Identification of a Stroma-Mediated Wnt/β-Catenin Signal Promoting Self-Renewal of Hematopoietic Stem Cells in the Stem Cell Niche


  • Author contributions: J.A.K. and Y.J.K.: performed research, collected data; G.P.: performed research, collected data, interpreted data; M.K.: performed research, collected data; Y.O.P.: designed research; H.K. and S.H.L.: performed research; I.S.C.: analyzed data, performed statistical analysis; J.S.L.: designed research; E.H.J.: collected data, analyzed data; I.H.O.: designed research, analyzed, and interpreted data, wrote the manuscript. J.A.K., Y.J.K., and G.P. contributed equally to this work.

  • First published online in STEM CELLSExpress February 26, 2009


With contrasting observations on the effects of β-catenin on hematopoietic stem cells (HSCs), the precise role of Wnt/β-catenin signals on HSC regulation remains unclear. Here, we show a distinct mode of Wnt/β-catenin signal that can regulate HSCs in a stroma-dependent manner. Stabilization of β-catenin in the bone marrow stromal cells promoted maintenance and self-renewal of HSCs in a contact-dependent manner, whereas direct stabilization in hematopoietic cells caused loss of HSCs. Interestingly, canonical Wnt receptors and β-catenin accumulation were predominantly enriched in the stromal rather than the hematopoietic compartment of bone marrows. Moreover, the active form of β-catenin accumulated selectively in the trabecular endosteum in “Wnt 3a-stimulated” or “irradiation-stressed,” but not in “steady-state” marrows. Notably, notch ligands were induced in Wnt/β-catenin activated bone marrow stroma and downstream notch signal activation was seen in the HSCs in contact with the activated stroma. Taken together, Wnt/β-catenin activated stroma and their cross-talk with HSCs may function as a physiologically regulated microenvironmental cue for HSC self-renewal in the stem cell niche. STEM CELLS 2009;27:1318–1329


Hematopoietic stem cells (HSCs) constitute a rare subpopulation in hematopoietic tissues that sustain the production of blood cells throughout life and can reconstitute bone marrow after myeloablation or on transplantation into myeloablated recipients. The properties of homeostatic maintenance and expansion of HSCs are dependent on their unique ability to execute self-renewing divisions [1]. Multiple intrinsic regulators of HSC function are now recognized, including key transcription factors and signaling molecules [2]. Growing evidence also now points to the microenvironment of HSCs in the bone marrow, so-called stem cell niches, as playing a crucial role in the regulation of HSC maintenance and expansion [3, 4]. However, the signals that can activate the stem cell niche in response to physiological requirements for HSC self-renewal and the downstream molecules involved remain poorly understood.

Wnts are secreted glycoproteins associated with the cell surface or extracellular matrix that influence diverse biological processes, including embryonic induction and cell fate specifications. In the “canonical” Wnt signaling pathway, Wnt binding to seven-pass transmembrane Frizzled (Fz) family receptors and single-pass coreceptors, LDL-receptor-related protein 5,6 (LRP 5,6), induces β-catenin stabilization and its entry into the nucleus where it activates T-cell factor/lymphoid enhancer binding factor (TCF/LEF) target genes. In the absence of Wnt, β-catenin is destabilized by a destruction complex composed of Axin and the serine-threonine kinase, glycogen synthase kinase 3β (GSK 3β) [5].

The expression of Wnt and its receptors in hematopoietic tissues and their effects on hematopoietic progenitor cells point to roles for Wnt in hematopoietic function [6]. Recently, studies have provided further evidence that Wnt signaling regulates self-renewal of HSCs. Purified Wnt3a protein [7] and transduction of the constitutively active form of β-catenin into HSCs of transgenic bcl-2 mice [8] caused a phenotypic increase of HSCs during culture in vitro and enhanced reconstitution levels in vivo. Similarly, administration of a GSK 3β inhibitor, which inhibits the β-catenin destabilizing complex, resulted in an enhanced and accelerated repopulation of transplanted HSCs in vivo [9]. However, in a recent study on mice with conditional inactivation of β-catenin, normal hematopoietic development and repopulating activity was observed, suggesting that β-catenin activity is dispensable for HSC function [10]. Furthermore, mice with in vivo stabilized β-catenin exhibited defective hematopoietic repopulation during steady-state or stimulated conditions in a myeloablated host, which were nevertheless accompanied by expansion of phenotypically defined HSCs and defects in differentiation [11, 12]. Thus, the precise role of β-catenin in hematopoiesis remains unclear. In particular, the possible role of Wnt/β-catenin pathways in the regulation of stem cell niches and resulting influence on HSC self-renewal has not been explored. In this study, by differentially stabilizing β-catenin in hematopoietic versus stromal cells, we identify a new pathway for Wnt/β-catenin signaling that is mediated through the stroma that may function as a physiological cue for HSC self-renewal in the stem cell niche.

Materials and Methods

Animal and Cells

C57BL/6J-Ly 5.2 (BL6) mice or C57BL/6J-Pep3b-Ly5.1 (Pep3b) mice were used as recipients or donors in congenic transplantation. Experiments were undertaken with approval from the Animal Experiment Board and the Institutional Review Board of the Catholic University of Korea. 293T cells or mouse L cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco, Grand Island, NY, supplemented with 10% fetal bovine serum (FBS). Enrichment of murine bone marrow cells (BMCs) by lineage depletion (Lin), 5-fluorouracil treatment (5-FU BMCs) or sorting for LinSca-1+c-kit+ (LSK) cells were performed as described [13]. Mesenchymal stromal cells (MSCs) were obtained from murine bone marrow by serial passage of adherent cells in medium containing 10% FBS until they became negative for CD45 [14], or by flowcytometric sorting for a CD45 population from the adherent cells.

