Cyclooxygenase-2 deficiency impairs muscle-derived stem cell-mediated bone regeneration via cellular autonomous and non-autonomous mechanisms

This study investigated the role of cyclooxygenase-2 (COX-2) expression by donor and host cells in muscle-derived stem cell (MDSC)-mediated bone regeneration utilizing a critical size calvarial defect model. We found that BMP4/green fluorescent protein (GFP)-transduced MDSCs formed significantly less bone in COX-2 knock-out (Cox-2KO) than in COX-2 wild-type (WT) mice. BMP4/GFP-transduced Cox-2KO MDSCs also formed significantly less bone than transduced WT MDSCs when transplanted into calvarial defects created in CD-1 nude mice. The impaired bone regeneration in the Cox-2KO MDSCBMP4/GFP group is associated with downregulation of BMP4-pSMAD1/5 signaling, decreased osteogenic differentiation and lowered proliferation capacity after transplantation, compared with WT MDSCBMP4/GFP cells. The Cox-2KO MDSCBMP4/GFP group demonstrated a reduction in cell survival and direct osteogenic differentiation in vitro. These effects were mediated in part by the downregulation of Igf1 and Igf2. In addition, the Cox-2KO MDSCBMP4/GFP cells recruited fewer macrophages than the WT MDSC/BMP4/GFP cells in the early phase after injury. We concluded that the bone regeneration capacity of Cox-2KO MDSCs was impaired because of a reduction in cell proliferation and survival capacities, reduction in osteogenic differentiation and a decrease in the ability of the cells to recruit host cells to the injury site.

Introduction

Muscle-derived stem cells (MDSCs) are capable of undergoing multipotent differentiation into myogenic, osteogenic and chondrogenic lineages, both in vitro and in vivo. Previous work by our team has shown that MDSCs transduced with BMP2 or BMP4 can promote ectopic bone formation and bone healing in vivo (1–3). It was found that MDSC-mediated bone regeneration was induced by direct differentiation of donor cells and the attraction of host cells via paracrine cellular effects imparted by the donor cells. We also found that cyclooxygenase-2 (COX-2) was dynamically expressed during the process of MDSC-mediated bone regeneration by both inflammatory cells and non-inflammatory mesenchymal cells (i.e. the transplanted MDSCs) (4); however, the role of dynamic COX-2 expression in the process remains unclear.

COX-2 is a rate-limiting enzyme that catalyzes the synthesis of prostaglandins from arachidonic acid and has primarily been found in areas of inflammation; hence, COX-2 has been targeted for the development of many selective and non-selective non-steroidal anti-inflammatory drugs. The discovery of these COX-2 inhibitors has greatly contributed to the development of numerous analgesic medications for pain relief; however, over the past decade, it has been shown that COX-2 inhibitors may interfere with fracture healing, although the specifics of its involvement in this complex process is currently unclear.

COX-2 knock-out (Cox-2KO) mice have been shown to undergo abnormal endochondral ossification in a femoral defect model and they also exhibit delayed and reduced bone fracture healing during both endochondral and intramembranous bone formation, though normal bone development in these mice does not appear to be affected significantly (5–7). More recently, it was demonstrated that COX-2 expression in the injury milieu was important for periosteal induced fracture healing and is expressed by both inflammatory and non-inflammatory periosteal cells (8). The impaired fracture healing in Cox-2KO mice has been associated with decreased cellular proliferation, reduced expression of MMP9 and inhibition of angiogenesis at the injury site. These adverse effects could be reversed using an E-type prostanoid receptor (EP) 4 agonist but not an EP2 agonist, indicating that COX-2-PGE2-EP4 is the major signaling pathway involved in the reduction of bone healing capacity seen in Cox-2KO mice (9). Likewise, numerous studies have shown that COX-2-selective inhibitors can impair fracture healing in a variety of mouse, rat and rabbit fracture models; however, there are also reports that COX-2 inhibitors have no obvious adverse effects on bone fracture healing (10–14). COX-2 gene therapy has been used to efficiently promote bone fracture healing in vivo utilizing a local injection of a human retroviral-COX-2 vector which can target proliferating periosteal cells. On the other hand, COX-2 ex vivo gene therapy failed to promote bone marrow stem cell-mediated bone repair in a critical size calvarial bone defect model (15). Furthermore, impaired fracture healing in aged mice has been associated with reduced intrinsic COX-2 expression (16).

The COX-2-PGE2 pathway is required for regulating energy homeostasis via the upregulation of UCP1 to induce the transition of brown adipose tissue into white adipose tissue, and is also implicated in regulating increased energy consumption. It also serves as a downstream effector of β-adrenergic signaling (17,18).

The role of COX-2 in stem cell function has also been reported. COX-2 is a major immune regulatory factor of human mesenchymal stem cells (19) and we have demonstrated in previous studies that both murine and human stem cells express COX-2 endogenously (4,20) and dynamically during the bone formation process. We showed that COX-2 was highly expressed in chondrocytes during the chondrocytic stage of MDSC-mediated endochondral bone regeneration (4); however, the role of COX-2 in MDSC-mediated bone regeneration is still unclear. Because COX-2 inhibitors are often used clinically for the relief of musculoskeletal pain, it is important to determine whether COX-2 inhibition will affect stem cell-mediated bone formation.

Results

MDSCs regenerated less bone in COX-2-deficient mice

MicroCT analysis demonstrated that BMP4/green fluorescent protein (GFP)-transduced MDSCs (isolated from C57BL/10J mice) could partially heal a critical size bone defect in wild-type (WT) mice; however, almost no bone regeneration was observed when the cells were implanted in Cox-2KO mice (P < 0.01) (Fig. 1A and B). Histological analyses showed that few donor cells (GFP) were present in the WT group (black arrows and enlarged in the insert), and no GFP-positive donor cells were present in the defect area in Cox-2KO mice (Fig. 1C). We observed that many lymphocytes were present at the regeneration site in the WT mice (red arrow), whereas the Cox-2KO mice formed scar tissue over the defect area. Herovici’s staining showed the formation of type I collagen (red color) in the defect area of the WT host group while almost no type I collagen was found in the Cox-2KO group, only type III collagen (blue) was present in the defect area.

Figure 1.

Bone regeneration capacity of BMP4/GFP-transduced MDSCs in WT and Cox-2KO mice. (A) MicroCT 3D reconstruction of a calvarial defect 8 weeks after transplantation of BMP4/GFP-transduced MDSCs. Partial bone formation was found in the defect areas in the WT mice whereas almost no bone formation was observed in the defect areas in the Cox-2KO mice. (B) Quantification of three-dimensional new bone volume in the defect areas indicated that the BMP4/GFP-transduced MDSCs regenerated significantly less bone in Cox-2KO mice compared with the WT mice. *P < 0.05 compared with the WT mice, N = 4. Wilcoxon rank sum test. (C) Few donor cells were detected in the defect area of the WT mice (Inset in C) and no donor cells were found in the defect area of Cox-2KO mice. Black arrows show GFP-positive cells and the red arrow shows lymphocyte infiltration in the area of the defect. No GFP-positive cells were found in the defect areas of Cox-2KO mice. (D) Herovici’s staining shows the expression of COL1 (in red) in the trabecular bone matrix within the defect area of WT mice. No COL1-positive matrix was observed in the defect area of the Cox-2KO mice and only a thin layer of blue COL3 matrix in the defect area was observed in the Cox-2KO mice. BM, bone marrow; Br, brain.

