Sphingolipid Modulation Activates Proteostasis Programs to Govern Human Hematopoietic Stem Cell Self-Renewal

1Humans have an enormous demand (∼10e12 cells daily) for hematopoietic output. Meeting this need over a lifetime raises an inherent risk for malignancy due to DNA replication errors and potential metabolic and cellular stressors that cause cellular damage. Yet hematologic malignancies remain relatively rare. The hierarchical organization of blood is thought to mitigate the risk of transformation. All mature lineages descend from a pool of rare and largely quiescent long-term hematopoietic stem cells (LT-HSCs) that in turn generate highly proliferative but more short-lived populations, including short-term HSCs (ST-HSCs) and committed progenitors. Only LT-HSCs are able to regenerate the entire blood system upon transplantation and maintain long-term output due to their unique capacity for self-renewal. Whereas dormancy in LT-HSCs minimizes the possibility of detrimental mutations passing to all progeny, damage arising within the rapidly proliferating downstream progenitors is ultimately purged upon terminal differentiation. However, recent studies have found that dormancy is not the only feature that uniquely protects LT-HSCs. Compared to downstream progenitors, stressed or damaged LT-HSCs respond by activating pathways that result in culling rather than pathways that serve to repair or resolve the damage, as is typical for progenitors and most mammalian cell types. How this unique LT-HSC fate choice is made is poorly understood, particularly upon quiescence exit when LT-HSCs must also dynamically shift metabolic state. The unfolded protein response (UPR) and macroautophagy, hereafter referred to as autophagy, have been identified as critical mediators for HSC stress responses. More broadly, these new findings revealed an unexplored layer of metabolic and organelle biology that influences LT-HSC regulation of self-renewal, quiescence, proliferation, and lineage commitment that extends beyond the traditional focus on transcription factor networks governing fate decisions.

2The energy and macromolecule requirements of HSCs are distinct from progenitors, and genetic ablation of key metabolic regulators in mouse models often lead to HSC exhaustion; as such, stringent control of cellular metabolism is fundamental for HSC function. Recent data suggest that HSCs contain integrated cellular networks coordinating proteostasis with dynamic biosynthetic and metabolic states to govern stem cell fate decisions, especially in the transition from dormancy to cellular activation. Presumably, quality control mechanisms ensure HSC organelle and proteostatic health upon damage incurred from cell cycle entry and other basal metabolic stressors, such as reactive oxygen species (ROS). However, metabolic as well as translation initiation heterogeneity cannot be attributed entirely to cellular identity differences between LT-HSCs and committed progenitors; dormant HSCs exhibit lower biosynthetic activity and protein synthesis than active HSCs, suggesting HSC metabolic requirements are highly adaptive to cellular state. Additionally, the dysregulation of quality control mechanisms (e.g., autophagy) that occur upon HSC aging results in metabolic changes, suggesting that HSC fate is intimately entangled with metabolic and proteostatic regulation. Enhanced endoplasmic reticulum (ER) function in human HSCs with enforced expression of the chaperone ERDJ4/DNAJB9 confers protection against the ER stress that is induced upon xenotransplantation, thereby preserving self-renewing HSCs. Moreover, we recently showed higher levels of activating transcription factor 4 (ATF4) and components of the pro-survival integrated stress response (ISR) in human HSCs are cytoprotective during homeostasis, despite their lowered threshold for culling with strong stress stimuli. Thus, a complex picture emerges suggesting the nature of the stress stimulus is important to fine-tune quality control responses for HSC persistence or culling. However, these studies raise a number of key questions: is there a single stimulus that activates both autophagy and the UPR to confer cytoprotective functions to HSCs; is there a coordinated cellular stress response upon stress induction; and how do proteostasis programs impact HSC self-renewal?

3Proper management of lipid composition is integral for maintaining cellular membrane dynamics for cell division and signaling in cell lines. Although lipid homeostasis is critical for cellular and organismal health, the exploration of lipid metabolism in HSC function is limited. Although lipostatic stress as well as proteostatic stress is known to converge on the activation of the UPR for pro-survival in human cells, and autophagy is a crucial element for maintaining lipid homeostasis in mice, there is a critical lack of understanding on the impact of altering lipid composition to HSC function beyond the connection of fatty acid metabolism and mitochondrial function. Ceramide (Cer), the central component of sphingolipids (SpLs), has been proposed to be part of a lipid biostat that regulates cellular stress and activates stress responses. DEGS1 (delta 4-desaturase, Sphingolipid 1, or DES1) is an ER-membrane-spanning protein and the final enzyme in de novo SpL synthesis, which converts dihydroceramide (dhCer) to Cer; both genetic ablation and inhibition with the synthetic retinoid fenretinide/N-(4-hydroxyphenyl) retinamide (4HPR) are sufficient to activate autophagy in mouse cells or human cell lines. The potent bioactive lipid sphingosine-1-phosphate (S1P) is generated from metabolism of Cer and exerts pleiotrophic signaling roles in proliferation, survival, and migration of immune cells and HSCs, as well as regulation of lymphocyte lineage commitment and hematopoietic stem and progenitor cell (HSPC) function. Although S1P signaling regulates mouse hematopoiesis, whether other SpLs are important to HSC function is unknown.

4Here, we uncovered a transcriptional signature of genes governing SpL synthesis that distinguishes human HSCs and committed progenitors. This distinct SpL wiring in the human hematopoietic hierarchy was confirmed with sphingolipidomic analysis of sorted mature and HSPC populations. When human HSCs are placed in ex vivo conditions thought to promote cord blood (CB) HSC activation and expansion, they actually lose HSC function due to impaired proteostatic programs. By contrast, inhibition of DEGS1 in human HSCs with 4HPR treatment before quiescence exit in ex vivo culture induced a coordinated response of proteostatic cellular stress programs, including autophagy to maintain HSC self-renewal. Despite ex vivo culture, HSCs following SpL modulation functionally show higher self-renewal relative to cultured cells without treatment pointing to a linkage between SpLs, proteostatic quality control programs, and HSC self-renewal in the transition from quiescence to cellular activation.

