Elevated endocytic trafficking mediated by GPRASP2 maintains HSC fidelity

Abstract

Endolysosomal trafficking supports cellular homeostasis through coordinated regulation of extrinsic signaling inputs. Hematopoietic stem cell (HSC) function requires a tightly regulated balance between quiescence for long-term preservation and rapid activation for blood production. Although lysosomal regulation of metabolism and quiescence has been linked to stem cell maintenance, the contribution of endocytic routing to lysosomal function remains incompletely understood. Here, we show that quiescent HSCs rely on elevated endocytosis to maintain self-renewal. This activity is mediated in part by GPRASP2, a post-endosomal sorting protein. HSCs enriched in GPRASP2 are functionally dormant yet molecularly primed for activation however, in response to proliferative cues, dormant HSCs exhibit reduced signaling and proliferation. Disruption of GPRASP2-mediated endocytosis induces rapid proliferation and increased expression of signaling constituents, consistent with a model in which elevated endocytosis attenuates signaling through receptor internalization. Thus, we identify endocytosis as a mechanism by which HSCs limit functional exhaustion arising from chronic activation.

Introduction

Hematopoietic stem cells (HSCs) reside in the bone marrow (BM) at the apex of the hematopoietic system and give rise to all mature blood cell types^{1, 2}. HSCs must reconcile two opposing demands: long-term preservation of a self-renewing stem cell pool that supports lifelong hematopoiesis, and the ability to exit quiescence in response to hematopoietic stressors to produce downstream progeny. The phenotypically and transcriptionally defined HSC pool is molecularly and functionally heterogeneous and contains HSC subsets with distinct self-renewal capacity and lineage potential^3-6. Thus, HSCs exhibit diverse responses to hematopoietic stress, which may result from the selective activation of cells with distinct predetermined functions⁷. Indeed, a small subset of “deeply quiescent” HSCs rarely cycle or contribute to homeostatic blood production but can be functionally recruited under conditions of stress^8-10. Exquisite regulation of the balance between quiescence and cell cycle activation appears to preserve HSC integrity, as HSCs with a low division history show superior self-renewal and stemness in transplantation studies^{8, 9, 11-16}. However, how HSCs mechanistically govern the balance between quiescence and activation in response to external cues remains poorly understood.

Transcriptionally diverse HSCs support the model that cell intrinsic properties drive the response to external cues, but the mechanistic basis of this heterogeneity is only starting to come into focus⁷. Although the endolysosomal system has recently been implicated as a critical regulator of HSC function, the primary focus of this work has been the lysosome and its role in mediating efficient clearance of mitochondria and autophagosomes, which is essential for stem cell maintenance. Lysosomal activity and metabolic state are also tightly correlated in HSCs^17-19. Endocytosis can support cellular homeostasis by enabling the uptake of essential nutrients and signaling molecules, remove waste and pathogens from the extracellular space, and regulate the density of membrane proteins critical for cell function^{20, 21}. Like lysosomal activity, tight regulation of endocytosis is critical to maintaining HSC function; disruption of endocytosis by loss of caveolae or hyperactivation during ex vivo culture is associated with loss of function^{22, 23}. Further, transcription factor EB (TFEB), a master lysosomal regulator, controls lysosomal activity and results in decreased cell-surface expression of TFR1 and increased quiescence in human HSCs²². Thus, endocytosis of key cell-surface receptors and their subsequent trafficking to the lysosome may modulate and limit the cell cycle entry of HSCs in response to extrinsic cues^{24, 25}. Despite growing recognition that endocytosis and lysosomal flux are important for maintaining HSCs, little is known about how endolysosomal trafficking regulates cell surface receptor turnover in HSCs and how this pathway integrates with stemness programs.

G-protein coupled receptor (GPCR)-associated sorting proteins (GPRASPs) are post-endosomal sorting proteins that can bind the cytoplasmic tail of GPCRs and target them to the lysosome for degradation^26-28. Loss of GPRASP1 and GPRASP2 has been shown to prevent degradation of target GPCRs, accompanied by hyperactivation of downstream signaling effectors^{29, 30}. Specifically, GPRASP1 can link GPCRs to the ESCRT machinery that regulates endosome maturation and degradation via the lysosome³¹. We recently reported GPRASP1/2 as key regulators of C-X-C chemokine receptor type 4 (CXCR4) membrane availability in hematopoietic stem and progenitor cells (HSPCs) and B cells^{32, 33}. Loss of Gprasp2 expression stabilizes CXCR4 at the cell surface, enhancing the homing and engraftment capacity of HSPCs after transplantation and disrupting the normal trafficking of B cells through germinal centers. Despite its established role in regulating HSPC CXCR4 availability, the molecular basis by which GPRASP2 governs HSC behavior during homeostasis and in response to stress has yet to be determined.

Here, we demonstrate that GPRASP2 expression marks deeply quiescent HSCs with vast stemness and self-renewal potential in the heterogenous HSC pool. We further show that quiescent HSCs are molecularly and functionally enriched for endolysosomal activity. Mechanistically, loss of GPRASP2 disrupts endolysosomal activity and transcriptionally activates intermediate signaling constituents, leading to increased cell cycling and a loss of self-renewal. Further, we find that quiescent HSCs are uniquely sensitive to perturbations in endolysosomal activity compared to downstream HSPCs, reflecting a critical reliance on this pathway to maintain quiescence and self-renewal. Elevated endolysosomal activity in dormant HSCs correlated with increased selective uptake of HSC-derived extracellular vesicles previously shown to increase HSC reconstitution. Cumulatively, these data reveal endolysosomal trafficking mediated by GPRASP2 as a functional regulator of quiescent, durably self-renewing HSCs.

Results

Gprasp2 expression correlates with molecular programs of quiescence, stemness and signaling.

Previous work in our lab established that Gprasp2 is highly expressed in LT-HSCs compared to downstream HSPCs³². To better understand if native Gprasp2 expression patterns in LT-HSCs correlated with stemness and self-renewal, we interrogated human and mouse published datasets of LT-HSCs (Extended Data Fig. 1a-b). Surprisingly, transcriptionally defined LT-HSCs were heterogeneous with respect to Gprasp2 expression (Extended Data Fig. 1a-b). Importantly, phenotypic mouse (Lineage⁻SCA-1⁺c-Kit⁺CD150⁺CD48⁻) and human (CD34⁺CD38⁻CD90⁺CD45RA⁻CD49f⁺) LT-HSCs expressing Gprasp2 were enriched for stemness programs and depleted for programs relating to cell cycle and differentiation (Extended Data Fig. 1a-b). Further, when comparing the transcriptional profiles of quiescent (freshly isolated) and activated (four days in culture) human LT-HSCs, GPRASP2 is enriched in quiescent cells (Extended Data Fig. 1c). Immunocytochemistry validated heterogeneous expression of GPRASP2 amongst mouse LT-HSCs (Extended Data Fig. 1d). These data suggest that Gprasp2^high and Gprasp2^low HSCs may represent functionally distinct HSC subsets.

To test this hypothesis, we generated a Gprasp2-tdTomato reporter mouse (Fig. 1a). We validated the fidelity of tdTomato as a reporter of Gprasp2 gene and protein expression in bone marrow (BM) HSPCs (defined as Lineage⁻SCA-1⁺c-Kit⁺) and LT-HSCs and found that GPRASP2 expression robustly correlates with tdTomato fluorescence (Fig. 1b, Extended Data Fig. 2a-c). Analysis of BM and peripheral blood (PB) from Gprasp2-tdTomato mice reveals a hierarchical expression pattern for Gprasp2 across the blood system, with the highest level of expression in LT-HSCs, consistent with prior results³² (Fig. 1c-d, Extended Data Fig. 2d-e). Next, we examined the expression levels of cell surface receptors associated with stemness in LT-HSCs differentially expressing Gprasp2-tdTomato (i.e., Gprasp2^high and Gprasp2^low; Fig. 1e-f). We found Gprasp2^high LT-HSCs enriched in stemness markers (e.g. SCA-1, CD150, EPCR) and depleted in differentiation markers (e.g. CD34, CD48) relative to Gprasp2^low LT-HSCs (Fig. 1f). To investigate the underlying molecular programming, Gprasp2^high and Gprasp2^low BM LT-HSCs were sorted for bulk RNA sequencing (RNAseq). Relative to Gprasp2^low LT-HSCs, Gprasp2^high LT-HSCs showed increased expression of 1798 genes and decreased expression of 494 genes (Extended Data Fig. 2f). Gene set enrichment analysis (GSEA) of Gprasp2^high LT-HSCs revealed low transcriptional and metabolic activity, consistent with previously described quiescent and highly transplantable LT-HSC subsets (Fig. 1g). We next performed gene set variation analysis (GSVA) to create a transcriptional signature for Gprasp2^high LT-HSCs and compare it to published signatures of LT-HSC subsets previously shown to have superior stemness and self-renewal potential^{12, 13, 34-37}. The transcriptional profile of Gprasp2^high LT-HSCs highly correlated with published signatures of dormant, serially engrafting LT-HSCs and negatively correlated with that of transient, actively cycling LT-HSCs (Fig. 1h).

Figure 1. Mouse LT-HSCs enriched for stemness molecular programs express high levels of Gprasp2.

a, Schematic of Gprasp2-tdTomato allele generation. b, LT-HSCs were sorted based on low, medium, or high tdTomato MFI and western blot (WB) performed to measure GPRASP2 expression. Representative WB and quantification. c, Representative flow cytometry of Gprasp2-tdTomato mean fluorescent intensity (MFI) in hematopoietic cells isolated from Gprasp2-tdTomato mice. d, Quantification of Gprasp2-tfTomato in hematopoietic cells from male (n=4) and female (n=3) reporter mice normalized to LT-HSCs from three independent experiments. e, Representative flow cytometry gating of Gprasp2^high and Gprasp2^low LT-HSCs. f, Flow cytometry of stemness-associated cell surface markers on Gprasp2^high and Gprasp2^low LT-HSCs (n=7, two independent experiments, Mann-Whitney test). g, Gprasp2^high and Gprasp2^low LT-HSCs were sorted from Gprasp2-tdTomato mice and bulk RNA sequencing was performed (n=3). GSEA of DEGs from these samples. h, GSVA of DEGs from (g) compared to published LT-HSC transcriptional signatures by Pearson correlation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Gprasp2 is a robust reporter of durable, self-renewal potential.

To test if Gprasp2 expression functionally correlates with stemness and self-renewal, we serially transplanted Gprasp2^high and Gprasp2^low cells (Fig. 2a-b). CD45.2⁺ Gprasp2^high or Gprasp2^low LT-HSCs and CD45.1⁺ LT-HSCs were transplanted at a 1:4 ratio into lethally irradiated CD45.1⁺/CD45.2⁺ recipients (Fig. 2a-e). Gprasp2^high LT-HSCs displayed a significant competitive advantage over both bulk CD45.1⁺ competitor LT-HSCs and Gprasp2^low LT-HSCs (Fig. 2c, Extended data Fig. 3a). Gprasp2^high LT-HSCs also repopulated all BM HSPC compartments more robustly than Gprasp2^low LT-HSCs (Fig. 2d).

