TGF-β as a candidate bone marrow niche signal to induce hematopoietic stem cell hibernation

Dormancy or hibernation of hematopoietic stem cells (HSCs), which is indispensable for HSC maintenance, is known to occur solely in the particular bone marrow (BM) microenvironment known as the HSC niche. Most of the HSCs are in the G₀ phase in the BM niche. However, HSCs are recruited into the cell cycle at long intervals, on average every 1 to 2 months.^1,2  Thus, the capacity to enter and to leave a hibernation-like state is one of the properties of “stemness.” The so-called stromal cells in the HSC BM niche, including osteoblasts, fibroblasts, adipocytes, and endothelial cells, produce several secreted and membrane-bound growth factors.³  Several signaling pathways have been characterized that keep HSCs in hibernation or undifferentiated states. These include the Ang-1–Tie-2 signal,⁴  the Notch ligand–Notch signal,⁵  the N-cadherin homotypic signal,⁶  and the transforming growth factor-β (TGF-β) signal.⁷  However, the precise molecular mechanisms underlying HSC hibernation remain largely elusive

Mouse BM HSCs are enriched exclusively in CD34⁻c-Kit⁺ Sca-1⁺ lineage marker–negative (Lin⁻) (CD34⁻KSL) cells, a population representing 0.004% of BM mononuclear cells, whereas CD34⁺KSL cells are progenitors with short-term repopulating capacity.⁸  We have recently reported that HSCs use the PI3K-Akt-FoxO signaling pathway to regulate their hibernation state, as does C elegans in dauer formation.⁹  Akt is inactive in the cytoplasm of freshly isolated hibernating CD34⁻KSL HSCs, and FoxOs, its downstream targets, are active in their nuclei. In contrast, Akt is active in cycling CD34⁺KSL progenitors and phosphorylated FoxOs are excluded to the cytoplasm. Of note is our discovery that lipid raft status finely tunes cytokine signal levels and regulates Akt activity. Lipid raft microdomains are cholesterol- and glycosphingolipid-enriched patches in the plasma membrane into which various functional molecules are distributed. Lipid rafts act as platforms for cellular functions that include cytokine signaling, membrane trafficking, and cytoskeleton organization.¹⁰  Because larger rafts have greater potential for concentration of transducers and for exclusion of negative regulators, lipid raft size controls signal intensity and functional outcomes. HSCs freshly isolated from the BM niche lacked lipid raft clustering (LRC). However, cytokine-induced LRC was essential for augmentation of HSC cytokine signals to levels sufficient for cell-cycle re-entry. Conversely, inhibition of LRC in HSCs attenuated cytokine signals, leading to repression of Akt followed by sustained nuclear accumulation of FoxOs, and induced a hibernation-like state in CD34⁻KSL HSCs ex vivo.

The FoxO subfamily of transcription factors is involved in diverse physiologic processes.¹¹  Upon activation by growth factors, the serine/threonine kinase Akt directly phosphorylates FoxO1, FoxO3, and FoxO4, resulting in their nuclear exclusion and subsequent degradation. In the absence of growth factors or in the presence of stressful stimuli, FoxOs are translocated to the nucleus and up-regulate the expression of a series of target genes, thereby promoting cell-cycle arrest, stress resistance, or apoptosis.¹²  Mice that were conditionally deleted of FoxO1, FoxO3, and FoxO4 in adult hematopoietic system exhibited defective long-term repopulating activity that correlated with increased cell cycling and apoptosis of HSCs.¹³  Levels of reactive oxygen species (ROSs) were intriguingly increased in FoxO-deficient HSCs; in vivo treatment with the antioxidative agent N-acetyl-l-cysteine (NAC) rescued the FoxO-deficient HSC phenotype. Even in mice deficient for a single FoxO gene, FoxO3a, much milder but similar defects were observed.¹⁴  These results suggest that FoxOs play essential roles in the establishment of resistance to physiologic oxidative stress, a resistance necessary to ensure the quiescence, survival, and function of HSCs. These findings demonstrate a tight correlation between lipid raft status and Akt-FoxO signaling in the context of HSC hibernation and survival and indicate that LRC plays a key role in HSC emergence from hibernation and that LRC-inhibitory signals from the BM niche are critical in the induction and maintenance of HSC hibernation.

