An age-progressive platelet differentiation path from hematopoietic stem cells causes exacerbated thrombosis

SUMMARY

Platelet dysregulation is drastically increased with advanced age and contributes to making cardiovascular disorders the leading cause of death of elderly humans. Here we reveal a direct differentiation pathway from hematopoietic stem cells into platelets that is progressively propagated upon aging. Remarkably, the aging-enriched platelet path is decoupled from all other hematopoietic lineages, including erythropoiesis, and operates as an additional layer in parallel with canonical platelet production. This results in two molecularly and functionally distinct populations of megakaryocyte progenitors. The age-induced megakaryocyte progenitors have profoundly enhanced capacity to engraft, expand, restore, and reconstitute platelets in situ and upon transplantation, and produces an additional platelet population in old mice. The two pools of co-existing platelets cause age-related thrombocytosis and dramatically increased thrombosis in vivo. Strikingly, aging-enriched platelets are functionally hyper-reactive compared to the canonical platelet populations. These findings reveal stem cell-based aging as a mechanism for platelet dysregulation and age-induced thrombosis.

eTOC/In Brief:

With advanced age, platelets generated by a differentiation pathway that shortcuts the canonical progenitor cascade to directly make megakaryotic precursor cells from hematopoietic stem cells cause thrombocytosis and are more prone to thrombosis compared to canonically derived platelets.

Graphical Abstract

INTRODUCTION

Aging is characterized by the stereotypical decline in tissue function and is the primary risk factor for major diseases. In particular, while platelets (Plts) are critical for controlled hemostasis, their dysregulated production and activation during aging contribute to the pathogenesis of Plt-related disorders in the elderly. Tipping the homeostatic balance towards inadequate Plt production is associated with higher risk for bleeding disorders, whereas Plt overproduction and hyper-reactivity lead to pathologic clot formation in thrombotic diseases such as deep vein thrombosis and ischemic stroke ^1–8. Plts are extremely short-lived (~4 days in mice and ~8 days in humans) and therefore continuously derived from hematopoietic stem cells (HSCs) in the bone marrow (BM) via a gradual differentiation cascade of intermediate progenitor cells towards megakaryocyte progenitors (MkPs) ⁹. While the differentiation path from HSCs to MkPs may be altered upon in vitro or in vivo stress and upon aging ^10–12, we and others have shown that in young, unperturbed mice, MkPs and Plts, like all other adult hematopoietic lineages, arise via a Flk2+ differentiation stage (Figure 1A, top row of Figure 1B) ^13,14. HSCs in the young adult “FlkSwitch” mouse model express Tomato (Tom), whereas multipotent progenitors (MPPs) and downstream progenitor and mature cells irreversibly switch to GFP expression due to Cre-mediated Tom excision (“floxing”). The FlkSwitch model labels several cell populations with combined platelet/erythroid potential, and we have demonstrated that Plts are GFP+ during development regardless of whether they are derived via Tom+ fetal HSCs or co-existing, developmentally restricted GFP+ HSCs ^13,15–18. Analogous differentiation paths and progenitor populations can also be discerned in humans and persist throughout ontogeny ^19,20. Several independent studies have contributed to the consensus conclusion that hematopoietic aging is characterized by the deterioration of HSC function, most notably demonstrated by their reduced reconstitution capacity compared to young HSCs (yHSCs) ^21–24. This age-related functional decline of old HSCs (oHSCs) stands in stark contrast to our recently uncovered surprising gain of function of old MkPs (oMkPs) compared to young MkPs (yMkPs) ^25,26. This led us to hypothesize that age-related alterations in MkPs contribute to dysregulation of Plts in the elderly. Here, we employed the FlkSwitch model as a powerful tool for tracking hematopoietic differentiation pathways to demonstrate an unexpected cellular mechanism for the etiology of Plt-related disorders upon aging.

Figure 1. Aging of FlkSwitch mice leads to progressive enrichment of Tomato+ megakaryocyte progenitors and platelets.

A. Schematic of the mTmG and Flk2-Cre constructs that serve as the basis for the color composition of hematopoietic cells in the FlkSwitch mouse model. Cre expression in Flk2+ cells leads to irreversible deletion of Tomato and a switch to GFP-expression in all descendent cells.

B. Young FlkSwitch mice show very high, equal floxing of mature cells (3-month old, top), whereas Tom+ Plts, but not erythroid, GM, B or T cells, increase in aged mice (24-month old, bottom).

C. The proportion of GFP+ Plts progressively decrease beyond 12 months of age, whereas the vast majority of erythroid, GM, B and T cells remain GFP+ for life. Quantification of data from B and additional intervening time points are shown. Data represent mean ± SEM of 6 independent experiments, n=21 mice. Statistics: unpaired t-test compared to baseline 3 months old. ***P<0.0005

D. HSCs remain Tom+ for life. Tom versus GFP expression in young and old HSCs.

E. MkPs are GFP+ in young mice and Tom+ MkPs are enriched in old mice. Tom versus GFP expression in young and old MkPs.

F. All progenitors except MkPs maintain efficient switching to GFP-expression throughout life. Fold difference in percent GFP+ classical myeloid, erythromyeloid, and lymphoid progenitor cells compared to MPPs in the BM of young and old FlkSwitch mice.

G. Multipotent progenitor subfractions in young (Y) and old (O) FlkSwitch mice remain GFP+ during aging. Percent GFP labeling of MPP2, MPP3, MPP4, and ST-HSC, gated as in Figure S2. Data represent mean ± SEM of 3 independent experiments, n=5 young mice, n=5 old mice. Statistics: t-test. Comparisons of young to old were not statistically different.

H. Percent GFP+ cells in hematopoietic populations from ten individual Old FlkSwitch mice.

I. Aging-induced shortcut megakaryopoiesis. Schematic of youthful differentiation pathways in FlkSwitch mice (left) and altered megakaryopoiesis in old FlkSwitch mice (right). In young adult and old FlkSwitch mice, HSCs express Tom. Only aging mice have Tom+ MkPs and Plts; cells of all other hematopoietic lineages remain GFP+ throughout life.

RESULTS

Platelets and Megakaryocyte Progenitors Diverge from a Flk2-Dependent Differentiation Pathway During Aging

To determine the potential changes to lineage specification during aging, we aged FlkSwitch mice and tracked Tom+ and GFP+ hematopoietic cells over time. “Tom+” cells were defined as Tom+GFP- to indicate lack of current and past Cre expression, whereas “GFP+” cells include both Tom-GFP+ and Tom+GFP+ cells to account for lingering Tom protein in cells where GFP expression directly indicates that Cre-mediated recombination has occurred (Figure 1B, D, S1) ¹³. As previously reported, all mature cell populations in the peripheral blood (PB) of young mice were predominantly GFP+ ¹³. Surprisingly, aging led to the distinct production of Tom+ Plts, but not Tom+ erythroid, granulocyte/macrophage (GMs), B or T cells (Figure 1B). The proportion of GFP+ Plts in the PB progressively decreased beyond 12 months of age to reach ~50% of the total Plt pool by 20+ months in each individual mouse analyzed, whereas the vast majority of erythroid, myeloid, B, and T cells remained GFP+ for life (Figure 1C). Examination of bone marrow (BM) populations revealed high fidelity of the FlkSwitch paradigm established with young mice ^13,14: the vast majority of HSCs remained Tom+ for life, and Flk2+ MPPs (MPP^F) efficiently switched to GFP expression (Figure 1D,F, Table S1). Moreover, the overwhelming majority of classical myeloid progenitor cells, including common myeloid progenitors (CMPs), granulocyte/macrophage progenitors (GMPs), megakaryocyte-erythroid progenitors (MEPs), and erythroid progenitors (EPs) were GFP+, as were myeloid populations using alternative markers (Figure 1F, 1H, S1) ²⁷. A striking exception was observed for MkPs: nearly half of MkPs in old, but not young, mice were Tom+ (Figure 1E-F). The divergence of MkP/Plt generation from erythroid production was particularly surprising, as these two lineages share critical molecular regulators as well as several progenitor populations with combined erythroid and Plt repopulation capacity ^28–31. The uniqueness of this reduced floxing pattern was further underscored by investigation of tissue-resident macrophages (trMacs). Brain and lung trMacs are known to be minimally GFP-labeled in young adult FlkSwitch and other Flk2-driven lineage tracing mice ^32–34. Interestingly, GFP labeling of trMacs significantly increased in aged FlkSwitch mice (Figure S2), contrasting the Flk2-divergent specification of Plts during aging. The megakaryopoiesis-specific shortcut was observed in every individual mouse that we have aged and analyzed to date (Figure 1H and Table S1). Analysis of multipotent progenitor pools believed to exist between HSCs and Flk2+ MPPs failed to identify a clear candidate intermediate, including the proposed megakaryocytic/erythroid-biased MPP2 (Figure 1G, S2) ³⁵, indicating that the Tom+ MkPs may derive directly from the Tom+ aged HSCs. Consistent with this idea, only HSC and MkP pools were significantly expanded upon aging (Figure S1D) ^25,36,37. Together, these data demonstrate that the canonical Flk2+ differentiation programs of all lineages are robustly maintained throughout life, and that only the Plt lineage deviates from the classical Flk2+ pathway during aging by initiation of a pathway in parallel to canonical Plt differentiation (Figure 1I).

Young and Old HSCs similarly differentiate through the Flk2+ pathway upon iso- and heterochronic transplantation

To determine whether the Plt-specific pathway was imposed by the aged environment or manifested by heritable changes in oHSCs, we performed heterochronic and isochronic transplantation experiments of HSCs from FlkSwitch mice and monitored the differentiation paths by tracking Tom:GFP ratios of donor-derived mature cells in the PB (Figure S3A–B). As we have previously shown, the vast majority of circulating cells derived from Tom+ yHSCs were GFP+ (>90% of B and T cells were GFP+, and Plts and GMs plateaued at ~60%; the remaining donor-derived cells are Tom+) after transplantation into young hosts (Y-Y) ¹³ (Figure S3C, top). The extent of GFP-labeling of erythromyeloid cells is determined at the level of MPP^{F 13}, therefore equally affecting GMs and Plts. To determine whether oHSCs deviate from this baseline Tom:GFP ratio established in Y-Y transplants, and therefore retain the shortcut Plt differentiation potential, we transplanted old FlkSwitch HSCs into young recipients. Interestingly, oHSCs gave rise to Plts, GMs, B, and T cells with similar Tom:GFP ratios (O-Y) as yHSCs (Y-Y) in young recipients (Figure S3C, bottom), indicating that transplanted oHSCs did not retain their divergent Plt specification in the young niche. Additionally, similar Tom:GFP patterns were observed when either yHSCs (Y-O) or oHSCs (O-O) were transplanted into old recipients (Figure S3D). Thus, the aged niche did not appear to induce the divergent Plt-specific differentiation of transplanted yHSCs, and transplantation of oHSCs did not recapitulate the divergent Plt-specific differentiation path observed in situ. Instead, transplantation appeared to reset oHSCs into a youth-like Plt differentiation path, regardless of recipient age (Figure S3C versus S3D) and despite reduced overall repopulation capacity of oHSCs relative to yHSCs in both young and old recipients (Figure S3E and S3F). These experiments demonstrate that the differentiation paths of both yHSCs and oHSCs are relatively unaffected by the recipient age in adoptive transfer experiments.