Plasmid, Conditioned Medium, and Retroviral Vector

The TOPFLASH reporter construct containing eight TCF/LEF binding and FOPFLASH reporter containing mutated TCF/LEF binding sites were previously described [15]. The Wnt 3a conditioned medium (Wnt 3a-CM) was produced from L-cells as previously described [7] and filtered before use. The activity of Wnt 3a-CM was verified by induction and nuclear localization of β-catenin in stromal cells (supporting information Fig. S3C, S3D). A stable form of the β-catenin gene (S37A) [16] was cloned into the murine stem cell virus vector (MSCV) expressing green fluorescent protein (GFP) under the phosphoglycerate kinase (PGK) promoter (MSCV-PGK-GFP, abbreviated as MPG).

Retroviral Transduction and Ex Vivo Culture

Transduction of the retroviral vectors into hematopoietic cells was performed as previously described [13]. Briefly, cells were prestimulated for 48 hours in the serum-free medium containing BIT® (Stemcell Technology, Vancouver, Canada) in the presence of 100 ng/ml Flt-3 ligand (FL) (R&D Systems, Minneapolis, MN,, 100 ng/ml Stem Cell Factor (SCF) (R&D systems) and 50 ng/ml Thrombopoietin (TPO) (CytoLab/PeproTech, Rehovort, Israel) followed by three infections in medium containing the same cytokines (FL+SCF+TPO). MSCs expressing the stable form of β-catenin were established by infecting murine MSCs with MPG or MPG-β-catenin, followed by sorting for transduced (GFP+) cells. MSCs were irradiated (1,500 cGy) before use and cocultures with hematopoietic cells were performed for 5 days as described [14] in the presence or absence of a trans-well filter (Corning, Corning, NY, or γ-secretase inhibitor ll (Calbiochem, Darmstadt, Germany). The cell cycle of cocultured hematopoietic cells were analyzed by staining with Hoechst 33342 (10 μM) and Pyronin Y (1 μg/ml) as described [17] along with antibodies against CD45 (BD Pharmingen, San Diego, CA,[lowen]us.shtml), lineage markers (StemCell Technology), Sca-1-PEcy7 (BD Pharmingen), c-kit-APC (eBioscience).

In Vivo Repopulation and Competitive Repopulating Unit Assay

Repopulation and differentiation of HSCs in congenic mice models was performed as previously described [13]. For quantitative measurements of HSC numbers, competitive repopulating unit (CRU) assays were performed as previously described [18]. Briefly, serial dilution of cells were transplanted into lethally irradiated (900 cGy) mice with 1 × 105 helper cells, and recipient mice with 1%, or more, of donor-lymphoid and myeloid engraftments were scored as positive; 1 CRU was defined as the cell dose that resulted in negative engraftment in 37% of the test mice.

Microarray Analysis for Gene Expression in β-Catenin Transduced MSCs

Analysis of gene expression for β-catenin/stroma in comparison to MPG/stroma was performed with Illumina BeadChip array hybridization analysis. Biotin-labeled cRNA samples were prepared according to Illumina's recommended labeling procedure. Briefly, 500 ng of total RNA was used for cDNA synthesis, followed by an amplication/labeling step to synthesize biotin-labeled cRNA using the Illumina TotalPrep RNA Amplification kit (Ambion, Austin, TX). Labeled, amplified material (1,500 ng per array) was hybridized to a version 2 of the Illumina Mouse-6 BeadChip (48K) according to the manufacturer's instructions (Illumina, San Diego, CA). Arrays were developed by Amersham fluorolink streptavidin-Cy3 (GE Healthcare Bio-Sciences, Little Chalfont, U.K.) and scanned with an Illumina Bead Array Reader confocal scanner (BeadStation 500GXDW; Illumina). Array data processing and analysis was performed using Illumina BeadStudio software. Average linkage hierarchical cluster analysis was carried out using a 1-Pearson correlation as similarity metric with the use of GeneCluster/TreeView program ( One hundred and thirty-four unique genes were selected by t test (p < .005) with false discovery rate statistical confidence whose expression was induced three-fold or greater in β-catenin/stroma when compared with MPG/stroma. Gene Ontology program ( was then used to categorize genes in subgroups based on their biologic function and on their localization to extracellular regions.

Reverse Transcription Polymerase Chain Reaction and Real-Time Quantiative Polymerase Chain Reaction

Total RNAs purified from BMCs or MSCs were subjected to the reverse transcription polymerase chain reaction (RT-PCR) using the primer sets described (supporting information Table S2). For real-time quantiative polymerase chain reaction (RQ-PCR), cDNA was amplified in the MyiQ iCycler (Stratagene, Cedar Creek, TX). The threshold cycle (Ct) value for each gene was normalized to the Ct value of glyceraldehyde 3 phosphate dehydrogenase (GAPDH). The relative mRNA expression was calculated by using the formula; 2−ΔΔCt, where ΔCt = Ctsample − CtGAPDH and ΔΔCt = ΔCtsample − ΔCtreference group.

Western Blotting and Immunostaining

Western blots were performed using an anti-active-β-catenin antibody (clone 8E7; Upstate, Lake Placid, NY), an antibody against total β-catenin (Sigma, St. Louis, MI), an anti-human dll-1 antibody (AdipoGen, Seoul, Korea), an anti-murine dlk-1 antibody (clone 105B, AdipoGen) or an antibody against jagged-1 (Cell Signaling, Danvers, MA). The Annexin V assay was performed according to the manufacturer's instructions. For bone marrow immunohistochemistry, femurs in a paraffin block (5 μm) were deparaffinized and pretreated with proteinase-K for 5 minutes for antigen retrieval. Endogenous peroxidase was blocked with 0.3% hydrogen peroxide for 5 minutes. The slides were then incubated with each antibody overnight at 4°C, washed and incubated with a secondary antibody (horseradish peroxidase) for 30 minutes at room temperature and visualized with DAKO REAL EnVision Detection System (DAKO, Glostrup, Denmark, and diaminobenzidine (DAB), followed by hematoxylin counterstaining. For double staining, slides that had been visualized with DAB were rinsed with PBS/Tween-20, denatured, then stained for a second antigen by incubating with rabbit polyclonal anti-N-Cadherin (Abcam, Cambridge, U.K., or rat anti-mouse CD45 (BD Biosciences, Flanklin Lakes, NJ; 1:20) for 60 minutes. Staining for N-cadhrein was visualized with the Universal Alkaline Phosphatase Immunostaining Kit (Diagnostic Biosystem, Pleasanton, CA) and for CD45, with Rat on Mouse AP-Polymer Kit (Biocare Medical) using Vector Blue Alkaline Phosphatase Substrate Kit III (Vector Laboratory, Burlingame, CA).

