid":"","institution":"The Open University, UK","correspondingAuthor":false,"prefix":"","firstName":"Hailemariam","middleName":"","lastName":"Asres","suffix":""},{"id":367099312,"identity":"23ca7a74-971e-4e8a-b0d1-f727cc6b9b46","order_by":7,"name":"U Venkatesh","email":"","orcid":"","institution":"Madras Diabetes Research Foundation","correspondingAuthor":false,"prefix":"","firstName":"U","middleName":"","lastName":"Venkatesh","suffix":""},{"id":367099313,"identity":"b5799260-0d8e-40de-b1ad-16dd8a4a39db","order_by":8,"name":"B Binukumar","email":"","orcid":"","institution":"National Institute of Mental Health \u0026 Neuro Sciences","correspondingAuthor":false,"prefix":"","firstName":"B","middleName":"","lastName":"Binukumar","suffix":""},{"id":367099314,"identity":"078f377c-19fb-471a-be34-4610b88abdf7","order_by":9,"name":"Venkatesan Radha","email":"","orcid":"","institution":"Madras Diabetes Research Foundation","correspondingAuthor":false,"prefix":"","firstName":"Venkatesan","middleName":"","lastName":"Radha","suffix":""},{"id":367099315,"identity":"77a9eab3-e5f7-4bcb-8c17-5da2190bf827","order_by":10,"name":"Polina Popova","email":"","orcid":"","institution":"Almazov National Medical Research Centre","correspondingAuthor":false,"prefix":"","firstName":"Polina","middleName":"","lastName":"Popova","suffix":""},{"id":367099316,"identity":"73985571-d0df-4156-92d9-16925abe2d41","order_by":11,"name":"Sapna Sharma","email":"","orcid":"","institution":"Helmholtz Zentrum München Deutsches Forschungszentrum für Gesundheit und Umwelt (GmbH)","correspondingAuthor":false,"prefix":"","firstName":"Sapna","middleName":"","lastName":"Sharma","suffix":""},{"id":367099317,"identity":"0cd5245e-a294-4dd6-b626-22327baf9792","order_by":12,"name":"Harald Grallert","email":"","orcid":"","institution":"Helmholtz Zentrum München Deutsches Forschungszentrum für Gesundheit und Umwelt (GmbH)","correspondingAuthor":false,"prefix":"","firstName":"Harald","middleName":"","lastName":"Grallert","suffix":""},{"id":367099318,"identity":"e9c9ae0d-5c2e-48c0-800f-56ca526a68c0","order_by":13,"name":"Ranjit Mohan Anjana","email":"","orcid":"","institution":"Madras Diabetes Research Foundation","correspondingAuthor":false,"prefix":"","firstName":"Ranjit","middleName":"Mohan","lastName":"Anjana","suffix":""},{"id":367099319,"identity":"8b2ddbea-d436-4d77-85d2-965f1e4bacc3","order_by":14,"name":"Muthuswamy Balasubramanyam","email":"","orcid":"","institution":"Madras Diabetes Research Foundation","correspondingAuthor":false,"prefix":"","firstName":"Muthuswamy","middleName":"","lastName":"Balasubramanyam","suffix":""},{"id":367099320,"identity":"960077cd-2f90-410c-808d-1708cd8ff35e","order_by":15,"name":"Nikhil Tandon","email":"","orcid":"","institution":"All India Institute of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Nikhil","middleName":"","lastName":"Tandon","suffix":""},{"id":367099321,"identity":"3eaf12ec-f9c9-4c61-b9f8-9f826cff2845","order_by":16,"name":"Viswanathan Mohan","email":"","orcid":"","institution":"Madras Diabetes Research Foundation","correspondingAuthor":false,"prefix":"","firstName":"Viswanathan","middleName":"","lastName":"Mohan","suffix":""},{"id":367099322,"identity":"808be0c8-8b37-4e92-b95f-b67978f5bd10","order_by":17,"name":"Ponnusamy Saravanan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAklEQVRIiWNgGAWjYJADxgcgkg0IwUCCCC3MBiRrYYMqIqDFnIH94eeKmsPy5uxnj1Xz1NxL7GM/lsDwo4YhcWYDdi2WDTzGkmeOHTbc2ZOXdpvnWHFiG0/aAcaeYwyJs3HYYnCAh0GygS2NccOBHLPbOWwJuW0S7A0MvA0MifNwamF//LPhX5r9hvNvzIpz/kG0MP7Fq4XBTLKxzSZxw40cM+bcNpAWtgPMIFtwOcyymcfMsrHPJnnDjTfG0n/7EuqBfkk4LHNMwhiX983Z2x/fbPgmYbvhfI7hxxnfEozl248ZPnxTYyM74wAOhzFjEz2ALyINcMqMglEwCkbBKIABAGumV7U2/CjYAAAAAElFTkSuQmCC","orcid":"","institution":"University of Warwick, UK","correspondingAuthor":true,"prefix":"","firstName":"Ponnusamy","middleName":"","lastName":"Saravanan","suffix":""}],"badges":[],"createdAt":"2024-10-17 06:56:13","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-5280336/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5280336/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":66980781,"identity":"bd2f71b3-ee88-4b93-87ad-98293401a293","added_by":"auto","created_at":"2024-10-18 17:23:44","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":6431942,"visible":true,"origin":"","legend":"\u003cp\u003eGDM prediction using conventional risk factors. \u003cstrong\u003eA:\u003c/strong\u003e Composite risk score (venous or point-of-care HbA1c, age, BMI, and family history of diabetes) for GDM prediction. \u003cstrong\u003eB:\u003c/strong\u003e Conventional risk predictors (age, waist-to-height ratio, family history of diabetes, point-of-care HbA1c, and blood pressure) for GDM prediction.\u003c/p\u003e","description":"","filename":"Figure1.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/bb82b32ebd7a031e201767b9.png"},{"id":66980780,"identity":"8a6b3a64-2917-4dfd-abc0-02df257bfa07","added_by":"auto","created_at":"2024-10-18 17:23:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":74002106,"visible":true,"origin":"","legend":"\u003cp\u003eThe distribution and significance of differentially methylated CpGs at early pregnancy among the study participants are illustrated using Manhattan plots and heat maps. \u003cstrong\u003eA:\u003c/strong\u003e Manhattan plots show the chromosomal distribution of CpG sites with significant differential methylation levels between NGT and GDM. The x-axis represents chromosomal locations, while the y-axis displays the negative log10 p-value values. The dotted horizontal line indicates the genome-wide significance threshold (p\u003csub\u003efdr\u003c/sub\u003e-value \u0026lt; 0.000001). Labelled 5 CpGs (highlighted among DMPs in red) are identified as robust GDM predictive biomarkers in the RF machine learning model. \u003cstrong\u003eB:\u003c/strong\u003e Heat map visualizing the methylation levels of 5 differentially methylated CpGs in NGT compared to GDM.\u003c/p\u003e","description":"","filename":"Figure2.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/ec7f2ec2ab2462835bcaa51a.png"},{"id":66980771,"identity":"ba3f6243-6fcb-4dd8-b560-cb937444684e","added_by":"auto","created_at":"2024-10-18 17:23:41","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":21260931,"visible":true,"origin":"","legend":"\u003cp\u003eDNA Methylation levels of the best-performed panel (5 CpGs) at early pregnancy in NGT vs GDM (\u003cstrong\u003eA-E\u003c/strong\u003e). Bar graphs depict Mean ± SD. *p-value\u0026lt;0.001 compared to NGT.\u003c/p\u003e","description":"","filename":"Figure3.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/e7dbd7871294c3c51c1c6b74.png"},{"id":66980783,"identity":"ba803d30-cb91-4c67-99a1-a4e2e7d48915","added_by":"auto","created_at":"2024-10-18 17:23:44","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":12075834,"visible":true,"origin":"","legend":"\u003cp\u003eROC analysis of the model utilizing 5 CpGs and conventional risk factors at early pregnancy. \u003cstrong\u003eA:\u003c/strong\u003eROC showing the novel DNA methylation signatures of 5 CpG panel as candidate biomarker predictors of GDM. \u003cstrong\u003eB:\u003c/strong\u003e ROC curves and AUC of different prediction models. Model 1a: Conventional risk predictors (age, waist-to-height ratio, family history of diabetes, point-of-care HbA1c, and blood pressure), Model 1b: Novel candidate biomarker panel of 5 CpGs at early pregnancy. Model 1c: Full GDM prediction model\u003cstrong\u003e - \u003c/strong\u003eintegrated clinically accessible maternal risk factors (conventional risk predictors) with novel DNA methylation signatures of 5 CpG panel (1a \u0026amp; 1b combined).\u003c/p\u003e","description":"","filename":"Figure4.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/ec4adc8243bfdd286e6b65d8.png"},{"id":66980775,"identity":"da0d7cf7-dabd-4f33-985f-92031d62ab4f","added_by":"auto","created_at":"2024-10-18 17:23:43","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":11491146,"visible":true,"origin":"","legend":"\u003cp\u003eGO and pathway analysis of differentially methylated CpGs and their annotated genes. \u003cstrong\u003eA:\u003c/strong\u003e Bar plot presenting the top 10 terms for biological process, molecular function, and cellular component from the GO analysis of genes annotated to DMPs. \u003cstrong\u003eB:\u003c/strong\u003e KEGG pathway enrichment analysis of Differentially Methylated Genes (DMGs) is displayed. Node size indicates the number of genes, and node color reflects the significant level of pathway enrichment.\u003c/p\u003e\n\u003cp\u003eBiological Processes (BP), Cellular Components (CC), Molecular Functions (MF)\u003c/p\u003e","description":"","filename":"Figure5.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/62695a919e9b0f495b0d8211.png"},{"id":66980768,"identity":"1a250b23-a93f-4b25-93fa-3bcdb776b46e","added_by":"auto","created_at":"2024-10-18 17:23:40","extension":"tif","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2730368,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Figure 1: \u003c/strong\u003eSchema of the study design - recruitment of study participants.\u003c/p\u003e","description":"","filename":"SupplementaryFigure1.tif","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/777d7544b271565747fa7c26.tif"},{"id":66980785,"identity":"e03444a9-f8c5-4b5b-bd90-c9f30c2a3a87","added_by":"auto","created_at":"2024-10-18 17:23:44","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":3509834,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Figure 2: \u003c/strong\u003eModel development using conventional risk factors.