Age-enriched MkPs are transcriptionally distinct

To investigate the molecular programs of aging-induced MkP heterogeneity, we compared the bulk transcriptomes of the three distinct murine MkP populations: GFP+ yMkP, GFP+ oMkP, and Tom+ oMkP. Young and Old GFP+ MkPs remained relatively similar with age, distinguished by only 73 differentially expressed genes (DEGs) (Figure 2A). This indicates a remarkable resilience of GFP+ oMkPs to aging-induced changes known to be characteristic of the BM environment ^38–41. In contrast, Tom+ oMkPs were substantially different from both GFP+ yMkPs (1,071 DEGs) and oMkPs (187 DEGs, Table S2), and associated with Gene Ontology (GO) terms such as Regulation of Response to Stress and Immune Response (Figure 2B). Interestingly, several genes associated with HSC-selective expression were increased in the Tom+ oMkPs compared to GFP+ oMkPs: Mllt3, Fhl1, Plscr2, Nupr1, Cdkn1c, Hoxb5, Cd105, and Ly6a (Figure 2C-D, and S4C) ^42–47. We corroborated our gene expression data by using flow cytometry to demonstrate that mRNA changes resulted in differential expression of cell-surface proteins of representative upregulated (CD105) and downregulated (CD119) genes in Tom+ vs GFP+ oMkPs (Figure 2D). These data demonstrate that Tom+ oMkPs represent an aging-enriched hematopoietic progenitor with a distinct molecular profile. To investigate the mechanisms by which aging induces the divergent Plt pathway, we compared expression profiles of young and old MkPs to oHSCs. Interestingly, compared to GFP+ yMkPs and GFP+ oMkPs, Tom+ oMkPs were transcriptionally more similar to oHSCs (Figure 2E, S4A). Hierarchical clustering analysis of gene expression patterns also confirmed similarity of the three MkP population and oHSCs and revealed a subset of genes that were highly expressed in both Tom+ oMkP and oHSCs, but not GFP+ MkPs (Figure 2F). Tom+ oMkPs also clustered closest to oHSCs when we compared our RNAseq data with HSC- and MkP-specific genesets available from Gene Expression Commons (GEXC) (Figure S4B) ⁴⁸. Together, these results demonstrate that oHSCs are transcriptionally more similar to Tom+ oMkPs compared to old and young GFP+ MkPs, supporting the shortcut pathway from oHSCs into Tom+ oMkPs identified by lineage tracing.

Figure 2. Bulk RNA sequencing revealed a distinct molecular profile of Tom+ MkPs.

A. Volcano plots showing DEGs between MkPs. Dotted lines indicate P=0.05.

B. GO term analysis of DEGs in Tom+ MkPs compared to GFP+ MkPs.

C. Tom+ oMkPs highly express genes associated with HSC function. mRNA levels of specific DEGs by RNAseq read count in Tom+ oMkPs compared to GFP+ oMkPs. **P<0.005, ***P<0.0001.

D. Differential cell-surface expression of CD105 and CD119 by old MkPs. Flow cytometry analysis to test RNAseq results. Data represent 3 independent experiments with n=4 young, n=4 old mice. Statistics: paired t-test. *P<0.05.

E. Tom+ oMkPs are located closest to oHSCs in transcriptional space. Principal Components 1 and 2 capture 93.6% of the total transcriptional variance across MkPs and HSCs, demonstrating that Tom+ oMkPs are most similar to oHSCs in PC1 and according to Euclidean distance calculated from centroids (diamonds).

F. Kmeans-clustered heatmap of gene expression Z-score for DEGs between Tom+ oMkPs (up) and GFP MkPs (down) demonstrates shared transcriptional signal in oHSCs and Tom+ oMkPs that is diminished or absent in other MkP populations.

Single-cell transcriptomics identified Tom+ MkPs as a distinct age-enriched cluster

To understand lineage relationships and cell states of the shortcut Plt pathway at the clonal level, we performed single-cell RNA sequencing (scRNAseq) on Lineage-c-Kit+CD150+ and Lineage-c-Kit+CD150-cells from young and old FlkSwitch mice. Unsupervised clustering identified 21 unique clusters which were annotated based on gene expression profiles with established lineage-associated markers of hematopoietic cells (Figure 3A, S5A). MkPs selectively express Meg-lineage genes: Selp, Itga2b (CD41), Itgb3 (CD61), and CD9 (Figure S5A). Iterative unsupervised clustering of MkPs revealed a subcluster that was enriched in old mice (Figure 3B). This age-enriched MkP subpopulation was uniformly Tom+, while all other MkPs and progenitor populations were almost exclusively GFP+ (Figure 3C-E), substantiating the discovery of this population by lineage tracing (Figure 1). Moreover, this Tom+ MkP population was enriched for a gene expression signature similar to that identified by bulk RNAseq (Figure 3F, S5A, Table S2). Pseudobulk analysis derived from the single-cell transcriptome of the three MkP populations also showed concordance with the DEGs identified by our bulk RNAseq analysis (Figure 3G, S5B, Table S2, S3A–F). Thus, the transcriptomic signatures of the Tom+ oMkPs were largely recapitulated in both datasets.

Figure 3. Single-cell transcriptomics provided independent evidence of the aging-induced, distinct cluster of Tom+ MkPs.

A. UMAP representation of scRNAseq data from combined cKit+CD150+/− cells from young and old FlkSwitch mice.

B. UMAP occupancy of young and old cells in the scRNAseq data. Circled area represents a population of cells enriched in old mice.

C. Old MkPs, but not young MkPs or young or old GMPs, express Tomato. Quantification of the number of GFP+ and Tom+ MkPs and GMPs by scRNAseq data demonstrating strong concordance with Tom/GFP profiles by flow cytometry.

D. GFP and Tom-expression in the scRNAseq data demonstrating that the population of cells only present in old BM is Tom+ (red), whereas all other cells in both young and old FlkSwitch mice express GFP (green).

E. GFP and Tom-expression among annotated MkPs in the scRNAseq data. Only old MkPs possess Tom-expressing cells.

F. Dot plot of selected genes among young and old MkP populations. MkP lineage genes (left) and the top 10 DEGs (right, top five upregulated and top five downregulated) between MkPs. Data is scaled by gene.

G. Volcano plots showing DEGs between MkPs based on pseudobulk analysis of scRNAseq. Dotted lines indicate where P=0.05.

H-J. Two “classical” trajectory inference models, DPT and PAGA, are unable to transcriptomically identify the aging-induced branchpoint.

H. PAGA analysis. Tom+ MkPs are shown as a probable link from HSCs.

I. DPT analysis. DPT values (left) and cell cluster annotation (right).

J. Investigation of GFP+ and Tom+ MkPs via DPT further indicates no apparent transcriptome-based branch point.

K. DPT analysis of scRNAseq data revealed that Tom+ oMkPs are generated faster than GFP+ MkPs from HSCs. HSCs and MkPs are displayed as subsets of all annotated clusters. The median for each group is indicated by the line within the box plot. HSC n=189 cells, Tom+ MkP n=402 cells, and GFP+ MkP n=832 cells. Statistical analysis was conducted with a one-way ANOVA adjusted for multiple comparisons (Tukey). ****P<0.0001

We next sought to reconstruct the hierarchy of the shortcut Plt pathway using trajectory analysis. We utilized two commonly employed algorithms, Diffusion Pseudotime (DPT) ⁴⁹ and Partition-based graph abstraction (PAGA) ⁵⁰, neither of which was capable of identifying unique transcriptional paths driving aging-induced Plt differentiation (Figure 3H-J). This is likely because, although GFP+ and Tom+ oMkPs are uncoupled in ontogeny, they reconverge on similar transcriptomes as functional MkPs and are more similar to each other than other cell populations (Figure 3A-E). As an illustration of this, the number of DEGs between them only represent 10.03% of detected transcripts, likely preventing trajectory inference modalities from predicting distinct differentiation paths. In contrast, DPT plotted by annotated cluster revealed that Tom+ MkPs arose faster in pseudotime compared to their GFP+ counterparts, consistent with (but not proving) a bypass pathway with fewer intermediate progenitor cell states (Figure 3K). Additionally, Plt-lineage genes and top DEGs between oGFP+ and oTom+ MkPs illustrate shared lineage programs yet divergence in gene expression (Figure 3F). Complementing our native long-term lineage tracking data (Figure 1), this demonstrates that unperturbed aging hematopoiesis is associated with the acquisition of a transcriptome that drives old HSCs toward Plt production.

Age-enriched MkPs are functionally enhanced compared to both the canonically-derived coexisting Old MkPs and Young MkPs

Our observations that transplanted yHSCs and oHSCs exhibit similar floxing efficiency point to oMkPs as putative major perpetuators of the aging-induced divergent Plt pathway. We recently demonstrated that the bulk population of oMkPs have a remarkable expansion capacity compared to yMkPs ²⁵. While these alterations could logically have been attributable simply to the process of either cell-intrinsic aging or to an aging phenotype imposed by the environment, our RNAseq data where GFP+ MkPs were more similar throughout aging than to Tom+ MkPs that co-exist with GFP+ MkPs in the aged environment prompted us to consider alternative explanations. To investigate potential functional consequences of the aging-induced shortcut Plt path, we first compared the in vitro expansion capacity of GFP+ yMkPs, GFP+ oMkPs, and Tom+ oMkPs (Figure 4A). Whereas the GFP+ yMkPs and GFP+ oMkPs were indistinguishable in their low capacity to expand in culture, Tom+ oMkPs expanded significantly more compared to both GFP+ MkP populations (Figure 4B). Phenotypic analysis revealed that cultures initiated by Tom+ oMkP contained significantly more phenotypic MkPs compared to the two GFP+ MkPs (Figure 4C). These observations prompted in vivo assessment of the three MkP populations by transplantation (Figure 4D-I). We first examined their clonal potential by in vivo spleen colony assays (Figure 4D) ^51,52. Notably, we observed a greater number of colony-forming units of the spleen (CFU-S) derived from Tom+ oMkPs compared to GFP+ MkPs (Figure 4E-G). Given the unicellular origin of each colony, these CFU-S data, therefore, revealed that a substantial fraction of Tom+ oMkPs demonstrate greater in vivo clonal potential compared to GFP+ MkPs. We next investigated how the engraftment and multi-lineage reconstitution potential of MkPs might be altered by the aging-induced shortcut Plt pathway (Figure 4H). We previously demonstrated that the bulk population of oMkPs robustly reconstitute Plts. Therefore, we were surprised to find that the GFP+ fraction of oMkPs minimally contributed to Plt donor-chimerism and at no greater capacity than yGFP+ MkPs (reaching 1–6% donor contribution) (Figure 4I). In direct contrast to both GFP+ MkP populations, the Tom+ oMkPs were remarkably capable of robust, yet transient, reconstitution of Plts (Figure 4I). As expected due to lack of Flk2 and therefore Cre expression by MkPs, GFP+ MkPs gave rise to GFP+ Plts and Tom+ MkPs gave rise to Tom+ Plts (Figure S5C). Consistent with our previously reported results, all three MkP populations lacked the capacity to reconstitute B and T cells, while some erythroid and GM chimerism was observed (Figure S6A) ²⁵. Similar to a recent study ⁵³, we found that yMkPs can be phenotypically subdivided by CD48 expression, with detection of a small population of CD48^-/lo MkPs in young, unperturbed mice (Figure S6B). Importantly, however, the CD48+ and CD48^-/lo yMkP populations were functionally indistinguishable from each other by in vitro analysis (Figure S6C) and did not have differential Plt reconstitution capacity upon transplantation (Figure S6D), but the CD48^-/lo population did produce significantly more erythroid cells (Figure S6E). These data from WT mice are consistent with the FlkSwitch model in that the vast majority of MkPs in young mice can be viewed as one population under native conditions. Collectively, the surprising gain of functional capacity upon aging that directly contrast the behavior of aged HSCs appears to be harbored entirely within the age-enriched Tom+ MkPs ²⁵. This enhanced capacity can be mechanistically explained by their specification from HSCs via shortcut differentiation, with inheritance of sustained expression of stem cell-promoting genes such as Hoxb5, Mllt3, and c-Kit (Figure 2C) that, collectively, contribute to the observed increase in lineage potential and engraftment capacity. The superior ability of age-enriched MkPs to expand, engraft, and reconstitute Plts compared to the co-existing canonical MkPs suggests a selective and pathway-specific consequence of aging. Moreover, the functional similarities of GFP+ yMkPs and GFP+ oMkPs further support a model where the canonical, Flk2+ differentiation path tempers major aging-imposed changes to molecular and functional properties throughout life.