One of the niche signaling molecules, TGF-β, acts as a negative regulator of hematopoietic stem and progenitor cell proliferation in vitro.⁷  Upon association with TGF-β, TGF-β type II receptor (TβRII) forms a complex with TGF-β type I receptor (TβRI). Subsequently, the activated TGF-β receptor complex phosphorylates receptor-activated Smads (R-Smad2) and R-Smad3. R-Smads eventually heterodimerize with the common mediator Smad4, and the resulting complex translocates to the nucleus and recruits transcriptional cofactors to control expression of genes, including those involved in the cell cycle. It has been reported that TGF-β1–null mice and inducible TβRII knockout models develop a transplantable lethal inflammatory disorder affecting multiple organs.^15,16  However, mice deficient in the TβRI, activin receptor-like kinase 5 (ALK-5), show no defects in HSC quiescence or in maintenance of the HSC pool.¹⁷  Mice deficient for the TGF-β type II receptor have not been well characterized with respect to HSC hibernation.¹⁶  TGF-β signaling deficiency so far has not revealed any effect on HSC proliferation and differentiation in vivo. Therefore, the outcome of TGF-β signaling is believed to be context dependent in hematopoiesis and the regulation of hematopoietic stem and progenitor cells is more complicated in the BM microenvironment in vivo than is seen in liquid cultures ex vivo.

In this study, we screened candidate niche signals that inhibit lipid raft clustering and identified that TGF-β efficiently inhibits cytokine-mediated LRC. We further characterized its role in HSC hibernation.

C57BL/6 (B6-Ly5.2) mice were purchased from Japan SLC (Shizuoka, Japan). C57BL/6 mice congenic for the Ly5 locus (B6-Ly5.1) were purchased from Sankyo-Lab Service (Tsukuba, Japan). C57BL/6 Ly5.1 × Ly5.2 F1 mice were bred and maintained in the Animal Research Facility of the Institute of Medical Science, University of Tokyo. Animal care in our laboratory was in accord with the guidance of Tokyo University for animal and recombinant DNA experiments.

Mouse CD34⁻KSL HSCs and CD34⁺KSL progenitor cells were purified from BM cells of 2-month-old mice. In brief, low-density cells were isolated on Lymphoprep (1.086 g/mL; Nycomed, Oslo, Norway). The cells were stained with an antibody cocktail consisting of biotinylated anti–Gr-1, –Mac-1, –B220, -CD4, -CD8, and –Ter-119 monoclonal antibodies (Phar-Mingen, San Diego, CA). Lineage-positive cells were depleted with antibiotin MicroBeads (Miltenyi Biotec, Bergisch Gladbach, Germany). The remaining cells were further stained with fluorescein isothiocyanate (FITC)–conjugated anti-CD34, phycoerythrin (PE)–conjugated anti–Sca-1, and allophycocyanin (APC)–conjugated anti–c-Kit antibodies (PharMingen). Biotinylated antibodies were detected with streptavidin-APC Cy7 (Molecular Probes, Eugene, OR). Analysis and cell sorting were performed on a MoFlo using Summit software (Dako, Glostrup, Denmark) and results were analyzed with FlowJo software (TreeStar, Ashland, OR).¹⁸ 

