Restricted Hematopoietic Progenitors and Erythropoiesis Require SCF from Leptin Receptor+ Niche Cells in the Bone Marrow

SUMMARY

Hematopoietic stem cells (HSCs) are maintained in a perivascular niche in bone marrow, in which leptin receptor⁺ (LepR) stromal cells and endothelial cells synthesize factors required for HSC maintenance, including stem cell factor (SCF). An important question is why LepR⁺ cells are one hundred times more frequent than HSCs. Here, we show that SCF from LepR⁺ cells is also necessary to maintain many c-kit⁺-restricted hematopoietic progenitors. Conditional deletion of Scf from LepR⁺ cells depleted common myeloid progenitors (CMPs), common lymphoid progenitors (CLPs), granulocyte-macrophage progenitors (GMPs), megakaryocyte-erythrocyte progenitors (MEPs), pre-megakaryocyte-erythrocyte progenitors (PreMegEs), and colony-forming units-erythroid (CFU-Es), as well as myeloid and erythroid blood cells. This was not caused by HSC depletion, as many other restricted progenitors were unaffected. Moreover, Scf deletion from endothelial cells depleted HSCs, but not progenitors. Early erythroid progenitors were closely associated with perisinusoidal LepR⁺ cells. This reveals cellular specialization within the niche: SCF from LepR⁺ cells is broadly required by HSCs and restricted progenitors while SCF from endothelial cells is required mainly by HSCs.

Graphical Abstract

In Brief

Bone marrow hematopoietic stem cells reside in perivascular niches associated with sinusoids and require factors from leptin receptor⁺ stromal cells and endothelial cells. Morrison and colleagues show that c-kit⁺-restricted hematopoietic progenitors also require Stem Cell Factor made by LepR⁺ cells, but not endothelial cells. At least some of these restricted progenitors reside in perisinusoidal niches.

INTRODUCTION

In adult mammals, hematopoiesis occurs primarily in the bone marrow, where hematopoietic stem cells (HSCs) and restricted hematopoietic progenitors are maintained throughout life. HSCs are maintained in a perivascular niche, in which leptin receptor⁺ (LepR⁺) stromal cells and endothelial cells are necessary sources of factors for HSC maintenance, including stem cell factor (SCF), Cxcl12 (Ding et al., 2012; Ding and Morrison, 2013; Greenbaum et al., 2013; Oguro et al., 2013), and pleiotrophin (Himburg et al., 2018). Approximately 80% of dividing and non-dividing HSCs in bone marrow are adjacent to sinusoidal blood vessels (Kiel et al., 2005; Acar et al., 2015). The niche cells we identified based on LepR expression have also been identified by others based on expression of high levels of Cxcl12 (Sugiyama et al., 2006; Omatsu et al., 2010), low levels of the Nestin-GFP transgene (Kunisaki et al., 2013), and Prx1-Cre (Greenbaum et al., 2013). Adipocytes are also a critical source of SCF for HSCs in fatty bone marrow and after myeloablation (Zhou et al., 2017).

LepR⁺ cells arise perinatally in the bone marrow (Zhou et al., 2014). They are very rare in the first few weeks after birth but then expand in number, ultimately representing around 0.3% of young adult bone marrow cells (Ding et al., 2012; Zhou et al., 2014). In contrast, HSCs represent only around 0.003% of bone marrow cells (Harrison and Zhong, 1992). Why are LepR⁺ cells 100 times more numerous than HSCs? One possibility is that LepR⁺ cells are also a niche cell for certain restricted hematopoietic progenitors, which are far more abundant than HSCs. Like HSCs, a number of restricted progenitors express the SCF receptor, c-kit, and require SCF for their maintenance, including common myeloid progenitors (CMPs) (Akashi et al., 2000), common lymphoid progenitors (CLPs) (Kondo et al., 1997), megakaryocyte-erythrocyte progenitors (MEPs) (Akashi et al., 2000), granulocyte-macrophage progenitors (GMPs) (Akashi et al., 2000; Pronk et al., 2007), pre-megakaryocyte-erythrocyte progenitors (PreMegEs) (Pronk et al., 2007), and erythrocyte progenitors (pre-colony forming unit-erythrocytes [PreCFU-Es] and CFU-Es; Pronk et al., 2007). Together, these restricted progenitors represent around 1.5% of mouse bone marrow cells. Because 95% of the cells that express high levels of Scf in normal young adult bone marrow are LepR⁺ (endothelial cells express much lower levels of Scf; Zhou et al., 2014), we wondered whether c-kit⁺-restricted progenitors might also depend upon SCF from LepR⁺ cells.

In contrast to the HSC niche, less is known about niches for restricted hematopoietic progenitors. A subset of early lymphoid progenitors resides near the endosteum, where they depend upon Cxcl12 synthesized by osteoblasts (Visnjic et al., 2004; Zhu et al., 2007; Ding and Morrison, 2013; Greenbaum et al., 2013). Other subsets of lymphoid progenitors depend upon perivascular niches (Tokoyoda et al., 2004; Sugiyama et al., 2006), including lymphoid progenitors that depend upon interleukin-7 (IL-7) from LepR⁺ cells (Cordeiro Gomes et al., 2016). During regeneration after myeloablation, GMPs form clusters close to megakaryocytes and megakaryocyte depletion prolongs the persistence of GMP clusters (Hérault et al., 2017). Erythroblastic islands near macrophages are sites of terminal erythroid differentiation, and clusters near sinusoids contain more mature erythroblasts (Mohandas and Prenant, 1978; Yokoyama et al., 2003).

SCF is required for the maintenance of HSCs and c-kit⁺-restricted progenitors as well as erythropoiesis, mast cell development, and lymphopoiesis (Russell, 1979; Huang et al., 1990; Zsebo et al., 1990; Flanagan et al., 1991; Barker, 1994; Rodewald et al., 1995; Waskow et al., 2002). Experiments in culture suggest that SCF acts by promoting survival and proliferation, as well as regulating the differentiation of various hematopoietic stem and progenitor cells, including erythroid lineage cells (Toksoz et al., 1992; Muta et al., 1995; Kondo et al., 1997; Wu et al., 1997; Domen and Weissman, 2000).

Here, we show that c-kit⁺-restricted hematopoietic progenitors, including erythroid lineage progenitors, require SCF synthesized by LepR⁺ cells, but not endothelial cells. This suggests that LepR⁺ cells are much more common than HSCs in the bone marrow because they create niches for many early restricted progenitors in addition to HSCs.

RESULTS

Restricted Progenitors Require SCF from LepR⁺ Cells

By qPCR analysis, LepR⁺ stromal cells had 172- ± 82-fold (mean ± SD) and 35- ± 17-fold higher levels of Scf mRNA as compared to unfractionated bone marrow cells and endothelial cells, respectively (Figure 1A). To test whether LepR⁺ cells or endothelial cells are a necessary source of SCF for restricted progenitor maintenance in the bone marrow, we conditionally deleted Scf using Lepr-cre or Tie2-cre. Although Tie2-cre also recombines in hematopoietic cells, bone marrow hematopoietic cells do not express detectable Scf and conditional deletion of Scf from hematopoietic cells has no effect on HSC frequency or hematopoiesis (Ding et al., 2012). The presence of a single null allele of Scf in Scf^Δ/FL mice reduced Scf transcript levels in unfractionated bone marrow cells to 52% ± 12% of the level in Scf^FL/FL control mice (Figure 1B). Conditional deletion of the second Scf allele in Lepr-cre;Scf^Δ/FL mice reduced Scf transcript levels in bone marrow cells to 12% ± 2% of Scf^FL/FL control mice (Figure 1B). Conditional deletion of Scf from endothelial cells in Tie2-cre;Scf^Δ/FL mice reduced Scf transcript levels in bone marrow cells to 44% ± 1% of Scf^FL/FL control mice (Figure 1B). Conditional deletion of Scf from both endothelial cells and LepR⁺ cells in Tie2-cre;Lepr-cre;Scf^Δ/FL mice reduced Scf transcript levels in bone marrow cells to 2.3% ± 0.9% of Scf^FL/FL control mice (Figure 1B). Transcripts that encode the soluble form and the membrane-bound form of SCF were both depleted in LepR⁺ cells from Lepr-cre;Scf^Δ/FL mice and Tie2-cre;Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL controls (Figures S1A–S1C). This is consistent with published data indicating that LepR⁺ cells and endothelial cells are the major sources of Scf in normal young adult bone marrow (Ding et al., 2012; Oguro et al., 2013).

