Low Level of c-Kit Expression Marks Deeply Quiescent Murine Hematopoietic Stem Cells

Abstract

Although c-kit is expressed highly on murine hematopoietic stem cells (HSCs) and essential for bone marrow (BM) hematopoiesis, the significance of the high level of expression of c-kit on HSCs was not well determined. We show here that CD150⁺CD48⁻Lineage⁻Sca-1⁺c-kit⁺ HSCs in adult BM are distributed within the range of roughly a 20-fold difference in the expression level of c-kit, and that c-kit density correlates with the cycling status of the HSC population. This predisposition is more evident in the BM of mice older than 30 weeks. The HSCs in G₀ phase express a lower level of c-kit both on the cell surface and inside the cells, which cannot be explained by ligand receptor binding and internalization. It is more likely that the low level of c-kit expression is a unique property of HSCs in G₀. Despite functional differences in the c-kit gradient, the HSCs are uniformly hypoxic and accessible to blood perfusion. Therefore, our data indicate the possibility that the hypoxic state of the HSCs is actively regulated, rather than them being passively hypoxic through a simple anatomical isolation from the circulation. STEM CELLS 2011;29:1783–1791

INTRODUCTION

A fine-tuned self-renewal activity and the continuous generation of all types of blood cells throughout life are the unique properties of hematopoietic stem cells (HSCs). The activity of HSCs is regulated through both environmental cues and intrinsic changes. Stem cell factor (SCF; also known as steel factor, mast cell growth factor, and kit-ligand) is one of the main extrinsic factors that directly target HSCs through the specific receptor, c-kit, and regulate the fate of HSCs [1].

Murine HSCs express c-kit and most of the HSC activity is identified in the highly positive c-kit fraction in steady-state adult bone marrow (BM) [2]. In vitro, it was found that ligand receptor interactions promote the survival of HSCs and, synergistically with other cytokines, their proliferation [3]. However, the function of c-kit is less well-characterized, in vivo.

Although HSCs expand in fetal liver in the absence of SCF [4], c-kit mutant HSCs are defective in expansion in transplanted recipients [5–7] and in maintenance in steady-state adult BM [8]. The cycling of HSCs is also enhanced in c-kit mutated animals [8], which may be the cause of HSC exhaustion [9]. Therefore, uncompromised c-kit signaling is required for the proper regulation of HSCs in vivo. However, in the steady-state, HSCs are basically devoid of signaling from c-kit [10] and how c-kit regulates hematopoiesis is unclear. HSCs are assumed to reside in the BM microenvironment or niches and c-kit binding with membrane-associated SCF is likely to be a key regulatory mechanism for keeping HSCs in niches [11]. Administration of a neutralizing antibody against c-kit eliminates HSCs from niches and induces cytopenia [12, 13]. In steady-state BM, HSCs are quiescent or slowly cycling [14–16], indicating that the majority of HSCs are not driven to proliferate by their binding to membrane-associated SCF.

Although HSCs are highly enriched in the highly positive c-kit fraction, there still is diversity in the level of c-kit expression. HSCs with different levels of c-kit expression might behave differently, as c-kit signaling requires lipid raft clustering to form the receptor dimerization [10, 17]. Furthermore, the chance to generate dimers would be low if the receptors were scarce. Alternatively, the diversity could be an outcome of their response to SCF, as c-kit receptors bound to SCF are internalized and not recycled to the surface of a cell [18].

In this study to explore the significance of highly expressed c-kit in HSCs, we subfractionated phenotypically defined HSCs based on the level of c-kit expression and compared the activity of the cells in in vitro and in vivo. We conclude that HSCs with lower expression levels of c-kit are more dormant and sustain an uncompromised potential as HSCs. The relationship between HSC dormancy and niche properties was also investigated.

MATERIALS AND METHODS

Mice

C57BL/6 and C57BL/6-Tg(CAG-EGFP) mice were purchased from SHIMIZU Laboratory Supplies (Kyoto, Japan). Unless otherwise mentioned, the mice used were 10–15 weeks of age. All animal experiments were approved by the animal experiment committee at Kansai Medical University.