Statistical Analysis

The significance of the differences between groups was analyzed using the Student's t test (p < .05).


β-Catenin Activation is Compartmentalized in the Stroma Under Physiological Conditions Stimulating Hematopoiesis

To gain insight on the possible role of the hematopoietic microenvironment in Wnt/β-catenin mediated regulation, we were first interested to see whether Wnt/β-catenin signals can target the stromal compartment of the microenvironment in the normal regulatory process of hematopoiesis. Therefore, we first examined the distribution of Wnt/β-catenin signaling molecules in the stromal and hematopoietic compartments of bone marrow by comparing expression patterns of genes in nonhematopoietic (CD45) bone marrow MSCs and hematopoietic cells. To our surprise, receptors for Wnt/β-catenin signals (Fz 1, 2, 7, 8) [19] were predominantly expressed in stromal cells rather than in hematopoietic cells, whereas canonical Wnt ligands (Wnt 1, 2b, 4, 10b) that activate β-catenin signaling [5] were predominantly expressed in the hematopoietic compartment (Fig. 1A and 1B and supporting information Fig. S1). In contrast, Wnt 5a, a noncanonical Wnt ligand related to the inhibition of canonical pathways [20], was enriched in stromal cells and the receptors Fz 4, 6 that are involved in noncanonical protein kinase-C activation [21] were enriched in 5-FU BMCs. Furthermore, a subset of genes involved in canonical Wnt-signal reception, LRP5, LRP6 [22], Ryk [23], Dickkopf2 [24], and Ror2 [25], was more highly expressed in stromal cells, whereas the Dickkopf4 [26], SFRP1, and SFRP2 genes [27] that inhibit the signaling, were enriched in 5-FU BMCs. These results reveal a compartmentalization of Wnt signaling molecules in the hematopoietic microenvironment where canonical Wnt signals preferentially localize to stromal cells.

Figure 1.

Stromal compartmentalization of Wnt/β-catenin receptor molecules and β-catenin in bone marrow. (A): Compartmentalized expression pattern of Wnt-related genes in hematopoietic and stromal cells. Reverse transcription polymerase chain reaction (RT-PCR) analysis of indicated molecules in normal BM, 5-FU BMCs, and MSCs sorted for CD45(−). (B): Real-time PCR analysis of expression levels. Relative expression levels of indicated genes in untreated normal BM, 5-FU BM, and CD45(−) MSCs were calculated by normalization to the levels in normal BM. Shown are the mean folds with SEM from three experiments. (C): Localization of β-catenin in the non-hematopoietic CD45(−) stromal component of bone marrows. Bone marrow trabecular regions were double stained with β-catenin (brown, DAB), CD45 (blue, vector blue) and hematoxylin eosin counterstain (cobalt). (D, E): Colocalization of N-cadherin and β-catenin in the endosteal stroma of bone marrow. Bone marrows were single stained with β-catenin or N-cadherin (DAB) (D), or double stained with β-catenin (DAB) and N-cadherin (vector blue) (E). Abbreviations: BM, bone marrow; 5-FU BM, 5-fluorouracil-treated bone marrow; MSCs, mesenchymal stromal cells; PC, positive control for each molecule obtained from RNA mixture of mouse brain and testis.

To further examine compartmentalization of Wnt/β-catenin signaling molecules under in vivo conditions, we looked at β-catenin accumulation in the trabecular region of bone marrows where most long-term HSCs reside and their self-renewal occurs [3, 4]. As shown in Figure 1C–1E, accumulation of β-catenin (total form) was predominantly observed in the endosteal lining of the bone marrows and colocalized with the non-hematopoietic (CD45) N-cadherin+ cell population. This indicates that the endosteal osteoblastic niche, referred to as SNO (spindle shaped, CD45N-cadherin+ osteoblast) [3, 4] may function as a primary target of Wnt/β-catenin signaling in the bone marrow microenvironment. Of note, accumulation of the active form (unphosphorylated) of β-catenin was selectively observed in the endosteal stroma of bone marrows in “stimulated” (by systemic injection of Wnt-3a) and “stressed” (by radiation) marrows, and not in homeostatic “steady-state” marrows (Fig. 2). These results reveal that β-catenin activation occurs in a compartmentalized fashion in the bone marrow microenvironment in conditions associated with stimulating HSC self-renewal suggesting that β-catenin is stabilized in the stroma during the physiological regulation of hematopoiesis.

Figure 2.

Physiological regulation of stromal β-catenin accumulation in bone marrow. (A): Endosteal accumulation of the active form of β-catenin with Wnt 3a stimulation. Mice were intravenously injected with Wnt 3a-CM or control-CM at indicated time (12 hours interval) before examination and their trabecular bone marrows were immunostained with antibody against active form (unphosphorylated) β-catenin, counterstained, and visualized by DAB (brown color). Images at low (×100) and higher (×400) magnification of the dotted inlets are shown. (B): Selective accumulation of the active form of β-catenin in the marrows under stressed condition. The trabecular region of femurs from steady-state or irradiated (5 days prior) mice were immunostained with antibody against the active form of β-catenin protein, counterstained and visualized with DAB (brown color). Shown are the representative images at low magnification (×40) and higher magnification (×200) of the dotted inlets and ×400 magnification. Note the high basal level of red blood cells in the irradiated BM. Abbreviations: BM, bone marrow; DAB, diaminobenzidine.