\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eModel 1a:\u003c/strong\u003e Age + BMI + FH of diabetes + venous HbA1c\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eModel 1b: \u003c/strong\u003eAge + BMI + FH of diabetes + POC HbA1c\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eModel 1c: \u003c/strong\u003eAge + waist-height ratio + FH of diabetes + venous HbA1c\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eModel 1d: \u003c/strong\u003eAge + waist-height ratio + FH of diabetes + POC HbA1c\u003c/p\u003e","description":"","filename":"SupplementaryFigure2.tif","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/fd4f60f4092ba8d2dcf6cd7b.tif"},{"id":66981007,"identity":"42d40954-add4-40ed-aec3-79fdc92589fc","added_by":"auto","created_at":"2024-10-18 17:31:43","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":2575072,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Figure 3: \u003c/strong\u003eDNA methylation levels of a panel of 5 CpGs during the OGTT visit (A-E). Bar plot depicts Mean ± SD. *p-value \u0026lt;0.01, **p-value \u0026lt;0.001, *** p-value \u0026lt;0.0001 compared to NGT.\u003c/p\u003e","description":"","filename":"SupplementaryFigure3.tif","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/0ee0d3663b495aecbb8f0a44.tif"},{"id":66980778,"identity":"51f74da1-3e6b-4526-87d2-923479ffa5d0","added_by":"auto","created_at":"2024-10-18 17:23:43","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":6471548,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Figure 4:\u003c/strong\u003e Correlation analysis of each CpGs at early pregnancy with glucose levels at OGTT.\u003c/p\u003e\n\u003cp\u003eSpearman correlation matrix showing correlation coefficients between the 5 CpGs at early pregnancy and the OGTT fasting glucose levels at OGTT visit. *p\u0026lt;0.05, **p\u0026lt;0.01.\u003c/p\u003e","description":"","filename":"SupplementaryFigure4.tif","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/3bacc90095d5b67a50ce6332.tif"},{"id":66980784,"identity":"2bc587bd-27e2-4d61-b72f-70f8291c87e1","added_by":"auto","created_at":"2024-10-18 17:23:44","extension":"tif","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":6127780,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Figure 5:\u003c/strong\u003e Spearman correlation matrix showing correlation coefficients between the 5 CpGs at early pregnancy and the OGTT 60-minute glucose levels at OGTT visit. *p\u0026lt;0.05, **p\u0026lt;0.01.\u003c/p\u003e","description":"","filename":"SupplementaryFigure5.tif","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/1d0a0838e329652f6cba7d5a.tif"},{"id":66980776,"identity":"fda32bd7-43f6-4f31-ae39-5b2fb5560076","added_by":"auto","created_at":"2024-10-18 17:23:43","extension":"tif","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":6307116,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Figure 6:\u003c/strong\u003e Spearman correlation matrix showing correlation coefficients between the 5 CpGs at early pregnancy and the OGTT 120-minute glucose levels at OGTT visit. *p\u0026lt;0.05, **p\u0026lt;0.01.\u003c/p\u003e","description":"","filename":"SupplementaryFigure6.tif","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/ff5ea230a87814b4d065608a.tif"},{"id":66980770,"identity":"542db4fb-e81c-4390-8b50-9098c0cbae6b","added_by":"auto","created_at":"2024-10-18 17:23:41","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":20660,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Table 1: \u003c/strong\u003eMachine learning data analysis framework\u003c/p\u003e","description":"","filename":"SupplementaryTable1.docx","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/f69159ea32ed638a66bcfa0f.docx"},{"id":66980779,"identity":"fc6c3a91-d356-4c09-9c27-5fc009f96642","added_by":"auto","created_at":"2024-10-18 17:23:43","extension":"xlsx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":122016,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementaryTable 2: \u003c/strong\u003eList of 527 significant CpGs from EWAS analysis.\u003c/p\u003e","description":"","filename":"SupplementaryTable2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/fc6c46c2d222602b6d655995.xlsx"},{"id":66981008,"identity":"44870dcb-1576-4cb7-9c35-a9976b8018a8","added_by":"auto","created_at":"2024-10-18 17:31:44","extension":"docx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":17129,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementaryTable 3: \u003c/strong\u003ePerformance of the\u003cstrong\u003e \u003c/strong\u003eCpG panels from different.\u003c/p\u003e\n\u003cp\u003eRandom Forest (RF), Least Absolute Shrinkage and Selection Operator (LASSO), Recursive Feature Elimination (RFE), Support Vector Machine (SVM).\u003c/p\u003e","description":"","filename":"SupplementaryTable3.docx","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/faa69fe3e14a6644147040f0.docx"},{"id":66980774,"identity":"f6d408f4-c078-41e1-8706-30b35c4521d1","added_by":"auto","created_at":"2024-10-18 17:23:43","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":17543,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementaryTable 4: \u003c/strong\u003eDiscovery of best-performed biomarker panel (5 CpGs) for GDM prediction using RF Model.\u003c/p\u003e","description":"","filename":"SupplementaryTable4.docx","url":"https://assets-eu.researchsquare.com/files/rs-5280336/v1/8e669f4a2688eec0bcba5d33.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eFirst-trimester DNA Methylome profiling identifies novel predictors for gestational diabetes mellitus in Indian women\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGestational diabetes mellitus (GDM) is a prevalent metabolic disorder characterized by glucose intolerance with onset or first recognition during pregnancy. It is a significant public health concern globally\u0026nbsp;[1], affecting up to 20% of pregnancies in some\u0026nbsp;populations.\u0026nbsp;GDM not only increases the risk of adverse outcomes during pregnancy but also has long-term health implications for both mother and child\u0026nbsp;[2]. Women with GDM are at higher risk of developing type 2 diabetes mellitus (T2DM)\u0026nbsp;[3], cardiovascular disease\u0026nbsp;[4, 5], and obesity\u0026nbsp;[6]\u0026nbsp;later in life, while their offspring are more likely to face an elevated risk of macrosomia and other cardiometabolic disorders later in life\u0026nbsp;[7, 8].\u0026nbsp;GDM is typically diagnosed in the late stages of pregnancy (24-28 weeks)\u0026nbsp;[9, 10], but excess fetal adiposity may already be evident\u0026nbsp;[11, 12].\u0026nbsp;\u0026nbsp;Yet predictive markers for GDM remain an unmet demand in clinical practice.\u003c/p\u003e\n\u003cp\u003eRecent efforts have focused on developing prediction models\u0026nbsp;[13]\u0026nbsp;using conventional clinical and blood-based markers such as maternal age, body mass index (BMI)\u0026nbsp;[14], family history of diabetes, HbA1c, cardiometabolic\u0026nbsp;[15], metabolomic\u0026nbsp;[16, 17], and small non-coding circulating RNA\u0026nbsp;[18]\u0026nbsp;biomarkers for early \u0026nbsp;prediction of GDM.\u0026nbsp;However, the predictive accuracy of these models is, at best, moderate, particularly when used in the early stages of pregnancy. Our recent prospective pregnancy cohort of 2115 pregnant women from the STratification of Risk of Diabetes in Early pregnancy (STRiDE) study used composite risk score models using HbA1c showed better predictability than simple clinical risk factors\u0026nbsp;[19]. However, the predictive ability of a \u0026lsquo;single-threshold\u0026rsquo; was still moderate and a 2-threshold approach can reduce the burden of oral glucose tolerance test (OGTT) by at least 50%\u0026nbsp;[20]. Hence, if a composite risk score using genomic or epigenetic markers could be developed,\u0026nbsp;this could result in precision diagnosis and management of GDM.\u003c/p\u003e\n\u003cp\u003eRecent advances in epigenetics, specifically DNA methylation, offer promising avenues for the early identification of GDM\u0026nbsp;[21, 22]. DNA methylation is a form of epigenetic modification that can influence\u0026nbsp;gene expression and has been implicated in various metabolic and inflammatory processes associated with GDM\u0026nbsp;[23]. Recent Epigenome-Wide Association Studies (EWAS) have identified specific DNA methylation patterns associated with GDM\u0026nbsp;[24]. While these findings are promising, the studies were conducted in late pregnancy,\u0026nbsp;limited to candidate genes\u0026nbsp;and remain inconclusive\u0026nbsp;[22, 24]. In addition, none have shown any potential for prediction of GDM. Integrating the analysis of DNA methylation biomarkers with maternal risk factors is crucial for developing effective GDM prediction models. There is a lack of studies that have explored the combined evaluation of DNA methylation biomarkers alongside maternal risk factors, highlighting a significant gap. Our present work is an attempt to fill that gap in knowledge.\u003c/p\u003e\n\u003cp\u003eWe hypothesized that DNA methylation changes identified through EWAS in early pregnancy, combined with common clinical risk factors and HbA1c levels, could accurately predict GDM diagnosed via OGTT at 24-28 weeks. A nested cohort study sample of 258 women with fasting glucose levels below 5.1mmol/l from the prospective STRiDE study was selected for the EWAS analysis.