Figure 1. SCF from LepR⁺ Stromal Cells Is Required to Maintain c-kit⁺-Restricted Progenitors in Bone Marrow.

(A) qRT-PCR analysis of Scf transcript levels in LepR⁺ stromal cells, endothelial cells, and unfractionated cells isolated from bone marrow. Data are normalized to Scf transcript levels in unfractionated bone marrow cells.

(B) qRT-PCR analysis of Scf transcript levels in unfractionated bone marrow cells from mice of the indicated genotypes. The same bar colors are used for the same genotypes throughout the figure.

(C) Bone marrow cellularity from two tibias and two femurs.

(D–G) The frequencies of (D) HSCs, (E) MPPs, (F) HPC-1 cells, and (G) HPC-2 cells in the bone marrow (see Figure S1E and Table S3 for markers and sorting gates).

(H) Representative flow cytometry plots of CMPs, MEPs, and GMPs in the bone marrow along with the percentages of bone marrow cells in each population.

(I–L) The frequencies of (I) CLPs, (J) CMPs, (K) MEPs, and (L) GMPs in the bone marrow.

(M) Percentage of bone marrow cells that formed myeloid colonies in culture.

(N) The frequencies of Mac-1⁺Gr1⁺ myeloid cells in the bone marrow.

(O–R) Numbers of (O) myeloid cells, (P) B cells, (Q) T cells, and (R) platelets per μL of blood.

All data represent mean ± SD. The number of mice analyzed per genotype is shown on each bar in each panel. Statistical significance was assessed using one-way ANOVAs with Tukey’s correction for multiple comparisons (A–G and I–Q) and Kruskal-Wallis test (R; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). See also Figures S1 and S2 and Table S3.

Relative to Scf^Δ/FL controls, bone marrow cellularity was not significantly different in Tie2-cre;Scf^Δ/FL mice but was significantly lower in Lepr-cre;Scf^Δ/FL mice and in Tie2-cre;Lepr-cre;Scf^Δ/FL mice (Figure 1C). Relative to Scf^Δ/FL controls, the frequencies and absolute numbers of HSCs in the bone marrow were significantly reduced in Tie2-cre;Scf^Δ/FL mice, Lepr-cre;Scf^Δ/FL mice, and Tie2-cre;Lepr-cre;Scf^Δ/FL mice (Figures 1D and S1G). The markers used for the isolation of each cell population are shown in Table S3, and the flow cytometric gates are shown in Figures S1D and S1E. Consistent with published studies (Ding et al., 2012; Oguro et al., 2013), competitive transplantation assays confirmed that HSC activity was significantly reduced in bone marrow from Tie2-cre;Scf^Δ/FL mice, Lepr-cre;Scf^Δ/FL mice, and Tie2-cre;Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL controls (Figure S1F). The frequency and number of multipotent progenitors (MPPs) in the bone marrow were not significantly reduced in Tie2-cre;Scf^Δ/FL mice but were significantly reduced in Lepr-cre;Scf^Δ/FL mice and in Tie2-cre;Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL controls (Figures 1E and S1H). This suggests that MPPs depend upon SCF synthesized by LepR⁺ cells, but not by endothelial cells.

To begin to assess c-kit⁺ progenitors downstream of HSCs and MPPs, we first examined hematopoietic progenitor cells HPC-1 and HPC-2. Single-cell functional assays in vitro and in vivo suggest that these populations contain heterogeneous mixtures of restricted progenitors (Oguro et al., 2013). Relative to Scf^Δ/FL control mice, the frequencies of HPC-1 and HPC-2 cells were not significantly altered by conditional deletion of Scf from LepR⁺ cells and/or endothelial cells (Figures 1F and 1G). Due to a reduction in bone marrow cellularity after deletion of Scf from LepR⁺ cells (Figure 1C), the absolute numbers of HPC-1 cells were significantly reduced in the bone marrow of Lepr-cre;Scf^Δ/FL mice and Tie2-cre;Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL controls (Figure S1I). The absolute numbers of HPC-2 cells were significantly reduced in the bone marrow of Tie2-cre;Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL controls (Figure S1J). However, the frequencies and absolute numbers of HPC-1 and HPC-2 cells did not significantly differ between Lepr-cre;Scf^FL/FL mice and Scf^FL/FL controls (Figures S2M and S2N). The functions of HPC-1 and HPC-2 cells also did not significantly differ between Lepr-cre;Scf^Δ/FL mice, Tie2-cre;Lepr-cre;Scf^Δ/FL mice, and Scf^Δ/FL controls based on the percentage of cells that formed colonies in culture, the size of the colonies, or the re-constituting activity after competitive transplantation into irradiated mice, with the exception that HPC-2 cells from mutant mice tended to form slightly larger colonies and to give higher levels of myeloid reconstitution after transplantation (Figures S1K–S1P). Overall, these results suggest that HPC-1 and HPC-2 cells exhibit a limited dependence on SCF from LepR⁺ cells.

Deletion of Scf from LepR⁺ cells, but not endothelial cells, did significantly reduce the frequencies and absolute numbers of a number of other restricted progenitors in the bone marrow, including CLPs, CMPs, MEPs, and GMPs (Figures 1H–1L and S1Q–S1T). Scf deletion from LepR⁺ cells, but not endothelial cells, also significantly reduced the frequency of bone marrow cells that formed myeloid colonies in culture (Figure 1M), the frequency of Mac-1⁺Gr1⁺ cells in the bone marrow (Figure 1N), and the numbers of myeloid cells in the blood (Figure 1O). Deletion of Scf from LepR⁺ cells tended to reduce platelet counts in the blood, but the difference was not statistically significant (Figure 1R). Deletion of Scf from endothelial cells did not significantly affect the frequencies of CLPs, CMPs, MEPs, or GMPs, and there was no added effect of deleting in both LepR⁺ cells and endothelial cells beyond the effect of deleting only in LepR⁺ cells (Figures 1H–1O). Many c-kit⁺-restricted progenitors, therefore, require Scf from LepR⁺ cells, but not endothelial cells.

These results did not depend on heterozygosity for the Scf null allele, as similar results were observed when we compared Lepr-cre;Scf^FL/FL mice to Scf^FL/FL controls: Lepr-cre;Scf^FL/FL mice exhibited significant depletion of HSCs, CLPs, CMPs, MEPs, GMP, PreMegEs, and CFU-Es in the bone marrow based on both cell frequency and absolute number (Figures S2J–S2U). Bone marrow cells from Tie2-cre;Scf^FL/FL mice exhibited significant depletion of HSCs, but not restricted progenitors (Figures S2A–S2I). These mice did not exhibit evidence of systemic SCF depletion, as coat pigmentation was normal (Figures S2V and S2W).

Although CLPs were depleted by deleting Scf from LepR⁺ cells, downstream B lineage progenitors were not depleted when Scf was deleted from LepR⁺ cells and/or endothelial cells, including proB cells, preB cells, and differentiated B cells (Figures S1V–S1Z). PreB colony-forming cells were also not depleted (Figure S1U). The numbers of B and T cells in the blood were also not significantly affected by deletion of Scf from LepR⁺ cells and/or endothelial cells (Figures 1P and 1Q). Therefore, many restricted progenitors and differentiated cells were present in normal frequencies despite the depletion of HSCs and many c-kit⁺-restricted progenitors in these mice.