Distinct Effects of β-Catenin Stabilization on HSCs Depending on Their Targeting Site

On the basis of the physiological compartmentalization of β-catenin in the stromal component, we hypothesized that the regulatory effect of Wnt/β-catenin signals may be exerted through the stroma and therefore, that the effect may be distinct depending on whether Wnt/β-catenin signals are stabilized in the hematopoietic or stromal compartments. To test our hypothesis, we decided to compare the effects on HSCs of stabilizing β-catenin in stroma with those of stabilizing β-catenin in hematopoietic cells.

For this, a retroviral vector (MPG) encoding a stable form of (S37A) β-catenin [16] gene was constructed and its ability to transactivate TCF/LEF target genes was confirmed with the TOPFLASH reporter [15] (supporting information Fig. S2). Using these retroviral vectors, we first examined the effect of expressing stable form β-catenin in bone marrow MSCs in coculture with HSCs. The established bone marrow stromal cells retained the ability to support HSCs without their own hematopoietic contribution in coculture, as well as the ability to reconstitute bone marrow stroma after intrafemoral injection into irradiated mice marrow as previously shown [28] (supporting information Fig. S3A, S3B). Their ability to respond to canonical Wnt signals was also confirmed as induction and nuclear localization of β-catenin following stimulation with Wnt-3a CM were readily detectable (supporting information Fig. S3C, S3D). After transduction of these stromal cells with control vector (MPG/stroma) or β-catenin (β-catenin/stroma), the surface phenotypes of the transduced cells revealed no phenotypic difference between MPG/stroma and β-catenin/stroma cells (supporting information Fig. S4A). Although β-catenin/stroma exhibited a modest increase in chemically induced osteogenic differentiation compared to MPG/stroma, no significant alteration was seen in the composition of osteoblastic cells in the transduced cells during culture (supporting information Fig. S4B–S4G). When 5-FU BMCs were cocultured on the stromal feeder cells for 5 days, the hematopoietic cells cocultured with β-catenin/stroma cells compared to MPG/stroma exhibited a higher frequency of phenotypically defined HSCs (LinSca-1+c-kit+) and increased LSK cell expansion relative to the input cells (0.5% ± 0.04% vs. 1.2% ± 0.1% for frequency, and 2.4 ± 0.3 vs. 6.1 ± 0.8-folds expansion of input cells for MPG/stroma versus β-catenin/stroma group, respectively; p < .05; Fig. 3A, 3B). When examined for LSK cell population, no significant difference in apoptosis was seen (Fig. 3C), but higher proportions of LSK cells were in G0 state in the coculture with β-caten/stroma (52.2% ± 4.6% vs. 74.2% ± 2.7% for MPG/stroma vs. β-catenin/stroma group, respectively; p < .05; Fig. 3D, 3E). However, there was no significant difference in the S/G2/M phase of LSK cells or total cell numbers in the culture (supporting information Fig. S5), showing β-catenin/stroma enhanced the maintenance of an undifferentiated state during the mitotic division of cocultured cells. Similarly, the ex vivo maintenance of HSCs, as defined by CRU [18] was higher when 5-FU BMCs were cocultured with β-catenin/stroma than with MPG/stroma, as assessed by a two-fold higher recovery of CRUs in the limiting dilution assays performed immediately after a 5-day short-term culture, a period set to minimize accessory cell effects during the culture [17] (1/37,200 vs. 1/82,900 for MPG/stroma vs. β-catenin/stroma group, respectively; Fig. 4A). To further characterize the effects of β-catenin/stroma on the repopulating activity of HSCs, 5-FU BMCs were cocultured with transduced stroma and transplanted into irradiated recipient mice at a nonlimiting dose to assess average reconstitution levels. As shown, transplants from the β-catenin/stroma coculture displayed significantly higher levels of repopulation than those from the MPG/stroma coculture, for up to 24 weeks post-transplantation (Fig. 4B, 4C). The enhanced repopulation was not associated, however, with significant alteration in lymphomyeloid differentiation (Fig. 4D), suggesting the enhancement of repopulation occurs at the level of multilineage repopulating cells. Limiting dilution transplantation of reconstituted mice marrows into secondary recipients provided further support for this possibility, as it revealed an 18-fold higher number of regenerated CRUs in marrows that had received cells cocultured on β-catenin/stroma than in marrows receiving cells on control stroma (1,155 CRUs vs. 65 CRUs for β-catenin/stroma and control stroma, respectively; Fig. 4E). However, the enhancing effect of β-catenin on hematopoietic reconstitution was not seen when HSCs were cocultured with stroma separated by a trans-well filter, suggesting that the effect was dependent on direct cell-to-cell contact between HSCs and β-catenin activated stroma (Fig. 4F). Taken together, these results show that the stabilization of β-catenin in the stromal compartment leads to a higher maintenance of the undifferentiated state and/or increased self-renewal of HSCs in contact with the stroma.

Figure 3.

Effect of β-catenin stabilization in stromal cells on the in vitro hematopoietic activity of cocultured HSCs. (A, B): 5-Fluorouracil-treated bone marrow cells (5-FU BMCs) were cocultured on MPG/stroma or β-catenin/stroma for 5 days and analyzed for LSK cell population. Shown are the % LSK cells of the hematopoietic (CD45+) cells after coculture (A) and the ratio of absolute numbers of LSK cells in each culture relative to the initial input LSK cell numbers (B) with an error bar indicating SEM (three experiments; *, p < .05). (C): Effects of coculture on the apoptosis of LSK cells. After 4 day's coculture, CD45+LSK cells were gated and AnnexinV (+) in PI (−) cells were examined. Shown are the mean with error bar indicating SEM (n = 3). (D, E): Effect on the cell cycle of cocultured hematopoietic cells. After coculture for 4 days, cells were stained for CD45 and LSK markers, followed by staining with Hoechst 33,342 and pyronin Y. The cell cycling status (G0 or S/G2/M) was examined for gated CD45+LSK populations. Shown are the representative flow cytometry plots (D) and % of CD45+LSKcells in G0 populations or S/G2/M population (E) (two experiments, n = 5; **, p < .01). Abbreviations: HSCs, hematopoietic stem cells; LSK, LinSca-1+ c-kit+; MPG, murine stem cell virus vector (MSCV) expressing green fluorescent protein (GFP) under the phosphoglycerate kinase (PGK) promoter (MSCV-PGK-GFP); PI, propidium iodide.