Hematopoietic Growth Factors

Recombinant human (rh) granulocyte colony-stimulating factor (G-CSF) and rh thrombopoietin (TPO) were generous gifts from Kyowa Hakko Kirin Co. (Tokyo, Japan). Recombinant murine (rm) SCF, rh interleukin (IL)-6, and rh soluble IL-6 receptor (sIL-6r) were purchased from Prospec (Rehovot, Israel, www.prospecbio.com).

Antibodies

The antibodies RA3-6B2 (B220), H129.19 (CD4), 53-7 (CD5), 53-6.7 (CD8α), M1/70 (CD11b, Mac1), A7R34 (CD127), RB6-8C5 (Gr1), and Ter-119 were purchased from BD Biosciences Pharmingen (San Diego, CA, www. bdbiosciences.com); 2.4G2 (CD16/32) was purchased from Beckman Coulter (Fullerton, CA, www.beckmancoulter.com); E13-161.7 (Sca-1) and TC15-12F12.2 (CD150) were purchased from BioLegend (San Diego, CA, www.biolegend.com); HM48-1 (CD48) was purchased from eBioscience (San Diego, CA, www.ebioscience.com); M-19 (Ki-67) and polyclonal goat-anti-rat were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, www.scbt.com); 11E6 (phospho-Akt1(Ser473): pAkt) was purchased from Millipore (Billerica, MA, www.millipore.com); and 2B8 (CD117, c-kit) was purchased from Biolegend and eBioscience.

Cell preparation, Flow Cytometry, and Intracellular Staining

Preparation and analyses of the BM cells were performed as reported previously [19, 20] with minor modifications. Briefly, BM cells were collected by crushing femurs and tibias in a mortar. Lineage (Lin)⁺ cells were stained on ice with antibodies against B220, CD4, CD5, CD8α, CD11b, CD127, Gr1, and Ter119. The cells were then incubated at 4°C with immunomagnetic beads conjugated with sheep-anti-rat Fc (Dynabeads, Dynal, Oslo, Norway, www.invitrogen.com/site/us/en/home/brands/Dynal.html). Following this, the Lin⁺ cells were magnetically depleted and residual Lin antibody-bound cells were visualized with a secondary goat-anti-rat antibody conjugated with phycoerythrin-Cy7. The cells were subsequently stained with anti-Sca-1-fluorescein isothiocyanate (FITC) or -pacific blue, anti-CD150-PE, anti-c-kit-APC (allophycocyanin), and anti-CD48-biotin antibodies and visualized with Streptavidin-APC-Cy7. The stained cells were analyzed on a FACSCantoII (BD Biosciences, San Jose, CA, www.bdbiosciences.com). The cells with low viability were excluded from the analyses by detecting the 7-amino-actinomycin D (7AAD) positive staining. To detect intracellular c-kit, pAkt, or Ki-67, the BM cells stained as described above were further stained for the intracellular antigens using a BD Cytofix/Cytoperm kit (BD Bioscience Pharmingen). For the cell cycle analyses, 7AAD was used after Ki-67 staining as a DNA stain.

Real-Time Reverse Transcriptase Polymerase Chain Reaction

Total RNAs were isolated from sorted c-kit^low and c-kit^high cells using RNeasy Plus Micro kit (Qiagen, Valencia, CA, www.qiagen.com). cDNA syntheses were carried out using iScript DNA synthesis kit (Bio-Rad Laboratories, Hercules, CA, www.bio-rad.com). Real-time polymerase chain reaction (PCR) was performed on Rotor-Gene Q real-time PCR cycler (Qiagen) by the use of Power SYBR Green PCR Master Mix (Applied Biosystems, Warrington, UK, www.appliedbiosystems.com). The primer sequences were listed in Supporting Information Table 1. Real-time PCR was performed three times in triplicate. Mean value from three tubes in each PCR was considered as one data. Individual data were equalized for the glyceraldehyde-3-phosphate dehydrogenase levels and mean values from three experiments using the c-kit^high cell mRNAs were adjusted to one.