Erythropoiesis Requires Scf from LepR⁺ Cells

The observation that CMPs and MEPs were depleted after Scf deletion from LepR⁺ cells (Figures 1J and 1K) raised the possibility that LepR⁺ cells might be a critical source of SCF for erythropoiesis in the bone marrow. Using CD150 and CD105 as markers, the CMP and MEP populations can be divided into a series of restricted erythroid progenitors, including PreMegEs, PreCFU-Es, and CFU-Es (Pronk et al., 2007; gates and markers are in Figure S3A and Table S3). Relative to Scf^Δ/FL controls, Scf deletion from endothelial cells did not significantly affect the frequencies of any of these cell populations, but deletion from LepR⁺ cells significantly reduced the frequencies and absolute numbers of PreMegEs and CFU-Es in the bone marrow (Figures 2A–2E and S3B–S3D). Colony forming assays confirmed that Scf deletion from LepR⁺ cells, but not endothelial cells, depleted CFU-Es (Figure 2F). LepR⁺ cells are thus an important source of SCF for the maintenance of early erythroid progenitors.

Figure 2. Scf Deletion from LepR⁺ Cells Depletes Erythroid Progenitors and Causes Anemia.

(A) Schematic of erythroid lineage cells in the bone marrow.

(B) Representative flow cytometry plots for PreMegE, PreCFU-E, and CFU-E progenitors in the bone marrow of mice of each genotype along with the percentages of bone marrow cells that fell within each population.

(C–E) Frequencies of (C) PreMegE, (D) PreCFU-E, and (E) CFU-E cells in the bone marrow from mice of each genotype (see legend above these panels).

(F) Frequency of CFU-E colonies formed by bone marrow cells in methylcellulose culture.

(G and H) Frequencies of (G) proerythroblasts and (H) basophylic erythroblasts in bone marrow.

(I–L) Red blood cell (RBC) counts (I), reticulocyte frequency (J), mean corpuscular volume (MCV) (K), and mean corpuscular hemoglobin (MCH) (L) content in the blood.

All data represent mean ± SD. The number of mice analyzed per genotype is shown on each bar in each panel. Statistical significance was assessed using one-way ANOVAs with Tukey’s correction for multiple comparisons (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). See also Figure S3 and Table S3.

Deletion of Scf from LepR⁺ cells, but not endothelial cells, also depleted mature erythroid lineage cells, including proerythroblasts and basophylic erythroblasts in the bone marrow, and reduced erythroid cell counts in the blood (Figures 2G–2I, S3E, and S3F). Consistent with this, Lepr-cre;Scf^Δ/FL mice and Tie2-cre;Lepr-cre;Scf^Δ/FL mice (but not Tie2-cre;Scf^Δ/FL mice) exhibited significant increases in reticulocyte frequency in the blood, as well as mean corpuscular volume (MCV) and mean corpuscular hemoglobin (MCH) (Figures 2J–2L), changes consistent with hypoplastic macrocytic anemia. SCF produced by LepR⁺ cells is, therefore, necessary for normal erythropoiesis.

When bone marrow erythropoiesis does not fulfil physiological demands, erythropoiesis expands into the spleen (Paulson et al., 2011). LepR⁺ cells do not appear to be an important source of SCF in the spleen (Inra et al., 2015). Consistent with the reductions in bone marrow hematopoiesis in Lepr-cre;Scf^Δ/FL mice and Tie2-cre;Lepr-cre;Scf^Δ/FL mice, we observed significantly increased extramedullary hematopoiesis in these mice as reflected by increased splenic cellularity (Figure S3G) and increased frequencies and absolute numbers of CMPs, MEPs, GMPs, PreCFU-Es, CFU-Es, proerythroblasts, and basophylic erythroblasts in the spleen (Figures S3H–S3V).

Scf Deletion from LepR⁺ Cells Accelerates Erythroid Progenitor Differentiation

SCF can promote the survival and proliferation, and delay the differentiation, of erythroid progenitors (Muta et al., 1995; Joneja et al., 1997; Wu et al., 1997; Munugalavadla et al., 2005; McIver et al., 2016). We did not detect any effect of Scf deletion from LepR⁺ cells and/or endothelial cells on the frequencies of apoptotic PreMegE, PreCFU-E, or CFU-E cells based on staining for activated-caspase-3 (Figures S4A–S4C). It remains possible there was an increase in cell death we did not detect, for example, if dying cells were cleared quickly. We also detected a significant increase in plasma erythropoietin (EPO) levels after Scf deletion from LepR⁺ cells (Figure S4D), potentially protecting erythroid progenitors from apoptosis after Scf deletion (Waskow et al., 2004). Scf deletion from LepR⁺ cells and/or endothelial cells also did not affect the rate of proliferation of PreMegE (Figure 3A) or CFU-E (Figure 3C) cells based on incorporation of a 2-hr pulse of 5-bromo-2′-deoxyuridine (BrdU) in vivo; however, we did observe increased BrdU incorporation by PreCFU-E cells after Scf deletion from LepR⁺ cells (but not endothelial cells; Figure 3B).

Figure 3. Scf Deletion from LepR⁺ Cells Accelerates Erythroid Differentiation.

(A–C) Incorporation of a 2-hr pulse of BrdU into (A) PreMegE, (B) PreCFU-E, and (C) CFU-E cells in the bone marrow of mice with the indicated genotypes.

(D) Representative flow cytometry plots for CD71 versus c-kit expression in PreCFU-E and CFU-E cells from the bone marrow of mice with the indicated genotypes.

(E) Percentages of CD71⁺ PreCFU-E cells from the bone marrow.

(F) Mean fluorescence intensity for c-kit staining on PreCFU-E cells.

(G) Percentages of CD71⁺ CFU-E cells.

(H) Mean fluorescence intensity for c-kit staining on CFU-E cells.

(I and J) Gene set enrichment analysis showing significant enrichment of genes associated with erythroblast differentiation in PreCFU-Es (I) and CFU-Es (J) from Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL controls. FDR, false discovery rate; NES, normalized enrichment score.

(K) Hemoglobin alpha chain 1, Aminolevulinate dehydratase, and Spectrin beta transcript levels in PreCFU-Es and CFU-Es from mice of the indicated genotypes (see legend on right) by RNA-seq analysis (n = 3).

(L) Myb and Gata2 transcript levels in PreCFU-Es and CFU-Es by RNA-seq analysis (n = 3).

Data represent mean ± SD. The numbers of mice analyzed per genotype is shown in the bars in each panel. Statistical significance was assessed using one-way ANOVAs (A–C, E, and G) or repeated-measures one-way ANOVAs (F and H) with Tukey’s correction for multiple comparisons and the DEseq2 Program with Benjamini-Hochberg corrections for multiple comparisons (K and L; *p < 0.05; **p < 0.01). See also Figure S4 and Table S1.

The commitment and differentiation of erythroid progenitors are regulated by the rate of proliferation (Pop et al., 2010; Tusi et al., 2018). Because PreCFU-Es give rise to more rapidly proliferating CFU-Es in wild-type mice, the increased rate of division by PreCFU-Es after Scf deletion (Figures 3A–3C) raised the possibility that Scf deletion accelerated erythroid differentiation. To test this, we evaluated the expression of CD71 (the transferrin receptor) and c-kit in PreCFU-Es and CFU-Es. During the maturation of PreCFU-Es into CFU-Es, CD71 expression increases while c-kit expression decreases (Munugalavadla et al., 2005; Tusi et al., 2018; compare PreCFU-E and CFU-E in Scf^FL/FL control mice in Figure 3D). Scf deletion from LepR⁺ cells increased CD71 expression and decreased c-kit expression in PreCFU-Es and reduced c-kit expression in CFU-Es (Figures 3D–3H). This suggests that Scf deletion from LepR⁺ cells accelerated the differentiation of erythroid progenitors. Consistent with this possibility, SCF can delay the differentiation of erythroid progenitors in culture (Muta et al., 1995; Joneja et al., 1997; Wu et al., 1997; Munugalavadla et al., 2005; McIver et al., 2016).

