Discoidin Domain Receptor 1 Regulates Runx2 during Osteogenesis of Osteoblasts and Promotes Bone Ossification via Phosphorylation of p38

Discoidin domain receptor 1 (Drd1) is a collagen-binding membrane protein, but its role in osteoblasts during osteogenesis remains undefined. We generated inducible osteoblast-specific Ddr1 knockout (OKOΔDdr1) mice; their stature at birth, body weight and body length were significantly decreased compared with those of control Ddr1f/f-4OHT mice. We hypothesize that Ddr1 regulates osteogenesis of osteoblasts. Micro-CT showed that compared to 4-week-old Ddr1f/f-4OHT mice, OKOΔDdr1 mice presented significant decreases in cancellous bone volume and trabecular number and significant increases in trabecular separation. The cortical bone volume was decreased in OKOΔDdr1 mice, resulting in decreased mechanical properties of femurs compared with those of Ddr1f/f-4OHT mice. In femurs of 4-week-old OKOΔDdr1 mice, H&E staining showed fewer osteocytes and decreased cortical bone thickness than Ddr1f/f-4OHT. Osteoblast differentiation markers, including BMP2, Runx2, alkaline phosphatase (ALP), Col-I and OC, were decreased compared with those of control mice. Ddr1 knockdown in osteoblasts resulted in decreased mineralization, ALP activity, phosphorylated p38 and protein levels of BMP2, Runx2, ALP, Col-I and OC during osteogenesis. Overexpression and knockdown of Ddr1 in osteoblasts demonstrated that DDR1 mediates the expression and activity of Runx2 and the downstream osteogenesis markers during osteogenesis through regulation of p38 phosphorylation.


Introduction
Bone development is a dynamic process within higher organisms. In mammals, skeletogenesis is a two-step ossification process comprising endochondral ossification and intramembranous ossification. The endochondral ossification process starts with mesenchymal stem cells (MSCs) committing to chondrocytes, in which proliferation, differentiation, hypertrophy, terminal differentiation and apoptosis result in the formation of bone; osteoblasts and osteoclasts comprise the majority of cells in mature bone [1]. By contrast, in the intramembranous ossification process, MSCs in cortical and trabecular bone are directly differentiated into osteoblasts [2] to form flat bone structures, including the skull, clavicle, jaw, sternum, ribs, and collar bone. At the periosteum, osteoprogenitor cells differentiate into osteoblasts and are responsible for mineralization of the bone matrix during bone formation. Runx2 (Runt-related transcription factor 2, also known as Cbf1) is a key regulator during osteoblast differentiation, which induces preosteoblasts to become mature osteoblasts. Specific deletion of Runx2 in osteoblasts leads to defective intramembranous ossification and postnatal bone formation [3,4]. Runx2 regulates the expression of genes at different stages of osteogenesis, such as the early stage during which preosteoblasts commit to osteoblasts by secreting alkaline phosphatase (ALP) and type I collagen (Col-I) and expressing the differentiation marker osteopontin (OPN) and the later stage during which the mature osteoblast markers osteocalcin (OC), which is a calcium binding protein, bone sialo protein (BSP) and matrix metalloproteinase-13 (MMP13) are expressed to promote mineralization [3][4][5][6][7][8][9]. The basic condition of bone growth is the dynamic balance of cartilage and cortical bone formation.
Discoidin domain receptors (DDRs) are a family of receptor tyrosine kinases that includes DDR1 and DDR2, which are transmembrane collagen-binding receptors. DDR1 is commonly expressed in a variety of mice and in humans, and DDR1 gene maps indicate 17 exons at chromosome 6 (6p21.3) in the human genome, which can be expressed as five distinct DDR1 isoforms. However, DDR2 gene maps show 19 exons at chromosome 1 (1q23.3) but only one isoform [10,11]. Fibrillar collagens are ligands for both DDRs; thus far, DDR1 and DDR2 could be activated by collagen, including types I−III, V and X [11][12][13]. In particular, DDR1 also binds type VIII collagen and basement membrane type IV collagen [11]. Global DDR1 knockout in mice causes dwarfism, and the body weight is more decreased than in wild-type mice. Micro-CT results show that the global DDR1 knockout mice are short and thinner in fibular bone but the growth plate length, chondrocyte proliferation and apoptosis are unaltered compare with wild-type mice [10]. However, the mechanism of global Ddr1 knockout that causes dwarfism in mice and whether DDR1 regulates the function of osteoblasts during development remain undefined. DDR1 is also expressed at the periosteal collar of the clavicle in embryonic and postnatal mice [10]. Global Ddr1-null mice have abnormal development in multiple organ systems, which may affect the interpretation of DDR1 function, such as loss of auditory function, lactation impairment, cochlear duct defect, reproduction defects, and proteinuria [11][12][13][14][15]. There are no reports that show the role of Ddr1 in osteoblast function during osteogenesis. Therefore, we firstly created the Ddr1f/f mice, and used the Cre-LoxP system to generate the inducible osteoblast-specific Ddr1 knockout mice (OKO∆Ddr1) in this study. Our results showed that there were no other developmental abnormalities compared with global Ddr1 knockout mice. The OKO∆Ddr1 mice showed a shorter stature at birth than the control Ddr1 f/f-4OHT mice. Therefore, we hypothesize that Ddr1 may play a role in regulating the osteogenesis of osteoblasts. To elucidate the role of Ddr1 in regulating osteoblasts during bone development, we used micro-CT analysis, H&E staining and IHC staining to evaluate the expression of osteoblast differentiation markers in the femurs in the OKO∆Ddr1 mice in vivo. We also used an in vitro model of Ddr1 knockdown in osteoblasts, which showed decreased mineralization, ALP activity, levels of phosphorylated p38 and protein levels of BMP2, Runx2, ALP, Col-I and OC during osteogenesis. In this study, by using overexpression and knockdown of Ddr1, we demonstrated that, via p38, DDR1 mediates the expression and activity of Runx2 and the downstream osteogenesis markers OC, Col-I, and ALP during osteogenesis.