Cox-2KO MDSCs displayed a reduction in bone formation capacity when transplanted into a critical size defect created in CD-1 nude mice

We transduced two populations of WT MDSCs, and their BMP4 secretion levels at the time of transplantation were: 23.9 ± 3.6 and 23.3 ± 0.9 ng/million cells/24 h, respectively. We also transduced two populations of COX-2KO MDSCs with retro-BMP4/GFP, and their BMP4 secretion levels at the time of transplantation were: 22.9 ± 4.1 and 26.4 ± 6.0 ng/million cells/24 h, respectively. We characterized the cells by using real-time (RT)-PCR and flow cytometry. We found that both WT MDSCBMP4/GFP cells and Cox-2KO cells expressed NG2, CD140a, CD140b, CD44, CD73, stem cell antigen 1 (Sca-1) and paired box gene 7 (Pax-7); however, the expression levels of these cell markers varied among the populations. Furthermore, both WT MDSCBMP4/GFP and Cox-2KO MDSCBMP4/GFP expressed similar levels of osteogenesis-related genes, including Runt-related transcription factor X 2 and osterix; however, we found that alkaline phosphatase (Alp), a late-expressing osteogenic marker was expressed at a lower level in Cox-2KO MDSCBMP4/GFP cells than in WTMDSCBMP4GFP cells. Col1A1 expression also varied between the different populations of cells. Variation in the expression of Col1A1 was also evident between the different populations of WT MDSCBMP4/GFP cells and Cox-2KO MDSCBMP4/GFP cells; moreover, Alp and Col1A1 expression levels were very low compared with those of osteogenic transcription factors. We also investigated the expression of the cell survival genes Bcl-2, Bcl-XL and Sirt2, and found that Bcl-2 was expressed at a higher level in one WT MDSCBMP4/GFP population compared with the other three cell populations. There were no statistical differences between cell populations with respect to the expression of Bcl-XL. There was a statistical difference in Sirt1 expression between the WT MDSCBMP4/GFP#2 and Cox-2KO MDSCBMP4/GFP#1; however, no differences in expression were observed between other populations of cells (Supplementary Material, Fig. S1). Gene expression levels are also shown in Supplementary Material, Figure S2A. Flow cytometry analysis indicated heterogeneity in CD73 expression between cell populations; however, CD44 expression levels were similarly high in all groups (Supplementary Material, Fig. S2B).

We then transplanted the cells in vivo to investigate the bone regeneration capacity of WTMDSC/BMP4/GFP and Cox-2KOMDSCBMP4/GFP cells. In this experiment, we choose nude mice to exclude the effect of immune rejection, as COX-2 KO mice strains are generated from two different backgrounds of mice. We found that BMP4/GFP-transduced Cox-2KO MDSCs formed significantly less bone in the CD-1 nude mice than did BMP4/GFP-transduced WT MDSCs at 2-, 4- and 6-weeks post-implantation (Fig. 2A and B, showing one population from WT and Cox-2KO cells). Herovici’s staining revealed the formation of a very dense trabecular bone matrix –type I collagen (red color) in the WT MDSCBMP4/GFP group, whereas the density and quantity of type I collagen in the Cox-2KO MDSCBMP4/GFP group were greatly reduced (Fig. 2C). GFP immunohistochemistry showed that the majority of osteoblasts and osteocytes were derived from the donor cells as revealed by diaminobenzadine (DAB) staining of GFP (brown) in the WT MDSCBMP4/GFP group. In contrast, the Cox-2KO MDSCBMP4/GFP group had fewer cells that differentiated into osteoblasts and osteocytes. Surprisingly, we found that some of the cells forming strong GFP-positive patches actually exhibited muscle fiber morphology (Fig. 2D). H&E staining demonstrated typical trabecular bone consisting of bone matrix and bone marrow, where hematopoietic cells and blood vessels were obvious, in the defect area of the WT MDSCBMP4/GFP group. The Cox-2KO MDSCBMP4/GFP group formed much less bone and some of the cells may have undergone myogenic differentiation, indicating that there was a disruption of osteogenic differentiation signaling, even though the cells were transduced with BMP4 (Fig. 2E).

Figure 2.

The role of donor COX-2 in MDSCBMP4/GFP-mediated bone regeneration. (A) MicroCT 3D reconstruction showed almost complete healing of the defect using BMP4/GFP-transduced WT MDSCs at 2 weeks and completely healed at 6 weeks; but only partial healing by BMP4/GFP-transduced Cox-2KOMDSCs even by 6 weeks. (B) Three-dimensional new bone volume quantification by microCT indicated significantly less bone in the defect area of the Cox-2KO MDSCBMP4/GFP group compared with the WT MDSCBMP4/GFP group at all time points. **P < 0.0, ***P < 0.001 compared with the WT MDSC BMP4/GFP group, (N = 6). (C) Herovici’s staining showed dense COL1-positive trabecular bone (in red) in the WT MDSCBMP4/GFP group. In contrast, only sparse trabecular bone was found in the Cox-2KO MDSCBMP4/GFP group. (D) GFP immunohistochemistry in the WT MDSCBMP4/GFP group revealed that the donor cells were present within bone trabeculae where osteoblasts and osteocytes usually reside. In contrast, fewer GFP-positive cells were found in the bone defect area. Instead, GFP-positive muscle fibers were detected in the defect area. (E) H&E staining demonstrates typical trabecular bone in the WT MDSCBMP4/GFP group. In contrast, lower trabecular bone formation was observed in the defect area in the Cox-2KO MDSCBMP4/GFP group, but some muscle fibers could be identified in the defect area in this group (Inset).

Impaired bone regeneration capacity of Cox-2KO MDSCs is associated with a reduction in chondrogenic-osteogenic differentiation signaling

In order to investigate the mechanism(s) involved in the impairment of bone regeneration exhibited by the BMP4/GFP-transduced Cox-2KO MDSCs, we analyzed bone regeneration at 3, 7 and 14 days after cell transplantation in the calvarial bone defect area. Von Kossa staining revealed thick trabecular bone formation after 14 days in the WT MDSCBMP4/GFP group, whereas only a thin layer of trabecular bone was formed in the COX-2KO MDSCBMP4/GFP group (Fig. 3A). Alcian blue staining revealed reduced chondrogenic differentiation and endochondral bone formation at 7 and 14 days in the Cox-2KO MDSCBMP4/GFP group compared with the WT MDSCBMP4/GFP group (Fig. 3B). Phosphorylated SMAD1/5 (pSmad1/5) was also significantly decreased in the Cox-2KO MDSCBMP4/GFP group at 7 days after cell transplantation, compared with WT MDSCBMP4/GFP cells (Fig. 3C and D). Immunofluorescent staining also indicated that the pSmad1/5/GFP positive cells in the Cox-2KO MDSCBMP4/GFP group were significantly lower than that of WTMDSCBMP4/GFP group (Supplementary Material, Fig. S3A and B). Western-blot results also showed a significant decrease in the expression of osteocalcin (an osteoblastic matrix marker) in the Cox-2KO MDSCBMP4/GFP group compared with the WT MDSCBMP4/GFP group (Fig. 3E and F).

Figure 3.

Downregulation of BMP4-pSMAD1/5 signaling affects bone regeneration mediated by Cox-2KO MDSCs. (A) Von Kossa staining at the 14-day time point revealed thicker mineralized bone tissue in the defect area of the WT MDSCBMP4/GFP group; only a thin layer of mineralized bone tissue was detected in the defect area of the Cox-2KO MDSCBMP4/GFP group (brown black staining). (B) Alcian blue staining at the 7- and 14-day time points. Fewer Alcian blue-positive cells (blue) were found in the Cox-2KO MDSCBMP4/GFP group than in the WTMDSCBMP4/GFP group at 7 days. By 14 days, less Alcian blue-positive matrix was found in the trabecular bone than that of WTMDSCBMP4/GFP group. Black arrow points to the cells of a myofiber-like structure. (C) Western-blot images of pSmad1/5 at day 7. (D) The pSmad 1/5 expression was significantly reduced in the Cox-2KO MDSCBMP4/GFP group compared with the WT MDSCBMP4/GFP group. (E) Western-blot image of osteocalcin at day 14. (F) Quantitative analysis revealed a reduced osteocalcin expression in the Cox-2KO MDSCBMP4/GFP group compared with WT MDSCBMP4/GFP group. *P ≤ 0.05 compared with the WT MDSCBMP4/GFP group.