5We undertook transcriptome analysis of highly resolved subpopulations of the human hematopoietic hierarchy and found that lipid signaling and metabolism genes involved in SpLs are differentially expressed (false discovery rate [FDR] < 0.05; fold change [FC] > 1.5) in LT-HSCs and ST-HSCs (as defined in,, ) compared to committed progenitors (Figures 1A and S1A). Previous lipid measurements of mammalian cells indicated that SpLs contribute only ∼10% of the cellular lipidome, mostly represented by structural sphingomyelins (SMs) and glycosphingolipids. Overlaying the differentially expressed SpL genes (Figure 1A) onto the metabolic pathway showed many of the SpL genes highly expressed in HSCs centered around those involved in the synthesis of the low abundant bioactive dhCer and Cer species (Figure S1A). To assess whether there is distinct SpL biosynthesis across the cell types comprising the human hematopoietic hierarchy, especially at the level of these less abundant SpLs, we isolated CD34+CD38− stem-enriched (stem) and CD34+CD38+ progenitor-enriched (progenitor) cells and 5 mature blood lineages (B and T lymphocytes, monocytes, neutrophils, and erythrocytes) from CB by flow cytometry. These populations were subjected to Cer, dhCer, sphingosine, S1P, dhSph, dhS1P, hexosylceramides (HexCer) (Cer containing glucose or galactose), and SM measurement using liquid chromatography mass spectrometry (LC-MS) (Figures 1B and S1B–S1H). SMs were the most abundant SpLs in our analysis (Figures 1B and S1H; 72%–94%), consistent with previous lipidome profiling in mammalian cells. Importantly, our profiling identified the accumulation of S1P specifically in erythrocytes (Figure S1E), confirming this lineage-specific association and the robustness of our sphingolipidome profiling. We found no significant differences in SpL content between stem and progenitor cells except in the amount of dhCer carrying the C16:0 fatty acid, providing evidence for differential wiring of de novo SpL synthesis at the lipid level in HSPCs (Figure 1C). By contrast, the mature lineages showed significant differences from stem and/or progenitor cells (Figures S1C–S1H). Importantly, we saw that B cells, neutrophils, and erythrocytes were significantly different in their ratio of Cer/dhCer from stem cells (Figure 1D). In contrast, T cells and monocytes did not differ in the Cer/dhCer ratio, raising the question of whether Cer homeostasis regulates HPSC fate and lineage commitment decisions. DEGS1 expression levels are significantly increased in LT-HSCs, ST-HSCs, and granulocyte-monocyte progenitors (GMPs) following 6 h of cytokine stimulation, suggesting increasing de novo SpL-generated Cer pools may be an early event in the transition from quiescence to cellular activation (Figure 1E). To determine whether alterations in the Cer/dhCer ratio were functionally relevant in HSPC, we modulated their ratio through DEGS1 perturbation and asked whether DEGS1 was required for in vivo repopulation. A lentiviral knockdown (KD) construct to DEGS1 was generated that decreased DEGS1 gene expression to 37% of shControl (shCtrl) in a cell line model (Figure S1I). CB stem cells were transduced in vitro with either shCtrl or shDEGS1 vectors co-expressing blue fluorescent protein (BFP) and transplanted into mice. At 4 weeks, we found DEGS1 KD significantly decreased human CD45+ BFP+ chimerism by 2-fold relative to BFP+ input and resulted in lineage skewing with an increase in myeloid cells at the expense of B lymphoid cells (Figures 1F, 1G, and S1J–S1O). In summary, we have uncovered considerable diversity in SpL composition across the human hematopoietic hierarchy and found that human HSPCs require DEGS1 in vivo.

6To determine whether DEGS1 is required as part of the proper transition from quiescence to cellular activation and/or self-renewal maintenance in human HSCs, its function must be inhibited at the start of quiescence exit. As lentiviral transduction of CD34+ cells requires some period of cytokine pre-stimulation to induce cellular activation, we used the irreversible DEGS1 inhibitor 4HPR at 2 μM at the start of cytokine activation. This dose caused 50% maximal reduction in total cells of lineage-depleted CB (lin-CB) cells after 3 days in culture (Figure S2A). 4HPR treatment for 8 days altered SpL composition, resulting in decreased Cer and increased dhCer compared to controls (Figures 2A, S2B, and S4A–S4D). Bromodeoxyuridine (BrdU) incorporation in LT-HSCs, ST-HSCs, and GMPs was decreased by 4HPR treatment at day 3, suggesting modulating Spls in HSPCs decreases proliferation in vitro (Figure S2C). Nonetheless, 4HPR treatment still permits quiescence exit and does not significantly alter cell viability (Figures S2D and S2E). Similarly, there was no difference in cell cycle distribution between shCtrl and shDEGS1 in LT-HSCs or in GMPs isolated from mice at 4 weeks transplantation, suggesting DEGS1 KD does not lock cells into quiescence in vivo (Figure S2F). Next, we utilized the colony-forming cell assay (CFC) to assess in vitro progenitor function to determine whether DEGS1 inhibition alters functional output from HSPC subpopulations. 4HPR treatment significantly increased clonogenic output specifically from LT-HSCs (50% over control), but not from ST-HSCs or GMPs (Figure 2B). Treatment of LT-HSCs and ST-HSCs with 4HPR led to an increased proportion of granulocyte-macrophage (GM) and monocyte (M) colonies (Figure 2C). Flow cytometry confirmed the morphologic CFC assessment that 4HPR enhanced myeloid output at the expense of erythroid (Figures 2D and 2E). As 4HPR resembles all-trans retinoic acid (ATRA) and may have additional effects independent of DEGS1, CFC assays were performed on LT-HSCs, ST-HSCs, and GMPs treated with ATRA but did not mimic 4HPR treatment (Figure S2G). Although 4HPR effects appeared independent of the retinoid pathway, other potential off-target effects on HSC function were possible. Thus, we turned to a genetic approach with DEGS1 KD and transduced LT-HSCs, ST-HSCs, and GMP with shCtrl or shDEGS1 vectors and sorted BFP+ transduced cells for CFC assays. We found that shDEGS1 LT-HSCs showed similar but more modest changes in colony distribution and loss of erythroid output (Figures 2F and 2G). The more modest effects of DEGS1 KD may be reflective of the more modest but measurable changes in Cer and dhCer levels compared to control KD in cells isolated from xenografts (Figures 2H, 2I, and S2H). Thus, DEGS1 modulation pharmacologically or via KD alters lineage balance in vitro and 4HPR modulation of SpL homeostasis selectively enhances clonogenic efficiency only in LT-HSCs.

7To directly test whether 4HPR modulation of SpL impacts LT-HSC function, we utilized the xenograft assay to selectively readout self-renewal capacity from culture-generated progeny with HSC function. Our approach to test 4HPR was guided by ex vivo expansion methods that are used clinically to expand CB-derived HSPCs, as single CB units do not contain enough HSCs to enable HSC transplantation (HSCT) into adult recipients in the clinic. Thus, we added 4HPR into an expansion scheme with lin− CB cells containing LT-HSCs, ST-HSCs, and committed progenitors that was similar to recently developed state-of-the-art methods for CB transplantation. Three initial cell concentrations corresponding to high (16), medium (3.2), and limiting (0.65) long-term repopulating cell (LTRC) doses were cultured ex vivo for 8 days mimicking fed-batch conditions and then their progeny were analyzed by flow cytometry and assessed for engraftment potential by xenotransplantation (Figure 3A; see STAR Methods). 4HPR-treated cultures had fewer CD34+ cells, enhanced myeloid differentiation, and decreased erythroid differentiation (Figures 3B–3F). Despite fewer CD34+ cells at transplantation, human CD45 chimerism for control and 4HPR-treated progeny were similar in all cell doses following xenotransplantation (Figure 3G). Similarly, lineage distribution for B lymphoid, myeloid, and erythroid lineages (Figures 3H–3J) and the percentage of lin− CD34+ cells (Figure 3K) from the 16-LTRC dose was comparable for control and 4HPR-treated groups. Flow cytometry analysis of the HSC hierarchy in individual mice confirmed all HSPC subpopulations analyzed exhibited similar experimental variation between mice transplanted with control and 4HPR-treated cells (Figures S3A–S3I). These data indicate that, despite reduced CD34+ cell output, the quality and quantity of functionally defined repopulating HSCs from 4HPR-cultured cells at 16 weeks xenotransplantation was similar to controls where far higher numbers of CD34+ cells were transplanted. Moreover, lineage differentiation effects resulting from ex vivo modulation of SpLs in HPSC by 4HPR are transient and reversible following xenotransplantation because the number of HSPCs regenerated in the mice was equivalent to controls. In contrast, when DEGS1 is genetically modulated via KD, we observed a severe decrease (4.5-fold decrease over shCtrl; Figure S3J) in the number of CD34+ at 4 weeks in vivo. Hence, reducing DEGS1 activity transiently or persistently restrains the generation of CD34+ progeny.

