Osteolineage niche cells initiate hematopoietic stem cell mobilization

Abstract

Recent studies have implicated bone-lining osteoblasts as important regulators of hematopoietic stem cell (HSC) self-renewal and differentiation; however, because much of the evidence supporting this notion derives from indirect in vivo experiments, which are unavoidably complicated by the presence of other cell types within the complex bone marrow milieu, the sufficiency of osteoblasts in modulating HSC activity has remained controversial. To address this, we prospectively isolated mouse osteoblasts, using a novel flow cytometry-based approach, and directly tested their activity as HSC niche cells and their role in cyclophosphamide/granulocyte colony-stimulating factor (G-CSF)-induced HSC proliferation and mobilization. We found that osteoblasts expand rapidly after cyclophosphamide/G-CSF treatment and exhibit phenotypic and functional changes that directly influence HSC proliferation and maintenance of reconstituting potential. Effects of mobilization on osteoblast number and function depend on the function of ataxia telangiectasia mutated (ATM), the product of the Atm gene, demonstrating a new role for ATM in stem cell niche activity. These studies demonstrate that signals from osteoblasts can directly initiate and modulate HSC proliferation in the context of mobilization. This work also establishes that direct interaction with osteolineage niche cells, in the absence of additional environmental inputs, is sufficient to modulate stem cell activity.

Figure 1

Phenotypic isolation of osteoblasts. (A) Strategy for isolation of bone-associated cells. (B) Flow cytometric analysis of single-cell suspensions from collagenase-treated bone (Bone) or BM aspirates (BM). Data are presented as dot plots showing staining for hematopoietic markers (CD45/Ter119) and Opn. The frequency of Opn⁺CD45⁻Ter119⁻ cells is shown in the upper left. “-CNTRL” indicates representative background fluorescence seen in the absence of Opn secondary antibody (rat-αFITC) or when an isotype control (rat-αIgG) only is used. This is representative of 10 to 15 independent experiments. (C) QRT-PCR analysis of mRNA isolated from double-sorted Opn⁺ cells (OPN⁺) compared with irrelevant bone marrow cells (BM) or Opn⁻ (OPN⁻) cells. Relative expression (mean ± SD) over a β-actin normalization control is shown for osteocalcin (OC), alkaline phosphatase (ALP), osteoactivin (OA), Runx2 (RX2), and osterix (OS), vascular endothelial growth factor receptor 1 and 2 (VEGFR1, VEGFR2), tyrosine kinase with Ig and EGF homology domains 1 and 2 (TIE1, TIE2), endothelial nitric oxide synthase (eNOS); n = 3 independent experiments for osteolineage markers; n = 2 independent experiments for endothelial markers). Each transcript was measured in triplicate. (D) Flow cytometric analysis of parathyroid hormone receptor-1 (PTHR-1) expression by Opn⁺CD45⁻Ter119⁻ cells (gray histogram) or BM cells (black line). Both PTHR1^hi and PTHR1^lo OPN⁻CD45⁻Ter119⁻ cells exhibited equal levels of activity of bone-specific ALP (Figure S2), consistent with an equivalent state of osteolineage differentiation of PTHR1^hi and PTHR1^lo cells^, and probably reflecting known differences in PTHR1 expression by functionally equivalent osteoblasts present at local areas of bone formation or of bone resorption. (E) QRT-PCR analysis of mRNA isolated from individual double-sorted Opn⁺CD45⁻Ter119⁻ cells. Data are plotted as the percentage of osteoblasts (mean ± SD) with increased expression of OC, ALP, OA, RX2, or OS compared with a control population of single-sorted total BM cells (n = 2 independent experiments).

Figure 2

Opn⁺CD45⁻Ter119⁻ cells isolated from bone exhibit osteoblast function. (A) Flow cytometric analysis of ALP activity, determined using a modified version of the fluorogenic ELF-97 assay (Invitrogen). Histograms show kinetics (1 minute, 15 minutes, and 30 minutes) of detection of ALP activity in purified Opn⁺ cells (black line) compared with the ALP activity in a control cell population of total BM cells (gray filled, n = 4). Numbers in the upper right indicate frequency of ALP⁺ cells among Opn⁺ cells/frequency of ALP⁺ among BM control cells. This is representative of 3 independent experiments (*P < .05 at each time point indicated). (B) Sorted osteoblasts form bone nodules in culture. Ten OPN⁺ cells were cultured in α-MEM/15% serum (d1 of culture, 10 times) and at day 15 (d15, 4 times) osteogenesis was induced by the addition of 20 mg/mL ascorbic acid plus 50 μM β-glyceraldehyde. Cell clusters form after 7 days (d22 of culture, 4 times) and distinct bone nodules are visible by 15 days (d30 of culture, 4 times). Mineralization was confirmed at d30 by the von Kossa method. Micrographs shown are representative of 3 independent experiments (n = 20 wells/experiment). Plates were viewed using a UPLanF1 lens at 10×/0.30 Ph1 (first micrograph) or a UPLan FLN lens at 4×/0.13 (all other micrographs). The efficiency of bone nodule formation was 89% plus or minus 3% of seeded wells over the 3 experiments performed. (C) Differentiating osteoblasts in bone nodule cultures showed increasing expression of ALP mRNA (graph, middle panel, representative of 3 to 5 different wells sampled individually at each time point indicated; *P < .05) and increasing ALP activity (histogram, right panel, representative of 3 individual wells sampled at each time point, day 15, day 20, or day 30, after culture initiation). Data are representative of 3 independent experiments; error bars represent SD.