Results
2.1. Generation of Osteoblast-Specific Knockout Mice (a1(I) Collagen-CreERT; Ddr1 f/f ; OKO∆Ddr1 Mice) To investigate the role of Ddr1 in osteoblasts during bone ossification, we first generated 4-OHT inducible osteoblast-specific Ddr1 knockout (a1(I)-CreERT; Ddr1 f/f ) mice on the OKO background. To confirm the osteoblast-specific knockout of Ddr1 in OKO∆Ddr1 mice, we conducted gene expression, Western blot, and IHC staining assays in Ddr1 f/f-4OHT and OKO∆Ddr1 mice. The qPCR results showed that the gene expression of Ddr1 in OKO∆Ddr1 mice was significantly decreased by approximately 95% compared with that in Ddr1 f/f-4OHT mice ( Figure 1A). Additionally, the protein level of DDR1 was decreased by approximately 90% in OKO∆Ddr1 mice compared with Ddr1 f/f-4OHT mice ( Figure 1B). To further demonstrate that we specifically knocked out Ddr1 in osteoblasts, we performed IHC staining for DDR1 in the femurs of 4-week-old mice, the results of which showed that DDR1 staining was obvious in the cortical bone, periosteum and articular chondrocytes in Ddr1 f/-4OHT mice. However, DDR1 staining was decreased by approximately 90% in cortical bone in OKO∆Ddr1 mice but was still observed in articular chondrocytes ( Figure 1C,D). By contrast, DDR2 staining in cortical bone and articular chondrocytes was similar in Ddr1 f/f-4OHT and OKO∆Ddr1 mice and showed no significant quantitative difference ( Figure 1C,D). These results indicated that we had succeeded in the creation of osteoblast-specific Ddr1 knockout (OKO∆Ddr1) and that this model did not affect the expression of Ddr2, suggesting that there is no compensatory effect of Ddr2 expression. We stained the osteoclasts by TRAP staining, and also immunochemistry stain of TRAcP as an osteoclast marker. The results showed that knockout Ddr1 in osteoblasts had no effect on osteoclast activity and the expression of osteoblasts (S1A, 1B). These results indicated that knockout Ddr1 in osteoblasts would not influence the osteoclast phenotype, and demonstrated the osteoblast-specific Ddr1 knockout in mice. osteoblasts, which showed decreased mineralization, ALP activity, levels of phosphorylated p38 and protein levels of BMP2, Runx2, ALP, Col-I and OC during osteogenesis. In this study, by using overexpression and knockdown of Ddr1, we demonstrated that, via p38, DDR1 mediates the expression and activity of Runx2 and the downstream osteogenesis markers OC, Col-I, and ALP during osteogenesis.