Cox-2KO MDSCs exhibited less-efficient osteogenic and chondrogenic differentiation capacities in vitro

In order to investigate whether the impaired bone regeneration capacity of Cox-2KOMDSCs is cell autonomous, we performed in vitro multipotent differentiation experiments on the BMP4/GFP-transduced WT and Cox-2KO MDSCs. Our results indicate that osteogenesis in the BMP4/GFP-transduced Cox-2KO MDSCs was significantly reduced as demonstrated by microCT (Fig. 4A and B). Von Kossa staining showed that the mineralization pattern in Cox-2KO MDSCBMP4/GFP pellets was confined to the edges of the pellets. In contrast, the WT MDSCBMP4/GFP underwent diffuse mineralization throughout the pellets (Fig. 4C). Osteocalcin immunohistochemistry also demonstrated much lower terminal osteogenic differentiation in Cox-2KO MDSCBMP4/GFP cells than in WT MDSCBMP/GFP cells (Fig. 4D). Herovici’s staining indicated the presence of collagen type 1 at the periphery of the osteogenic pellets in the WT MDSCBMP4/GFP group, but almost no type I collagen (red staining) was observed in the Cox-2KO MDSCBMP4/GFP pellets (Fig. 4E). The results of chondrogenic differentiation indicate that BMP4/GFP-transduced WT MDSCs formed typical chondrocytes as shown by Alcian blue staining. In contrast, only a few cells developed into chondrocytes in the Cox-2KO MDSC BMP4/GFP group (Fig. 4F). Collagen 2A1 (Col2A1) staining demonstrated chondrogenic differentiation in the pellets of the WT MDSCBMP4/GFP group, whereas only a few Col2A1-positive cells were found in the Cox-2KOMDSCBMP4/GFP pellets (Fig. 4G); however, the results of myogenic differentiation showed that the BMP4/GFP-transduced Cox-2KO MDSCs formed larger fast myosin heavy chain (fMHC)-positive myotubes (orange color or red color), whereas the BMP4/GFP-transduced WT MDSCs underwent limited fusion. The majority of the cells that underwent myogenic differentiation remained mononucleated in the WT MDSCBMP4/GFP group (Fig. 4H). Quantification of fMHC-positive cells demonstrated that there were significantly more fMHC-positive cells in the Cox-2KO MDSC group, (normalized to the total number of cells, i.e. total number of DAPI-positive cells), than in the WT-MDSCBMP4/GFP cells (Fig. 4I).

Figure 4.

Multipotent differentiation of BMP4/GFP-transduced MDSCs. (A) 3D microCT images of pellet mineralization. (B) Quantification of the mineralized matrix volume showed a significant decrease in the Cox-2KO MDSCBMP4/GFP group compared with WT MDSCBMP4/GFP group. (C) Von Kossa staining of the pellet sections showed extensive mineralization in the WT MDSCBMP4/GFP group, but only focal mineralization on the edge of the pellet in Cox-2KO MDSCBMP4/GFP group. (D) Osteocalcin immunohistochemistry revealed abundant osteocalcin-positive cells in the pellets generated by the WT MDSCBMP4/GFP group compared with fewer positive cells in the Cox-2KO MDSCBMP4/GFP group. (E) Herovici’s staining for collagen type 1. BMP4-transduced WT-MDSCs showed collagen type 1 as red in the periphery of the pellets. Almost no collagen type 1 was identified in the BMP4/GFP-transduced Cox-2KOMDSCs. Collagen type 3 appears dark blue and other tissues appear yellow. (F) Alcian blue staining of the chondrogenic pellets. BMP4/GFP-transduced WT MDSCs pellets contained typical (hypertrophic) chondrocytes and stained positive with Alcian blue while the Cox-2KOMDSCBMP4/GFP group minimally formed a chondrogenic matrix. (G) Strong COL2A1-positive matrix and chondrocytes were detected throughout the pellets in the WTMDSCBMP4/GFP group compared with focal COL2A1 positivity in the Cox-2KO MDSCBMP/GFP group. (H) Myogenic differentiation. Fast myosin heavy chain (fMHC) and GFP double immunofluorescent staining showed that the BMP4/GFP-transduced MDSCs could still undergo myogenic differentiation. The BMP4/GFP-transduced Cox-2KO MDSCs were more myogenic and formed more multinucleated myotubes than BMP4/GFP-transduced WT MDSCs. (I) Quantification of fMHC-positive cells including orange (fMHC and GFP double-positive) or red color cells (fMHC + GFP dim) are counted. *P < 0.05 compared with the WT MDSCBMP4/GFP group.

Cox-2-deficient MDSCs evoked a reduced inflammatory response and postponed inflammatory resolution during MDSC-mediated bone regeneration

The number of CD68⁺macrophages was significantly reduced in the bone defect area implanted with Cox-2KOMDSCBMP4/GFP cells compared with the WTMDSCBMP4/GFP cells at day 3. There were no differences between the two groups at day 7; however, the number of CD68⁺macrophages was significantly higher on day 14 in the Cox-2KOMDSCBMP4/GFP group (Fig. 5A and B). There were no differences in the number of Gr-1⁺neutrophils in the defect areas at any of the time points (Fig. 5C and D). These results indicate that COX-2 expression in the donor cells plays a role in attracting macrophages early after injury and implantation of the stem cells.

Figure 5.

Effect of COX-2 deficiency on the chemo-attraction of host inflammatory cells. (A) CD68 immunofluorescence staining for macrophages at different time points. (B) Quantification of CD68⁺macrophages in the defect area at different time points indicated that the infiltration of macrophages was reduced in the Cox-2KOMDSCBMP4/GFP group compared with WT MDSCBMP4/GFP group 3-day post-implantation; however, the observation of increased macrophage infiltration at the 14-day time point indicated a sustained occupancy of macrophages in the defect area. (C) Gr-1 immunofluorescence for neutrophils. (D) Quantification of Gr1⁺neutrophils at 3, 7 and 14 days. No statistically significant differences were found between the two groups at any time point. *P < 0.05; **P < 0.01 compared with the WTMDSCBMP4/GFP group. De, defect; Br, brain. Scale bar = 100 µm for all pictures.

Cox-2-deficient donor cells do not significantly affect the chemo-attraction of host angiogenic cells

In order to investigate whether COX-2 gene deficiency affects the paracrine effects imparted by the transplanted MDSCs on the chemo-attraction of host endothelial cells and angiogenesis, we performed CD31 staining. There were no significant differences in the number of CD31-positive cells detected between the two groups on days 3 or 7; however, we did observe a reduction in CD31-positive cell number in the Cox-2KOMDSCBMP4/GFP group on day 14 (Fig. 6A and B). CD31-positive cells were not colocalized with GFP, which indicated that they were host derived. Similarly, our in vitro experiment showed that conditioned medium from both WT and Cox-2KO MDSCs could promote MS1 murine endothelial cell proliferation after 72 h of culture; however, there was no significant difference between the WT and Cox-2KO groups in terms of promoting endothelial cell proliferation (Fig. 6C).

Figure 6.

Effect of COX-2 deficiency on the chemo-attraction and proliferation of angiogenic cells by the BMP4/GFP-transduced MDSCs. (A) CD31 staining for endothelial cells at different times points. De, defect; Br, brain. Insets are enlarged images of the boxed areas to show CD31⁺capillary vessels. (B) Quantification of CD31-positive cells in the defect area indicated no differences at 3 and 7 days, but a reduction in the Cox-2KO MDSCBMP4/GFP group at 14 days. (C) MS1 endothelial cell proliferation with conditioned media from un-transduced WT MDSCs and Cox-2KO MDSCs. At 48 h, one population of WT MDSCs and one population of Cox-2KO MDSCs enhanced endothelial cell proliferation in the proliferation medium (PM) compared with the unconditioned medium. *P < 0.05, **P < 0.01. At 72 h, two populations of WT MDSCs and two populations of Cox-2KO MDSCs enhanced endothelial cell proliferation in the proliferation medium (PM) compared with the unconditioned medium. No statistical differences in proliferation were observed between WT and Cox-2KO MDSCs. #P < 0.05, ##P < 0.01. Scale bar = 100 μm for all pictures.

