Multi-omics Analysis of Umbilical Cord Hematopoietic Stem Cells from a Multi-ethnic Cohort of Hawaii Reveals the Transgenerational Effect of Maternal Pre-Pregnancy Obesity

Overview of the maternal pre-pregnancy cohort with baby cord blood

In this study, baby cord blood samples from 72 pregnant women (34 obese; 38 non-obese) who delivered at Kapiolani Medical Center for Women and Children in Honolulu, Hawaii (2015-2018) were collected. The study was approved by the Western IRB (WIRB Protocol #20151223). Patients meeting the inclusion criteria were identified from pre-admission medical records with pre-pregnancy BMI > 30.0 (maternal obesity) or 18.5-25.0 (normal weight). All participants involved in this study provided written informed consent before the collection of cord blood samples. Pregnant women undergoing elected C-sections at >37 weeks gestation were included, in order to minimize confounding events during the labor. Patient exclusion criteria included pregnant women with preterm rupture of membranes, labor, multiple gestations, pregestational diabetes, hypertensive disorders, cigarette smokers, infection of human immunodeficiency virus or hepatitis B virus, and chronic drug use. Demographic and phenotypic information was recorded, including maternal and paternal age, ethnicity, gestational weight gain, gestational age, parity, and gravidity. For newborns, Apgar scores were documented at 1 minute and 5 minutes post-delivery. The Apgar score serves as a comprehensive assessment of a newborn’s health, with a normal range considered to be between 7 to 10 ¹⁴.

Sample preparation and methylation profiling

The baby cord blood sample was collected in the operating room under sterile conditions at the time of the C-section (Pall Medical Cord Blood Collection Kit containing 25ml citrate phosphate dextrose). The umbilical cord was first cleansed with chlorhexidine swabs before cord blood collection. The total volume of collected blood was measured and recorded before aliquoting to conical tubes for centrifugation. The tubes were centrifuged at 200g for 10 min, and plasma was collected. The plasma volume was replaced with 2% FBS/PBS. Negative selection reagents were added to the blood and incubated for 20 min (Miltenyi Biotec, Auburn, CA). The cord blood was diluted with an equal volume of 2% FBS/PBS. A 20ml aliquot of the diluted blood was layered over a density gradient (15ml Lymphoprep) and centrifuged at 1200g for 20 min. The top layer containing an enriched population of stem cells was collected, centrifuged at 300g for 8 min, and then washed in 2% FBS/PBS. Red blood cells were lysed using ammonium chloride (9:1) with incubation on ice for 10 min, washed twice, and then resuspended in 100µl of 2% FBS/PBS. Cells were stained with Lineage FITC and CD34 APC for 45 min on ice, washed twice, and then sorted using the BD FacsAria III. Hematopoietic stem cells (CD34+/CD38-/Lin-) were collected and stored at -80°C until DNA/RNA extraction.

DNA and RNA were extracted simultaneously using the AllPrep DNA/RNA extraction kit (Qiagen). DNA purity and concentration were quantified in Nanodrop 2000 and Picogreen assay. Bisulfite conversion of 500 ng DNA was performed using the EZ DNA Methylation kit (Zymo), followed by sample processing for Infinium HumanMethylation450 bead chips (Illumina) according to the manufacturer’s instructions. Bead chips were analyzed at the Genomics Shared Resource at the University of Hawaii Cancer Center.

Bulk RNA sequencing

A total of 50 RNA samples were prepared for bulk RNA Sequencing. RNA concentration and RIN score were assayed using Nanodrop 2000 and Agilent Bioanalyzer. A total of 200 ng of high-quality RNA (RIN>7) was subjected to library construction (polyA) and sequenced on HS4000 (2x100) at the Yale Center for Genome Analysis, Connecticut to obtain at least 25M paired reads per sample.

Methylation data pre-processing

R version 3.6.3 was used for all downstream analyses. Raw intensity data (.idat) were extracted using the ‘ChAMP’ package (version 2.16.2) in R ^15–18. Quality control and processing were performed following the ChAMP pipeline. Background controls were subtracted from the data, and raw data that did not pass detection P-value of 0.05 were removed. For each CpG site, the methylation score was initially calculated as the beta value, a fluorescence intensity ratio between 0 and 1. CpG sites whose probes had known underlying SNPs and association with XY chromosomes were removed from analysis due to potential confounding. After BMIQ normalization ¹⁹, the batch effect due to non-biological technical variation caused by experiment handling was removed using the ComBat function in the ChAMP package, confirmed by the singular value decomposition (SVD) plot. The M-values for differential analysis were transformed from beta-values using lumi (ver 3.1.4) in R ^20–23. A total of 410,765 CpG sites remained for downstream analysis after probe filtering, normalization, and batch removal.

