All single cell RNA-seq data used in this study are publically available, including 2 datasets (GSE135202, GSE162950, 7 samples) for AGM (Zeng et al., 2019; Calvanese et al., 2022), 4 datasets (GSE155259, GSE162950, CRA002443, E-MTAB-7407, 31 samples) for FL (Popescu et al., 2019; Wang et al., 2020; Roy et al., 2021; Calvanese et al., 2022), 3 datasets (GSE155259, HRA002414, E-MTAB-9389, 20 samples) for FBM (Jardine et al., 2021; Roy et al., 2021; Zheng et al., 2022), GSE245108 (16 samples) for young adult bone marrow (YBM) (Zhang et al., 2024), GSE253355 (12 samples) for aged/old adult bone marrow (OBM) (Bandyopadhyay et al., 2024). The detailed information on datasets used in this study can be found in Supplementary Table S1.

The processed raw count matrix of each dataset was downloaded from GEO, ArrayExpress or NGDC. If the count matrix is not available, raw sequencing data were downloaded and reads were mapped to the human reference genomes (refdata-gex-GRCh38-2020-A) using CellRanger (v7.1). All count matrixes were loaded into Seurat (v4.3) (Hao et al., 2021) for downstream analysis. All samples were filtered to retain high-quality cells based on the following criteria: cells with 500–8,000 detected genes, a total UMI count ≤50,000, and a mitochondrial gene expression percentage ≤10%. Samples were grouped into 5 groups: AGM (4–6 weeks), FL (4–19 weeks), FBM (10–19 weeks), YBM (21–34 years old) and OBM (52–74 years old). We integrated samples in each group and removed batch effects using Harmony (v1.2) (Korsunsky et al., 2019). Marker genes were identified using FindAllMarkers with default parameters. Cell types were annotated according to the marker genes (Supplementary Table S2) provided in related literature and the DISCO database (Li et al., 2022).

To investigate the transcriptional dynamics during HSC maturation, we reconstructed the developmental trajectory using Monocle (v2.32) (Qiu et al., 2017). HLF + cells were extracted from the HSPC populations in the AGM, FL, and FBM datasets to focus on the HSC lineage. HLF is highly enriched in and specific to HSCs compared to other hematopoietic populations (Lehnertz et al., 2021; Calvanese et al., 2022). Differentially expressed genes (DEGs) along the HSC maturation trajectory were identified using the differentialGeneTest function in Monocle. Genes with a p-value or q-value ≥0.05 were filtered out to obtain a set of significant DEGs. These DEGs were then clustered and visualized using a heatmap generated by the plot_pseudotime_heatmap function.

CellChat (v2.1, https://github.com/jinworks/CellChat) (Jin et al., 2021; Jin et al., 2024) was employed to infer cell-cell communication via ligand-receptor interactions using annotated Seurat objects containing expression data and cell type annotations from the AGM, FL, FBM, YBM, and OBM. The analysis focused on protein-mediated interactions, with the CellChat database curated to exclude non-protein signaling pathways. Interaction number and strength was calculated for individual ligand-receptor pairs between each pair of cell types. Specifically, for each ligand-receptor pair, CellChat first calculates the communication probability by multiplying the average expression of the ligand in the sender cell type with the average expression of the receptor in the target cell type. To determine the overall interaction strength between two cell types, the interaction strengths of individual ligand-receptor pairs are aggregated. The analysis concentrated on interactions where HSPC were either sending or receiving signals. For each significant ligand-receptor pair, if the ligand is expressed by HSPCs and the receptor is expressed by another cell type, HSPCs are designated as the signal-sending cell in that interaction. Conversely, if the receptor is expressed by HSPCs and the ligand is expressed by another cell type, HSPCs are designated as the signal-receiving cell. This allows for a systematic categorization of HSPC signaling roles based on the expression patterns of ligands and receptors. Comparative analysis of CellChat results was conducted across developmental stages (AGM, FL, and FBM) and the aging continuum (FBM, YBM, and OBM). Changes in interaction strength of specific ligand-receptor pairs, cell type pairs, and signaling pathways were examined across various stages of HSC maturation and aging. The results are visualized using the “ligand-receptor dotplot” and “information flow plot” functions. The LR dotplot displays the communication probabilities of selected ligand-receptor pairs in different stages as a heatmap, allowing for easy identification of stage-specific changes in interaction strength. The information flow for a signaling pathway is calculated by summing up the communication probabilities among all pairs of cell groups in the inferred network. To obtain the relative information flow, CellChat normalizes these raw values by dividing each value by the sum of all information flow values across pathways. This normalization step ensures that the relative information flow values sum up to 1, creating a probability distribution that reflects the relative contribution of each pathway to the overall communication network.