Figure 4.

Effect of stromal β-catenin stabilization on the hematopoietic activity in vivo. (A): CRU frequencies of hematopoietic cells immediately after 5 days culture in vitro. 5-Fluorouracil-treated bone marrow cells (5-FU BMCs) cocultured on each stromal feeder cells were transplanted into lethally irradiated mice together with recipient marrow cells (1 × 105 cells) at serial dilution doses as indicated. Twelve weeks after transplantation, repopulation in the recipient blood was analyzed and Poisson statistics were applied on a percentage of mice with negative engraftments to determine CRU frequencies. Upper and lower limits within a 95% confidence interval (CI) of CRU frequencies are shown to represent ±2 SEM. (B, C): Effect of β-catenin-activated stroma on the repopulating activity of cocultured hematopoietic stem cells. 5-FU BMCs (Ly5.1) were cocultured on stromal cells transduced with MPG or MPG-β-catenin for 5 days. Cultured cells (equivalent to 1 × 105 input cells) were then transplanted into lethally irradiated recipient mice (Ly5.2). Shown are the representative flow cytometry profiles (B) and mean % donor-derived leukocytes (Ly5.1+) cells in the recipient blood at the time indicated after transplantation (C) (n = 4 for each). *, p < .05. (D): The myeloid (Mac-1/Gr-1) and lymphoid (B220) lineages of donor-derived reconstituted cells in (C) were analyzed and the percent of each lineage in reconstituted cells are shown with SEM. (E): Number of donor-derived CRUs regenerated in the recipient bone marrows. The reconstituted primary recipient mice in (C) were pooled and transplanted by serial dilution into secondary recipient mice. Each CRU frequency was determined 16 weeks after secondary transplantation by applying Poisson statistics and total CRU numbers were calculated assuming that the two femurs and tibiae represent 25% of total marrow [13]. Values are the mean for the total number of donor-derived (Ly5.1) CRUs per mouse with error bars representing the upper and lower limits of a 95% CI. (F): Effect of transwell filter separation during coculture. Lin(−) BMCs were plated on the transwell filter (4 × 103 cells per ml) in the coculture with each transduced stroma (MPG/non-contact (NC) or β-catenin/NC) for 5 days and transplanted into irradiated mice (transplant of 5 × 103 initial cells per mouse). Shown are the mean % donor-derived cells ± 2 SEM of donor cells in the recipient blood 16 weeks after transplantation (two experiments; n = 10 each). Abbreviations: CRU, competitive repopulating unit; MPG, murine stem cell virus vector (MSCV) expressing green fluorescent protein (GFP) under the phosphoglycerate kinase (PGK) promoter (MSCV-PGK-GFP).

We next examined the effect of expressing the stable form of β-catenin in the hematopoietic compartment. Gene transfer efficiency into 5-FU BMCs reached 70%–80%, as determined by the percent of green fluorescent protein GFP(+) cells 2 days after transduction. Accumulation of the stable form of β-catenin protein was detected in cells transduced with β-catenin and not in the control transduced (MPG) cells (Fig. 5A). There was no observed difference in apoptotic cells as determined by Annexin V binding to the undifferentiated fraction (LinSca-1+), nor increase in the percent LSK cells in β-catenin transduced cells when compared with control transduced cells (Fig. 5B, 5C). Nevertheless, when β-catenin transduced cells were transplanted into lethally irradiated recipients, levels of donor-derived transduced (GFP+) cells were dramatically decreased compared to the levels of GFP(+) cells in recipients of equivalent numbers of control transduced (MPG) cells (Fig. 5D). This decrease corresponded to an 11-fold lower frequency of CRUs for the β-catenin transduced cells when compared with the control transduced cells at the time of transplant (1/550,000 vs. 1/48,000 for β-catenin and MPG transduced, respectively; Fig. 5E). Thus, the predominant effect of directly expressing the stable form of β-catenin in hematopoietic cells was the loss of HSCs; this loss was not accompanied by the increase in the % of LSK cells that had been observed when β-catenin was stabilized in vivo [11, 12].

Figure 5.

Effect on hematopoietic function of β-catenin stabilized directly in hematopoietic cells. (A): Accumulation of the stable form of β-catenin protein after retroviral transduction of 5-fluorouracil treated bone marrow cells (5-FU BMCs) as detected by antibody against total form of β-catenin or HA tag. (B, C): Effects of β-catenin transduction on survival and differentiation. Lin bone marrow cells were transduced with each retroviral vector and cultured in medium containing identical cytokines without virus for an additional 48 hours. These cultured cells were then analyzed for % LSK cells of transduced (GFP+) or AnnexinV binding in GFP+LinSca-1+ cells. Shown are the mean ± SEM obtained from four independent experiments. (D, E): Comparison of engraftment levels of 5-FU BMCs transduced with MPG or β-catenin. 5-FU BMCs from Pep3b (Ly5.1) mice were transduced with one of the retroviral vectors and transplanted into lethally irradiated recipient mice (Ly5.2) and analyzed for engraftment. Shown are the representative flow cytometry plots for mice transplanted with 8 × 104 of BMCs (D), and CRU frequencies in the transduced BMCs calculated by employing Poisson statistics on percent negatively engrafted mice with 95% CI to represent ±2 SEM (E). Abbreviations: CRU, competitive repopulating unit; GFP, green fluorescent protein; HA, hemagglutinin; LSK, LinSca-1+c-kit+.

Taken together, our results reveal a distinct regulatory effect of β-catenin on HSCs depending on the primary site of activation and suggest that the regulatory effect of Wnt/β-catenin signals to promote maintenance or self-renewal of HSCs, as reflected by their compartmentalization in the microenvironment, are predominantly stroma-mediated.