Source of variation analysis and confounding adjustment

To eliminate potential confounding factors of pre-pregnant maternal obesity among the 13 clinical factors, we conducted a source of variation analysis to identify the clinical factors that significantly contribute to the methylation level variation, as done before ^24,25. The variables with F statistics greater than 1 (the error value) were determined as confounders and subjected to confounding adjustment. These factors include the baby’s sex, net weight gain during the pregnancy, maternal age, maternal ethnicity, paternal ethnicity, gravidity, and gestational age. To adjust for confounding, a linear regression model is built using the ‘limma’ package to fit methylation M values of each CpG site, using the confounding factors above. The remaining residuals on the M values were considered to be confounding-adjusted for the subsequent bioinformatics analysis of DNA methylation.

Bioinformatics analysis of differential methylation (DM)

A moderated t-test from the ‘limma’ R package (version 3.42.2) ²⁶ was used for detecting DM CpG sites between healthy controls and cases with M values. The p-values were adjusted for multiple hypotheses testing using Benjamini-Hochberg FDR. CpG sites with FDR <0.05 were considered statistically significant. To minimize the effect of the gestational age, CpG sites located within the gestational-age-related differentially methylated regions (DMR) were removed. A total of 130 DMRs related to gestational age were identified using linear regression analysis performed with bumphunter ²⁷ across eight public datasets including a total of 248 patients.: GSE31781 ²⁸, GSE36829 ²⁹, GSE59274 ^28,30, GSE44667 ³¹, GSE74738 ³², GSE49343 ³³, GSE69502 ³⁴, and GSE98224 ^35,36. The complete list of DMRs was included in Supplementary Table 1. Hypermethylation and hypomethylation states were defined by the values of log2 Fold Change (log₂FC) of M values in cases compared to controls: hypermethylation if bigger than 0, and hypomethylation if less than 0. Corresponding genes and feature locations of these differential CpG sites were annotated using IlluminaHumanMethylation450kanno.ilmn12.hg19 (ver 0.6.0) ³⁷ .

KEGG pathway enrichment analysis

‘gometh’ function from R package “missMethyl” (version 1.26.1) ^38–41 was used for KEGG pathway enrichment ^42–44 with DNA methylation data. DM sites were used for pathway enrichment within five supergroups from KEGG pathways: Metabolism, Genetic Information Processing, Environmental Information Processing, Cellular Processes, and Organismal Systems. Pathways with adjusted p-values less than 0.05 were considered significant. Pathway scores for protein pathways (KEGG: Transcription, Translation, Folding, sorting and degradation) and immune pathways (KEGG: Immune system) were calculated with averaged beta values from the promoter region CpG sites.

Weighted co-expression network analysis

Firstly, we adjusted all beta values with clinical confounders, then summarized the DM CpG sites at the gene level by averaging the beta values in the promoter regions (those in the TSS200 and TSS1500 promoter regions). Next, we transformed adjusted beta values to adjusted M values for the downstream adjacency matrix construction. We used adjusted M values for the weighted gene co-expression network analysis (WGCNA) with R package ‘WGCNA’ (version 1.70-3) ^45,46. The soft threshold for the weighted adjacency matrix with an adjusted R²>0.8 was 7. The topological overlap matrix was constructed for hierarchical clustering. Modules were identified by the dynamic tree-cut algorithm. The networks were exported to Cytoscape with an edge weight greater than 0.03 in each module. The genes with the highest betweenness and degree in the WGCNA network were identified as the hub genes for different modules.

Protein-protein interaction network analysis

For the protein-protein interaction (PPI) network analysis, DM genes are used as the inputs and were mapped on the STRING database (version 10) ⁴⁷. Significantly functionally associated protein pairs were identified using PANDA (Preferential Attachment based common Neighbor Distribution derived Associations) (version 0.9.9) ⁴⁸. KEGG pathways associated with these protein pairs (in terms of genes) were found using PANDA. The bipartite network graph was visualized using Cytoscape (version 3.8.1) ⁴⁹.