Figure 2. Gprasp2 defines a hierarchy of LT-HSCs with durable stemness and self-renewal.

a, Experimental schematic of competitive serial transplantation of Gprasp2^high and Gprasp2^low LT-HSCs. b, Schematic of recovery of Gprasp2^high and Gprasp2^low LT-HSCs from recipient mice for serial transplantation. c, %CD45.2⁺ PB of primary, secondary, and tertiary recipients transplanted with Gprasp2^high and Gprasp2^low LT-HSCs and competitor cells. For primary transplants, n≥12 mice per condition, three independent experiments; for secondary transplants, n≥11 mice per condition, three independent experiments; for tertiary transplants; n≥4 mice per condition, two independent experiments. d, Frequency of CD45.2⁺ cells in BM hematopoietic progenitors from primary recipients in (c). e, Representative flow cytometry of the Gprasp2-tdTomato MFI in BM LT-HSCs of primary, secondary, and tertiary recipients from (c) relative to LT-HSCs isolated from control C57Bl/6 mice (non-reporter animals) and steady state (i.e. non-transplanted) Gprasp2-tdTomato mice. f, Experimental schematic of serial non-competitive transplantation of Gprasp2^high and Gprasp2^low LT-HSCs. g, %CD45.2⁺ PB of recipient mice transplanted with Gprasp2^high and Gprasp2^low LT-HSCs. n≥9 mice per condition, three independent experiments. h, Complete blood counts of mice from (g). Steady-state line represents the average blood counts from ten unmanipulated C57Bl/6 mice. i, Experimental schematic of single index-sorted Gprasp2^high and Gprasp2^low LT-HSCs transplanted with 150, 000 supporting BM cells. j, %CD45.2⁺ PB of recipient mice transplanted with single Gprasp2^high or Gprasp2^low LT-HSCs. n≥47 mice per condition from three independent experiments. Dotted line represents 0.5% CD45.2⁺ PB. Engraftment below 0.1% CD45.2⁺ PB is not shown. k, Correlation between the %CD45.2⁺ PB of mice engrafted with single Gprasp2^high or Gprasp2^low LT-HSCs (from j) and Gprasp2-tdTomato MFI levels of the transplanted cell as indicated by index sorting. Solid line represents the linear regression for graphed data. Dotted line represents 0.5% CD45.2⁺ PB. Engraftment below 0.1% CD45.2⁺ PB is not shown. r = Pearson correlation coefficient. l, Pearson correlation of % CD45.2⁺ peripheral blood with MFI of index sorting markers (n=17). m, Gprasp2^high LT-HSCs segregated by repopulation kinetics. *p < 0.05, **p < 0.01, ***p < 0.001. Mann-Whitney test unless otherwise indicated.

In reconstituted bone marrow, Gprasp2^high LT-HSCs gave rise to LT-HSCs expressing high and low levels of Gprasp2, while Gprasp2^low LT-HSCs only gave rise to Gprasp2^low LT-HSCs (Fig. 2e), suggesting that Gprasp2^high LT-HSCs may be at the top of the primitive hierarchy relative to Gprasp2^low LT-HSCs. To test if persistent Gprasp2 expression tracks with serial transplantation potential, we re-sorted Gprasp2^high and Gprasp2^low LT-HSCs from primary recipients and performed secondary competitive transplants (Fig. 2c, Extended data Fig. 3b-c). Gprasp2^high LT-HSCs derived from Gprasp2^high LT-HSCs again displayed a competitive advantage over CD45.1⁺ LT-HSCs in blood output. In contrast, Gprasp2^low LT-HSCs generated by the same Gprasp2^high LT-HSCs (i.e., Gprasp2^low from Gprasp2^high) show little to no repopulating potential. Similarly, Gprasp2^low LT-HSCs (Gprasp2^low from Gprasp2^low) do not repopulate the blood of secondary recipients (Fig. 2c, Extended data Fig. 3b-c). These inheritable repopulating dynamics are preserved in tertiary transplants, where only LT-HSCs that preserve high Gprasp2 expression maintain durable reconstitution and self-renewal potential (Fig. 2c, Extended data Fig. 3d-e). These data imply that heritability of long-term self-renewal in HSCs directly correlates with maintenance of Gprasp2 expression. Interestingly, Gprasp2^high LT-HSCs also have enhanced ex vivo expansion potential compared to Gprasp2^low LT-HSCs (Extended Data Fig 3. f-i). Thus, high Gprasp2 expression is a robust reporter of HSCs capable of durable and multi-lineage serial blood repopulating activity, while loss of Gprasp2 expression correlates strongly with a loss of blood repopulating potential.

Gprasp2 defines a hierarchy of LT-HSCs

As Gprasp2^low LT-HSCs transcriptomes correlated with LT-HSCs previously characterized as “high output”¹², we hypothesized that Gprasp2^low LT-HSCs may functionally resemble these cells. To test this, we transplanted CD45.2⁺ Gprasp2^high or Gprasp2^low LT-HSCs into sublethally irradiated CD45.1⁺ recipients without competitor HSCs (Fig. 2f). Gprasp2^low LT-HSCs showed early blood output that waned overtime post-transplant. In contrast, blood output from Gprasp2^high LT-HSCs was delayed relative to Gprasp2^low cells, but increased overtime (Fig. 2g). Further, the magnitude of total white blood cell (WBC) output by Gprasp2^low cells at all time points was significantly less than that of Gprasp2^high cells by three weeks post-transplant (Fig. 2h). Acute, transient blood production followed by a rapid decline in output is consistent with previously described “high-output’” LT-HSCs¹². Importantly, these differences are not evenly observed across blood lineages, as Gprasp2^low myeloid output rapidly declined between four- and eight-weeks post-transplant, while low lymphoid output was maintained (Extended data Fig. 3j). Notably, these lineage repopulating dynamics are consistent with the pattern of reconstituted BM progenitors post-transplant: common lymphoid progenitors (CLPs) are the only Gprasp2^low-derived progenitor apparent in the BM, suggesting that Gprasp2^low LT-HSCs rapidly exhausted following transplant and only the longest-lived progenitors are detectable at 20 weeks post-transplant (Extended data Fig. 3k). We next sorted Gprasp2^high and Gprasp2^low LT-HSCs from Gprasp2^high LT-HSC primary recipients and performed secondary transplants (Fig. 2f). A similar pattern emerged in secondary transplants: Gprasp2^high LT-HSCs robustly engrafted, while Gprasp2^low LT-HSCs produced a transient burst of myeloid cells that rapidly declined by four weeks post-transplant (Extended data Fig. 3l-m). Again, the magnitude of blood production was greater from serially transplanted Gprasp2^high LT-HSCs at all timepoints post-transplant (Extended data Fig. 3l). Thus, Gprasp2^high LT-HSCs yield robust and long-term multilineage output that is greater in magnitude than that of Gprasp2^low LT-HSCs, suggesting that Gprasp2 expression reflects a functional hierarchy in the immunophenotypic HSC pool. In contrast, Gprasp2^low LT-HSCs represent a transient, non-durable population within the phenotypically defined LT-HSC pool.

Robust blood output correlates with high Gprasp2 expression at a clonal level.

To rigorously interrogate the activity and potential of individual blood repopulating clones in Gprasp2^high and Gprasp2^low LT-HSC subsets, BM LT-HSCs from CD45.2⁺ Gprasp2-tdTomato reporter mice were index sorted as single cells and then transplanted with 150,000 CD45.1⁺ BM cells into lethally irradiated mice (Fig. 2i). Clones that yielded >0.5% multi-lineage PB repopulation were deemed ‘engrafted’. About 30% (20/66) of Gprasp2^high LT-HSCs showed >0.5% engraftment; 17 of these persisted for at least 16 weeks. No Gprasp2^low LT-HSCs displayed >0.5% PB engraftment (0/48) (Fig. 2j, Extended Figure 4a, Extended Data Table 1). Remarkably, clone engraftment correlated more robustly with Gprasp2 expression than any other HSC cell surface marker (Fig. 2k-l). Heterogeneity in output kinetics was observed among the transplantable Gprasp2^high clones; nine engrafted clones displayed ‘delayed’ engraftment kinetics, with low output at early timepoints that steadily increased through week 16 (Fig. 2m, Extended Data Fig. 4b-e). Eight clones were deemed ‘transient’ due to rapid output kinetics that declined over time (Extended Data Fig. 4b-d). These data are consistent with previous studies that also observed functional heterogeneity amongst individual repopulating clones^{5, 12, 38}. Most Gprasp2^high clones with delayed kinetics were serially transplantable (i.e. durable) (Extended Data Fig. 4e-h), with 75% of these clones repopulating LT-HSCs in secondary recipients (Extended Data Fig. 4h). Thus, Gprasp2 expression marks a heterogeneous population of engraftable clones. Combined index sorting and ex vivo colony formation assays confirmed the correlation between Gprasp2 expression and multipotency at the clonal level (Extended Data Fig. 4i-j).

Gprasp2^high LT-HSCs resist proliferative cues

Our transcriptional profiling of Gprasp2^high LT-HSCs suggested a deep state of quiescence (Fig. 1g, Extended Data Fig. 1a-b).¹⁷ Indeed, more Gprasp2^high LT-HSCs are in G0 than Gprasp2^low LT-HSCs (Fig. 3a, Extended Data Fig 5a). To explore this further, we utilized Gprasp2-tdTomato; ROSA26^+/rtTA; pTRE-H2BGFP^+/GFP mice, which allow for doxycycline-inducible Histone-2B-GFP expression to track in vivo division kinetics³⁹. Mice were treated with doxycycline water for six weeks and then ‘chased’ for 150 days before analysis of BM HSPC GFP retention (Fig. 3b). 100% of BM LT-HSCs were GFP+ 24 hours after doxycycline water removal (Extended Data Fig. 5b-c). Gprasp2^high LT-HSCs displayed significantly more GFP retention after chase than Gprasp2^low LT-HSCs, consistent with molecular profiling implicating Gprasp2^high LT-HSCs as LT-HSCs programmed for dormancy (Fig. 3b-c, Fig. 1g)¹³. Gprasp2^high LT-HSCs also have lower mitochondrial membrane potential, despite similar mitochondrial content (Fig. 3d, Extended Data Fig. 5d). To investigate if Gprasp2^high LT-HSCs are resistant to proliferative cues, we used time-lapse imaging to observe time to first and second division of Gprasp2^high and Gprasp2^low LT-HSCs exposed to Stem Cell Factor (SCF) and Thrombopoietin (TPO) ex vivo (Fig. 3e). Cells were stained with SCA-1, Lysobrite, and CD71 to interrogate markers of stemness and differentiation as well as lysosomal inheritance⁴⁰. Gprasp2^high LT-HSCs displayed a delayed time to first and second division relative to Gprasp2^low LT-HSCs (Fig. 3f-g). We next compared lysosomal inheritance in daughter cells after division, which can be symmetrically or asymmetrically inherited and is predictive of daughter cell fate⁴⁰. Gprasp2^high LT-HSCs more frequently engaged in symmetric cell divisions and displayed higher expression levels of SCA-1 pre- and post-division, relative to Gprasp2^low LT-HSCs (Fig. 3h, Extended Data Fig. 5e-g). To confirm the reduced sensitivity of Gprasp2^high LT-HSCs to proliferative cues in vivo, we treated Gprasp2-tdTomato mice with polyinosinic:polycytidylic acid [poly(I:C)], which mimics a viral infection and triggers HSC activation and proliferation (Fig. 3i)^{41, 42}. Poly(I:C) treatment induced a loss of Gprasp2 expression in LT-HSCs (Extended Data Fig. 5h). Notably, more Gprasp2^low LT-HSCs were activated to cycle compared to Gprasp2 (Fig. 3j, Extended Data Fig. 5h). Therefore, Gprasp2^high LT-HSCs are a dormant population that is resistant to activation by external cues.

Figure 3. Gprasp2^high marks deeply quiescent HSCs.

a, Flow cytometry of Ki67/DAPI in LT-HSCs to assess cell cycle status of Gprasp2^high and Gprasp2^low LT-HSCs (n=8, three independent experiments). b, In vivo LT-HSCs division kinetics tracking using Gprasp2-tdTomato crossed with ROSA26^rtTa/rtTA; pTRE-H2B^GFP+/GFP+ mice treated with doxycycline for six weeks compared to untreated littermates. c, Flow cytometry of H2B-GFP MFI in Gprasp2^high and Gprasp2^low LT-HSCs compared to untreated littermates and C57BL/6 mice (n=6, four independent experiments). d, Flow cytometry for TMRM in Gprasp2^high and Gprasp2^low LT-HSCs (n=4, two experiments). e, Schematic of cell generations observed pre- and post-division using time-lapse imaging of Gprasp2^high and Gprasp2^low LT-HSCs. f, Time-to-first division of Gprasp2^high and Gprasp2^low LT-HSCs (two biological replicates, n≥277). g, Cell cycle length of first generation of Gprasp2^high and Gprasp2^low LT-HSCs (two biological replicates, n≥454). h, Sister ratio of Lysobrite MFI in Gprasp2^high and Gprasp2^low LT-HSCs daughter cells post division (n≥269). i, Experimental schematic for hematopoietic stress model. Gprasp2-tdTomato mice were intraperitoneally injected with PBS or poly(I:C) (5 mg/kg). BM was analyzed for LT-HSC frequency and cell cycle (Ki67/DAPI) 24 hours after treatment. j, Cell cycle analysis of mice from (i) (n=5, two independent experiments). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Mann-Whitney test unless otherwise indicated.