8To determine whether 4HPR treatment caused a potential dissociation between phenotypic and functional HSPCs during culture, we quantitatively assess the HSC self-renewal capacity of 4HPR-treated CB progeny using serial transplantation approaches. In contrast to previous strategies that focused on the expansion of CD34+ cells to improve the clinical utility of CB as a HSCT source, SpL modulation by 4HPR here restrained proliferation and resulted in 3.3-fold fewer CD34+ cells relative to control treatment (Figure 3F). Therefore, we investigated in parallel the effects of two known CB CD34+ agonists UM171 and SR1 alone and in combination with 4HPR. First, we assessed the clonogenic activity of LT-HSCs in CFC assays and found that 4HPR significantly increased clonogenic activity in combination with either UM171 or SR1 to levels similar to 4HPR alone (Figure 4A). In contrast, UM171 or SR1 alone did not increase clonogenic activity of LT-HSCs. Next, we cultured 2 independent pools of lin−CB with DMSO control, 4HPR, UM171+SR1 (U+S), and UM171+SR1+4HPR (3-Factor) for 8 days and found 4HPR treatment restricted the percentage and number of CD34+ cells, and 3-Factor treatment resulted in an enhanced percentage of CD34+ cells comparable to U+S treatment (Figure 4B). The total number of cells following 8 days culture was similar between 4HPR and 3-Factor treatment (Figure S4A). We performed LC-MS analysis and found the SpLs measured in 3-Factor progeny phenocopied 4HPR treatment and were similarly perturbed (Figures S4B–S4D). We further analyzed CD34+ cells at day 8 culture for cultured LT-HSCs (cLT:CD90+CD45RA−), ST-HSCs (cST:CD90−CD45RA−), and cultured committed progenitors (cProg:CD90−CD45RA+; Figure 4C). 4HPR alone significantly limited the percentage of cProg cells while increasing cST cells in the CD34 compartment and trended to increased cLT cells relative to control; 3-Factor treatment resulted in significant enrichment of both cLT and cST populations (Figures 4C and 4D). The functional identity of these subpopulations was confirmed in an independent experiment by xenotransplantation for 16 weeks; as expected, cProg cells were unable to give rise to human CD45 grafts (Figures S4E–S4G). Hence, these data show 4HPR acts dominantly over UM171 or SR1 to enhance LT-HSC clonogenic ability, and 3-Factor treatment during ex vivo culture increases immunophenotypic HSC subsets.

9To determine whether 4HPR alone or 3-Factor enhances in vivo HSC self-renewal following ex vivo treatment, lin− CB progeny cultured with control, 4HPR, U+S, or 3-Factor for 8 days were serially transplanted into NSG mice (Figures 3G and S4H). In line with previous work on UM171 and SR1, U+S or 3-Factor ex vivo treatment significantly increased primary engraftment of CB cells (Figure S4H; ). Next, human CD45+ cells engrafted at 16-LTRC doses were serially transplanted in a limiting dilution assay (LDA) to enumerate functional LTRCs from the primary recipients for 5 independent pools of CB treated with 4HPR; the total period of engraftment was 32 weeks. 4HPR-treated cells showed a significant 2.5-fold increase in LTRC frequency over control upon secondary transplantation (Figures 4E and 4F; p = 0.039). In the 2 experiments where U+S were also assessed, 4HPR alone resulted in 2.1-fold increase in LTRC frequency, similar to the frequency obtained by combining the five experiments (Figures 4E, 4F, S4I, and S4J). However, the LTRC frequency of U+S-treated cells was similar to control. Importantly, 3-Factor-treated cells gave a 4.7-fold augmentation in LTRC frequency over cultured control cells (p = 0.007). Collectively, these transplantation data showed that CB expansion with 4HPR alone and in combination with U+S successfully sustained a pool of functional HSCs during ex vivo culture with enhanced in vivo HSC self-renewal capacity relative to controls.

10To elucidate the biological pathways altered by 4HPR treatment in human HSPCs that contributed to lineage differentiation effects and maintenance of HSC self-renewal during cytokine activation, we performed RNA sequencing (RNA-seq) on lin− CB cells at days 2 and 4 following treatment with control, 4HPR, U+S, and 3-Factor (n = 3). GSEA pathway analysis identified 473 significantly upregulated gene sets and 80 significantly downregulated gene sets (FDR < 0.05) in 4HPR compared to controls at day 2 (Figure S5A; Table S1). Many of the upregulated pathways following 4HPR treatment centered on cell cycle progression, which is consistent with the decreased S phase progression in HSPC subsets we observed following 4HPR treatment (Table S1; Figure S2C). Lipid metabolism pathways were significantly altered with an upregulation of Cer/SpL biosynthesis and a decrease in cholesterol/sterol biosynthetic pathways (Figure 5A), possibly a feedback response due to SpL homeostasis disruption by 4HPR. Interestingly, a strong cellular stress induction theme was highlighted in the pathway analysis (FDR < 0.05; Figure S5A) following 4HPR treatment, including ER stress/UPR/ATF4, protein folding, ROS, and autophagy. These pathways were grouped into 6 modules, and GSEA showed a significant enrichment of these pathways for 4HPR treatment relative to control (Figures 5A and 5B; shaded gray on Figure S5A; gene lists in Table S2).

11We asked whether some of these stress response programs following 4HPR-induced lipostatic stress are potentially intrinsic HSC maintenance mechanisms necessary for cellular activation. When we analyzed ER stress/UPR, protein folding, ROS, and autophagy in uncultured LT-HSCs compared to uncultured GMPs by GSEA, only autophagy was not significantly enriched, suggesting some stress response programs are more activated in progenitors than HSCs at homeostasis (Figure S5B). To distinguish cell cycle priming as opposed to lineage priming differences between LT-HSCs and GMPs, we compared ST-HSCs to LT-HSCs at homeostasis, and we found a significant enrichment of protein folding (p = 0.001) and near significant enrichment of ER stress/UPR (p = 0.079) in LT-HSCs, suggesting these programs may be required for quiescence exit and/or self-renewal (Figure 5C). Additionally, we analyzed published single-cell RNA-seq data of HSCs from bone marrow. Within CD34+CD38−CD45RA− cells, we identified two clusters, cell-cycle-primed and non-primed, using a clustering algorithm developed to robustly find distinguishing gene expression features in stem cells (; see STAR Methods). We specifically examined LT-HSCs, which were represented in both clusters, and found that cell-cycle-primed LT-HSCs have higher CD38 surface expression, cell cycle programs, and CDK6 and DEGS1 expression than non-primed LT-HSCs (Figures S5C, S5D, 5D, and 5E). The latter had enrichment for a set of dormancy-associated genes found in label-retaining dormant mouse LT-HSCs (Figure S5E; ). We found ER stress/UPR and protein folding enriched in cell-cycle-primed LT-HSCs, suggesting the activation of ER proteostasis programs as well as upregulation of DEGS1 gene expression is an early event in HSC activation.

12Next, we asked how the gene expression changes enacted by 4HPR treatment compared to combination treatment with U+S during culture and changed over time. Cells subjected to 3-Factor treatment at day 2 showed negative enrichment for cholesterol/sterol biosynthesis pathways and at day 4 also showed enrichment of ER stress and UPR pathways as observed with 4HPR alone (Table S1). Gene expression changes for the top 4 genes in the selected pathways are similar following 4HPR treatment and 3-Factor treatment, but 3-Factor-treated samples show fewer significantly altered gene sets than with 4HPR treatment alone at both days 2 and 4 (Figures 5F, S5F, and S5G). These data point to activation of autophagy and ROS gene sets as resulting exclusively from 4HPR treatment. Although cytokine treatment induces cell cycle entry but no cell division in LT-HSCs at day 2 (progenitors have divided), most LT-HSCs have completed one cellular division only by day 4 (Figure S2D; ). Nonetheless, gene expression changes in ER stress/UPR and metabolic gene sets persist from day 2 to day 4 (Table S1; Figures S5F–S5H), suggesting 4HPR treatment is provoking a sustained HSC maintenance response in ex vivo culture independently of cell cycle transit following quiescence exit.