Cox-2KO MDSCs exhibited reduced proliferation capacity in vivo

We performed staining for Ki67 in order to investigate whether the proliferation capacities of the MDSCs were impaired after in vivo transplantation due to a lack of COX-2 expression. Ki67 staining indicated that there were significantly fewer Ki67 ⁺ GFP⁺cells present in the Cox-2KO MDSCBMP4/GFP group than in the WTMDSCBMP4/GFP group at all time points. Moreover, the ratio of Ki67⁺GFP⁺cells to the total number of GFP⁺cells (Ki67 ⁺ GFP⁺/GFP⁺) was also significantly lower in the Cox-2KOMDSCBMP4/GFP group compared with the WT MDSCBMP4/GFP group at all analyzed time points (Fig. 7A and C). Furthermore, western-blot analysis of the cell survival gene phosphorylated Akt (pAkt) also demonstrated significantly decreased levels of pAkt in the defect area at 7 days after cell transplantation (Fig. 7D and E).

Figure 7.

Effect of COX-2 expression on the proliferation of MDSCs during endochondral bone formation in vivo. (A) Ki67 immunofluorescence for cell proliferation. De, defect; Br, brain. (B) Quantification of Ki67 ⁺ GFP⁺/GFP⁺ratio indicated there were smaller percentages of proliferating donor cells in the Cox-2KOMDSCBMP4/GFP group in comparison to the WT MDSCBMP4/GFP group. (C) Quantification of overall Ki67 ⁺ GFP⁺cells in the defect area indicated that Cox-2KO MDSCs did not proliferate as effectively as WTMDSCs in the injury milieu. *P < 0.05; **P < 0.01; ***P < 0.001 compared with the WTMDSCBMP4/GFP group. (D) Western-blot images of pAkt. (E) Quantification of band density indicated a significant decrease of pAkt at 7 days in the defect area of Cox-2KOMDSCBMP4/GFP group compared with the WT MDSCBMP4/GFP group at 7-day post-implantation. *P ≤ 0.05.

Cox-2KO MDSCs exhibit reduced resistance to oxidative stress

In order to determine why BMP4/GFP-transduced Cox-2KO MDSCs did not proliferate efficiently in vivo, we exposed BMP4/GFP-transduced WT and Cox-2KOMDSCs to 250 and 400 µm H₂O₂ and measured cell survival using a live cell imaging system. We found that the cell survival rate of the Cox-2KOMDSCBMP4/GFP group was significantly lower than the WTMDSCBMP4/GFP group at both H₂O₂ concentrations at all time points over a 24-h period (Fig. 8A and B).

Figure 8.

Cell survival and gene expression under oxidative and inflammatory stress conditions. (A and B) Cell survival rate in the presence of 250 and 400 μm H₂O₂. The survival rate of Cox-2KOMDSCBMP4/GFP cells was significantly lower than that of WTMDSCBMP4/GFP cells at both concentrations of H₂O₂ at all time points compared with WTMDSCBMP4/GFP. **P < 0.01; ***P < 0.001. (C and D) Gene expression of Cox-2KOMDSCBMP4/GFP and WTMDSCBMP4/GFP cells in the presence of 250 μm H₂O₂ for 4 h. The expression of Pdgf1, Igf1, Igf2 and Sod1 was significantly downregulated and the expression of Ucp1 was upregulated in the Cox-2KOMDSCBMP4/GFP group compared with WTMDSCBMP4/GFP group in the presence of H₂O₂. *P < 0.05; **P < 0.01; ***P < 0.001. (E and F) Gene expression after exposure to 100 ng/ml of TNFα. The mRNA expression levels of Mcp1, Hgf, Vegfa, Fgf2 and Cat mRNA were significantly increased and those of Igf1, Igf2 and Gpx1 were significantly decreased in the Cox-2KO MDSCBMP4/GFP group compared with the WT MDSCBMP4/GFP group. *P < 0.05; **P < 0.01; ***P < 0.001.

Cox-2KO MDSCs exhibit downregulation of Igf1 and Igf2 under oxidative and inflammatory stress conditions

Semi-quantitative RT-PCR results indicated that the Cox-2KOMDSCBMP4/GFP group had lower expression levels of platelet-derived growth factor β (Pdgfβ), insulin-like growth factor 1 (Igf1), insulin-like growth factor 2 (Igf2) and copper zinc superoxide dismutase 1 (Sod1); however, displayed an upregulation of uncoupling protein 1 (Ucp1) compared with the WT MDSCBMP4/GFP group when exposed to 250 μm H₂O₂. No significant differences in the expression of monocyte chemotactic protein 1 (Mcp1), hepatic growth factor (Hgf), vascular endothelial cell growth factor A (Vegfa), fibroblast growth factor 2 (Fgf2), transforming growth factor β 1 (Tgfβ1), glutathione peroxidase 1 (Gpx1) or catalase (Cat) was detected (Fig. 8C and D). The Igf-1, Igf2, Ucp1 expression were also verified with quantitative RT-PCR and the results were consistent with semiquantitiative PCR results (Supplementary Material, Fig. S4A). Furthermore, when the Cox-2KO MDSCBMP4/GFP cells were exposed to 100 ng/ml of TNFα, the Mcp1, Hgf, Vegfa, Fgf2 and Cat, expression levels were higher compared with the WT MDSCBMP4/GFP cells; however, Igf1, Igf2 and Gpx1 were downregulated (Fig. 8E and F). We further performed quantitative PCR for Mcp1, Hgf, Fgf2, Igf-1, Igf-2, Gpx1 and Cat and found statistical results similar to those of semi-quantitative PCR (Supplementary Material, Fig. S4B). In addition, we found that Igf1 and Igf2 were consistently lower in the Cox-2KO MDSCBMP4/GFP group than in the WTMDSCBMP4/GFP group when exposed to either H₂O₂ or TNFα. Therefore, we further tested the expression of these factors in untransduced and BMP4/GFP-transduced WT MDSCs and Cox-2KO MDSCs without any stress treatment. Interestingly, we found that Igf1 and Igf2 expression was consistently lower in Cox-2KO MDSCs than in WT MDSCs. There were no differences in the expression of any of the other genes tested between the groups; therefore, the downregulation of Igf1 and Igf2 could be inferred to be an intrinsic alteration in the Cox-2KO MDSCs, because this downregulation was observed in both transduced and untransduced cells (Supplementary Material, Figs S5 and S6).

Discussion

The bone regeneration mediated by the MDSCs in the calvarial bone defect model is a process that recapitulates postnatal bone fracture repair. Previously, we showed that murine MDSCs transduced with BMP4 could efficiently regenerate bone in a critical size bone defect model (1,21). The transplanted cells not only directly differentiated into osteoblast cells, but also influenced the host cell response in the injury milieu via paracrine factors (a non-autonomous cellular effect) (4). One of the factors that we found to be dynamically expressed during the bone regeneration process in both the transplanted donor cells and host cells was COX-2; however, the role that COX-2 plays in the bone regeneration process has not been thoroughly elucidated and prompted the current investigation. In this study, we found that BMP4/GFP-transduced MDSCs formed significantly less bone in the defect area of Cox-2KO host mice compared with the WT mice. Strikingly, when we transplanted BMP4/GFP-transduced WT MDSCs and Cox-2KO MDSCs into a defect area created in CD-1 nude mice, we found that bone regeneration was dramatically less in the Cox-2KO MDSCBMP4/GFP group than in the WT MDSCBMP4/GFP group, even though both cell populations secreted similar levels of BMP4.

Numerous studies have investigated the role of COX-2 in fracture healing; however, far fewer studies have focused on the role that COX-2 plays in tissue regeneration, especially stem cell-mediated bone regeneration. The role of the COX-2-PGE2 signaling pathway in bone fracture repair has been well described over the past decades (5–7); however, the cells that actually express COX-2 and play a role in bone fracture healing remain unknown. In a previous study, we showed that COX-2 was dynamically expressed during MDSC-mediated endochondral bone formation and that both inflammatory and donor cells expressed COX-2 by day 3 after the creation of a parietal bone defect and implantation of the donor cells. By day 7, COX-2 was highly expressed in chondrocytes derived from the transplanted MDSCs. When bone formation began, COX-2 expression was downregulated in the osteoblasts derived from transplanted MDSCs. By day 21, trabecular bone was formed, and COX-2 expression was markedly downregulated in both donor and host cells (4).