Wnt/β-Catenin Activation in Stroma Induces Notch Signal Activation in HSCs

The observed stroma-mediated effect of β-catenin stabilization on HSCs in the specific conditions associated with stimulating HSCs suggested that a cross-talk may occur between stroma and HSCs, in which stromal cells activated by β-catenin may, in turn, stimulate HSCs. Therefore, we next explored the downstream effect of Wnt/β-catenin activated stroma in the microenvironment. A recent study demonstrated that induction of jagged-1, a serrate family notch ligand in the osteoblastic niche leads to self-renewal and expansion of HSCs in contact with stroma [3]. Promoters of jagged-1 and δ-like one have also been shown to be targeted by β-catenin [29, 30]. Therefore, we were interested to see whether Wnt/β-catenin signals activated in stroma also lead to a corresponding activation of notch in the hematopoietic microenvironment. As shown, both the transcript and protein levels of jagged-1 were induced in β-catenin stabilized stromal cells compared to control transduced stroma (Fig. 6A). Similarly, induction of jagged-1 was observed when stromal cells were stimulated in vitro with Wnt3a-CM compared to cells that had been stimulated with control-CM (Fig. 6B). Moreover, the δ-family notch ligand, δ-like one (dll-1) was also induced in β-catenin transduced cells and in the Wnt3a-CM stimulated cells (Fig. 6C, 6D). To further test induction of notch ligands under in vivo conditions, we examined mice bone marrows injected with Wnt3a-CM. As shown, significant induction of jagged-1 and dll-1 were seen in the trabecular endosteum of Wnt3a injected mice compared to marrows from mice injected with control CM (Fig. 6E, 6F). These results indicate that stromal cells activated by Wnt/β-catenin signals induce notch ligands and that notch signal activation may be involved in a microenvironmental cross-talk between activated stroma and HSCs.

Figure 6.

Induction of notch ligands in the stroma activated by Wnt/β-catenin signal. (A, C): Induction of jagged-1 (A) and dll-1 (C) in β-catenin transduced stromal cells as determined by Western blot and real-time polymerase chain reaction (PCR) analysis. Shown are the representative immunoblots (left) and the real-time PCR analysis (right) for relative expression levels of the indicated genes in β-catenin/stroma relative to the MPG/stroma calculated by normalizing each gene expression against glyceraldehyde 3 phosphate dehydrogenase (GAPDH) using the formula described in Material and Methods. (B, D): Induction of jagged-1 (B) and dll-1 (D) in stroma by stimulation with Wnt-3a CM. Stromal cells were stimulated with Wnt-3a CM or control-CM and expression of jagged-1 or dll-1 was examined by Western blot 24 hours after or by real-time PCR analysis 6 hours after stimulation. Shown are the representative immunoblots (left) and real-time PCR analysis (right) for relative expression levels of indicated genes in Wnt 3a-CM-stimulated cells relative to the cells stimulated with control-CM (Right) (n = 3). (E, F): Selective induction of jagged-1 (E) and dll-1 (F) in the endosteum of trabecular bone marrows of mice stimulated with Wnt 3a. Mice were intravenously injected with Wnt3a-CM or control CM and their bone marrows were examined 24 hours after for indicated notch ligand by immunohistochemistry. Shown are the representative images at indicated low and higher magnification (400×) of the dotted insets. Arrows indicate positive staining (brown color, DAB) of each antibody in the endosteum of trabecule. Abbreviation: CM, conditioned medium.

To identify other possible factors involved in this cross-talk, we also performed a microarray-based bioinformatics screen to identify extracellular genes that are induced in β-catenin/stroma compared to MPG/stroma. Although group of genes encoding for soluble growth factors for HSCs such as growth arrest-specific 6 (Gas-6) [31], CXCL5 [32] and proliferin-2 [33] were also induced in the β-catenin transduced stroma, we sought for membrane-bound or extracellular matrix-associated molecules as a candidate factor to mediate the contact-dependent effects. We identified one trans-membrane growth factor, δ-like one homolog (dlk-1) that was significantly induced by β-catenin activation (supporting information Table S1, Fig. 7A). Dlk-1 has been shown to exert a cis-inhibitory effect on cells expressing its protein products [34] but to activate notch signals in neighboring cells [35]. We also observed induction of dlk-1 in the Wnt3a-CM stimulated stroma and trabecular endosteum injected with Wnt 3a-CM (supporting information Fig. S6A, S6B). Another gene, microfibril-associated glycoprotein2 (MAGP-2/MFAP-5), that encodes for an extracellular matrix protein that can facilitate shedding of jagged-1 in the cell membrane [36] was also induced in our screen for genes induced in β-catenin transduced stromal cells (supporting information Table S1, Fig. 7A). Taken together, these results show that Wnt/β-catenin activation in the stroma leads to induction of several genes encoding for notch ligands and for molecules involved in notch signal regulation. This suggests that Wnt/β-catenin activated stroma may trigger downstream activation of notch signals in adjacent HSCs.

Figure 7.

