Maternal plasma lipids are involved in the pathogenesis of preterm birth

Yile Chen, Bing He, Yu Liu, Max T. Aung, Zaira Rosario-Pabón, Carmen M. Vélez-Vega, Akram N. Alshawabkeh, José F. Cordero, John D. Meeker, Lana X. GarmireGigaScience 2022 · DOI

TL;DR

This study identifies specific maternal plasma lipids, particularly fatty acids, as key players in the development of preterm birth. By analyzing a prospective birth cohort, researchers developed a predictive model for preterm birth, highlighting potential new avenues for early detection and intervention.

What Problem Does This Paper Address?

Preterm birth, defined as delivery before 37 weeks of gestation, remains a leading cause of infant mortality and morbidity worldwide. Despite its significant impact, the precise biological mechanisms leading to preterm birth are not fully understood. Previous research has pointed to various risk factors, including socioeconomic status and environmental exposures, but the endogenous molecular pathways involved are still being elucidated. This paper aims to bridge this knowledge gap by investigating the role of maternal plasma metabolites, specifically lipids, in the pathogenesis of preterm birth, seeking to identify potential biomarkers and causal relationships.

Methods

The study employed an untargeted metabolomic approach using liquid chromatography-mass spectrometry (LC-MS/MS) on maternal plasma samples. Key analytical techniques included:

  • Weighted Gene Correlation Network Analysis (WGCNA): Used to identify modules of co-expressed metabolites and their association with preterm birth. This method helps to uncover complex relationships within the metabolome.
  • Machine Learning Classification: Several algorithms (RPART, PAM, GBM, LOG, RF, SVM, LDA) were evaluated to build predictive models for preterm birth using identified metabolites. Random Forest (RF) was selected for its high accuracy.
  • Granger Causality Test: Applied to infer potential causal relationships between specific metabolites and the occurrence of preterm birth.
  • Differential Metabolomics Analysis: Used to identify metabolites with significantly different levels between preterm birth cases and controls, adjusting for potential confounding clinical factors.

Datasets

The study utilized data from the Puerto Rico PROTECT cohort, a prospective birth cohort designed to investigate environmental and social contributors to preterm birth.

  • Samples: Maternal plasma samples were collected from 31 women who experienced preterm birth and 69 controls who delivered at full term. Samples were collected between gestational weeks 24-28.
  • Metabolite Identification: A total of 333 metabolites were identified and annotated.
  • Machine Learning Evaluation: The dataset was split into training (80%) and testing (20%) sets for model evaluation.

Key Insights and Ideas

  • Lipids as Central Players: The study strongly implicates lipids, particularly fatty acids (FAs), in the pathogenesis of preterm birth. This reinforces and expands upon previous findings.
  • FA/CAR-Enriched Module: WGCNA identified a specific module significantly associated with preterm birth, enriched in fatty acids and carene. This suggests a coordinated metabolic pathway disruption.
  • Predictive Biomarker Panel: A panel of 17 metabolites, primarily unsaturated FAs, demonstrated high accuracy (AUC = 0.92) in predicting preterm birth using a Random Forest model. This panel showed specificity for preterm birth over other clinical factors.
  • Causality of Suberic Acid: The Granger causality analysis suggests that decreased levels of suberic acid may lead to an increase in certain fatty acids, potentially promoting preterm birth. This points to a specific metabolic dysregulation.
  • Spontaneous Preterm Birth (sPTB) Signatures: Many of the identified preterm birth markers were also robust predictors of spontaneous preterm birth, suggesting shared underlying mechanisms.

Key Results

  • 333 Metabolites Identified: The untargeted metabolomic analysis identified and annotated 333 metabolites in maternal plasma.
  • 38 Differentially Abundant Metabolites: After controlling for confounders, 38 metabolites showed significant changes uniquely associated with preterm birth.
  • 17 Preterm Biomarkers: A subset of 17 metabolites were identified as strong preterm birth biomarkers.
  • High Predictive Accuracy: The Random Forest model using these 17 biomarkers achieved an AUC of 0.92 for predicting preterm birth in the testing dataset.
  • sPTB Prediction: A model using 16 metabolites predicted spontaneous preterm birth with an AUC of 0.89.
  • Causality: Decreased suberic acid was predicted to causally influence several fatty acids, and upregulated hexanoylcarnitine and certain FAs were predicted as direct causal metabolites of preterm birth.

Limitations and Future Work

  • Dietary Confounding: The study acknowledges that specific dietary records were not collected, which could be a confounding factor.
  • Inferred Mechanisms: Biological mechanisms are inferred from maternal blood measurements, not directly from relevant tissues like the placenta.
  • Validation Needed: The identified biomarker panel is considered suggestive and requires validation in independent cohorts.
  • Future Directions: The authors suggest further investigation into the role of FAs, lipid metabolism, and lipid peroxidation in preterm birth. They also highlight the need to validate the predictive biomarker panel and explore the role of genetic polymorphisms in enzymes like CYP4F/A.

Summary updated

Abstract

Altogether, this study demonstrates the involvement of lipids, particularly fatty acids, in the pathogenesis of preterm birth.
BibTeX
@inproceedings{chen2022maternal,
  author = {Yile Chen and Bing He and Yu Liu and Max T. Aung and Zaira Rosario-Pabón and Carmen M. Vélez-Vega and Akram N. Alshawabkeh and José F. Cordero and John D. Meeker and Lana X. Garmire},
  title = {{Maternal plasma lipids are involved in the pathogenesis of preterm birth}},
  booktitle = {GigaScience},
  year = {2022},
  doi = {10.1093/gigascience/giac004},
}
Figure