Other investigators have also characterized the identity of COX-2-expressing cells in fracture healing. Recently, a study using Col2 Cre and Prx1 Cre transgenic mice that were crossed with Cox-2^f/f to generate COX-2 conditional knock-out mice, in which COX-2 is specifically knocked-out in chondrocytes and mesenchymal cells, respectively, demonstrated that fracture healing was impaired, thus indicating the importance of COX-2 expression in these cells for bone fracture healing (22). Another recent study by O’Connor’s group (23) investigated the five arachidonic acid metabolism enzymes immunohistochemically during fracture healing and found that the expression level of COX-2 was very low in periosteal osteoblasts and in newly formed bone; however, COX-2 was abundantly expressed by osteoclasts in the early stage of fracture healing, and was highly expressed in chondrocytes. Therefore, the authors concluded that the expression of COX-2 and other arachidonic acid enzymes in osteoblasts was not essential for osteoblast function. COX-2 and other enzymes expressed in leukocytes, chondrocytes and osteoclasts, which are in close proximity to osteoblasts and mesenchymal cells during fracture healing, may regulate osteoblast proliferation and differentiation via cell-dependent signaling; therefore, the authors suggested a new paradigm where osteoclasts express COX-2, which is involved in the regulation of fracture healing. Coincidentally, another study revealed that local injection of lenti-COX-2 virus at the site of a bone fracture could accelerate fracture healing by enhancing the recruitment of mesenchymal stem cells, reducing cartilage callus formation and enhancing angiogenesis at the fracture healing site (24). More interestingly, these authors also found that COX-2 gene therapy enhanced the bone remodeling enzymatic activities of TRAP and Cathepsin K; therefore, they concluded that COX-2 gene therapy might convert endochondral bone fracture healing to intramembraneous bone fracture healing. On the other hand, it is possible that initial bone resorption is enhanced because of an increase in osteoclast activity and not necessarily because of bypassing the process of endochondral bone formation. These studies further emphasize that COX-2 expression is not limited to osteoblasts.

In order to determine the role of COX-2 in MDSC-mediated bone regeneration, we used a strategy to define the donor (mesenchymal cells or osteogenic progenitors) and host cells separately. First, we demonstrated that when normal MDSCs expressing BMP4/GFP were transplanted into Cox-2KO mice, new bone formation was significantly reduced, compared with when these cells were implanted into WT mice. These results were consistent with the previous studies outlined above, which showed that COX-2 expression by other cells types beyond osteogenic progenitor cells also play a role in new bone formation; however, in the current study, there was much less newly formed bone compared with our previous report, even in the WT mice (4,21). We believe this reduction in bone regeneration may have been the result of a graft-host response, because this was an allogenic transplantation, and indeed, we found many lymphocytes in the area of newly formed bone in the WT mice (Fig. 1C).

Furthermore, we investigated the role of COX-2 expression in osteogenic progenitor cells (mesenchymal stem cells) during MDSC-mediated bone regeneration by transplanting BMP4/GFP-transduced WTMDSCs and Cox-2KOMDSCs into CD-1 nude mice. We used CD-1 nude mice to minimize immune rejection of the transplanted cells by animal host, because the COX-2 mice are generated from mice of mixed background (129P2:B6). In this critical size defect model, the transplanted MDSCs served as the osteogenic progenitor cells after removal of the host periosteum during the creation of the critical size calvarial bone defect. We found that the BMP4/GFP-transduced Cox-2KO MDSCs also formed significantly less bone compared with that of the WT MDSCs even though they expressed similar levels of BMP4. This result indicates that the expression of COX-2 by the donor cells (osteogenic progenitors) plays a role in the bone formation process. Indeed, we found that BMP4/GFP-transduced Cox-2KO MDSCs did not differentiate into osteogenic lineages as efficiently as BMP4/GFP-transduced WT MDSCs. In fact, some of the cells actually differentiated into a myogenic lineage as evidenced by the formation of GFP-positive myofibers within the calvarial defect. This phenomenon could be explained by the fact that MDSCs are muscle derived and tend to intrinsically undergo myogenic differentiation.

In order to address whether these cellular effects were autonomous, we performed an in vitro multipotent assay and found a dramatic reduction in the osteogenic and chondrogenic potentials of the Cox-2KO cells while retaining their myogenic propensity despite their expression of BMP4. Similarly, another study has shown that COX-2-deficient periosteum derived stem cells also had a diminished capacity for in vitro osteogenic and chondrogenic differentiation, both with and without BMP2 stimulation (22). These latter results support the essential role of autonomous expression of COX-2 by osteogenic progenitor cells during the process of new bone formation. On the other hand, another study found that ex vivo COX-2 gene therapy failed to promote bone regeneration (15); however, based on a newly published study, this negative result may be explained by the fact that the expression of COX-2 in the osteoprogenitor cells requires spatiotemporal control (22). Hence, the persistent, constitutive expression of large amounts of COX-2 would have failed to promote bone regeneration. The cell autonomous role of COX-2 is also supported by our results, in that Cox-2KOMDSCs did not proliferate as efficiently as the WTMDSCs after transplantation. Because there were no differences in the proliferation capacities between the Cox-2KO and WTMDSCs in vitro when cultured in proliferation medium, we concluded that COX-2 expression was important for maintaining the abilities of the cells to tolerate environmental stress in vivo. This interpretation is further validated by our in vitro findings in which Cox-2KOMDSCs exhibited significantly reduced survival capacities compared with WTMDSCs in the presence of oxidative stress.

We have previously shown that transplanted MDSCs can differentiate into chondrocytes, osteoblasts and osteocytes (4); however, we do not know whether this is sequential differentiation or whether the cells differentiate into multiple lineages. The current study showed that the formation of intermediate cartilage was reduced in parallel to a reduction in new bone formation. This finding was consistent with our in vitro results whereby Cox-2KOMDSCs did not effectively differentiate into chondrogenic cells. Initially, it was thought that COX-2 deficiency induced the redirection of endochondral bone formation toward intramembraneous bone formation. However, we found residual Alcian blue-positive cartilage within the Cox-2KOMDSCBMP4/GFP group; therefore, we believe that the decrease in cell autonomous osteogenic and chondrogenic differentiation subsequently decreased new bone formation. Additionally, we found that Cox-2KOMDSCs did not survive as well as WTMDSCs under oxidative stress conditions in vitro, further supporting our belief that the impaired bone formation capacity of the Cox-2KO MDSCs in vivo is, at least in part, because of autonomous cellular mechanisms.

Furthermore, we asked the question whether COX-2 deficiency could influence the paracrine effects that MDSCs impart on the host cells and performed CD68 (macrophages) and Gr-1 (neutrophils) immunostaining. We found that the Cox-2KOMDSCs attracted fewer macrophages to the injury site during the early stages of bone healing, which is most likely attributable to COX-2 deficiency. Importantly, we ruled out the involvement of macrophage chemokine-monocyte chemotactic protein-1, a major macrophage chemotactic factor, as the cause of the differential attraction of the macrophages by measuring its expression levels in vitro and found similar expression levels by both groups regardless of whether they were cultured under oxidative or inflammatory stressors.

We also tested the role that COX-2 plays in angiogenesis by performing CD31 staining, another event that is critical for new bone formation. There were no significant differences between the Cox-2KO MDSCBMP4/GFP and WT MDSCBMP4/GFP groups in their ability to recruit endothelial cells to the injury site during the early stages of the healing process (3- and 7-day post-surgery); however, a decrease in vascular endothelial cell accumulation was observed at day 14 when trabecular bone formation had begun. This finding is most likely not due to the paracrine effects imparted by the cells, but rather the decline of vascular endothelial cells may have been a downstream consequence of impaired new bone formation. This hypothesis was confirmed by our in vitro results which showed that the conditioned medium from the Cox-2KO MDSC cultures was similar to the conditioned medium obtained from the WT MDSC cultures in terms of their ability to promote endothelial cell proliferation.

