Stromal cell–derived factor 1/CXCR4 signaling is critical for the recruitment of mesenchymal stem cells to the fracture site during skeletal repair in a mouse model

Authors


Abstract

Objective

Stromal cell–derived factor 1 (SDF-1; CXCL12/pre–B cell growth-stimulating factor) is a dominant chemokine in bone marrow and is known to be involved in inflammatory diseases, including rheumatoid arthritis. However, its role in bone repair remains unknown. The purpose of this study was to investigate the role of SDF-1 and its receptor, CXCR4, in bone healing.

Methods

The expression of SDF-1 during the repair of a murine structural femoral bone graft was examined by real-time polymerase chain reaction and immunohistochemical analysis. The bone graft model was treated with anti–SDF-1 neutralizing antibody or TF14016, an antagonist for CXCR4, and evaluated by histomorphometry. The functional effect of SDF-1 on primary mesenchymal stem cells was determined by in vitro and in vivo migration assays. New bone formation in an exchanging-graft model was compared with that in the autograft models, using mice partially lacking SDF-1 (SDF-1+/−) or CXCR4 (CXCR4+/−).

Results

The expression of SDF1 messenger RNA was increased during the healing of live bone grafts but was not increased in dead grafts. High expression of SDF-1 protein was observed in the periosteum of the live graft. New bone formation was inhibited by the administration of anti–SDF-1 antibody or TF14016. SDF-1 increased mesenchymal stem cell chemotaxis in vitro in a dose-dependent manner. The in vivo migration study demonstrated that mesenchymal stem cells recruited by SDF-1 participate in endochondral bone repair. Bone formation was decreased in SDF-1+/− and CXCR4+/− mice and was restored by the graft bones from CXCR4+/− mice transplanted into the SDF-1+/− femur, but not vice versa.

Conclusion

SDF-1 is induced in the periosteum of injured bone and promotes endochondral bone repair by recruiting mesenchymal stem cells to the site of injury.

Skeletal injuries remain among the most prevalent clinical problems, especially in the aging society. This problem is compounded in patients with rheumatoid arthritis (RA), who tend to have lower bone mineral density due to the combination of their disease-induced generalized osteopenia (1) and the consequences of steroid-induced osteoporosis (2). Although simple fractures are often treated effectively, osteoporotic hip fractures are associated with 24% mortality (3). The structural bone loss that occurs in compound fractures and periprosthetic osteolysis also exemplifies the serious clinical problems requiring massive structural bone reconstruction. In reconstructive surgery in such cases, autologous bone grafting is the gold standard due to the abilities of the live donor tissue, which serves as both the biologic scaffold and the source of osteogenesis via its mesenchymal stem cells (MSCs) (4).

Due to morbidity issues in terms of donor sites and the limited availability of bone autografts, allogeneic bone graft transplantation has become the standard of care. However, because the allograft bone is dead and void of osteogenic and osteoinductive properties (5), the long-term clinical results are poor. A recent study showed a 56.6% survival rate for structural allografts at 216 months (6). In order to overcome these limitations, it is of great importance to develop novel biologic strategies, which first requires elucidation of the molecular signals responsible for successful bone repair.

MSCs are pluripotent cells that differentiate into multiple cell lineages and can promote structural and functional repairs in many organs including bones, making MSCs an attractive candidate for cell-based bone regeneration (7). Many experimental and clinical studies have attempted to regenerate bones with MSCs, but the results present several notable shortcomings, such as vulnerability to infection, uncertainty regarding the differentiation capability in specific in vivo situations, the high cost of ex vivo cell handling, concern regarding the limited number of cells that can actually contribute to bone formation, and possibly even malignant transformation of the cells during ex vivo cell expansion (8–10). During the course of organ regeneration, however, it has been demonstrated that both local MSCs derived from the injured tissue and circulating MSCs collaborate in the healing of damaged organs. Circulating MSCs “sense” a tissue injury, migrate to the sites of damage, and undergo tissue-specific differentiation (11). However, the mechanisms responsible for MSC migration to the site of bone injury have not yet been revealed.