Gprasp2 is required for LT-HSC self-renewal and division kinetics

Our findings indicate that high Gprasp2 expression identifies quiescent LT-HSCs with durable self-renewal and high blood output potential (Fig. 2-3, Extended Data Fig. 3-5). To test if Gprasp2 is required for this phenotype, we targeted Gprasp2 in Gprasp2^high and Gprasp2^low LT-HSCs via lentiviral transduction with Gprasp2-shRNAs (Fig. 4a). After validating shRNA knockdown and transduction efficiency (Extended Data Fig. 6a-d), we transcriptionally profiled Gprasp2^high and Gprasp2^low LT-HSCs following knockdown of Gprasp2. Few transcriptional changes were apparent in Gprasp2^low LT-HSCs, with only 24 differentially expressed genes (DEG) relative to control (Extended Data Fig. 6e-f). In contrast, 866 genes were upregulated and 1139 downregulated when Gprasp2 was knocked down in Gprasp2^high LT-HSCs (Extended Data Fig. 6e-f). GSEA revealed upregulation of cell cycle genes, programs relating to lysosomal transport, and signaling programs, including JAK-STAT and small GTPase-mediated signaling (Fig. 4b-c, Extended Data Fig. 6f-g). To follow up on the observed changes in cell cycle genes, we performed time-lapse imaging of Gprasp2^high LT-HSCs treated with Gprasp2- or control-shRNAs. Consistently, time-lapse imaging of single Gprasp2^high LT-HSCs after Gprasp2 knockdown showed shorter division times relative to control shRNA treated cells (Fig. 4d, Extended Data Fig. 6h). Next, to assess the effect of Gprasp2 knockdown on self-renewal, we interrogated shRNA-treated Gprasp2^high and Gprasp2^low LT-HSCs in ex vivo LT-HSC expansion cultures⁴³. Fewer phenotypic LT-HSCs (defined here as Lineage⁻c-Kit⁺SCA-1⁺EPCR⁺CD150⁺)⁴⁴ expanded in cultures of Gprasp2^high LT-HSCs treated with Gprasp2-shRNAs than those treated with control-shRNAs (Fig. 4e). These data support a model in which GPRASP2 is required for LT-HSCs transcriptional identity and quiescence, and high Gprasp2 expression denotes the most dormant and functional HSCs.

Figure 4. HSC quiescence and stemness requires Gprasp2.

a, Experimental schematic of lentiviral knockdown of Gprasp2 in Gprasp2^high and Gprasp2^low LT-HSCs followed by bulk RNA sequencing, time-lapse imaging, and ex vivo LT-HSC expansion. b, GSEA of DEGs in Gprasp2^high LT-HSCs treated with Gprasp2-shRNA (FDR≤0.05). c, Heat map of expression of genes involved in GTPase Mediated signal transduction in Gprasp2^high control and Gprasp2-shRNA treated LT-HSCs. d, Cell cycle length of Gprasp2^high LT-HSCs treated with either control or Gprasp2-shRNA (n≥167 cells across three biological replicates, one experiment). e, Ex vivo expansion of Gprasp2^high and Gprasp2^low LT-HSCs with lentiviral knockdown of Gprasp2 for seven days post-transduction. Here, LT-HSCs were defined as Lineage⁻cKIT⁺SCA-1⁺EPCR⁺CD150⁺ (n=3 biological replicates, two technical replicates, three independent experiments). ***p < 0.001. Mann-Whitney test unless otherwise indicated.

LT-HSC self-renewal relies on Gprasp2-mediated endocytosis

To gain further insight into the distinct molecular programs of Gprasp2-defined LT-HSC subsets, we performed low-input proteomic profiling of sorted Gprasp2^high and Gprasp2^low LT-HSCs (Fig. 5a)⁴⁵. Proteomic profiling of Gprasp2^high LT-HSCs confirmed depletion of cell-cycle programs and regulators, including CDK6, CDK4, MKI67, and proteins relating to translation, relative to Gprasp2^low LT-HSCs, consistent with deeply quiescent LT-HSCs (Fig. 5b, Extended Data Fig. 7a). Curiously, the correlation between RNA and protein expression is smaller in Gprasp2^high LT-HSCs, which has been reported as a feature of HSCs relative to downstream progenitors (Extended Data Fig. 7b)⁴⁶. Strikingly, despite quiescent programming, Gprasp2^high LT-HSCs also show a strong enrichment for signaling proteins, including a broad representation of G-protein signaling components (GNAI1-3, GNAQ/11/12/13, GNB1-5, GNG2-12), second messengers (PLCB1–3, PRKACA/B), small GTPases (KRAS, RRAS, RAC1/2, CDC42), and MAP Kinases (MAP2K1, MAPK3) (Fig. 5b, Extended Data Fig. 7c). A closer look at these changes revealed that Gprasp2^high LT-HSCs express higher levels of receptors and upstream signaling constituents, while the downstream signaling elements are lacking, suggesting a lack of active signal transduction (e.g., general upregulation of phosphatases and partial repression of kinases, Extended Data Fig. 7c). Indeed, phosphoflow analysis showed lower phosphorylation levels for P38 and ERK1/2 in Gprasp2^high LT-HSCs, even after in vivo activation by poly(I:C) (Fig. 5c).

Figure 5. Gprasp2 deficiency accelerates stem cell exhaustion via endocytosis dysregulation.

a, Proteomic profiling of Gprasp2^high and Gprasp2^low LT-HSCs by mass spectrometry (n=3). b, Gene Ontology enrichment of differentially expressed proteins (DEPs; FDR ≤ 0.05) showing upregulated and downregulated pathways in Gprasp2^high LT-HSCs. c, Phosphorylation level of P38 and ERK1/2 in Gprasp2^high and Gprasp2^low LT-HSCs with and without poly(I:C) in vivo treatment (5ug/kg; 24 hours) (N=5). d, Representative flow cytometry and quantification of dextran-pHrodo green uptake in Gprasp2^high and Gprasp2^low LT-HSCs treated with 10,000 MW dextran (compared to 4ºC negative control) to measure endocytosis (n=6, 3 independent experiments). e, Schematic showing lentiviral knockdown of Gprasp2 in Gprasp2^high and Gprasp2^low LT-HSCs followed by dextran incubation and flow cytometry. f, dextran-green MFI of transduced Gprasp2^high and Gprasp2^low LT-HSCs to measure endocytosis (n=5, three independent experiments). g, Experimental schematic of LT-HSCs sorted and cultured in LT-HSC expansion or maintenance media with or without endolysosomal inhibitors (genistein 1 μM, chlorpromazine 1 μM, bafilomycin 1 nM, chloroquine 2 μM, Dyngo 1 μM). h, Numbers of LT-HSCs following seven days of expansion culture (n=12, five experiments). i, Numbers of LT-HSCs following 48 hours of maintenance culture (n=6, three experiments). j, Numbers of LT-HSCs following seven days of expansion culture of Gprasp2^high and Gprasp2^low LT-HSCs (n=6, three experiments). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Mann-Whitney test unless otherwise indicated.

Interestingly, proteins relating to vesicular trafficking (CLTC, AP2 complex subunits, DNM1-3, ARRB1, RAB5A, RAB7, RAB11A) were strongly enriched in Gprasp2^high LT-HSCs (Fig. 5b, Extended Data Fig. 7a), which is consistent with our observation that expression of endocytosis and endosomal trafficking genes was also perturbed in Gprasp2-deficient Gprasp2^high LT-HSCs (Fig. 4b). Endocytosis contributes to the control of receptor density on the cell surface^{20, 25, 47}. Differences in vesicle trafficking could explain the discrepancy between the abundance of cytokine receptors and a lack of signaling transduction in Gprasp2^high LT-HSCs. GPRASP2 reportedly participates in endosomal sorting of GPCRs to the lysosome³¹, and co-immunoprecipitation assays reveal a physical interaction in hematopoietic cells between GPRASP2, clathrin, and SCYL1, which is involved in clathrin-mediated endocytosis (Extended Data Fig. 7d-e)⁴⁶. Thus, we interrogated endocytic uptake in Gprasp2^high and Gprasp2^low LT-HSCs using phRodo-dextran. Endocytosis was higher in Gprasp2^high LT-HSCs than Gprasp2^low LT-HSCs, while lysosomal mass was unchanged (Fig. 5d, Extended Data Fig. 7f). Next, we tested if elevated levels of endocytosis in Gprasp2^high LT-HSCs depend on Gprasp2 by treating Gprasp2^high and Gprasp2^low LT-HSCs with Gprasp2- or control-shRNAs (Fig. 5e). GPRASP2 loss reduced endocytic uptake in Gprasp2^high LT-HSCs, but not in Gprasp2^low LT-HSCs, suggesting that GPRASP2 contributes to endocytosis in Gprasp2^high LT-HSCs (Fig. 5f). To interrogate a role for endolysosomal trafficking in HSC self-renewal, we exposed LT-HSCs to chemical inhibitors of this pathway during ex vivo expansion culture, including chlorpromazine (clathrin-mediated endocytosis)⁴⁸, bafilomycin (endosome/lysosome acidification)⁴⁹, genistein (caveolae formation)⁵⁰, Dyngo-4a (dynamin assembly)⁵¹, and chloroquine (PICALM expression)⁵² (Fig. 5g, Extended Data Fig. 7g). Cells were cultured in drug concentrations that modestly, but significantly, reduced endocytic uptake by LT-HSCs while also preserving cell viability (Extended Data Fig. 7h-j). Chemical inhibition of the endolysosomal pathway specifically inhibited phenotypic LT-HSCs expansion during ex vivo culture (Fig. 5h, Extended Data Fig. 7k). Phenotypic LT-HSC maintenance was also reduced by endolysosomal inhibition when cells were cultured in differentiation media (Fig. 5i, Extended Data Fig. 7l). These effects were unique to LT-HSCs, as MPPs were unaffected by chemical inhibition of the endolysosomal pathway (Extended Data Fig. 7m). Indeed, when Gprasp2^high and Gprasp2^low LT-HSCs were tested separately, only the expansion of Gprasp2^high LT-HSCs was significantly hindered by endocytosis inhibition (Fig. 5j).

CD48^lowSCA-1^high sorting strategy captures Gprasp2^high LT-HSCs

Given that high Gprasp2 expression reliably enriches for mouse LT-HSCs displaying robust stemness and self-renewal potential, we sought correlative cell surface markers that would allow enrichment of Gprasp2^high LT-HSCs without needing genetic reporter mice⁵³. Index sorting revealed that Gprasp2 expression correlates well with SCA-1 and negatively correlates with CD48 (Fig. 6a). Indeed, back gating LSK CD48⁻CD150⁺ HSCs to SCA-1 versus CD48 revealed that CD48^lowSCA-1^high LT-HSCs and CD48^highSCA-1^low LT-HSCs were enriched and depleted in Gprasp2 expression, respectively (Fig. 6b-d). We termed these CD48^lowSCA-1^high LT-HSCs as “Grand SLAMs” (Gprasp2 SLAM HSCs) and tested their ex vivo LT-HSC expansion capacity as a proxy for self-renewal potential (Fig. 6e). Grand SLAMs exhibited superior ex vivo expansion capacity relative to CD48^highSCA-1^low LT-HSCs (Fig. 6f) and displayed lower in vivo cycling kinetics (Fig. 6g), like Gprasp2^high LT-HSCs (Fig. 3). Next, 200 CD45.2⁺ Grand SLAMs or CD48^highSCA-1^low LT-HSCs were transplanted into lethally irradiated CD45.1⁺ recipient mice (Fig. 6h). Grand SLAMs functionally phenocopied Gprasp2^high LT-HSCs via increased CD45.2⁺ blood output (Fig. 6i, Extended Data Fig. 8a), increased WBC production (Fig. 6j), and robust repopulation of all HSPCs in recipient BM, relative to CD48^highSCA-1^low LT-HSCs (Fig. 6k). 200 CD45.2⁺ Grand SLAMs or CD48^highSCA-1^low LT-HSCs isolated from primary recipients were transplanted into lethally irradiated CD45.1⁺ secondary recipient mice (Extended Data Fig. 8b-c). As seen with Gprasp2^high LT-HSCs, Grand SLAMs displayed durable self-renewal and blood repopulating potential in secondary transplants (Extended Data Fig. 8d-f). Thus, Gprasp2^High LT-HSCs can be enriched from mouse BM without the use of an engineered genetic reporter allele by one additional, simple gating step; LSK CD150⁺CD48⁻ à CD48^lowSCA-1^high will capture most Gprasp2^high LT-HSC (i.e. Grand SLAMs).