The markers and antibodies used were the DNA marker 4,6-diamidino-2-phenylindole (DAPI), Alexa-488–conjugated cholera toxin B subunit (CTxB), Alexa-647–conjugated goat anti–rabbit IgG, goat anti–mouse IgG, and Alexa-488–conjugated goat anti–rabbit IgG (Molecular Probes, Carlsbad, CA), rabbit anti–phospho-Akt and rabbit anti-FOXO3a (Upstate Cell Signaling, Charlottesville, VA), rabbit anti-p57 (Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti–phospho-Smad2/3 (CHEMICON, Temecula, CA), and rabbit anti–phospho-Src (Y418; Biosource, Camarillo, CA). Individual CD34⁻ KSL cells were sorted into a serum-free culture-medium drop supplemented with 50 ng/mL mouse SCF and/or 50 ng/mL human TPO on slide glasses. The sorted cells were incubated at 37°C for the indicated time periods. After fixation with 2% paraformaldehyde and blocking in 10% goat serum for 1 hour at room temperature, cells were incubated with a primary antibody for 12 hours at 4°C. The cells were then washed and were incubated with a secondary antibody for 30 minutes at room temperature. Immunofluorescence was observed with a Leica TCS SP2 AOBS confocal microscope (Wetzlar, Germany) or with an Olympus Laser Scanning Cytometer 2 (LSC2; Tokyo, Japan).

CD34⁻KSL cells were clonally deposited into 96-well microtiter plates containing 200 μL S-Clone SF-03 (Sanko Junyaku, Tokyo, Japan) supplemented with 5 × 10⁻⁵ M 2-β-mercaptoethanol, 10% FCS, and the indicated cytokines (20 ng/mL mouse SCF, 50 ng/mL human TPO, 20 ng/mL mouse IL-3, and 2 U/mL human EPO) in the presence or absence of 5 ng/mL human TGF-β1, TGF-β2, TGF-β3, latent TGF-β1, Activin-A, and Nodal (R&D Systems, Minneapolis, MN). Survival and cell division of HSCs were monitored by microscopy. To allow colony formation, single HSCs were cultured in the presence of SCF, TPO, IL-3, EPO, and anti–TGF-β blocking antibody (R&D Systems) for 11 days. Colonies were recovered, cytospun onto glass slides, and subjected to May-Grünwald-Giemsa staining for morphologic examination.

Competitive repopulation assays were performed using the Ly5 system. In brief, single cultured HSCs or pooled single cultured HSCs (B6-Ly5.1) were mixed with 2 × 10⁵ BM competitor cells (B6-F1) and were transplanted into B6-Ly5.2 mice irradiated at a dose of 9.5 Gy. After transplantation, peripheral blood cells of the recipients were stained with biotinylated anti-Ly5.1 (A20) and FITC-conjugated anti-Ly5.2. The cells were simultaneously stained with PE-Cy7–conjugated anti-B220 antibody, a mixture of APC-conjugated anti–Mac-1 and –Gr-1 antibodies, or a mixture of PE-conjugated anti-CD4 and -CD8 antibodies (PharMingen). Biotinylated antibody was developed with streptavidin Alexa-594 (Molecular Probes, Carlsbad, CA). The cells were analyzed on a fluorescence-activated cell sorting (FACS) Vantage (BD. Franklin Lakes, NJ). Percentage chimerism was calculated as (percentage Ly5.1 cells) × 100/(percentage Ly5.1 cells + percentage F1 cells). When percentage chimerism of donor-derived cells was more than 1.0 (summed over myeloid, B-lymphoid, and T-lymphoid lineages), recipient mice were considered to be multilineage reconstituted (positive mice).

Semiquantitative RT-PCR was carried out using normalized cDNA and quantitative PCR with TaqMan rodent GAPDH control reagent (Perkin-Elmer Applied Biosystems, Foster City, CA) as previously described.¹⁹  Cycling parameters were as follows: denaturation at 95°C for 15 seconds, annealing at 58°C for 15 seconds, and extension at 72°C for 30 seconds. Amplification proceeded for 38 cycles. PCR products were separated on 1.4% agarose gels and visualized by ethidium bromide staining.

Whereas HSCs are exposed to a variety of secreted and membrane-bound growth factors in the BM,³  lipid rafts are diffusely distributed on freshly isolated CD34⁻KSL HSCs but are highly polarized or clustered on CD34⁺KSL progenitor cells.⁹  This suggests that LRC is tightly inhibited on HSCs in the BM niche. We hypothesized that some signals from the niche act against LRC and keep HSCs in hibernation. Several niche signals are associated with HSC dormancy. These include the Ang-1–Tie-2 signal, the Notch ligand–Notch signal, the N-cadherin homotypic signal, the TGF-β signal, and so on. We tested the effects of these signals on LRC in HSCs.