Figure 4. Age-enriched MkPs have greater expansion, clonal, and platelet reconstitution potential upon transplantation compared to GFP+ Young or Old MkPs.

A. Schematic of in vitro expansion assay of MkPs from FlkSwitch mice. 1000 MkPs were plated per well and quantified by flow cytometry after 3 days of expansion.

B. Tom+ oMkPs displayed greater expansion capacity in vitro compared to both GFP+ yMkPs and GFP+ oMkPs. Left: Representative images of MkP in vitro cultures at day 3. Right: Quantification revealed significantly greater number of cells from Tom+ MkPs compared to both GFP+ MkPs. Data represent mean ± SEM of 3 independent experiments, n=9 GFP+ yMkP wells, n=9 GFP+ oMkP wells and n=17 Tom+ oMkP wells. Statistics: one-way ANOVA,Tukey’s multiple comparisons test. ****P<0.001

C. Gating strategy and summary of immunophenotypic analysis of cultured MkPs from B. Tom+ oMkP cultures contain significantly more phenotypic MkPs compared to the two GFP+ MkPs.

D. Schematic of MkP CFU-S analysis. 2,000 MkPs from FlkSwitch mice were transplanted into lethally irradiated mice and spleens were analyzed for CFU-S at day 8.5.

E-F. Tom+ MkPs harbors greater clonal potential compared to GFP+ MkPs. Representative fluorescence microscopy images (E) and enumeration (F) of colony number per spleen positive for colonies. Data represent mean and SEM from 3 independent experiments. GFP+ yMkP n=4; GFP+ oMkP n=5; Tom+ oMkP n=8. Statistics: one-way ANOVA, Tukey’s multiple comparisons test. *P<0.05.

G. Summary of CFU-S experiments. All transplanted mice: GFP+ yMkP n=7; GFP+ oMkP n=6; Tom+ oMkP n=10.

H. Schematic of MkP transplantation from FlkSwitch mice. 22,000 MkPs were isolated by FACS and transplanted into young WT recipient mice. PB analysis by flow cytometry was done to monitor repopulation of mature cells.

I. Tom+ oMkPs demonstrated greater contribution to Plts compared to both GFP+ yMkPs and GFP+ oMkPs. Donor-derived Plts in PB of recipients presented as percent donor-chimerism. Data represent mean ± SEM of 3 independent experiments, n=6 GFP+ yMkP recipients, n=4 GFP+ oMkP recipients, and n=13 Tom+ oMkP recipients. Statistics: unpaired two-tailed t-test. T-tests between GFP+ yMkP and GFP+ oMkP were not statistically significant. T-tests between GFP+ oMkP and Tom+ oMkP *P<0.05, **P<0.005, ***P<0.0005.

Age-enriched MkPs rapidly restore acutely depleted Plts

The drastic increase in thrombocytosis-fueled Plt pathologies and the remarkable Plt reconstitution capacity of age-enriched MkPs led us to hypothesize that the shortcut Plt pathway may also harbor highly potent physiological control of Plt homeostasis. To directly test whether the age-enriched MkPs endow old mice with enhanced Plt restoration capacity, we elicited acute thrombocytopenia in FlkSwitch mice and measured Plt differentiation kinetics during Plt recovery. Plt-specific stress was induced by subjecting young and old FlkSwitch mice to a single anti-Plt antibody injection (Figure 5A). As expected ^54–56, this led to extremely rapid and robust depletion of circulating Plts in young mice, with gradual recovery starting at ~3 days (~72 hours, Figure 5B). Similarly, Plt counts also dropped drastically in old mice; notably, however, Plt numbers were restored more rapidly compared to young mice (Figure 5B). Analysis of the Tom:GFP ratio of Plts suggested that a significantly greater proportion of the newly produced Plts were derived via the shortcut pathway (Figure 5C-D). The Plt challenge also provoked alterations to BM cellularity at 24 hours post-depletion, with significant and trending reduction of MkPs in young and old mice, respectively, while no major changes were observed in HSC and MPP cellularity (Figure 5E-F). Interestingly, no major alterations to the frequencies of MkPs that were GFP+ were observed in either young and old FlkSwitch mice at 24 hours post-depletion (Figure 5G). The enumeration and floxing of PreMegE populations were also unaffected upon Plt depletion (Figure 5H). To determine if the Tom+ MkPs may have a competitive advantage to respond to the Plt challenge, we quantified in vivo 5-ethynyl-2′-deoxyuridine (EdU) incorporation upon Plt depletion (Figure 5I). The challenge elicited an overall increase in proliferation of both HSCs and MkPs, while MPPs were unaltered compared to steady-state (Figure 5J-K). Importantly, in old mice, consistently marked increase in proliferation was exhibited by Tom+, but not GFP+ oMkPs (Figure 5K). The hyper-responsiveness of functionally enhanced age-enriched Tom+ oMkPs is consistent with our observations of their superior in vitro expansion and profound reconstitution capacity upon transplantation (Figure 4A-I). Taken together, these data suggest that HSCs and MkPs exhibit a dynamic response to rebuild the circulating Plt supply with an aginginduced shift in cellular mechanisms. Compared to GFP+ oMkPs, the age-enriched Tom+ oMkPs appear to serve as dominant first responders in the rapid rescue of acute thrombocytopenia.

Figure 5. Age-enriched MkPs serve as first responders to acute platelet depletion.

A. Schematic of anti-GPIbα-mediated Plt depletion and subsequent time-course analysis of cellular response. BM analysis was performed 24 hours post-depletion.

B. A single injection of anti-GPIbα antibodies led to rapid and robust Plt depletion followed by gradual restoration of circulating Plt numbers. Plt numbers/microliter PB are indicated at each time point. Grey inset shows Plt numbers with a different y-axis scale during the period of lowest Plt numbers.

C-D. Analysis of Tom:GFP Plts demonstrated reduction in floxing in young (C) and old (D) mice during Plt restoration. Data in C-D represent mean ± SEM of 4 independent experiments with n=6 young mice, n=4 old mice. Statistics: unpaired t-test. *P<0.05, **P<0.005, ***P<0.0005, ****P<0.00005.

E-F. Quantification of BM cellularity 24 hours post-depletion revealed a significant decrease in frequencies of yMkPs, while yHSCs and yMPPs were unaltered (E). A similar pattern was observed in old mice, albeit an insignificant trend in oMkP cellularity upon Plt-depletion (F).

G. Floxing of MkPs were unaltered 24 hours post-depletion compared to control mice.

H. pMegE numbers and floxing were unaltered in the BM 24 hours post-depletion compared to control mice.

I. Schematic of EdU-incorporation analysis. Proliferation rates in control and Plt-depleted mice were pulsed for 24 hours with EdU, with or without simultaneous administration of anti-GPIbα antibodies.

J-K. Plt depletion selectively induced MkP proliferation. Short-term in vivo EdU incorporation revealed that HSCs and MkPs, but not MPPs, in young and old mice respond to Plt depletion, with Tom+ oMkPs dominating the response in old mice (K). Data represent mean ± SEM of 4 independent experiments with n=4 young mice, n=4 old mice. Statistics: paired t-test. *P<0.05, ***P<0.0005.

Age-enriched platelets participate in exacerbated clot formation upon vascular injury

Age-related dysregulation of both Plt numbers and activity poses tremendous health risks for aging humans. Here, we show that the well-documented significant increase in Plt counts in aging mice is due to an additive accumulation of Tom+ Plts to the canonical GFP+ Plts, while changes to the absolute number of other circulating cells result from GFP+ differentiation pathways (Figure 6A) ^8,25,57. Remarkably, despite being specified by two distinct paths, immunophenotyping of oPlts revealed uniform expression of well-established Plt surface proteins, including CD41, CD9, CD42a, and CD42b (Figure 6B). Measurement of Plt lifespan by flow cytometric analysis of the disappearance of fluorescently labeled Plts in vivo (Figure 6C) revealed that GFP+ and Tom+ oPlts also had indistinguishable half-lives (~48 hours) and a maximal lifespan of 144 hours (Figure 6D).

Figure 6. The age-induced shortcut platelet pathway contributes to platelet hyper-reactivity in old FlkSwitch mice.

A. Aging leads to thrombocytosis due to accumulation of Tom+ Plts. Absolute quantification of circulating cells presented as total cells/microliter of PB. While changes to the absolute number of other mature cells in the PB result from cells derived via GFP+ differentiation pathways, the numerical increase in Plts is a consequence of the Tom+ differentiation path in aged FlkSwitch mice. Statistics: unpaired t-test. *P<0.05, **P<0.005, ***P<0.0005. T-tests between Tom+ yPlts and Tom+ oPlts: ***P<0.0005. T-tests between all other Tom+ young and old cells were insignificant.

B. Tom+ oPlts express traditional Plt markers. Frequency of Plts expressing known Plt surface markers: CD41, CD9, CD42a, CD42b.

C. Schematic of experimental design for measuring Plt lifespan. Using flow cytometry, the lifespan of GFP+ and Tom+ Plts was monitored by tracking the disappearance of in vivo CD42c-DyLight649 labeling over time.

D. The lifespan of Tom+ and GFP+ Plts in old FlkSwitch are equivalent. Top: Representative histograms of Plt-labeling and B220+ labeled B-cell controls (grey=unlabeled). Bottom: Summary of Plt lifespan kinetics. Data represent 2 independent experiments with n=4 young WT for B cell analysis and n=8 old FlkSwitch mice for Plt analysis. Statistics: one-way ANOVA, Tukey’s multiple comparisons test. **P<0.001, **** P<0.0001

E. Schematic of the laser-induced thrombosis model to monitor thrombus formation upon vascular injury.

F. Tom+ oPlts contribute to excessive thrombus formation in old mice. Representative images of clot formation in young (top) and old (bottom) FlkSwitch mice displaying participation of GFP+ and Tom+ Plt in thrombi. Also see Supplemental Movies.