NicheNet (v2.1) (Browaeys et al., 2020) was employed to predict ligands potentially influencing the DEGs between YBM and OBM HSPC using integrated and annotated scRNA-seq data from YBM and OBM. Genes expressed in fewer than 10% of HSPC or fewer than 5% of sender cells within the HSC niche were excluded. The gene set of interest consisted of DEGs between YBM and OBM HSPC, while the background gene set included all genes expressed in OBM HSPC. Ligand activity scores, ligand expression fold change in sender cells, and the regulatory potential of targeted DEGs were used to prioritize and select the regulatory networks of interest.

The Gene Ontology (GO) enrichment analysis was performed using clusterProfiler (v4.6) (Xu et al., 2024). The analysis was conducted separately for upregulated and downregulated DEGs. GO terms with an adjusted p-value (q-value) ≤ 0.05 were considered significantly enriched.

To comprehensively characterize the molecular and cellular changes within the hematopoietic niche during development and aging, we analyzed 11 single-cell RNA sequencing (scRNA-seq) datasets encompassing the main hematopoietic microenvironments. These datasets included samples from the AGM region, FL, FBM, YBM, and OBM. After stringent quality control measures, batch effect correction, and data integration, we constructed five single-cell atlases of the HSC niche, comprising 23,732 cells from AGM, 217,693 cells from FL, 134,261 cells from FBM, 97,614 cells from YBM, and 82,742 cells from OBM (Figures 1A–E). Using markers from literature and databases, we annotated 26 major cell types in these cell atlases, including HSPC, other hematopoietic cells, endothelial cells, MSC, epithelial cells, muscle cells, neural cells, etc. Notably, the proportions of these cell types exhibited marked variations across the five stages (Figure 1F). These changes in niche cell composition likely play a crucial role in regulating HSC function and fate at distinct developmental stages. For example, the exclusive presence of primordial germ cells (PGC) in the AGM dataset suggests that these cells may contribute to the unique properties of the AGM niche in supporting HSC emergence and expansion. PGCs have been reported to secrete factors such as BMP4 and WNT3A, which are known to promote HSC development and self-renewal (Zuo et al., 2021; Esfahani et al., 2024). Similarly, the unique presence of hepatocytes in the FL dataset indicates their specific role in supporting HSC expansion and differentiation during fetal hematopoiesis. Hepatocytes have been shown to produce cytokines like SCF and TPO, which are essential for HSC proliferation and survival (Lee et al., 2021; Lee et al., 2022). To validate the robustness of our cell type annotations, we examined the expression patterns of canonical markers for each identified cell type across the five groups. Remarkably, the expression profiles of these markers remained highly consistent in the corresponding cell types across all developmental and aging stages (Figure 1G).

We reconstructed the developmental trajectory of HSC in the AGM, FL, and FBM regions and compared the communication patterns among these three regions to comprehensively understand the evolving niche interactome and its adaptation to the needs of maturing HSC (Supplementary Figures S1A–C). DEGs that show regular changes along the HSC maturation trajectory can be categorized into upregulated and downregulated genes (Supplementary Figure S1D; Supplementary Table S3). Upregulated genes, including CD48, CD52, CD74, CD83, and transcription factors like JUN, EGR1, and GATA2, play significant roles in facilitating HSC maturation. CD molecules such as CD48 and CD74 are involved in cell adhesion and signaling, crucial for HSC niche interactions and promoting maturation (Boles et al., 2011; Becker-Herman et al., 2021). Upregulated genes are primarily associated with immune response regulation and hematopoietic activation Supplementary Figure S1E; Supplementary Table S4).