\u0026nbsp;\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eStudy participants:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe STRiDE study population consisted of 2115 Indian pregnant women from the STRiDE study, a prospective, longitudinal cohort, that recruited at the first antenatal visit but before 16 weeks of gestation. The study was conducted between 2016 and 2019 across seven clinical sites in three cities of South India, and the details of the cohort were published elsewhere\u0026nbsp;[19]. \u0026nbsp;For the present nested cohort study within the original STRiDE cohort, a total of 258 pregnant women (who had fasting glucose levels of \u0026lt;5.1 mmol/L) in early pregnancy were randomly selected. Participants with similar age and BMI were selected to minimise the influence of these common risk factors. At 24\u0026ndash;28 weeks (OGTT visit) of gestation, a 75g OGTT was conducted, and women were classified into Normal Glucose Tolerance [NGT] (n = 147) and GDM (n = 111). The International Association of Diabetes and Pregnancy Study Groups (IADPSG) criteria were used to diagnose GDM\u0026nbsp;[25]. DNA methylation profiling was carried out in all enrolled study participants in this nested cohort at early pregnancy, and at OGTT VISIT (NGT n=96, GDM n=93). All the samples (cases/control) were randomised before the experiment. The schema of the study design is shown in \u003cstrong\u003eSupplementary Figure 1\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eWritten informed consent was obtained from all participants. The study was approved by the Institutional Ethics Committee of the National Institute of Mental Health and Neuro Sciences (NIMHANS), Bengaluru and the Madras Diabetes Research Foundation (MDRF), Chennai, India.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDNA isolation:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGenomic DNA was isolated from whole blood samples obtained from all the enrolled participants, employing a phenol-chloroform extraction method\u0026nbsp;[26]. Subsequently, the quantification of DNA samples was performed using the Qubit\u0026reg; 2.0 Fluorometer with the Qubit dsDNA HS Assay Kit, along with NanoDrop One C (ThermoFisher Scientific, USA).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBisulfite conversion, DNA methylation profiling \u0026amp; Bioinformatics:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBisulfite conversion was performed using the EZ DNA Methylation Kit from Zymo Research (USA) and subsequently analyzed with the HumanMethylationEPIC 850k BeadChip assays (Illumina, San Diego, CA, USA), following the manufacturer\u0026apos;s instructions on an iScan reader (Illumina). To minimize potential batch effects, both case and control samples were processed and run concurrently on the same BeadChip.\u003c/p\u003e\n\u003cp\u003eCpGs with poor signal quality (detection p-value \u0026gt; 0.05), those located on the X and Y chromosomes, common SNPs, and cross-reactive probes were excluded from the analysis\u0026nbsp;[27, 28]. This resulted in a final dataset comprising 258 samples and 597,856 CpGs for EWAS. Prior to conducting the EWAS, data normalization was performed using the quantile normalization method. The Champ.DMP function from the ChAMP (Chip Analysis Methylation Pipeline for Illumina HumanMethylation450 and EPIC) package was utilized to execute the EWAS, employing a linear regression model with CpGs as dependent variables and the grouping variable for GDM status (present/absent)\u0026nbsp;[29]. The output included key parameters such as fold change, p-value, and Benjamin-Hochberg adjusted p-value as p\u003csub\u003efdr\u003c/sub\u003e-value \u0026lt; 0.000001 to identify differentially methylated CpGs, resulting in a total of 808 CpGs. Ultimately, we selected 527 CpG sites that were annotated to genes for further data analysis. We used Random Forest (RF), Least Absolute Shrinkage and Selection Operator (LASSO), and Recursive Feature Elimination (RFE) for feature selection. \u0026nbsp;Two classifiers - RF and Radial Kernel Support Vector Machines (SVM) - were applied to construct biomarker panels using the CpG lists derived from the above feature selection methods as input variables. The panel construction employed a forward stepwise strategy aimed at optimizing the AUC, incrementally adding significant CpGs one by one to enhance model performance. The overall framework for developing these biomarker panels is illustrated in \u003cstrong\u003eSupplementary Table 1\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eThe analysis was conducted across six distinct scenarios, combining the three feature selection techniques with the two classifiers. For each scenario, ten iterations were conducted by randomizing the samples in the training and testing datasets, ensuring reproducibility through fixed seeds. Panels of varying sizes were constructed, starting with a single CpG and expanding to ten CpGs per panel, culminating in ten unique sets of panels for each scenario, all optimized for AUC in each iteration. The stability of each biomarker panel was evaluated through bootstrap validation, comprising 1000 runs. This validation process generated aggregated metrics, including AUC, sensitivity, specificity, and accuracy, along with their corresponding standard deviations (SD) from the 1000 iterations. The panel exhibiting the lowest SD was identified as the most robust model for each scenario, as a lower SD indicates enhanced model stability. Ultimately, this systematic approach resulted in a compilation of the best biomarker panel for each scenario. The final panel was selected based on achieving the highest AUC among the six panels identified.\u003c/p\u003e\n\u003cp\u003eAll analyses were executed using R (version 4.3.2), employing the \u0026quot;train\u0026quot; function from the caret package to implement both the RF and SVM classifiers. For the RF model, parameters such as \u0026quot;ntree\u0026quot; (number of trees) and \u0026quot;mtry\u0026quot; (number of randomly selected CpGs at each tree) were manually optimized, whereas the SVM model parameters, C (the margin between the classifier line and support vectors) and sigma (degree of non-linearity), were automatically tuned through 10-fold cross-validation with five repeats.\u0026nbsp;Recent studies have shown a strong correlation between cell types in peripheral blood during inflammatory processes, particularly in pregnancy\u0026nbsp;[30, 31]. Therefore, we did not perform Houseman cell-type adjustment in our analysis. This decision helps to avoid violations of the no multicollinearity assumption in EWAS, which can lead to reversed associations, especially when investigating inflammation-related phenotypes such as those observed in pregnancy. Additionally, it minimizes the risk of obtaining spurious significant results\u0026nbsp;[31, 32].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional enrichment GO and KEGG analysis:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis for the 527 combined differentially methylated positions (DMPs) and their annotated genes were performed using the SRplot online tool\u0026nbsp;[33, 34]. For both the GO and KEGG analyses, the top 10 pathways were selected for further examination. A significant threshold of a false discovery rate (FDR) of less than 0.05 was established to identify biologically relevant pathways and processes associated with the DMRs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical Analysis:\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNumeric variables were summarized with mean \u0026plusmn; standard deviation (SD) for continuous variables and n (%) for categorical variables. The Kolmogorov\u0026ndash;Smirnov test confirmed normality, allowing for parametric tests due to the normally distributed nature of the data. Continuous variable averages were compared using the independent t-test, while proportions were assessed using the Chi-square test or Fisher\u0026rsquo;s exact test as appropriate. Spearman correlation analysis was performed to explore the association between the best panel of CpGs at early pregnancy and OGTT glucose levels at OGTT visit.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAfter obtaining the best panel of CpGs building step-by-ML approaches, further logistic regression models were fitted to have a better understanding of these CpGs and other maternal risk factors related to GDM. Each logistic regression model\u0026rsquo;s predictive performance was assessed by plotting\u0026nbsp;the receiver operating characteristic curve (ROC), from which AUC with 95% CI was calculated as the primary predictive measure. Statistical analyses were conducted using R (Version 4.3.2), with p-values \u0026lt; 0.05 deemed statistically significant.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel developments: -\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003ea. Conventional risk predictors considering clinical accessibility:\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOverall, we considered well-documented clinically accessible maternal risk factors and blood-based biomarkers associated with the development of GDM, which are available in the original full cohort of 2115 pregnant women and included the same in this nested GDM case-control subset as potential conventional risk predictors (age, waist-height ratio, family history (FH) of diabetes, point of care (POC) HbA1c and blood pressure). These predictors are routinely assessed at relatively affordable clinical costs. First, we built logistic regression models using the maternal risk factors at early pregnancy, with the outcome being the presence or absence of GDM based on OGTT results at OGTT visit. Model 1a fitted with how age, BMI, FH of diabetes and venous HbA1c predicted GDM. \u0026nbsp; Model 1b tested the association of age, BMI, FH diabetes, and POC HbA1c instead of venous HbA1c adjusted for each other. As both venous and POC HbA1c performed equally well and POC HbA1c can be used in rural settings without the need for a lab for HbA1c measurement, we retained the POC HbA1c in the final model. Additional analyses using waist-height ratio were presented instead of BMI to assess the potential role of waist circumference (Model 1c \u0026amp; 1d) (\u003cstrong\u003eSupplementary Figure 2\u003c/strong\u003e). \u0026nbsp;Further, we considered other variables significantly associated with GDM in our dataset (age, waist-height ratio, FH of diabetes, POC HbA1c and blood pressure) as conventional risk predictors. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eb. Candidate biomarkers predictors:\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThese include EWAS-derived novel DNA methylation biomarkers, which are not yet available for clinical use and require specialized laboratories for measurement.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003ec. Full model predicting GDM risk in the clinical context\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo develop a clinically efficient comprehensive GDM prediction model, we integrated conventional risk predictors with novel epigenetic candidate biomarkers predictors.\u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eEarly pregnancy and OGTT characteristics of the whole STRiDE cohort and the nested case-control study participants are shown in \u003cstrong\u003eTable 1\u003c/strong\u003e. \u0026nbsp;The mean gestational age in early pregnancy was around 10 weeks. Women who developed GDM at 24-28 weeks had marginally higher systolic blood pressure and glycated haemoglobin levels at early pregnancy. The composite risk score model of the whole STRiDE study cohort (n=2115), combining age, BMI, FH of diabetes and POC HbA1c had an AUC of 0.61 for predicting GDM. The nested EWAS cohort (n=258) using the same composite risk score had a very similar AUC of 0.63 \u003cstrong\u003e(Figure 1A)\u003c/strong\u003e suggesting that this is representative of the larger whole STRiDE study cohort of 2,115 individuals.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs\u0026nbsp;waist-height ratio may be a better metabolic predictor than BMI in South Asians \u0026nbsp;[35], we replaced BMI with waist-height ratio and added systolic BP to assess the predictive ability of GDM.\u0026nbsp;This prediction model, combining age, waist-height ratio, FH of diabetes, POC HbA1c, and blood pressure, had an AUC of 0.71 for GDM \u003cstrong\u003e(Figure 1B).\u003c/strong\u003e All these risk factors are either routinely measured or can be done at relatively affordable clinical costs. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of DNA Methylation signatures:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter initial processing, our dataset included 597,856 DNA methylation probes from each study participant. We performed a discovery\u0026nbsp;EWAS, applying a significance threshold of p\u003csub\u003efdr\u003c/sub\u003e-value \u0026lt;0.000001 to identify CpG sites significantly associated with GDM. This analysis identified 527 DMPs related to GDM \u003cstrong\u003e(Supplementary Table 2).\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeveloping DNA Methylation signature panel as candidate biomarkers and its performance metrics for early\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;GDM\u003c/strong\u003e \u003cstrong\u003eprediction:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo leverage the strengths of each method and ensure a consistent selection of the most informative CpGs for GDM prediction out of the 527 available at\u0026nbsp;early pregnancy, we applied three machine learning techniques: RF, RFE, and LASSO. The top 40 CpGs were chosen based on their feature-important scores using RF, which utilises a feature-importance ranking algorithm. Then, we used RFE, which systematically removes the least significant features, and found out that the model performed best with 40 CpGs. Lastly, we selected 21 CpGs as the most predictive features using LASSO, which penalises less significant features by shrinking their coefficients towards zero. All three feature selection methods were applied to the same dataset.\u0026nbsp;Using these selected CpGs from each approach, we developed prediction models for GDM utilizing the classifiers \u0026ndash; RF and SVM methodologies, incorporating bootstrap validation with 1000 repetitions.\u003c/p\u003e\n\u003cp\u003eThe\u0026nbsp;RF\u0026nbsp;method demonstrated the best performance, identifying a panel of five CpG sites at\u0026nbsp;early pregnancy\u0026nbsp;(cg10139015, cg23507676, cg11001216, cg04539775, cg04985016) for GDM prediction. In the test dataset, this panel achieved an AUC of 0.98, with a sensitivity of 0.98 and a specificity of 0.83. During bootstrap validation, the panel maintained robust performance, yielding an AUC of 0.89, a sensitivity of 0.83, and a specificity of 0.80.\u0026nbsp;\u003cstrong\u003eSupplementary Table 3\u0026nbsp;\u003c/strong\u003esummarises the performance metrics for the various feature selection methods in test and validation datasets.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe identified panel of five CpG sites, along with their annotated genes and the chromosome-wide distribution of 527 DMPs, is illustrated in the Manhattan plot (\u003cstrong\u003eFigure 2A\u003c/strong\u003e).\u0026nbsp;\u003cstrong\u003eFigure 2B\u003c/strong\u003e presents the hierarchical clustering of the five CpG sites at\u0026nbsp;early pregnancy\u0026nbsp;showing the differential expression patterns among GDM and NGT separately. Detailed information regarding these five CpGs, including their annotated genes and corresponding chromosomes are summarized in\u0026nbsp;\u003cstrong\u003eSupplementary Table 4.\u003c/strong\u003e \u003cstrong\u003eFigure 3A-E\u003c/strong\u003e shows that women who were diagnosed to have GDM at OGTT, had higher methylation levels of all these\u0026nbsp;five CpGs\u0026nbsp;compared to women who were NGT in early pregnancy. This higher methylation pattern of 5 CpGs were similar when measured at the time of OGTT (\u003cstrong\u003eSupplementary Figure 3\u003c/strong\u003e\u003cstrong\u003eA-E\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrediction of GDM combining clinical, POC HbA1c and DNA methylation signatures:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe conducted an ROC analysis to illustrate the performance of the candidate biomarker panel using the 5 CpG sites in early pregnancy to predict GDM at OGTT. The 5 CpG marker panel demonstrated a robust AUC of 0.82 (95% CI: 0.77\u0026ndash;0.87), with a sensitivity of 82% and a specificity of 77% \u003cstrong\u003e(Figure 4A).\u003c/strong\u003e Combining these\u0026nbsp;novel\u0026nbsp;DNA methylation signatures with conventional clinical risk predictors and POC HbA1c has increased the AUC of the predicted GDM risk to 0.86\u0026nbsp;(95% CI: 0.81\u0026ndash;0.91), suggesting that the full GDM prediction model at early pregnancy improved the discrimination\u0026nbsp;\u003cstrong\u003e(Figure 4B).\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAssociation of novel candidate biomarkers with glycemic variables at OGTT:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess whether there are any differential associations between the identified\u0026nbsp;candidate biomarker panel of 5 CpGs and glucose levels at OGTT, we tested the individual CpGs with fasting, 60 min and 120 min glucose levels at the time of OGTT.\u0026nbsp;Four CpGs (cg10139015, cg23507676, cg11001216, cg04985016) were positively correlated with fasting glucose levels (p-value\u0026lt;0.05) (\u003cstrong\u003eSupplementary Figure 4)\u003c/strong\u003e. All the 5 CpGs showed a positive correlation with 60-mins (p-value \u0026lt;0.01) (\u003cstrong\u003eSupplementary Figure 5)\u003c/strong\u003e and 120-mins glucose levels (p-value \u0026lt;0.01) (\u003cstrong\u003eSupplementary Figure 6)\u003c/strong\u003e. Interestingly, among the 5 CpGs, cg23507676 showed a higher magnitude of\u0026nbsp;positive relationship with fasting, 60- and 120-minute glucose levels during the OGTT.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional analysis GO and KEGG:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo explore the potential mechanisms involved in GDM, we conducted GO and KEGG analyses on the annotated genes of 527 DMPs. The DMP classification included Biological Processes (BP), Cellular Components (CC), and Molecular Functions (MF). The top ten GO enrichment results indicated a significant predominance in BP, highlighting processes such as nucleotide catabolism, coenzyme A biosynthesis, Ras protein signal transduction, and small GTPase-mediated signaling. In the CC analysis, notable enrichments were observed in complexes such as the GABA-A receptor, early endosome, guanyl nucleotide exchange factor, and chloride channel complexes. The MF analysis revealed enrichments in activities including molecular adaptor, GABA-A receptor, guanyl-nucleotide exchange factor, ion channel, and Ras GTPase binding\u0026nbsp;\u003cstrong\u003e(Figure 5A)\u003c/strong\u003e. The most significant enriched pathways, including morphine addiction, endocytosis, TH17 cell differentiation and phosphatidylinositol signaling pathways, were identified through KEGG analysis for the 527 DMPs\u0026nbsp;\u003cstrong\u003e(Figure 5B).\u003c/strong\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this EWAS study which investigated 850,000 CpGs from peripheral blood samples of a prospective early pregnancy cohort, we show the following novel findings: 527 CpGs were differentially methylated and associated with the development of GDM. We identified a panel of 5 CpGs using modern machine learning techniques as candidate biomarkers with the best discriminatory power for GDM prediction. \u0026nbsp;We developed a clinically efficient comprehensive early GDM prediction model comprising the\u0026nbsp;clinically accessible maternal risk factors, POC HbA1c and novel DNA methylation candidate biomarkers achieving the highest predictive ability for GDM with an AUC of 0.86.