Stromal cell–derived factor 1 (SDF-1)/pre–B cell growth-stimulating factor belongs to the CXC subfamily of chemokines such as CXCL12, which was initially identified as a bone marrow stromal cell–derived factor (12) and as a bone marrow stromal cell–derived pre–B cell stimulatory factor (13). SDF-1 plays many important roles through activation of a G protein–coupled receptor, CXCR4 (13–15), and the interaction of SDF-1/CXCR4 and hematopoietic stem cells (HSCs) has been extensively reported. In bone marrow, endothelial cells and stromal cells express SDF-1, which not only acts as a chemoattractant for HSCs to a bone marrow niche but also supports their survival and proliferation (16, 17). Furthermore, during the last decade, accumulating data have supported an emerging hypothesis that SDF-1/CXCR4 also plays a pivotal role in the biologic and physiologic functions of MSCs (18, 19). SDF-1 is up-regulated at sites of injury and serves as a potent chemoattractant to recruit circulating or residing CXCR4-expressing MSCs, which are necessary for tissue-specific organ repair or the regeneration of the liver (20), heart (21, 22), brain (23), kidney (24), and skin (25). Moreover, the local delivery of SDF-1 into injured tissue promotes the recruitment of circulating mesenchymal stromal and progenitor cells to lesions in the heart (21, 22) and brain (23). However, the involvement of the SDF-1/CXCR4 axis on MSCs in bone repair has not been elucidated.

In this study, we attempted to investigate the hypothesis that SDF-1 plays an important role in endochondral bone repair. Using mouse segmental bone graft models, both live and dead, we demonstrated that SDF-1 recruits MSCs to bone repair sites during the early phase of bone repair. Our results not only lead to further understanding of the physiologic mechanisms of bone regeneration but also suggest new strategies for the therapeutic use of SDF-1 to promote successful bone healing.

MATERIALS AND METHODS

Reagents.

Recombinant murine CXCL12/SDF-1α (rMuSDF-1) and monoclonal anti-human/mouse SDF-1/CXCL12 neutralizing antibody (clone 79014) were purchased from R&D Systems (Minneapolis, MN). Goat anti-human SDF-1 polyclonal antibody (sc-6193) and normal goat IgG (sc-2028) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Phycoerythrin (PE)–conjugated CD34 antibody (clone RAM34), PE-conjugated CD45 antibody (clone 30-F11), and fluorescein isothiocyanate (FITC)–conjugated CD44 antibody (clone IM7) were purchased from eBioscience (San Diego, CA), and FITC-conjugated CD29 antibody (clone Ha2/5) was purchased from BD PharMingen (San Diego, CA).

Murine live and dead bone graft models.

All animal studies were conducted in accordance with principles and procedures approved by the Kyoto University Committee of Animal Resources. Murine segmental live and dead bone graft models were created using 6-week-old C57BL/6 mice, as previously described (26). The mice were killed on days 0, 0.5, 1, 2, and 3 after surgery for RNA extraction, and on days 7, 14, and 28 for histologic and radiologic analysis. The live bone graft model was also created in SDF+/− and CXCR4+/− mice.

RNA extraction, semiquantitative polymerase chain reaction (PCR), and quantitative real-time PCR.

Total RNA was extracted from mouse tissue samples. The periosteum and liver were snap-frozen in liquid nitrogen and total RNA was extracted using TRIzol (Gibco BRL, Carlsbad, CA), which was purified using the RNeasy Mini Kit (Qiagen, Valencia, CA). To isolate total RNA from the graft surgery specimens, the bone graft and surrounding tissue were homogenized with Polytron (Kinematica, Tokyo, Japan), followed by RNA extraction. The RNA was reverse-transcribed and amplified by PCR, as previously described (27). The PCR products were evaluated by electrophoresis on a 2% agarose gel (Sigma, St. Louis, MO) for semiquantitative PCR. Quantitative PCR was performed using LightCycler SYBR Green I Master and a LightCycler qPCR machine (Roche Diagnostics, Penzberg, Germany), according to the manufacturer's instructions.

All gene expression data were normalized against actin. The primer sequences and expected fragment sizes of the PCR products are presented in Table 1. The same primer sets were used for both semiquantitative PCR and quantitative PCR.

Table 1. Primer sequences used for polymerase chain reaction and amplification
GenePrimer nucleotide sequenceProduct size, bpAccession no.
SDF1Forward 5′-CGCCAGAGCCAACGTCAAGC-3′107NM007393
 Reverse 5′-TTCGGGTCAATGCACACTTG-3′  
CSCR4Forward 5′-CAGCATCGACTCCTTCATCC-3′118NM015784
 Reverse 5′-GGTTCAGGCAACAGTGGAAG-3′  
PeriostinForward 5′-GAAGAACGAATCATTACAGG-3′189L12029
 Reverse 5′-CGGGTGTGTCTCCCTGAAGC-3′  
ActinForward 5′-AGATGTGGATCAGCAAGCAG-3′125NM009911
 Reverse 5′-GCGCAAGTTAGGTTTTGTCA-3′  

Primary cell culture.