Since the Cox-2KO MDSCs exhibited an impaired capacity for bone regeneration because of their poor proliferation capacity and reduced resistance to oxidative stress, we performed gene expression analysis in which the cells were exposed to H₂O₂ (oxidative) and TNFα (inflammatory) stressors and found a consistent reduction of Igf1 and Igf2 expression under both treatment conditions. We further investigated whether the decreased expression of Igf1 and Igf2 was the result of H₂O₂ and TNFα exposure by testing the expression of these genes in both BMP4/GFP-transduced and untransduced cells without stressing the cells. We found that both genes were expressed at reduced levels in both groups, and therefore concluded that this reduction in Igf1 and Igf2 expression was because of the COX-2 deficiency in the MDSCs. As both of these genes have been shown to be very important for bone development and bone mass maintenance, their reduction in expression is at least one of the reasons why the Cox-2KOMDSCs exhibit an impaired capacity for new bone formation. IGF1 is the most abundant growth factor in the bone matrix and is important for the maintenance of bone mass throughout adulthood. IGF1 released from the bone matrix stimulates osteoblastic differentiation of mesenchymal stem cells and is mediated by the activation of the Igf1-mTOR signaling pathway (25). Osteocyte-derived Igf1 also plays a major role in bone formation by mechanical stimulation through the activation of Wnt signaling (26) and has been shown to promote osteoblast, osteoclast and chondrogenic differentiation via the activation of ephrin B2/EphB4-mediated cell-cell communication (27). In addition, IGF1 has been shown to promote bone defect healing and regeneration when appropriately delivered to the injury site (28). Igf2 is another gene that is important for osteogenic and osteoclastic differentiation in bone. Mice with a knock-out of their Igf2-promoter 2 (Igf2-P2) exhibit thin, short bones with a reduction in mineralization and their medullar cavities with the addition of altered bony remodeling and repair. These abnormalities are associated with a reduction in the number of embryonic mesenchymal chondroprogenitors, adult mesenchymal stem cells and osteoprogenitors. Osteopontin is a target of the IGF2 signaling pathway during the differentiation of osteoprogenitors into osteoblasts and is impaired in Igf2-P2 mutant mice. Igf2-P2 mutant mice also display an impaired ability to form giant osteoclasts because of a defective microenvironment (29). Igf2 is capable of inducing osteoblastic differentiation of parthenogenic embryonic stem cells in vitro and promoting bone regeneration in a critical size calvarial bone defect model (30). Igf2 also regulates osteoblastic differentiation by upregulating CXC chemokine ligand 7 (CXCL7) expression in stromal cells and stromal cell-derived factor 1 (SDF1) expression in osteoblastic cells (31). The local injection of IGF2 into the femurs of mice was found to significantly lengthen the treated femurs and elevate their bone mineral densities (32). Furthermore, it has been demonstrated that Igf2 promotes human MSC osteoblastic differentiation via cross-talk with integrin 5 (33). Due to the important roles of Igf1 and Igf2 during the process of bone development, bone formation and bone remodeling, we strongly suspect that the decreased Igf1 and Igf2 expression in the Cox-2KO MDSCs contribute, at least partially, to their impaired bone regeneration capacity.

In summary, this study revealed that COX-2 expression by both transplanted donor cells and host cells are important for MDSC-mediated bone regeneration, although the donor cells appear to exert a greater influence in this process. The impaired bone regeneration capacity of the Cox-2KO MDSCs was because of a reduction in cell proliferation and survival capacities, less efficient osteogenic differentiation (an autonomous cellular effect) and a reduction in the donor cells’ ability to recruit host inflammatory cells to the injury site (a non-autonomous cellular effect). The downregulation of Igf1 and Igf2 in the Cox-2KO cells may contribute to the impairment of bone regeneration mediated by Cox-2KO MDSCs.

Materials and Methods

Cell isolation and retro-viral BMP4/GFP gene construction and transduction

One population of MDSCs was isolated from 3-week-old C57BL/10J mice using a modified preplate technique (34). Three populations of MDSCs were also isolated from the skeletal muscle of 5-week-old B6;129P2-Ptgs2^tm1Unc(WT/WT) (COX-2 WT) and B6;129P2-Ptgs2^tm1Unc (KO/KO) Cox-2KO mice, respectively. A human BMP4 gene was engineered into a retro-viral vector upstream of a GFP gene. These genes were separated by an internal ribosome entry site (IRES) which allowed for individual expression of the BMP4 target gene and GFP. The retrovirus was then packaged in GP293 cells (ATCC, Manassas, VA, USA). MDSCs were transduced with retro-BMP4/GFP 3 times in a 24-h cycle in the presence of polybrene (8 µg/ml) at a multiplicity of infection of 5 (Sigma–Aldrich, Milwaukee, WI, USA). The cells were then expanded several times, after which the GFP⁺cells were selected by fluorescence-activated cell sorting (FACS, BD FACSAria IIu, Bedford, MA, USA). The secretion of BMP4 in the supernatant of the BMP4/GFP-transduced MDSCs was measured with a Human BMP-4 Quantikine ELISA Kit (DBP400, R&D System, Minneapolis, MN, USA), and the amount of BMP4 released was calculated (ng/million cells/24 h). The marker profile of WTMDSCBMP4/GFP and Cox-2KOMDSCBMP4/GFP was also characterized by qPCR and flow cytometry analysis including NG2 (verdican), CD140a (PDGFRα), CD140b (PDGFRβ), CD44, CD73, Sca-1, Pax-7, Runt-related transcription factor X2 (Runx2), osterix (Osx), alkaline phosphatase (Alp) and collagen type 1 (Col1A1).

Creation of a critical size calvarial defect and cell implantation

The Institutional Animal Care and Use Committee of the University of Pittsburgh approved all animal protocols. For this study, we utilized a previously described 5-mm critical size cranial defect model created in the right parietal bone of mice (35). Fibrin sealant (Tisseel Baxter) was used as a scaffold for all in vivo experiments. Following the creation of the defect, BMP4/GFP-transduced MDSCs (5 × 10⁵) in 10 µl PBS were mixed with 15 µl of thrombin and immediately implanted into the defect. Finally, 15 µl of fibrin sealant (Tisseel, Baxter, Mississauga, ON, USA) was added and allowed to solidify for 1–2 min prior to closing the scalp with sutures. The animals were then allowed to recover in an oxygen chamber.

Evaluation of the role of host COX-2 expression in MDSC-mediated bone healing

BMP4/GFP-transduced MDSCs isolated from normal C57BL/10J mice were transplanted into the skull defects created in WT and Cox-2KO mice (n = 4), respectively. Bone regeneration was evaluated via micro-computed tomography (microCT) biweekly for 8 weeks. After week 8, the mice were sacrificed and skull tissues were harvested and fixed in 4% neutral buffered formalin (NBF) for 48 h. Tissues were then decalcified in 10% ethylenediamine tetraacetic acid (EDTA) disodium dihydrate and 1% sodium hydroxide and embedded in paraffin. Histological sections of 5 μm in thickness were cut for analyses. Herovici’s staining was performed to detect type I collagen (Col1) and immunohistochemistry for GFP was performed to determine the contribution of donor cells in the newly formed bone.

Evaluation of the role of donor COX-2 in MDSC-mediated bone healing

Approximately 5 × 10⁵ BMP4/GFP-transduced WT and Cox-2KO MDSCs were implanted into the skull defect created in CD-1 nude mice (n = 6) as described above, and bone regeneration was evaluated via microCT biweekly for a total of 6 weeks. After 6 weeks, the mice were sacrificed and skull tissues were harvested and fixed in 4% NBF and processed as described above. Herovici’s and H&E staining and GFP immunohistochemistry were performed.

Micro-computed tomography

Bone regeneration within the calvarial defect was monitored post-surgically via microCT (vivaCT 40, Scanco Medical, Fabrikweg 2, Switzerland) biweekly using 30 µm voxel size, 70 kvp, 110 µA scanning parameters. After obtaining two-dimensional image slices, the view of interest (VOI) was uniformly delineated and three-dimensional reconstructions were created using an appropriate threshold that remained constant throughout all analyses. De novo three-dimensional new bone volume was quantified using the software provided by the manufacturer by counting every slice of the new bone area. Three-dimensional new bone volume was measured using Gauss Sigma 0.8 and Gauss Support 1 with a threshold of 163. The microCT parameters and terminology utilized followed the guidelines of the American Society of Bone and Mineral Research (36).