To further test this, we performed RNA sequencing (RNA-seq) analysis on PreCFU-E and CFU-E cells from Scf^FL/FL, Scf^Δ/FL, and Lepr-cre;Scf^Δ/FL mice. Only four genes significantly differed among the cells isolated from Scf^FL/FL versus Scf^Δ/FL control mice, and no Gene Ontology terms were enriched among these few differentially expressed genes. In PreCFU-Es and CFU-Es isolated from Lepr-cre;Scf^Δ/FL mice, hundreds of genes were differentially expressed relative to cells from Scf^Δ/FL controls. Among the gene sets that significantly increased in expression in PreCFU-E cells from Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL mice, 4 of the top 5 categories related to erythrocyte differentiation (Table S1). Among the gene sets that significantly increased in expression in CFU-E cells from Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL mice, 2 of the top 5 related to erythrocyte differentiation (Table S1).

We also performed gene set enrichment analysis using two curated sets of genes that are upregulated in erythroblasts as compared to earlier erythroid progenitors (Wong et al., 2011; Tusi et al., 2018). Both gene sets were significantly enriched among the genes upregulated in PreCFU-Es and CFU-Es from Lepr-cre;Scf^Δ/FL mice (Figures 3I and 3J). Genes that increase in expression during erythrocyte differentiation, including hemoglobin alpha chain 1, aminolevulinate dehydratase (Nilsson et al., 2009), and spectrin beta (Chen et al., 2009) were significantly increased in PreCFU-Es and CFU-Es from Lepr-cre;Scf^Δ/FL mice as compared to Scf^Δ/FL mice (Figure 3K). Conversely, genes that decrease in expression during erythrocyte differentiation, including c-Myb (Lyon and Watson, 1995) and Gata2 (Leonard et al., 1993), were significantly decreased in PreCFU-Es and CFU-Es from Lepr-cre;Scf^Δ/FL mice (Figure 3L).

Early Erythroid Progenitors Reside in Perisinusoidal Niches with LepR⁺ Cells

To test whether erythroid progenitors are closely associated with LepR⁺ cells in the bone marrow, we performed deep confocal imaging of cleared, bisected tibias (Acar et al., 2015). We localized a mixed population of early erythroid progenitors (PreCFU-E plus CFU-E cells) as Lin⁻c-kit⁺Sca1⁻CD41⁻CD105⁺ cells (hereafter referred to as c-kit⁺CD105⁺ progenitors) in digitally reconstructed 3D images from Lepr-cre;Rosa26-tdTomato mice (Figures 4A and 4B). We found that 93% ± 3.7% of c-kit⁺CD105⁺ progenitors were immediately adjacent to LepR⁺ cells or processes from LepR⁺ cells, significantly more than would be expected by chance (Figure 4C). We also localized these cells in bisected bones from Cdh5-creER;Rosa26-tdTomato mice to assess the localization relative to sinusoids versus arterioles (Acar et al., 2015). We found that 98% ± 0.7% of c-kit⁺CD105⁺ progenitors were most closely associated with sinusoidal blood vessels, usually within 5 mm, and the vast majority of these cells were distant from arterioles, usually more than 35 mm (Figures 4D–4F). PreCFU-E and CFU-E erythroid progenitors thus reside in perisinusoidal niches in the bone marrow, closely associated with LepR⁺ cells.

Figure 4. Lineage^–Sca1^–c-kit⁺CD105⁺ Erythroid Progenitors Reside in Perisinusoidal Niches Adjacent to LepR⁺ Stromal Cells.

(A) Representative tiled composite confocal image of a 3-μm optical section from a bisected tibia from a Lepr-cre;Rosa26-tdTomato mouse stained with anti-bodies against Lineage markers and Sca1 (blue), c-kit (green), CD105 (white), and tdTomato (red).

(B) Higher magnification image from (A) (white square), showing a Lineage⁻Sca1⁻c-kit⁺CD105⁺ cell (arrow) adjacent to a perisinusoidal LepR⁺ cell (red; * marks the sinusoid lumen).

(C) Distances between Lineage⁻Sca1⁻c-kit⁺CD105⁺ cells or random spots and the closest LepR⁺ cells in tibias from Lepr-cre;Rosa26-tdTomato mice (n = 5).

(D and E) Distances between Lineage⁻Sca1⁻c-kit⁺CD105⁺ cells or random spots and the closest sinusoids (D) or arterioles (E) in tibias from Cdh5-creER;Rosa26-tdTomato mice (n = 3).

(F) Percentages of Lineage⁻Sca1⁻c-kit⁺CD105⁺ cells or random spots that are closest to sinusoids versus arterioles in tibias from Cdh5-creER;Rosa26-tdTomato mice (n = 3).

In (C)–(F), data represent mean ± SD. In (C)–(E), statistical significance was assessed using Kolmogorov-Smirnov tests.

DISCUSSION

Few studies have functionally identified the cellular sources of factors required for the maintenance of restricted hematopoietic progenitors in vivo (Ding and Morrison, 2013; Greenbaum et al., 2013; Cordeiro Gomes et al., 2016). Here, we show that many c-kit⁺-restricted progenitors in the bone marrow depend upon SCF synthesized by LepR⁺ cells. Early erythroid progenitors, and likely other c-kit⁺-restricted progenitors, reside in perisinusoidal niches adjacent to LepR⁺ cells (Figure 4). HSCs also reside mainly in perisinusoidal niches in which they depend upon SCF, Cxcl12, and pleiotrophin synthesized by LepR⁺ cells and endothelial cells (Ding et al., 2012; Ding and Morrison, 2013; Himburg et al., 2018). This raises the possibility that many restricted progenitors co-occupy the same niches as HSCs. It is unclear why HSCs depend upon SCF synthesized by endothelial cells (Ding et al., 2012; Ding and Morrison, 2013; Greenbaum et al., 2013; Xu et al., 2018), but most restricted progenitors do not. There are many potential differences in the presentation of SCF by endothelial cells versus LepR⁺ cells.

The depletion of restricted progenitors after Scf deletion from LepR⁺ cells appears to be independent of HSC depletion. PreCFU-E cells, proB cells, preB cells, and mature B cells were not significantly depleted from the bone marrow after Scf deletion from LepR⁺ cells (Figures 2D, S3C, and S1V–S1Z). Therefore, not all restricted progenitors were depleted by Scf deletion, as would be expected if restricted progenitor depletion had been caused by HSC depletion. Moreover, Scf deletion from endothelial cells depleted HSCs without affecting the frequencies or absolute numbers of restricted progenitors (Figures 1D–1L and S1I–S1Z). Therefore, HSC depletion does not, in itself, lead to the depletion of restricted progenitors. Furthermore, the frequencies and functions of HPC-1 and HPC-2 progenitors that appear to be immediately downstream of MPPs, at early stages of lineage restriction, were not reduced by Scf deletion from LepR⁺ cells (Figures 1F, 1G, S1K–S1P, S2M, and S2N).

Mice with a loss of Scf function have bone marrow hypocellularity, HSC depletion, and hypoplastic macrocytic anemia (Russell, 1979; Huang et al., 1990; Zsebo et al., 1990; Barker, 1994). Our data show that deletion of Scf from LepR⁺ cells recapitulated all of these phenotypes, demonstrating that LepR⁺ cells are a major source of SCF for erythropoiesis in the bone marrow.