Detection of Hypoxic Cells

Analyses were performed using the Hypoxyprobe-1 Plus Kit (Hypoxyprobe, Burlington, MA, www.hypoxyprobe.com) according to the manufacturer's instructions. Briefly, mice were injected with 60 mg/kg of pimonidazole and sacrificed 30–90 minutes later. The BM cells were stained for surface antigens and then fixed, permeabilized, and stained with MAb1-FITC. Pimonidazole noninjected mouse BM cells were used as a control to visualize the MAb1-FITC background signal.

Hoechst 33342 Perfusion and FACS Analyses

Basically, the experiments were performed according to the preceding studies [21, 22]. Briefly, mice were anesthetized with pentobarbital and injected with two doses (0.8 mg/25 g body weight) of Hoechst 33342 (Anaspec, Fremont, CA, www.anaspec.com), 10 and 5 minutes before harvesting BM or blood samples. The BM cells were promptly harvested and diluted with Ca²⁺ and Mg²⁺-free phosphate-buffered saline containing 2% fetal calf serum. As preliminary tests, we compared the diluents with or without verapamil and/or reserpine; however, these did not affect the readout. We subsequently excluded them from the washing and analyzing processes. After the cell surface staining step, the cells were analyzed on a FACSAria (BD Biosciences) using a 375-nm laser and 450/40 and 670LP filters for detecting the Hoechst dye. The detection of Hoechst blue fluorescence had a higher resolution than that of Hoechst red; therefore, we mainly analyzed the data based on Hoechst blue signals (Supporting Information Fig. 4).

In Vitro Culture and Analyses

The sorted BM cells were incubated at 37°C for 9–12 days in X-vivo 15 (Lonza, Walkersville, MD, www.lonza.com) containing 1% bovine serum albumin, 20 ng/ml rmSCF, 100 ng/ml rhTPO, 10 ng/ml rhG-CSF, 50 ng/ml rhIL-6, and 100 ng/ml rhsIL-6r. Following this, the cells were collected and stained with antibodies against Sca-1, c-kit, CD11b, and CD48 and then analyzed by FACSCantoII.

In Vivo Repopulation Assays

The sorted BM cells from the C57BL/6-Tg(CAG-EGFP) mice were intravenously transplanted into lethally irradiated C57BL/6 recipients together with 200,000 competitor BM cells per mouse. Peripheral blood (PB) nucleated cells were analyzed 6, 15, and 22 weeks after transplantation for green fluorescent protein (GFP), B220, CD4/8, Gr-1, and CD11b expression by FACS. At week 23, the BM cells were analyzed for GFP expression and the Lin⁻Sca-1⁺c-kit (LSK) phenotype, and half-femur equivalent BM cells from the primary recipients were transplanted into secondary recipients and their PB was analyzed 13 weeks later.

RESULTS

Distribution of the HSC Population on the Basis of c-Kit Expression Levels

HSCs are known to be highly enriched in the CD150⁺CD48⁻LSK population [23]. Most of the CD150⁺CD48⁻LSK cells strongly expressed c-kit and fell within the range of less than a 10-fold difference in the c-kit expression level by means of fluorescent intensity (FI) (Fig. 1, right-lower panel, gate A; we first established Sca-1⁺c-kit⁺ gate in Lin⁻ gated cells to define LSK cells and applied it to CD150⁺CD48⁻Lin⁻ cells). A small subpopulation with a much lower level of c-kit expression (20- to 30-fold lower than the level of the highest c-kit expression) that looked like an extension from the main population was observed (dotted gate B). We did not find any cells negative for c-kit(Sca-1⁺). To explore functional differences based on the level of c-kit expression within the HSC population, by making the cells included in the two gates the subjects under investigation, we defined the lowest 10% of c-kit expressers in the CD150⁺CD48⁻LSK population as c-kit^low cells and the highest 10% as c-kit^high cells. There was a tendency for the CD150⁺CD48⁻LSK cells in elderly mice to show a lower c-kit expression than in younger mice when using CD150⁺CD48⁻Lin⁻Sca-1⁻c-kit⁺ cells as a measure (Supporting Information Fig. 1).

