Consistency and Variability of the Human Milk Oligosaccharide Profile in Repeat Pregnancies

Abstract

Human milk oligosaccharides (HMOs) are a set of complex carbohydrates and the third largest solid component of human milk, after lactose and lipids. To date, over 150 HMOs have been identified and the diversity of structures produced by lactating women is influenced by maternal genetics as well as other maternal, infant, and environmental factors. While the concentrations of individual HMOs have been shown to vary between individuals and throughout the course of lactation, the variability of HMO concentration profiles following different pregnancies occurring in the same woman is presently unknown. As such, the objective of this study was to compare HMO concentrations in human milk samples provided by the same women (n = 34) following repeat pregnancies. We leveraged existing human milk samples and metadata from the UC San Diego Human Milk Research Biorepository (HMB) and measured the concentrations of the 19 most abundant HMOs using high-performance liquid chromatography with fluorescence detection (HPLC-FL). By assessing dissimilarities in HMO concentration profiles, as well as concentration trends in individual structures between pregnancies of each participant, we discovered that HMO profiles largely follow a highly personalized and predictable trajectory following different pregnancies irrespective of non-genetic influences. In conclusion, this is the first study to assess the interactions between parity and time following delivery on variations in HMO compositions.

Keywords: HPLC; human milk; human milk oligosaccharides; lactation; parity.

Figure 1

Concentrations (μmol/mL) and relative abundance (%) of 19 HMOs in samples from participants (n = 34) who provided human milk following two different pregnancies. First and subsequent samples were paired for each participant. Participants are numbered 1–34 and grouped by secretor status. Participants are ordered along the x-axis by increasing absolute difference in the times of lactation when samples were provided.

Figure 2

Principal coordinate analysis (PCoA) based on Bray–Curtis dissimilarity measures of the HMO composition of milk samples (n = 68) provided following different pregnancies. Each circle represents the HMO composition of a given sample. Circles with the same color represent samples from the same subject but different pregnancies. Closed circles denote samples provided within a month postpartum, while open circles denote longer than a month.

Figure 3

Bray–Curtis dissimilarity metrics comparing the HMO profiles of milk samples provided by (a,b) the same participant and (c,d) randomized participants with the same secretor status. Dissimilarity metrics are graphed in the context of (a,c) the difference in the time of lactation following birth (t2 − t1) and (b,d) the absolute difference in the time of lactation following birth (|t2 − t1|). Each circle represents a single participant. Black trendlines were fitted with either (a,c) local polynomial regression (locally weighted scatterplot smoothing (loess)) or (b,d) a generalized linear model. The grey areas around the trendlines represent the 95% confidence interval. The strength of the correlation between Bray–Curtis dissimilarity metrics and absolute difference in lactation time was evaluated using Spearman rank correlation coefficient, ρ.

Figure 4

Change in the concentration (μmol/mL) of 19 HMOs, as well as HMO-bound sialic acid (Sia), HMO-bound fucose (Fuc), and total quantified HMOs (SUM), over the course of lactation following first and subsequent deliveries. Trendlines were fitted with a local polynomial regression (locally weighted scatterplot smoothing (loess)). The grey and transparent orange areas around the trendlines represent the 95% confidence interval of the first and subsequent deliveries, respectively.

Figure 5

Percentage (%) change in the concentration of 19 HMOs, as well as HMO-bound sialic acid (Sia), HMO-bound fucose (Fuc), and total quantified HMOs (SUM), from first to subsequent delivery ((t2 − t2)/t1), compared to the difference in time of lactation when samples were collected (t2 − t1). Each circle represents a single participant and is colored by the participant’s secretor status. Black trendlines were fitted with a local polynomial regression (locally weighted scatterplot smoothing (loess)). The grey area around the trendlines represents the 95% confidence interval.

Figure 6

HMOs that were significantly affected by maternal–infant characteristics. (a) Effect on the concentration of FDSLNH was determined using Kendall–Theil Sen Siegel nonparametric linear regression. The trendline was fitted with a local polynomial regression (locally weighted scatterplot smoothing (loess)). The grey area around the trendline represents the 95% confidence interval. The beta (β) coefficient, median absolute deviation (MAD), and p-value (p) are denoted in the figure. (b) Concentration of LNFP3 in milk samples provided following first and subsequent deliveries stratified by the combination of infant sex. Each point represents a milk sample. Lines connect the two samples provided by each participant. Effect on the concentration of LNFP3 was determined using a multivariate regression. The β coefficient, standard error (SE), and p of the regression are denoted above the figure.