Figure 6. Gprasp2^high LT-HSCs can be isolated via gating on CD48^lowSCA-1^high LT-HSCs.

a, Pearson correlation between surface markers and Gprasp2 from single index sorted LT-HSCs (n≥ 235, three independent experiments). b, Flow cytometry gating strategy of LT-HSCs with additional gating of CD48^lowSCA-1^high/CD48^highSCA-1^low cells. c, Representative flow cytometry of LT-HSCs showing Gprasp2-tdTomato MFI of Gprasp2^high and Gprasp2^low LT-HSCs, CD48^lowSCA-1^high LT-HSCs, and CD48^highSCA-1^low LT-HSCs. d, Quantification of Gprasp2-tdTomato MFIs from (c) (n=9, three independent experiments). e, Experimental schematic of ex vivo expansion of CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs. f, Ex vivo expansion of CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs for two weeks, as defined by numbers of Lineage⁻cKIT⁺SCA-1⁺EPCR⁺ CD150⁺ cells (n=3 biological replicates, n≥3 technical replicates, three independent experiments). g, Flow cytometry of H2B-GFP MFI in CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs compared to untreated littermates and C57BL/6 mice (n=11, four independent experiments). h, Experimental schematic of transplantation of 200 CD48^lowSCA-1^high or CD48^highSCA-1^low LT-HSCs into lethally irradiated recipients. i, % CD45.2⁺ PB of mice transplanted with CD48^lowSCA-1^high or CD48^highSCA-1^low LT-HSCs. n≥13 mice per condition, three independent experiments. j, Total WBCs of recipients in (h). (n≥3 per condition, one experiment). Steady-state line is the average WBC counts from n=10 CD45.1⁺/CD45.2⁺ unmanipulated mice. k, CD45.2⁺ BM progenitors in recipients from (h). (n≥8, two experiments). l, Dextran-pHrodo green MFI in CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs (N=8). m, Experimental schematic for EV uptake assays. EVs were isolated from HSPCS or LT-HSCs and labeled with PKH26. EV uptake by CD48^lowSCA-1^high or CD48^highSCA-1^low LT-HSCs was analyzed by flow cytometry. n, Percentage of cells positive for uptake of LT-HSC- or HSPC-derived EVs (N≥4). *p < 0.05, **p < 0.01, ***p < 0.001. Mann-Whitney unless otherwise indicated.

Taking advantage of this new gating strategy, we confirmed that Grand SLAMs display higher endocytosis rates than CD48^highSCA-1^low LT-HSCs (Fig. 6l), as seen for Gprasp2^high LT-HSCs (Fig. 5d). Extracellular vesicles (EVs) are part of the autocrine and paracrine crosstalk between HSCs and their microenvironment⁵⁴. It has been recently reported that HSC-derived EVs, the production of which depends on fatty acid oxidation and cholesterol, enhance HSC transplantation and stimulate SCF production from bone marrow stromal cells⁵⁵. To test the premise that elevated endocytosis in Grand SLAMs promotes increased uptake of HSC-derived EVs, we tested Grand SLAM EV uptake (Fig. 6m). Grand SLAMs presented significantly higher EV uptake than CD48^highSCA-1^low LT-HSCs when HSCs were the source of EVs (Fig. 6n, Extended Data Fig. 8g). Strikingly, these differences disappear when EVs from HSPCs are used, suggesting selective EV endocytic uptake depending on cell of origin (Fig. 6n, Extended Data Fig. 8g). No differences were observed in the amount and size of EVs produced by either Grand SLAMs or CD48^highSCA-1^low LT-HSCs (Extended Data Fig. 8h-j). Thus, the specific and unidirectional nature of these changes implies that the greater endocytic activity in Grand SLAMs is a unique feature of the most primitive and functional HSCs, which contributes to their identity.

Overall, we discovered that dormant HSCs have elevated endocytosis, which is necessary for maintaining LT-HSC self-renewal. Further, GPRASP2 marks dormant LT-HSCs with superior stemness and self-renewal capacity. Despite functional dormancy, LT-HSCs with high GPRASP2 are enriched in signaling intermediate proteins, suggesting a poised state for activation. Disrupted endocytosis led to increased gene expression relating to cell signaling, more rapid cell cycling, consistent with a model in which elevated endocytosis attenuates signaling inputs to maintain dormancy. Lastly, our Gprasp2-reporter model and Grand SLAM gating strategy provide new tools to interrogate HSC dormancy and endocytosis, and we show that Grand SLAMs preferentially uptake more HSC-derived EVs—which have been shown to improve HSC function—than CD48^high SCA-1^low HSCs, suggesting elevated endocytosis may support a variety of functions critical to HSC maintenance.

Discussion

Our work reveals that dormant HSCs exhibit elevated endocytosis, and loss of endocytosis uniquely disrupts HSC self-renewal and growth. Paradoxically, dormant HSCs are enriched in intermediate signaling proteins yet maintain dormancy in the bone marrow and are less responsive to proliferative cues. GPRASP2, an endosomal sorting protein enriched in dormant HSCs, regulates elevated endocytosis, and disruption of GPRASP2 induces more rapid cell cycle entry in HSCs, accompanied by increased expression of signal transduction genes, suggesting a more responsive cell state. Functionally, HSCs with enriched endocytosis also uptake more HSC-derived EVs, which have been shown to increase HSC blood reconstitution⁵⁵. Collectively, these data support a model in which elevated endocytosis preserves HSC dormancy and fidelity by limiting responsiveness to proliferative cues while in parallel facilitating uptake of extrinsic signals reinforcing HSC fidelity.

Our data show that dormant HSCs are enriched for proteins involved in endocytosis and vesicular trafficking and exhibit elevated endocytic activity, and that disruption of endocytosis impairs HSC maintenance and self-renewal. Endolysosomal trafficking is a highly conserved cellular function fundamental to growth, survival, and signaling⁵⁶. However, much of the current understanding of endocytosis derives from studies in proliferating or activated cells, and consequently the relationship between endocytosis and cellular quiescence remains incompletely understood. Across cell types, including HSCs, lysosomes function as signaling and metabolic hubs that regulate quiescence and activation, suggesting that upstream endocytic trafficking may contribute to the maintenance of stem cell dormancy^{17, 19, 22}. Consistently, neural stem cells and embryonic fibroblasts show enhanced receptor trafficking and increased lysosomal degradation supports quiescence, and toggling lysosomal function affects the depth of quiescence^{57, 58}. Indeed, endocytosis regulates pluripotency maintenance of embryonic stem cells and reduces Wnt signaling by internalization of its receptors in colorectal cancer cells^{59, 60}. Moreover, higher endocytosis rates have been recently reported in human LT-HSCs compared to downstream progenitors, and increased lysosomal biogenesis in LT-HSCs leads to reduced Tfr1 receptor expression, suggesting coordinated regulation between lysosomes and endocytic receptor internalization^{22, 61}. Together, these findings are consistent with prior work showing that lysosomal degradation supports quiescence and position endocytosis as an upstream contributor to lysosomal regulation, providing an additional mechanistic component underlying stem cell dormancy. However, its exact role and the identity of molecular players remain poorly understood.

Towards this, our work reveals that LT-HSC quiescence and self- renewal are also impacted by endocytosis via a mechanism that includes endosomal sorting protein GPRASP2. A modest disruption of endocytosis reduced LT-HSC self-renewal and expansion ex vivo, supporting the model that tight regulation of endocytosis is necessary for LT-HSC function. GPRASP2 enrichment captures a high fraction of dormant HSCs, which are paradoxically enriched in many intermediate signaling proteins (Fig. 5a, Extended Data Fig. 7c). However, we found that despite increased signaling intermediates, dormant HSCs maintained quiescence and lower cell signaling in response to proliferative cues (Fig. 3j, Fig. 5c). Thus, although LT-HSCs preserve integrity via quiescence, they are also poised to respond to extrinsic cues. Consistent with this, GPRASP2 deficiency, which reduces endocytosis, resulted in more rapid cell cycling and elevated expression of signaling constituents, but a loss of self-renewal (Fig. 4). These data are consistent with a model in which dormant HSCs enriched in GPRASP2 are poised with signaling intermediates but use elevated endolysosomal activity to regulate receptors on the plasma membrane to prevent activation, concordant with published works showing many cells use endocytosis to regulate signaling and plasma membrane composition²⁰. Further, our findings indicate elevated endocytosis may also support extrinsic input of cues maintaining HSC stemness and self-renewal as HSCs with increased endocytosis also uptake more HSC-derived EVs (Fig. 6m-n), which have been shown to improve HSC reconstitution capacity⁵⁵. While signaling attenuation preserves LT-HSC integrity, effective maintenance of stemness requires a finely tuned capacity to receive extrinsic cues.

Lastly, we demonstrate that high GPRASP2 expression denotes potent, durable, self-renewal LT-HSCs. Indeed, GPRASP2 correlates with durable blood reconstitution more faithfully than current common immunophenotypic LT-HSC markers (Fig. 2k-l). Most importantly, because Gprasp2^high LT-HSCs can be isolated via a modest adjustment to the conventional SLAM-marker based isolation of LT-HSCs (i.e. Grand SLAMs, Fig. 6b), any laboratory can adopt this approach and significantly improve the functional purity of sorted LT-HSCs without the use of additional antibodies, specialized sorting schemes, or genetic reporters.

Importantly, these studies, when considered in the context of our previous work, highlight the difficulty of drawing direct conclusions under ectopic conditions. We previously reported Gprasp2-deficient HSPCs show superior reconstitution capacity post-transplant, which is abrogated in secondary transplants³². These minor timing discrepancies between Gprasp2-knockdown and Gprasp2^low progenitors (i.e., HSPCs treated with Gprasp2-shRNAs or HSCs naturally expressing low levels of GPRASP2, respectively) are potentially due to the effect of an ectopic manipulation of a broader progenitor population (i.e., transduced HSPCs, which mostly consist of non-HSC progenitors), while our new reporter-based approach allows us to explore the same biological process during homeostasis. Thus, this new system offers the opportunity for nuances and dynamics that are not apparent in ectopic systems.

In summary, we propose that elevated endocytosis supports HSC dormancy by attenuating signaling inputs, a process regulated by the endosomal sorting protein GPRASP2. Endolysosomal trafficking of receptors allows HSCs to maintain a quiescent state while remaining poised with signaling intermediates that enable rapid activation during emergency hematopoiesis. Endolysosomal processing may contribute to additional mechanisms supporting HSC fidelity, including EV uptake, suggesting that endocytosis integrates both intrinsic signaling restraint and extrinsic niche-derived inputs to maintain long-term stem cell function.

Methods

Mice

C57BL/6J, B6.SJL-Ptprc^a Pepc^b/BoyJ C57BL/6.SJL-PtprcaPep3b/BoyJ, C57BL/6J-Ptprc^em6Lutzy/J, ROSA26^rtTa/rtTA (B6.Cg-Gt(ROSA)26Sor^{tm1(rtTA*M2)Jae}/J (stock #006965), Col1a1^{tetO-H2B-GFP/+} (Col1a1^{tm7(tetO-HIST1H2BJ/GFP)Jae}/J) (stock #016836) mice were obtained from Jackson Laboratory (Bar Harbor, Maine). Gprasp2-tdTomato mice were generated by the Center for Advance Genome Engineering at St. Jude Children’s Research Hospital. All mice were used between 6-16 weeks old. All animals were housed in a pathogen-free facility, and all experiments were carried out according to procedures approved by the St. Jude Children’s Research Hospital and University of Michigan Institutional Animal Care and Use Committee.