13We had previously identified a cytoprotective ATF4-ISR program downstream of the UPR in response to metabolic stress in human HSCs. Because we retrieved enrichment in the ER stress/UPR functional model for an “ATF4 activated genes” gene set by 4HPR treatment, we asked whether HSC subsets following DEGS1 KD also showed evidence for activation of this UPR/ATF4 program. We quantified phosphorylated eukaryotic translation initiation factor 2α (eIF2α) in LT-HSCs and ST-HSCs from 4-week xenografts engrafted with shCtrl- or shDEGS1-transduced HSPCs by confocal microscopy (Figures 5G and 5H; 3 CB from experiment 2 of Figures S1J–S1M). Upon UPR activation by ER stress, eIF2α is phosphorylated to attenuate eIF2B, resulting in global translation inhibition except for the specific transcriptional programs, e.g., ATF4, licensed by the UPR. As LT-HSCs are activated, with 60% in cycle at 4 weeks transplant, the presence of anti-pEIF2S1/phospho-eIF2α staining in shCtrl LT-HSC and ST-HSC reflects activation of an UPR/ISR program (Figures 5G, 5H, S2F, and S5I;, ). Importantly, shDEGS1 LT-HSCs have significantly increased pEIF2S1 staining over shCtrl, suggesting LT-HSCs are more sensitive to lipostatic stress than ST-HSCs in vivo (Figures 5G, 5H, S5J, and S5K). Thus, these data suggest SpL modulation via pharmacological inhibition ex vivo and via lentiviral KD in vivo are activating an UPR/ATF4 program during cellular activation to restore homeostasis.

14Metabolic dysregulation through increase in oxidative stress and mitochondrial dysfunction are known activation signals for ATF4-ISR programs. Pathways related to redox homeostasis, mitochondria gene sets, and Pink-Parkin mediate mitophagy, suggesting 4HPR treatment has activated the UPR/ATF4 to enact metabolic remodeling to restore homeostasis (Table S1). Therefore, we analyzed reactive oxygen species (ROS) and mitochondrial membrane potential following 2 days of 4HPR treatment in the progeny of CB stem and progenitor cells and found a decrease for both parameters following 4HPR treatment as compared to controls (Figures 5I, 5J, S5L, and S5M). Such metabolic reprogramming by 4HPR persists to day 3 when the progeny of HSPCs, including LT-HSCs treated with 4HPR, have exited quiescence and are actively cycling (Figures S5L, S5M, and S2D). In summary, as DEGS1 functions at the ER to convert dhCer to Cer, which is then further processed at the Golgi apparatus to form complex SpLs, these data are consistent with 4HPR inhibition of DEGS1 as a lipostatic stress stimulus capable of activating coordinated transcriptional stress programs to remodel the cellular metabolism of cultured HSPCs for stress recovery and ultimately preserve functional HSCs.

15As activating autophagy can promote HSC function, we asked whether 4HPR treatment leads to autophagy activation for HSC maintenance during ex vivo culture. Gene expression analysis showed SpL modulation by 4HPR upregulated autophagy gene sets in lin− CB. However, autophagy has not previously been characterized in human LT-HSCs in quiescence or upon cellular activation. Thus, we examined basal autophagy in CB LT-HSCs and GMPs with the autophagosome marker LC3II by confocal microscopy and found similar LC3II intensity consistent with murine data (Figures 6A, S6A, and S6B; ). To ascertain whether in vivo cellular activation disrupts basal autophagy levels in human HSCs, we compared LC3II staining in LT-HSCs isolated from granulocyte-colony-stimulating factor (G-CSF) mobilized peripheral blood (mPB) (Figures 6B and S6C) and found a significant decrease in LC3II intensity in mPB LT-HSCs compared to CB LT-HSCs. Next, we asked whether 4HPR treatment specifically activated autophagy in human CB stem cells and not progenitors, because only murine stem cells appear capable of activating autophagic flux in response to metabolic stress, with two independent assays: (1) LC3II staining with and without bafilomycin A1 (BAF), an agent that inhibits autophagosome turnover by blocking lysosome acidification (Figures 6C–6F, S6D, and S6E), and (2) flow cytometry with Cyto-ID (Figures 6G, 6H, S6G, and S6H). LC3II staining showed both control and 4HPR-treated stem and progenitor cells at day 2 exhibit basal autophagy in culture (Figures 6D, 6E, and S6D). However, only stem, but not progenitor, cells showed autophagy activation by an increase in LC3II foci area in the presence of 4HPR and BAF (Figures 6D–6F). Similarly, Cyto-ID staining with control or 4HPR-treated cells (0.2 μM or 2 μM) showed that 4HPR significantly amplified autophagic flux only in stem, but not in progenitor, cells (Figure 6G). We confirmed cytokine withdrawal activates autophagy (as measured by Cyto-ID) in human stem, but not progenitor, cells like in mouse (Figures 6H, S6F, and S6G), which is further enhanced by 4HPR (Figure S6H). Collectively, these data indicate 4HPR activated autophagy exclusively in stem cells during ex vivo culture.

16To mechanistically determine whether HSPCs require both autophagy and the UPR/ISR for survival in response to SpL modulation with 4HPR in culture, we pharmacologically inhibited autophagy with BAF alone or in combination with integrated stress response inhibitor (ISRIB) in lin− CB. ISRIB is a potent ISR inhibitor by maintaining activity of eIF2B and retaining protein translation despite EIF2α phosphorylation. At day 8, the percentage of total number of cells relative to control, CD34+ cells, CD14+ myeloid cells, and GlyA+ erythroid cells were quantified by flow cytometry analysis and compared to vehicle (Figures 7A–7D). ISRIB appeared to have no significant effect on CB lin− cells treated with or without 4HPR by the parameters we measured but significantly decreased proliferation when autophagy was also inhibited. Although BAF treatment was equally potent in restraining proliferation and differentiation particularly to the erythroid lineage in both control and 4HPR-treated CB cells, as previously shown in embryonic stem cells and K562 cells, the addition of ISRIB with BAF significantly limited the number of viable cells and erythroid cells with 4HPR treatment relative to BAF+ISRIB alone and trended to significance in cells treated with 4HPR+BAF relative to 4HPR+BAF+ISRIB (Figures 7A and 7D). However, BAF treatment in lin− CB cells was context dependent as addition of BAF at day 2, perhaps when culture-induced stress was lower, had minimal effect on cell growth except for erythroid differentiation (Figures 7E and S7B–S7F; ). Finally, lin− CB cells were treated with the same combination of drugs as in Figure 7A, beginning at quiescence for 20 h, and then Cyto-ID was used to assess whether autophagic flux activation requires coordinate activation of the UPR/ISR (Figure 7F). As these cells are a mix of stem and progenitor cells, no difference between control and 4HPR-treated cells was observed. However, upon BAF addition to accumulate autolysosomes, a loss of flux demonstrated by an increase in Cyto-ID mean fluorescence intensity (MFI) was observed in 4HPR+BAF cells relative to 4HPR cells alone or control + BAF-treated cells (Figure 7F). At this time point, BAF treatment successfully blocked lysosome acidification as measured by flow cytometry with LysoTracker (Figure S7G). Importantly, the accumulation of autolysosomes seen between cells treated with 4HPR+ISRIB and cells treated with 4HPR+ISRIB+BAF was abolished, suggesting some aspect of the ISR may be required to activate autophagic flux with 4HPR treatment (Figure 7F). In summary, these data, along with the serial transplantation assays, support a model (Figure 7G) where SpL modulation by 4HPR in ex vivo culture activates a coordinated stress response, including autophagy and ER stress programs, to restore cellular homeostasis and maintain stemness in 4HPR-treated cells during the transition from quiescence to cellular activation.

