Human milk oligosaccharide composition is affected by season and parity and associates with infant gut microbiota in a birth mode dependent manner in a Finnish birth cohort

Abstract

Background: Human milk oligosaccharides (HMOs), their determinants, infant gut microbiota and health are under extensive research; however, seldom jointly addressed. Leveraging data from the HELMi birth cohort, we investigated them collectively, considering maternal and infant secretor status.

Methods: HMO composition in breastmilk collected 3 months postpartum (n = 350 mothers) was profiled using high-performance liquid chromatography. Infant gut microbiota taxonomic and functional development was studied at 3, 6, and 12 months (n = 823 stool samples) via shotgun metagenomic sequencing, focusing on HMO metabolism via glycoside hydrolase (GH) analysis. Maternal and infant secretor statuses were identified through phenotyping and genotyping, respectively. Child health, emphasizing allergies and antibiotics as proxies for infectious diseases, was recorded until 2 years.

Findings: Mother's parity, irritable bowel syndrome, gestational diabetes, and season of milk collection associated with HMO composition. Neither maternal nor infant secretor status associated with infant gut microbiota, except for a few taxa linked to individual HMOs. Analysis stratified for birth mode revealed distinct patterns between the infant gut microbiota and HMOs. Child health parameters were not associated to infant or maternal secretor status.

Interpretation: This comprehensive exploration unveils intricate links between secretor genotype, maternal factors, HMO composition, infant microbiota, and child health. Understanding these nuanced relationships is paramount for refining strategies to optimize early life nutrition and its enduring impact on long-term health.

Funding: Sweet Crosstalk EU H2020 MSCA ITN, Academy of Finland, Mary and Georg C. Ehrnrooth Foundation, Päivikki and Sakari Sohlberg Foundation, and Tekes.

Keywords: Child health; Early life exposures; FUT2; Glycoside hydrolases; HMOs; Maternal factors.

Fig. 1

Cohort overview and study design (a) scheme of the sampling strategy and cohort characteristics. (b) Human milk oligosaccharide (HMO) composition and concentrations in mothers' milk, showing the absolute HMO composition across 350 breast milk samples (panel 1) and the average concentrations of HMOs by secretor status (panel 2) expressed in mg/ml.

Fig. 2

Association between technical and maternal factors with HMO composition. Association between (a) technical factors, (b) maternal factors and HMO compositions. Coefficients of association were obtained from linear models on the z-score transformed HMO concentrations, with the milk group included as fixed effect in the models. Associations with FDR adjusted q-values <0.05 were considered statistically significant, with all original unadjusted p-values <0.05. Tile coloring reflects direction and magnitude of coefficient, grey tile represents non-significant association. The other technical and maternal variables considered in this analysis, that showed no significant associations with HMOs concentrations were not plotted (complete list in Supplemental File 1).

Fig. 3

Association between HMO composition and infant gut microbiota taxonomic composition at 3 months. Overview of the infant gut microbiota composition at 3 months colored by (a) FCT or (b) milk groups. PCoA on the species-level taxonomic composition of the fecal samples using Bray–Curtis distance. (c) Association between HMO compositions and the infant gut microbiota at 3 months. Coefficients of association were obtained from linear models on the log transformed species-level taxonomic composition, models including parity, birth mode and IA (maternal intrapartum antibiotics) as fixed effects. Only species reaching a minimum relative abundance of 1% in 10% of the samples are shown. The associations with FDR adjusted q-values <0.15 were considered statistically significant, with original unadjusted p-values <0.05. Tile coloring reflects direction and magnitude of coefficient, grey tile represents non-significant association.

Fig. 4

Association between HMO composition and infant gut microbiota CAZy functional potential at 3 months. (a) Overview of the infant gut microbiota CAZy functional potential at 3 months for CAZy sub-families involved in HMO degradation. Samples are plotted by FCT, and the CAZy counts in cpm were compared using Wilcoxon test. FDR p-values (q-values) are reported as follows: ∗∗∗q-value <0.001, ∗∗q-value <0.01, ∗q-value <0.05. (b) Association between HMO composition and infant gut microbiota functional potential for the total infant cohort at 3 months. Coefficients of association were obtained from linear models on the CAZy sub-families (log transformed gene counts per million reads). Models included parity, birth mode and IA as fixed effects. The associations with FDR adjusted p-values (q-values) < 0.15 were considered statistically significant, with original unadjusted p-values <0.05. Tile coloring reflects direction and magnitude of coefficient, grey tile represents non-significant association. Only CAZy subfamilies with Human milk oligosaccharides and host glycan as predicted substrate were plotted, however, all significant associations are available in Supplemental File 2.

Fig. 5

Association between HMO composition and infant gut microbiota taxonomic composition stratified by delivery mode. (A) Overview of the infant gut microbiota composition at 3 months colored by maternal secretor status. PCoA on the species-level taxonomic composition of the fecal samples using Bray–Curtis distance stratified by delivery mode. (B) Association between HMO compositions and infant gut microbiota stratified by delivery mode. Coefficients of association were obtained from linear models on the log transformed species-level taxonomic composition, including parity as fixed effects. Only species reaching a minimum relative abundance of 1% in 10% of the stratified sample groups are shown. The associations with FDR adjusted p-values (q-values) < 0.15 were considered statistically significant, with original unadjusted p-values <0.05. Tile coloring reflects direction and magnitude of coefficient, grey tile represents non-significant association.