Primary mouse bone marrow–derived stromal cells (BMSCs), harvested as previously described (28), were cultured in α-minimum essential medium (α-MEM) (Sigma) containing 10% fetal bovine serum, 100 units/ml penicillin, and 100 μg/ml streptomycin (Gibco BRL) at 37°C in a 5% CO2 incubator with a humidified atmosphere. After 3 days, the culture was rinsed 3 times with phosphate buffered saline (PBS) to remove nonadherent cells. When the cells became confluent, they were trypsinized and passaged at a ratio of 1:3, which was regarded as postpassage 1. The cell surface antigens were analyzed by flow cytometry. To label mouse BMSCs, cells were incubated in culture medium supplemented with bromodeoxyuridine (BrdU Cell Proliferation Kit; Amersham, Piscataway, NJ) for 24 hours before the cells were harvested.

In vitro chemotaxis assay.

In vitro cell migration was assayed using inserts with an 8-μm pore membrane, as previously described (29). The upper wells were loaded with 1 × 105 mouse BMSCs in α-MEM containing 0.1% bovine serum albumin. For the chemotaxis assay, rMuSDF-1 at different concentrations (0 ng/ml, 10 ng/ml, and 100 ng/ml) in 500 μl of medium was applied to the lower chambers. For the chemokinesis and inhibition assays, rMuSDF-1 (100 ng/ml) and TF14016 (50 nM or 100 nM), an antagonist of CXCR4 (30), respectively, were applied to the upper chambers. To evaluate involvement of the Gi protein, 100 ng/ml of pertussis toxin, a specific inhibitor of the Giα subunit, was added to the chambers. After 24 hours of incubation, migrated cells were counted under a light microscope.

In vivo chemotaxis assay.

To further investigate whether MSCs migrate to the site of bone healing in vivo, BrdU-labeled mouse BMSCs were transplanted intravenously into the graft-recipient mice, and the distribution of these cells around the graft bone was evaluated. The labeled cells were collected in PBS at a cell concentration of 1.0 × 106/ml, and a 200-μl aliquot of this cell suspension was injected into each mouse through the tail vein, after the graft surgery. To assess the inhibitory effect of the receptor antagonist, these mice also received continuous administration of TF14016 dissolved in PBS (10 mM) or vehicle, using Alzet micro-osmotic pumps (model 1002; Durect, Cupertino, CA) delivering 12 μg/day of TF14016, as described previously (31). The mice were killed on day 7 for histomorphometric analysis. The migrated BrdU-positive cells were immunostained, and the number of migrated cells was counted. Cell counting was performed by 3 blinded observers, and the mean ratio of BrdU-positive chondrocytes–to–total chondrocytes in the formed soft callus was calculated.

In vivo treatment for loss-of-function studies in the mouse model.

For loss-of-function studies of the live graft model, mice received mouse anti–SDF-1 neutralizing antibody or TF14016. Anti–SDF-1 neutralizing antibody or control IgG was injected intraperitoneally. Pulse injections of 84 μg of anti–SDF-1 neutralizing antibody diluted in PBS were given on days 2, 4, 7, and 10 after surgery (336 μg total). The mice were killed on day 14 for histologic analysis. TF14016 (10 mM [84 μg total]) or vehicle was administrated continuously, as described above, and the mice were killed on day 7.

Exchanging–live bone graft models in SDF+/− mice and CXCR4+/− mice.

To investigate the functional role of SDF-1, segmental graft bones were exchanged between 6-week-old SDF-1+/− mice and CXCR4+/− mice. Segmental live bone derived from a CXCR4+/− mouse was transplanted into a host SDF-1+/− mouse, and this exchange represented the “C-to-S” model. The reverse approach (segmental live bone derived from an SDF-1+/− mouse transplanted into a host CXCR4+/− mouse) represented the “S-to-C” model. These mice were killed on day 14 for histologic and radiologic analysis.

Histologic analysis and measurement of the area of new bone formation.