17Here, we provide direct evidence that sphingolipid composition regulates HSC self-renewal and lineage commitment. Although short-term culture normally results in marked CD34+ cell expansion at the expense of LT-HSCs, our data showed that modulation of SpL metabolism during culture activates cellular stress proteostasis programs that collectively promote the proper metabolic transition from quiescence to cellular activation, resulting in LT-HSC maintenance. Four findings emerge from our work: (1) the sphingolipidome has distinct composition within the various mature, progenitor, and HSC subsets that comprise the human hematopoietic hierarchy; (2) KD or pharmacological dysregulation of DEGS1 is sufficient to alter the lineage determination of HSPCs; (3) dysregulation of Cer and dhCer homeostasis serves as a signal to activate cellular stress proteostasis programs, including the UPR and autophagy in human HSPCs; and (4) transient SpL modulation by 4HPR during ex vivo expansion culture results in the maintenance of functionally defined LT-HSCs that possess serial repopulating ability following xenotransplantation.

18Membranes are increasingly being recognized as playing central roles in cellular homeostasis, and Cer pools have a special role in cellular lipid metabolism. Here, we establish SpL homeostasis as another aspect of lipid metabolism beyond fatty acid oxidation that is pivotal to physiological human hematopoiesis. Cer and the complex SpLs are essential components of multiple cellular membranes, including the ER, Golgi, mitochondria, lysosomes, autophagosomes, and at the plasma membrane. Cer is generated either via de novo synthesis or from the salvage pathway through catabolic membrane recycling. We found that LT-HSCs and ST-HSCs favor increased expression of de novo synthesis pathway genes compared to more activated progenitors. We undertook sphingolipidome mapping of the human hematopoietic hierarchy and found distinct composition of sphingolipid species between five mature blood lineages and HSPCs, particularly around the interface of cer to dhCer homeostasis. By inhibiting the de novo SpL metabolic enzyme DEGS1 that adds a double bond to dhCer to form Cer, we modulated these de novo Cer pools and found that cellular stress programs were initiated. Our findings point to SpL remodeling as an early event in cellular activation of quiescent HSPCs.

19Studies from the obesity and aging fields have identified mild induction of ER stress/UPR and autophagy as improving healthspan and lifespan, possibly through effects on proteostasis and organelle homeostasis within stem cell populations. Recently, we reported that ATF4 activates the ISR upon amino acid deprivation to increase HSC survival. Moreover, physiological aging of HSCs appears to result in part through suppression of autophagy programs that maintain metabolic homeostasis. Although a number of ATF4 target genes are known to be involved in autophagy regulation, the linkage between the UPR, autophagy, and ATF4/ISR remains unclear. Our present study points to SpL modulation by 4HPR as a single stimulus that concurrently activates each of these three arms to restore cellular homeostasis following metabolic perturbations, such as those occurring during culture. Together, these function as a set of coordinated quality control responses, potentially via a cytoprotective mechanism that promotes the persistence of a more regenerative HSC pool following ex vivo cytokine activation from quiescence. We speculate that alterations of Cer/DhCer are part of a “lipid biostat” for stress that is in part enacted by the very distinct biophysical properties of Cer and dhCer. As alluded to in yeast and mouse cells, altering lipid homeostasis may have major consequences on membrane structure, and this alteration has been proposed to activate the UPR. DEGS1 KD was sufficient to activate the ISR arm of the UPR in HSC subpopulations in vivo (Figures 5G and 5H). Perhaps UPR sensing of various proteostatic and lipostatic stress is blunted during aging and together with erosion of autophagy flux in aged HSCs responsible for some aging phenotypes. Would SpL modulation be proficient to activate these cellular stress protection programs and rejuvenate HSCs during aging, as suggested for rapamycin?

20Our data indicate a need to evaluate potential adverse effects of obesity, which is linked to an increase in Cer levels, on SpL homeostasis in human hematopoiesis. Obesity is also linked to disruption of the murine bone marrow hematopoietic niche and altering HSC function, which persists in cells transplanted from obese mice into a normal environment. We wonder whether obesity reprograms HSC stress responses via SpL lipostatic stress. Moreover, SpL dysregulation is associated with incomplete autophagy in a number of human neuropathies resulting from germline mutations in SpL genes, which manifest as lysosomal lipid storage disorders. Interestingly, SpL gene expression was higher in neural stem cells than activated progeny, and the lysosome, which serves as a major site for SpL degradation, was proposed to be critical for quality control of neural stem cells during aging. Determining the coordination of SpL metabolism, autophagy regulation, and lysosomal biogenesis may prove informative for understanding how adult stem cells maintain life-long self-renewal. These data predict that SpL metabolism regulates adult stem function in both blood and brain and perturbation of lipostatic stress sensing in the lysosome or other organelles may contribute to human disease and aging.

21The study of HSC stress responses has implications beyond understanding stemness control. The ability to control the cellular properties of human HSCs has important clinical implications, particularly for HSCT. Although 50,000 HSCT procedures occur annually, two-thirds of patients who need HSCT lack matched donor tissue. Limitations of CB as a source for HSCT include the delay in neutrophil recovery following transplantation and long-term sustainability of the donor graft. We found that DEGS1 inhibition regulates the erythroid-myeloid axis in vitro, where there was no alteration of S1P levels, suggesting such lineage commitment changes resulted from dysregulation of SpL composition, not S1P signaling. Although we focused on a cell-autonomous aspect of this biosynthetic pathway, our results are also consistent with recent work in the field on the role of S1P signaling axes on lymphoid and erythroid/platelet lineage commitment. We observed a decrease in S1P levels in shDEGS1 cells isolated from the xenograft where there was also disruption of B lymphoid engraftment. Intrinsic genetic perturbation of S1P signaling is sufficient to dysregulate lymphoid commitment and the HSPC hierarchy in mice. However, the erythroid-to-myeloid skewing observed with 4HPR treatment could also result from an ER stress-mediated emergency granulopoiesis-like program; this lineage bias may aid in faster neutrophil recovery in patients following transplant of an expanded CB. Another major limitation in wider clinical CB HSCT is the small number of HSCs present in a single unit and the inability of many cytokine cocktails to expand LT-HSCs. Our studies highlight the metabolic and proteostatic stress perturbations that arise upon cytokine activation that collectively impair LT-HSC function at the expense of generation and expansion of downstream progeny. SpL modulation by 4HPR during ex vivo culture activates cytoprotective proteostasis programs that result in preservation of significantly higher numbers of functional LT-HSCs in the immunophenotypic HSC pool. Notably, we show here that the search for small molecules to increase LT-HSC donor graft sustainability from CB can be achieved with molecules that do not necessarily expand CD34+ cells. Thus, ex vivo CB expansion regimens, including 4HPR, to activate a lipostatic stress program together with UM171 and SR1 may provide the best compromise between generating large numbers of ST-HSCs and progenitors for rapid engraftment while still sparing functional LT-HSCs to ensure long-term hematopoiesis.

22Further information and requests for resources and unique/stable reagents generated in this study should be directed to and are available without restriction from the Lead Contact, John E. Dick (jdick@uhnresearch.ca).