Notch signal activation in the microenvironmental cross-talk with HSC. (A): Real-time polymerase chain reaction (PCR) confirmation of the indicated notch-related genes identified in the microarray-based screening. Shown are the relative expression levels of the indicated genes in β-catenin/stroma relative to the MPG/stroma calculated as described in Material and Methods (three experiments). (B): Activation of notch signals in cocultured hematopoietic cells. 5-Fluorouracil-treated bone marrow cells (5-FU BMCs) cocultured on each indicated stroma for 5 days were sorted for CD45+LinSca-1+ cells and induction of the indicated genes was examined by real-time PCR. Shown are the expression levels in the coculture with β-catenin/stroma relative to the MPG/stroma with SEM obtained from three independent experiments. (C): Effect of notch signal inhibitor on the cocultured hematopoietic cells. 5-FU BMCs were cocultured on each indicated stroma for 5 days in the presence of γ-secretase inhibitor ll (30 μM) or control solvent dimethyl sulfoxide. The cultures were then analyzed for LSK phenotype in the cocultured hematopoietic cells (CD45+). Shown are the ratios of the total numbers of LSK cells in each culture relative to the initial input LSK cell numbers with an error bar indicating SEM (four experiments; *, p < .05). (D): Proposed model integrating the observations for Wnt/β-catenin effects on HSCs. Activation of β-catenin may exert a positive regulatory effect on HSCs through the stroma with notch activation, but cause a negative effect when directly stabilized in HSCs (models 1 and 2). Glycogen synthase kinase 3β (GSK 3β) inhibitor administered in vivo may cause both a positive effect on stroma and a negative effect on HSC, thereby causing a mixed phenotype (model 3). Similarly, mice with conditional accumulation of β-catenin may display the results of both a positive effect on stroma during stromal contact in the trabecule (expansion of LSK) and a negative effect on HSCs when contact is lost in the marrow (loss of HSCs) (model 4). Abbreviations: HSCs, hematopoietic stem cells; LSK, LinSca-1+c-kit+; MPG, murine stem cell virus vector (MSCV) expressing green fluorescent protein (GFP) under the phosphoglycerate kinase (PGK) promoter (MSCV-PGK-GFP).

To test this possibility, we examined notch signal activation in HSCs in contact with activated stroma. As shown in Figure 7B, the undifferentiated hematopoietic cell population (CD45+LinSca-1+) purified from a coculture with β-catenin/stroma showed significant induction of the notch down-stream genes Hes-1 and deltex-1 [37] as well as induction of bmi-1, a gene whose expression level is tightly linked to HSC self-renewal [38, 39]. Moreover, inhibition of notch signals with the γ-secretase inhibitor abrogated the phenotypic changes of HSCs induced by β-catenin activated stroma, suggesting that the effects of β-catenin activated stroma on HSCs are dependent on functional down-stream notch signals (Fig. 7C). Activation of notch signals has previously been shown to enhance self-renewal of HSCs [40], whereas their inhibition leads to a failure to maintain the undifferentiated state and the self-renewal properties of HSCs [3, 41].

Taken together, our results show that Wnt/β-catenin signals activated in the stromal compartment leads to induction of notch ligands and that cross-talk between activated stroma and HSCs results in activation of notch signals and enhanced maintenance of HSCs.


There is a major interest in understanding the microenvironmental regulation of stem cell behavior and its physiological significance. In the current study, we show that canonical Wnt/β-catenin signals target the stromal compartment of the hematopoietic microenvironment and trigger distinct regulatory effects on HSCs. We show that physical contact between HSCs and β-catenin-activated stroma caused enhanced maintenance and expansion of HSCs as shown in phenotypic (LSK cells) and functional (in vivo competitive repopulating cells) assays. Higher maintenance of HSCs in β-catenin-activated stroma was most evident during in vivo reconstitution in a myeloablated host environment. Only limited self-renewal and moderately higher frequencies of HSCs were detected during short-term in vitro culture, possibly because of less optimal conditions for maintenance of HSCs in vitro and limited mitotic division of the cells under these conditions [42]. Alternatively, it is also possible that HSCs in contact with β-catenin-activated stroma are maintained in an undifferentiated but primed state for self-renewal such that under the in vivo conditions stimulating reconstitution they exhibit a higher level of self-renewal. In contrast, direct stabilization of β-catenin in hematopoietic cells resulted in the loss of HSCs. These results reveal a clear difference in the effects on HSCs of stabilized β-catenin depending on the site of activation in stromal or hematopoietic compartments. Although contrasting effects of β-catenin on HSC was also reported in a study model using bcl-2 transgenic mice [8], it should be considered that bcl-2 expression was shown to cause an expansion of HSCs and increase of their repopulating activities [43]. Furthermore, consistent with our finding that the effect of Wnt/β-catenin to promote the maintenance or self-renewal of HSCs is stroma-mediated, we detected compartmentalization of Wnt/β-catenin signal molecules in the hematopoietic microenvironment. Analysis of the gene expression patterns of Wnt signaling molecules revealed that canonical Wnt receptors were predominantly compartmentalized on stromal cells. Moreover, Wnt signaling molecules, such as the coreceptors LRP 5,6 [22], Ryk, a tyrosine kinase coreceptor of Wnt1, 3a [23], dickkopf-2, a molecule that binds to LRP 6 and activates canonical Wnt signaling [24], and Ror-2, an orphan receptor tyrosine kinase that potentiates canonical pathway signaling [25] were also enriched in stromal cells compared to 5-FU BMCs. In contrast, Wnt ligands involved in canonical pathways (Wnt1, 2b, 4, 10b) were enriched in hematopoietic cells of 5-FU BMCs. Interestingly, molecules involved in the noncanonical Wnt pathway that have been shown to have inhibitory effects on the canonical pathways [20] were reciprocally localized from those molecules involved in canonical pathways. However, the significance of this finding needs further studies, since it is possible that Wnt interacts with multiple Fz receptor families and their effects could be dependent on the receptor context [44, 45].