Specimens were processed as paraffin-embedded sections with a thickness of 5–7 μm and stained with hematoxylin, eosin, and Alcian blue. The areas of new bone formation were measured by computer tracing, as previously described (26). For immunohistochemical analysis, deparaffinized sections were blocked for endogenous peroxidase activity with 0.3% hydrogen peroxide in methanol for 20 minutes. Goat anti–SDF-1 polyclonal antibody (dilution 1:100) or control IgG antibody was applied and incubated for 30 minutes at room temperature. The reaction products were visualized using the Vectastain ABC Kit and the DAB Peroxidase Substrate Kit (both from Vector, Burlingame, CA), according to the manufacturer's instructions. Immunohistochemical detection of BrdU-labeled cells was performed using the Cell Proliferation Kit (Amersham), according to the manufacturer's instructions.

Statistical analysis.

Data are presented as the mean ± SD and were analyzed with Student's t-test. P values less than 0.05 were considered significant.

RESULTS

Up-regulated expression of SDF1 messenger RNA (mRNA) in the acute phase of live bone graft healing.

To investigate the involvement of SDF-1 in the acute phase of skeletal repair, we used mouse segmental live and dead bone graft models (26). Histologic analysis showed that the area of new bone formation was significantly reduced around the dead bone graft compared with the live graft on days 7, 14, and 28, by 68%, 73%, and 83%, respectively (additional information available from the corresponding author). Radiologic analysis showed sufficient callus formation around the live graft, but few calluses formed around the dead allograft (data not shown). Next, total RNA was extracted from live and dead bone grafts on days 0, 0.5, 1, 2, and 3 (n = 4 for each time point), and the expression levels of SDF1 and CXCR4 mRNA were analyzed by quantitative PCR. SDF1 mRNA expression increased by 3.0-fold in the live bone grafts on day 2 when compared with day 0 (P < 0.05) and was 3.1-fold higher than that in the dead grafts at the same time point (P < 0.05). Expression further increased by 4.1-fold on day 3, although the increase had no statistical significance (P = 0.090). In contrast, no significant increase in SDF1 mRNA expression was identified in the dead graft model (Figure 1A). CXCR4 mRNA levels during live and dead graft healing were similar at all time points (data not shown). These results indicate the involvement of SDF-1 in the acute phase of endochondral bone healing.

Figure 1.

Expression pattern of stromal cell–derived factor 1 (SDF-1) in a live graft bone during the acute phase of endochondral bone healing. A, Time course of SDF1 mRNA expression in live and dead bone graft models, as analyzed by real-time quantitative polymerase chain reaction (PCR) (n = 4 at each time point). Expression levels are the fold index versus the day 0 level in each model. Values are the mean and SEM results from triplicate real-time PCR analyses. ∗ = P < 0.05. B, Immunohistochemical staining for SDF-1. Bottom panels show higher-magnification views of the boxed areas in the top panel. Arrowheads indicate the periosteum of the live graft (left) and the host bone (right) of the same section on day 2 after the live graft surgery. Bar in top panel = 500 μm; bar in bottom panel = 200 μm. C, Expression of SDF1 and CXCR4 mRNA in the periosteum and the liver. Results are representative of 3 independent experiments.

High expression of SDF-1 protein in the periosteum of live bone grafts.

To examine the localization of SDF-1 protein expressed during the acute phase of live bone graft healing, immunohistochemical analysis was performed. The expression of SDF-1 protein was observed at the growth plate and the endosteum (results not shown), as previously reported (32). We also detected high expression of SDF-1 protein at the periosteum of the bone graft on day 2 (Figure 1B, bottom left). In contrast, no protein expression was observed at the periosteum of the host bone in the same section (Figure 1B, bottom right). To confirm whether SDF1 mRNA was expressed in the periosteum at the transcript level, we extracted total RNA from the periosteum of femoral shafts and observed the expression of SDF1 and CXCR4 mRNA (by PCR), as was observed in the liver as control (Figure 1C), supporting the notion that the periosteal cells are capable of expressing SDF-1 transcript and protein when injured. These results collectively demonstrate that the periosteum of the bone graft is the main source of SDF-1 production during the acute phase of structural bone healing.

In vitro and in vivo migratory effect of SDF-1 on mouse BMSCs.