To investigate the dynamic niche interactions during HSC ontogeny and maturation, we employed CellChat (Jin et al., 2021) to infer cell-cell communication via ligand-receptor interactions. Ligand-receptor analysis targeting HSPC revealed dynamic cell communication regulation during development and maturation (Figure 2A). We identified numerous specific ligand-receptor pairs in the AGM and FBM microenvironments, with FL serving as an intermediate stage, exhibiting similar ligand-receptor relationships to both AGM and FBM. Aggregating the number and strength of interactions between cells in HSC niches revealed an increase in these interactions during HSC maturation (Figure 2B). Examining the interactions of each cell type targeting HSPC separately, we observed that the interaction strength of all cell types targeting HSPC increased from AGM to FL, while most cell types exhibited increased interaction strength from FL to FBM (Figure 2C).

Mono/Mac-derived SPP1 binding to CD44 on HSPC exhibits the highest interaction strength in AGM but is not detected in FL and FBM (Figure 2D). This binding activates PI3K/Akt and MAPK signaling pathways, promoting HSC proliferation and survival during early hematopoietic development (Herishanu et al., 2011; Li et al., 2023). Additionally, Mono/Mac-derived SPP1 binds to ITGA5-ITGB1 and ITGA4-ITGB1 co-receptors in AGM but not in FL and FBM. Activation of FAK and Src family kinases downstream of ITGA5-ITGB1 supports HSC adhesion to the extracellular matrix and differentiation, while NF-kB and MAPK signaling pathways activated by ITGA4-ITGB1 contribute to HSC homing and maintenance (Khurana et al., 2016; Krenn et al., 2022). These interactions highlight the diverse regulatory roles of macrophage-derived SPP1 and its co-receptor complexes in governing HSC emergence rather than expansion and maturation in FL and FBM. Mono/Mac-secreted resistin (RETN) binding to CAP1 (calcyphosin) receptor on HSPC significantly increases in interaction strength in FBM compared to FL. RETN-CAP1 binding activates the cAMP/PKA pathway in various cell types, including immune cells and adipocytes (Lee et al., 2014). Additionally, the interaction between amyloid precursor protein (APP) on endothelial cells and CD74 on HSPC progressively strengthens from AGM to FL and FBM (Figure 2A). CD74 can regulate the expression of transcription factors including KLF4, IRF8, and CEBPA, which are known to regulate HSC maintenance (Becker-Herman et al., 2021). The interaction between pleiotrophin (PTN) on MSC and nucleolin (NCL) on HSPC shows a significant increase in interaction strength from the AGM to the FL, with the interaction stabilizing from the FL to the FBM. PTN-NCL binding primarily activates the Wnt and BMP signaling pathway, which is crucial for guiding tissue renewal and regeneration (Reister et al., 2019; Reister et al., 2021).

We also compared the relative information flow of each cell-cell communication signaling among AGM, FL and FBM. In the AGM niche, several pathways are notably specific and crucial for HSC emergence, including ANGPTL, VISFATIN, CDH5, JAM, CD34, NOTCH, KLK, VEGF, MPZ, IGF, ANGPT, FGF, NECTIN, and ncWNT (Figures 2E, F). Many of these pathways, such as NOTCH signaling, are derived from endothelial cells and play a vital role in the EHT process and HSC fate determination (Lomeli and Castillo-Castellanos, 2020; Cai et al., 2024). These pathways ensure a well-structured niche that supports the initial emergence and development of HSC, preparing them for further migration and adhesion in the FL. In the FL, CD48 signaling is highly specific, with CD48 serving as a co-stimulatory molecule targeting CD244A on HSPC. CD48 is primarily expressed in HSPC, granulocyte, cDC1, cDC2, CLP, and T/NK cells (Figure 2G). The CD48-CD244A interaction activates cytokine signaling such as IFNγ, supporting HSC proliferation and function (Boles et al., 2011). The signaling pathways specific to the FBM that target HSPC including LT, TWEAK, CXCL, SELE, ADGRE, and CypA. CypA (cyclophilin A or PPIA) is involved in protein folding and stabilization, influencing HSC survival and differentiation through modulation of cellular stress responses. Depletion of PPIA can accelerate HSC aging (Maneix et al., 2024).