Figure 1. CD150⁺CD48⁻Lin⁻Sca-1⁺c-kit hematopoietic stem cells (HSCs) showed different levels of c-kit expression. The Sca-1⁺c-kit⁺ gate in the Lin⁻ cells (right-upper panel) was applied to CD150⁺CD48⁻Lin⁻ cells (right-lower panel, gate A). Some of the Sca-1⁺ cells were lower for c-kit expression (gate B). Therefore, we adopted a wider gate encompassing the two gates as an HSC gate. n = 3. Abbreviations: 7AAD, 7-amino-actinomycin D; FSC, forward scatter; SSC, side scatter.

Lower c-Kit Expressers Were Also Low in Intracellular c-Kit Expression and Did Not Contain Detectable Levels of pAkt

Expression of the c-kit receptor on cell surfaces is known to decline via internalization when the cells are incubated with SCF. The internalized receptors are degraded and not recycled to the cell surfaces [18]. If a lower level of c-kit expression is a consequence of c-kit receptor internalization, the c-kit^low cells should harbor a relatively higher amount of intracellular c-kit and c-kit^high cells harbors a lower amount. To clarify whether or not this was the case, we first stained c-kit on the surface and then inside CD150⁺CD48⁻Lin⁻Sca-1⁺ cells using the same clone (2B8) of the antibody against the extracellular domain of c-kit, conjugated with different fluorochromes. Contrary to the hypothesis, the c-kit^low cells were low in intracellular c-kit and the higher expression on the cell surfaces correlated with the higher concentration inside the cells (Fig. 2A). Degradation did not seem to be the primary reason for the low intracellular expression level of c-kit in c-kit^low cells because stimulation with SCF downregulated the receptor on the cell surfaces, but it was easily detectable inside the cells in vitro (Fig. 2B). Thus, internalization was not the major cause of the low c-kit expression level. c-kit mRNA levels detected by real-time reverse transcriptase-PCR did not differ between the c-kit^low and c-kit^high cells. Therefore, transcriptional regulation seemed not to be decisive for the level of c-kit, as well (Fig. 2C).

Figure 2. The c-kit^low hematopoietic stem cells were also low in intracellular c-kit expression. (A): Freshly isolated bone marrow cells were analyzed for the CD150⁺CD48⁻Lin⁻Sca-1⁺c-kit phenotype and intracellular c-kit expression. The same clone (2B8) of the antibody against c-kit, but conjugated with different fluorochromes (APC and Alexa Fluor 488), was used to detect surface and intracellular c-kit expression separately. n = 3. (B): Sorted CD150⁺CD48⁻Lin⁻Sca-1⁺ cells were incubated in a serum-free medium containing 50 ng/ml of recombinant murine stem cell factor for 5 hours at 37°C and analyzed for surface and intracellular c-kit expression. The incubation downregulated surface c-kit expression but subcellular c-kit expression did not seem to be affected. (C):c-Kit mRNA levels did not differ significantly between c-kit^high and c-kit^low cells. mRNA levels of p21 and p57 were significantly higher in c-kit^low cells compared with c-kit^high cells. *, p < .01.

The LSK cells, which include proliferating early progenitors, were found to be partially positive for a possible c-kit downstream signaling molecule, phosphorylated Akt (pAkt) (Fig. 3, left panel). In contrast, no pAkt was detected in c-kit^low cells, whereas c-kit^high cells were slightly positive (Fig. 3, center panel). Although the presence of pAkt did not necessarily mean that the c-kit downstream signal was mediated, it did indicate a functional difference between c-kit^low and c-kit^high cells. We also investigated pSTAT5 expression as a candidate signaling intermediate downstream of c-kit. However, the phosphorylated protein was not detected in any of the freshly isolated CD150⁺CD48⁻LSK cells (data not shown).