Quantitative RT-PCR

Total RNA was isolated (RNeasy Micro kit; QIAGEN) and reversed transcribed into cDNA (HighCapacity cDNA Reverse Transcriptional kit with RNase Inhibitor; Invitrogen). qRT-PCR was performed using Fast SYBR Green Master Mix (Applied Biosystems) on an ABI StepOnePlus thermal cycler (Applied Biosystems) according to the manufacturer’s instructions. The following primer sets were used for specific gene expression analysis:

Gprasp2: Forward: 5’-TGCTAGGCCCAAAACTGAAAC-3’

Reverse: 5’-CATTCGGTGTCTTGTTCCAGA-3’

Tbp: Forward: 5’-GAAGAACAATCCAGACTAGCAGCA-3’

Reverse: 5’-CCTTATAGGGAACTTCACATCACAG-3’

Western blotting

Low-input western blots on sorted Gprasp2^high/med/low LT-HSCs (ie. lower, middle, and upper 30% of the Gprasp2-tdTomato histogram, Fig. 1e) were performed using a serum GPRASP antibody gifted from Dr. Simonin⁶², and β-ACTIN (Abcam, Cambridge, UK) as previously described⁶³. Briefly, LT-HSCs were sorted directly into PBS with 2% fetal bovine serum containing protease and phosphatase inhibitors (Invitrogen, Carlsbad, California). Next, an appropriate volume of 4X LDS sample buffer containing reducing agent was added to each sample followed by polyacrylamide gradient gel electrophoresis, immunoblotting, and analysis.

Generation and validation of Gprasp2-tdTomato mice

The C57BL/6J-Gprasp2-tdTomato mouse model was created using CRISPR-Cas9 technology and direct zygote injection at the St. Jude Transgenic Core Unit. Briefly, a mixture of the 40ng/μL 3X NLS SpCas9 protein (St. Jude Protein Production Core), 20ng/ μL chemically modified sgRNA (Synthego, Redwood City, CA), and 10ng/ul donor ssDNA (IDT, Coralville, IA) was injected into male pronucleus of fertilized zygotes as previously described⁶⁴. Animals were genotyped at the Center for Advanced Genome Engineering (St. Jude) by PCR and gel electrophoresis followed by sequencing. Animals positive for the desired knock-in modification were backcrossed to C57BL/6J mice 4 times and then breed to homozygosity. Pups were screened via PCR using specific primers to amplify the adjacent sequences between the fluorescent protein and the respective 3’UTR of Gprasp2. The following primer sets were used for specific gene expression analysis:

5’ junction: Forward: 5’- TGAAAATGCTGTTGAACTTGTCTGA-3’

Reverse: 5’- CGGGATTTTCCTCCACGTCC-3’

3’ junction: Forward: 5’- ACCACCTGTTCCTGTACGGC-3’

Reverse: 5’- CACATATGACTTGCTTCACTGGG-3’

Validation of Gprasp2 mRNA in HSPCs and protein expression in LT-HSCs in correlation with tdTomato MFI was performed by qPCR and low-input western blot, respectively.

Flow cytometry of peripheral blood and bone marrow

Peripheral blood (PB) was collected from the retro-orbital sinus and incubated in red blood cell (RBC) lysis buffer as previously described⁶⁵. PB was stained with anti-Gr1-PerCP-Cy5.5 (RB6-8C5), anti-B220-PerCP-Cy5.5 (RA3-6B2), anti-CD11b-PerCP-Cy5.5 (M1/70), anti-B220-PE-Cy7 (RA3-6B2), anti-CD4-PE-Cy7 (RM4-5), anti-CD8-PE-Cy7 (53-6.7), anti-CD45.2-V500 (104) and anti-CD45.1-FITC (A20) antibodies (all at 1:1000 dilution, BD Biosciences, San Jose, CA). 4',6-diamidino-2-fenilindol (DAPI) (Sigma-Aldrich, St. Louis, MO) was used for dead cell exclusion.

Bone marrow (BM) was extracted from mice tibias, femurs, pelvic bones and spine. Bones were crushed and incubated in RBC lysis buffer (Sigma-Aldrich St. Louis, MO) for five minutes on ice. BM HSPCs were visualized by flow cytometry after staining with the following antibodies: LT-HSC/ST-HSC/MPP2/MPP3/MPP4 [B220-BV605 (RA3-6B2), CD3-BV605 (145-2C11), CD4-BV605 (GK1.5), CD8-BV605 (53-6.7), CD19-BV605 (6D5), Gr-1-BV605 (RB6-8C5), Ter119-BV605 (TER-119), SCA-1-BV785 (E13-161.7), c-Kit-APC-780 (2B8), CD150-PE-Cy7(TC15-12F12.2), CD48-Alexa Fluor 700 (HM48-1), Flt3-PerCPe710 (A2F10.1), CD45.1-FITC (104), and CD45.2-APC (A20)]; CMP/GMP/MEP [B220-BV605 (RA3-6B2), CD3-BV605 (145-2C11), CD4-BV605 (GK1.5), CD8-BV605 (53–6.7), CD19-BV605 (6D5), Gr-1-BV605 (RB6-8C5), Ter119-BV605 (TER-119), SCA-1-BV785 (E13-161.7), c-Kit-APC-780 (2B8), FcR II/III-Alexa Fluor 700 (93), CD34-eFluor660 (RAM34), CD45.1-APC (104), and CD45.2-FITC (A20)]; and CLP [B220-BV605 (RA3-6B2), CD3-BV605 (145-2C11), CD4-BV605 (GK1.5), CD8-BV605 (53-6.7), CD19-BV605 (6D5), Gr-1-BV605 (RB6-8C5), Ter119-BV605 (TER-119), SCA-1-BV785 (E13-161.7), c-Kit-APC-780 (2B8), IL-7R-PE-Cy7 (A7R34), CD45.1-APC (104), and CD45.2-FITC (A20)]; for backgating on stemness markers in LT-HSCs: [B220-BV605 (RA3-6B2), CD3-BV605 (145-2C11), CD4-BV605 (GK1.5), CD8-BV605 (53-6.7), CD19-BV605 (6D5), Gr-1-BV605 (RB6-8C5), Ter119-BV605 (TER-119), SCA-1-BV785 (E13-161.7), c-Kit-APC-780 (2B8), EPCR-PerCPe710 (RCR-16), CD229-BV711 (LY9.7.144), CD244.2-A647 (M2B4 (B6)458.1). DAPI (Sigma-Aldrich, St. Louis, MO) was used for dead cell exclusion.

All population frequencies and CD45.2⁺ cells post-transplant were analyzed using an LSR Fortessa (BD Biosciences, San Jose, CA), Northern Lights (Cytek, Freemont, CA), or an Aurora (Cytek, Freemont, CA). Data analysis was performed with FlowJo version 10 (LLC, Ashland, OR).

HSPC cell isolation

BM was extracted from femurs, pelvic bones, tibias and spine via crushing. c-Kit⁺ cells were isolated via magnetic enrichment using anti-CD117 microbeads (Miltenyi Biotec, Carlsbad, CA) and an autoMACs magnetic cell separator (Miltenyi Biotec, Carlsbad, CA). For LT-HSC/MPP cell isolation, cells were stained with anit-B220-BV605 (RA3-6B2), anti-CD4-BV605 (GK1.5), anti-CD8-BV605 (53-6.7), anti-Gr-1-BV605 (RB6-8C5), anti-Ter119-BV605 (TER-119), anti-SCA-1-BV785 (E13-161.7), anti-c-Kit-APC-780 (2B8), anti-CD150-PE-Cy7 (TC15-12F12.2) and anti-CD48-Alexa Fluor 700 (HM48-1) antibodies. DAPI (Sigma-Aldrich, St. Louis, MO) was used for dead cell exclusion. For isolation of Gprasp2^high/low LT-HSCs, Gprasp2-tdTomato mice were used, and gating was consistent with Fig. 1e. Cell isolation was performed using cell sorting on a FACSAria III and a FACSDiscover S8 (BD Biosciences, San Jose, CA).

shRNAs

shRNAs were designed as previously described^{66, 67}. Gene knockdown efficiency in LT-HSCs was previously assessed³².

Gprasp2 shRNA:

5’TGCTGTTGACAGTGAGCGCCAGTAAAGTTAGTGTGATTTATAGTGAAGCCACAGATGTATAAATCACACTAACTTTACTGTTGCCTACTGCCTCGGA-3’

Lentiviral vector preparation and HSC viral transduction

Vesicular stomatitis virus glycoprotein (VSVG)–pseudotyped lentivirus was prepared using a four-plasmid system [transfer vector- (expressing mCherry), Gag/Pol-, Rev/Tat-, and VSVG envelope plasmid] by co-transfection of 293T cells using TransIT 293 (Mirus, Madison, WI). Viral supernatants were cleared and titered 48 hours post-transfection.

Non-tissue culture treated 96-well plates were coated with Retronectin (Takara Bio USA, Madison, WI), according to the manufacturer’s instructions. Lentiviral particles corresponding to a multiplicity of infection of 25 were spin-loaded onto plates for 1 hour,1000 G at room temperature. Wells were washed with PBS and LT-HSCs were resuspended in 200 μL PVA-based media as previously described and incubated at 37°C in 5% O₂, 5% CO₂⁴³. To collect cells for further experimentation after transduction, media was slowly removed, and cells washed and resuspended in PBS + 2% FBS. Transduced cells were then sorted based on mCherry or EGFP expression.

Ex vivo LT-HSC culture and expansion/maintenance assay

For LT-HSC expansion culture, up to 1,500 LT-HSCs were plated in PVA-based media and maintained as previously described⁴³ with or without endolysosomal inhibitors (genistein, (Sigma-Aldrich, St. Louis, MO); chlorpromazine, (Sigma-Aldrich, St. Louis, MO); bafilomycin A1, (Sigma-Aldrich, St. Louis, MO); chloroquine, (Sigma-Aldrich, St. Louis, MO); Dyngo-4a, (Abcam, Cambridge, UK)). Seven days after plating, cultured were counted using a hemocytometer and LT-HSCs (Lineage⁻SCA-1⁺cKIT⁺CD150⁺EPCR⁺) were enumerated by flow cytometry via staining with B220-PerCP (RA3-6B2), CD3-PerCP (145-2C11), CD4-PerCP (GK1.5), CD8-PerCP (53-6.7), CD19-PerCP (6D5), Gr-1-PerCP (RB6-8C5), Ter119-PerCP (TER-119), SCA-1-PerCP-Cy5.5 (E13-161.7), c-Kit-APC-780 (2B8), CD150-PE-Cy7 (TC15-12F12.2) and EPCR-APC (eBio1560) antibodies, and DAPI (Sigma-Aldrich, St. Louis, MO) for viability.

For LT-HSC maintenance culture, up to 1500 LT-HSCs or HSPCs were plated in StemSpan^™ Serum-Free Expansion Medium (SFEM, STEMCELL Technologies) supplemented with mouse 10 ng/mL SCF, 20 ng/mL TPO, 20 ng/mL Insulin-like Growth Factor 2, and 10 ng/mL human Fibroblast growth factor as previously described³² for 48 hours with or without endolysosomal inhibitors (genistein, (Sigma-Aldrich, St. Louis, MO); chlorpromazine, (Sigma-Aldrich, St. Louis, MO); bafilomycin A1, (Sigma-Aldrich, St. Louis, MO); chloroquine, (Sigma-Aldrich, St. Louis, MO); Dyngo-4a, (Abcam, Cambridge, UK)). After 48 hours, the cells were collected and LT-HSCs were enumerated via flow cytometry staining.

Bone Marrow/LT-HSC Transplantation

For competitive transplants, recipient mice (CD45.2/CD45.1 or CD45.1) were lethally irradiated with two split doses of 5.5 Gy. 100 Gprasp2^high/low CD45.2+ LT-HSCs and injected with 400 CD45.1⁺ LT-HSCs intravenously. CD45.2⁺ PB was assessed by flow cytometry every four weeks for 20 weeks, and CD45.2⁺ BM HSPCs were assessed at 20 weeks post-transplant. To minimize biological variation, femur, tibia, pelvic bones and spine were collected and pooled from at least three individual donors for collection of BM via crushing. For secondary and tertiary transplants, tibia, pelvic bones and spine were collected and pooled from all recipients from the same experimental condition. After BM collection, cells were sorted again for CD45.2⁺ Gprasp2^high/low LT-HSCs and CD45.1⁺ LT-HSCs and transplanted into lethally irradiated recipients intravenously. PB and BM were assessed as described above.