Stemness score computation

The stemness score was based on Shannon entropy and scaled plasticity, as proposed previously ⁵⁰. Shannon entropy has been widely applied in developmental biology, particularly in stem cell research ^51–53. The formulas are shown below: N is the total number of CpG sites. CpG is represented by the beta value on each CpG probe. The stemness score was calculated for all samples using all remaining 410,765 CpG sites after the preprocessing. A Wilcoxon rank test was performed between the stemness scores of the healthy and maternally obese groups.

Bulk RNA-seq data processing

The Illumina universal adapter regions of raw RNA-seq data were first trimmed using BBMap (version 38.91) ⁵⁴. All raw sequences passed the quality control using fastqc (version 0.11.8) ⁵⁵. The trimmed .fastq files were aligned by STAR (version 2.7.0f) ⁵⁶ to the human Ensembl genome (Homo_sapiens.GRCh38.dna.primary_assembly.fa) and Ensembl annotation (Homo_sapiens.GRCh38.94.gtf). The gene expression counts were calculated using featureCount⁵⁷ from Subread (ver 1.6.4) ⁵⁸.

Differential expression (DE) of RNA-Seq data

The statistically significant DE genes between healthy controls and maternally obese cases were found adjusting for the same clinical confounders for methylation analysis using the ‘DESeq2’ (version 1.26.0) ⁵⁹ and ‘limma-voom’ function from ‘limma’ package ²⁶. The p-values were adjusted for multiple hypotheses testing using BH adjustment. Genes with adjusted p-values less than 0.05 were considered statistically significant.

Correlation analysis between bulk RNA-seq and methylation data

A subset of 47 patients have done both methylation and RNA-seq assyas. Differential expression analysis was done on the bulk RNA-seq data using the ‘limma’ package, with a threshold of BH adjusted p<0.05 to be differential genes. Pearson correlation coefficients (PCC) were calculated between gene expression and methylation beta values from the promoter regions, among the same patients. As mostly a negative correlation between gene expression and DNA methylation in the promoter region is expected, genes with a high negative correlation (PCC<-0.2) were used for pathway enrichment using TOPPFUN ^60–62. Top genes of interest were selected with the absolute value Fold Change>1.5 in gene expression and gene-methyl correlation <-0.3 for hyper- and hypo-methylated CpG sites.

Metabolomics analysis

Metabolomics data were acquired from a previously published study involving 87 patients in the same cohort from three batches (metabolomics workbench study ID ST001114) ⁶³. Targeted metabolites were generated with LC-MS, and untargeted metabolites were generated with GC-MS. After removal of compounds missing in more than 10% samples, a total of 185 metabolites remained, including 10 amino acids (AA), 40 acylcarnitines (AC), 35 acyl/acyl phosphatidylcholines (PC aa), 38 acyl/alkyl phosphatidylcholines (PC ae) and 62 untargeted metabolites. The raw metabolite data were log transformed, standardized, normalized using variance stabilization normalization (VSN), and batch corrected with ComBat function in sva pacakge⁶⁴. Differential metabolites were identified by limma, with clinical confounders adjustment.

Multi-omics integration on metabolomics, epigenomics, and transcriptomics

A subset of 42 patients have the matched methylation, gene expression, and metabolomics data. We applied multi-omics integration with Data Integration Analysis for Biomarker discovery using Latent cOmponents (DIABLO) implemented in the mixomics package ⁶⁵. DIABLO finds the correlated consensus latent variables among different omics in the supervised manner. Top DIABLO features for each omic were selected based on the loading values. We integrated the pathway level methylation, gene, and metabolite interaction using pathview ⁶⁶.

Overview of study design and cohort characteristics

This study aims to investigate the transgenerational effect of pre-pregnancy maternal obesity on offspring. A total of 72 patients who elected to deliver full-term babies through C-sections were recruited from Kapiolani Medical Center for Women and Children in Honolulu, Hawaii from 2016 to 2018. Among them, 38 deliveries are in the healthy control group and 34 are cases with pre-pregnancy maternal obesity. We excluded natural virginal births, to avoid its potential confounding effect on multi-omics profiles. We also carried out stringent recruitment selection criteria, including matching the mothers’ ages as much as possible, as well as similar net gestational weight gain to minimize its confounding effect over maternal pre-pregnancy maternal obesity. The overall study design is shown in Figure 1. Briefly, upon collecting the blood samples, umbilical cord blood hematopoietic stem cells (uHSCs) were enriched by FACS sorting with CD34+CD35-LIN-markers (see Methods). We extracted DNA and RNA from these uHSCs for Illumina 450k array based DNA methylation and bulk RNA-Seq sequencing respectively. The plasma from these cord blood samples was subject to untargeted metabolomics assays using GC-MS and targeted metabolomics assays using LC–MS ⁶³. Given the rationale that DNA methylation could be the mediator for exerting the transgenerational effect of maternal obesity ^67,68, we carried out multi-omics data integration analysis in the DNA methylation-centric manner.