The expression of MHC-I and MHC-II molecules on HSPC is a hallmark of their initial maturation (Supplementary Figures S1E, F). In FL and FBM, MHC-I signaling enables HSPC to target T/NK cells, with the interaction strength decreasing over time (Figure 2H). This decline may reflect a reduced role for mature HSC in directly engaging T/NK cells, potentially indicating a shift in their functional requirements during maturation (Calvanese et al., 2022). MHC-II signaling in FL and FBM involves interactions with various cells, including DCs and other antigen-presenting cells, with a general decrease in interaction strength (Figure 2I). This broad interaction profile highlights the role of MHC-II in shaping the immune environment around HSC, potentially facilitating their integration into the immune system. Notably, the MIF signaling pathway exhibits a progressively increasing and robust interaction across the AGM, FL, and FBM regions during HSC maturation (Figures 2A, J, K). MIF (macrophage migration inhibitory factor) secreted by all cell types targets HSPC mainly through the CD74-CD44 co-receptor complex (Becker-Herman et al., 2021), demonstrating the growing importance of this signaling axis as HSC transition through their developmental stages.

To investigate the alteration of cell-cell communication in the HSC niche during aging, the interaction was evaluated in FBM, YBM, and OBM, and then compared among the three regions to identify changes in communication patterns between OBM and YBM, with the FBM serving as a reference for the relatively stable development of HSC during fetal stage and young adulthood. The overall interaction strength is decreased from FBM to YBM, which may suggest a decline in supportive interactions necessary for maintaining HSC function (Figures 3A, B). The altered cell-cell communication patterns observed in the aging HSC niche have significant implications for HSC function and biological behavior. As HSC age from YBM to OBM, the increased interaction strength may reflect an adaptive response aimed at preserving HSC functionality (Figures 3C, D). The consistent increase in interaction strength from plasma cells and T/NK cells targeting HSPC highlights the critical role of immune cells in modulating the aging HSC niche. The interaction of T/NK cells with HSPC through the GZMA-PARD3 and GZMA-F2R ligand-receptor pairs, specific to OBM and absent in YBM (Figure 3D), induces nitric oxide production, enhancing CXCL12-CXCR4–induced motility and rapid stem and progenitor cell mobilization (Gur-Cohen et al., 2015). The presence of plasma-HSPC interactions such as WNT10A-(FZD6+LRP6), WNT10A-(FZD6+LRP5), and WNT5B-FZD6 in OBM, but not in YBM (Figure 3E), indicates a shift towards more complex signaling in older HSC niches. These Wnt signaling pathways have been implicated in regulating HSC self-renewal, differentiation, and aging. For example, increased Wnt signaling has been shown to induce HSC aging by promoting symmetric division and reducing HSC quiescence (Abidin et al., 2015; Liu et al., 2019). Comparing the ligand-receptor pathways of FBM, YBM, and OBM revealed an enrichment of interleukin signals (IL1, IL2, and IL4) in OBM (Figure 3F), likely contributing to HSC aging by promoting an inflammatory and less regenerative niche (He and Wang, 2021; Caiado and Manz, 2024). The increased exposure of HSCs to inflammatory cytokines in the aging niche may therefore contribute to their functional decline and reduced ability to maintain blood homeostasis.