G. Tom+ and GFP+ cells are major contributors to greater thrombus formation in old FlkSwitch mice, while the smaller thrombi in young FlkSwitch mice consist exclusively of GFP+ cells. Dynamics of clot formation at time points post-vascular injury were quantified by MFI of thrombi, comparing GFP+ (left) or Tom+ (right) accumulation. Data represent mean ± SEM of 3 independent experiments with n=3 young, and n=3 old mice, with 9–12 injuries per mouse. Statistics: unpaired t-test of area under the curve. *P<0.01, ****P<0.0001,

H. Fibrin content increased in thrombi in old mice. Dynamics of fibrin formation in thrombi at time points post-vascular injury were analyzed by change in MFI conferred by A647-conjugated anti-fibrin antibodies. Representative images of fibrin formation (left) and quantification of fibrin MFI within thrombi (right) in FlkSwitch mice. Data represent 3 independent experiments with n=3 young, and n=3 old mice. Statistics: unpaired t-test of area under the curve. ****P<0.0001

I. Platelet-leukocyte aggregate formation is greater in old compared to young blood. Quantification of myeloid cell aggregation with Plts (Gr1+CD41+, left) and B-cell aggregation with Plts (B220+CD41+, right) with and without thrombin-mediated activation (0.1 U/mL). + indicate thrombin-stimulated. Data represent 3 independent experiments with n=7 Y mice, n=8 Y+, mice n=22 O mice, n=8 O+ mice. Statistics: one-way ANOVA, Tukey’s multiple comparisons test. *P<0.05, ***P<0.0005, ****P<0.0005

J-K. oPlts demonstrated higher activation of integrin αIIb/β3 (left) and P-selectin (right) surface display upon stimulation by ADP (10 μM) (J) and thrombin (0.1 U/mL) (K). Quantification of Plt frequency and MFI within CD9+ Plts. Data represent 3 independent experiments with n=5 young mice, n=5 old mice. Statistics: unpaired t-test. *P<0.05, **P<0.005, ***P<0.0005. Quantification of P-selectin+ Plt frequency (left) and P-selectin MFI (right) within CD9+ Plts. Data represent 3 independent experiments with n=7 young mice, n=7 old mice. Statistics: unpaired t-test. *P<0.05, ***P<0.0005, ****P<0.0005 Y, young; O, old

To assess Plt functionality, we evaluated the role of Tom+ and GFP+ Plts in clot formation in vivo using an intravital microscopy laser ablation model of hemostasis (Figure 6E) ^58–60. We introduced laser-induced injuries to cremaster arteriole walls of FlkSwitch mice to initiate and directly visualize thrombus formation, then quantified the accumulation of Tom+ and GFP+ thrombus constituents in real time. Vascular injury to young FlkSwitch mice resulted in the formation of small, unstable thrombi composed of only GFP+ cells (Figure 6F, top panel, Figure S7, Movie S1). In contrast, immediately after rupture of the vascular walls in old FlkSwitch mice, we observed the formation of dramatically large, stable thrombi at the site of injury, with participation of both GFP+ and Tom+ Plts (Figure 6F, bottom panel, Figure S7, Movie S2). Larger thrombus formation in the old FlkSwitch model were substantiated by significantly greater fluorescence of GFP+ and Tom+ cells within the clots of the old FlkSwitch mice compared to young FlkSwitch mice (Figure 6G). Additionally, vascular insult also resulted in greater accumulation of fibrin within the clot of old mice compared to young mice, indicative of more stable thrombi (Figure 6H). These experiments uncovered that laser-induced clot formation is drastically amplified in old mice, and that both Tom+ and GFP+ Plts contribute to the enlarged thrombi.

The hyper-reactivity of Plts from old mice was also evident by Plt-leukocyte aggregation (PLA) assays that quantify mature myeloid and B-cell interaction with Plts. Circulating PLAs appeared at significantly higher frequencies in old compared young blood (Figure 6I, Figure S7). Upon in vitro stimulation by thrombin, we also observed a greater elevation in PLA formation in old compared to young blood (Figure 6I, Figure S7). Thus, PLA formation is elevated in old blood under both basal and stimulated conditions, further emphasizing the potent thrombotic response in aged blood.

We then sought to measure the mechanisms of activation of aged Plts by examining the expression of cell-surface proteins that are critical for Plt activation. Upon stimulation, the integrin αIIb/β3 switches from low affinity to high affinity for its ligand (fibrinogen) to promote adhesion and thrombus growth ⁶¹. Additionally, P-selectin is translocated from intracellular granules to the external membrane in activated Plts ^62,63. In vitro analysis of Plt activation revealed conformational change and elevated expression of αIIb/β3 and P-selectin, respectively, on oPlts compared to yPlts upon both adenosine diphosphate (ADP, Figure 6J) or thrombin (Figure 6K) stimulation. Thus, αIIb/β3 and P-selectin expression are consistent with hyper-reactivity of old Plts in response to stimulatory agonists. Together, these data suggest that the two Plt populations in aged mice contribute to physiological alterations in hemostasis by enhancing Plt participation in clot formation.

A hyper-reactive platelet subpopulation accumulates upon aging

Based on our observations of functionally enhanced Tom+ MkPs (Figures 4 and 5) and the hyper-reactivity of oPlts (Figure 6), we hypothesized that the aging-induced Plt differentiation pathway leads to Tom+ Plts that are functionally distinct from GFP+ Plts. To test this hypothesis, we compared the functional differences between Tom+ and GFP+ oPlts in agonist-induced activation assays. Strikingly, thrombin-induced real-time neo-exposure of P-selectin was more rapid and robust in the Tom+ oPlts compared to both yPlts and GFP+ oPlts (Figure 7A). Similar to P-selectin, Endothelial cell specific adhesion molecule (ESAM) also translocate from alpha-granules to the cell-surface following Plt activation⁶⁴. Consistent with P-selectin neo-exposure, Tom+ oPlts also exhibited greater membrane translocation of ESAM upon thrombin activation compared to the coexisting GFP+ oPlts and to yPlts (Figure 7B). Interestingly, side scatter measurements by flow cytometry showed undiscernible kinetics and levels of degranulation between the three Plt populations upon activation (Figure 7C), thereby uncoupling degranulation from hyper-reactivity.

Figure 7. Age-enriched platelets are functionally hyper-reactive compared to GFP+ old platelets.

A. Tom+ oPlts undergo more rapid P-selectin neo-exposure compared to GFP+ oPlts and yPlts in response to 0.2 U/mL thrombin. Real time flow cytometry analysis of P-selectin exposure upon activation ex vivo. Representative flow plot and corresponding quantification of MFI rate and kinetics. (A-C) Underlaying faded lines represent sample controls without thrombin stimulation. Data represent 4 independent experiments with n=4 young WT, and n=10 old FlkSwitch mice. Statistics: one-way ANOVA, Tukey’s multiple comparisons test. **P<0.001, **** P<0.0001

B. Tom+ oPlts undergo more rapid ESAM neo-exposure compared to GFP+ oPlts and yPlts in response to 0.2 U/mL thrombin. Real time flow cytometry analysis of ESAM exposure upon activation ex vivo. Representative flow plot and corresponding quantification of MFI rate and kinetics. Data represent 3 independent experiments with n=3 young WT, and n=8 old FlkSwitch mice. Statistics: one-way ANOVA, Tukey’s multiple comparisons test. *P<0.05, ** P<0.001

C. Degranulation of Tom+ oPlts, GFP+ oPlts, and yPlts during activation is equivalent. Real time flow cytometry analysis of side scatter measurements of Plts upon stimulation. Representative flow plot and corresponding quantification of MFI rate and kinetics. Data represent 4 independent experiments with n=4 young WT, and n=10 old FlkSwitch mice. Statistics: one-way ANOVA, Tukey’s multiple comparisons test.

D. Representative images of adherent and spreading Plts on immobilized fibrinogen at 5, 15, 30, and 45 minutes after 0.2 U/mL thrombin stimulation.

E. Tom+ and GFP+ oPlts undergo four distinct phases of Plt spreading. Representative images of (I) adhesion, (II) filopodia, (III) filopodia and lamellipodia and (IV) only lamellipodia.

F. Tom+ oPlts more efficiently undergo shape change to achieve full spreading upon stimulation compared to GFP+ oPlts. Distribution of GFP+ and Tom+ oPlt phases of spreading in 5, 15, 30, and 45 minutes after thrombin stimulation. Data represent a summary of multiple images at each time point: 5 min=7 images/mouse, 15 min=7 images/mouse, 30 min=5 images/mouse, 45 min=5 images/mouse. Statistics: two-Way ANOVA, Sidak’s multiple comparisons test *P≤0.05, **P<0.001, ***P<0.0001

G. Tom+ and GFP+ oPlts generate equivalent numbers of filopodia structures. Enumeration of filopodia/Plt at phase II of spreading. Each data point represents one platelet.

H. Tom+ oPlts exhibit greater spreading capacity compared to GFP+ oPlts. Enumeration of average spreading area per Plt at 5, 15, 30, and 45 minutes after thrombin stimulation. Data represent 4 independent experiments with n=5 old FlkSwitch mice. Each data point is average of multiple images at each time point: 5 min=7 images/mouse, 15 min=7 images/mouse, 30 min=5 images/mouse, 45 min=5 images/mouse. Statistics: two-way ANOVA, Sidak’s multiple comparisons test *P≤0.05, **P<0.001.

Plt activation also triggers shape-changing cytoskeletal rearrangements to mediate thrombus formation and stabilization. Therefore, we asked whether Tom+ oPlts are also functionally divergent from the GFP+ oPlts in their dynamic capacity to spread on immobilized fibrinogen. Upon thrombin stimulation, GFP+ and Tom+ oPlts underwent the four distinct phases of Plt spreading: (I) adhesion, (II) formation of filopodia, (III) filopodia and lamellipodia, and (IV) only lamellipodia (Figure 7D-F). Remarkably, compared to GFP+ oPlts, the Tom+ oPlts progressed towards the later phases of spreading (phases III and IV) more rapidly throughout the course of the experiment (Figure 7F). Additionally, while both oPlts generated similar numbers of filopodia during spreading (Figure 7G), the Tom+ oPlts exhibited greater total spreading area compared to GFP+ oPlts (Figure 7H), substantiating the more rapid response quantified in Figure 6F. Collectively, these experiments revealed that compared to the more tempered reactivity of GFP+ Plts, the age-enriched Tom+ Plts serve as the pro-thrombotic culprits of age-related hyper-reactivity.

DISCUSSION

Discovery of a platelet-specific HSC differentiation path in aging

We utilized the unmanipulated, native hematopoietic system of the FlkSwitch mouse to investigate HSC lineage output from young adulthood into aging. Our study revealed that only the Plt lineage deviates from the classical Flk2+ pathway during aging, with Tom+ HSCs directly differentiating into accumulating Tom+ MkPs that generate pro-thrombotic Tom+ Plts by bypassing the co-existing canonical differentiation cascade of Flk2Cre-marked GFP+ progenitors. Consequently, the aging-induced Plt differentiation pathway causes rapid thrombopoiesis from a stress-responsive pool of aging-enriched MkPs that leads to the production of a second Plt population that causes steady-state thrombocytosis, exacerbated injury-induced thrombosis, and hyper-reactivity in response to Plt agonists.

Native versus stress-induced megakaryopoiesis

Functional and molecular heterogeneity of HSPCs has been well documented ^{30,35,65–67}, with stress and aging implicated in selectively promoting megakaryopoiesis ^12,68. Previous reports have concluded that HSCs may be primed for rapid and unilineage differentiation into Plt-committed cells ^{10–12,20,30,69}. Here, we unequivocally demonstrate that a bypass path to Plt production progressively develop upon aging (Figure 1). In contrast, a direct HSC-MkP path in young mice is incompatible with lineage tracing data by us and others (Figure 1) ^13,14. This aging-dependency is further supported by the absence of functionally enhanced MkPs in young mice (Figure 4 and S6). Potential reasons for perceived detection of direct HSC-MkP differentiation in other studies include the use of in vitro differentiation systems; inferring differentiation from transcriptional priming; reliance complex mathematical models; and use of inducible systems that cause HSC proliferation and potential lineage bias. Both polyinosinic:polycytidylic acid and tamoxifen, two of the most commonly used agents of inducible lineage tracing systems, have been reported to selectively induce HSC cycling and differentiation into the megakaryocyte lineage ^12,70,71. The strength of our model lies in its simplicity: we simply enumerate functionally defined cells as Tom+ or GFP+ without inducing perturbations and without making assumptions or inferences by mathematical modeling (Figure 1). Cells that are GFP+ have (at some point) expressed sufficient levels of Flk2-regulated Cre recombinase; cells that remain Tom+ have not. Under native conditions in young adult mice, HSCs differentiate via canonical progenitor cells; aging may serve as a stress mechanism that progressively activates a latent, but primed, transcriptional program that ultimately triggers the bypass Plt path (Figure 1). Thus, two functionally and molecularly distinct subsets of aged MkPs, independently produced by HSCs via two separate but parallel differentiation paths, mediate Plt production during aging. We showed that thrombocytosis is driven by age-induced heterogeneity of the MkP population (Figure 6A), and that an increasing proportion of the Plt homeostatic responsibilities of HSCs appear to be delegated to MkPs upon aging (Figures 4 and 5), possibly due to the age-related functional decline of HSCs (Figure S3). Our prospective isolation of age-enriched MkPs pinpoints the cellular mechanism of our recently reported increase in MkP repopulation capacity upon aging ^25,26 and identifies the ageenriched Tom+ MkPs as a potent source of thrombopoiesis.