Nonetheless, consistent with compartmentalized gene expression patterns, immunohistochemical examination of trabecular bone marrows showed β-catenin localization predominantly to the endosteal lining of CD45 N-cadherin+ cells, suggesting that the stromal microenvironment serves as a primary target site of β-catenin activation. Importantly, endosteal accumulation of the active form of β-catenin was detected selectively under the “stimulated” and “stressed” conditions that can stimulate HSCs in vivo. The regulated accumulation of β-catenin to the stroma suggested that cross-talk occurs between β-catenin-stabilized stroma and HSCs to transmit stromal activation signals to HSCs. In a search for the factors that mediate such a cross-talk we found that jagged-1 and dll-1, notch ligands with TCF/LEF binding sites in their promoters [29, 30], were induced by β-catenin activation. Furthermore, demonstration that dll-1 and jagged-1 were specifically induced in the trabecular endosteum of mice injected with Wnt3a-CM provided further evidence of a link between physiological activation of Wnt/β-catenin signals and induction of notch ligands. Interestingly, a microarray-based bioinformatics screen for membrane-associated genes induced in β-catenin activated stroma provided further support for the involvement of notch signal activation in the cross-talk between stroma and HSCs as two membrane-associated genes related to notch activity, dlk-1 and MAGP-2 were identified in the screen. The protein products of dlk-1 do not have a receptor binding [delta/serrate/lag2 (DSL)] domain and their function remains poorly understood. However, in a recent study, dlk-1 was shown to exert a cis-inhibitory effect on autonomous notch activity in cells expressing dlk-1 protein during drosophila development [34], but to associate with notch signal activation in neighboring cells during mammalian thymus development [35]. In addition, stromal expression of dlk-1 was shown to increase primitive hematopoietic cells during coculture [46]. Moreover, the extracelluar matrix protein, MAGP-2, was shown to facilitate specific shedding of membrane-bound jagged-1 and its oligomerization to activate notch signals [36]. Therefore, it is possible that stabilization of β-catenin in the bone marrow stroma leads to an induction of notch ligands in the stroma in a regulated manner. Interestingly, recent study have shown that notch signals are dispensable for normal maintenance of HSCs and that only basal levels of notch signals are exposed to HSCs in the bone marrow environment [47]. Consistent to the observation, we observed almost undetectable levels of notch ligands in the in situ bone marrow sections examined, unless stimulated by Wnt 3a ligand (Fig. 6E, 6F). In contrast, activation of notch signals in HSCs [40, 48] or induction of notch ligands in the stroma [3] could enhance self-renewal of HSCs. Therefore, it is possible that the phenotypes detected following Wnt/β-catenin activation of stroma are mediated by up-regulation of notch ligands and activation of notch signaling in HSCs. In support of this model, significant induction of the down-stream notch signals, Hes-1 and deltex-1, was seen in the hematopoietic cells in contact with β-catenin-activated stroma, and the phenotypic change to these hematopoietic cells, higher maintenance of undifferentiation, was abrogated by inhibition of down-stream notch signals, indicating that the phenotype is dependent on the notch signal activation in HSCs. However, possible influence of γ-secretase type ll inhibitor for signals other than notch signals should be also considered and in vivo effects of notch signals on the phenotypic or functional changes of HSCs should be explored. Similarly, at present, we can not completely rule out the possibility that other growth factors induced by β-catenin in the stroma were also involved in the cross-talk with HSCs, since growth factors such as Gas-6 [31], CXCL5 [32] and proliferin-2 [33] had been shown to maintain HSCs in the stroma-based culture. Further studies are necessary to elucidate the underlying signals involved in the cross-talk in the microenvironment.

Of note, recent studies have shown that, in addition to endosteal niche, perisinusoidal vascular niche constitute another microenvironment for HSCs, since large part of phenotypically defined HSCs were shown to reside in the vicinity of MECA-32+ sinusoidal cells in the bone marrow [49, 50]. Although the functional distinction between the endosteal and perisinusoidal niche remains unclear, subsets of human perisinusoidal stroma(CD45MCAM/CD146+) or cultured stromal cells could reconstitute both endosteal and perivascular niche [28, 51]. Interestingly, we observed similar β-catenin expression in the subset of perisinusoidal stroma and MCAM/CD146+ cells in the perisinusoidal (MECA-32+) region of murine bone marrows and (supporting information Fig. S7A, S7B). In addition, we observed that these perisinusoidal stromal cells (LinCD45CD31 CD146+) also responded to in vitro and in vivo stimulation with Wnt 3a (supporting information Fig. S7C–S7F). Therefore, it is possible that stromal targeting of Wnt-β-catenin signals in the bone marrow may include both the perisinusoidal and the endosteal compartments of the microenvironment. Further studies on the specific role of each niche for HSCs are necessary to better evaluate the biological effects of Wnt /β-catenin signals on the perisinusoidal and endosteal microenvironments.

Of note, our results demonstrating distinct biological effects for stroma-mediated Wnt/β-catenin signals may allow the integration of findings from previous studies (schematic illustration in Fig. 7D). For example, the different effects observed following administration of GSK 3β in vivo and those observed in mice with conditional in vivo stabilization of β-catenin may reflect both positive regulatory effects on HSCs through the stroma and negative effects on HSCs from direct β-catenin stabilization. However, the effect of β-catenin disruption in the microenvironment, unfortunately, could not be determined since mice die within 23 days of induction [10], whereas the effect on HSCs resulted contrasting level of competitive repopulation of HSCs depending on the study model [10, 52]. Further studies are needed to evaluate both the autonomous and paracrine actions of Wnt/β-catenin signals in the hematopoietic microenvironment.

Importantly, our findings on the stroma-mediated effect of Wnt/β-catenin signals on HSCs provide new insight on the significance of the microenvironment as a mediator of regulatory signals for HSCs. Therefore, the stem cell niche may be an attractive target for stem cell manipulation as inferred from recent success in HSC expansion and mobilization by using a hormone, parathyroid hormone, that selectively target stroma [53] Further studies are also warranted to study the role of the microenvironment in various disease conditions.


In summary, our findings provide new insight on a mode of Wnt-β-catenin signaling that is mediated by stroma and the cross-talk between activated stroma and HSCs that may serve as a physiologically regulated microenvironmental cue for HSC self-renewal in the stem cell niche. Our data provide a new understanding of the regulation of HSCs and helps to integrate results from previous studies.


This research was supported by a grant for high-performance cell therapy R&D project (0405-DB01-0104-0006) from the Ministry of Health, Welfare and Family and a grant from Korea Science and Engineering Foundation (KOSEF) (2008-05981) and Stem Cell Research Program (to E.J.) of Korean Government (Mininstry of Education, Science, and Technology [MEST]). Microarray study was supported by Shared Research Equipment Assistance Program (S.-H.L. and I.-S.C.) by Korea Basic Sicence Institute, MEST. We thank Dr. Keith Humphries (Terry Fox Lab, Vancouver, BC, Canada) for his kind reading of the manuscript.


The authors indicate no potential conflicts of interest.