SDF-1 has previously been shown to mediate the mobilization and migration of bone marrow–derived stem and progenitor cells in the sites of organ injury (20–25). Thus, we hypothesized that SDF-1 works as a chemoattractant of MSCs to bone-healing sites in endochondral bone repair. To verify this hypothesis, we performed in vitro and in vivo cell migration assays using primary mouse BMSCs. First, we evaluated the phenotypic characterization of primary cultured mouse BMSCs by flow cytometry. Although mouse BMSCs at postpassage 3 consisted of cells of heterogeneous origin, including CD34-positive and CD45-positive cells, which was consistent with a previous report (33), CD45-negative cells gradually overgrew with passaging, and the cells became homogeneously negative for CD34 and CD45 at postpassage 21 (additional information available from corresponding author). Moreover, mouse BMSCs at postpassage 24 expressed both SDF1 and CXCR4 mRNA, as confirmed by PCR, confirming the proficient reaction of these cells to SDF-1. We also confirmed that these cells maintained the capability of differentiation into osteogenic, chondrogenic, and adipogenic cells (data not shown). Based on these results, we decided to use mouse BMSCs from postpassage 21 to postpassage 25 for subsequent assays.

We examined the in vitro chemotactic effect of SDF-1 on mouse BMSCs, using a chemotactic chamber assay. Two different doses (10 ng/ml and 100 ng/ml) of rMuSDF-1 induced the migration of mouse BMSCs, by 80% and 224%, respectively, in a dose-dependent manner (Figure 2). Treatment with 10 nM and 100 nM TF14016, an antagonist for CXCR4, reduced the effect by 45% and 52%, respectively. This result suggests that the migration induced by SDF-1 is mediated by CXCR4. The chemokinetic assay, in which 100 ng/ml of rMuSDF-1 was added to both the upper and lower chambers, showed reduced migration (by 50%), indicating that mouse BMSCs migrated in response to a chemical gradient of SDF-1 (Figure 2). Furthermore, when 100 ng/ml of pertussis toxin was added to the chambers, the SDF-1–induced migratory effect was completely inhibited, indicating that this effect occurred via activation of Gi protein (results not shown). Taken together, the results demonstrate the functional effect of the SDF-1/CXCR4 axis on the in vitro migration of mouse BMSCs.

Figure 2.

In vitro chemotaxis assay. Mouse bone marrow–derived stromal cells (1.0 × 105/100 μl cultured in α-minimum essential medium with 1% [weight/volume] bovine serum albumin) were applied to the upper wells. The upper and lower wells were filled with medium supplemented with the indicated doses of recombinant murine stromal cell factor 1 (SDF-1) or TF14016. Cells that migrated to the undersurface of the membrane were fixed with methanol, stained with crystal violet, and counted under light microscopy in 4 randomly chosen fields at 400× magnification. Values are the mean and SEM results from >3 independent experiments. ∗ = P < 0.05.

Next, to investigate whether SDF-1 can recruit MSCs toward the sites of bone repair in the live bone graft model, we performed an in vivo migration assay. We first observed that 94.1 ± 1.0% (mean ± SD) of BrdU-labeled mouse BMSCs maintained the capability to adhere to the culture plate; 94.0 ± 1.8% of these cells were BrdU positive by immunocytochemical assay (data not shown). After the surgery, 2 × 105 BrdU-labeled mouse BMSCs were transplanted intravenously into the mice, and migration of the cells to sites around the graft bone was evaluated immunohistochemically on day 7. For the inhibition assay, the mice were treated with TF14016 or control vehicle. In the specimens treated with vehicle, positive BrdU staining was detected in the developing callus formed around the live graft bone, through normal skeletal repair (Figure 3A, top), indicating migration of transplanted mouse BMSCs to the graft bone. In contrast, treatment with TF14016 significantly reduced the number of migrated cells (Figure 3A, bottom), suggesting that in vivo migration of mouse BMSCs is indeed regulated by the SDF-1/CXCR4 axis. Furthermore, we observed that 25.2% of total chondrocytes in the fracture callus were BrdU positive in vehicle-treated specimens, while 15.4% were BrdU positive in TF14016-treated specimens (Figure 3B). These results demonstrate that migrated cells mediated by the SDF-1/CXCR4 axis actually differentiate into chondrocytes and participate in endochondral bone repair, which strongly corroborates our hypothesis that SDF-1 has a crucial role in the in vivo recruitment of MSCs toward bone repair sites during the early phase of bone repair.

Figure 3.