Unexpected microenvironment-independent mechanisms of stem and progenitor cell aging

Surprisingly, the canonical GFP+ MkPs maintain youthful properties throughout life, with relatively attenuated expansion and engraftment capacities that are indistinguishable from young GFP+ MkPs. The functional similarities of the young and aged GFP+ MkPs were unexpected, as we detected significant changes in gene expression (Figures 2 and 3) and because the well-documented changes to the BM environment were expected to affect cell function ^38,72–76. Similar resilience to external factors has recently been reported for HSCs upon attempts to influence HSC aging by several heterochronic strategies ⁷⁷. In stark contrast to the age-resilient GFP+ oMkPs, the Tom+ oMkPs, despite co-existing in the same aged environment as the GFP+ oMkPs, have dramatically enhanced Plt production capacity in vitro and upon transplantation, and in response to Plt demand (Figures 4 and 5). This uncouples the aging microenvironment from MkP functional capacity and demonstrates that MkP properties are influenced to a greater extent by their differentiation path than by the environment of the aged BM. In addition to uncovering hallmarks of megakaryopoiesis, the surprising gain of function of age-enriched MkPs confers a new perspective on the potential mechanisms behind the well-established functional decline of oHSCs.

The aging-induced platelet differentiation pathway confers pro-thrombotic functional consequences on platelet progeny

Elevated and biased Plt production is a hallmark of aging hematopoiesis, but its consequences in hemostatic function remain unclear. Here, we demonstrate profound functional differences between young and old Plts and show that this effect is mediated by the age-enriched Tom+ Plts (Figures 6 and 7). The production of Tom+ Plts directly contributes to mechanisms of activation, spreading, and clot formation upon vascular damage. Our observations show that Tom+ Plts from old mice are functionally hyperreactive across various clotting stages, which is crucial for platelet-platelet interactions and thrombus stabilization and known to accelerate thrombus formation and reduce bleeding time. These changes drive impaired hemostasis in the aged blood. Together, our identification of dual differentiation pathways for Plt production during aging reveal a significant role of the aging-induced Plt pathway to harbor unique mechanisms that elevate thrombotic potential upon the process of aging.

Controlling aging-induced thrombosis via stem and progenitor cells

While drastic increases in Plt dysregulation and adverse thrombotic events in aging populations have been clear for decades, distinct differentiation paths and cell populations upon aging have not been envisioned as underlying mechanisms of disease susceptibility. The emergence of Plt subpopulations during aging provides evidence that Plt heterogeneity is a determinant of age-related Plt diseases. Remarkably, functional Plts can be produced via two alternative pathways, one of which is aging-induced and decoupled from erythropoiesis. The production of mature cells via distinct differentiation paths offers a paradigm of stem cell aging that is currently unexplored. Our identification of the cellular origins and mechanisms of aging-enriched Plts provides compelling therapeutic opportunities for targeting hematopoietic stem cells and megakaryocyte progenitors to control both Plt generation and functional reactivity throughout life. Therefore, our findings may profoundly impact the millions of elderly people at risk for experiencing adverse thrombotic events.

Limitations of the Study

Our discoveries utilized double-transgenic, aged mice, making it difficult and time-consuming to reproduce the findings in wild-type (wt) strains or to investigate the role of specific regulators by mouse genetics strategies. Additionally, more stringent definition of MkPs in the aged mice is needed to determine whether the transient erythroid and myeloid repopulation from oMkPs is due to contamination of co-transplanted cells within the MkP phenotypic pool or whether oMkPs gain additional lineage potential at the clonal level upon aging that is apparent only upon transplantation. Moreover, single-cell trajectory analysis did not reveal two distinct paths from HSCs to MkPs, despite the unequivocal evidence from lineage tracing. Improved computational algorithms may overcome this shortcoming in the future.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact

• Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Camilla Forsberg (cforsber@ucsc.edu).

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

• The bulk RNA and scRNAseq datasets have been deposited in the Gene Expression Omnibus (GEO) and will be publicly available upon publication. Accession numbers are listed in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

KEY RESOURCES TABLE

REAGENT OR RESOURCE	SOURCE	IDENTIFIER
Antibodies
Alexa Fluor® 488 Rat anti-mouse Ly-6C Antibody	BioLegend	128021
Alexa Fluor® 700 Rat anti-mouse CD48 Antibody	BioLegend	103426
BD Calibrite-APC Beads	BD Biosciences	340487
Anti-mouse CD45.2-Pacific Blue	BioLegend	109820
Rat anti-mouse F4/80 Biotin	BioLegend	123106
Rat anti-mouse F4/80 BV605	BioLegend	123133
fibrinogen antibody	Novus Biologicals	NBP1−33582
Goat anti-Rat PE-Cy5	Invitrogen	A-10691
GPIbalpha (CD42b) - Dylight 649	Emfret Analytics	M040−3
JON/A Rat anti-mouse αIIb/β3	Emfret Analytics	M023−2
Rat anti-Mouse	BioLegend	124212
Rat anti-Mouse	BioLegend	100702
Rat anti-Mouse B220 - A700	BioLegend	103231
Rat anti-Mouse B220 - Purified	BioLegend	103201
Rat anti-Mouse B220-APC-Cy7	BioLegend	103224
Rat anti-Mouse B220-BV605	BioLegend	103244
Rat anti-mouse B220-Pacific Blue	BioLegend	103227
Rat anti-Mouse CD105-Pacific Blue	Biolegend	120412
Rat anti-Mouse CD11c APC/Cy7	BioLegend	117323
Rat anti-Mouse CD150-BV786	BioLegend	115937
Rat anti-Mouse CD150-PECy7	BioLegend	115914
Rat anti-Mouse CD19-BV785	BioLegend	115543
Rat anti-Mouse CD3 - Purified	BioLegend	155602
Rat anti-Mouse CD3 − A700	BioLegend	100216
Rat anti-Mouse CD3-APC	eBioscience	17−0031-82
Rat anti-Mouse CD3-PE	BioLegend	100308
Rat anti-Mouse CD34-Bio	ebioscience	13−0341-82
Rat anti-Mouse CD4 -A700	BioLegend	100429
Rat anti-Mouse CD4 -Purified	BioLegend	100401
Rat anti-mouse CD41-Alexa Fluor700	BioLegend	133926
Rat anti-Mouse CD41-APC	BioLegend	133914
Rat anti-Mouse CD41-APC Cy7	BioLegend	133927
Rat anti-mouse CD41-BV421	BioLegend	133912
Rat anti-Mouse CD41-PECy7	BioLegend	133916
Rat anti-Mouse CD42c-DyLight649	Emfret Analytics	X649
Rat anti-Mouse CD48 BV785	BioLegend	103449
Rat anti-Mouse CD48-APC	BioLegend	103412
Rat anti-Mouse CD5 - A700	BioLegend	100635
Rat anti-Mouse CD5 - Purified	BioLegend	100602
Rat anti-Mouse CD61 PeCy7	BioLegend	104318
Rat anti-Mouse CD61-Alexa 647	BioLegend	104314
Rat anti-Mouse CD71 Biotin	BioLegend	113803
Rat anti-Mouse CD8 - A700	BioLegend	100730
Rat anti-mouse fibrin-AF647	Holinstat Lab	N/A
Rat anti-Mouse CD9 A647	BioLegend	124809
Rat anti-Mouse CD9-Biotin	BioLegend	124803
Rat anti-Mouse cKit-APC Cy7	BioLegend	105826
Rat anti-Mouse cKit-PECy7	BioLegend	105814
Rat anti-mouse ESAM-APC	BioLegend	136207
Rat anti-Mouse Flk2-PE	eBioscience	12−1351-83
Rat anti-mouse GPIbα	Emfret Analytics	R300
Rat anti-Mouse GPIX (CD42a) - purified	Emfret Analytics	M051−0
Rat anti-Mouse Gr-1 - Purified	BioLegend	108401
Rat anti-Mouse Gr-1 -A700	BioLegend	108421
Rat anti-Mouse Gr1-Pacific Blue	BioLegend	108430
Rat anti-Mouse Ly6c A700	BioLegend	128023
Rat anti-mouse Ly6C mouse - Pacific Blue	BioLegend	128013
Rat anti-mouse Ly6G APC	BioLegend	127613
Rat anti-Mouse Mac-1 - Purified	BioLegend	101201
Rat anti-Mouse Mac-1(CD11b) - A700	BioLegend	101222
Rat anti-Mouse Mac1- PECy7	BioLegend	101216
Rat anti-Mouse Sca-1 A700	BioLegend	108142
Rat anti-Mouse Sca-1 APC	BioLegend	108112
Rat anti-Mouse Sca1-PB	BioLegend	122520
Rat anti-Mouse Slam-PECy7	BioLegend	115914
Rat anti-Mouse Ter119 - A700	BioLegend	116220
Rat anti-Mouse Ter119 − Purified	BioLegend	116201
Rat anti-Mouse Ter119-PECy5	BioLegend	116210
Rat anti-mouse CD16/32 PECy7	BioLegend	101318
Rat anti-mouse CD41 PECy7	BioLegend	133916
Rat anti-mouse CD49b PECy7	BioLegend	103518
Rat anti-mouse cKit APC/Cy7	BioLegend	105826
Rat anti-mouse IL7ra PeCy7	BioLegend	135014
Rat anti-mouse P-selectin-PECy7	BioLegend	148309
Strepavidin BV605	BioLegend	405229
Streptavidin BV 785	BioLegend	405249
CD117-Miltenyi Beads	Miltenyi	130−091-2
Bacterial and virus strains
Biological samples
Chemicals, Peptides, and Recombinant Proteins
IMDM+GlutaMax	Fisher	31980−030
Equal Fetal	Atlas	EF-0500-A
rmTPO	Peprotech	315−14 - 50ug
rmIL-6	Peprotech	216−16
rmSCF	Peprotech	250−03
rm IL-3	Peprotech	213−13
Primocin	Invivogen	ant-pm-2
non-essential amino acids	Invitrogen/Gibco	11140−050
normal mouse serum	Invitrogen	10410
bovine calf serum	Invitrogen/Gibco	BP1600
normal rat serum	Invitrogen/Gibco	10710C
HyClone™ Dulbecco’s Phosphate Buffered Saline (DPBS) solutions,10X	VWR	82013−31
Thrombin	Sigma	T4648−1KU
500 μl 1-step Fix/Lyse Solution	Invitrogen	00−5333-57
adenosine diphosphate (ADP)	Sigma	A2754−1G
porcine heparin	Thermo Scientific Chemicals	AC411210010
human fibrinogen	Sigma-Aldrich	F3879
PFA, paraformaldeyde	Electron Microscopy Sciences	15712-S
DMEM high glucose no glutamax	Invitrogen	11960069
Trizol Reagent	Thermo fisher	15596018
Heparin sodium	Thermo Scientific™	AC411210010
Hoechst 33342, Trihydrochloride, Trihydrate	ThermoFisher	H3570
Tyrode’s Solution (with HEPES & 0.25% BSA, without Calcium) - #BSS-365	Boston Bioproduct	BSS-365
Tyrodes solution	Sigma	T1788−100ML
Critical commercial assays
Nextera Library Prep	Illumina	FC-131−10
Illumina HiSeq 4000	Illumina	QB3-Berkeley Genomics at University of California Berkeley
3’ CellPlex Kit Set A	10X Genomics	1000261
Chromium Next GEM Single Cell 3ʹ Kit v3.1	10X Genomics	1000269
Novaseq 6000	Illumina	University of California Davis DNA Technologies and Expression Analysis Core
Click-iT® EdU Flow Cytometry Assay Kit	Life Technologies	C10634
Deposited data
Bulk RNAseq data	GEO	GSE166704, GSE226318
SCRNAseq data	GEO	GSE255019
Experimental models: Cell lines
Experimental models: Organisms/strains
C57Bl6	JAX	000664
C57Bl6 from the NIH aging colony	NIH-ROS	N/A
C57BL/6-Tg(UBC-GFP)30Scha/J	JAX	004353
Flk2-Cre	T. Boehm laboratory, reference number 78, obtained under fully executed Material Transfer Agreement	N/A
B6.129(Cg)-Gt(ROSA)26Sortm4(ACTBtdTomato,-EGFP)Luo/J (mTmG) (also called ROSAmT/mG)	JAX	007676
Flk2-Cre x mTmG; Flk2-Cre males (Benz et al., 2008) crossed to homozygous RosamT/mG females (Jax Strain 007676)	In house	N/A
Oligonucleotides
Recombinant DNA
Software and algorithms
Cell Ranger Cloud pipeline	10X Genomics	https://www.10xgenomics.com/support/software/cell-ranger/latest
Scanpy 1.9.6	Reference #88	https://scanpy.readthedocs.io/en/stable/
Seurat 4.3	Reference #96	https://github.com/satijalab/seurat/releases
DESeq2	Reference #97	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
Diffusion PseudoTime Inference	Reference #49	http://www.helmholtz-muenchen.de/icb/dpt
Other
High Sensitivity chip	Agilent Technologies	5067−4626
heparinized-capillaries	Hirschmann	9000205
EDTA-coated capillaries	Sarstedt	19.447.001

