PGY-4 Indiana University School of Medicine INDIANAPOLIS, Indiana, United States
Background: The incidence of gestational diabetes mellitus (GDM) has steadily increased in the US. Currently, GDM is diagnosed using the standard oral glucose tolerance test at 24-28wk, well after onset of maternal hyperglycemia. This delay in diagnosis highlights the need for early biomarkers that can predict onset of GDM. Our goal was to identify specific proteins and microRNAs (miRNAs), which are small non-coding RNA with diverse biological functions, in plasma collected during first trimester that can serve as biomarkers for GDM. Objective: We hypothesize that there are differences in the plasma miRNAs/proteins early in pregnancy ( < 20 weeks gestation) in individuals who later developed GDM compared to those who did not, and these targets can be used in combination with clinical characteristics to predict onset of GDM. Design/Methods: We obtained 116 unique plasma samples (39 GDM; 77 non-diabetic (ND)) from the Hoosier Moms Cohort, a prospective, longitudinal pregnancy cohort (NCT03696368). Plasma proteomics were measured using Olink Explore 3072 (Uppsala, Sweden) and processed as recommended by service vendor. ANOVA F-testing was used to identify differentially expressed (DE) proteins. miRNA sequencing was performed using the extraction-free HTG EdgeSeq System. DE miRNA was defined using both DESeq2 and Limma-Voom to decrease type-1 error rates. Logistic regression models were constructed with DE targets and clinically important variables (fetal sex, maternal BMI, maternal hypertension). Predictive model was plotted along a receiver operating characteristic (ROC) curve. This proposal is supported by K08HD109636. Results: Macrophage receptor with collagenous structure (MARCO), an extracellular membrane receptor that is implicated in internalization of exosomes, was the only protein with significant negative correlation with GDM (OR 0.01, p< 0.001). For miRNA, 12 common targets were identified by both DESeq2 and Limma-Voom, and three were removed due to collinearity. MARCO protein, miRNA mir-664b-3p, and mir-18b-3p remained consistently predictive of GDM (p-value < 0.001) after adjusting for covariates. These targets had good diagnostic performance (AUC 0.878) and improved with addition of clinical covariates (fetal sex, maternal BMI, maternal hypertension) (AUC 0.900).
Conclusion(s): Our study indicates that miRNA and protein markers can be incorporated into predictive models for early detection of GDM. Our future steps will focus on refining and applying the predictive miRNA/protein model to different populations to assess generalizability, including the utility of miRNA in predicting neonatal outcomes.