\u0026nbsp;Pathway analyses indicated potential involvement of pathways such as endocytosis, phosphatidylinositol signaling, Th17, Th1 and Th2 cell differentiation in GDM. Findings from this large-scale epigenomic study in an Indian pregnancy cohort highlight the potential of early pregnancy novel epigenetic biomarkers and conventional maternal risk factors\u0026nbsp;in predicting GDM, and\u0026nbsp;to our knowledge, this is the first comprehensive study to report this in GDM.\u0026nbsp; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eStudies on GDM prediction models are advancing, with significant improvements seen by including early trimester conventional risk factors, blood-based inflammatory markers, metabolites related to glucose, lipid, and amino acid metabolism, genetic markers, miRNA signatures, extrachromosomal circular DNA, and cell-free DNAs\u0026nbsp;[17, 36\u0026ndash;40]. In our recent STRiDE study, we developed composite risk score models that combined HbA1c with factors such as age, BMI, and FH of diabetes during early pregnancy, showing promising results for GDM prediction\u0026nbsp;[19]. We improve the predictive ability of our approach by the addition of 5 CpGs in this study in a representative STRiDE-India cohort. Unlike many previous studies, we focused on developing prediction models based on clinically accessible maternal risk factors, which are crucial for practical implementation, and extended our analysis by incorporating novel DNA methylation biomarkers identified through EWAS. Our panel consisting of five CpGs, along with these easily accessible clinical risk factors and POC HbA1c demonstrated strong predictive ability for GDM around 10 weeks of gestation with robust sensitivity and specificity. Our GDM prediction model is more accurate than previous ones\u0026nbsp;[21, 22, 24]. \u0026nbsp;These findings are particularly important, as early lifestyle interventions based on accurate predictions could help in reducing the burden of GDM and its complications in Indian women. However, this will require future randomized controlled trials (RCTs) based on our composite risk score model including these CpGs.\u003c/p\u003e\n\u003cp\u003eThe top 5 CpG sites identified in our study were annotated to the BLK, A1BG, TRIP12, CLSTN3, and PSMC3 genes, all of which were implicated in various metabolic disorders. Among the 5 CpGs identified in the present study, the CpG cg23507676 annotated to the BLK had the highest correlation with OGTT fasting, 60- and 120-min glucose levels. This CpG\u0026nbsp;annotated to the\u0026nbsp;BLK gene, has been shown to play a vital role in insulin synthesis and secretion in pancreatic beta-cells by upregulating transcription factors like Pdx1 and Nkx6.1\u0026nbsp;[41]. Mutations, such as p.Ala71Thr compromise this function, reduce insulin output and were linked to BLK MODY and T2DM, although not all studies support BLK as a cause for MODY\u0026nbsp;[42]. Studies also have reported the association of this variant with a higher risk for T2DM and obesity\u0026nbsp;[43].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe CpG\u0026nbsp;cg11001216 annotated to the\u0026nbsp;A1BG protein levels, were shown to be downregulated in exosomes among individuals with obese T2DM and reduced in the inflamed livers of obese mice\u0026nbsp;[44, 45]. \u0026nbsp; A1BG has also been linked to diabetic kidney disease\u0026nbsp;[46, 47], and higher levels were noted in PCOS individuals\u0026nbsp;[48], indicating a complex relationship with metabolic disorders. TRIP12 (cg04985016)\u0026nbsp;plays a vital role in pancreatic cell homeostasis by stabilizing transcription factors like PTF1A, essential for insulin secretion\u0026nbsp;[49]. \u0026nbsp; The CLSTN3 (cg04539775), which encodes calsyntenin-3, is associated with obesity and dysfunction in white adipose tissue (WAT)\u0026nbsp;[50]. The variant rs7296261 in CLSTN3 has been linked to an increased risk of obesity, highlighting its role in metabolic dysregulation\u0026nbsp;[51]. Studies were reported that altered expression of PSMC3 in T2DM impairs pancreatic \u0026beta;-cell function, affecting insulin production and disease progression\u0026nbsp;[52]. While these genes are linked to metabolic disorders, further studies are needed to explore their pathophysiological roles as GDM and T2DM have shared common etiological pathways\u0026nbsp;[53]. Our findings provide proof of concept and would be useful for future mechanistic studies examining the potential role of these DNA Methylation signatures in the etiology of GDM in other ethnic groups. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur KEGG pathway analyses of 527 CpG-annotated genes revealed key pathways related to NK cell-mediated cytotoxicity, phosphatidylinositol signaling, Th17 cell differentiation, and Th1/Th2 differentiation. While NK cells play a critical role in immunity by eliminating pathogens; however, in diabetes, their function declines, especially under hyperglycemia, leading to reduced numbers and activity of NK cells\u0026nbsp;[54]. Studies have reported that in obesity, high-fat diets activate NK cells through increased NCR1 ligands on adipocytes, promoting inflammation and insulin resistance\u0026nbsp;[55, 56]. Studies have demonstrated that Th1/Th2 differentiation plays an important role in shaping the immune response in GDM. Women with GDM show lower Th1 levels and higher Th2, Tregs, and Th17 cells, indicating an immune adaptation that balances inflammation and supports fetal development\u0026nbsp;[57, 58]. These findings highlight the interplay between immune regulation and metabolic disorders. These cellular pathway alterations implicated in GDM, diabetes, insulin resistance, and inflammation were driven by shared underlying biological mechanisms. Our findings suggest that the altered methylation patterns identified in the present study are related to these cellular pathways and it may have a mechanistic role in the development of GDM. However, additional confirmatory and mechanistic studies are warranted to establish their biological relevance.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur study\u0026apos;s strength lies in the comprehensive longitudinal assessment of novel DNA methylation signatures through the EWAS approach along with a combined analysis of clinically accessible maternal risk factors during early pregnancy and in the OGTT visit within a large prospective early pregnancy cohort in India. This approach offered us to perform a comprehensive analysis of GDM prediction and how they are impacted by GDM. Our study is the first to report these findings in GDM.\u0026nbsp;It is also of interest that the panel of DNA methylome biomarkers presented in our study could be tested in interventional studies to demonstrate the molecular benefits of lifestyle modification for GDM. In clinical practice, various methods are commonly used to analyze DNA methylation, including targeted methylation sequencing, DNA methylation arrays, methylation-specific PCR (MSP), pyrosequencing, and methylation-sensitive restriction enzyme (MSRE) assays as predictive, diagnostic, and prognostic tools\u0026nbsp;[59, 60].\u0026nbsp;In this context, our study paves the way for developing in vitro diagnostic kits (IVDs) utilizing the five CpGs for GDM prediction and we propose to try to develop a point of care (POC) kit incorporating these 5 CpGs.\u003c/p\u003e\n\u003cp\u003eHowever, our study also has some limitations. Firstly, due to the study design, the associations found cannot imply causality. Secondly, there is a need for additional studies across ethnically diverse cohorts to validate\u0026nbsp;the clinical usefulness of our discriminating and predictive models in other pregnancy cohorts. Thus, though\u0026nbsp;we used sophisticated ML approaches like RF, SVM and bootstrap validations, validating these prediction models in external cohorts is needed to confirm our findings. We propose to test these in other cohorts which we have already approached for this purpose. Finally,\u0026nbsp;the sample size for this study was modest, although admittedly, most GDM studies tend to be small.\u0026nbsp;Additionally, the functional implications of the identified methylation changes need to be experimentally validated to establish their biological relevance.\u0026nbsp;Given the limited studies on DNA methylation in GDM prediction, our findings, despite these limitations, are significant in providing insights into the design of large-scale longitudinal studies for population-based screening for GDM prediction.\u003c/p\u003e\n\u003cp\u003eIn conclusion, we have identified a panel of DNA methylation signatures consisting of 5 CpGs, which has a robust predictive value for early detection of GDM especially when integrated with clinically accessible maternal risk factors in a large\u0026nbsp;longitudinal pregnancy cohort of 2,115 pregnant women. These findings provide proof of concept for developing a CpG panel for GDM prediction and may form the basis for future diagnostic and preventive strategies for GDM.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements/Funding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDr. Gokulakrishnan Kuppan is a current recipient of a DBT-Wellcome Trust India Alliance Intermediate Clinical \u0026amp; Public health Fellowship (Grant Number\u0026nbsp;IA/CPHI/18/1/503964) and acknowledges the funding received from the\u0026nbsp;DBT-Wellcome Trust India Alliance for this study.\u0026nbsp;STRiDE study was funded by MRC-DBT Newton fund (MRC Newton Fund MR/N006232/1).