Table 1.
Sample data.
Fig 1.
Single-cell RNA sequencing reveals major cell types in human milk after leukocyte enrichment.
(A) UMAP projection of 23,443 CD45 ⁺ DRAQ5 ⁺ cells isolated from fresh milk samples (100-350 mL) collected from 9 healthy lactating donors at various stages of mature lactation. Cells were sorted following exclusion of debris, doublets, and dead cells, and processed using the 10x Genomics Chromium 3′ GEX V3 platform. Libraries were sequenced on a NovaSeq to a target depth of ≥50,000 reads/cell. Post-alignment with Cell Ranger (v7.1), data were processed using Seurat (v3.1.1) with SCTransform normalization, CCA integration, PCA dimensionality reduction, and graph-based Louvain clustering. (B) Dot plot of canonical lineage marker genes used for cell type annotation. Dot size reflects the percentage of cells expressing each gene within a cluster, and color intensity indicates the average scaled expression.
Fig 2.
High-resolution sub-clustering of myeloid cells in human milk.
(A) UMAP projection of cells classified within the myeloid compartment, reanalyzed as in Fig 1 using PCA, UMAP, and Louvain clustering. Sub-clusters were assigned based on canonical marker expression and validated using supervised classifiers trained on public datasets and deep-learning deconvolution via Unicell. Epithelial-like subpopulations were retained based on dual expression of immune and epithelial markers. (B) Dot plot showing expression of representative genes across identified sub-clusters. Dot size indicates the proportion of expressing cells per cluster; color indicates mean scaled expression.
Fig 3.
Inter-individual variation in myeloid sub-cluster composition across human milk donors.
Stacked bar plot shows the proportional distribution of cells from each donor sample across all myeloid sub-clusters.
Table 2.
Key Myeloid Sub-cluster DEGs*.
Fig 4.
Scaled expression of top DEGs across broad mononuclear myeloid cell types and sub-clusters.
(a) Broad myeloid cell type DEGs. (b) Sub-cluster DEGs. Analysis was performed using the FindMarkers function in Seurat. Genes were considered significantly upregulated in each cluster if positive log2 fold-change (avg_log2FC > 0) and adjusted p-value < 0.05. The top positive DEGs for each sub-cluster were selected and their expression scaled (z-score normalization across all clusters) to visualize relative expression levels. Columns represent annotated sub-clusters; rows represent top DEGs enriched in each cluster. Color intensity corresponds to scaled average expression; genes with higher relative expression in a given sub-cluster appear darker. Sub-cluster of DEG origin is denoted by the color-coded bar at left.
Table 3.
Top Myeloid Sub-cluster Reactome pathways.
Fig 5.
Percent contributions of myeloid sub-clusters to the top 20 Reactome pathways ranked by weighted z-score.
Pathways are ordered from greatest to least weighted z-score and restricted to positive enrichment values. Weighted z-score was determined by multiplying the total additive z-score (restricted to positive contributions) by the number of contributing sub-clusters.
Fig 6.
Correlation analysis of canonical monocytes and macrophage markers with various lipid-handling transcriptomic signatures.
Monocyte (A; S100A9, FCGR1A, ITGAM) and macrophage (B; FCGR3A, CD68, MAFB, CD14) marker genes are shown in red, lipid signatures (INSIG1, SLC7A5, QSOX1 (“ISQ”) for monocytes or APOE, LIPA, APOC1, NCEH1, SOAT1, A2M for macrophages) are shown in blue, and co-expression is shown in fuchsia. A threshold of 0.5 was applied to emphasize cells with concurrent enrichment in both dimensions, highlighting regions where identity and metabolic state overlap in the UMAP space.
Fig 7.
Outgoing signaling patterns inferred from milk-derived leukocyte sub-clusters.