23Human cord blood samples were obtained with informed consent from Trillium and Credit Valley Hospital according to procedures approved by the University Health Network (UHN) Research Ethics Board. Mononuclear cells were obtained by centrifugation on Lymphoprep medium (Stem Cell Technologies) and were depleted of Lin+ cells (lineage depletion) by negative selection with the StemSep Human Progenitor Cell Enrichment Kit according to the manufacturer’s protocol (Stem Cell Technologies). Lin- CB cells were stored viably at −80°C or −150°C. Human mobilized peripheral blood samples (mPB) were obtained with informed consent from Princess Margaret Cancer Centre according to procedures approved by the UHN Research Ethics Board. Following mononuclear cell isolation, CD34+ cells from CB and mPB were enriched by positive selection with the CD34 Microbead kit (Miltenyi) and LS column purification with MACS magnet technology (Miltenyi).

24Lin– cells were thawed by dropwise addition of X-VIVO + 50% fetal calf serum supplemented with DNase (100 μg/mL final concentration, Roche) and resuspended at a density of 5 × 106 cells/mL. Cells were then stained with the following antibodies (all from BD, unless stated otherwise): FITC–anti-CD45RA (1:50, 555488), PE–anti-CD90 (1:50, 555596), PECy5–anti-CD49f (1:50, 551129), V450–anti-CD7 (1:33.3, 642916), PECy7–anti-CD38 (1:200, 335790), APC–anti-CD10 (1:50, 340923), APCCy7–anti-CD34 (1:200, custom made by BD). Cells were sorted on FACS Aria III (Becton Dickinson), consistently yielding > 95% purity. LT-HSC were sorted based on the following markers: CD34+CD38-CD45RA-CD90+CD49f+. ST-HSC were sorted as CD34+CD38-CD45RA-CD90-CD49f- and GMP as CD34+CD38+CD10-CD7-CD45RA+. shCtrl or shDEGS1 previously frozen 4 week xenograft samples were thawed as described above. Mouse cells were depleted with the mouse depletion kit (Miltenyi) and LS column purification with MACS magnet technology (Miltenyi). Cells were then stained with the following primary antibodies: FITC–anti-CD45RA PE–anti-CD90, PECy5–anti-CD49f, BV711-CD19 (1:50, 563036), Alexafluor700–anti-CD7 (1:50, 561603), either APC–anti-CD10 or Alexafluor700–anti-CD10 (1:50, 563500) BV510-CD45 (1:50, 563891), biotin-Flt3L (1:50, 624008), and APCCy7–anti-CD34. Cells were washed and then stained with streptavidin-Qdot605 (1:100, Q10101MP). LT-HSC were sorted based on the following markers: CD45+BFP+C19-CD34+CD38-CD45RA-CD90+CD49f+. ST-HSC were sorted as CD45+BFP+CD19-CD34+CD38-CD45RA-CD90-CD49f- and GMP as CD45+BFP+CD19-CD34+CD38+CD10-CD7- CD45RA+.

25To profile the SpL composition of primitive CB cells, 1.05-1.1 million CD34+CD38- and 1.9-2.9 million CD34+CD38+ cells were isolated from 3 pools of previously frozen lin- CB (22 million – 50 million cells) by flow cytometry, washed with PBS, and frozen as cell pellets. For the SpL distribution of mature CB lineages, freshly isolated CB bags from five individuals were processed for mononuclear cells and a small fraction stained with V450-anti-CD15, APC-cy7-anti-CD34, PE-Cy7-anti-GlyA, PECy5–anti-CD14, FITC-anti-CD3, and V500-anti-CD45 to sort for erythrocytes (GlyA+CD45-) prior to ammonium chloride lysis. After lysis, cells were stained at 10 million/ml and T cells (CD45+CD3+CD19-CD14-CD15-), B Cells (CD45+CD3-CD19+CD14-CD15-), monocytes (CD45+CD3-CD19-CD14+CD15-), and neutrophils (CD45+CD3-CD19-CD14-CD15+) were isolated by flow cytometry on the Aria Fusion or Aria RITT. Cell numbers ranged from 1-4 million cells. Subsequent lipid extraction and mass spectrometry for sphingomyelin species, hexosylceramide species, ceramide species, dihydroceramide species and sphingoid species were performed by the Lipidomics Facility of Stony Brook University Medical Center. As the morphology between the subpopulations are quite distinct, particularly between the mature lineages, normalization of the sphingolipidome data was an important consideration. Normalization to cellular inorganic phosphate (Pi) was chosen to minimize the potential confounding effects of differences in cellular size and protein content between the profiled populations. Two pools CD34- lin- CB cultured with DMSO control or 4HPR for 8 days were collected and 4 million expanded cells were washed twice in PBS and frozen as cell pellets. CD45+BFP+ cells lacking the CD34+CD19- population were sorted from shCtrl (1.8-2.7 million cells) or shDEGS1 (0.4-0.5 million cells) previously frozen 4 week xenograft samples (Primary transplant data in Figures S1I–S1O, Experiment 1, n = 2 biological replicates), washed and frozen as cell pellets. Subsequent lipid extraction and mass spectrometry for ceramide species, dihydroceramide species and sphingoid species were performed by the Lipidomics Facility of Stony Brook University Medical Center. Inorganic phosphate levels were measured for lipid normalization across samples.

26shRNA sequences were predicted using the Sherwood algorithm and ordered as Ultramer DNA oligos (IDT). Subsequently, shRNAs were amplified using AmpliTaq Gold 360 Polymerase (ThermoFisher, 4398813) using FW primer: 5′- GGATCCTGTTTGAATGAGGCTTCAGTACTTTACAGAATCGTTGCCTGCACATCTTGGAAACACTTGCTGGGATTACTTCT-3′ and RV primer: 5′-AGTAACGCGTAAAGTGATTTAATTTATACCATTTTAATTCAGCTTTGTAAAAATGTATCAAAGAGATAGCAAGGTATTCAGTTTTAGTAAACAAGATAATTGCTCCTAAAGTAGCCCCTTGAAGTCCGAGGCAGTAGGCA-3′. The PCR product was digested with BamH1-HF and Mlu1-HF (NEB) and subcloned into the pLBC2 lentiviral vector, downstream of SFFV-tBFP. Viral production, titration and transduction of CD34+CD38- CB cells were done as previously described.

275′-TGCTGTTGACAGTGAGCGAGGTCATGAAACTTACTCATAATAGTGAAGCCACAGATGTATTATGAGTAAGTTTCATGACCCTGCCTACTGCCTCGGA-3′

285′-TGCTGTTGACAGTGAGCGAGGTCATGAAACTTACTCATAATAGTGAAGCCACAGATGTATTATGAGTAAGTTTCATGACCCTGCCTACTGCCTCGGA-3′

295′-TGCTGTTGACAGTGAGCGCAGGAATTATAATGCTTATCTATAGTGAAGCCACAGATGTATAGATAAGCATTATAATTCCTATGCCTACTGCCTCGGA-3′

305′-TGCTGTTGACAGTGAGCGCAGGAATTATAATGCTTATCTATAGTGAAGCCACAGATGTATAGATAAGCATTATAATTCCTATGCCTACTGCCTCGGA-3′

31In order to assess shRNA knock-down efficiency, MOLM13 cells were infected at a multiplicity of infection of 0.3. Transduced cells were sorted for BFP+ expression and total RNA was isolated and DNase treated using the RNeasy Micro Kit (QIAGEN, 74004). RNA quality (RIN > 9) was verified using the Bioanalyzer RNA 6000 Pico Kit (Agilent) and cDNA was synthesized using SuperScript VILO (ThermoFisher, 11754050). qPCR was performed on the Roche Lightcycler 480 using Power SYBR Green (ThermoFisher, 4367659). All signals were quantified using the ΔCt method and were normalized to the levels of GAPDH.