The increased targeting of various cell types by HSC through the TGF-β pathway from YBM to OBM, absent in FBM, suggests a significant alteration in the regulatory environment associated with HSC aging (Figure 3G). TGF-β signaling has been shown to play a critical role in maintaining HSC quiescence and preventing excessive proliferation (Chambers et al., 2007). The altered TGF-β signaling in the aging niche may disrupt this balance, leading to increased HSC cycling and eventual exhaustion. The enhancement of HSC targeting to T/NK cells via the MHC-I pathway from YBM to OBM (Figure 3H) may reflect increased immune surveillance and clearance activity (Weiskopf et al., 2016). Additionally, the chronic exposure of HSCs to immune stress may lead to telomere shortening and other age-related cellular changes, further contributing to their functional decline. The weakening of MIF pathway interactions (Figures 3A, I) and the reduction in interactions between HSPC and other cell types via the galectin pathway from YBM to OBM (Figures 3J, K) suggest a broader decline in the efficacy of these signaling pathways within the aging HSC niche. The increase in CXCL12-CXCR4 interactions between MSC and HSPC from YBM to OBM (Figure 3L) indicates an enhanced retention and localization of HSC within the niche. While this increased retention may initially serve to protect HSCs against oxidative stress accumulated during aging (Zhang et al., 2016), it may also lead to altered stem cell behavior and function over time. The prolonged exposure of HSCs to the aging niche environment, characterized by chronic inflammation and altered signaling, may ultimately contribute to their functional decline and impaired regenerative capacity.

To characterize the changes in HSC at the transcriptomic level during aging and the alterations in the niche interactome of old HSC compared to young HSC, DEGs of HSPC between OBM and YBM were identified (Figure 4A; Supplementary Table S5). Upregulated genes involved in ribosomal function (RPS18, RPS27, RPL27A) and iron metabolism (FTH1, FTL) indicate increased ribosomal activity and attempts to manage oxidative stress, respectively (Yi et al., 2024). Enrichment analysis of upregulated and downregulated genes revealed increased inflammatory and stress responses associated with aging and cellular senescence (Figure 4B; Supplementary Table S6), such as response to lipopolysaccharide and positive regulation of cytokine production, suggesting heightened immune activation and inflammation (Kim et al., 2016). The over-representation of the “myeloid cell differentiation” term in the OBM suggests that there is an increased activity or propensity towards myeloid lineage commitment in the hematopoietic system during aging (Supplementary Table S6). This finding is consistent with the well-documented phenomenon of age-related myeloid skewing, where the balance of hematopoietic output shifts towards the myeloid lineage at the expense of lymphoid cell production (Oduro et al., 2012; Dorshkind et al., 2020). Analysis of upregulated and downregulated ligands/receptors showed a consistent trend with CellChat analysis (Figure 4C), with most altered interactions during HSC aging belonging to secreted signaling (Figure 4D).

NicheNet was used to infer the activity of ligands targeting the DEGs in HSPC and identify ligands driving HSC aging (Figure 4E). SPP1 was found to be significantly upregulated in MSC, leading to increased expression of its target genes JUN and JUND in HSPC (Figures 4F, G), consistent with the intensification of SPP1 signaling activity targeting HSPC from YBM to OBM during HSC aging (Figure 3A). The increased expression of JUN and JUND may contribute to HSC dysfunction by exacerbating inflammation and skewing myeloid differentiation (Matteini et al., 2021). MMP9, most significantly upregulated in Mono/Mac and granulocytes (Figure 4F), affects HSC by increasing the expression of FOS, JUN, and VEGFA (Figure 4G), facilitating inflammatory signaling and altering the HSC niche, contributing to age-related functional decline (Saw et al., 2019). Increased FCAR expression in Mono/Mac, GMP, and granulocytes leads to elevated TGFB1 levels in HSPC, which may impair HSC function and promote aging-related changes (Blank and Karlsson, 2015). These observations indicate that changes in ligand expression from niche cells significantly impact HSC function, promoting the aging process through mechanisms involving inflammation, oxidative stress, and cellular remodeling. Inhibiting these niche-derived ligands targeting HSPC could offer a promising strategy to mitigate HSC aging and preserve their function.