For non-competitive transplants, CD45.1⁺ recipient mice were lethally irradiated with two split doses of 4.75 Gy. 200 Gprasp2^high/low CD45.2⁺ LT-HSCs or 200 CD48^high/low SCA-1^high/low LT-HSCs were injected intravenously. PB was collected at 2.5 weeks, 3, weeks, 3.5 weeks, and 4 weeks, then every 4 weeks until 20 weeks post-transplant. In addition to assessing CD45.2⁺ PB via flow cytometry, complete blood counts were also performed at each time point on a Hematology Analyzer (Forcyte, Los Angeles, CA). CD45.2⁺ BM HSPCs were analyzed at 20 weeks post-transplant and Gprasp2^high/low or CD48^high/low SCA-1^high/low LT-HSCs were sorted and 200 LT-HSCs were injected into lethally irradiated recipients for secondary transplants. PB and BM were assessed in secondary transplants the same as primary transplant.

For single cell transplantation, single CD45.2⁺ Gprasp2^high/low LT-HSCs were index-sorted (fluorescence and scatter parameters for each single cell were recorded and indexed to its deposition location) into 96 well plates with RPMI 1640 supplemented with 2% FBS. 150 000 CD45.1⁺ whole bone marrow (WBM) cells were added to each well, and each well was injected into a single lethally irradiated CD45.2/CD45.1 mouse. CD45.2⁺ PB was assessed by flow cytometry every four weeks for 16 weeks. CD45.2⁺ BM HSPCs were examined 16 weeks post-transplant. For secondary transplants, WBM was isolated from primary recipients with detectable LT-HSCs and 5 x 10⁶ cells were injected into 3 new lethally irradiated recipients to control for variability. PB and BM were assessed as in primary transplants.

Immunofluorescence

About 1000 sorted LT-HSCs were placed on CD44 (10 μg/mL; IM7; Biolegend, San Diego, CA) coated coverslips and incubated at 37°C for 1 hour. Alternatively, HPC7 cells were placed on fibronectin (50ng/μL) coated coverslips and incubated at 37°C for 1 hour. For all immunofluorescence, cells were then gently washed with PBS and fixed with 4% paraformaldehyde (Electron Microscopy Sciences 15710-S) in HEPES-KOH buffer (pH 7.5) for 15 minutes at room temperature (RT). Following three PBS washes, cells were permeabilized and blocked using 0.3% Triton X-100 in blocking buffer (0.5% BSA, 50 mM ammonium chloride in PBS) for 30 minutes. After washing, cells were stained with a rabbit anti-GPRASP2 antibody (1:200 dilution; Atlas Antibodies, Stockholm, Sweden) in blocking buffer for 1 hour at RT, washed again and incubated with goat anti-mouse Alexa Fluor 647 (1:300 dilution; Abcam ab150119, Cambridge, UK) in blocking buffer for another hour at RT. After three washes with PBS, coverslips were mounted using ProLong Diamond with DAPI (Thermo Fisher P36962, Waltham, MA), and images were acquired using a Leica SP8 confocal microscope. Image processing was performed using Fiji software.

Single LT-HSC Colony Forming Assay

For single cell transplantation, single Gprasp2^high/med/low LT-HSCs were index-sorted (fluorescence and scatter parameters for each single cell were recorded and indexed to its deposition location) into 96 well plates with Metholcult^™ M3434 Methylcellulose-Based Medium (Stemcell Technologies, Vancouver, BC). Well plates were cultured at 37 °C, 5% CO₂ and after nine days each well was scored according to the manufacturer’s instructions for colony size and type.

Time-lapse imaging

Gprasp2^high/low LT-HSCs were sorted and plated immediately or Gprasp2^high LT-HSCs were sorted and transduced for 72 hours in PVA culture conditions before plating for imaging. Time-lapse experiments were conducted at 37 °C, 5% O₂, 5% CO₂ on μ-slide VI channel slides (Ibidi, Fitchburg, Wisconsin) coated with10 μg/mL anti-CD43–biotin (Life Technologies). Cells were cultured in phenol-red-free IMDM supplemented with 20% BIT, 100 ng/ml SCF, 100 ng/ml TPO, 2 mM L-glutamine, 50 μM 2-mercaptoethanol, 50 U/mL penicillin and 50 μg/ml streptomycin. Images were acquired using a Nikon ECLIPSE Ti2 inverted microscope and Spectra III fluorescent light source (Lumencor, Beaverton, OR). Time intervals of bright-field and fluorescent image acquisition were chosen to minimize photo-toxicity. Images were acquired using 10x Plan Apochromat λD objective (NA 0.45). Fluorescent images were acquired using optimized filter sets: GFP (488/10; 495LP; 532/18), BV480 (395/25; 420LP; 435/26), tdTomato (546/10; 555LP; 572/15), and LysoBrite^™ NIR (620/30; 645LP; 660/13). Single cell tracking and image quantification were performed using software as previously described⁴⁰. Cells were stained with SCA-1-BV480 (D7; BD Biosciences, San Deigo, CA), CD71-Alexa Fluor 488, (C2; BD Biosciences, San Diego, CA) and LysoBrite^™ NIR (AAT Bioquest, Pleasantown, CA) for Gprasp2^high/low LT-HSC imaging and SCA-1-BV480 and LysoBrite^™ NIR for Gprasp2^high LT-HSC imaging with shRNA treatment.

Bulk RNAseq

Cell populations were sorted directly into lysis buffer and total RNA was isolated (RNeasy Micro kit; QIAGEN). Sequencing was performed at Vantage (Vanderbilt University Medical Center, Nashville, TN). Vantage performed mRNA enrichment and cDNA library preparation utilizing the stranded mRNA (polyA-selected) library preparation kit. Sequencing was performed at Paired-End 150 bp on the Illumina NovaSeq 6000 targeting an average of 50M reads/sample. RNA-SEQ FASTQ sequences were mapped to the GRCm38 (mm10) mouse genome using STAR (version 2.7.9a)⁶⁸. Gene counts were calculated by STAR using Gencode annotation v22⁶⁹. Differential expression analysis was performed using DESEQ2 using log fold shrinkage with the ashr method^{70, 71}. For the Gprasp2^high vs Gprasp2^low comparison, data from individual mice were paired. Heatmaps of expression data used CPM normalized counts. For comparison of Gprasp2 shRNA to Control shRNAGprasp2^high/low LT-HSCs were analyzed by DESEQ2 separately. Paired analysis was performed based on pools of cells. Heatmaps of expression data used rlog normalized counts. PCA analysis was performed for general visualization and clustering of samples. The DESEQ2 Wald statistic was used to rank genes from differential expression analysis of both the Gprasp2^high vs Gprasp2^low and the Gprasp2 shRNA to Control shRNA Gprasp^high comparisons. The ranked gene lists were used as input for pre-ranked gene set enrichment analysis in GSEApy (v1.1.4) using the “MSigDB_Hallmark_2020” and “GO_Biological_Process_2023” gene set libraries obtained from Enrichr⁷⁶. Enriched pathways were filtered with an FDR cutoff of 0.05, and the resulting plots were generated in Python using Seaborn⁷².

scRNAseq Analysis

Analysis of publicly available datasets (Nestorowa et al. 2016, GEO:GSE81682 and Zhang et al. ArrayExpress: E-MTAB-9922) was performed using Seurat⁷³. Gene symbols were derived from Ensembl IDs using BioMart⁷⁴. Centered log ratio transformation (CLR) was used for normalization. Differential expression analysis was performed on Gprasp2^high/low LT-HSCs. For GSE81682, the LTSHC population normalized counts were divided into Gprasp2^high (Gprasp2 expression >2) and Gprasp2^low (Gprasp2 expression=0). For E-MTAB-9922, values of 0.4 and 0 were used for cutoffs respectively. For each gene in each Gprasp2 population, the expression value in all samples were summed. Summed expression data was used as input for NOISEQ differential expression analysis⁷⁵. NOISEQ rankings were used as input for GSEA analysis as described in bulk RNA-SEQ methods description.

LT-HSC proteomic profiling

20 000 Gprasp2^high/low LT-HSCs from three independent biological replicates were sorted into PBS. Cells were centrifuged at 350 g for 10 minutes and PBS was removed, leaving ~5 μL. Samples were snap-frozen on dry ice and placed in −80°C. Cell pellets were lysed in lysis buffer (50 mM HEPES, pH 8.5, 8 M urea and 0.5% sodium deoxycholate). 0.5ug of protein for each sample was digested with LysC (Wako). Samples were diluted to 2M Urea concentration with 50mM HEPES (pH 8.5) and further digested with Trypsin (Promega). The peptides in the digests were labeled with 18-plex Tandem Mass Tag (TMTpro) reagents (Thermo Scientific) following the manufacturer’s recommendation. The unreacted labels were quenched with 5% Hydoxylamine. The labeled samples were mixed equally and desalted using C18 cartridge (Harvard Apparatus) and dried by speed vac. The pooled sample was fractionated on an offline HPLC (Agilent 1220) using basic pH reverse phase liquid chromatography (pH 8.0, XBridge C18 column, 1 mm x 50 mm, 3.5 μm particle size, Waters). The fractions were dried and resuspended in 5% formic acid and analyzed by acidic pH reverse phase LC-MS/MS analysis. The peptide samples were loaded on a nanoscale capillary reverse phase C18 column (Thermo PepMap RSLC, 50 μm ID x 15 cm, 1.9 μm C18) by a HPLC system (Thermo Ultimate 3000) and eluted by a 90 min gradient. The eluted peptides were ionized by electrospray ionization and detected by an inline Orbitrap Exploris 480 mass spectrometer (Thermo Scientific). The mass spectrometer is operated in data-dependent mode with a survey scan in Orbitrap (60,000 resolution, 2 X 10⁵ AGC target and 50 ms maximal ion time) and MS/MS high resolution scans (60,000 resolution, 1 X 10⁵ AGC target, 110 ms maximal ion time, 34 HCD normalized collision energy, 1.5 m/z isolation window, and 15 s dynamic exclusion).

LT-HSC mass spectrometry (MS) data analysis

The MS/MS raw files were processed by JUMP, a tag-based hybrid search algorithm that integrates de novo sequence tag generation with database matching to achieve high sensitivity and specificity⁷⁶. The data was searched against the UniProt mouse database concatenated with a reversed decoy database for evaluating false discovery rate. Searches were performed using a 25 ppm mass tolerance for precursor ions and 25 ppm mass tolerance for fragment ions, fully tryptic restriction with two maximal missed cleavages, three maximal modification sites, and the assignment of a, b, and y ions. TMT tags on Lysine residues and N-termini (+304.2071453 Da) were used for static modifications and Met oxidation (+15.99492 Da) was considered as a dynamic modification. MS/MS spectra were filtered by mass accuracy and matching scores to reduce protein false discovery rate to ~1%. Proteins were quantified by summing reporter ion counts across all matched PSMs using JUMP⁷⁶ and were imported into the JUMPshiny⁷⁷, an interactive R Shiny platform for differential abundance analysis. JUMPshiny performs quality assessment by the intensity distribution, principal component analysis, and sample correlation metrics. The protein intensities were log2 transformed and normalized by trimming the top and bottom 15% of the data with median adjustment. Differential protein abundance analysis was performed using the limma R package by fitting linear models and applying empirical Bayes moderation to estimate the log2 fold change and calculate both unadjusted and FDR adjusted p-values⁷⁷. Pathway and gene set enrichment analyses were performed in R using Bioconductor packages clusterProfiler, DOSE, ReactomePA, and msigdbr, with mouse gene annotation provided by org.Mm.eg.db. Enrichment analyses were conducted using both over-representation analysis (ORA) and gene set enrichment analysis (GSEA) across Gene Ontology (GO), KEGG, WikiPathways, Reactome, and MSigDB gene set collections. The input to the enrichment analyses was a proteomics table including the protein UniProt IDs, log2 trimmed median normalized protein intensities for each sample and the limma statistical results. Proteins were indexed by the UniProt identifiers and were mapped to mouse Entrez Gene IDs using clusterProfiler::bitr and the org.Mm.eg.db mouse database. All proteins were retained in downstream analyses; one-to-many UniProt – Entrez mappings were preserved.

