Leveraging an integrated single cell atlas to understand perinatal bone marrow development

Abstract The perinatal period is an emerging yet understudied stage of hematopoietic ontogeny. It has been shown that the murine bone marrow does not become the site of substantive hematopoietic output until after birth, in part, due to a paucity of hematopoietic niche supporting mesenchymal stroma cells. Despite this, few studies have examined the mechanistic underpinnings of hematopoietic niche emergence in the perinatal period. To address this, we generated a novel single cell atlas of mouse bone marrow development containing single cell RNA sequencing (scRNAseq) datasets derived from late gestation fetal (E16.5) to aged adult (18 months) mice. This atlas consists of 16 published and 2 unpublished (P10 and 18 months) datasets for a total of 14 unique developmental timepoints and over 150,000 cells of hematopoietic and stromal origin. We performed dataset integration using Harmony, anchor-based canonical correlation analysis (CCA), and scVI methods. Following dataset integration, cell clustering revealed 52 distinct clusters. Because CXCL12-abundant reticular (CAR) cells are derived from mesenchymal progenitors, we focused our subsequent computational analyses on these 23 clusters. Our atlas resolved previously described mesenchymal progenitor populations including early (EMP) and late (LMP) mesenchymal progenitors, 5 different CAR cell clusters, and a novel fetal-perinatal progenitor population that expresses fibroblast and mesenchymal stem cell markers such as Cd34, Col4a1, Dlk1, Cd248, and Nid1. We observed few CAR-committed progenitors and CAR cells in the fetal-perinatal marrow, which expanded in the postnatal marrow. Hematopoietic niche supporting factors including Kitl, Cxcl12, and Angpt1 were not significantly expressed until P4 whereas Ptn was more highly expressed in fetal-perinatal marrow. Previous lineage tracing studies demonstrated that Fgfr3 and Dlx5 expressing perichondral cells become CAR cells. We identify Fgfr3+ and Dlx5+ mesenchymal cells in our atlas, but these were distinct from other progenitor populations (e.g. EMP, LMP). Subsequent trajectory analyses suggest that these represent separate mesenchymal trajectories of CAR cell development. Indeed, re-analysis of scRNAseq data from early embryonic hindlimb samples suggests that EMP and LMP do not arise until late gestation in the murine bone marrow. After characterizing the developmental heterogeneity of bone marrow stroma, we sought to define drivers of the fetal to postnatal transition using differential gene expression and gene set enrichment analyses (GSEA) coupled with inferred transcription factor activities using SCENIC. Mesenchymal progenitors in the fetal-perinatal period possessed gene signatures involved in embryonic processes such as the epithelial-mesenchymal transition whereas postnatal progenitors were enriched for genes associated with adipogenesis indicating that an adipogenic-primed progenitor program is restricted in fetal life. Further, mesenchymal progenitors from aged bone marrow exhibited a significant adipogenic bias. At birth and during the early postnatal period, significant changes to cellular metabolism, oxygen tension, and exposures sterile inflammatory stimuli occur. Gene expression associated with hypoxia, interferon-γ, and lipid metabolism were enriched in the postnatal marrow. Preliminary data with Seahorse assays suggest that there are developmental differences in bone marrow metabolism with fetal stroma possessing higher rates of glycolytic activity. These divergent transcriptional programs were associated with differences in transcription factor activities. For example, hypoxia inducible factor (Hif1a, Epas1), Ebf3, Foxc1, Vdr activity was increased in postnatal marrow whereas Ebf2 and thyroid receptor-α (Thra) activity was increased in fetal-perinatal marrow. Together, our novel single cell atlas represents a powerful tool to define developmentally regulated phenomena within the bone marrow microenvironment. This dataset provides further evidence that the perinatal period is a distinct developmental stage of hematopoietic ontogeny characterized by dynamic changes to CAR cells and their progenitors. We also computationally identify hypoxia signaling, metabolic shifts, and sterile inflammation as potential drivers of hematopoietic niche emergence in the perinatal period. The extent to which these contribute to the observed differences in transcriptomes of fetal and postnatal stroma requires additional study.