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mouse lines

All animals were housed and bred in the AAALAC accredited vivarium at University of California Santa Cruz and maintained under approved IACUC guidelines and overseen by a veterinarian, a vivarium manager, and trained personnel. The UCSC vivarium is a barrier facility, with mice housed in limited access, temperature-controlled rooms in microisolator cages with no more than five mice per cage for adult animals. Cages were inspected by trained personnel daily to ensure animal health, and ad libitum access to water and food. The following mice were utilized for these experiments: C57Bl6 (JAX, cat# 664), aged C57Bl6 (NIH-ROS), UBC-GFP (JAX, cat# 004353), and male FlkSwitch mice (Flt3-Cre x mTmG mice) ^78,79. Young mice were between 8–16 weeks of age and old mice were 20+ months of age, except old transplant recipient mice which were 18+ months of age. All WT C57Bl6 and UBC-GFP mice were randomized based on sex.

METHOD DETAILS

Flow Cytometry

Bone marrow stem and progenitor cell populations and mature cell subsets were prepared and stained as previously described ^{15,25,28,80–82}. Briefly, the long bones (femur and tibia) from mice were isolated, crushed with a mortar and pestle, filtered through a 70μm nylon filter and pelleted by centrifugation to obtain a single cell suspension. Cell labeling was performed on ice in 1X PBS with 5 mM EDTA and 2% FBS. HSCs (Lin- cKit+ Sca1+ Flk2- CD150+ Tom+) or MkPs (Lin-cKit+ Sca1-CD41+ CD150+ Tom+ or GFP+ or Lin- cKit+ Sca1- CD41+ CD150+ CD48+ or CD48^lo/−) from young or old FlkSwitch and WT mice were analyzed from unfractionated samples or isolated from c-Kit-enriched BM with CD117-microbeads (Miltenyi) using a FACS ARIA II (Becton Dickinson, San Jose, CA) as previously described ^13,25,28. Tissue-resident macrophages from brain and lung were analyzed as described ³². Brain microglia were analyzed as Live, CD45+ F4/80^hi CD11b^hi Ly6g- CD11c- and lung alveolar macrophages were analyzed as CD45+ F4/80^hi CD11b^mid SiglecF+. Cells with no history of Cre expression were defined as Tom+GFP- (“unfloxed”) cells, whereas cells with current or a history of Cre expression were defined solely by GFP expression, based on our previous demonstration that Tom+GFP+ cells have excised the loxP-flanked Tomato cassette ^13,15,16,18; that Tom+GFP+ and Tom-GFP+ cells are functionally indistinguishable ^13,15; and well-accepted field standards of loxP-stop-loxP-inducible single-color reporters ^14,53,83,84.

In vitro MkP Expansion

MkPs were sorted as described above (1000 per well from FlkSwitch mice or 2000 per cell from WT mice in 96-well U-bottom tissue culture plates) were cultured for 3 days in 200 μl/well containing IMDM medium (Fisher) supplemented with 10% FBS, 20 ng/ml rmTPO, 20 ng/ml rmIL-6, 50 ng/ml of rmSCF, and 5ng/ml rmIL-3 (cytokines from Peprotech), 1X Primocin (Invivogen), and 1X non-essential amino acids (Gibco). On day of analysis, a known number of APC-labeled spherobeads (BD Bioscience) were added. Cells were stained as described above and data was collected on either LSR II, FACSAria II (Beckton Dickinson) or CytoFlex LX (Beckman Coulter). Analysis was performed in FlowJo V9 or V10 (Beckton Dickinson). Cell expansion was calculated based on the number of beads recovered per beads added per well, as previously described ²⁵.

CFU-S Analysis

CFU-S experiments and analysis were performed as previously done ²⁸. 2,000 MkPs from young or old FlkSwitch mice were sorted and transplanted into lethally irradiated C57Bl6 mice. At day 8.5 post-transplant, spleens were harvested and visually analyzed for CFU-S formation. Images of spleens were acquired using Zeiss Axiozoom.

Transplantation Reconstitution assays

Reconstitution assays of FlkSwitch cells were performed by transplanting HSCs (200 per recipient) or MkPs (22,000 per recipient) from young or old mice into congenic, sublethally irradiated C57Bl6 mice via retro-orbital intravenous transplant as previously described ^25,85,86. Reconstitution of WT MkPs was performed by transplanting 22,000 CD48+ or CD48^lo/− MkPs from young male and female mice into sublethally irradiated (5 Gy) male and female UBC-GFP hosts via retro-orbital intravenous injection. 1X HBSS (Gibco) devoid of cells was instead injected for the Sham controls. Following transplantation, mice were bled via the tail vein at the indicated time points. Donor chimerism of Plts (single, FSC^low CD11b- Gr1- CD3- B220- Ter119-CD41+) and erythroid cells (single, CD11b- Gr1- CD3- B220- Ter119+ CD41-) was determined on the whole blood fraction. Donor chimerism of B cells (single, live, CD11b- Gr1- CD3 B220+), T cells (single, live, CD11b- Gr1- B220- CD3+), and GMs (single, live, CD3- B220- CD11b+ Gr1+) were quantified following ACK lysis (0.15 M NH₄Cl, 1 mM KHCO₃, and 0.1mM Na₂EDTA) of whole blood. Acquisition was performed on LSR II (Beckton Dickinson) or CytoFlex LX (Beckman Coulter) and analysis via FlowJo V10 (Becton Dickinson).

Bulk RNA-Sequencing and Analysis

The RNA-Seq libraries were generated from purified MkPs from young or old FlkSwitch mice. RNA-Seq libraries were generated using Nextera Library Prep, as we have previously done ^15,25,87. Young MkP samples analyzed previously²⁵ were compared to oMkP samples separated by Tom and GFP expression, and to oHSCs. Libraries were validated using the Bioanalyzer (Agilent 2100), sequenced using Illumina HiSeq 4000 as Paired-end reads at the QB3-Berkeley Genomics at University of California Berkeley and DESeq2 analysis was done with the help of Dr. Sol Katzman at the UCSC Bioinformatics Core. Gene Ontology (GO) enrichment analysis was performed with PANTHER classification system using list of DEGs by Tom+ MkPs relative to GFP+ MkPs extracted from DESeq2. Normalized RNA abundance counts where extracted from DESeq performed as described above and used to calculate Principal Components (PCs) with the `prcomp()` function in `R` with `rank` set to 50, resulting in 50 components that summate to 100% of the observed variance in the data. PCs 1 and 2 were plotted and used to calculate 1) centroids for each cell type by averaging the sample values in PC1 and PC2 space and 2) Euclidean distances from each centroid to the one calculated for Old HSCs. Furthermore, these normalized counts were scaled to Z-scores using the `scale()` function in `R` with `center` set to `TRUÈ. Z-scores were used to create a K-means clustered heatmap with `ComplexHeatmap` in `R`. Heatmaps were created on subsets of genes as described in relevant figures with log2FoldChange or Principal Component loading displayed as bar plots where relevant.

Single-Cell RNA sequencing

Cell preparation and sorting:

Cells were harvested and cKit-enriched from the long bones and hips of the hind legs of young and old mice as described above. cKit-enriched cells were then stained for lineage markers via a purified rat-anti-mouse antibody cocktail (CD3, CD4, CD5, CD8, CD11b, B220, Gr-1, and Ter119) on ice for 30 minutes. Cells were washed three times followed by incubation with goat-anti-rat Alexa Fluor 680 secondary on ice in the dark for 20 minutes. Cells were washed three times then resuspended in blocking solution (10% bovine calf serum (BCS), 10% normal mouse serum (Invitrogen), and 10% normal rat serum (Invitrogen) in 1X PBS (Cytiva). Cells were blocked for 10 minutes on ice in the dark. Following block, directly conjugated antibodies were directly added to the cells (CD117/cKit and CD150) and incubated on ice in the dark for 45 minutes, gently mixing the cells at the 30-minute mark. A small aliquot of cells was also used to perform a CD150 fluorescence minus one (FMO) control. Cells were washed three times, resuspended in 10% BCS in 1X PBS (wash buffer), with three drops of propidium iodide (diluted 1:5000 in wash buffer) added. Live, lineage-cKit+CD150- and live, lineage-cKit+CD150+ young and old cells were sorted on a BD FACSAria II (BD Biosciences) (Figure S5). The CD150- and CD150+ gates were touching to ensure a continuum of cells across CD150 expression were sorted. Cells were sorted into chilled 1X PBS (without Mg2+) containing 20% BCS and checked for sort purity. Following sorting, cells were washed and each of the four sorted populations (young CD150-, young CD150+, old CD150-, and old CD150+) were split into three aliquots for Cell Multiplexing Oligo (CMO) labeling (10X Genomics) following manufacturer’s instructions, with each individual group of cells receiving a single unique CMO tag. Cells were washed three times with wash buffer and manually counted via a hemocytometer. When possible, equal numbers of cells from each of the 12 CMO labeled aliquots were combined to artificially inflate the numbers of rare CD150+ cells and to equalize numbers between young and old mice. The pooled cells were then split into two replicates for further processing.

Sample sequencing and preparation:

Both pooled replicates were used in parallel in the Chromium Next GEM single Cell 3’ v3.1 protocol (10X Genomics) following manufacturer’s directions. GEM generation and barcoding was performed immediately post-sorting, with resulting samples stored at −20ºC per manufacturer’s instructions. All other steps were performed simultaneously for all samples from each of the two independent experiments. Following library construction, samples were assessed via an Agilent Bioanalyzer with a High Sensitivity chip (Agilent Technologies). Final sample libraries were sequenced on a Novaseq 6000 (Illumina) at the University of California Davis DNA Technologies and Expression Analysis Core, targeting 50,000 read pairs per cell for the gene expression libraries and 5,000 reads per cell for the CMO libraries.

Data analysis – quality control and data processing.