Figure 3. CD150⁺CD48⁻Lin⁻Sca-1⁺c-kit (LSK) hematopoietic stem cells were partially positive for pAkt. Freshly isolated bone marrow cells were analyzed for LSK (left panel) and CD150⁺CD48⁻LSK (center panel) phenotypes with intracellular pAkt. The LSK cells were partially positive, but only some of the c-kit^high cells expressed pAkt. A, c-kit^high cells; B, c-kit^low cells. n = 3.

The c-Kit^Low Cells Were Enriched with Cells in G₀

It has been suggested that the phosphatidylinositol 3-kinase–Akt signaling pathway takes part in regulating HSC quiescence [10, 24]. Therefore, we investigated the cell cycle status of these populations. The c-kit^low cells were found to be in G₀ more frequently than c-kit^high cells (Fig. 4; Table 1). Alternatively, the cells in G₀ had a lower c-kit expression than those in G₁ or SG₂M. This tendency did not change with age, although the HSC population was found relatively more frequently in G₁ and SG₂M phases in younger mice and more frequently in G₀ in older mice (Supporting Information Fig. 2). mRNA analysis indicated that the expression of cell cycle regulators p21 and p57 mRNA was higher in the c-kit^low cells (Fig. 2C), which is consistent with the higher frequency of G₀ cells in the population.

Figure 4. c-kit^low hematopoietic stem cells were enriched with G₀ cells. Freshly isolated bone marrow cells were analyzed for the CD150⁺CD48⁻Lin⁻Sca-1⁺c-kit phenotype and intracellular Ki-67 and DNA contents (7-amino-actinomycin D staining). The gating strategy for G₀, G₁, and SG₂M is illustrated at the lower left of the figure. Representative FACS plots indicate that c-kit^low cells were more enriched with G₀ cells than c-kit^high cells (Table 1). n = 3. Abbreviations: 7AAD, 7-amino-actinomycin D.

Table 1. Cell cycle profiles of c-kit^low and c-kit^high cells

Neither Hypoxia Nor Low Blood Perfusion Distinguished c-Kit^Low from c-Kit^High Cells

HSCs are thought to reside in discrete locations of the hematopoietic microenvironment called niches. These niches are assumed to be low in the partial pressure of oxygen. Kubota et al. [25] reported that the slowest cycling hematopoietic cells were to be found in the sinusoidal hypoxic zone close to the bone surface and distant from the central capillary-rich area of the BM cavity. These cells were reported to be low in c-kit expression, thus they share slow dividing and low c-kit expressing properties with the c-kit^low cells. Therefore, we examined whether or not the c-kit^low cells were in a hypoxic state. Pimonidazole is a nitroimidazole that forms stable adducts with thiol only in hypoxic cells. The BM cells were isolated from mice injected with pimonidazole and were investigated for hypoxia by identifying the cells bound with pimonidazole. Whereas the LSK cells were approximately 80% positive for pimonidazole, the CD150⁺CD48⁻LSK cells were almost 100% positive, which fits with the notion that HSCs reside in hypoxic niches. Therefore, the c-kit^low cells were in a hypoxic state; however, the c-kit^high cells were also in hypoxia and there was no correlation between the level of c-kit expression and the hypoxic status (Fig. 5A).

Figure 5. The level of c-kit expression in hematopoietic stem cells did not correlate with the hypoxia or blood perfusion status in bone marrow (BM). Both c-kit^low and c-kit^high cells were positive for pimonidazole and were distributed throughout the Hoechst low perfusion area. (A): Pimonidazole was injected 90 minutes before harvesting the BM. Cells in hypoxia were detected as pimonidazole positive. i, c-kit^high cells; ii, c-kit^low cells. n = 3. (B): Hoechst 33342 was injected 10 and 5 minutes before harvesting the BM. Isolated cells were analyzed for the CD150⁺CD48⁻Lin⁻Sca-1⁺c-kit phenotype and Hoechst 33342 staining. n = 3. Abbreviations: LSK, Lin⁻Sca-1⁺c-kit.

