Cohort description: The ‘Intergenerational programming of diabesity in offspring of women with gestational diabetes mellitus’ (InDiaGDM) study investigated the impact of maternal hyperglycemia on the risk of diabesity (diabetes + adiposity) in the offspring and the modifying influence of vitamin B12 and folate status25. Details of inclusion and exclusion criteria are shown in the flowchart (Supplementary figure 1). All participating women signed an informed consent, and the study was approved by the institutional ethics committee (KEM EC no. 1404, SKNMC /Ethics/App/2014/265).
Maternal and Neonatal Data. Maternal socio-demographic data, medical and obstetric information, and standard anthropometric measurements were taken at 28 weeks gestation. A fasting oral glucose tolerance test (OGTT, 75-gram anhydrous glucose, IADPSG26) was performed at that time. Those diagnosed with gestational diabetes mellitus (GDM) were appropriately advised. Delivery details (gestational age and type of delivery (e.g., vaginal, cesarean)) were recorded. Detailed neonatal anthropometry was performed within 24 hours of delivery (i.e., weight, length, head, and abdominal circumferences, triceps, and subscapular skin folds).
Biospecimen collection and determination of blood metabolites:The following measurements were made on maternal and cord blood: hemogram (complete blood count), plasma glucose, insulin, total and HDL cholesterol, triglycerides, total vitamin B12, total homocysteine and folate. Supplementary table 1 describes laboratory methods used for these measurements.
Definitions:Neonatal adiposity (sex-stratified) was defined as highest tertile (T3) of sum of skinfolds (SSF) of neonates born to normal glucose tolerant (NGT) mothers. T1 were classified as lean. Maternal adiposity was defined from sub-scapular skin folds at 28 weeks’ gestation: above median (adipose) and below median (lean). GDM was diagnosed using IADPSG criteria26. Maternal vitamin B12 and folate status were classified as adequate if above and inadequate if below the median concentration for the cohort.
Sample processing randomization and blinding:To avoid unintentional sampling bias, the sample IDs were randomly ordered five times and then mother-neonate pairs were grouped into batches of eight pairs (N=16). Laboratory staff were provided with a blinded (i.e., no group information or metadata) processing order for samples.
ADsEV isolation:Maternal and cord blood plasma samples at delivery were treated with 10μL [611U/mL] Thrombin (System Biosciences, Palo Alto, CA, USA) per 0.5mL plasma. After a 10-minute incubation at room temperature, and centrifugation in a standard microcentrifuge at 12,298 g for 10 minutes, the supernatant was removed and filtered through a 0.2μm cellulose acetate syringe filter (VWR International, Radnor, PA, USA). Total EVs were isolated from filtered plasma supernatants using ExoQuick exosome precipitation solution (System Biosciences, Palo Alto, CA, USA) according to manufacturer’s instructions. ADsEVs were positively selected with a fatty acid-binding protein antibody (FABP4) (0.5 mg/mL; Mouse IgG monoclonal antibody, Abgent cat. No AM2235b, Inc., San Diego, CA, USA) using a magnetic capture technique (EasySep™ "Do-It-Yourself" Positive Selection Kit II - StemCell Technologies, Vancouver, BC, Canada). After positive selection, the ADsEVs were resuspended in TRIzol™ Reagent (ThermoFisher Scientific, Waltham, MA, USA).
MicroRNA isolation and assessment: Exosomal RNA isolation was completed using TRIzol reagent (ThermoFisher Scientific, Waltham, MA USA) according to manufacturer’s instructions. The quantity and quality of the RNA were assessed using a Nanodrop 2000 (ThermoFisher Scientific, Waltham, MA USA) and Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA USA). Samples with OD260/OD280 ratio >1.5 and <2.2 were selected for miRNA profiling. The miRNA labeling was performed using miRNA Complete Labeling and Hyb Kit (Agilent Technologies, Santa Clara, CA, USA) with a sample input of 10uL of total RNA. Samples were hybridized to Agilent Human miRNA Microarray, Release 21.0, 8x60K. The microarray slides were scanned using an Agilent scanner (G2600D) and raw data extracted using Agilent Feature Extraction software (Version 11.5.1.1). The raw microarray data is available online on Gene expression Omnibus with accession ID: GSE217933.
Statistical analyses:Sample size and power calculations were done using “ssize” package in R to measure a reliable fold change of 2.0 in miRNA expression in the adipose group with respect to lean at a significance level of 0.05. A total of 120 individuals (~60 in each group) would provide a power of 80%.