• Download figure
• Open in new tab

Figure 1. Overview of the study design and analysis.

In the preparation step, cord blood plasma samples are collected for metabolome profiling and stem cell sorting. DNA and RNA extraction assays are performed on the enriched stem cells for the methylation and RNA-seq analyses. Downstream analyses are mainly focused on the methylation data. Bulk RNA-seq data were used for validations for methylation discoveries. (Created with BioRender.com)

The demographic details and clinical information of these patients are summarized in Table 1. The distributions of the most representative variables are shown in Figure 2. Among categorical demographic variables, the distribution of baby sex had no statistical difference between obese and health groups, whereas the ethnicity distributions among mothers and fathers, parity and gravidity are statistically different (P<0.05) between the two groups (Figure 2A-2E). Besides maternal pre-pregnancy BMI, other maternal parameters such as maternal age, gestational week, net weight gain and hemoglobin are also not statistically significantly different between the two groups per study design (Figure 2F-2I, Table 1). While mothers of Asian ethnicity are the majority in the control group, NHPIs account for the majority of the maternal-obese group, revealing the health disparity issue known in the state of Hawaii ⁶⁹. Moreover, the control group has lower parities and gravidities, compared to the cases. Babies born to obese mothers show significantly higher (P<0.05) body weights compared to the control group, as expected ⁷⁰. Other parameters including the baby gender, head circumference, body length, and APGAR score at 5 min after birth are not statistically significantly different between case and control groups (Figure 2J-2M).

• Download figure
• Open in new tab

Figure 2. Mother and newborns statistics of the multi-ethnic cohort from Hawaii.

(A-E) Categorical variables including baby sex, maternal ethnicity, paternal ethnicity, parity and gravidity between control and obese groups are shown in the barplots. P-values using Chi-square test are annotated comparing control and obese groups. (F-I) The distributions of maternal age, gestation age, maternal net weight gain during pregnancy, and maternal hemoglobin between control and obese groups are compared. Mean and standard deviation are shown in boxplot. P-values using t-test are annotated. (J-M) The distributions of baby weight, baby head circumference, baby length, and APGAR score after 5 minutes of delivery between control and obese groups are compared. Mean and standard deviation are shown in boxplot. P-values using t-test are annotated.

Global hypermethylation pattern revealed by CpG level methylation analysis

For scientific rigor, it is critical to adjust for confounding in DNA methylation association analysis ⁷¹. Thus we performed the source of variance (SOV) analysis on the beta values of the DNA methylation with respect to physiological and phenotypic information, in order to assess potential confounding factors systematically ^24,25,71. As shown in Figure 3A, marginal F-statistics in the SOV analysis show that the dominating contribution to DNA methylation variation is maternal pre-pregnancy obesity status, confirming the quality of the study design which aimed to minimize other confounders’ effect. The other minor confounding factors include baby sex, maternal age, maternal ethnicity, net weight gain during pregnancy, paternal ethnicity, gravidity, and gestational age (F-statistics>1). After adjusting these factors by linear regression, all have reduced F-statistics of less than 0.5 (Figure 3B) except maternal pre-pregnancy obesity, confirming the success of confounding removal.

• Download figure
• Open in new tab

Figure 3. DNA methylation analysis on uHSCs.

(A-B) Source of variance plot before and after confounding adjustment. F-statistics are reported for each clinical factor. F statistics greater than 1 are considered to have confounding effects in addition to the case/control difference due to pre-pregnancy maternal obesity. (C) Volcano plot of -log(BH adjusted p-values) against logFC. The cutoff line for adjusted p-value < 0.05 is shown as the red horizontal line. The hyper/hypo threshold is shown as a blue vertical line where logFC=0. Non-significant methylation CpG sites after the differential analysis were shown in gray. Significant CpG sites are colored. After the removal of gestational age-related CpG sites, 10,254 CpG sites are hypermethylated and 5,394 sites are hypomethylated. (D-E) Normalized location distribution of differentially methylated CpG sites according to their CpG features in terms of isle regions and gene regions based on the chip annotation. Isle regions include shelf, shore, island, and open-sea. Gene regions include gene body, intergenic region (IGR), TSS200, TSS1500, 5’UTR, 3’UTR, and 1st Exon.

