The cellular and immunological dynamics of early and transitional human milk

Abstract

Human milk is essential for infant nutrition and immunity, providing protection against infections and other immune-mediated diseases during the lactation period and beyond in later childhood. Milk contains a broad range of bioactive factors such as nutrients, hormones, enzymes, immunoglobulins, growth factors, cytokines, and antimicrobial factors, as well as heterogeneous populations of maternal cells. The soluble and cellular components of milk are dynamic over time to meet the needs of the growing infant. In this study, we utilize systems-approaches to define and characterize 62 analytes of the soluble component, including immunoglobulin isotypes, as well as the cellular component of human milk during the first two weeks postpartum from 36 mothers. We identify soluble immune and growth factors that are dynamic over time and could be utilized to classify milk into different phenotypic groups. We identify 24 distinct populations of both epithelial and immune cells by single-cell transcriptome analysis of 128,016 human milk cells. We found that macrophage populations have shifting inflammatory profiles during the first two weeks of lactation. This analysis provides key insights into the soluble and cellular components of human milk and serves as a substantial resource for future studies of human milk.

Fig. 1. Soluble analyte concentrations in HM vary between individuals and over time.

a Schematic representation of the pipeline used to investigate the soluble and cellular components of HM, n = 60 samples. Created with BioRender.com. b Hierarchical clustered heatmap of the z-scores for 55 soluble immune, growth and signaling analytes from week 1 and week 2 HM samples (n = 34 and n = 26 samples, respectively). Sample IDs are listed under the x-axis for each week.

Fig. 2. Soluble analytes in HM significantly changed over the first 2 weeks postpartum and infant gestational age.

a Dot plots of soluble analytes with significant differences in concentration between week 1 (red) and week 2 (blue) (n = 27 of the 55 soluble analytes, 34 samples from week 1 and 26 samples from week 2). Black lines are representative of the median of all samples. Dots are representative of individual samples. Analytes with no significant changes over the two timepoints are not shown. Soluble analytes are grouped into categories based on major known functions (black box outlines). Total protein concentrations were interpolated from a standard curve. b Dot plot of CD40L showing significant difference in concentration between preterm (red, n = 11) and term (blue, n = 23) at week 1. Black lines are representative of the median of all samples. Dots are representative of individual samples. Analytes with no significant changes over the two timepoints are not shown. Significance was determined with Mann–Whitney tests. *p < 0.05. c Dot plots of CST3, CD44var (v6), TIMP-1, and CD62L showing significant differences in concentration between preterm (red, n = 10) and term (blue, n = 16) at week 2. Significance was determined with Mann–Whitney tests. *P < 0.05; **P < 0.005; ***P < 0.0005; ****P < 0.00005.

Fig. 3. Levels of IgA in HM vary between individuals, and IgG is detectable in HM and has altered FcR binding characteristics compared to IgG from blood.

a Dot plot of IgA measured across all samples and separated by week 1 (red) and week 2 (blue) (n = 34 samples and 26 samples, respectively). Black lines are representative of the median of all samples at respective timepoints. Dots are representative of individual samples. Significance was determined with Wilcoxon–Mann–Whitney tests. **P < 0.005. b Bar graphs of the endpoint titers obtained by ELISA of serum and HM IgG binding to the receptors FcγR1α, FcγR3α, FcγR3b, FcγR2α, and FcγR2β at different concentrations (n = 16 samples and 16 HM samples). Endpoint titers are designated by the most dilute serum concentration detected above the minimum threshold of the background OD450 multiplied by 3. Dots are representative of individual samples. Bars are representative of the mean of all samples.

Fig. 4. HM soluble analyte expression could cluster samples into distinct groups.

a Plot of k-means clustering, using soluble analyte concentrations, show four distinct clusters: Cluster 1 (green, n = 26 samples), Cluster 2 (red, n = 5 samples), Cluster 3 (yellow, n = 28 samples), Cluster 4 (blue, n = 1 sample). b Bar graph of the frequency of total week 1 (red) and week 2 (blue) samples within each cluster (n = 60 samples). c Bar graph of the total frequency of samples from preterm (<37 weeks, pink) and term (>37 weeks, purple) pregnancies within each cluster (n = 60 samples). d Volcano plot of Bonferroni adjusted P values of the soluble analytes contributing to cluster formation, Cluster 1 (green), Cluster 2 (red), and Cluster 3 (yellow) dots are colorized above the −Log10(P-adj) significance threshold. The horizontal dotted line, y = 1.30, represents P = 0.05. (Inset) Heatmap of the top 5 down or upregulated significant analytes in each cluster generated by Log2FC (n = 59 samples).

Fig. 5. Human milk contains a diverse array of maternal cell types.

a UMAP of single-cell transcriptomes derived from HM (n = 128,016 cells). Clusters are labeled numerically (0–23) and by the dominant cell types. b Bar graph of the frequency of total cells within each cluster. Dominant cell types are listed below. c UMAP of all single-cell transcriptomes, highlighting (blue) milk synthesis gene LALBA. d UMAPs of all single-cell transcriptomes, highlighting (blue) lactocyte cells (LC) with LC1 markers (XLF6 and CLDN4) and LC2 markers (XDH and CSN1S1). e UMAPs of all single-cell transcriptomes, highlighting (blue) macrophage marker CD68, monocyte marker FCGR3A, T cell marker CD3D, NK cell marker NKG7, B cell marker CD79A (circled), and neutrophil marker CXCL8. f Heatmap of the z-scores based on gene expression of the top variable genes within each cluster and their main functionality (right).

Fig. 6. Clusters dominated by macrophages and epithelial cells differ between weeks 1 and 2.

a UMAPs of all single-cell transcriptomes, highlighting the cellular distribution from week 1 (red, n = 77,150 cells) and week 2 (blue, n = 50,866 cells) in the aggregate clustering profile. b Bar graph of the average frequency of cells from individuals at each timepoint. Disparate cell types have a higher frequency (>5% different) of cells from one timepoint over the other. The dominant cell type is listed below the bar.

Fig. 7. Macrophages shift from anti- to pro-inflammatory between weeks 1 and 2.

a UMAPs of all single-cell transcriptomes, highlighting of the macrophage-rich clusters with disparities between timepoints (black). b UMAP of macrophage-rich clusters highlighting week 1 only cells expressing macrophage marker CD68 (red, expression >50%). c UMAP of macrophage-rich clusters highlighting week 2 only cells expressing macrophage marker CD68 (blue, expression >50%). d UMAP of macrophage-rich clusters showing week 1 (red, 47%) and week 2 (blue, 39%) cells expressing anti-inflammatory marker TGFβ1. e UMAP of macrophage-rich clusters showing week 1 (red, 49%) and week 2 (blue, 40%) cells expressing anti-inflammatory marker IL1RN. f UMAP of macrophage-rich clusters showing week 1 (red, 8%) and week 2 (blue, 42%) cells expressing pro-inflammatory marker IL1β. g UMAP of macrophage-rich clusters showing week 1 (red, 21%) and week 2 (blue, 54%) cells expressing pro-inflammatory marker CCL4.

Fig. 8. Cellular expression of genes encoding most abundant soluble analytes found in HM.

UMAP plots of the cell expression patterns of markers for genes encoding the most abundant analytes identified from the soluble HM protein analysis are displayed to the left of corresponding bar graphs showing the total frequency of the gene across cell types identified in human milk. The genes encoding soluble analytes are displayed in descending order of soluble abundance in HM.