STAR★METHODS

CONTACT FOR REAGENTS AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Sean Morrison (sean.morrison@utsouthwestern.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

Scf^GFP (Kitl^tm1.1Sjm/J, RRID:IMSR_JAX:017860), Scf^FL (Kitl^tm2.1Sjm/J, RRID:IMSR_JAX:017861), Scf^Δ (Kitl^tm2.2Sjm/J, RRID:MGI: 5306348) were generated in our laboratory (Ding et al., 2012), Tie2-cre (B6.Cg-Tg(Tek-cre)1Ywa/J, RRID:IMSR_JAX:008863) (Kisanuki et al., 2001) mice, Lepr-cre (B6.129(Cg)-Lepr^tm2(cre)Rck/J, RRID:IMSR_JAX:008320) (DeFalco et al., 2001) mice, and Rosa26-tdTomato (B6.Cg-Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}/J, RRID:IMSR_JAX:007909) mice were obtained from Jackson Laboratory. Cdh5-creER (B6.CBA-Tg(Cdh5-cre/ERT2)1Rha, RRID:IMSR_TAC:13073) transgenic mice were provided by Dr. Ralf H. Adams (Sörensen et al., 2009). Mice were all backcrossed at least six times onto a C57Bl/Ka background and were maintained on this background. Germline Scf loss-of-function was achieved using the Scf^GFP or the Scf^Δ allele. We used 8–16 week-old age-matched mice in all experiments. To induce Cre recombinase activity, Cdh5-creER;Rosa26-tdTomato mice received an intraperitoneal injection of 2mg of tamoxifen dissolved in corn oil every second day for 10 days (5 injections total). All mice were housed in AAALAC-accredited, specific-pathogen-free animal care facilities at the University of Texas Southwestern Medical Center (UTSW). All procedures were approved by the UTSW Institutional Animal Care and Use Committee.

Primary cell culture

To assess the frequency of colony-forming units erythroid (CFU-E), 1×10⁵ total bone marrow cells or 2×10⁵ total spleen cells were plated in Methocult GM M3434 medium supplemented with 10ng/ml of TPO and 1X penicillin/streptomycin in 6-well plates. To assess the frequency of colony-forming units granulocyte/macrophage (CFU-GM), 1×10⁴ total bone marrow cells were plated in Methocult GF M3534 medium supplemented with 10ng/ml Flt-3 ligand, 25ng/ml of GM-CSF, and 1x penicillin/streptomycin in 6-well plates. To assess the frequency of CFU-preB, 1×10⁵ total bone marrow cells were plated in Methocult M3630 medium supplemented with 20ng/ml of Flt-3 ligand, 100ng/ml SCF, and 1X penicillin/streptomycin in 6-well plates. CFU-E colonies were counted after 3 days using an inverted microscope. CFU-GM and CFU-preB colonies were counted after 8 days using an inverted microscope. To assess the colony-forming capacity of HPC-1 cells and HPC-2 cells, 200 cells were double sorted by flow cytometry and plated in Methocult GM M3434 medium supplemented with 10ng/ml of TPO and 1X penicillin/streptomycin in 6-well plates. Colonies were counted after 8 days using an inverted microscope.

METHOD DETAILS

Flow cytometry analysis and sorting of hematopoietic cells

Bone marrow was flushed from two tibias and two femurs using staining medium (HBSS supplemented with 2% bovine serum) and dissociated into a single cell suspension by gently triturating with a 23 gauge needle. Spleens were mechanically dissociated by crushing through a 40 μm cell strainer using staining medium. Cells were counted, and then stained with antibodies at 4°C for 30 minutes (except for anti-CD34 that was incubated for 90 minutes). For staining of HSCs and restricted hematopoietic progenitors, cells were first stained with antibodies against lineage markers (biotin-conjugated antibodies against CD2, CD3, CD5, CD8a, B220, Ter119, and Gr1), washed with staining medium and then resuspended in fluorophore-conjugated antibodies against c-kit, Sca1, CD150, CD48, CD34, CD16/32, Flt3, IL7Rα, and efluor450-conjugated streptavidin. For erythroid progenitors, cells were stained with fluorophore-conjugated antibodies against lineage markers (CD2, CD3, CD5, CD8a, B220, Ter119, and Gr1), c-kit, Sca1, CD150, CD41, CD16/32, CD71, and CD105. For analysis of lymphoid and myeloid cells, cells were stained with fluorophore-conjugated antibodies against Mac-1, Gr1, B220, CD43, and IgM. For erythroblasts, cells were stained with fluorophore-conjugated antibodies against CD71 and Ter119. Cells were then analyzed using a FACSAria II SORP (BD Biosciences) or a FACSAria Fusion SORP (BD Biosciences) cytometer.

To sort erythroid progenitors, HPC-1 cells and HPC-2 cells, tibias, femurs, pelvises, and spines were collected and crushed using a mortar and pestle. Cells were resuspended in staining medium and filtered through a 40 μm cell strainer. The cells were stained with APC-efluo780 conjugated anti-c-kit antibody and c-kit⁺ cells were enriched using anti-APC paramagnetic microbeads (Miltenyi Biotec). For erythroid progenitors, cells were then stained with fluorophore-conjugated antibodies against lineage markers (CD2, CD3, CD5, CD8a, B220, Ter119, and Gr1), Sca1, CD150, CD41, CD16/32, CD105. For HPC-1 and HPC-2, cells were stained with fluorophore-conjugated antibodies against lineage markers (CD2, CD3, CD5, CD8a, B220, Ter119, and Gr1), Sca1, CD150, and CD48. Cells were isolated by two successive rounds of sorting to ensure purity using a FACSAria II cytometer.

For peripheral blood analysis, 50 μL of blood were collected from the tail vein or from the heart before euthanasia and mixed with 5μL of 0.5M EDTA to prevent clotting. Platelets, RBC, MCV and MCH were evaluated using a Hemavet HV950 (Drew Scientific). To determine reticulocyte frequency in the blood, 2μL of blood were stained in 1ml of 0.1μg/ml Thiazole Orange in HBSS containing 2mM EDTA and 0.02% sodium azide. Samples were incubated for 1h at room temperature, and then analyzed by flow cytometry. For myeloid and lymphoid cells analysis, red blood cells were lysed and samples were washed in 5ml of staining medium. Cells were then stained with fluorophore-conjugated antibodies against B220, CD3, Mac-1 and Gr1 and analyzed using a FACS Canto cytometer. Dead cells were identified and gated out of all analyses by staining with DAPI or with propidium iodide after antibody staining.

Flow cytometric sorting of stromal cells

Bone marrow stromal cells were enzymatically digested from flushed tibias and femurs as previously described (Zhou et al., 2014). Briefly, tibias and femurs were gently flushed and then digested in HBSS plus calcium and magnesium supplemented with DNase I (200U/ml) and Liberase^DL (250μg/ml), agitating for 30 minutes at 37°C. Bone marrow fragments were allowed to sediment for ~1min, then the cell suspension was collected. Cell were washed once in staining medium and incubated with a biotin-conjugated goat anti-LepR antibody for 90 min at 4°C. Cells were washed and then incubated with fluorophore-conjugated antibodies against CD45 and Ter119 and PE-conjugated streptavidin. Endothelial cells were stained by intravenous injection of 10μg of anti-VE-cadherin antibody 5–10 minutes before killing the mice. Cells were purified with two successive rounds of sorting to ensure high purity using a FACSAria II SORP cytometer.