In vivo chemotaxis assay in the live bone graft model. A, Sections obtained on day 7 after surgery for histologic and immunohistochemical analysis, stained with Alcian blue and hematoxylin and eosin (left), and anti-bromodeoxyuridine (anti-BrdU) antibody (middle). Sections from phosphate buffered saline (PBS)– and TF14016-treated mice are shown in the upper and lower panels, respectively. Sections obtained for immunohistochemical analysis were counterstained with hematoxylin. Panels on the right show higher-magnification views of the boxed areas in the middle panels. Results are representative of 3 independent experiments. Bars in left panels = 200 μm; bars in middle and right panels = 50 μm. B, Percentages of BrdU-positive chondrocytes among total chondrocytes in the newly formed callus. Values are the mean and SEM results from 3 independent specimens. ∗ = P < 0.05 versus PBS. Color figure can be viewed in the online issue, which is available at http://www.arthritisrheum.org.

Reduced new bone formation by the inhibition of SDF-1.

To elucidate whether SDF-1 is indeed required for successful bone repair, we performed loss-of-function studies using 2 different reagents, anti–SDF-1 neutralizing antibody and TF14016. After the graft surgery, the mice received a series of intraperitoneal injections of anti–SDF-1 neutralizing antibody or continuous administration of TF14016, and then were killed for histologic analysis on days 14 and 7 after the surgery, respectively. In comparison with the control specimen, new bone formation was significantly reduced in anti–SDF-1 antibody– and TF14016-treated live bone grafts, by 81% and 57%, respectively, indicating that SDF-1 plays a critical role in normal bone repair (Figure 4).

Figure 4.

Loss-of-function study. A, Histologic sections obtained from a live bone graft model following systemic administration of anti–stromal cell–derived factor 1 (anti–SDF-1) neutralizing antibody (Ab) (middle) or control IgG (top) on day 14 after the surgery, stained with Alcian blue and hematoxylin and eosin. B, Histologic sections obtained from a live bone graft model following treatment with TF14016 (middle) and phosphate buffered saline (PBS) (top) on day 7. Results are representative of 4 independent experiments. Bars = 1 mm. Graphs (bottom) show areas of new bone formation around graft bones, as measured by computerized tracing. Values are the mean and SEM results from 4 independent specimens. ∗ = P < 0.05 versus PBS. Color figure can be viewed in the online issue, which is available at http://www.arthritisrheum.org.

Reduced bone formation in SDF-1+/− and CXCR4+/− mice and restored bone formation in an exchanging–bone graft model.

To further investigate the in vivo functional role of SDF-1, we created the live bone graft model in SDF-1+/− and CXCR4+/− mice, and new bone formation was evaluated on day 14. Radiologic analysis showed reduced callus formation in the live bone graft models of both SDF-1+/− and CXCR4+/− mice, compared with wild-type mice, in which sufficient callus was formed around the live bone graft (Figure 5A). In fact, histologic evaluation revealed that the area of new bone formation was significantly reduced in both SDF-1+/− and CXCR4+/− mice, by 55% and 65%, respectively (Figure 5B), which is consistent with the results described above.

Figure 5.

Autograft models in wild-type (WT), heterogeneous stromal cell–derived factor 1 (SDF-1+/−), and CXCR4+/− mice, and exchanging–bone graft models, in which a segmental live bone from a CXCR4+/− mouse was transplanted into a host SDF-1+/− mouse (C-to-S) or vice versa (S-to-C). A, Radiographs of bone graft models on day 14 after graft surgery. Results are representative of 4 independent specimens. B, Bone area measurements on day 14. Values are the mean and SEM results from 4 independent specimens. ∗ = P < 0.05.

Then, we created an exchanging–live bone graft model between SDF-1+/− and CXCR4+/− mice, in which a CXCR4+/− mouse–derived live bone segment was transplanted into an SDF-1+/− mouse and vice versa. Figure 5B shows that the CXCR4+/− mouse–derived live bone graft showed the capability of restoring the decreased bone formation in SDF-1+/− mice by 52%. In contrast, the SDF-1+/− mouse–derived live bone graft was not able to restore the impaired bone formation in the CXCR4+/− mouse, indicating a functional role of SDF-1 in skeletal bone repair.