GSEA was performed using all quantified proteins, without applying a significance cutoff, to identify coordinated shifts in pathway-level expression. In contrast to ORA, GSEA rankings were derived directly from the sample-level log2-normalized protein intensities, not from limma significance calls. Protein-level statistics were computed across samples for the sample group comparison to generate a pre-ranked gene list, which was then used as input for GSEA. GSEA was performed using clusterProfiler functions gseGO (GO Biological Process (BP), Molecular Function (MF), and Cellular Component (CC)), gseKEGG for KEGG pathways, gseWP for WikiPathways, gsePathway for Reactome pathways, and GSEA for MSigDB collections. All GSEA analyses used the fgsea algorithm for pre-ranked gene lists, with gene set size limits of 10–500 genes, exponent = 1, and ε = 0 (or 1×10⁻¹⁰ for MSigDB). Enrichment significance was assessed using a nominal p-value cutoff of 0.05, with Benjamini–Hochberg correction applied to obtain FDR-adjusted p-values. GSEA results were ordered by absolute normalized enrichment score (∣NES∣) for interpretation and reporting. Pathway directionality was inferred from the normalized enrichment score (NES), reflecting coordinated up- or down-regulation of pathway member proteins.

Immunoprecipitation/ MS

HPC7s were cultured as previously described⁷⁸. Murine Gprasp2 cDNA was cloned into a pLV-EF1α vector with an N-terminal 3X FLAG and HPC7s were transduced with the GPRASP2-3X FLAG vector or an empty vector control. GFP⁺ HPC7 cells were sorted and expanded. FLAG-tagged GPRASP2 proteins were immunoprecipitated using Anti-FLAG M2 Magnetic Beads (Sigma-Aldrich, St. Louis, MO) according to the manufacturer’s protocol. Briefly, cells were washed three times and lysed for 30 minutes on ice followed by five minutes of sonication (30 seconds on, 30 seconds off) using a bioruptor (Diagenode, Denville, NJ). Lysate was spun at 20 000 g for 15 minutes and supernatant was collected and quantified. One milligram of protein was pre-cleared using control agarose resin (Thermofisher, Waltham, MA) for one hour and then incubated with Anti-FLAG M2 magnetic beads for 5 hours at 4°C with end-over-end mixing. Beads were then washed three times for 10 minutes and then elution was performed two times for 30 minutes with 3xFLAG peptide (Sigma, St. Louis, MO). Immunoprecipitate was then speed vacuumed in a Concentrator Plus (Eppendorf, Enfield, CT) for one hour at RT; 4X LDS buffer was added, samples were heated for five min at 55°C and loaded onto a 10% SDS-PAGE gel and stained with Gelcode Blue Stain (Thermofisher, Waltham, MA). Protein in gel bands were reduced with dithiothreitol (DTT) (Sigma) and alkylated by iodoacetamide (IAA) (Sigma, St. Louis, MO). The gel bands were then washed, dried, and rehydrated with a buffer containing trypsin (Promega, Madison, WI). Samples were digested overnight, acidified and the peptides were extracted. The extracts were dried and reconstituted in 5% Formic acid. The peptide samples were loaded on a nanoscale capillary reverse phase C18 column (Thermo Easy Spray (ES904) 75id, 2um C18, 150mm) by a HPLC system (Thermo Ultimate3000) and eluted by an acetonitrile gradient. The eluted peptides were ionized and detected by an inline mass spectrometer (Thermo Orbitrap Exploris 480). The MS and MS/MS spectra were collected over a 75-min liquid chromatography gradient. The mass spectrometer is operated in data-dependent mode with a survey scan in Orbitrap (60,000 resolution, 1 X 10⁶ AGC target and 50 ms maximal ion time) and MS/MS high resolution scans (15,000 resolution, 1 X 10⁵ AGC target, 150 ms maximal ion time, 32 HCD normalized collision energy, 1.5 m/z isolation window, and 8 s dynamic exclusion). Validation by western blot was performed by using 5% of elution against 3 μg input lysate. Nupage LDS buffer with reducing agent and phosphotase/protease inhibitors were added to samples and heated at 95°C for 7 minutes. Samples were loaded on a 4-20% SDS-PAGE gel and blotted using rabbit GPRASP2 primary antibody (1:1000; Atlas antibodies, Stockholm, Sweden), rabbit anti-Histone-3 (Abcam, Cambridge, UK) and anti-rabbit-HRP (1:5000; Cell Signaling Technologies, Danvers, MA) and developed using Clarity Western ECL Substrate (Bio-Rad, Hercules, CA) for chemiluminescence.

MS Database Search and Analysis

Database searches were performed using JUMP² search engine against a composite target / decoy Uniprot mouse protein database³. Searches were performed using a 25-ppm mass tolerance for precursor and fragment ions, fully tryptic restriction with two maximal missed cleavages. Carbamidomethylation of Cysteine (+57.02146 Da) was used for static modification and Met oxidation (+15.99492 Da) was considered as a dynamic modification. All matched MS/MS spectra were filtered by mass accuracy and matching scores to reduce protein false discovery rate to <1%. Spectral counts, matching to individual proteins reflect their relative abundance in one sample after the protein size is normalized. The spectral counts between samples for a given protein is used to calculate the p-value which is derived by G-test⁴.

Endocytosis activity assay

Mouse bone marrow was stained for LT-HSC markers (Lineage⁻SCA-1⁺cKIT⁺CD150⁺CD48⁻) and then incubated with 50 μg/mL pHRodo Dextran Green 10 000 MW (Invitrogen, Waltham, MA) for 20 minutes at RT. Samples were then rinsed with ice-cold PBS to stop endocytic uptake and placed on ice. Samples were run on a flow cytometer immediately after the assay to prevent increased signal from lysosomal acidification. 4°C controls were used as negative controls.

For transduced LT-HSCs, following 48 hours of transduction, 50 μg/mL pHRodo Dextran Green 10 000 MW (Invitrogen, Waltham, MA) was added directly to cells in 96 well plates with antibodies for LT-HSC markers (Lineage⁻SCA-1⁺cKIT⁺CD150⁺EPCR⁺) for 10 minutes at 37°C. Immediately following incubation, media was carefully removed and ice-cold PBS was added to stop endocytic activity. Samples were centrifuged at 350 g for 5 minutes and PBS was removed and an additional PBS wash was performed. Samples were run on a flow cytometer immediately after washing to prevent increased signal from lysosomal acidification.

Poly(I:C) treatment and cell cycle analysis

5mg/kg Poly (I:C) (Invivogen, San Diego, CA) was administered intraperitoneally to Gprasp2-tdTomato mice 24 hours prior to assessment. BM was collected and after staining for surface antigens, cells were fixated with Cytofix/Cytoperm kit (BD Biosciences, San Diego, CA) and stained with Ki-67 and DAPI for cell cycle analysis as previously described⁷⁹.

Dox administration and in vivo cycling assay

Gprasp2-tdTomato-TetOP-H2B-GFP mice were given doxycycline (dox; Sigma D9891, 2 mg/mL, supplemented with sucrose at 10 mg/mL) in drinking water for 6 weeks (8- to 12-week-old mice at the beginning of treatment) as previously described⁹. After 6 weeks, dox was removed and efficiency of transgene expression was validated in LT-HSCs via flow cytometry 24 hours after dox removal. To assay in vivo cycling kinetics, mice were chased after dox treatment for 150 days and then BM was assayed via flow cytometry. Littermate controls that did not receive dox treatment expressing the transgene were used to control for leakiness of H2B-GFP.

Extracellular vesicle (EV) isolation

Sorted LSK cells (Lineage⁻ Sca-1⁺ c-Kit⁺) or SLAM-HSCs (CD150⁺ CD48⁻ LSK) were cultured in StemSpan SFEM II supplemented with 50 ng/mL TPO and 50 ng/mL SCF for 48 hours at 37 °C in a humidified 5% CO₂ incubator. After 48 hours, EVs were isolated from the culture supernatant using SmartSEC Mini EV Isolation System, and 100 μL of EV eluate was collected⁸⁰.

EV release and uptake analysis

To evaluate EV release, EVs were isolated from 3,000 Sca-1^high CD48^low HSCs and 3,000 CD48^highSCA-1^low HSCs sorted from the same mouse. The isolated EVs were diluted 1:10 in filtered H₂O and subjected to nanoparticle tracking analysis (NTA) using the NanoSight Pro (Malvern Panalytical) and NS Xplorer software (version 1.1.0.6, Malvern Panalytical) to measure EV size, distribution, and concentration⁵⁵. For EV uptake analysis, EVs extracted from 5,000 LSK cells and SLAM-HSCs were used. Magnetic streptavidin beads were collected and washed from a 40 μL bead suspension according to the manufacturer’s instructions. The washed beads were then incubated on ice for 2 hours with a mixture of biotin-anti-CD9, biotin-anti-CD63, and biotin-anti-CD81 antibodies (3.3 μL each per reaction) to pre-load the beads. The antibody-loaded beads were subsequently incubated with SEC-purified EVs at 4 °C overnight to allow the formation of EV-bead complexes. After two washes with Beads Wash Buffer, the EV-bead complexes were incubated with PKH26 dye at room temperature for 5 minutes. Excess dye was quenched and removed using 2% BSA/PBS, followed by three additional washes with Beads Wash Buffer. The EV-bead complexes were then rotated at room temperature for 2 hours with 100 μL Exosome Elution Buffer to release PKH26-labeled EVs. The EVs were mixed with 400 μL StemSpan SFEM II and concentrated using Amicon Ultra Centrifugal Filters (14,000 × g, 5 minutes), yielding a final volume of 30 μL concentrated EVs. Concentrated EVs extracted from four mice were pooled and subsequently used for EV characterization and uptake assays. Beads without EVs were subjected to the same PKH26 labeling and concentration procedures to generate a control solution for threshold setting. As EV recipients, 1,000 CD48^lowSCA-1^high HSCs and 1,000 CD48^highSCA-1^low HSCs sorted from the same mouse were suspended in 20 μL of StemSpan SFEM II supplemented with TPO and SCF. For each population, 10 μL of PKH26-labeled EVs (1×10⁹ particles/mL) was added, followed by incubation for 24 hours. After incubation, HSCs were collected, and the proportion and fluorescence intensity of PKH26 uptake (detected in the PE channel) were quantified by gating on SLAM-HSCs using BD FACSAria II.

Statistics

Data are summarized by mean, standard deviation, median, and range by groups. Normality of all samples was checked by Shapiro-Wilk test, and then a two sample t test (one sample t test if control group being normalized) was applied to groups following normal distribution, or two sample exact Mann-Whitney test (one sample Wilcoxon test if control group being normalized) was applied to groups not fulfilling normal assumption, to examine whether there was significant difference between two groups. False Discovery Rate (FDR) method was employed to adjust for multiple testing problem. Analyses were performed in R-3.5.3.

Extended Data

Extended Data Figure 1. Gprasp2 expression is enriched in quiescent mouse and human LT-HSCs.

a, Transcriptomic analysis of mouse LT-HSCs (Lineage⁻SCA-1⁺c-KIT⁺CD48⁻CD150⁺) from Nestorowa et al., 2016 segregated based on Gprasp2 expression. UMAP projection of hematopoietic progenitors showing annotated clusters and Gprasp2 expression. GSEA plots of Gprasp2^high LT-HSCs compared to Gprasp2^low LT-HSCs. b, Transcriptomic analysis of human HSCs (CD34⁺CD38⁻CD45RA⁻CD90⁺CD49f⁺) from Zhang et al., 2022 segregated based on Gprasp2 expression. UMAP projection of LT-HSCs showing GPRASP2 expression. GSEA plots of Gprasp2^high human LT-HSCs compared to Gprasp2^low human LT-HSCs. c, Volcano plot from bulk RNA sequencing of human HSCs (CD34⁺CD38⁻CD45RA⁻CD90⁺CD49f⁺) that were either freshly isolated (quiescent) or cultured for four days (activated) from García-Prat et al., 2021. d, Confocal imaging of GPRASP2 levels in mouse LT-HSCs. Representative image and quantification of GPRASP2 mean fluorescence intensity (MFI) are shown (n=72 cells from a single experiment). Open arrow indicates LT-HSC with low GPRASP2, and closed arrows indicate LT-HSC with high GPRASP2. Scale bars=10 μm.