325′-CAAACATTCCAAACCAGCGAT-3′5′-GCAGTTGCATTAACCACTCAA-3′

335′-CAAACATTCCAAACCAGCGAT-3′

345′-GCAGTTGCATTAACCACTCAA-3′

355′-ACATCGCTCAGACACCATG-3′5′-TGTAGTTGAGGTCAATGAAGGG-3′

365′-ACATCGCTCAGACACCATG-3′

375′-TGTAGTTGAGGTCAATGAAGGG-3′

38All animal experiments were done in accordance to institutional guidelines approved by the University Health Network Animal care committee. Aged match female or male NSG mice (NOD.Cg PrkdcscidIl2rgtm1Wjl /SzJ; Jackson Laboratory) 10-12 weeks of age were sublethally irradiated with 250 rads 1 day before intrafemoral injection. Following 4 weeks xenotransplantion of lentiviral DEGS1 knockdown into male NSG, mice were euthanized and the injected femur and other bones were flushed separately in Iscove’s modified Dulbecco’s medium (IMDM) and human chimerism and transduced cells marked by BFP expression were assessed by flow cytometry on the BD Celesta and the following antibodies: PE–anti-CD19 (349209), PE–anti-GlyA (Beckman Coulter, A07792), V500–anti-CD45 (560777), APCCy7–anti-CD34, and BV786–anti-CD33 (740974). For primary transplant analysis of ex vivo cultured cells, female mice were euthanized at 16 weeks after transplantation. The injected femur and other bones were flushed separately in Iscove’s modified Dulbecco’s medium (IMDM) and human chimerism was assessed with the following antibodies: PE–anti-CD19 (349209), PE–anti-GlyA (Beckman Coulter, A07792), PECy5–anti-CD45 (Beckman Coulter, A07785), PECy7–anti-CD14 (1:200; Beckman Coulter, A22331), APC–anti-CD33 (551378) and V450–anti-CD15 (642917). For secondary LDA assays, primary transplant mice from each treatment group were individually thawed, and 1/3 of each mouse pooled and stained with APC–anti-CD45 and FITC–anti-CD45. Human CD45+ cells were sorted from each pooled sample and injected at indicated doses into irradiated female NSG mice and engraftment assessed at 16 weeks post-transplant. For LDA experiments, injected femur and non-injected femurs were isolated and flushed separately and analyzed with the following: APC–anti-CD45, FITC–anti-CD45, PE-anti-CD19 and APC–anti-CD33. A mouse was considered engrafted if CD45+ > 0.1 and multilineage. LTRC frequency was estimated using the ELDA software (http://bioinf.wehi.edu.au/software/elda/; ). The remaining 2/3 of bone marrow from each individual mouse was first enriched for CD34+ cells by positive selection with the CD34 Microbead kit (Miltenyi) and LS column purification with MACS magnet technology (Miltenyi). CD34+ enriched cells were then directly analyzed by flow cytometry with the following antibodies: FITC–anti-CD45RA, APC–anti-CD90, PECy5–anti-CD49f, V450–anti-CD7, PECy7–anti-CD38, V421–anti-CD10, APCCy7–anti-CD34, PE-anti-CD19, V500-anti-CD45.

39LT-HSC, ST-HSC, or GMP were sorted directly into methylcellulose (cat. No H4034, Stem Cell Technologies), supplemented with FLT3 Ligand (20 ng/ml) and IL6 (50 ng/ml). Fenretinide/4HPR (Tocris Biosciences, cat. # 1396), StemRegenin1 (Stem Cell Technologies, cat. # 72344), and UM171 (Xcessbio, cat. #60223-2), or DMSO vehicle were added following sorting at the following concentrations, DMSO (control), 2 μM 4HPR, 35nM UM171 or 500 nM SR1, such that DMSO was always < 0.1% and equal between treatment and control groups. ATRA was a kind gift from G. Keller. Samples were mixed and plated onto 35 mm dishes in duplicates. Colonies were allowed to differentiate for 10-11 days and morphologically assessed for colonies in a blind fashion by a second investigator. At day 14, colonies from replicate plates were pooled and resuspended in PBS/FBS and stained with FITC–anti-CD45RA, APC–anti-CD90, PECy5–anti-CD14, APCCy7–anti-CD34, PE-anti-CD235a (GlyA) for flow cytometry analysis on a BD Canto or Celesta. For shDEGS1 experiments, sorted LT-HSC, ST-HSC, or GMP were cultured in low cytokine media for 4 hours and then transduced with shCtrl or shDEGS1 lentivirus. At day 3 post-transduction, BFP+ cells were sorted directly into methylcellulose as above for colony scoring 10 days later (day 13 post-transduction). At day 16 post-transduction, colonies were analyzed on the Celesta for BFP and with APC–anti-CD90, PECy5–anti-CD14, APCCy7–anti-CD34, PE-anti-GlyA, and BV786–anti-CD33.

40Lin- CB were thawed via dropwise addition of X-Vivo based thawing media (X-Vivo, 50% FBS+1% DNase), resuspended in cytokine media, viable cells counted, and placed into appropriate cell concentrations for liquid culture: StemPro (Stem Cell Technologies) supplemented with StemPro nutrients (Stem Cell Technologies), L-glutamine (GIBCO), Pen/Strep (GIBCO), human LDL (Stem Cell Technologies, 50 ng/mL) and the following cytokines (all from Miltenyi): SCF (100 ng/mL), Flt3L (20 ng/mL), TPO (100 ng/mL), IL-6 (50 ng/mL), IL-3 (10 ng/mL), GM-CSF (20 ng/mL), except EPO (3 units/mL, from Jansen). The experimental scheme for ex vivo culture of viable lin- CB cells post thawing is the following: 3 initial cell doses in a LDA fashion with 62,500 cells (high), 12,500 cells (medium), and 2,500 cells (limiting) are cultured per well in a 24 well plate with 0.5 mls expansion media plus vehicle control or drug(s) on day 0. We mimicked fed-batch growth conditions where drug was added with fresh media every second day for 8 days to reduce auto-inhibitory signaling and ensure nutrients and drugs are not limiting during the culture period: fresh media plus drugs are added on day 2 (0.5 mls), day 4 (1 ml), and feeding and transferring to a 6 well plate on day 6 (2 mls). On day 8, all progeny are collected and 10/11th are unbiasly transplanted via intrafemoral injection into 5 NSG mice/condition. 1 LTRC cell is roughly equivalent to 700 lin- CB cells as previously calculated in xenotransplantation. This translates to the equivalent of the following day 0 cell doses: ∼11,300 lin- cells or ∼16 initial LTRC/mouse (high), ∼2270 lin- cells or∼3.2 LTRC/mouse (medium) and 454 lin- cells or ∼0.65 LTRC/mouse (limiting). The remaining 1/11th of day 8 progeny are analyzed by flow cytometry on a BD canto with a plate reader to enumerate cell numbers and lineage distribution with the following: V450-anti-CD15, APC-cy7-anti-CD34, PE-Cy7-anti-GlyA, PECy5–anti-CD14, or FITC-anti-CD15, PE-Cy5-anti-CD34, PE-anti-GlyA, PECy7–anti-CD14. LT-HSC, ST-HSC and GMP were sorted and ∼500 cells/well were cultured in 100 μL of cytokine media with DMSO on a 96 well plate on day 0. On day 4, an additional 100 μL of ex vivo media with compounds were added. Cells were stained with APCCy7–anti-CD34 on day 8 on a BD Canto with plate reader. For inhibition of autophagy and the ISR, 20nM Bafilomycin A1 (Sigma, B1793) and 500 nM ISRIB (Selleckchem, S7400), dissolved in DMSO, were added to 20,000 lin- CB cultured in cytokine media as indicated and analyzed for viable cells and lineage as previously and with CytoID (Enzo Life Sciences) or LysoTracker blue (ThermoFisher Scientific, L7525).