Lipid raft distribution was assessed using cholera toxin subunit B (CTxB) to label endogenous GM1 ganglioside, a component of lipid rafts. Freshly isolated CD34⁻KSL HSCs were incubated with Ang-1, TGF-β1, or the Notch ligand Jagged-1 for 1 hour, and were then stimulated with SCF and TPO, cytokines supportive of HSC self-renewal and survival, for 30 minutes. As reported, stimulation of HSCs by cytokines induced LRC. Jagged-1 had no antagonistic effect at all on LRC. Ang-1 partially inhibited LRC, but not enough to inhibit activation of the Akt-FoxO pathway and to induce cell- cycle arrest (data not shown). In contrast, TGF-β strikingly inhibited SCF- and TPO-induced LRC (Figure 1A). TGF-β1 also inhibited Akt activation and caused sustained nuclear accumulation of FOXO3a (Figure 1A). Lipid raft protein components include transmembrane antigens/receptors, GPI-anchored proteins, cytoskeletal proteins, Src-family kinases, G-proteins, and other proteins involved in signal transduction. Treatment of HSCs with PP2, an inhibitor specific for Src-family kinases, inhibited LRC (Figure S1, available on the Blood website; see the Supplemental Materials link at the top of the online article). Intriguingly, TGF-β1 similarly suppressed the cytokine-mediated activation of c-Src tyrosine kinase (Figure 1B). Hibernating HSCs express high levels of cytoplasmic cyclin D1 and p57 cyclin–dependent kinase inhibitor (CDKI). Cytokine stimulation induces nuclear translocation of cyclin D1 and disappearance of p57, which is supposedly due to rapid protein degradation.⁹  TGF-β1 inhibited nuclear translocation of cyclin D1 (data not shown) and kept p57 in the cytoplasm (Figure 1C). These effects of TGF-β1 on HSCs were comparable with that of β-cyclodextrin (MβCD), which inhibits LRC by depleting plasma membrane cholesterol.⁹ 

We then asked whether TGF-β induces quiescence in HSCs. Single CD34⁻KSL HSCs were cultured in the presence of SCF, TPO, and TGF-β1. In the presence of SCF and TPO, more than 85% of HSCs proliferated robustly, but in the absence of SCF or TPO, none survived more than 24 hours (data not shown). In contrast, addition of TGF-β1 in culture strongly suppressed colony formation of single HSCs in a dose-dependent manner (Figure S2). Detailed observation revealed that addition of 5 ng/mL TGF-β1 in culture strongly suppressed division of single HSCs that, however, remained alive. During 5-day culture, 57% of single HSCs stayed dormant, that is, persisted as living single cells, and 22% of single HSCs divided only once (Figure 2A). Similar results were obtained when HSCs were cultured under another HSC-supporting cytokine condition, SCF plus IL-11 (Figure S3). After culture medium was changed to an optimal medium supplemented with SCF, TPO, IL-3, and EPO, 72.2% of single HSCs, which had stayed dormant for 5 days, gave rise to colonies; of these, 47% were neutrophil/macrophage/erythroblast/megakaryocyte (nmEM) colonies, derived from colony-forming units–nmEM (CFU-nmEMs) with multipotency, that is, a full range of differentiation capacity along myeloid lineages (Figure 2B). Thus, 33.9% of surviving single HSCs could retrospectively be inferred to have been CFU-nmEMs. Even after 7 days of culture, most single HSCs retained multipotency (Figure 2B). These data demonstrate that TGF-β can keep HSCs in hibernation without loss of higher-order biologic potential ex vivo. Furthermore, when we compared the activities of TGF-β1, TGF-β2, and TGF-β3 with respect to induction of the hibernating state, some HSCs survived as single cells for more than 15 days in culture in the presence of TGF-β3 (Figure S4).