Extended Data Figure 2. Validation of Gprasp2-reporter model and flow cytometry gating strategies for isolation of blood and bone marrow cells.

a, Flow cytometry gating strategy for isolation of mouse HSPCs and LT-HSCs. b, Representative post-sort purity check for LT-HSC Gprasp2-tdTomato high and low populations. c, Mouse bone marrow HSPCs were sorted based on low, medium, or high tdTomato MFI and subject to RT-qPCR for Gprasp2 transcript levels (n=3 independent biological replicates). A linear regression analysis was performed to correlate tdTomato intensity with Gprasp2 transcript levels. R2 values are shown in the table above. Representative flow cytometry gating strategy for BM hematopoietic progenitors (d) and peripheral blood cells (e). f, Volcano plot from bulk RNA sequencing comparing Gprasp2^high LT-HSCs to Gprasp2^low LT-HSCs from Gprasp2-tdTomato mice (n=3).

Extended Data Figure 3. Gprasp2 defines a hierarchy of LT-HSCs with durable stemness and self-renewal.

%CD45.2⁺ PB lineages of primary (a) and secondary (b) recipients of Gprasp2^high and Gprasp2^low LT-HSCs and competitor cells. For primary transplants, n≥12 mice per condition, three independent experiments. For secondary transplants, n≥11 mice per condition, three independent experiments. c, Frequency of CD45.2⁺ cells in BM hematopoietic progenitors from secondary recipients in (b). d, %CD45.2⁺ PB lineages of tertiary recipients of Gprasp2^high and Gprasp2^low LT-HSCs and competitor cells (n≥4 mice per condition, two independent experiments. e, Frequency of CD45.2⁺ cells in BM hematopoietic progenitors from tertiary recipients in (d) (n≥4 mice per condition, two independent experiments). f, Experimental schematic of ex vivo LT-HSC expansion of Gprasp2^high/low LT-HSCs. g, Total cell singlets from ex vivo expansion of sorted Gprasp2^high/low LT-HSCs for 2 weeks. (n=3 technical replicates, 1 experiment). h, LT-HSC fold expansion of sorted Gprasp2^high/low LT-HSCs for 2 weeks. LT-HSCs were defined by Lineage⁻cKIT⁺SCA-1⁺CD150⁺EPCR⁺ (n=3 technical replicates, 1 experiment). i, Representative gating strategy for LT-HSCs in ex vivo expansion culture. j, %CD45.2⁺ PB lineages of non-competitive primary recipients of Gprasp2^high and Gprasp2^low LT-HSCs. k, Frequency of CD45.2⁺ cells in BM hematopoietic progenitors from primary recipients in (j). n≥9 mice per condition, three independent experiments. l, %CD45.2⁺ total PB, %CD45.2⁺ PB lineages, and total blood counts of non-competitive secondary recipients of Gprasp2^high and Gprasp2^low LT-HSCs. n≥7 mice per condition, two independent experiments. Steady-state line represents the average blood counts from ten unmanipulated C57Bl/6 mice. m, Frequency of CD45.2+ cells in BM hematopoietic progenitors from secondary recipients in (l) (n≥4 mice per condition, two independent experiments). *p < 0.05, **p < 0.01, ***p < 0.001. Mann-Whitney test unless otherwise indicated.

Extended Data Figure 4. HSC stemness and pluripotency correlates with high Gprasp2 expression at a clonal level.

a, % CD45.2⁺ PB from Gr-1⁺ Cd11b⁺ myeloid cells, B220⁺ B cells, or CD4⁺ CD8⁺ T cells in lethally irradiated recipients from single Gprasp2^high/low LT-HSCs (n≥48 per condition, 3 experiments). b, Total % CD45.2⁺ PB, and % CD45.2⁺ PB from Gr-1⁺ Cd11b⁺ myeloid cells, B220⁺ B cells, or CD4⁺ CD8⁺ T cells in two representative Gprasp2^high clones showing durable engraftment kinetics. c, Total % CD45.2⁺ PB, and % CD45.2⁺ PB from Gr-1⁺ Cd11b⁺ myeloid cells, B220⁺ B cells, or CD4⁺ CD8⁺ T cells in a representative Gprasp2^high clone showing transient engraftment kinetics. d, Frequency of CD45.2⁺ BM hematopoietic progenitors (left panel) and frequency of CD45.2⁺ LT-HSCs (right panel) from durable recipients used for secondary transplant. e, Experimental schematic of secondary transplantation into 3 mice of 5 x 10⁶ WBM cells isolated from primary recipients of purified single Gprasp2^high/low LT-HSC transplantation with 150 000 supporting whole bone marrow cells. % CD45.2⁺ PB (f), and % CD45.2⁺ PB from Gr-1⁺ Cd11b⁺ myeloid cells, B220⁺ B cells, or CD4⁺ CD8⁺ T cells (g) in lethally irradiated recipients from secondary transplant from primary single LT-HSC recipient mice. Each line represents the repopulating activity of a specific donor (n≥2, 2 independent experiments). h, Frequency of CD45.2⁺ BM hematopoietic progenitors (left panel) and frequency of CD45.2⁺ LT-HSCs (right panel) from recipients from secondary transplant. i, Experimental schematic of purified LT-HSC index sorting from Gprasp2-tdTomato reporter mice into methylcellulose for colony forming assays. j, Colony size and colony type scoring 9 days after plating plotted by Gprasp2-tdTomato MFI (n=122; 1 experiment; Kruskal-Wallis test with post hoc Dunn’s multiple comparison). *p < 0.05, **p < 0.01.

Extended Data Figure 5. Gprasp2^high marks deeply quiescent HSCs.

a, Representative flow cytometry plot and gating strategy of LT-HSCs stained with Ki67/DAPI for cell cycle analysis. b, In vivo LT-HSCs division kinetics tracking using Gprasp2-tdTomato crossed with ROSA26^rtTa/rtTA; pTRE-H2B^GFP+/GFP+ mice treated with doxycycline for 6 weeks compared to untreated littermates. After 24 hours, bone marrow was analyzed to assess labeling efficiency. c, Representative histograms and bar chart of LT-HSC labeling efficiency of doxycycline-treated mice 24 hours after removing dox-treated water (n=3, two independent experiments). d, Mitotracker analysis via flow cytometry in Gprasp2^high/low LT-HSCs (n=4, 2 experiments). e, Mean intensity of CD71 before and after division (n≥277). f, Mean intensity of Lysobrite before and after division (n≥277). g, Mean intensity of SCA-1 before and after division (n≥277). h, Flow cytometry histograms of Gprasp2-tdTomato MFI in LT-HSCs after 24 hours and MFI in poly(I:C) treated mice compared to PBS controls (n≥5, 2 experiments). *p < 0.05, ****p < 0.0001. Mann-Whitney test unless otherwise indicated.

Extended Data Figure 6. Gprasp2 deficiency reduces quiescence and stemness maintenance.

a, Gprasp2 expression 48 hours after lentiviral knockdown in Gprasp2^high and Gprasp2^low LT-HSCs. Extracted from RNA sequencing data described in Fig. 5a (n=3 biological replicates, 3 experiments). b, Representative western blot for GPRASP2 protein expression 48 and 72 hours after lentiviral knockdown in mouse HSPCs. c, Quantification for (b). d, % of mCherry+ cells expression 48 and 72 hours after lentiviral knockdown in mouse HSPCs. e, Principal component (PC) analysis for RNAseq samples described in Fig. 4a. f, Number of DEGs between Gprasp2 shRNA vs. Control shRNA for RNAseq samples from Fig. 4a and volcano plot with DEG for Gprasp2^high Gprasp2 shRNA vs control shRNA. g, Heat map of expression of genes involved in in GTPase Mediated signal transduction in Gprasp2^low control and Gprasp2-shRNA LT-HSCs. h, Cell division time and fluorescence intensity (tdTomato, lysobrite, SCA-1) from time-lapse imaging samples described in Fig. 4a.

Extended Data Figure 7. Gprasp2 deficiency accelerates stem cell exhaustion via endocytosis dysregulation.

a, Principal component (PC) analysis for proteomic samples described in Fig. 5a (n=3). b, Correlation between RNA and protein expression in Gprasp2^high and Gprasp2^low LT-HSCs. Rho = Spearman's rank correlation coefficient. c, DEPs in Gprasp2^high compared to Gprasp2^low LT-HSCs grouped by their function in signaling cascades. d, Validation of GPRASP2-FLAG overexpression in HPC7 cells for co-immunoprecipitation/mass spectrometry experiments. e, Mass spectrometry measurements of proteins co-immunoprecipitated with GPRASP2 control and GPRASP2-FLAG overexpressing HPC7 cells. Intensity normalized to highest value (2 independent experiments). f, LAMP1 analysis via flow cytometry in Gprasp2^high/low LT-HSCs (n=4, 2 experiments). g, Schematic for mechanism of action of endocytosis inhibitors. h, Dextran-pHrodo green MFI in LT-HSCs treated with endocytosis inhibitors after 7 days in expansion media. i, Dextran-pHrodo green MFI in LT-HSCs treated with endocytosis inhibitors after 48 hours in maintenance media. j, Viability of LT-HSCs treated with endocytosis inhibitors after 7 days in expansion media. k, Total number of cells in cultures of LT-HSCs treated with endocytosis inhibitors after 7 days in expansion media. l, Total number of cells in cultures of LT-HSCs treated with endocytosis inhibitors after 48 hours in maintenance media. m, Number of HSPCs in cultures of LT-HSCs and MPPs treated with endocytosis inhibitors after 48 hours in maintenance media.

Extended Data Figure 8. Grand SLAMS (Gprasp2^high LT-HSCs) can be captured in CD48^low SCA-1^high LT-HSCs for purification and functional testing,

a, % CD45.2⁺ PB from Gr-1⁺Cd11b⁺ myeloid cells, B220⁺ B cells, or CD4⁺CD8⁺ T cells in lethally irradiated recipients from CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs (n≥11 per condition, 3 experiments). b, Experimental schematic for transplantation of 200 CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs into primary and secondary lethally irradiated recipients. c, Schematic of sorting strategy based on CD48 and SCA-1 expression in serial transplantation. d, % CD45.2⁺ peripheral blood in lethally irradiated recipients of 200 CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs in secondary recipients (n≥3 per condition, 1 experiment). e, % CD45.2+ PB from Gr-1⁺Cd11b⁺ myeloid cells, B220⁺ B cells, or CD4⁺CD8⁺ T cells in secondary CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSC recipient mice (n≥3 per condition, 1 experiment). f, WBC concentrations from secondary CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSC recipient mice (n≥3 per condition, 1 experiment). Steady-state line is the average WBC counts from n=10 CD45.1⁺/CD45.2⁺ unmanipulated mice. g, Representative plots for PKH26 intensity after uptake of LT-HSC- or HSPC-derived EVs. Dashed line represents the threshold for PHK26+ cells. h, Nanoparticle tracking analysis (NTA) size distribution profiles of EVs derived from CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs. Curves represent the mean values from six paired samples. i, EV concentration measured by NTA in CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs cultures (N=6). j, Mean EV size measured by NTA in CD48^lowSCA-1^high and CD48^highSCA-1^low LT-HSCs cultures (N = 6). *p < 0.05, **p < 0.01, ***p < 0.001. Mann-Whitney test unless otherwise indicated.

Data Availability

The single cell RNAseq datasets from Extended Data Fig. 1a-b were reanalyzed from GSE81682 and E-MTAB-9922, respectively. Bulk RNAseq data from Extended Data Fig. 1c was reanalyzed from GSE153917. Bulk RNAseq performed from Fig. 1d-e and Fig. 5b, Extended Data Fig. 6a, e-f can be found under accession number GSE316660. Bulk proteomics performed from Fig. 6b, Extended Data Fig. 7a-b can be found under PRIDE accession number PXD072835.