Figure 3

Cy/G-CSF treatment induces changes in the osteoblastic niche. (A) Experimental strategy. For HSC mobilization, mice were injected intraperitoneally with Cy (200 mg/kg) and then on successive days with human G-CSF (250 μg/kg/day) administered as a single daily subcutaneous injection. The day of Cy treatment was considered day −1 and the first day of G-CSF treatment as day 0. The timeline of Cy and G treatments for each experimental group is shown. For in vivo BrdU labeling experiments, mice in every experimental group (untreated, D0, D2, and D4) received 9 successive injections of BrdU (1 mg/mouse in 0.9% saline, intraperitoneally, indicated by green arrows), over a 4.5-day period. All experimental groups began BrdU treatment 12 hours after Cy injection in the first experimental group (D4) and received continued BrdU injections every 12 hours thereafter (indicated by green arrows). Thus, all animals were sacrificed (SAC) after the same time period of BrdU exposure. (B) Osteoblast (Opn⁺CD45⁻Ter119⁻) frequencies are increased in Cy/G-treated mice. Representative FACS plots are shown (n = 10), with the frequency of Opn⁺CD45⁻Ter119⁻ shown in each gate. (C) Total numbers of osteoblasts are increased in response to Cy/G (D2) treatment but return to normal numbers by D4. Data are plotted as means plus or minus SD. Differences were significant for untreated versus D0 (*P < .05) and untreated vs D2 (*P < .05). Differences were not significant for untreated versus D4 (*P > .05). (D) Osteoblasts proliferate in response to Cy/G treatment as demonstrated by increased in vivo uptake of BrdU in osteoblasts from mobilized (D0, D2, and D4) as opposed to untreated (untreated, far left) mice. Histograms indicate the percentage BrdU⁺ cells (determined by control staining of osteoblasts from mice that did not receive BrdU) among Opn⁺CD45⁻Ter119⁻ cells. This is representative of 2 independent experiments.

Figure 4

Cytokine mobilized osteoblasts promote HSC proliferation and maintain HSC reconstituting potential. (A) Experimental strategy. A total of 100 000 lineage-negative (Lin⁻) BM cells, which include HSC and hematopoietic progenitor cells (HPCs) from untreated mice, were exposed for 12 hours in vitro to 2000 OPN⁺ osteoblasts isolated from untreated mice (white) and analyzed for KTLS frequency and/or proliferation by loss of CFSE by FACS. (B) Frequency of c-kit⁺Thy1.1^loLin⁻Sca1⁺ (KTLS) HSC among Lin-depleted BM cells (HPC) maintained alone () or with untreated () or cytokine-modified (D2; ■) osteoblasts for 12 hours. KTLS HSC frequency among input HPCs is also shown (□). Data are presented as the average frequency of KTLS HSC (± SD) as determined by FACS in short-term (12 hours) HPC:osteoblast cocultures (*P < .05). Data are compiled from 6 independent experiments performed. (C) Increased proliferation of CFSE-labeled KTLS HSC among Lin-depleted BM cells (HPC) cultured with cytokine-modified (HPC:D2, panel 3) compared with untreated (HPC:untreated, panel 2) osteoblasts after 36 hours. Histograms represent CFSE labeling of only those cells falling within the KTLS gate and are representative of 7 independent experiments. (D) Average frequency (± SD) of divided KTLS cells per peak (1, 2, 3, as indicated in panel C) in coculture experiments (*P < .05). (E) Increased cell survival in HPC cultures containing untreated or D2 osteoblast, compared with HPC cultured alone. Cultures were performed as described previously, and stained with KTLS surface markers in addition to markers of cell death and viability (annexin V (early apoptosis marker) and PI (late apoptosis marker)). Data are representative of 2 independent experiments. (F) Increased total numbers of KTLS HSCs in HPC:D2 cultures. HSC numbers were determined for each culture by multiplying the frequency of KTLS cells (determined by FACS) by the total number of cells present. Data are shown as the mean absolute cell number for each condition over 7 independent experiments plus or minus the SD (*P < .05).