To obtain direct evidence of HSC activity, we performed competitive hematopoiesis repopulation assays in vivo. We again selected single HSCs that had not divided during 5-day clonal single-cell culture in the presence of SCF, TPO, and TGF-β. Single HSCs or pools comprising 20 individual HSCs were transplanted into lethally irradiated recipient mice. As a control, freshly isolated single HSCs or pools comprising 20 individual HSCs were similarly transplanted. Comparable proportions of freshly isolated and cultured hibernating (in the presence of TGF-β) single HSCs exhibited LTR activity (26% and 20%, respectively); establishment of chimerism also was comparable (13.5% and 9.5%, respectively) (Figure 2C). In contrast, when single HSCs were cultured in the presence of SCF and TPO without TGF-β, they robustly proliferated but lost LTR activity (data not shown). All recipient mice infused with pools of 20 freshly isolated HSCs showed donor cell repopulation. So did those infused with pools of 20 cultured single HSCs, although established chimerism declined compared with that established by freshly isolated HSCs. All these data strongly support the proposition that TGF-β induces hibernation in HSCs ex vivo without affecting HSC capacity to self-renew and to differentiate into a full range of hematopoietic cell lineages.

The activated TGF-β receptor complex phosphorylates Smad2 and Smad3. Smad2 and Smad3 use Smad4 as a partner to form a transcriptionally active complex. To obtain physiologic evidence that supports active TGF-β signaling in niche HSCs, we next examined TGF-β signals in freshly isolated HSCs. Smad2/3, the signaling molecules directly downstream from and activated by TGF-β receptors, were highly phosphorylated in freshly isolated CD34⁻KSL HSCs, where they accumulated in the nucleus. In contrast, Smad2/3 were scarcely phosphorylated in CD34⁺KSL progenitor cells (Figure 3A). Quantification of the levels of Smad2/3 phosphorylation by laser scanning microscopy showed a striking contrast between freshly isolated CD34⁻KSL HSCs and CD34⁺KSL progenitors (Figure 3A). These data strongly indicate that the TGF-β signaling pathway is active in HSCs in the BM niche, but not in progenitor cells. We observed, in keeping with this, that Smad2/3 in HSCs were rapidly dephosphorylated by cytokine stimulation. Pretreatment of HSCs with TGF-β, however, again counteracted cytokine stimulation and Smad2/3 remained phosphorylated (Figure 3A).

We previously reported that CD34⁻KSL HSCs express a high level of p57^Kip2, whereas CD34⁺KSL progenitor cells do not. Of note was that p57 as well as cyclin D1, D2, and D3 localize in the cytoplasm in HSCs.⁹  That TGF-β up-regulates p57 expression in human primitive hematopoietic cells to induce cell-cycle arrest intrigued us.²⁰  To verify this finding in mouse HSCs, we stimulated freshly isolated CD34⁻KSL HSCs with SCF and TPO for 12 hours, to down-regulate p57; we then treated the cells with TGF-β. Twelve hours after the addition of TGF-β, p57 was abundantly reinduced at both mRNA and protein levels, whereas expression of p21 and p27 did not change at all (Figure 3B). These data indicate that TGF-β regulates the expression of p57, which supposedly functions as a specific CDKI that binds to and suppresses the activity of the cyclin D/CDK complexes in HSCs.

Smad2 and Smad3 are activated not only by TGF-β, but also by Activin and Nodal.²¹  We evaluated the effects of these agents on HSC cell cycle. Single CD34⁻KSL HSCs were cultured in the presence of SCF, TPO, and Activin or Nodal. TGF-β strongly suppressed division of single HSCs; 65.7% of them stayed dormant during 2-day culture. Activin-A and Nodal were not efficient in suppressing division of single HSCs; they, respectively maintained dormancy in only 6.6% and 6.9% of HSCs during 2-day culture (Figure 4). These data establish that within its family TGF-β has a major role in maintenance of HSC hibernation.