Currently, it has been proposed that there are at least two types of niches existing in adult BM. Endosteal osteoblastic niches are located close to the endosteum and far from the central sinus; therefore, their accessibility to blood permeation, which supplies oxygen, is supposed to be limited. In contrast, vascular niches are adjacent to sinusoids comprised of a fenestrated endothelium [26, 27]. To understand the accessibility of HSCs to blood permeation from the BM vasculature, mice were injected with Hoechst 33342, and the Hoechst-perfused BM cells were investigated. Some of the LSK cells were Hoechst bright but the majority was Hoechst low. Compared with the LSK cells, the CD150⁺CD48⁻LSK cells had a lower level of Hoechst staining, which is in agreement with previous studies showing that HSCs were enriched in the least-perfused fraction of BM cells (Supporting Information Fig. 3) [21, 22]. However, there was no clear difference between the c-kit^low and c-kit^high cells (Fig. 5B).

The Lowest c-Kit Expressers Could Proliferate and Express the In Vitro HSC Phenotype

Although most of the CD150⁺CD48⁻LSK cells were included in the range of less than a 10-fold difference in the FI for c-kit, there was small number of cells expressing a much lower level of c-kit. To investigate whether or not these cells were a continuum of the CD150⁺CD48⁻LSK population, we sorted the Lin⁻Sca-1⁺c-kit^lowest (LSK^lowest) cells (Supporting Information Fig. 4), the gate for LSK^lowest overlaps with the lower part of gate B in Fig. 1 and the highest FI value of c-kit for the LSK^lowest gate was set to be lower for more than 2,000 when compared with the lowest FI value for the LSK gate) and placed them into a liquid culture. Although most of the cells disappeared during the 9–12 days of the culture period, a few growing aggregates of cells were observed out of 1,000–2,000 cells. The culture-generated cells from the LSK^lowest cells contained CD48⁻CD11b^low/−Sca-1⁺c-kit⁺ cells, which should include HSCs as defined in vitro [28]. The level of c-kit expression was equivalent to that of the same type of cells generated from the LSK cells (Supporting Information Fig. 4).

The Lowest c-Kit Expressing Cells Generated Highly Positive c-Kit LSK Cells and Long-Tem Reconstituted Myeloablated Recipients

To investigate whether or not the lowest c-kit expressers could generate highly positive c-kit HSCs and long-term reconstitute hematopoiesis, we sorted the LSK, LSK^lowest, and LSK⁻ cells and transplanted them with competitor BM cells into lethally irradiated mice (Fig. 6A). After transplantation for 22 weeks, the PB of the recipients was analyzed for donor-type lymphomyeloid cells (Fig. 6B, 6C). The level of reconstitution by 9,000 LSK^lowest cells was comparable with that by 500 LSK cells. Without gating through CD150⁺CD48⁻, most of the LSK^lowest cells showed neither clonogenic (Supporting Information Fig. 4) nor HSC activity (Fig. 6B, 6C). On the other hand, 50,000 LSK⁻ cells did not show any sign of reconstitution (Fig. 6B). The BM from the primary recipients of LSK^lowest cells was observed to contain donor-type LSK cells with a high level of c-kit expression (Fig. 6D). Secondary BM recipients also exhibited donor-type lymphomyeloid reconstitution in their PB (Fig. 6C). Therefore, the LSK^lowest cells contained HSCs and generated highly positive c-kit LSK HSCs in vivo.