Figure 5

Increased engraftment efficiency of CFP⁺ HSC exposed for 12 hours to D2 osteoblasts compared with HSC exposed to untreated osteoblasts. (A) Experimental strategy. A total of 100 000 Lin⁻ BM cells (HPC) from untreated mice were exposed for 12 hours in vitro to 2000 OPN⁺ osteoblasts isolated from D2 mobilized mice (which had received 1 dose of Cy + 2 daily doses of G; black) and tested for HSC function by measuring reconstitution potential. (B) Lethally irradiated recipient mice were transplanted with cells from 12-hour HPC only cultures (), or HPC:untreated () or HPC:D2 (▬) cocultures. The average engraftment was calculated by determining the percentage donor-derived myeloid (top; Mac-1⁺, Gr-1⁺), B (middle, B220⁺), and T (bottom, CD3⁺, CD4⁺, CD8⁺) cells at the indicated time points after transplant. Donor-derived cells were identified based on CFP transgene expression and were corrected for the percentage of CFP⁺ myeloid or lymphoid populations (80%-90%) in unmodified CFP⁺ transgenic hosts (*P < .05; ‡P = .0598). Error bars represent SD.

Figure 6

Effects of Cy/G treatment in ATM^−/− mice. (A) Osteoblast cell numbers are not increased in Cy/G treated ATM-deficient mice. The total numbers of osteoblasts isolated after Cy/G treatment were calculated based on the frequency of OPN⁺ cells (as determined by FACS) among total numbers of cells isolated from collagenase treated bones from either ATM^+/+ or ATM^−/− mice that were untreated (ATM WT untreated, ; ATM KO untreated, ▬) or treated with Cy/G (ATM WT D2, gray hatched; ATM KO D2, black hatched). Data are plotted as means ± SD (*P < .05). (B) Decreased osteoblast survival after Cy/G treatment in ATM-deficient mice. ATM WT or ATM KO mice were treated with Cy/G and osteoblasts were isolated from collagenase-treated bones and stained with osteoblast specific markers as well as cell viability markers (annexin V⁺, early apoptotic marker; and PI⁺, late apoptotic marker). Representative data for 2 independent experiments are shown. (C-E) Lower frequency and engraftment efficiency of wild-type HSC exposed for 12 hours to D2 ATM KO osteoblasts compared with HSC exposed to wild-type D2 osteoblasts. A total of 100,000 Lin⁻ BM cells (HPC) from wild-type, untreated mice were exposed for 12 hours in vitro to 2000 OPN⁺ osteoblasts isolated from ATM KO or WT D2 mobilized mice (ATM WT untreated, ; ATM KO untreated, ▬; ATM KO D2, black hatched; ATM WT D2, gray hatched) and tested for HSC function by measuring in vitro maintenance of HSC frequency (C,D) compared with an HPC only control (gray dotted bar in panel D) and in vivo reconstitution potential (E) as described previously (Figures 4,5). Data are shown compared with an HPC only control (gray dotted) and an uncultured HPC control (■). Data are plotted as mean (± SD; *P < .05) and represent chimerism at 16 weeks after transplant. (F) Equivalent frequency of apoptotic (annexin V⁺, PI⁻ early apoptotic) osteoblasts after 12-hour HPC:osteoblast coculture assays. Coculture assays were performed as described previously, and cells were stained for osteoblast specific markers and the cell death and viability markers annexin V (early apoptotic marker) and PI (late apoptotic marker). Representative data are shown for 2 independent experiments performed.

Figure 7

Model of osteoblast-mediated HSC mobilization and maintenance of reconstituting potential. (A) Osteoblasts (blue) within the normal BM microenvironment appear to negatively regulate HSC proliferation (red), at least in part through cell-cell contact. (B) Changes in the osteoblast compartment induced by Cy/G treatment (indicated by yellow star) alter the niche (red star) such that osteoblasts promote HSC proliferation and maintain HSC reconstitution potential. This alteration could arise in vivo directly, by regulation of fully differentiated osteoblasts by Cy/G, or indirectly, via (1) the stimulation/proliferation of more primitive osteoprogenitor cells (blue) or (2) the cross-regulation/activation of osteoblasts by osteoclasts (green) or other stromal elements (brown), that through an ATM dependent mechanism results in (3) a specialized subset of osteolineage cells that are sufficient to promote HSC self-renewal through the secretion of soluble factors (pink asterisks).