Next, we conducted differential methylation (DE) analysis on the confounding adjusted DNA methylation data (Methods). We observed a global hypermethylation pattern in pre-pregnancy obese samples, with 10,254 hypermethylated vs. 5394 hypomethylated CpG sites (Figure 3C). The top 20 differentially hypermethylated and hypomethylated CpG sites are reported in Table 2, respectively. These CpG sites are related to a wide variety of biological functions, including inflammation (CD69, ADAM12), transcription factors (ZNF222, HMGN4, LHX6, TAF3), proliferation and apoptosis (HDAC4, DHRS4, LRCH3, SAFB2, CRADD, EBF3, PRKAR1B). Some top DM CpG sites are associated with obesity directly, including HDAC4 ⁷² and PLEC1⁷³.

We further examined the distributions of these differentially methylated sites, relative to the CpG island regions and promoter proximity. (Figure 3D-E). A big fraction (38.7%) of the DM sites are located in CpG islands ^74,75, significantly higher than that from the Illumina 450K annotation (P<2E-16). CpG islands are more frequent in the hypermethylated sites (41.3%) as compared to these in the hypomethylated sites (33.6%), consistent with the global hypermethylation pattern. Relative to gene localization, DM sites are most frequent (32.1%) in the promoter regions (including 16.8% and 15.3% in TSS200 and TSS1500 respectively) as expected.

Functional analyses reveal the association between maternal obesity and cell cycle, immune function and metabolic changes in the cord blood of offspring

To investigate the biological functions related to the epigenome alternation, we conducted systematic analysis of DM sites employing multiple methods: KEGG pathway enrichment analysis, Weighted Gene Co-expression Network Analysis (WGCNA), and Protein-Protein Interaction (PPI) network analysis.

KEGG pathway enrichment analysis on hypermethylated CpG sites identified five significant pathways with FDR<0.05 (Figure 4A), including the cell cycle, ribosome, nucleocytoplasmic transport, ribosome biogenesis in eukaryotes, and mTOR signaling pathway. Cell cycle, ribosome, and nucleocytoplasmic transport pathways are essential to normal cell functioning. mTOR signaling pathway coordinates the nutrient-mediated metabolism, immune responses and cell cycle progression, and dysregulation of mTOR could lead to various diseases such as cancer and obesity ⁷⁶. There was no significantly enriched pathway emerging from hypomethylated CpG sites. The maternally obese group shows significantly higher methylation levels in KEGG protein synthesis and immune system pathway collections compared to the control group, indicating repression in immune response as well as translation and protein synthesis (Figure 4B-C). Similarly, we further explored the differential potential, or stemness, of umbilical cord Hematopoietic Stem Cells (uHSCs). We first confirmed the homogeneity of uHSCs by single-cell RNA sequencing UMAP plot (Supplementary Figure 2). We calculated the cell stemness scores using the DNA methylation beta values similar to others ⁷⁷. uHSCs derived from the maternally obese group exhibit significantly elevated stemness scores (P<0.01) in comparison to the control group (Figure 4D), confirming the results in KEGG pathway analysis.

• Download figure
• Open in new tab

Figure 4. Pathway and network analysis.

(A) KEGG pathway enrichment for hypermethylated CpG sites from promotor region. Enriched KEGG pathway names, adjusted p-values (-log10 transformed), and the size of enriched gene list are reported for CpG sites from TSS200+TSS1500 regions. The red dotted line shows the threshold cutoff for FDR at -log10(0.05). (B-C) Boxplots of averaged beta values for KEGG protein pathway collection and immune pathway collection with Wilcoxon P-values. (D) Violin plots of cell entropy scores between control and obese groups with Wilcoxon P-values. (E-F) WGCNA network analysis results. WGCNA modules are shown for both the control and the obese group. The top two modules with largest degrees are turquoise and brown modules. Each node represents a gene. Genes co-expressed in each module are annotated. (G)Protein-protein interaction (PPI) network. Bipartite graphs represent enriched KEGG pathways and associated genes with significant PPIs. Red nodes represent genes with hypermethylated CpG sites. Blue nodes represent genes with hypomethylated CpG sites. Yellow nodes represented the enriched KEGG pathways. Number of inter-pathway PPIs are annotated in the rectangular boxes.