\u0026nbsp;P. Popova was financially supported by the Ministry of Science and Higher Education of the Russian Federation (Agreement No. 075-15-2022-301). Dr M. Balasubramanyam acknowledges the funding from ICMR and DHR. The authors would like to thank all the study participants and staff of STRiDE study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflict of interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor\u0026rsquo;s contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eKG has conceptualized, written, and taken the lead in completing manuscripts through all stages of preparation and submission. KG, PS, MB, RMA, NT, HG, UR, VR, PP and VM have reviewed the project from the conceptualization stage and have contributed to each version of the manuscript. MD, RMA, and UR are members of the research team who participated in the conduct of the study. CT performed experiments. KS, YW, HA, VU, BB, and SS were involved in data analysis. PS was the principal investigator for the STRiDE study and one of the mentors for KG\u0026rsquo;s fellowship. All authors have contributed to the article critically for intellectual content and have provided final approval of the version to be published.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFinancial Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from the corresponding author, [KG], upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBuchanan TA, Xiang AH, Page KA (2012) Gestational Diabetes Mellitus: Risks and Management during and after Pregnancy. Nat Rev Endocrinol 8(11):639\u0026ndash;649. https://doi.org/10.1038/nrendo.2012.96\u003c/li\u003e\n\u003cli\u003eSaravanan P, Diabetes in Pregnancy Working Group, Maternal Medicine Clinical Study Group, Royal College of Obstetricians and Gynaecologists, UK (2020) Gestational diabetes: opportunities for improving maternal and child health. Lancet Diabetes Endocrinol 8(9):793\u0026ndash;800. https://doi.org/10.1016/S2213-8587(20)30161-3\u003c/li\u003e\n\u003cli\u003eDiaz-Santana MV, O\u0026rsquo;Brien KM, Park Y-MM, Sandler DP, Weinberg CR (2022) Persistence of Risk for Type 2 Diabetes After Gestational Diabetes Mellitus. Diabetes Care 45(4):864\u0026ndash;870. https://doi.org/10.2337/dc21-1430\u003c/li\u003e\n\u003cli\u003eYu Y, Soohoo M, S\u0026oslash;rensen HT, Li J, Arah OA (2022) Gestational Diabetes Mellitus and the Risks of Overall and Type-Specific Cardiovascular Diseases: A Population- and Sibling-Matched Cohort Study. Diabetes Care 45(1):151\u0026ndash;159. https://doi.org/10.2337/dc21-1018\u003c/li\u003e\n\u003cli\u003eKramer CK, Campbell S, Retnakaran R (2019) Gestational diabetes and the risk of cardiovascular disease in women: a systematic review and meta-analysis. Diabetologia 62(6):905\u0026ndash;914. https://doi.org/10.1007/s00125-019-4840-2\u003c/li\u003e\n\u003cli\u003eChu SY, Callaghan WM, Kim SY, et al (2007) Maternal Obesity and Risk of Gestational Diabetes Mellitus. Diabetes Care 30(8):2070\u0026ndash;2076. https://doi.org/10.2337/dc06-2559a\u003c/li\u003e\n\u003cli\u003eXiong X, Saunders LD, Wang FL, Demianczuk NN (2001) Gestational diabetes mellitus: prevalence, risk factors, maternal and infant outcomes. Int J Gynaecol Obstet Off Organ Int Fed Gynaecol Obstet 75(3):221\u0026ndash;228. https://doi.org/10.1016/s0020-7292(01)00496-9\u003c/li\u003e\n\u003cli\u003eLowe WL, Scholtens DM, Lowe LP, et al (2018) Association of Gestational Diabetes With Maternal Disorders of Glucose Metabolism and Childhood Adiposity. JAMA 320(10):1005\u0026ndash;1016. https://doi.org/10.1001/jama.2018.11628\u003c/li\u003e\n\u003cli\u003e(2013) Diagnostic Criteria and Classification of Hyperglycaemia First Detected in Pregnancy. World Health Organization, Geneva\u003c/li\u003e\n\u003cli\u003eAmerican Diabetes Association Professional Practice Committee (2024) 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes-2024. Diabetes Care 47(Suppl 1):S20\u0026ndash;S42. https://doi.org/10.2337/dc24-S002\u003c/li\u003e\n\u003cli\u003eVenkataraman H, Ram U, Craik S, Arungunasekaran A, Seshadri S, Saravanan P (2017) Increased fetal adiposity prior to diagnosis of gestational diabetes in South Asians: more evidence for the \u0026ldquo;thin-fat\u0026rdquo; baby. Diabetologia 60(3):399\u0026ndash;405. https://doi.org/10.1007/s00125-016-4166-2\u003c/li\u003e\n\u003cli\u003eU S, Hr M, Gc S (2016) Accelerated Fetal Growth Prior to Diagnosis of Gestational Diabetes Mellitus: A Prospective Cohort Study of Nulliparous Women. Diabetes Care 39(6). https://doi.org/10.2337/dc16-0160\u003c/li\u003e\n\u003cli\u003eMennickent D, Rodr\u0026iacute;guez A, Far\u0026iacute;as-Jofr\u0026eacute; M, Araya J, Guzm\u0026aacute;n-Guti\u0026eacute;rrez E (2022) Machine learning-based models for gestational diabetes mellitus prediction before 24-28\u0026nbsp;weeks of pregnancy: A review. Artif Intell Med 132:102378. https://doi.org/10.1016/j.artmed.2022.102378\u003c/li\u003e\n\u003cli\u003ePopova PV, Grineva EN, Gerasimov AS, Kravchuk EN, Ryazantseva EM, Shelepova ES (2015) The new combination of risk factors determining a high risk of gestational diabetes mellitus. Minerva Endocrinol 40(4):239\u0026ndash;247\u003c/li\u003e\n\u003cli\u003eParkhi D, Sampathkumar S, Weldeselassie Y, Sukumar N, Saravanan P (2023) Systematic Review of risk score prediction models using maternal characteristics with and without biomarkers for the prediction of GDM. 2023.10.23.23297401\u003c/li\u003e\n\u003cli\u003eYu J, Ren J, Ren Y, et al (2024) Using metabolomics and proteomics to identify the potential urine biomarkers for prediction and diagnosis of gestational diabetes. eBioMedicine 101. https://doi.org/10.1016/j.ebiom.2024.105008\u003c/li\u003e\n\u003cli\u003eRazo-Azamar M, Nambo-Venegas R, Meraz-Cruz N, et al (2023) An early prediction model for gestational diabetes mellitus based on metabolomic biomarkers. Diabetol Metab Syndr 15:116. https://doi.org/10.1186/s13098-023-01098-7\u003c/li\u003e\n\u003cli\u003eDinesen S, El-Faitarouni A, Frisk NLS, S\u0026oslash;rensen AE, Dalgaard LT (2023) Circulating microRNA as Biomarkers for Gestational Diabetes Mellitus\u0026mdash;A Systematic Review and Meta-Analysis. Int J Mol Sci 24(7):6186. https://doi.org/10.3390/ijms24076186\u003c/li\u003e\n\u003cli\u003eSaravanan P, Deepa M, Ahmed Z, et al (2024) Early pregnancy HbA1c as the first screening test for gestational diabetes: results from three prospective cohorts. Lancet Diabetes Endocrinol 0(0). https://doi.org/10.1016/S2213-8587(24)00151-7\u003c/li\u003e\n\u003cli\u003eBenhalima K, Van Crombrugge P, Moyson C, et al (2018) A Modified Two-Step Screening Strategy for Gestational Diabetes Mellitus Based on the 2013 WHO Criteria by Combining the Glucose Challenge Test and Clinical Risk Factors. J Clin Med 7(10):351. https://doi.org/10.3390/jcm7100351\u003c/li\u003e\n\u003cli\u003eLi N, Liu H-Y, Liu S-M (2024) Deciphering DNA Methylation in Gestational Diabetes Mellitus: Epigenetic Regulation and Potential Clinical Applications. Int J Mol Sci 25(17):9361. https://doi.org/10.3390/ijms25179361\u003c/li\u003e\n\u003cli\u003eLinares-Pineda T, Pe\u0026ntilde;a-Montero N, Fragoso-Bargas N, et al (2023) Epigenetic marks associated with gestational diabetes mellitus across two time points during pregnancy. Clin Epigenetics 15(1):110. https://doi.org/10.1186/s13148-023-01523-8\u003c/li\u003e\n\u003cli\u003eDłuski DF, Wolińska E, Skrzypczak M (2021) Epigenetic Changes in Gestational Diabetes Mellitus. Int J Mol Sci 22(14):7649. https://doi.org/10.3390/ijms22147649\u003c/li\u003e\n\u003cli\u003eWang X, Huang J, Zheng Y, et al (2021) Study on the relationship between DNA methylation of target CpG sites in peripheral blood and gestational diabetes during early pregnancy. Sci Rep 11(1):20455. https://doi.org/10.1038/s41598-021-99836-2\u003c/li\u003e\n\u003cli\u003eInternational Association of Diabetes and Pregnancy Study Groups Consensus Panel, Metzger BE, Gabbe SG, et al (2010) International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy. Diabetes Care 33(3):676\u0026ndash;682. https://doi.org/10.2337/dc09-1848\u003c/li\u003e\n\u003cli\u003eThangaraj K, Joshi MB, Reddy AG, Gupta NJ, Chakravarty B, Singh L (2002) CAG repeat expansion in the androgen receptor gene is not associated with male infertility in Indian populations. J Androl 23(6):815\u0026ndash;818\u003c/li\u003e\n\u003cli\u003ePidsley R, Zotenko E, Peters TJ, et al (2016) Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling. Genome Biol 17(1):208. https://doi.org/10.1186/s13059-016-1066-1\u003c/li\u003e\n\u003cli\u003eWu MC, Kuan P-F (2018) A Guide to Illumina BeadChip Data Analysis. Methods Mol Biol Clifton NJ 1708:303\u0026ndash;330. https://doi.org/10.1007/978-1-4939-7481-8_16\u003c/li\u003e\n\u003cli\u003eTian Y, Morris TJ, Webster AP, et al (2017) ChAMP: updated methylation analysis pipeline for Illumina BeadChips. Bioinforma Oxf Engl 33(24):3982\u0026ndash;3984. https://doi.org/10.1093/bioinformatics/btx513\u003c/li\u003e\n\u003cli\u003eHolbrook JD, Huang R-C, Barton SJ, Saffery R, Lillycrop KA (2017) Is cellular heterogeneity merely a confounder to be removed from epigenome-wide association studies? Epigenomics 9(8):1143\u0026ndash;1150. https://doi.org/10.2217/epi-2017-0032\u003c/li\u003e\n\u003cli\u003eDas J, Wadhwa N, Natchu UC, et al (2023) Genome-wide temporal landscaping of DNA methylation in pregnant women delivering at term: a GARBH-InI study. Epigenomics 15(9):543\u0026ndash;556. https://doi.org/10.2217/epi-2023-0145\u003c/li\u003e\n\u003cli\u003eBarton SJ, Melton PE, Titcombe P, et al (2019) In Epigenomic Studies, Including Cell-Type Adjustments in Regression Models Can Introduce Multicollinearity, Resulting in Apparent Reversal of Direction of Association. Front Genet 10:816. https://doi.org/10.3389/fgene.2019.00816\u003c/li\u003e\n\u003cli\u003eTang