Figure 6. The Lin⁻Sca-1⁺c-kit^lowest (LSK^lowest) cells could generate highly c-kit positive LSK cells in vivo. (A): The GFP⁺ BM cells were sorted as (a) 500 LSK, (b) 9,000 LSK^lowest, and (c) 50,000 Lin⁻Sca-1⁺c-kit⁻ cells and transplanted as illustrated. (B): Peripheral blood was analyzed for donor (GFP⁺) reconstitution in individual recipients at the indicated time points. (C): The lineage distribution (mean ± SD) in donor (GFP⁺) cells is shown in the panel. The data shown are from one of two similarly designed experiments with similar results. (D): The LSK (a) and the LSK^lowest cells (b) generated highly c-kit positive LSK cells in the BM of the recipients, 23 weeks later. Abbreviations: BM, bone marrow; BMT, bone marrow transplantation; GFP, green fluorescent protein; LSK, Lin⁻Sca-1⁺c-kit.

DISCUSSION

SCF and c-kit are the major regulators of HSC survival and proliferation in vitro [3, 29] and they are essential for BM hematopoiesis in vivo because c-kit mutant w/w mice and SCF mutant sl/sl mice were found to be perinatally lethal due to hematopoietic failure [30–32]. However, because of this lethality, the function of c-kit was less well-characterized in adult mice. As most HSCs strongly express c-kit [2], we focused on investigating the significance of this high level of expression of c-kit.

The c-kit expression varied within the CD150⁺CD48⁻LSK population. Although most of the cells were highly positive for c-kit, a small subpopulation with a lower level of c-kit expression was also observed (Fig. 1). As the LSK^lowest cells were able to long-term reconstitute hematopoiesis in transplanted recipients (Fig. 6), it is likely that HSCs were distributed from the highest to a roughly 20-fold lower level of c-kit by FI. It should be mentioned that because of the rapid advances in FACS technologies in the last decade or two, we were able to identify and separate small differences in the expression level of cell surface antigens. Therefore, previously defined c-kit bright cells by Orlic et al. [2], for example, might partially overlap with our c-kit^low and lowest populations. We could find neither c-kit⁻ (Sca-1⁺) cells in the CD150⁺CD48⁻Lin⁻ gate (Fig. 1) nor long-term repopulating cells in LSK⁻ cells (Fig. 6). However, this does not necessarily rule out the possibility of existence of very rare HSCs with undetectable c-kit levels [33]. In addition, we did not try manipulating the c-kit⁻ cells in vitro because we wanted to investigate freshly isolated HSCs and their in vivo status [34, 35].

When SCF is bound to c-kit, the complex is internalized and finally degraded [18]. If this is the mechanism that regulates the level of c-kit on the surface of HSCs, c-kit^low cells should contain intracellular c-kit more abundantly than c-kit^high cells. However, c-kit^low cells were found to contain a lower amount of intracellular c-kit and c-kit^high cells contain a higher amount (Fig. 2). This indicated that internalization was not the major mechanism for the regulation of c-kit expression. Rather, the different levels of c-kit expression seemed to reflect functional differences in the HSCs. Previous studies indicated that a negative cell cycle regulator transforming growth factor β (TGF-β) [36, 37] downregulated cell surface c-kit expression through a decrease in c-kit message stability [38]. This might explain why the c-kit^high cells were shifted to G₁ and SG₂M, whereas the c-kit^low cells accumulated with G₀ cells (Fig. 4). Indeed, freshly isolated CD34⁻LSK cells were reported to be positive for phosphorylated-smad2/3, which are TGF-β down-stream signaling molecules [39]. In this study, we did not observe differences in the c-kit mRNA expression levels between the c-kit^low and c-kit^high cells. Message degradation might not be observed without inhibiting mRNA neo-synthesis. However, the levels of p21 and p57 mRNA expression were higher, in the c-kit^low cells (Fig. 2C). Therefore, factors such as TGF-β might be the candidates responsible for maintaining the c-kit^low population in quiescent in vivo. It was recently reported that highly dormant bromodeoxyuridine label-retaining cells (LRCs) express low levels of c-kit [25]. These cells resemble the c-kit^low cells because the LRCs were reported to be dormant and in hypoxia, in addition to the low level of c-kit expression. Although the identity of the two populations is currently unknown, altogether these results indicate that low c-kit expression is a unique property of highly dormant HSCs. We observed that dormant HSCs greatly accumulated in the c-kit^low fraction in middle aged (7–16 months) mice (Supporting Information Fig. 2). Wilson et al. [40] showed that LRCs in middle aged mice were highly enriched with G₀ cells, whereas non-LRCs were shifted more toward G₁ and SG₂M phases. It would be interesting to discover to what extent the c-kit^low cells overlap with the LRCs.