Demographic characteristics: Normally distributed data are presented as mean (SD) and skewed data as median (25th and 75th percentiles). The significance of differences between groups was tested by Student’s t test (if normally distributed) or Mann-Whitney test (if skewed).
MicroRNA data: Raw microRNA expression data was log2 transformed and normalized using the 90th Percentile shift normalization method in GeneSpring GX (Version 14.5) software. MiRNAs that showed true signal in at least 10% of samples were selected for further analysis. To check if the average expression of all ADsEV miRNAs was upregulated or downregulated in the adipose group, mean z-scores of miRNA expression values were calculated for all miRNAs from lean and adipose groups in maternal plasma and cord blood and the distribution of z-scores between the groups were compared. We also investigated if the ADsEV miRNA expression in maternal plasma samples correlated with those from cord blood samples by applying a Spearman correlation test.
The central objective of the study was to compare differential expression of ADsEV miRNAs in maternal plasma and cord blood of adipose (T3) and lean neonates (T1). Differential expression of miRNAs between groups was calculated by Student’s T test for those normally distributed and by Mann-Whitney test for skewed. Fold changes were calculated by formula: FC = average (miRNA expression in adipose – median expression of miRNA in lean). This was done in 2 steps: 1. Without adjusting for confounding variables and 2. with adjusting for confounding variables (maternal GDM status, adiposity, and maternal vitamin B12 and folate status for maternal samples, and additionally neonatal gender for cord blood samples) bymultiple linear regression analysis. We used a discovery significance of p≤0.1 and a FC≥|1.2| and represented it as a volcano plot generated using ggplot225 package in R.
Functional analyses:Differentially expressed miRNAs with p-value≤0.1 were selected for miRNA target gene and functional enrichment using Mienturnet27 software (http://userver.bio.uniroma1.it/apps/mienturnet/; March 2019). Target enrichment was done using miRTarBase 8.0 and the functional enrichment was conducted by using the KEGG database. This tool computationally maps the given list of miRNAs to their target genes/proteins in their reference database (KEGG/miRTarBase) and fetches a list of target genes that are computationally predicted and/or experimentally confirmed. Relevant overrepresented classes of annotated genes with at least 1 miRNA-target interaction were selected for functional enrichment and are studied. MicroRNAs targeting biological pathways belonging to the following KEGG pathway categories are not shown because they are of no direct interest in this analysis: cancer, bacterial/viral/parasitic infectious diseases, excretory, immune and nervous system, immune and neurodegenerative diseases, environmental adaptation, aging, substance abuse, and drug resistance. Pathways with FDR p-value≤0.05 and related to adipogenesis, adipocyte differentiation and adipocyte signaling were considered relevant from the KEGG pathway categories like cell growth and development, signal transduction, metabolism, endocrine and cardiovascular health, digestive systems and studied in detail. Fold changes of each DE miRNA in adipose group are shown in a barplot. MiRNAs and their targeted genes are represented as a network plot, generated using ggnet28 package in R. A bubble plot of the functionally enriched miRNAs was generated using ggplot229, where colour and size of each bubble stands for its FDR p-value and the number of targeted genes, respectively. Adipogenesis related pathways targeted by both maternal and cord blood DE miRNAs are highlighted using a Sankey diagram generated using SankeyMatic30 software.
Differential expression of miRNAs in relation to maternal conditions (hyperglycemia and adiposity during pregnancy):To check the influence of maternal conditions (i.e., hyperglycemia and adiposity during pregnancy) on ADsEV miRNA profile, differential expression analysis was carried out in both maternal and cord blood ADsEV miRNA profiles between following groups: 1. GDM and NGT (unadjusted and adjusted for confounders: maternal adiposity, B12 and folate status) and 2. maternal adiposity (below and above median) (unadjusted and adjusted for confounders: GDM, B12 and folate status). The lists of DE miRNAs from the unadjusted and adjusted analyses were combined and duplicates were removed. DE miRNAs from both analyses were subjected to functional enrichment in Mienturnet and adipogenesis related pathways were screened. A Venn diagram was made using jvenn31 software to compare the adipogenesis related microRNAs from the three analyses: 1. Neonatal adiposity, 2. Maternal GDM, and 3. Maternal adiposity in both maternal and cord blood.