ESPE Abstracts

Background: Many studies have associated being born small for gestational age (SGA) [and by implication having suboptimal fetal growth (SFG)] to childhood cardiometabolic risk markers. However, not all growth-restricted pregnancies result in SGA. In the Manchester BabyGRO study, we focussed on pregnancies at risk of SFG with most babies born AGA, and using transcriptomic and metabolomic data we have identified pathways related to higher child systolic blood pressure (SBP). We aim to use these datasets to define a functionally-related network associated with fetal and child growth that predict higher childhood SBP.

Methods: RNA sequencing (for transcriptomics) and nuclear magnetic resonance (for metabolomics) were performed on fasted blood samples from 25 children, aged three-to-seven years with greater SFG-risk identified in fetal life. ∆fetal ([birthweight centile minus 23-week estimated fetal weight centile]/days) and ∆child ([weight centile minus birthweight centile]/years) were calculated. Using rank regression, all genes and then all metabolites were regressed against ∆fetal and then ∆child. We used all detected genes and metabolites to form a hypergraph, allowing investigation of complex relationships between all elements. The hypergraph central clusters for ∆fetal and ∆child analyses represent subsets of significant genes and metabolites sharing the most correlations, and indicating likely functional relationships. Random forest classification (RFC), was used to establish whether these central clusters could predict the highest quartile of childhood SBP, calculating out of bag (OOB) area under the curve (AUC) and error rate. Gene set enrichment analysis (GSEA) was undertaken using EnrichR.

Results: For ∆fetal, the central cluster contained 159 genes and one metabolite, glucose. RFC showed that these predicted the upper quartile of childhood SBP with an OOB AUC of 0.993 and error rate 8.6%. For ∆child, the central cluster contained 180 genes and no metabolites. These predicted the upper quartile of childhood SBP with OOB AUC 1.000 and error rate 2.9%. GSEA for ∆child highlighted integrin family cell surface interactions (P=0.002), integrin-mediated cell adhesion (P=0.013), interleukin-2 signalling and vascular wall cell surface interactions (both P=0.010) of central importance.

Conclusion: A functional network predominantly based on genes related to fetal and child growth predicts higher child SBP. This approach has causally implicated integrins, known to regulate endothelial phenotype and facilitate leucocyte homing, as potential mediators linking early growth to childhood BP. The metabolome was not strongly associated, reflecting a need to investigate samples from other tissues and/or select a more sensitive metabolomic technique.