Associations Between Maternal Microbiome and Incidence of Low-Birth Weight in Four Low- and Middle-Income Countries

OBJECTIVES: Recent studies suggest that maternal inflammation in pregnancy may be associated with poor obstetric and perinatal outcomes. The goal of this study was to examine relationships between maternal microbiome (taxonomic and functional abundance), maternal anthropometry and markers of inflammation on birth weight and incidence of low birth weight (<2500 g) in resource-limited settings. METHODS: This was secondary analysis from the Women First trial conducted in semi-rural regions of Guatemala, Pakistan, India and the DRC. Maternal weight was measured at 12 and 34 wk of pregnancy. Infant measures were collected within 48 h of delivery. Fecal samples at 12 and 34 wk were used for microbiome analysis (16S rRNA gene amplicon sequencing) and fecal calprotectin and myeloperoxidase assays. Sample sizes for microbiome data at 34 wk were n = 96–249 depending on the site. Linear mixed models using the MaAslin2 package were utilized to assess changes in microbiome associated with birth weight or LBW. Predictive models using gradient boosted machines (XGBoost) were developed using the H2o.ai engine. RESULTS: Microbiome composition (β-diversity) was significantly different between sites and between 12 and 34 wk pregnancy within each site (p < .0.05). No differences in β-diversity were observed at either time point between mothers with LBW infants relative to non-LBW infants. Shannon diversity at 12 wk in Pakistan was lower in mothers with LBW infants. Notable differences in genus-level abundance between LBW and non-LBW mothers (p < 0.05) were observed in Klebsiella and Rothia (Pakistan); Phascolarctobacterium (India); Howardella (DRC); and Bacteroides (Guatemala). However, no overlap in differentially abundant taxa was observed between sites suggesting distinct alterations associated with birth weight. Classification models to predict LBW among all sites including maternal anthropometry, inflammation and predicted microbial functions showed moderate performance (72% accuracy, 77% sensitivity and 47% specificity on 20% test data). CONCLUSIONS: Collectively, the findings indicate population-specific alterations in the maternal microbiome associated with infant LBW. A more granular analysis of microbiome, metabolome and maternal anthropometry has the potential for more robust prediction. FUNDING SOURCES: Bill & Melinda Gates Foundation, NIH/NICHD/ODS.