Evaluation of inflammation and angiogenesis during MDSC-mediated bone regeneration

In order to investigate whether the implantation of Cox-2KO MDSCs affect donor and host cell interactions during new bone formation, 36 male CD-1 nude mice were divided into two groups: one group received 5 × 10⁵ BMP4/GFP-transduced WT MDSCs, the other group received the same number of BMP4/GFP-transduced Cox-2KO MDSCs. After transplantation, six mice from each group were sacrificed at 3-, 7- and 14-day time points. Skull tissues were harvested and embedded in NEG50 freezing medium, snap-frozen in liquid nitrogen and stored at −80°C. Cryosections of 10 µm were cut for histology and immunofluorescent staining for CD68 (macrophages), Gr-1 (neutrophils), CD31 (endothelial cells) and Ki67 (proliferating cells). Alcian blue staining was performed at 7 and 14 days. Von Kossa staining was performed at day 14.

Histological evaluation of calvarial defect bone regeneration

Alcian blue staining was performed to determine the presence of endochondral bone formation according to an online protocol (IHC world, http://www.ihcworld.com/_protocols/special_stains/alcian_blue.htm). Von Kossa staining was performed to detect mineralization within the regenerated tissues (IHC world, http://www.ihcworld.com/_protocols/special_stains/alcian_blue.htm). Herovici’s staining was used to reveal the formation of type I collagen (37) and H&E staining was also performed.

Immunofluorescence and immunohistochemistry

Cryosections were fixed with 4% paraformaldehyde (PFA, Sigma–Aldrich, Milwaukee, WI, USA) and blocked with 5% donkey serum. Sections were subsequently incubated with a primary target protein antibody together with a GFP primary antibody overnight at 4°C. The following day, the sections were incubated with the corresponding secondary antibodies for 2 h at room temperature. Finally, sections were counterstained with 4, 6 diamino-2-phenylindole (DAPI) and cover-slipped with Permafluor mount medium (TA030FM, Fisher Scientific). The primary antibodies used in this study included: rabbit anti-GFP (ab290, Abcam, Cambridge, MA, USA, 1:1000 dilution), rat anti-mouse CD68 (ab 53444, Abcam, 1:100), rat anti-Gr-1 (BD557445, BD Biosciences;1:100 dilution), rat anti-mouse CD31 (BD553370, BD Bioscience, 1:300 dilution), rabbit anti-pSmad1/5 (ab92698, Abcam, 1:100 dilution), goat anti-GFP (ab6673, Abcam, 1:200 dilution) and goat anti-Ki67 (SC-7846, Santa Cruz Biotechnology, 1:50 dilution). The secondary antibodies included donkey anti-goat-594-dyLight (705-515-147, Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA, 1:200 dilution), donkey anti-rabbit-IgG 488 (Molecular Probe, 1:400 dilution) and donkey anti-rat IgG-DyLight-594 (705-515-150, Jackson ImmunoResearch Laboratories Inc., 1:200 dilution). Calvarial bone samples harvested 6 or 8 weeks after MDSC implantation were decalcified. Skull tissues were dehydrated through an ethanol gradient, cleared with xylene and paraffin embedded. Immunohistochemistry was then carried out on 5-µm paraffin sections. After deparaffinization, washing and blocking with 5% donkey serum in PBS, sections were incubated with rabbit anti-GFP antibody (ab290, Abcam, 1:1000 dilution) overnight. The following day, sections were treated with 0.5% H₂O₂ in PBS for 30 min at room temperature, washed in PBS and incubated with goat anti-rabbit-biotin (BA 1000, Vector Laboratories, Burlingame, CA, USA, 1:200 dilution) for 2 h at room temperature. After three washes, each slide was incubated with ABC reagents (PK 7200, Elite ABC kits, Vector Laboratories) for 2 h at room temperature. DAB staining (SK-4100, Vector Laboratories) was used to visualize the GFP-positive cells. Hematoxylin (H-3404, Vector laboratories) counterstaining was performed following the DAB reaction. Immunofluorescent images were acquired using Northern Eclipse imaging software and bright field pictures were acquired using Q-Capture on a Nikon ECLIPSE E800 microscope (Nikon, Melville, NY, USA). To quantify the cell numbers, a total of 6–8 pictures (200×) were taken for each sample using sections from three different axial levels (300 µm apart) and analyzed with Image J (NIH) cell counter software. The numbers of cells within the defect areas of the 200× fields were counted and the defect areas were measured on each of the pictures. The average cell number of each sample was then calculated. The cell number in the defect region was normalized to the 200× field area for each picture from all groups. The normalized number of cells positive for CD68, Gr-1, CD31 and Ki-67 in each group was expressed as the mean ± SD.

Multipotent differentiation assays

BMP4/GFP-transduced WT and COX-2KO MDSCs were tested for their osteogenic, chondrogenic and myogenic differentiation capacities in vitro.

Osteogenic differentiation

We utilized pellet cultures for the osteogenic differentiation assay. Approximately 2.5 × 10⁵ cells were placed into 15-ml tubes, centrifuged and resuspended in osteogenic medium containing high glucose DMEM (Invitrogen) supplemented with 10% FBS (Invitrogen), 1% penicillin/streptomycin (Invitrogen), 10 ^− 2 m beta glycerol phosphate (Sigma–Aldrich Inc., St. Louis, MO, USA), 50 µg/ml l-ascorbic acid-2 phosphate (Sigma–Aldrich Inc.) and 10 ^− 7 m dexamethasone (Sigma–Aldrich Inc.). The cells were then centrifuged at 500 g for 5 min to pellet the cells. Pellets were subjected to media changes every 2–3 days. The pellets were scanned by Viva CT 40 (Scanco Medical, Scanco USA Inc., Wayne, PA, USA) at 4 weeks to evaluate mineralized matrix deposition. At 4 weeks, pellets were fixed in 4% NBF and rinsed with PBS and embedded in NEG50 (Thermo Scientific, Kalamazoo, MI, USA) freezing medium, snap-frozen in liquid nitrogen and sectioned at 8 µm for histology. Von Kossa staining was performed as stated above to detect mineralization. Osteogenic differentiation was confirmed by osteocalcin immunohistochemistry using a goat anti-mouse osteocalcin (OC) antibody (sc23790, Santa Cruz Biotechnology Inc., 1:50 dilution).

Chondrogenic differentiation

Approximately 1.25 × 10⁵ cells were placed into 15-ml tubes (BD Biosciences), centrifuged at 800 g for 5 min, resuspended in chondrogenic medium (StemPro^® Chondrogenesis Differentiation Kit, Invitrogen, Life Technology) and centrifuged a second time to remove any residual growth medium and to pellet the cells. Finally, the cells were resuspended in 0.5 ml chondrogenic medium and centrifuged at 500 g for 5 min. Cell pellets were maintained in chondrogenic medium for 24 days with media changes every 2–3 days. Chondrogenic differentiation was verified by Alcian blue staining using an online protocol as stated above. COL2A1 immunohistochemistry using goat anti-mouse type II collagen antibodies (COL2A1) (SC-7764, Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA, 1:50 dilution) was performed as stated above.

Myogenic differentiation

Approximately 4 × 10⁴ cells were seeded in collagen-coated 24-well plates (Corning Inc., Corning, NY, USA). When the cells reached 100% confluence on the second day, the proliferation medium (PM) was shifted to myogenic medium containing high glucose DMEM (Invitrogen) supplemented with 2% FBS (Invitrogen) and 1% penicillin/streptomycin (Invitrogen). After 5 days, myogenic differentiation was determined via the use of an antibody against fast myosin heavy chain (fMHC, Sigma, 1:400) and GFP (Ab290, Abcam, 1:1000) and quantified by counting fMHC-positive cells including orange color or red color multinucleated myotubes and single cells against total number of DAPI-positive cells.