Generation of Osteoblast-Specific Knockout Mice (a1(I) Collagen-CreERT; Ddr1 f/f ; OKOΔDdr1 Mice)
To investigate the role of Ddr1 in osteoblasts during bone ossification, we first generated 4-OHT inducible osteoblast-specific Ddr1 knockout (a1(I)-CreERT; Ddr1 f/f ) mice on the OKO background. To confirm the osteoblast-specific knockout of Ddr1 in OKOΔDdr1 mice, we conducted gene expression, Western blot, and IHC staining assays in Ddr1 f/f-4OHT and OKOΔDdr1 mice. The qPCR results showed that the gene expression of Ddr1 in OKOΔDdr1 mice was significantly decreased by approximately 95% compared with that in Ddr1 f/f-4OHT mice ( Figure 1A). Additionally, the protein level of DDR1 was decreased by approximately 90% in OKOΔDdr1 mice compared with Ddr1 f/f-4OHT mice ( Figure 1B). To further demonstrate that we specifically knocked out Ddr1 in osteoblasts, we performed IHC staining for DDR1 in the femurs of 4-week-old mice, the results of which showed that DDR1 staining was obvious in the cortical bone, periosteum and articular chondrocytes in Ddr1 f/-4OHT mice. However, DDR1 staining was decreased by approximately 90% in cortical bone in OKOΔDdr1 mice but was still observed in articular chondrocytes ( Figure 1C,D). By contrast, DDR2 staining in cortical bone and articular chondrocytes was similar in Ddr1 f/f-4OHT and OKOΔDdr1 mice and showed no significant quantitative difference ( Figure 1C,D). These results indicated that we had succeeded in the creation of osteoblast-specific Ddr1 knockout (OKOΔDdr1) and that this model did not affect the expression of Ddr2, suggesting that there is no compensatory effect of Ddr2 expression. We stained the osteoclasts by TRAP staining, and also immunochemistry stain of TRAcP as an osteoclast marker. The results showed that knockout Ddr1 in osteoblasts had no effect on osteoclast activity and the expression of osteoblasts (S1A, 1B). These results indicated that knockout Ddr1 in osteoblasts would not influence the osteoclast phenotype, and demonstrated the osteoblast-specific Ddr1 knockout in mice.

Skeletal Dysplasia in OKO∆Ddr1 Mice
To study the role of Ddr1 in osteoblasts during postnatal skeletal development, we performed double staining of the skeleton with Alizarin red for mineralized matrix (red) and Alcian blue for cartilaginous matrix (blue). Although selectively knocking out Ddr1 in osteoblasts was not embryonic lethal and did not cause structurally significant abnormalities in OKO∆Ddr1 mice, the long bones of OKO∆Ddr1 mice were thinner than those of Ddr1 f/f-4OHT mice (Figure 2A). Knocking out Ddr1 in osteoblasts caused a decrease in the mineralization of the rib/sternum, which is driven by intramembranous ossification to form bones ( Figure 2B). We also observed that the length and diameter of endochondral ossification bones (such as forelimb, hindlimb, and vertebrae) were smaller in OKO∆Ddr1 mice than in Ddr1 f/f-4OHT mice ( Figure 2C). At 4 weeks of age, OKO∆Ddr1 mice exhibited significantly decreased body weight and body length compared to those of Ddr1 f/f-4OHT mice ( Figure 2D). These findings indicate that Ddr1 knockout in osteoblasts causes dysplasia, which suggests that DDR1 plays a critical role in osteoblast function during ossification. expressed relative to the Ddr1 f/f-4OHT mice was quantified. Each group n ≥ 6. (C) IHC staining of DDR1 and DDR2 in cortical bone and articular cartilage of femurs from 4-week-old mice; black arrow indicates DDR1-positive cells; black frame indicates the cortical region. (D) Quantitative results by tissue faxes normalized with Ddr1 f/f-4OHT . Magnifications of 400X are shown, with scale bars of 50 μm. Each group, n ≥ 6; * p ≤ 0.05, ** p ≤ 0.01.

Skeletal Dysplasia in OKOΔDdr1 Mice
To study the role of Ddr1 in osteoblasts during postnatal skeletal development, we performed double staining of the skeleton with Alizarin red for mineralized matrix (red) and Alcian blue for cartilaginous matrix (blue). Although selectively knocking out Ddr1 in osteoblasts was not embryonic lethal and did not cause structurally significant abnormalities in OKOΔDdr1 mice, the long bones of OKOΔDdr1 mice were thinner than those of Ddr1 f/f-4OHT mice (Figure 2A). Knocking out Ddr1 in osteoblasts caused a decrease in the mineralization of the rib/sternum, which is driven by intramembranous ossification to form bones ( Figure 2B). We also observed that the length and diameter of endochondral ossification bones (such as forelimb, hindlimb, and vertebrae) were smaller in OKOΔDdr1 mice than in Ddr1 f/f-4OHT mice ( Figure 2C). At 4 weeks of age, OKOΔDdr1 mice exhibited significantly decreased body weight and body length compared to those of Ddr1 f/f-4OHT mice ( Figure  2D). These findings indicate that Ddr1 knockout in osteoblasts causes dysplasia, which suggests that DDR1 plays a critical role in osteoblast function during ossification.