Sequencing data was initially processed via the 10X Genomics Cell Ranger Cloud pipeline, using the mouse genome (mm10) modified to allow for detection of reporter transgene RNA expression. Egfp and TdTom reference sequences were manually added to the mm10 mouse genome and uploaded to the 10X cloud-based Cell Ranger platform for alignment, but were regressed out prior to clustering. For the downstream analysis, cells with expression of fewer than 200 genes were excluded, and genes detected in fewer than three cells were removed from subsequent analysis steps. Mitochondrial and ribosomal gene transcripts, as well as Malat1 transcripts, were removed to reduce technical bias. This resulted in a dataset of 22,790 cells and 22,870 genes. Additionally, 13 cells with absent Kit RNA were filtered out. Doublets identified by DoubletDetection 2.4.0 (p_thresh=1e-16, voter_thresh=0.5) removed 157 cells in total (Gayoso et al. 2018). The UMI counts for each cell were normalized by the total counts across all genes (target_sum=1e4), and log-transformed with an added pseudo count of 1. Single-gene drop-out rates were estimated to be ~3% based on Tomato/GFP double-negative cells that passed all other criteria (800 cell out of 22,607 cells had neither Tomato or GFP transcripts). This low drop-out rate is unlikely to significantly influence data analyses.

Data analysis – Dimension reduction and unsupervised clustering.

The dimensionality reduction was conducted using UMAP to integrate the neighborhood graph, which preceded clustering using the Leiden algorithm. These analytical steps were synthesized in the Scanpy 1.9.6 framework ⁸⁸. We conducted iterative clustering to identify differentially expressed genes to obtain consensus annotation. Leiden clusters were generated at a resolution of 0.6 using total data set. Sub-clustering was performed with resolutions of 0.2 to separate the myeloid, megakaryocyte progenitors and lymphoid groups. Expression of cell cycle genes can confound the clustering of cells ⁸⁹, and therefore, we regressed out the effect of cell cycle genes (as well as Egfp and tdTomato) for the unsupervised clustering step.

Data analysis – GFP group classification using the Gaussian Mixture Model.

Normalized and log-transformed Egfp expression showed a bimodal distribution across young and old samples in our dataset. We applied a two-component Gaussian Mixture Model (GMM) to the Egfp data via the GaussianMixture function in Scikit-learn 1.3.2 (Fabian et al. 2011). From the inferred mean values of the two components, we derived the GFP positive and negative group threshold as the mean of these components. We note that GFP negative cells represent the Tom+ population.

Data analysis – Differentially expressed genes (DEGs) analyses.

Clusters were manually annotated from the normalized, log-transformed dataset based on distinct differentially expressed marker genes, using the sc.tl.rank_genes_groups function that employs the t-test_overestim_var method in the Scanpy pipeline. Once cells were clustered, differential gene expression was performed by comparing a single cluster against the remainder of the cells. Genes with a higher representation in single clusters by this method were examined. Known lineage-defining genes identified lead to initial cluster identification, with additional known gene profiles examined to confirm.

Cell population	Marker genes used for annotations47,90–94
HSC	Procr, Mecom, Robo4, Cish, Meis1, Ryk
MPP	Hlf, Gcnt2, Cd27
MkP	Pf4, Vwf, Itga2b, Itgb3, Cd9, Mpl
Lymphoid	Il7r, Dntt, Flt3, Ebf1, Ly6d, Cd19, VpreB
CMP/other	Gata2, Itga2b, Angpt1, Il6, Cd209a
GMP/neutrophil progenitor	Csf1r, Elane, Mpo, Cebpa, Cebpe, Tnfrsf1a, Irf8
MEP	Gata2, Pf4, Angpt1, Cd9
Early Erythroid	Klf1, Grb10, Gata1, Cd59a
Erythroid	Klf1, Hba-a1, Gata1, Epor, Gypa
DC progenitor	Cd209a, Irf8, Batf3, Id2

Data analysis – Pseudobulk DEGs comparison.

To conduct pseudobulk differential expression analysis for pairwise comparison, we restructured the single-cell dataset into sample-level datasets ⁹⁵. We extracted the sample and cell type clusters for young GFP+, old GFP+, and old Tom+ MkPs and aggregated single-cell level counts at the sample level for each cell type cluster. The dataset was aggregated across cluster and sample groups using the aggregateExpression function in Seurat 4.3 ⁹⁶. We retained genes with an aggregate count exceeding 10 across all samples, then normalized gene expression counts to account for variations in sequencing depth. The resulting data were used as input for DESeq2 count matrix ⁹⁷ and comparison to our true bulk RNAseq datasets.

Data analysis – Pseudotime and trajectory inference.

Unsupervised ordering of cells was performed using the diffusion pseudo-time (DPT) inference method ⁴⁹. To assess the gene expression changes underlying stem cell differentiation, we considered a discrete compartments model, using the HSC annotated cluster as root cells, and n=10 for the number of diffusion components. To investigate the progression from HSCs to various differentiated states, the PAGA method ⁵⁰ was applied through the Scanpy platform to model their potential developmental trajectories.

Laser-induced cremaster arteriole thrombosis model

The laser-induced thrombosis assays were performed as previously described ^59,60,98,99. Briefly, the cremaster muscle of anesthetized Young (8–12 weeks of age) or Old (>22 months of age) FlkSwitch mice were prepared under a dissecting microscope with constant superfusion of 37°C bicarbonatebuffered saline. Injury of the cremaster arterioles (30–50 μm diameter) was performed by a laser ablation system (Ablate! photo-ablation system; Intelligent Imaging Innovations, Denver,CO). Multiple laser injuries were performed in each mouse, with each new injury induced upstream of prior injuries. Images of thrombus formation were taken at 0.2-s intervals using a Zeiss Axio Examiner Z1 fluorescent microscope with a 6×3 objective and a high-speed sCMOS camera. The two populations of platelets were distinguished by the expression of Tomato or GFP fluorescence. Images were analyzed using Slidebook 6.0 (Intelligent Imaging Innovations). Alexa Fluor 647- conjugated anti-fibrin (0.3 μg/g) administered by a jugular vein cannula prior to vascular injury. All captured images were analyzed for the change of fluorescent intensity over the course of thrombus formation after subtracting fluorescent background defined on an uninjured section of the vessel using the Slidebook program.

Platelet-Leukocyte Aggregation Analysis

10 μl of heparinized blood was aliquoted into an antibody cocktail containing CD41-APC, B220-BV605, Ter119-A700, GR-1 PB and incubated at room temperature in the dark for 20 minutes. In samples to be stimulated by Thrombin, Thrombin (0.1 Units/ml) was added to the cocktail and incubated for 5 minutes at room temperature and protected from light. 500 μl 1-step Fix/Lyse Solution (Invitrogen # 00–5333-57) was added to all samples and allowed to incubate another 30 minutes in the dark at room temperature. Samples were analyzed by flow cytometry within 2 hours of obtaining blood samples.

In vivo Platelet Lifespan Analysis

Fluorescently conjugated anti-GPIbβ antibodies (CD42c-DyLight649, Emfret, clone NA, #X649) and anti-B220-PB (BioLegend, clone RA3–6B2, #103227) were administered (0.1 μg/ g body weight, IV) retro-orbitally to >20 months old male FlkSwitch mice. Platelet and B-cell labeling efficacy was determined 2 hours post-injection from 5 μL retro-orbitally (heparinized-capillaries, 5 μL, Hirschmann, #9000205) sampled heparin-blood (0.3 mg/mL, Thermo Scientific Chemicals, #AC411210010).Blood was washed (800x g, 5 min, RT) two times in Tyrode’s buffer without calcium and stained 30 min/RT with anti-CD41-Alexa Fluor700 antibodies (1:200, BioLegend, clone MWReg30, #133926) and anti-CD19-BV785 (1:200, BioLegend, clone 6D5, # 115543). Labelled platelets and B-cells were traced at indicated time points using flow cytometry (CytoFLEX LX, Beckman Coulter) after micro sampling from retro-orbital blood (5 μL).

Platelet Activation and Glycoprotein Expression Analysis

Platelets were obtained from whole blood collected in EDTA coated capillary tubes, washed with Tyrode’s buffer without calcium (137 mM NaCl, 0.3mM Na₂HPO₄, 2mM KCl, 12mM NaHCO₃, 5mM HEPES, 5mM Glucose, 0.35% BSA), and examined under resting conditions or after activation with 1 U/ml Thrombin (Sigma) or with 10 μM adenosine diphosphate (ADP, Sigma) for 5 minutes at 37°C, followed by 5 minutes at room temperature. Platelets were stained with anti-CD9, anti-αIIb/β3 (JON/A, Emfret), and anti-P-selectin and measured by flow cytometry.

Platelet Activation Neo-exposure Assay

50 μL whole blood (retro-orbital, EDTA-coated capillaries, Sarstedt #19.447.001) from male FlkSwitch or male C57BL/6J mice was collected in 300 μL porcine heparin (0.3 mg/mL, Thermo Scientific Chemicals, #AC411210010) in PBS and washed two-times (800x g, 5 min, RT) in 2 mL Tyrode’s buffer without calcium. Washed platelets in whole blood were resuspended in Tyrode’s buffer with calcium and stained with anti-CD41-BV421 antibodies (1:200, BioLegend, clone MWReg30, #133912) for 30 min at RT/dark. 5×10⁵ platelets from FlkSwitch and 5×10⁵ platelets from male C57BL/6J mice were combined into PBS containing anti-P-selectin/CD62P-PE-Cy7 antibodies (1:200, BioLegend, clone RMP-1, #148309), anti-ESAM-APC antibodies (1:200, BioLegend, clone 1G8/ESAM, #136207), 0.2 U/mL bovine thrombin (Sigma-Aldrich, #T4648–1KU) and were immediately acquired by flow cytometry for 10 min (10 μL/s, CytoFLEX LX, Beckman Coulter). Resting platelets were acquired for 10 min without thrombin/activator. P-selectin and ESAM neo-exposure kinetic (slope, median fluorescence intensity (MFI)/s) as well as degranulation (SSC-A, MFI/s) overtime and area under curve (AUC) were calculated with FlowJo (BD Bioscience, Version 10.8.1).

Platelet spreading assay

12 mm glass coverslips were coated with human fibrinogen (100 μg/mL, Sigma-Aldrich, #F3879) over night at 4°C. Coverslips were washed three-times with PBS and blocked with 1% BSA (Fisher Scientific, #BP1600) in PBS for 1 hour at room temperature (RT). Washed platelets were isolated from heparin-blood by two times washing (800x g, 5 min, RT) in Tyrode’s buffer without calcium and harvested in the supernatant after centrifugation at 150x g for 10 min in Tyrode’s buffer with calcium. Purity and number of washed CD41+ platelets were determined via flow cytometry (CytoFLEX LX, Beckman Coulter). Washed platelets were treated with bovine thrombin (0.2 U/mL) for 10 min at 37°C and then allowed to spread on fibrinogen-coated coverslips for indicated time points (5, 15, 30 and 45 min) at RT. Coverslips were washed three-times with PBS to remove non-adherent platelets and fixed with 4% PFA (Electron Microscopy Sciences, #15712-S) in PBS for 20 min. Washed coverslips were mounted (Fluoromount-G, SouthernBiotech, #0100–01) onto a microscopic slide. Images of platelets wer platelets from male C57BL/6J micee acquired with the AxioImager Z2 widefield microscope (Zeiss) using a 63x/1.4 NA oil (wd0.19mm) objective. GFP (filter set 38, EX BP 470/40: DC 495 : EM BP 525/50) and Tomato (filter set 43, EX BP 550/25 : DC 570 : EM BP 605/70)-positive platelets were acquired with 2×2 pixel binning and ApoTome mode with five phase images. Four different platelet spreading stages (stadium I; round resting; stadium II: filopodia; stadium III: filopodia and lamellipodia and stadium IV: lamellipodia) were identified in four to seven images per time point.