DISCUSSION

In this study, we demonstrated that SDF-1 recruits MSCs during endochondral bone repair. We compared gene expression levels during the healing of live bone grafts and dead bone grafts and observed increased expression of SDF1 during the acute phase in the live graft model, whereas no remarkable increase was detected around dead bone. This differential increase in SDF1 expression suggests that this molecule might be a key regulator involved in successful bone repair. We next investigated the in vitro and in vivo chemotactic potency of SDF-1 and observed that SDF-1 promotes the migration of MSCs in vitro in a dose-dependent manner. We also verified that the BrdU-labeled mouse BMSCs injected intravenously were recruited to the live bone graft lesion, and, in addition, that this migration was inhibited by treatment with TF14016, an antagonist for CXCR4. These results strongly favor the notion that SDF-1 is an essential molecule for the migration of MSCs to sites of bone repair in vivo. Indeed, the bone grafts from CXCR4+/− mice showed the capability of restoring decreased bone formation in SDF-1+/− mice, but not vice versa.

Although the role of circulating MSCs in bone healing remains controversial, a recent study formally demonstrated the participation of circulating osteogenic connective tissue progenitor cells in a parabiotic mouse model of fracture healing (34). Moreover, another interesting study demonstrated that circulating bone marrow–derived osteoblast progenitor cells (MOPCs) are recruited to the bone-forming site via the SDF-1/CXCR4 axis in a bone-forming model of bone morphogenetic protein 2–induced ectopic bone formation (35). The effects of SDF-1 on MSCs in normal bone repair, however, have not yet been proven.

In this study, we demonstrated the migration of intravenously transplanted MSCs to the site of bone repair in a live bone graft model. The mobilization of BrdU-positive cells was observed around the graft bone, and the number of migrated cells was decreased by TF14016. These data support the existence of circulating MSCs and involvement of the SDF-1/CXCR4 axis in the migration of cells to the sites of bone repair. In fact, 2 recent studies demonstrated the expression of CXCR4 at the surface of CD45-negative MOPCs and human bone marrow–derived stromal stem cells (35, 36), and the MSCs used in the present study were derived from bone marrow and were negative for CD45. The expression of CXCR4 in MSCs at the mRNA level was confirmed by reverse transcription–PCR. Regarding the in vivo migration of BrdU-positive cells, a counter argument could be that SDF-1 merely recruited macrophages that phagocyted the transplanted BrdU-positive cells. However, 25.2% of total chondrocytes in the endochondral callus were BrdU positive, demonstrating chondrogenic differentiation of the migrated cells.

These observations support the notion that the migrated cells were not hematopoietic but mesenchymal cells and, what is more, actually participate in endochondral bone formation. To our knowledge, this study is the first to show an in vivo chemotactic function of SDF-1 on MSCs and the subsequent commitment of these cells to endochondral bone repair. Surprisingly, the percentage of BrdU-positive cells was higher than expected. This may be because the migrated cells proliferated and differentiated more rapidly than the resident cells, but this possibility remains to be elucidated.

One of the noteworthy findings is that blocking of SDF-1 or CXCR4 is directly connected to the decreased volume of newly formed bone. The loss-of-function studies revealed that anti–SDF-1 neutralizing antibody remarkably decreased the area of new bone formation on day 14 (Figure 4), and that treatment with TF14016 not only inhibited the migration of BrdU-labeled mouse BMSCs (Figure 3B) but also decreased the area of new bone formation on day 7 (Figure 4). Moreover, compared with that in wild-type mice, bone formation was significantly reduced in both SDF-1+/− and CXCR4+/− mice (Figures 5A and B). A clinical study showed that the volume of bone formation is related to the number of progenitor cells that were transplanted for the treatment of human fracture nonunion (9). Moreover, another recent study, using the same murine bone graft model, demonstrated statistically significant correlations between the graft and callus volume and the ultimate torque and torsional rigidity of the site of bone repair (37). Taken together, the results from the present study indicate that the SDF-1/CXCR4 axis plays a key role in the recruitment of MSCs to sites of bone healing and promotes successful endochondral bone regeneration.

The location and type of cells expressing SDF-1 during bone healing represent an intriguing and crucial question. To answer this, we performed gene expression and immunohistochemical analysis and observed that the periosteum of the long bones indeed expressed SDF1 mRNA, and that the expression of SDF-1 protein was highly increased in the periosteum of the live grafts, while no remarkable increase was observed either in the periosteum of the host bone or in that of the dead grafts (Figures 1B and C). It was previously reported that SDF-1 is expressed at the endosteum and the growth plate of normal long bones in adults (32), but a recent study showed that SDF-1 is expressed at the periosteum during embryonic endochondral bone development, and that expression is substantially reduced after birth (38). Combined with these reports, our results may lead to an interesting hypothesis that, in endochondral bone repair, progenitor cells in the periosteum regain the embryonic state and recruit MSCs through the up-regulation of SDF-1 expression, which is consistent with a well-accepted hypothesis that fracture healing recapitulates normal bone growth.