Decreased Bone Formation in the Tibia of OKO∆Ddr1 Mice
At the end of endochondral ossification, hypertrophic cartilage is replaced by osteoblasts, osteoclasts, and blood vessels. At the perichondrium of cortical bone, osteoprogenitor cells differentiate into osteoblasts and form the bone collar of the cartilage anlage. In 2-and 4-week-old OKO∆Ddr1 mice, we observed that the long bones of the femur and tibia were slimmer than those in Ddr1 f/f-4OHT mice ( Figure 3A). To characterize how Ddr1 regulates osteoblast function during skeletal development, we performed micro-CT analysis. The reconstructed 3-D images revealed that the long bones were obviously thicker in Ddr1 f/f-4OHT mice than in OKO∆Ddr1 mice ( Figure 3B). The cross-sectional views and 3-D reconstruction of the trabecular bone (2-mm segment) showed less trabecular bone in Ddr1 f/f-4OHT mice than in OKO∆Ddr1 mouse tibia ( Figure 3C). Moreover, the 3-D structure of tibial cortical bone was thinner and smaller in Ddr1 f/f-4OHT mice than in OKO∆Ddr1 mice ( Figure 3C). To investigate whether Ddr1 knockout in osteoblasts affects bone mineralization in vivo, we analyzed the 3-D structure of tibias (2 mm length) in 4-week-old mice, and the inner region for trabecular bone and outer region for cortical bone were also analyzed. The trabecular bone volume (BV/TV) and trabecular number (Tb.N) were decreased and the trabecular separation (Tb.Sp) was increased in OKO∆Ddr1 mice ( Figure 3D). It had no significant differences on trabecular bone mineral density (Tb. BMD) and trabecular thickness (Tb.Th) ( Figure 3D). The results of the cortical bone quantitation showed a significant decrease in the bone volume and bone mineral content (BMC) but no effect on the bone mineral density (BMD) in OKO∆Ddr1 mice ( Figure 3E). Consistent with the micro-CT data, 4-week-old OKO∆Ddr1 mice showed a marked reduction in bone structure properties (mechanism loading to failure) such as max-load, break point, stiffness, area under the max curve, area under the break curve, modulus, and toughness in femur cortical bone compared to those in Ddr1 f/f-4OHT mice; these properties were assessed with the three-point bending analysis (Table 1). These results indicated that osteoblasts with Ddr1 knockout caused the bone structure to be significantly weaker than that of the control mice. Taken together, these results indicate that osteoblasts with Ddr1 knockout decrease the function of osteoblasts and decrease bone formation of cortical bone and trabecular bone, leading to bone ossification delay.

Knocking out Ddr1 in Osteoblasts Decreases Cortical Bone Thickness and Cell Number
To further understand the role of Ddr1-mediated osteoblast function and osteogenesis-related markers during bone ossification, we performed Safranin O/fast green and hematoxylin and eosin (H&E) staining. In the Safranin O/fast green staining assay, trabecular bone was less abundant in 4-week-old OKO∆Ddr1 mouse femurs ( Figure 4A), but quantification of the growth plate length showed no difference between Ddr1 f/f-4OHT and OKO∆Ddr1 mouse femurs ( Figure 4B). Upon performing H&E staining of the femur ( Figure 4C), we quantitated the cortical thickness and observed a decrease in 2and 4-week-old OKO∆Ddr1 mice compared with that in age-matched Ddr1 f/f-4OHT mice ( Figure 4D). Moreover, quantitation of the number of mature osteoblasts in cortical bone revealed fewer cells in 2-and 4-week-old OKO∆Ddr1 mouse tibias than in the tibias of Ddr1 f/f-4OHT mice ( Figure 4C). These results indicated that osteoblasts with Ddr1 knockout exerted decreased cortical and trabecular bone formation due to a decrease in osteoblast number, which ultimately resulted in delayed bone ossification; however, osteoblasts with Ddr1 knockout had no influence on the growth plate length, which means that Ddr1 knockout in osteoblasts may not affect the function of chondrocyte.