Endothelial cell proliferation assay

Approximately 1.5 × 10⁵ WT and Cox-2KO MDSCs were seeded in T175 coated flasks and cultured in 18 ml PM for 36 h. The medium from two populations of WT and Cox-2KO MDSC cultures were centrifuged at 800 g for 5 min, aliquoted and stored at −80 °C. MILE SVEN1 (MS1) mouse endothelial cells (CRL-2279, ATCC, Manassas, VA, USA) were cultured at 37 °C under 10% CO₂ in medium consisting of high glucose DMEM (Invitrogen, Life Technologies) supplemented with 5% fetal bovine serum and 1% penicillin/streptomycin and passaged every 2 days. The cells were trypsinized when they reached 70–80% confluence. For the proliferation experiments, cells were seeded in a flat bottom 96-well plate (Corning, NY, USA) at a density of 4 × 10³ cells per well in 0.2 ml of medium. Four replicates were used for each treatment group. To test whether the MDSCs had a paracrine effect on the endothelial cells in vitro, we tested the cell growth of the MS1 cells by culturing them in conditioned medium obtained from both WT and Cox-2KO MDSCs and non-conditioned fresh PM, at 37 °C and 5% CO₂ for 48 and 72 h. Two hours before finishing the experiment at each time point, 100 µl of medium was removed from each well and 20 µl of Cell Titer 96 AQueous One solution (TB 245, Promega Corporation, Madison, WI, USA) was added to each well. The plates were incubated for an additional 2 h and the absorbance was measured using a Tecan Infinite 200-plate reader at a wavelength of 490 nm (A490). The proliferation assay was repeated 4 times, and the results were reported from one experiment. Cell growth under each condition was compared with PM alone at each of the respective time points.

Semi-quantitative reverse transcription polymerase chain reaction (RT-PCR), real-time PCR, western-blot analysis and flow cytometry

Three populations of WT and Cox-2KOMDSCs were cultured in proliferation medium. At passage 18–20, cells were trypsinized with 0.1% Trypsin-EDTA (Invitrogen), centrifuged and then 2 × 10⁵ MDSCs from each population were lysed with 1 ml Trizol (Invitrogen). Two populations of BMP4/GFP-transduced WT and Cox2KO MDSCs were harvested at three different passages, respectively, centrifuged and lysed in Trizol (Invitrogen). Total RNA was extracted using the protocols provided by the manufacturer. Reverse transcription was performed using 100–1000 ng total RNA, depending on the experiment, and a Maxima First Strand cDNA synthesis kit was used for RT-PCR (Thermo Scientific). The cDNA was diluted with DNAase and RNAase free water and stored at −20 °C for further PCR amplification. The PCR primers were designed using Primer 3 (38,39), and PCR was performed in a 25 µl reaction with the Gotaq PCR system (Promega Corp., Madison, WI, USA). The PCR products were electrophoresed on a 1% agarose gel and the images were captured using GelDoc with QuantOne software (GelDoc system, BioRad, Hercules, CA, USA). Primer sequences are listed in Supplementary Material, Table S1. The band density of target gene and house-keeping gene (Gapdh, actin) was measured used the Quantone software by substracting the background and the relative gene expression level was calculated by normalizing the target gene level to that of Gapdh or actin. Quantitative real-time PCR were also performed on selected genes that showed difference in semi-quantitative PCR using Biorad SsoAdvanced™ Universal SYBR^® Green Supermix using CFX machine (BioRad) and Gapdh as the house-keeping gene control.

For western-blot assay, cell lysates were prepared in radioimmunoprecipitation assay (RIPA) buffer (#9806, Cell Signaling Technology, Inc., Danvers, MA, USA) supplemented with protease inhibitor (P8340) and phosphatase inhibitors (P5726 and P0044, 1:100, Sigma–Aldrich, St. Louis, MO, USA). The newly regenerated tissues within the defect areas of the mice from both groups at the 7- and 14-day time points were dissected and homogenized with RIPA buffer in liquid nitrogen. Protein concentration was quantified using a Bio-Rad Protein Assay Kit 2 (#500-0002, Bio-Rad, Hercules, CA, USA). Western-blot analyses were performed using a rabbit anti-pSMAD1/5 (#9516S, Cell Signaling, Technology, Inc., Danvers, MA, 1:1000 dilution), goat anti-mouse osteocalcin (OC) (sc23790, Santa Cruz Biotechnology Inc., 1:200 dilution), mouse anti-phosphorylated Akt (4051, Cell Signaling Technology, 1:1000 dilution) and mouse anti-beta actin (A5441, Sigma, 1:8000 dilution) primary antibodies to detect pSMAD1/5, osteocalcin, pAkt and actin, respectively. Horseradish peroxidase-conjugated rabbit anti-goat (#31402, Piece, 1:5000 dilution), rabbit anti-mouse (#31450, Piece, 1:10000 dilution), goat anti-rabbit (#31460, Piece, 1:5000 dilution) secondary antibodies were used to detect goat, mouse and rabbit primary antibodies. Supersignal Western Pico Chemilumnescent Substrate (Pierce) was used to reveal the target protein bands. A FOTO/analyst FX (Fotodyne Inc.) system was used to capture digital images. The band densities were quantified using Photoshop by subtracting the background, and then normalized to actin.

Flow cytometry analysis of cell surface markers of CD73, CD44 and CD45 were also performed using phycoerythrin (PE) rat anti-mouse CD73(BD Bioscience, 550741), PE-Cy5 rat anti-mouse CD44 (BD Bioscience 553135) and allophycocyanin (APC) rat anti-mouse CD45 (BD Bioscience 559884). Two populations of BMP4/GFP-transduced Cox-2WTMDSCs and Cox-2KOMDSCs were first incubated with FC-blocking reagent (130-092-575, Miltenyl Biotech) for 10 min and then incubated with fluorescence-conjugated primary antibodies for another 20 min at 4 °C. The cells were analyzed with Calibur (BD Bioscience) at the Flow Cytometery core of University of Texas Health Science Center at Houston.

Oxidative and inflammatory stress assays

For the oxidative stress experiment, 2 × 10³ BMP4/GFP-transduced WT and Cox-2KOMDSCs were seeded in 24-well plates and incubated overnight in PM. The following day the proliferation medium was removed and cells were rinsed with PBS one time, and oxidative stress medium was added to four wells each containing 250 or 400 µm hydrogen peroxide and propidium iodide (PI, 2 μg/ml), respectively. Only the dead cells become intercalated with PI which fluoresces red. The plates were set up in a live cell imaging system. Three locations from each well were randomly chosen for image capture. Bright field and red fluorescence images were taken every 10 min, and the dead cell number was counted at 0, 4, 8, 12, 16, 20 and 24 h time points using Image J software. The survival rate was calculated for each time point. Additionally, experiments were prepared in 6-well plates using 250 µm H₂O₂. Approximately 1 × 10⁴ BMP4/GFP-transduced WT and Cox-2KO MDSCs were seeded on the plates and the cells were allowed to adhere overnight. The second day, the cells were rinsed with PBS one time and switched to proliferation medium containing 250 µm H₂O₂ for 4 h, the cells were lysed with 1 ml Trizol (Invitrogen) and total RNA was extracted following the manufacturer’s protocol. Semi-quantitative RT-PCR was performed using the same protocol described above.

For the inflammatory stress experiments, 1 × 10⁴ BMP4/GFP-transduced WT and Cox-2KO MDSCs were seeded on 6-well plates in triplicate for each population. On the second day, the cells were rinsed with PBS one time and a proliferation medium containing 100 ng/ml TNFα was added to each well and treated for 4 h. Cells were then rinsed with PBS and lysed with 1 ml Trizol and total RNA was extracted for semi-quantitative RT-PCR analysis as stated above.

Statistical analysis

All values are expressed as the mean ± SD. Student’s t-test was used for comparing two groups. One-way analysis of variance (ANOVA) followed by Tukey’s post hoc T test was used for multiple group analyses. For those data that did not have a Gaussian distribution and unequal variances (Fig. 1B and Supplementary Material, Fig. S3), the Wilcoxon rank sum test was utilized. P < 0.05 was considered statistically significant.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We thank Ryan J. Warth, MD in the Department of Orthopaedic Surgery, University of Texas Health Science Center at Houston, and Dr Lavanya Rajagopalan, PhD for their editorial assistance. We also thank Lori Walton for her assistance in performing the Herovici’s and H&E staining.

Conflict of Interest statement. None declared.

Funding

This project was supported by a grant from the National Institutes of Health (5RO1-DE13420-09) awarded to J.H.