D, Chen M, Huang X, et al (2023) SRplot: A free online platform for data visualization and graphing. PloS One 18(11):e0294236. https://doi.org/10.1371/journal.pone.0294236\u003c/li\u003e\n\u003cli\u003eYu G, Wang L-G, Han Y, He Q-Y (2012) clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters. OMICS J Integr Biol 16(5):284\u0026ndash;287. https://doi.org/10.1089/omi.2011.0118\u003c/li\u003e\n\u003cli\u003eSharma A, Lee-\u0026Oslash;deg\u0026aring;rd S, Qvigstad E, et al (2022) \u0026beta;-Cell Function, Hepatic Insulin Clearance, and Insulin Sensitivity in South Asian and Nordic Women After Gestational Diabetes Mellitus. Diabetes 71(12):2530\u0026ndash;2538. https://doi.org/10.2337/db22-0622\u003c/li\u003e\n\u003cli\u003eOmazić J, Viljetić B, Ivić V, et al (2021) Early markers of gestational diabetes mellitus: what we know and which way forward? Biochem Medica 31(3):030502. https://doi.org/10.11613/BM.2021.030502\u003c/li\u003e\n\u003cli\u003eWang J, Huang P, Hou F, Hao D, Li W, Jin H (2024) Predicting gestational diabetes mellitus risk at 11\u0026ndash;13 weeks\u0026rsquo; gestation: the role of extrachromosomal circular DNA. Cardiovasc Diabetol 23(1):289. https://doi.org/10.1186/s12933-024-02381-1\u003c/li\u003e\n\u003cli\u003eJuchnicka I, Kuźmicki M, Niemira M, et al (2022) miRNAs as Predictive Factors in Early Diagnosis of Gestational Diabetes Mellitus. Front Endocrinol 13:839344. https://doi.org/10.3389/fendo.2022.839344\u003c/li\u003e\n\u003cli\u003ePopova PV, Klyushina AA, Vasilyeva LB, et al (2021) Association of Common Genetic Risk Variants With Gestational Diabetes Mellitus and Their Role in GDM Prediction. Front Endocrinol 12:628582. https://doi.org/10.3389/fendo.2021.628582\u003c/li\u003e\n\u003cli\u003eTang Z, Wang S, Li X, et al (2024) Longitudinal integrative cell-free DNA analysis in gestational diabetes mellitus. Cell Rep Med 5(8):101660. https://doi.org/10.1016/j.xcrm.2024.101660\u003c/li\u003e\n\u003cli\u003eBorowiec M, Liew CW, Thompson R, et al (2009) Mutations at the BLK locus linked to maturity onset diabetes of the young and beta-cell dysfunction. Proc Natl Acad Sci U S A 106(34):14460\u0026ndash;14465. https://doi.org/10.1073/pnas.0906474106\u003c/li\u003e\n\u003cli\u003eBonnefond A, Yengo L, Philippe J, et al (2013) Reassessment of the putative role of BLK-p.A71T loss-of-function mutation in MODY and type 2 diabetes. Diabetologia 56(3):492\u0026ndash;496. https://doi.org/10.1007/s00125-012-2794-8\u003c/li\u003e\n\u003cli\u003eSecolin R, Gonsales MC, Rocha CS, et al (2021) Exploring a Region on Chromosome 8p23.1 Displaying Positive Selection Signals in Brazilian Admixed Populations: Additional Insights Into Predisposition to Obesity and Related Disorders. Front Genet 12:636542. https://doi.org/10.3389/fgene.2021.636542\u003c/li\u003e\n\u003cli\u003evan Bilsen JHM, van den Brink W, van den Hoek AM, et al (2021) Mechanism-Based Biomarker Prediction for Low-Grade Inflammation in Liver and Adipose Tissue. Front Physiol 12:703370. https://doi.org/10.3389/fphys.2021.703370\u003c/li\u003e\n\u003cli\u003eWang Y, Wu Y, Yang S, Chen Y (2023) Comparison of Plasma Exosome Proteomes Between Obese and Non-Obese Patients with Type 2 Diabetes Mellitus. Diabetes Metab Syndr Obes Targets Ther 16:629\u0026ndash;642. https://doi.org/10.2147/DMSO.S396239\u003c/li\u003e\n\u003cli\u003eAltman J, Bai S, Purohit S, et al (2024) A candidate panel of eight urinary proteins shows potential of early diagnosis and risk assessment for diabetic kidney disease in type 1 diabetes. J Proteomics 300:105167. https://doi.org/10.1016/j.jprot.2024.105167\u003c/li\u003e\n\u003cli\u003eAhluwalia TS, R\u0026ouml;nkk\u0026ouml; TKE, Eickhoff MK, et al (2024) Randomized Trial of SGLT2 Inhibitor Identifies Target Proteins in Diabetic Kidney Disease. Kidney Int Rep 9(2):334\u0026ndash;346. https://doi.org/10.1016/j.ekir.2023.11.020\u003c/li\u003e\n\u003cli\u003eKim Y-S, Gu B-H, Choi B-C, et al (2013) Apolipoprotein A-IV as a novel gene associated with polycystic ovary syndrome. Int J Mol Med 31(3):707\u0026ndash;716. https://doi.org/10.3892/ijmm.2013.1250\u003c/li\u003e\n\u003cli\u003eHanoun N, Fritsch S, Gayet O, et al (2014) The E3 ubiquitin ligase thyroid hormone receptor-interacting protein 12 targets pancreas transcription factor 1a for proteasomal degradation. J Biol Chem 289(51):35593\u0026ndash;35604. https://doi.org/10.1074/jbc.M114.620104\u003c/li\u003e\n\u003cli\u003eBai N, Lu X, Jin L, et al (2022) CLSTN3 gene variant associates with obesity risk and contributes to dysfunction in white adipose tissue. Mol Metab 63:101531. https://doi.org/10.1016/j.molmet.2022.101531\u003c/li\u003e\n\u003cli\u003eLi W, Jiang Q, Chen S, Liu J (2024) Adipose-specific CLSTN3B gene associates with human obesity. Metab Open 22:100269. https://doi.org/10.1016/j.metop.2024.100269\u003c/li\u003e\n\u003cli\u003eMarques ES, Formato E, Liang W, Leonard E, Timme-Laragy AR (2022) Relationships between type 2 diabetes, cell dysfunction, and redox signaling: A meta-analysis of single-cell gene expression of human pancreatic \u0026alpha;- and \u0026beta;-cells. J Diabetes 14(1):34\u0026ndash;51. https://doi.org/10.1111/1753-0407.13236\u003c/li\u003e\n\u003cli\u003eLinares-Pineda TM, Fragoso-Bargas N, Pic\u0026oacute;n MJ, et al (2024) DNA methylation risk score for type 2 diabetes is associated with gestational diabetes. Cardiovasc Diabetol 23(1):68. https://doi.org/10.1186/s12933-024-02151-z\u003c/li\u003e\n\u003cli\u003eKim JH, Park K, Lee SB, et al (2019) Relationship between natural killer cell activity and glucose control in patients with type\u0026nbsp;2 diabetes and prediabetes. J Diabetes Investig 10(5):1223\u0026ndash;1228. https://doi.org/10.1111/jdi.13002\u003c/li\u003e\n\u003cli\u003eSpielmann J, Naujoks W, Emde M, et al (2020) High-Fat Diet and Feeding Regime Impairs Number, Phenotype, and Cytotoxicity of Natural Killer Cells in C57BL/6 Mice. Front Nutr 7:585693. https://doi.org/10.3389/fnut.2020.585693\u003c/li\u003e\n\u003cli\u003eLi Y, Wang F, Imani S, Tao L, Deng Y, Cai Y (2021) Natural Killer Cells: Friend or Foe in Metabolic Diseases? Front Immunol 12:614429. https://doi.org/10.3389/fimmu.2021.614429\u003c/li\u003e\n\u003cli\u003eSIFNAIOS E, MASTORAKOS G, PSARRA K, et al (2019) Gestational Diabetes and T-cell (Th1/Th2/Th17/Treg) Immune Profile. In Vivo 33(1):31\u0026ndash;40. https://doi.org/10.21873/invivo.11435\u003c/li\u003e\n\u003cli\u003eAt\u0026egrave;gbo J-M, Grissa O, Yessoufou A, et al (2006) Modulation of adipokines and cytokines in gestational diabetes and macrosomia. J Clin Endocrinol Metab 91(10):4137\u0026ndash;4143. https://doi.org/10.1210/jc.2006-0980\u003c/li\u003e\n\u003cli\u003eWagner W (2022) How to Translate DNA Methylation Biomarkers Into Clinical Practice. Front Cell Dev Biol 10:854797. https://doi.org/10.3389/fcell.2022.854797\u003c/li\u003e\n\u003cli\u003eLocke WJ, Guanzon D, Ma C, et al (2019) DNA Methylation Cancer Biomarkers: Translation to the Clinic. Front Genet 10:1150. https://doi.org/10.3389/fgene.2019.01150\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1: \u0026nbsp;Baseline characteristics at different visits of pregnancy\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"747\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVariables\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAll STRiDE\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eparticipants\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=2115)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAll (Nested case -control) participants\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=258)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNGT\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=147)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGDM\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=111)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\" valign=\"top\" style=\"width: 86.0776%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEarly Pregnancy (V1)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eAge (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e27\u0026nbsp;\u0026plusmn;\u0026nbsp;4.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e27 \u0026plusmn; 3.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e26 \u0026plusmn; 3.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e27 \u0026plusmn; 3.9\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eGestational Age (weeks) at inclusion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e10.5\u0026nbsp;\u0026plusmn; 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e10.4 \u0026plusmn; 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e10.2 \u0026plusmn; 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e10.6 \u0026plusmn; 3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003ePre-pregnancy weight (kg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e59.2\u0026nbsp;\u0026plusmn;\u0026nbsp;12.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e60 \u0026plusmn; 11.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e60 \u0026plusmn; 10.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e60 \u0026plusmn; 13.6\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eBody mass index (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e24.2\u0026nbsp;\u0026plusmn;\u0026nbsp;4.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e24.5 \u0026plusmn; 4.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e24.3 \u0026plusmn; 3.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e24.7 \u0026plusmn; 4.6\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eWaist Circumference (cm)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e86\u0026nbsp;\u0026plusmn;\u0026nbsp;10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e85.9 \u0026plusmn; 8.