TGFβ reportedly is produced not only by niche cells, but also by HSCs themselves.⁷  As expected, HSCs expressed a significant level of TGF-β1 and a low level of TGF-β3, but not Activin A or Nodal, indicating the presence of both autocrine and paracrine regulatory loops of TGF-β signaling (Figure 5A). Importantly, however, TGF-β is produced as an inactive form, latent TGF-β. We asked whether HSCs themselves could activate latent TGF-β to establish an autocrine TGF-β signaling loop. We seeded single CD34⁻KSL HSCs in the presence of SCF and TPO along with either active-form TGF-β or latent TGF-β, and allowed the HSCs to form colonies. TGF-β strongly suppressed colony formation, whereas latent TGF-β did not affect colony formation at all. These data indicate that HSCs can produce latent TGF-β but cannot activate it by themselves. Since TGF-β is produced by a variety of cells as an inactive form, the capacity to activate latent TGF-β could be a key property of BM niche cells.

The cell-cycle status of HSCs in the niche is supposed to be precisely regulated by a specific combination of niche signals. We have reported an unexpected role of lipid raft organization in the maintenance of HSC hibernation through regulating the PI3K-Akt-FoxO pathway that lies downstream of cytokine signaling.⁹  HSCs are exposed to a variety of secreted and membrane-bound growth factors in the niche. Nonetheless, our findings clearly demonstrate that lipid raft reorganization is strictly inhibited in HSCs in the niche. We inferred that nonclustered lipid raft microdomains finely tune cytokine signals and mediate them toward suitability for HSC survival in the hibernating state, and that some niche signals inhibit lipid raft reorganization to maintain HSC hibernation. These findings support a novel model in which HSC fate, that is, hibernation or cell-cycle re-entry, largely depends on lipid raft regulation.

We have now identified that TGF-β inhibits cytokine-induced LRC (Figure 1A). TGF-β suppresses Akt activation and induces nuclear accumulation of FoxO3a in HSCs. It also inhibits translocation of cyclin D1 into the nucleus and maintains high cytoplasmic accumulations of p57 (Figure 1D and data not shown). Through these mechanisms, TGF-β strongly inhibits cell division and maintains HSCs in the hibernating state ex vivo. Together with our finding that Smad2 and Smad3, which are activated by the TGFβ receptor complex, are selectively and highly phosphorylated in CD34⁻KSL HSCs, but not in CD34⁺KSL progenitor cells, these findings strongly indicate a physiologic role for TGF-β in HSC hibernation in the niche (Figure 3A). This notion is also supported by the study of C elegans, which indicated a critical role of Daf-7 as a positive regulator of Daf-16.²²  Daf-7 is a TGF-β–like molecule. Via its receptor and downstream signaling molecules (Daf-4, Daf-1, Daf-8, and Daf-14), it up-regulates Daf-16 expression and exerts a dauer larval gene program.²³ 

TGF-β is widely expressed in BM by elements that include osteoblasts and other stromal cells. Importantly, however, TGF-β is produced as a latent form. Latent TGF-β must be processed and activated. As shown in Figure 5B, HSCs are not able to activate latent TGF-β. That the BM niche is where TGF-β can be processed/activated and where TGF-β induces HSC hibernation is thus a tempting hypothesis. In contrast, Ang-1, another regulator of HSC hibernation, was much less effective than TGF-β in inhibiting cytokine-induced LRC and subsequent Akt activation (data not shown). Recently, the TPO signal was proposed as an essential component for HSC hibernation in the osteoblastic niche.²⁴  TPO efficiently induces LRC and activates the PI3K-Akt pathway in vitro. However, its signal is supposedly attenuated by inhibited LRC in hibernating HSCs in the niche. We assume that the attenuated TPO signal by inhibitory niche signals including TGF-β acts as a survival signal but not proliferation signal on HSCs and holds the key in keeping HSCs in hibernation. These data highlight lipid raft assembly and its regulation by TGF-β as a novel regulatory component of HSC hibernation. Our findings thus indicate that HSC hibernation is regulated by at least 2 different routes, the Ang-1–Tie-2 and TGF-β–Smad signaling pathways, and establish a central role for TGF-β in regulating the lipid raft–PI3K–Akt–FOXO pathway.