Bone marrow reconstitution assays

Recipient (CD45.1/CD45.2) mice were irradiated using an XRAD 320 X-ray irradiator (Precision X-Ray Inc.) with two doses of 540 rad at least 4h apart. For whole bone marrow transplantation, 500,000 unfractionated bone marrow cells from donor (CD45.2) and competitor (CD45.1) mice were mixed and injected intravenously through the retro-orbital venous sinus. Every 4 weeks until at least 16 weeks after transplantation, blood was collected from the tail vein and subjected to ammonium-chloride potassium chloride red cell lysis. Cells were then stained with antibodies against CD45.1, CD45.2, Mac-1, Gr1, B220, and CD3 to evaluate levels of donor cell engraftment in the myeloid, B, and T cell lineages. For transplantation of HPC-1 cells and HPC-2 cells, 200 HPC-1 or HPC-2 cells from donor (CD45.2) mice were mixed with 200,000 bone marrow cells from competitor (CD45.1) mice and injected intravenously through the retro-orbital sinus into CD45.1/CD45.2 recipients. Levels of donor cell engraftment were evaluated in the bone marrow of recipient mice at 14 days after transplantation by staining with antibodies against CD45.1, CD45.2, Mac-1, Gr1, B220, and CD3.

5-bromodeoxyuridine incorporation

Mice were intraperitoneally injected with 0.1mg/g body mass of 5-bromo-2′-deoxyuridine (BrdU) and killed 2 hours after injection. Bone marrow cells were obtained and c-kit⁺ cells were enrichment as described above, Lineage⁻c-kit⁺Sca1⁻ cells were sorted in purity mode into a 1.5ml eppendorf tube using a FACSAria II flow cytometer. Staining for BrdU incorporation was performed using the BD APC BrdU Flow Kit following the manufacturer’s instructions. BrdU levels in each sorted cell population were analyzed using a FACSAria II flow cytometer.

Analysis of apoptosis by caspase-3 staining

After isolation and counting of cells from the bone marrow, 5×10⁶ total bone marrow cells were stained with fluorophore-conjugated antibodies against lineage markers (CD2, CD3, CD5, CD8a, B220, Ter119, and Gr1), Sca1, CD150, CD41, CD16/32 and CD105. Cells were then washed and fixed using the BD Fixation/Permeabilization Solution Kit according to the manufacturer’s instruction. Cells were then stained using a rabbit anti-activated caspase-3 antibody. Data were acquired using a FACSAria II flow cytometer.

Measurement of EPO plasma levels by ELISA

Plasma was obtained from peripheral blood after centrifugation. EPO levels were measured using the Mouse Erythropoietin Quantikine ELISA Kit following the manufacturer’s instructions.

Deep imaging of half bones

Sample preparation and immunostaining of bisected (half) tibias were performed as previously described (Acar et al., 2015). Briefly, tibias from Lepr-cre;Rosa26-tdTomato or Cdh5-creER;Rosa26-tdTomato mice were fixed in 4% paraformaldehyde solution in PBS for 6 hours with gentle rocking at 4°C. After fixation, bones were embedded into OCT embedding medium. Each tibia was cut in two using a cryostat and then washed three times in PBS. Half bones were blocked using whole mount staining medium (PBS with 5% donkey serum, 0.5% NP-40, and 10% DMSO) overnight at room temperature. Half bones were then stained in whole mount staining medium with antibodies against lineage markers (FITC-conjugated CD3, B220, Ter119, Gr1, CD41), as well as goat anti-c-kit and rabbit anti-DsRed/Tomato antibodies for three days at room temperature. The samples were washed three times for 5 minutes in PBS and then overnight in PBS. Half bones were then stained with anti-Sca1 Alexa 488, anti-CD105 Alexa-647, mouse anti-FITC Alexa 488, donkey anti-goat Alexa-555 and donkey anti-rabbit DyLight-405 antibodies for three days at room temperature. The samples were washed three times for 5 minutes in PBS and then overnight in PBS. Bone clearing was performed in eppendorf tubes, gently rotating at room temperature. Samples were dehydrated through a methanol series before clearing with Murray’s clear (1:2 Benzyl Alcohol: Benzyl Benzoate; BABB) solution. Half bones were mounted in BABB and three dimensional images of the bone marrow were acquired using a Confocal/Multiphoton Zeiss LSM880 Inverted microscope with a Zeiss LD LCI Plan-Apo 253/0.8 multi-immersion objective lens.

Image analysis and quantification

Confocal half bone images were rendered in three dimensions and analyzed using Bitplane Imaris V8.2. Automated identification of Lin⁻c-kit⁺Sca1⁻CD105⁺ erythroid progenitors required a workflow including several steps. First, we subtracted the Lineage marker signal from the CD105 signal using the channel arithmetic function in Imaris to identify Lin⁻CD105⁺ cells. Second, we inverted the c-kit channel, setting the signal for positive cells above a predetermined threshold to zero, and then subtracted the inverted c-kit signal from the Lin⁻CD105⁺ channel using the channel arithmetic function. This allowed us to identify Lin⁻c-kit⁺Sca1⁻CD105⁺ erythroid progenitors in a single channel. Finally, we created a digital surface for Lin⁻c-kit⁺Sca1⁻CD105⁺ progenitors using the Imaris surface function, by manually selecting image intensity values that would distinguish Lin⁻c-kit⁺Sca1⁻CD105⁺ progenitors from other cells and from debris or staining artifacts. The original, unprocessed images were then manually checked to verify that each cell identified as a Lin⁻c-kit⁺Sca1⁻CD105⁺ progenitors had the appropriate staining patterns and morphology in each channel. To determine the distances of Lin⁻c-kit⁺Sca1⁻CD105⁺ cells to tdTomato⁺ cells (from Lepr-cre;Rosa26-tdTomato and Cdh5-creER;Rosa26-tdTomato mice), we used the Imaris Surface function to create a digital surface for LepR⁺ cells or endothelial cells based on Tomato signal intensity. To distinguish between arterioles and sinusoids in Cdh5-creER;Rosa26-tdTomato bone marrow samples, the tdTomato⁺ channel was first segmented manually based on blood vessel morphology and then masked to produce two channels, one representing the tdTomato⁺ signal from arterioles and the other representing the tdTomato⁺ signal from sinusoids. Three-dimensional distances between Lin⁻c-kit⁺Sca1⁻CD105⁺ cells and these digital surfaces were calculated using the Imaris Distance Transform MATLAB XTension. To generate random spot distributions in three dimensional space, we used the random number generator function of Microsoft Excel to create random combinations of 3 numbers representing X, Y, and Z spatial coordinates. These coordinates were then imported as 8mm spots using the Imaris Create Spots From File Python XTension. To ensure that random spots were only distributed over hematopoietic regions of the bone marrow we excluded regions of bone, blood vessel lumens, and poorly stained areas from the analysis.

RNA extraction and real time qPCR

For total bone marrow extracts, bone marrow was isolated from tibias by centrifugation at 3000 RPM for 3 minutes at 4°C. The cell pellet was then resuspended in Trizol and stored at −80°C. Total RNA was isolated following the manufacturer’s instructions. 1 μg of RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit. For sorted stromal and endothelial cells, 5,000–10,000 cells were sorted in staining medium, spun down, and resuspended in Trizol. Total RNA was extracted and reverse transcribed as above. Primers used for qPCR are listed in Table S2. Transcript levels were normalized to Actin (Actb) and fold change was calculated using the ΔCt method.

