Our results suggest that maternal factors, and most consistently maternal pre-pregnancy BMI, are associated with the microbial composition of mother’s milk. This is the first study to include maternal glucose tolerance status in the investigations of the association between maternal body size and the milk microbiota (Supplementary Table 1). Gestational diabetes, a well-known risk factor of maternal adiposity, is associated with a number of negative health outcomes, including increased risk of type 2 diabetes and metabolic syndrome in the mother, as well as large for gestational age, congenital malformations and hypoglycemia in the infant (25–28). These changes in infant health may be partially mediated by microbes transmitted from mother to infant, from both the birthing process as well as during direct breastfeeding. For this reason, it is imperative to study the relationship between these factors in order to best mitigate these maternal and infant outcomes as well as distinguish the role of body size versus glucose tolerance status on the milk microbiota and, hence, the infant.
According to a recent systematic review and dose-response meta-analysis, the risk of GDM increases by 4% for every unit increase in BMI; thus, to investigate BMI without also adjusting for GDM could lead to erroneous results (28). Our results suggest that these associations extend beyond the gut microbiota and that alterations in glucose tolerance status may also perturb the composition (Tables 2, 4) of mother’s milk microbiota. Previous studies have reported associations between both type 2 diabetes and insulin resistance and the gut microbiota composition (29). Potential mechanisms of action linking impaired glucose tolerance and the gut microbial community include lipopolysaccharide-triggered inflammation, impairment of GLP-1 and GLP-2 via bacterial production of short chain fatty acids, insulin resistance triggered by bacterial synthesis and absorption of branch-chained amino acids, or the metabolism of bile acids by bacteria and their organ-specific effects (29). It is unclear how impaired glucose tolerance may modulate the bacteria in human milk and the interplay present with maternal body size.
We did not find any differences in alpha diversity based on our maternal characteristics; however, we did find statistically significant differences in beta-diversity, with mother’s milk microbiota separating, or non-randomly clustering, based on pre-pregnancy BMI even after adjustment for other covariates (Supplementary Table 4, Supplementary Figs. 1 and 2). The human gut microbiota has been reported to cluster as a function of body size, but this has not yet been reported for the human milk microbiota (30). Our results demonstrating an association between maternal body size and microbial composition is consistent with other smaller scale studies on human milk microbiotas (Supplementary Table 1). Cabrera-Rubio et al. (2012) examined the association between maternal body size and the differential abundance of mother’s milk genera in a study of healthy Finnish women (n = 18) (20). They reported an increase in Staphylococcus in mother’s milk collected from obese women, which mirrors the findings in our study (Table 4).
Mode of delivery was also associated with changes in mother’s milk microbiota at both the phylum and genus levels (Table 2, 4). For example, we observed greater differential abundance of Staphylococcus in mother’s milk from women who underwent a (scheduled) C-section versus vaginal delivery (Table 4). Our results are similar to that reported by Cabrera-Rubio et al. (2012, 2016). These two small cross-sectional cohorts of healthy Finnish women (n = 18, 10) showed a non-statistically significant increase in Staphylococcus in milk observed among women who delivered their infant via a scheduled C-section versus a vaginal delivery (Table 4) (19, 20). The proposed mechanism whereby mode of delivery alters the milk microbiota is via the infant oral cavity, which is colonized during either vaginal delivery or C-section; from here, retrograde inoculation of bacteria can occur from the infant’s oral cavity into the mammary gland via the suckling process with direct breastfeeding (31–34).
The results of our multi-ethnic cohort revealed associations between ethnicity and specific bacterial taxa. Ethnicity and/or geographic location have been shown to be factors in determining various microbiomes of the body including the gut, oral cavity, respiratory tract, skin, and urogenital tract (35). Ethnicity and geographic location typically come with an overlay of dietary variation, making the impact of each variable challenging to separate. Only a few studies to date have assessed associations between ethnicity and the milk microbiota; however, the ethnic/geographic groups differed from our study as they generally examined Europe, Africa and the United States, making it challenging to compare findings (Tables 3, 5; Supplementary Tables 5, 6) (12,21,36,37).
We used a functional inference approach to characterize the microbial genetic potential in human milk. In agreement with what has been reported for other human-associated microbiotas, the functional capacity of the human milk microbiota is more stable than its taxonomic composition (38). We then assessed whether there were specific pathways that were associated with maternal characteristics and found that maternal BMI, specifically the obese sub-category, was significantly associated with an increase in the “Biosynthesis of secondary metabolites” KEGG pathway. Microbes produce secondary metabolites, which are small, bioactive molecules, not necessary for growth or development but are instead involved in microbe-host or microbe-microbe interactions (39,40). Indeed, many of the genes in the biosynthesis of secondary metabolites pathway encode for the biosynthesis of antibiotics (41). Increased production of these secondary metabolites could impact the overall microbial composition/function in these feeding infants. These potential alterations could represent a mechanism by which maternal BMI impacts infant health over both the short- and long-term and warrants future investigation.
Human milk is considered a low biomass sample and, for this reason, may be more affected by sample processing than higher biomass samples, such as stool. To address this concern, we used PCoA plots to visualize clustering, or lack thereof, of our milk samples and negative controls (Supplementary Fig. 5). Our negative controls were seen to cluster away from the samples, suggesting that our results do not arise from technical contaminates, which was confirmed using Adonis analyses to statistically corroborate that our samples clustered away from the negative controls (Weighted UniFrac R2 = 0.07, p = 0.0001; Bray-Curtis R2 = 0.10, p = 0.0001, Supplementary Fig. 5).
Strengths of the current study include its relatively large sample size, the diverse ethnicity of women included, clinical examination via an OGTT, and enrichment of the cohort with women of varying body size who had abnormal glucose tolerance status. These strengths allowed for a more fulsome investigation using multivariable statistics to determine how each maternal factor is independently associated with the milk microbiome. Limitations of the current study include a lack of disinfection of the mother’s breast, peri-areolar region, and/or nipple prior to milk sampling, and single time-point sampling. The microbes identified in the milk from the present study likely include bacteria from the skin microbiota. Practically, however, mothers do not disinfect their breast prior to pumping and storing milk for their infant, nor do they disinfect prior to breastfeeding. Therefore, the mother’s milk microbiota as collected in the current study is likely a more accurate depiction of what the infant would receive. Secondly, we could not adjust for all variables of interest to due sample size constraints, so we are likely missing additional determinants of both the milk microbiota and functional capabilities. Lastly, our study is limited by its cross-sectional design and thus we cannot assess how the milk microbiome changes over time. It is possible that the associations we identified between maternal factors and microbial composition in milk are not transitory and change across the course of lactation.