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e85.0 \u0026plusmn; 8.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e86.9 \u0026plusmn; 8.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eSystolic Blood Pressure (mm Hg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e102\u0026nbsp;\u0026plusmn;\u0026nbsp;10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e102 \u0026plusmn; 10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e101 \u0026plusmn; 10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e103 \u0026plusmn; 10*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eDiastolic Blood Pressure (mm Hg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e69\u0026nbsp;\u0026plusmn; 9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e68 \u0026plusmn; 7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e67 \u0026plusmn; 8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e69 \u0026plusmn; 7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eFasting plasma glucose (mg/dl)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e83\u0026nbsp;\u0026plusmn; 5.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e84 \u0026plusmn; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e83 \u0026plusmn; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e84 \u0026plusmn; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eHbA\u003csub\u003e1c\u0026nbsp;\u003c/sub\u003e(Venous, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e5.08\u0026nbsp;\u0026plusmn;\u0026nbsp;0\u0026middot;3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e5.09 \u0026plusmn; 0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e5.05 \u0026plusmn; 0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e5.14 \u0026plusmn; 0.3*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eHbA\u003csub\u003e1c\u0026nbsp;\u003c/sub\u003e(Point of care, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e5.15\u0026nbsp;\u0026plusmn;\u0026nbsp;0\u0026middot;5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e5.15\u0026nbsp;\u0026plusmn;\u0026nbsp;0\u0026middot;5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e5.13\u0026nbsp;\u0026plusmn;\u0026nbsp;0\u0026middot;4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e5.18\u0026nbsp;\u0026plusmn;\u0026nbsp;0\u0026middot;5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eFamily history of type 2 diabetes n [%]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e860 (40.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e148 (57.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e76 (51.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e67 (60.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" valign=\"top\" style=\"width: 100%;\"\u003e\n \u003cp\u003eSocio Economic Status (SES) n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eLower Class\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e499 (23.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e21 (8.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e7 (4.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e14 (12.6)\u0026nbsp;*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eMiddle Class\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e1177 (55.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e155 (60.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e87 (59.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e68 (61.3) *\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eUpper Class\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e439 (20.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e82 (31.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e53 (36.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e29 (26.1) *\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\" valign=\"top\" style=\"width: 86.0776%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOGTT Visit (V2)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eBody mass index (kg/m\u003csup\u003e2\u003c/sup\u003e)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e24.2 \u0026plusmn; 4.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e26.4 \u0026plusmn; 3.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e26.0 \u0026plusmn; 3.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e26.9 \u0026plusmn; 4.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eSystolic Blood Pressure (mm Hg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e102 \u0026plusmn; 10.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e104 \u0026plusmn; 11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e102 \u0026plusmn; 9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e105 \u0026plusmn; 12*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eDiastolic Blood Pressure (mm Hg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e68.8 \u0026plusmn; 9.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e68 \u0026plusmn; 8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e67 \u0026plusmn; 8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e69 \u0026plusmn; 8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eOGTT Fasting plasma glucose (mg/dl)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e81.5\u0026nbsp;\u0026plusmn; 8.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e83 \u0026plusmn; 7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e80 \u0026plusmn; 5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e86 \u0026plusmn; 7**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eOGTT 1hr Venous plasma glucose (mg/dl)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e137 \u0026plusmn; 30.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e150 \u0026plusmn; 33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e129 \u0026plusmn; 21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e177 \u0026plusmn; 27**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eOGTT 2hr Venous plasma glucose (mg/dl)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e118 \u0026plusmn; 26.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e131 \u0026plusmn; 30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e113 \u0026plusmn; 18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e154 \u0026plusmn; 27**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36.6801%;\"\u003e\n \u003cp\u003eHbA\u003csub\u003e1c\u0026nbsp;\u003c/sub\u003e(Venous, %)\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 16.4659%;\"\u003e\n \u003cp\u003e5.08 \u0026plusmn; 0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20.2142%;\"\u003e\n \u003cp\u003e4.86 \u0026plusmn; 0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12.7175%;\"\u003e\n \u003cp\u003e4.78 \u0026plusmn; 0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13.9224%;\"\u003e\n \u003cp\u003e4.97 \u0026plusmn; 0.4**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n"},{"header":"Graphical Abstract","content":"\u003cp\u003eGraphical Abstract is not available with this version.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[{"identity":"72e4d27d-1912-49cc-bf6f-c9992a855172","identifier":"10.13039/501100009053","name":"The Wellcome Trust DBT India Alliance","awardNumber":"IA/CPHI/18/1/503964","order_by":0},{"identity":"f47e3d86-9c62-4106-bb5f-05c0317259e5","identifier":"10.13039/100010897","name":"Newton Fund","awardNumber":"MR/N006232/1","order_by":1}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"National Institute of Mental Health and Neurosciences","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"DNA Methylation biomarkers, gestational diabetes mellitus, pregnancy, longitudinal study, first trimester, Asian Indians","lastPublishedDoi":"10.21203/rs.3.rs-5280336/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5280336/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGestational diabetes mellitus (GDM) poses significant health risks for both mothers and offspring, necessitating early detection strategies. We aimed to identify specific DNA methylation signatures in early pregnancy to enhance GDM prediction. We quantified the Epigenome-wide DNA methylation profiles in 258 individuals with normal fasting glucose at early pregnancy (\u0026lt; 16 weeks), of whom 43% (n=111) developed GDM. We identified a panel of five CpGs (cg10139015, cg23507676, cg11001216, cg04539775, cg04985016) that predicted GDM with an area under the curve (AUC) of 0.82 (sensitivity: 82% and specificity: 77%). Combining these CpGs with maternal risk factors (age, waist-height ratio, family history of diabetes, HbA1c, and blood pressure) achieved the highest predictive ability (AUC = 0.86). Our findings suggest that a panel of 5 CpGs exhibits robust predictive value, highlighting the feasibility of creating a CpG panel to predict GDM as early as the first trimester. This has the potential to enhance diagnostic and preventive strategies for GDM.\u003c/p\u003e","manuscriptTitle":"First-trimester DNA Methylome profiling identifies novel predictors for gestational diabetes mellitus in Indian women","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-10-18 17:23:26","doi":"10.21203/rs.3.rs-5280336/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1f55ed97-e9f8-4dc5-a86b-87b9af1285a9","owner":[],"postedDate":"October 18th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":39066673,"name":"Endocrinology \u0026 Metabolism"},{"id":39066674,"name":"Epigenetics \u0026 Genomics"}],"tags":[],"updatedAt":"2024-10-18T17:23:26+00:00","versionOfRecord":[],"versionCreatedAt":"2024-10-18 17:23:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5280336","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5280336","identity":"rs-5280336","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}