Because CD150⁺CD48⁻LSK cells are highly enriched with HSCs [23], and c-kit^low cells are mostly in G₀ and lack evidence of c-kit down-stream signaling, it may be reasonable to hypothesize that c-kit^low cells reside in BM endosteal niches. Endosteal niches are supposed to be hypoxic and distant from vasculature, whereas vascular niches are more accessible to blood oxygen [26, 27]. However, we found that both c-kit^low cells and c-kit^high cells became only slightly stained with Hoechst 33342 10 minutes after its administration (Fig. 5B). Therefore, CD150⁺CD48⁻LSK cells, regardless of the level of c-kit, were in a low but not a negative perfusion area. The methodology may have been limited in its sensitivity for distinguishing c-kit^low from c-kit^high cells, but it was still sensitive enough to demonstrate that both populations became readily perfused. This may not be surprising, considering that the administration of AMD3100 or IL-8/CXCL8 can mobilize HSCs to PB in 15 minutes or less [41, 42]. The state of hypoxia was investigated by injecting pimonidazole into mice and analyzing the BM cells with the pimonidazole-specific antibody MAb1. As almost all of the CD150⁺CD48⁻LSK cells were recognized by MAb1 (i.e., hypoxic), both c-kit^low and c-kit^high cells were in hypoxia and indistinguishable from each other (Fig. 5A). The level of c-kit expression correlated with the cell cycle status (Fig. 4). Therefore, regardless of being dormant or active, the HSCs were in hypoxia but still accessible to blood permeation, suggesting that an active regulation of hypoxia existed. Our data did not discriminate the G₀ cell-enriched and hypoxic endosteal niches from blood accessible vascular niches [26]. Spatial information in the BM of the ckit^low and c-kit^high cells needs to be presented in combination with our 7-color FACS data to solve this ambiguity. The possibility remains that endosteal and vascular niches are in close proximity to each other or even overlapping, which was recently proven to be the case in transplantation models [43].

CONCLUSION

We have shown here that the expression level of c-kit correlated with the cycling status of HSCs and that c-kit^low cells were enriched with cells in G₀. Our data indicated that hypoxic status and accessibility to blood perfusion of HSCs were not related to the level of c-kit expression (i.e., cell cycle status) and, therefore, did not separate the G₀ cell-enriched and hypoxic endosteal niches from blood-accessible vascular niches.

Acknowledgements

We thank Kyowa Hakko Kirin Co. (Tokyo, Japan) for generously providing recombinant human (rh) granulocyte colony-stimulating factor and rh thrombopoietin. This work was supported by JSPS KAKENHI (20591158 [to Y. Sasaki]; 19591144 and 21591251 [to Y. Sonoda]); the Strategic Research Base Development Program for Private Universities, the Science Frontier Program, and the 21st Century Center of Excellence (COE) program from the Ministry of Education, Culture, Sports, Science and Technology; a grant from the Promotion and Mutual Aid Corporation for Private Schools of Japan, a grant from Kansai Medical University (Research grant B); a grant from the Japan Leukemia Research Foundation; a grant from the Takeda Science Foundation; a grant from the Terumo Life Science Foundation (to Y. Sonoda); and grants from Mitsubishi Pharma Research Foundation (to Y. Sasaki. and Y. Sonoda).

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST

The authors indicate no potential conflicts of interest.