41For assessment of proliferation using BrdU incorporation assays, indicated subpopulations were cultured for three days with DMSO vehicle or 2 μM 4HPR in expansion media when BrdU was added to cells for 4 to 8 hours. BrdU staining was performed with the APC BrdU Flow Kit (BD PharMingen) with APC-anti-BrdU according to the manufacturer’s protocol. Sorted LT-HSC, ST-HSC or GMP were cultured in ex vivo culture media with vehicle or 4HPR as indicated and fixed and stained for Ki67-Hoechst assays as described. LT-HSC and GMP were sorted from shCtrl or shDEGS1 previously frozen 4 week xenograft samples (Primary transplant data in Figures S1I–S1O, Experiment 1, n = 2 CB) and fixed and stained for Ki67-Hoechst analysis. Samples were analyzed on a BD LSRII cytometer with an UV laser.

421x104 sorted CD34+CD38- or CD34+CD38+ cells were cultured in ex vivo culture media on a 96 well suspension plate with DMSO or 2 μM 4HPR. Bafilomycin A1 was added at 30 hours post-culture. Cells were collected at 48 hours post-culture and fixed with 4% paraformaldehyde. Cells were stain with a mouse monoclonal antibody to LC3II (NanoTools, 5F10) and DAPI. Slides were visualized on a Zeiss LS700 confocal microscope and images collected. Quantitation of LC3II foci area as a percentage of total cell area was done using R and ImageJ software (n = 3 CB). A range of 50-150 cells were quantified per condition. LT-HSC and GMP were purified sorted from CD34+ CB or mPB samples, fixed with 4% paraformaldehyde and stained with LC3 and DAPI as above. Following image collection on a Zeiss LS700 confocal microscope, LC3II integrated density (total signal intensity) was quantified using R and ImageJ software (n = 3). LT-HSC and ST-HSC were sorted from shCtrl or shDEGS1 previously frozen 4 week xenografts samples (n = 3, Experiment 2, Figures S1I–S1O), fixed with 4% paraformaldehyde and stained with a rabbit monoclonal antibody to EIF2S1 (phospho S51, Abcam, ab196460) and DAPI and images collected as above.

431x104 sorted CD34+CD38- or CD34+CD38+ cells were cultured in ex vivo media on a 96 well suspension plate with DMSO, 0.2 μM or 2 μM 4HPR for either 2 days or 3 days. Autophagy flux was measured using CytoID analysis kit (Enzo Life Sciences). Cytokine withdrawal experiments to induce autophagy were done by replacing media containing only drugs without cytokines for 6 hours on cells and then analyzing with CytoID. Staining for ROS and mitochondrial membrane potential in active mitochondria was performed by incubating cells at 37°C with 5 μM CellROX deep red (C10422) and 1 μM TMRE (T668), following the manufacturer’s protocols (ThermoFisher) and directly analyzed on a BD celesta.

44Three pools of lin- CB at the 16 LTRC dose (62,500 cells/well, 24 well plate) were cultured with ex vivo culture media and the following 4 treatment conditions: DMSO (control), 2 μM 4HPR, 35nM UM171 + 500 nM SR1 (U+S) or combination of 4HPR+UM171+SR1 (3-Factor). Cultured cells at day 2 and day 4 were collected, pelleted, washed twice with PBS and resuspended in RNeasy microRNA plus kit (QIAGEN) and frozen for subsequent RNA isolation. Sufficient cells were cultured to isolate a minimum of 500 ng RNA. Nextera libraries were generated without amplification and subjected to 125 bp, paired-end RNA-sequencing on the Illumina HiSeq 2500 with an average of ∼57 million reads/sample at the Center for Applied Genomics, Sick Kids Hospital. Reads were trimmed to remove the adapters and were then aligned to hg19 using TopHat (v2.1.1) and gene counts were generated using HTSeq (v 0.6.1). Read counts were retrieved for each sample and processed using edgeR to estimate differential expression between the treated and ctrl samples and between 3-Factor and U+S and 4HPR. Genes with count per million equal or less than 0.25 in at least one fifth of the samples were removed from further analysis. 15085 genes remained in the analysis. edgeR dispersion parameters were estimated for the whole dataset and a generalized model was applied to compare each treated-control pair. Cord blood batch was included in the model. Tests were corrected for multiple hypothesis testing using the Benjamini-Hochberg method. MDS plots, hierarchical clustering and heatmaps were generated using logarithm of base2 of count per million of TMM normalized counts. A score to rank genes from top upregulated to downregulated was calculated using the formula -sign(logFC) ∗ -log10(pvalue). The rank file from each comparison was used in pathway analysis (GSEA) (http://software.broadinstitute.org/gsea/index.jsp) using 2000 permutations and default parameters against a pathway database containing Msigdb c2 and c3, NCI, IOB, NetPath, HumanCyc, GO BP and Panther (http://baderlab.org/GeneSets, version April 2017). NES and FDR results are located in Table S1. EnrichmentMap version 2.1.0 in Cytoscape 3.4.0 was used to visualize enriched pathway gene sets at FDR ≤ 0.05 with a Jaccard coefficient set to 0.25. For GSEA analysis in Figure 5B and Velten et al. scRNA-seq, lists of genes were returned by taking the union of all gene sets for the pathways in Figure 5A.

45Index-sorted single cell (sc) RNA-seq data of human HSPCs were obtained from GEO (GSE75478). Raw count data and FACS surface marker annotations were loaded into the Scanpy single cell analysis suite. FACS surface marker measurements were subject to logicle transformation through the flowutils package (https://github.com/whitews/FlowUtils), and used to distinguish CD34+CD38- plates from CD34+CD38+ plates for each individual. Among the CD34+CD38- plates, CD34+CD38-CD45RA- HSPCs were identified and their gene expression counts were subjected to pooling normalization using the scran package in R. Normalized CD34+CD38-CD45RA- HSPCs for each individual were subsequently clustered using the ‘Self Assembling Manifolds’ (SAM) algorithm, which iteratively re-scales gene expression to identify subpopulations with subtle, yet consistent, differences. SAM returned two distinct clusters for each individual, with one cluster demarcated by higher CDK6 expression, expression of cell cycle genes, and surface CD38 expression as measured through FACS. Cells in the second cluster expressed fewer distinct genes but displayed higher relative expression of genes associated with HSC dormancy from murine studies (, using genes with logFC > 2 and FDR < 0.01 in dHSC versus aHSC and MPP). We thus labeled the clusters ‘cell cycle-primed’ and ‘non-primed’, respectively. Among CD34+38−45RA- cells from both individuals, we confirmed that the cell cycle-primed cluster had significantly higher expression of DEGS1 and of genes associated with ER Stress/UPR and protein folding, but comparable expression for genes associated with ROS and autophagy, consistent with our results for LT-HSCs. Signature scores for gene expression programs were derived by comparing the relative expression of that gene set with a random set of genes in the transcriptome as implemented in Scanpy. LT-HSC were represented in both the cell cycle-primed cluster and the non-primed cluster. To examine LT-HSC specifically, we identified CD34+38−45RA- cells that were within the top 40th percentile for both CD49f and CD90 surface marker expression.

46GraphPad Prism was used for all statistical analyses except RNA-seq. Unless otherwise indicated, mean ± SD values are reported in the graphs. Statistical significance was determined with Student t tests. p < 0.05 (∗), p < 0.01 (∗∗), p < 0.001(∗∗∗), and p < 0.0001(∗∗∗∗).

47The accession number for the 4HPR RNA-seq data in this paper is GEO: GSE125214 and the accession number for uncultured LT-HSC, ST-HC, and GMP RNA-seq data is GEO: GSE125345. Raw data are available on EGA under accession number EGAS00001003756.