Next, we applied WGCNA to cluster co-regulation of gene-level methylation, by averaging CpG sites to affiliated genes (see Methods). Five co-expression modules are identified, using the M-values adjusted for clinical confounders (Supplementary Figure 1A), and all modules show positive correlations with maternal obesity except one. The largest turquoise module (Figure 4E) is related to cell cycle, protein synthesis, and transport and vesicle trafficking pathways through pathway enrichment analysis. Some hub genes in this module are identified, including INTU, ANAPC7, and AGBL5. These genes were reported essential for maintaining cell polarity (INTU)⁷⁸, proliferation (ANAPC7) ⁷⁹ and glycemic control (AGBL5) ⁸⁰. The brown module (Figure 4F) is enriched with immune response pathways, in which TLR6 is identified as a hub gene. The other yellow module is related to ion homeostasis, and the gray module is related to the p53 pathway, apoptosis, cell senescence, and ER stress (Supplementary Figure 1B). The only negatively correlated blue module is associated with axon guidance and VEGF signaling pathway (Supplementary Figure 1B).

Furthermore, we examined the PPI network, using the gene-level DNA methylation as surrogates (Figure 4G). The PPI analysis identifies 14 unique pathways (FDR < 0.05) predominantly associated with hypermethylated CpG sites in the TSS200 and TSS1500 regions. The top five largest pathways included ribosome, proteasome, cell cycle, axon guidance, RNA polymerase, and neuroactive ligand-receptor interaction. Taken all three types of systematic analyses together, cell cycle, immune function and protein synthesis are ubiquitously highlighted, suggesting that these biological functions in cord blood stem cells are negatively impacted by maternal obesity.

Multi-omics analysis reveals disruptions in cell cycle and metabolic pathways

To systematically investigate the epigenetic, transcriptomic, and metabolomic alterations induced by maternal obesity, we performed multi-omics integration analysis on this cohort. We employed DIABLO, a supervised integration method that extracts features associated with maternal obesity, based on the correlations in the embedding space ⁶⁵. Figure 5A-C shows that methylation data provide the clearest separation between obese and control groups, confirming the value of the earlier DNA methylation centered analysis.

• Download figure
• Open in new tab

Figure 5. Multi-omics integration analysis

(A-C) Omics-specific sample plots from DIABLO showing the separation of obese and control samples in methylation data, gene expression data, and metabolomics data respectively. (D-F) Importance plot of top 25 features in methylation, gene expression and metabolomics modalities with the highest loadings extracted from the embedding space. The color represents the condition which features contribute the most.

The top 25 features from each omic with the highest feature weights (loadings) following integrated canonical correlation analysis are demonstrated in Figure 5D-F. The methylation features with the highest weights related to maternal obesity include CpG sites involved in cell-cycle control, glucose metabolism, and adipogenesis (FOXO1⁸¹), DNA repair (LIG3, SMUG1), erythropoietin pathway and differentiation (EPO, CSNK2A1, CSF1), which are hypermethylated in the obese group. Hypomethylation of LEP (encoding leptin) was also observed as a top feature, aligning with prior findings that maternal obesity is associated with elevated maternal leptin levels, a known marker of adipose tissue ⁸². These featured CpG sites indicate repression in fat metabolism and DNA repair and reduced differentiation potential. In the transcriptomic space, many genes related to mRNA splicing (SRRM1, IGF2BP1, IGF2BP2, CNOT4) have increased expression levels due to maternal obesity. Among the metabolite features, essential sugars (glucose, xylose), poly-unsatuated fatty acids (oleic acids, DHA, arachidonic acid), and phosphatidylcholine (PCs) are mostly decreased in the obese group; whereas most acylcarnitines (C) are elevated. The metabolic changes show an overall accumulation of saturated fatty acid, but repression on fat breakdown, glucose and unsaturated fatty acid generation. As poly-unsatuated fatty acids (eg. arachidonic acid) have important anti-inflammatory effects, the results indicate a pro-inflammatory environment in offspring born of pre-pregnant obese mothers.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work the author(s) used GPT-4.0 in order to improve the readability. After using this tool/service, the author(s) reviewed and edited the content thoroughly and take(s) full responsibility for the content of the publication.