Platelet Depletion and Cell Cycle Analysis

Acute thrombocytopenia was induced in young and old FlkSwitch mice by depleting platelets with a single injection of anti-GPIbα (2 μg/g body weight) intraperitoneally (R300, Emfret). Retro-orbital bleeds using heparin-coated capillary tubes were performed daily to monitor platelet recovery. Platelets were measured by flow cytometry and defined as single, FSC^low Ter119- CD61+ Tom/GFP+ cells. In a separate cohort of mice, HSPCs in the BM was analyzed 24 hours post-anti-GPIbα treatment. In the cell cycle experiments, EdU (50mg/kg body weight) was administered intraperitoneally and concurrent with anti-GPIbα administration. Following a 24-hour EdU pulse, HSCs, MPPs, and MkPs were purified by flow cytometry and EdU incorporation analyses were performed according to the manufacturer’s protocols (Click-iT^® EdU Flow Cytometry Assay Kits, Life Technologies).

QUANTIFICATION AND STATISTICAL ANALYSIS

Number of experiments, n, and what n represents can be found in the legend for each figure. Statistical significance was determined by two-tailed unpaired T-test, unless noted in Figure Legends. All data are shown as mean ± standard error of the mean (SEM) representing at least three independent experiments.

Supplementary Material

1

Movie S1. In vivo laser ablation injury and real-time monitoring of thrombus formation in a Young FlkSwitch mouse. Note formation of a small clot consisting of GFP+, but not Tom+, cells, related to Figure 6.

2

Movie S2. In vivo laser ablation injury and real-time monitoring of thrombus formation in an Old FlkSwitch mouse. Note formation of a large clot consisting of both Tom+ and GFP+ cells, related to Figure 6.

3

Table S1. Old, but not young, FlkSwitch mice have MkP- and Plt-specific decreases in GFP-expressing cells, related to Figure 1. The percentage of GFP+ cells within each indicated population is shown for seven young and ten old FlkSwitch mice, indicating that every individual old mouse has decreased percentage of GFP+ MkPs and Plts.

Table S2. The top 25 up- and downregulated genes between Tom+ and GFP+ oMkPs based on bulk RNAseq, with concordance with scRNAseq indicated, related to Figures 2 and 3.

5

Excel Table S3A-F. Lists of differentially expressed genes between Tom+ and GFP+ MkPs from young and old mice, related to Figures 2 and 3.

6

Figure S1. Gating strategies for hematopoietic stem and progenitor populations in young and old FlkSwitch bone marrow, related to Figure 1.

A. Representative FACS plots of HSC, MPP, GMP, CMP, and MEP.

B. Representative FACS plots of Common Lymphoid Progenitors (CLP).

C. Representative FACS plots of MkPs, CFU-E, pCFU-E, pGMP, pMegE, and “GMP”.

D. Increased HSC and MkP frequencies in old mice. Frequency of HSCs and MkPs in Lineage- BM cells. Data represent 6 independent experiments, n = 8 young mice, n = 18 old mice. Statistics: t-test. ** P< 0.005, *** P< 0.0005

E. FACS plots representing Tom and GFP expression by HSCs and classical myeloid progenitors, and CLPs from panel A and B, respectively.

F. FACS plots representing Tom and GFP expression by MkPs and alternatively defined erythromyeloid progenitors from panel C.

7

Figure S2. Tissue resident macrophage GFP-labeling in FlkSwitch mice increases across tissues with age while multipotent progenitor GFP-labeling is maintained upon aging, related to Figure 1.

A. Flow cytometry plots defining brain microglia (CD45+F4/80^hiCD11b^hiLy6g−CD11c− cells) and GFP labeling in young and old FlkSwitch mice.

B. Quantification of GFP labeling of brain microglia in young and old FlkSwitch mice.

C. Flow cytometry plots defining tissue resident alveolar macrophages (CD45+F4/80^hiCD11b^mid SiglecF+ cells) and.GFP labeling in the lungs of young and old FlkSwitch mice.

D. Quantification of GFP labeling of tissue resident alveolar macrophages in young and old FlkSwitch mice. Data represent mean ± SEM of 3 independent experiments, n = 6 young mice, n = 12 old mice. Statistics: t-test. *P<0.05, ***P<0.0005

E-F: Gating strategies for Multipotent Progenitor subfractions in FlkSwitch mice during aging.

E. FACS plots representing the GFP and Tom expression by subfractions of Multipotent Progenitors: MPP2 (Lin-cKit+Sca1+Flk2-CD48+CD150+), MPP3 (Lin-cKit+Sca1+Flk2-CD48+CD150-), MPP4 (Lin-cKit+Sca1+Flk2+CD48+CD150-) and ST-HSC (Lin-cKit+Sca1+Flk2-CD48-CD150-).

F. The frequency of MPP subpopulations do not change substantially upon aging. Data represent 3 independent experiments, n = 5 young mice, n = 5 old mice. Statistics: t-test. Comparisons of Y vs O were not statistically different. Y, young; O, old.

8

Figure S3. Old HSCs poorly reconstitute nucleated cells and platelets compared to young HSCs and did not retain the aging-induced shortcut platelet differentiation pathway upon transplantation, related to Figure 1.

A and B. Schematic of heterochronic and isochronic transplantation of HSCs from FlkSwitch into conditioned, non-fluorescent mice. PB was monitored for GFP/Tom+ fluorescence of donor-derived mature cells, presented as %GFP in recipient mice in C and D.

C and D. Percentage of GFP+ donor-derived cells were equivalent in young (C) or old (D) recipients of transplanted young or old HSCs for >16 weeks post-transplant. Cells that are not GFP+ are Tom+. Data represent mean ± SEM of 5 independent experiments with n= 15 Y-Y mice, n= 21 O-Y mice, n= 13 Y-O mice, n= 13 O-O mice. Statistics: unpaired two-tailed t-test between Plts and GMs. *p<0.05. E and F. Percentages of donor chimerism in young (E) or old (F) recipient mice transplanted with oHSCs from FlkSwitch mice demonstrate feeble long-term reconstitution of nucleated cells (GM, B, and T cells) and Plts compared to yHSCs. Data represent mean ± SEM of 5 independent experiments with n= 15 Y-Y mice, n= 21 O-Y mice, n= 13 Y-O mice, n= 13 O-O mice. Statistics: unpaired two-tailed t-test between total cell (black) or Plt (magenta) chimerism by oHSCs and yHSCs. *p<0.05, **p<0.005, ***p<0.005.

Y-Y, Young HSCs transplanted into young recipients; O-Y, Old HSCs transplanted into young recipients; Y-O, Young HSCs transplanted into old recipients; O-O, Old HSCs transplanted into old recipients.

9

Figure S4. Heatmaps of RNAseq analysis, related to Figure 2.

A. Kmeans-clustered heatmap of gene expression Z-score for top 30 genes by absolute loading in Principal Component 1 (top) or 2 (bottom) derived for all samples.

B. Kmeans-clustered heatmap of gene expression Z-score for genes with MkP- or HSC-specific expression according to the GEXC database. MkP- or HSC- specific genesets were targeted from GEXC with the search term of “active in MkP, while inactive in all other cells” and “active in HSC, while inactive in all other cells”, respectively.

C. Levels of Sca1 (Ly6a) mRNA in GFP+ yMkPs, GFP+ oMkPs, and Tom+ oMkPs based on bulk RNAseq (left) and pseudo-bulk scRNAseq (right) readcounts. Statistical analysis was conducted with a one-way ANOVA adjusted for multiple comparisons (Tukey). *P<0.05.

10

Figure S5. Characterization of hematopoietic cell populations identified by scRNAseq analysis and MkP reconstitution of platelets, erythroid cells, GMs, B, and T cells, related to Figure 3.

A. UMAP of all HSPCs from scRNAseq colored by MK and other lineage gene expression.

B. Pseudobulk analysis of the scRNAseq data largely recapitulates the bulk RNAseq data. Volcano plots of all concordant DEGs between scRNAseq and bulk RNAseq datasets.

C. Flow cytometry analysis of PB cells from MkP transplantation experiments showing GFP-expressing donor-derived cells from GFP+ MkPs and Tomato-expressing donor-derived cells from Tom+ oMkPs.

11

Figure S6. Tom+ MkPs from old FlkSwitch mice have greater myeloid reconstitution potential compared to GFP+ Young or Old MkPs, related to Figure 4.

A. Analysis of donor-derived erythroid cells, GM, B and T cells in the PB of recipients presented as percent donor chimerism. Tom+ oMkPs demonstrated greater contribution to erythroid (beyond day 30 post-transplant) and GM (all post-transplant timepoints) donor-to-host chimerism in the recipient mice compared to both GFP+ yMkPs and GFP+ oMkPs. Little to no B and T cell chimerism was observed. Data are from the same cohorts of mice presented in Figure 4, and represent mean ± SEM of 3 independent experiments, n = 6 GFP+ yMkP recipients, n = 4 GFP+ oMkP recipients, and n = 13 Tom+ oMkP recipients. Statistics: unpaired two-tailed t-test. T-tests between GFP+ yMkP and GFP+ oMkP were not statistically significant. Erythroid chimerism: T-tests between GFP+ yMkP and Tom+ oMkP #p<0.05 at Day 14 and Day 28. T-tests between GFP+ oMkP and Tom+ oMkP *p<0.05, **p<0.005, ***p<0.0005.

B. Phenotypic MkPs can be stratified based on CD48 expression. Representative flow cytometry plot (left) and quantification of MkP CD48 subtypes (right). Data represent more than 20 independent experiments, n=39 young WT mice. Statistics: unpaired two-tailed t-test, ****P<0.0001.

C. Equivalent in vitro growth capacity by CD48+ or CD48^lo/- MkPs. Quantification of cell expansion revealed no difference between young CD48+ or CD48^lo/- MkPs. Data represent 7 independent experiments, n=13 CD48+ yMkPs and n=8 CD48^lo/- yMkPs (each data point is the average of three replicate wells). Statistics: unpaired two-tailed t-test, ns = not statistically significant. D. Equivalent Plt reconstitution by CD48+ or CD48^lo/- yMkPs. Data represent mean ± SEM of 5 independent experiments, n=8 CD48+ yMkPs, n=3 CD48^lo/- yMkPs, and n=8 Sham. Statistics: Repeated measures two-way ANOVA with the Geisser-Greenhouse correction, adjusted for multiple comparisons by Tukey’s ad hoc test.

E. Similar to GFP+ yMkPs, CD48+ or CD48^lo/- yMkPs from WT mice contribute minimally to GM, B, and T cell reconstitution following transplantation. CD48^lo/- yMkPs specifically contribute to transient erythroid cell output. Percent donor chimerism was assessed via sequential tail bleeds at the indicated time points. Data from same cohort as D. **P<0.01, ****P<0.0001 from statistics described for D.

12

Figure S7. The age-enriched Tom+ platelets contribute to exaggerated clot formation in old FlkSwitch mice and gating strategy for Platelet-Leukocyte Aggregation analysis, related to Figures 6 and 7.

A-B: Images of uninjured FlkSwitch mice (A) and laser-induced clot formation (B) in Young and Old FlkSwitch mice. Thrombi in Old mice contain both Tom+ and GFP+ cells, whereas thrombi in young mice are exclusively GFP+.

C-D. Gating strategy of platelet-leukocyte aggregates in vitro analysis to determine frequency of platelet aggregates with distinct leukocyte subtypes within stimulated murine blood. Gr1+CD41+ events (C) and B220+CD41+ events (D). Stimulation by thrombin (0.1 U/mL) increased the double-positive events in young and old blood. Platelet-leukocyte aggregation remained higher in old blood compared to young blood, with and without stimulus by thrombin.

HIGHLIGHTS.

• Aging leads to two parallel platelet specification paths from HSCs

• The shortcut platelet pathway is perpetuated by highly expansive MkPs

• Canonical MkPs are resilient to age-induced changes

• The aging-induced shortcut path causes thrombocytosis and platelet hyperreactivity