Although TGF-β has been well characterized as a negative regulator of hematopoietic stem and progenitor cell proliferation in vitro,⁷  mice models deficient for TGF-β signaling molecules, including ALK-5 TβRI, show no defects in maintenance or quiescence of HSCs.¹⁷  These discrepancies may be at least partly explained by the considerably low mRNA expression of ALK-5 in HSCs compared with that in E12.5 total embryo (Figure 5A), which is indicative of alternative TβRIs in HSCs. Mice deficient for the TβRII have not been well characterized with respect to HSC hibernation because of lethal inflammatory disorder affecting multiple organs.¹⁶  Furthermore, overlapping receptor and Smad usage by different TGF-β superfamily ligands (TGF-βs, BMPs, and Activins) accounts for their functional redundancies, making their signals more complicated in vivo than is seen in liquid cultures ex vivo. Of note is that Smad4, which acts at a common level of convergence for all TGF-β superfamily signals, has recently been identified as critical for maintenance of self-renewing HSCs.²⁵  Thus, the physiological role of the TGF-β awaits further evaluation.

The negative regulation of lipid raft assembly is poorly understood. In this regard, the direct inhibition of lipid raft reorganization by TGF-β is notable. In the present study, TGF-β significantly inhibited cytokine-induced activation of c-Src, one of the lipid raft components, in HSCs (Figure 1B). This effect was comparable with that exerted by PP2, an inhibitor specific for Src-family kinases (Figure S1). Intriguingly, TGF-β has been reported to down-regulate protein expression of Src-family kinases.^26,27  Although the precise mechanism whereby TGF-β inhibits c-Src activation remains obscure, Src-family kinases could be key targets for TGF-β in affecting lipid raft reorganization.

TGF-β reportedly induces cell cycle arrest through up-regulating CDKIs, including p15INK4B (p15), p21, and p27, in various cell types.^28,29  However, TGF-β has been demonstrated to induce growth arrest in HSCs independently of p21 or p27.³⁰  We here have presented evidence that TGF-β specifically up-regulates the expression of p57, but not p15, p21, or p27, in HSCs (Figure 3B and data not shown). In accord with our findings, p57 has been identified as the only CDKI significantly up-regulated by TGF-β in vitro in human CD34⁺ primitive hematopoietic cells; of importance was the observation that knockdown of p57 expression blocked the cytostatic effect of TGF-β.²⁰  All these findings indicate the importance of TGF-β in regulating p57 expression and in maintaining HSC hibernation.

In this study, we also evaluated the role of autocrine TGF-β signaling loop in HSCs. Our data indicated that HSCs can produce latent TGF-β but cannot activate it by themselves. Latent TGF-β does not affect HSC cell growth at all. Intriguingly, a cell cycle–independent role of autocrine TGF-β has been reported on human primitive hematopoietic progenitor cells under a culture condition without supporting cells that can activate latent TGF-β.³¹  Whether the latent TGF-β can transduce alternative signals via an as-yet-unrecognized pathway in HSCs is a tempting question to be addressed.

Our findings stress the critical role of lipid rafts in regulating the cell-cycle status of HSCs and demonstrate a novel interplay between the lipid raft–PI3K–Akt–FoxO and the TGF-β–Smad signaling pathways in HSCs hibernating in the BM niche. Smad proteins activated by TGF-β could form a complex with FoxO proteins FoxO1, FoxO3a, and FoxO4.³²  TGF-β–mediated interaction between these 2 signaling pathways thus could hold a key role in the regulation of gene expression that controls HSC hibernation.

We thank Y. Morita and Y. Yamazaki for technical help and advice and Dr A. S. Knisely for critical reading of the paper.

This work was supported in part by grants from the Ministry of Education, Culture, Sport, Science and Technology, Japan, Japan Science and Technology Corporation (JST), the Naito Foundation (Tokyo, Japan), and the Terumo Lifescience Foundation (Kanagawa, Japan).

Contribution: S.Y., A.I., and H.E. designed the research and analyzed data; S.Y., A.I., and H.N. wrote the paper; K.E. contributed vital new reagents; and S.Y. and S.T. performed research and analyzed data.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Hiromitsu Nakauchi, 4-6-1 Shirokanedai, Minato-ku, Tokyo, Japan; e-mail: nakauchi@ims.u-tokyo.ac.jp.