Another important issue is regulation of the expression of SDF-1 in bone repair. Increased expression of SDF1 mRNA was observed on days 2 and 3 in our models, although other studies had demonstrated an increase in SDF-1 expression within 24 hours after injury (22, 24, 25). The discrepancy in the peak point of SDF-1 expression may relate to the types of injuries. Many of the reported animal models are vascular injury types, in which oxygen tension changes rapidly (21–24). Because SDF-1 is reportedly regulated by a hypoxia-specific transcriptional factor, hypoxia-inducible factor 1, the expression of SDF-1 may increase rapidly after the blood supply is stopped in those models (25). However, any trophic vasculature was not apparently affected in our live bone graft model (33). Furthermore, it can be expected that infiltration from the neighboring bone marrow retains the blood supply to the graft bone to some extent. These unique conditions in our model may bring about the gradual hypoxic change of the injured lesion, resulting in the relatively delayed increase in SDF1 expression. Indeed, SDF1 was not up-regulated during the acute phase of the murine rib fracture model, in which the blood supply can be preserved well enough, and proper endochondral bone repair was observed (data not shown). The regulatory mechanism of SDF-1 expression in bone repair should be pursued further in future studies.

Finally, to prove the functional role of SDF-1 in MSC migration, we created an exchanging–bone graft model between heterozygous SDF-1 mice and CXCR4 mice. SDF-1−/− and CXCR4−/− mice die perinatally, with severe abnormalities affecting many organs including the hematopoietic system, cardiovascular system, and brain, although without major skeletal deformities. In contrast, SDF-1+/− and CXCR4+/− mice have been reported to grow up and appear normal in terms of skeletal development, similar to wild-type mice (13, 39). We observed that a CXCR4+/− mouse–derived bone graft restored impaired bone formation in SDF-1+/− mice. In contrast, an SDF-1+/− mouse–derived bone graft was not able to restore the potency of bone formation in CXCR4+/− mice. The immunohistochemical study showed SDF-1 being expressed in the periosteum of the graft bone, not host bone (Figure 1B), and these data strongly support the hypothesis that SDF-1 expressed from the graft bone, not from the host bone, can recruit MSCs to the sites of bone healing and plays a critical role in skeletal bone repair.

In conclusion, this study demonstrates, for the first time, that the SDF-1/CXCR4 axis plays a crucial role in the migration of MSCs and contributes to endochondral bone repair. SDF-1 is highly expressed in the periosteum of the live bone grafts during the acute phase of healing after surgery. SDF-1 recruits MSCs toward the graft lesion and allows them to participate in proper bone repair. Blockage of this axis inhibits cell migration and results in decreased callus formation. Moreover, CXCR4+/− mouse–derived bone grafts expressing SDF-1 restored impaired bone formation in SDF-1+/− mice, but not vice versa. Further understanding of the function of SDF-1 and the mechanism of endochondral bone repair is likely to lead to the development of a new, less-invasive strategy for the therapeutic use of SDF-1 to achieve successful bone repair.

AUTHOR CONTRIBUTIONS

Dr. Ito had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Kitaori, Ito, Schwarz.

Acquisiton of data. Kitaori, Ito, Tsutsumi, Oishi, Nakano, Fujii, Nagasawa.

Analysis and interpretation of data. Kitaori, Ito, Nakamura.

Manuscript preparation. Kitaori, Ito, Schwarz.

Statistical analysis. Kitaori.

Discussion. Kitaori, Ito, Schwarz, Yoshitomi, Nagasawa, Nakamura.

Acknowledgements

We are grateful to Prof. K. Tashiro (Department of Genomic Medical Sciences, Graduate School of Medical Science, Kyoto Prefectural University of Medicine) for thoughtful discussions. We thank Dr. Y. Kanegae (Laboratory of Molecular Genetics, Institute of Medical Science, University of Tokyo) and Drs. T. Aoyama and K. Fukiage (Department of Tissue Regeneration, Institute for Frontier Medical Sciences, Kyoto University) for their valuable technical assistance.

Ancillary