RNA-Seq library preparation and data analysis

5,000–10,000 erythroid progenitors were sorted in tubes containing RLT buffer from the RNeasy Micro Kit and stored at −80°C. Total RNA was extracted following the kit instructions, and RNA integrity and concentration were measure using a Bioanalyzer. Two to six ng of RNA were used to generate a cDNA library using the SMARTer Stranded Total RNA-Seq Kit v2 - Pico Input Mammalian. Samples were sequenced with a NextSeq 500 (Illumina) using a 75bp single-read setup. Adaptor removal and quality trimming were performed using Trimmomatic. Reads were aligned to the mouse mm10 reference genome using STAR 2.5.2b. Gencode vM18 annotation was used as reference to build STAR indexes and alignment annotation. For each sample, a BAM file including mapped and unmapped reads with spanning splice junctions was produced. Secondary alignment and multi-mapped reads were further removed using in-house scripts. Only uniquely mapped reads were retained for further analyses. Quality control metrics were performed using RseqQC using the mm10 gene model provided. These steps include: number of reads after multiple-step filtering, ribosomal RNA reads depletion, and defining reads mapped to exons, UTRs, and intronic regions. Picard tool was implemented to refine the QC metrics (http://broadinstitute.github.io/picard/). Gene level expression was calculated using HTseq version 0.9.1 using intersection-strict mode by gene. Counts were calculated based on protein-coding gene annotation from the Gencode vM18 annotation file. RPKM (reads per kilobase of transcript per million reads mapped) values were calculated using edgeR. Length was curated using the protein-coding gene annotation from the Gencode vM18 annotation file. RPKM values were filtered for downstream differential and co-expression analyses using a “by condition” RPKM cutoff. Briefly, a gene is considered expressed if the RPKM > 0 in all three biological replicates (e.g., in Scf^Δ/FL) in any of the cell types analyzed. Differential expression between control and mutant samples was assessed using the DESeq2 package in R. All differentially expressed genes with an FDR ≤ 0.05 and log₂(fold change) R ≥ |0.3| (1.23 fold change) were retained. A permutation test was also applied using 1000 permuted experiments. None of these permuted analyses showed the same genes differentially expressed (permutation p < 0.001). The functional enrichment of GO terms among differentially expressed genes was assessed using ToppGene. A Benjamini-Hochberg FDR (p < 0.05) was applied for a multiple comparisons adjustment. Gene set enrichment analysis (GSEA) and visualization were performed using the fgsea package in R (https://github.com/ctlab/fgsea) using two gene signatures for genes highly expressed in erythroblasts (Wong et al., 2011) and proerythroblasts (Tusi et al., 2018).

QUANTIFICATION AND STATISTICAL ANALYSIS

Data in all figures were obtained in at least three independent experiments using different mice, as indicated in each figure panel. Data are shown as mean ± standard deviation. Flow cytometry data were analyzed using Flowjo v.10.4 (Flowjo, LLC).

Prior to analyzing the statistical significance of differences among treatments, we tested for normality in the data distributions using the D’Agostino-Pearson omnibus test for samples with n > 8 or the Shapiro-Wilk normality test for smaller samples. We also tested for similar variability among treatments using an F-test (when comparing two samples) or Levene’s median tests (when comparing more than two samples). When data were normally distributed and variability was similar among treatments, we used parametric tests to assess statistical significance. When the data significantly deviated from normality (p < 0.01) or variability was significantly different (p < 0.01) among treatments, we log2 transformed the data and tested for normality and variability. If the log2 transformed data did not significantly deviate from normality and variability among treatments did not significantly differ, we performed parametric tests to assess the statistical significance of differences among treatments using the log2 transformed data. If the log2 transformed data did significantly deviate from normality or similar variability among treatments, we performed non-parametric tests on the non-transformed data.

Statistical analyses were performed using one-way ANOVAs with Tukey’s corrections for multiple comparisons and Kruskal-Wallis non-parametric tests when comparing more than two treatments (Graphpad Prism v.6). We used unpaired Student’s t tests with Welch’s correction when comparing two treatments (Graphpad Prism v.6). For transplantation experiments in Figure S1F, we used a non-parametric mixed model, nparLD, to test for differences in donor cell reconstitution levels among cells of different genotypes. In Figures 4C–4E, we applied a normalized two-sample Kolmogorov-Smirnov test as follows. Distances to erythroid progenitors or random spots within each biological replicate were pooled to generate 100 bins of distances, each bin corresponding to 1% of the maximum distance for each biological replicate. The frequency of erythroid progenitors or random spots in each bin was calculated for each biological replicate, then used to perform a two-sample Kolmogorov-Smirnov test using MATLAB.

DATA AND SOFTWARE AVAILABILITY

The accession number for the RNA-seq data generated in this paper from sorted PreCFU-E and CFU-E cells from Scf^FL/FL, Scf^Δ/FL and Lepr-cre;Scf^Δ/FL are accessible at GEO: GSE122468. Software used to analyze the data is described in the Method Details section.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
α-mouse Gr1 Biotin	Tonbo Biosciences	clone RB6–8C5; RRID:AB_2621652
α-mouse Gr1 FITC	Tonbo Biosciences	clone RB6–8C5; RRID:AB_2621721
α-mouse Gr1 PE-efluor610	eBioscience	clone RB6–8C5; Cat. 61-5931-80
α-mouse Gr1 BV510	Biolegend	clone RB6–8C5; RRID:AB_2650931
α-mouse Gr1 PE	Tonbo Biosciences	clone RB6–8C5; RRID:AB_2621803
α-mouse CD2 Biotin	Biolegend	clone RM2–5; RRID:AB_312650
α-mouse CD2 FITC	Tonbo Biosciences	clone RM2–5; RRID:AB_2621657
α-mouse CD2 PE	Tonbo Biosciences	clone RM2–5; RRID:AB_2621728
α-mouse CD5 Biotin	Biolegend	clone 53–7.3; RRID:AB_312732
α-mouse CD5 FITC	Biolegend	clone 53–7.3; RRID:AB_312734
α-mouse CD5 PE	Biolegend	clone 53–7.3; RRID:AB_312736
α-mouse CD3 Biotin	Biolegend	clone 17A2; RRID:AB_2563946
α-mouse CD3 FITC	Biolegend	clone 17A2; RRID:AB_312661
α-mouse CD3 PE	Tonbo Biosciences	clone 17A2; RRID:AB_2621731
α-mouse CD3 Percp-Cy5.5	Tonbo Biosciences	clone 145–2C11; RRID:AB_2621872
α-mouse CD8a Biotin	Tonbo Biosciences	clone 53–6.7; RRID:AB_2621638
α-mouse CD8a FITC	Tonbo Biosciences	clone 53–6.7; RRID:AB_2621671
α-mouse CD8a PE	Tonbo Biosciences	clone 53–6.7; RRID:AB_2621741
α-human/mouse B220 Biotin	Tonbo Biosciences	clone RA3–6B2; RRID:AB_2621644
α-human/mouse B220 FITC	Tonbo Biosciences	clone RA3–6B2; RRID:AB_2621690
α-human/mouse B220 APC	Tonbo Biosciences	clone RA3–6B2; RRID:AB_2621574
α-human/mouse B220 Percp-Cy5.5	Tonbo Biosciences	clone RA3–6B2; RRID:AB_2621892
α-human/mouse B220 PE	Tonbo Biosciences	clone RA3–6B2; RRID:AB_2621764
α-mouse CD45.1 PE-Cy7	Tonbo Biosciences	A20; RRID:AB_2621850
α-mouse CD45.2 VioletFluor 450	Tonbo Biosciences	104; RRID:AB_2621950
α-mouse Ter119 Biotin	Tonbo Biosciences	clone TER-119; RRID:AB_2621651
α-mouse Ter119 FITC	Tonbo Biosciences	clone TER-119; RRID:AB_2621720
α-mouse Ter119 PE	Tonbo Biosciences	clone TER-119; RRID:AB_2621802
α-mouse Ter119 PE-Cy7	Biolegend	clone TER-119; RRID:AB_2281408
α-mouse Ter119 BV650	Biolegend	clone TER-119; RRID:AB_11204244
α-mouse CD71 FITC	eBioscience	clone R17217; RRID:AB_465124
α-mouse CD71 BV421	Biolegend	clone R17217; RRID:AB_10899739
α-mouse CD71 PE-Cy7	Biolegend	clone R17217; RRID:AB_2203383
α-mouse c-kit APC-eFluor780	eBioscience	clone 2B8; RRID:AB_1272177
α-mouse Sca1 Percp-Cy5.5	eBioscience	clone D7; RRID:AB_1272177
α-mouse Sca1 Alexa 488	eBioscience	clone D7; RRID:AB_493270
α-mouse CD150 PE	Biolegend	clone TC15–12F12.2; RRID:AB_313682
α-mouse CD150 PE-Cy7	Biolegend	clone TC15–12F12.2; RRID:AB_439796
α-mouse CD150 APC	Biolegend	clone TC15–12F12.2; RRID:AB_493461
α-mouse CD48 APC	Biolegend	clone HM48–1; RRID:AB_571996
α-mouse CD48 PE-Cy7	Biolegend	clone HM48–1; RRID:AB_2075049
α-mouse CD48 Alexa Fluor700	Biolegend	Clone HM48–1; RRID:AB_10612754
α-mouse CD34 FITC	eBioscience	clone RAM34; RRID:AB_465022
α-mouse IL7Rα (CD127) PE	Tonbo Biosciences	clone A7R34; RRID:AB_2621780
α-mouse IL7Rα (CD127) PE-Cy7	Tonbo Biosciences	clone A7R34; RRID:AB_2621859
α-mouse Flt3 (CD135) PE-Cy5	Biolegend	clone A2F10; RRID:AB_2107049
α-mouse Flt3 (CD135) APC	Biolegend	clone A2F10; RRID:AB_2107050
α-mouse CD43 PE	BD Biosciences	clone S7; RRID:AB_394748
α-mouse IgM APC	eBioscience	clone Il/41; RRID:AB_469458)
α-mouse Mac-1 (CD11b) APC-efluor780	eBioscience	clone M1/70; RRID:AB_1603193
α-mouse CD16/32 (FcγRIII/II) BV510	Biolegend	clone 93; RRID:AB_2563692
α-mouse CD16/32 (FcγRIII/II) Alexa Fluor700	eBioscience	clone 93; RRID:AB_493994
α-mouse CD41 FITC	Biolegend	clone MWReg30; RRID:AB_2129746
α-mouse CD41 Alexa Fluor700	Biolegend	clone MWReg30; RRID:AB_2572129
α-mouse CD105 BV421	BD Biosciences	clone MJ/78; RRID:AB_2734710
α-mouse CD105 APC	Biolegend	clone MJ/78; RRID:AB_2687060
α-mouse CD105 Alexa 647	Biolegend	clone MJ/78; RRID:AB_2728140
α-mouse CD45 FITC	Tonbo Biosciences	clone 30-F11; RRID:AB_2621689
Goat α-mouse LepR antibody, Biotin	R&D System	BAF497; RRID:AB_2296953
α-mouse VE-cadherin (CD144) efluor660	eBioscience	clone BV13; RRID:AB_11219483
Streptavidin efluor450	eBioscience	RRID:AB_10359737
Streptavidin PE	Biolegend	RRID:AB_2571915
Rabbit anti-active caspase 3 BUV395	BD Biosciences	Cat. 564095
goat α-c-kit antibody	R&D Systems	Cat. BAF1356; RRID:AB_355961
rabbit α-dsRed/tdTomato antibody	Takara	Cat. 632496
mouse α-FITC Alexa 488 antibody	Jackson ImmunoResearch	Cat. 200-542-037; RRID:AB_2339038
F(ab’)2 Fragment Donkey Anti-Rabbit IgG (H+L) Dylight 405	Jackson ImmunoResearch	Cat. 711-476-152; RRID:AB_2632566
donkey α-goat IgG (H+L) Alexa 555	Thermo Fisher Scientific	Cat. A21432; RRID:AB_2535853
Chemicals, Peptides, and Recombinant Proteins
HI bovine serum (FBS)	Thermo Fisher Scientific	Cat. 26170043
Hank’s Balanced Salt Solution (HBSS)	Thermo Fisher Scientific	Cat. MT21022CV
Anti-APC magnetic microbeads	Miltenyi Biotech	Cat. 130-090-855
Thiazole Orange	Sigma-Aldrich	Cat. 390062
DNase I	Sigma-Aldrich	Cat. 10104159001
LiberaseDL	Sigma-Aldrich	Cat. 5466202001
5-bromo-2′-deoxyuridine	Sigma-Aldrich	Cat. B5002
NP-40	Sigma-Aldrich	Cat. I8896
Donkey serum	Jackson ImmunoResearch	Cat. 017-000-121
DMSO	Sigma-Aldrich	Cat. D4540
Trizol	Invitrogen	Cat. 15596018
Methocult GM M3434	Stemcell Technology	Cat. 03434
Methocult M3630	Stemcell Technology	Cat. 03630
Methocult GF M3534	Stemcell Technology	Cat. 03534
Recombinant murine thrombopoietin (TPO)	Peprotech	Cat. 315–14
Recombinant murine SCF	Peprotech	Cat. 250–03
Recombinant murine GM-CSF	Peprotech	Cat. 315–03
Recombinant murine Flt-3 ligand	Peprotech	Cat. 250–31
Tamoxifen	Sigma-Aldrich	Cat. T5648
Corn oil	Sigma-Aldrich	Cat. C8267
Benzyl alcohol	Sigma-Aldrich	Cat. 305197
Benzyl benzoate	Sigma-Aldrich	Cat. B6630
Critical Commercial Assays
BD APC BrdU Flow Kit	BD Biosciences	Cat. 552598
BD Fixation/Permeabilization Solution Kit	BD Biosciences	Cat. 554714
Mouse Erythropoietin Quantikine ELISA Kit	R&D Systems	Cat. MEP00B
High-Capacity cDNA Reverse Transcription kit	Applied Biosystem	Cat. 4368814
RNeasy Micro Kit	QIAGEN	Cat. 74004
SMARTer Stranded Total RNA-Seq Kit v2 - Pico Input Mammalian	Takara	Cat. 634413
Deposited Data
RNaseq analysis	this paper	GEO: GSE122468
Experimental Models: Organisms/Strains
ScfFL (Kitltm2.1Sjm/J, RRID:IMSR_JAX:017861)	Ding et al., 2012	N/A
ScfGFP (Kitltm1.1Sjm/J, RRID:IMSR_JAX:017860)	Ding et al., 2012	N/A
ScfΔ (Kitltm2.2Sjm/J, RRID:MGI:5306348)	Ding et al., 2012	N/A
Tie2-Cre (B6.Cg-Tg(Tek-cre)1Ywa/J, RRID:IMSR_JAX:008863	Kisanuki et al., 2001	N/A
LepR-Cre (B6.129(Cg)-Leprtm2(cre)Rck/J, RRID:IMSR_JAX:008320)	DeFalco et al., 2001	N/A
Rs26-tdTomato (B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato) Hze/J, RRID:IMSR_JAX:007909)	Jackson Laboratory	JAX:007909
Cdh5-CreER (B6.CBA-Tg(Cdh5-cre/ERT2)1Rha, RRID: IMSR_TAC:13073)	Sörensen et al., 2009	N/A
Oligonucleotides
qRT-PCR and genotyping primers	IDT	Table S2
Software and Algorithms
FlowJo v.10.4	Flowjo, LLC	N/A
Imaris Bitplane v.8.2	Bitplane	N/A
Graphpad Prism v.6	Graphpad	N/A

Highlights.

• c-kit⁺-restricted progenitors in bone marrow require SCF from leptin receptor⁺ cells

• c-kit⁺-restricted progenitors in bone marrow do not require SCF from endothelial cells

• Erythropoiesis in bone marrow requires SCF from leptin receptor⁺ niche cells

• Early erythroid progenitors localize adjacent to perisinusoidal leptin receptor⁺ cells