Knocking out Ddr1 in Osteoblasts Caused Downregulation of the Osteogenesis-Related Marker Gene Expression
To investigate whether Ddr1 knockout in osteoblasts influences chondrocyte function at the chondro-osseous interface, we stained for chondrocyte terminal differentiation markers, such as type X collagen (Col-X) and matrix metalloproteinase-13 (MMP13). The IHC results showed that knocking out Ddr1 in osteoblasts had no effect on either Col-X or MMP13 in the hypertrophic zone of the growth plate and the chondro-osseous interface ( Figure 5B,C). We next detected that the gene expression of osteogenesis related markers, such as Runx2, and BMP2 was significantly lower in OKOΔDdr1 mice than in Ddr1 f/f-4OHT mice ( Figure 5A). The Runx2 protein level was significantly decreased in 4-week-old OKOΔDdr1 mouse trabeculae ( Figure 5B). ALP is a hallmark of early development [16]. In OKOΔDdr1 mice, the gene expression of ALP was decreased compared with that in Ddr1 f/f-4OHT mice ( Figure 5A). In osteogenesis, CoI-I and OC are markers of late osteoblast differentiation [17]. The gene expression of Col-I and OC was decreased by ~80% in OKOΔDdr1 mice ( Figure 5A). According to these results, we demonstrated that osteoblasts with Ddr1 knockout

Knocking out Ddr1 in Osteoblasts Caused Downregulation of the Osteogenesis-Related Marker Gene Expression
To investigate whether Ddr1 knockout in osteoblasts influences chondrocyte function at the chondro-osseous interface, we stained for chondrocyte terminal differentiation markers, such as type X collagen (Col-X) and matrix metalloproteinase-13 (MMP13). The IHC results showed that knocking out Ddr1 in osteoblasts had no effect on either Col-X or MMP13 in the hypertrophic zone of the growth plate and the chondro-osseous interface ( Figure 5B,C). We next detected that the gene expression of osteogenesis related markers, such as Runx2, and BMP2 was significantly lower in OKO∆Ddr1 mice than in Ddr1 f/f-4OHT mice ( Figure 5A). The Runx2 protein level was significantly decreased in 4-week-old OKO∆Ddr1 mouse trabeculae ( Figure 5B). ALP is a hallmark of early development [16]. In OKO∆Ddr1 mice, the gene expression of ALP was decreased compared with that in Ddr1 f/f-4OHT mice ( Figure 5A). In osteogenesis, CoI-I and OC are markers of late osteoblast differentiation [17]. The gene expression of Col-I and OC was decreased by~80% in OKO∆Ddr1 mice ( Figure 5A). According to these results, we demonstrated that osteoblasts with Ddr1 knockout delayed bone formation through downregulation of osteogenesis-related markers, indicating that Ddr1 in osteoblasts plays a critical role in regulating osteoblast activity during osteogenesis.

DDR1 Promoted Mineralization in MC3T3-E1 Cells
Our results showed that knocking out Ddr1 delayed the expression of osteogenesis-related markers in vivo, which ultimately resulted in delayed bone ossification. To better illustrate the role of Ddr1 in mediating osteoblast functions during development, we performed gain-and loss-offunction assays in preosteoblast MC3T3-E1 cells transduced with different Ddr1 lentiviruses. Mineralization capacity was decreased with Ddr1 knockdown in MC3T3-E1 cells as indicated by Alizarin red staining ( Figure 6A), and ALP activity was decreased in MC3T3-E1 cells with Ddr1 knockdown ( Figure 6B). On the other hand, overexpression of Ddr1 in MC3T3-E1 cells increased both osteoblast mineralization ( Figure 6C) and ALP activity ( Figure 6D). These results indicated that Ddr1 positively regulates the mineralization capacity in preosteoblast MC3T3-E1 cells. We also detected the protein levels of osteogenesis-related markers and revealed that the expression of Runx2, BMP2, ALP, Col-I, and OC was decreased in MC3T3-E1 cells with Ddr1 knockdown ( Figure 6E) and increased in MC3T3-E1 cells overexpressing Ddr1 ( Figure 6F). These results indicated that Ddr1 regulates Runx2, BMP2, ALP, Col-I, and OC expression to mediate osteoblast differentiation and to promote mineralization in MC3T3-E1 cells.

DDR1 Promoted Mineralization in MC3T3-E1 Cells
Our results showed that knocking out Ddr1 delayed the expression of osteogenesis-related markers in vivo, which ultimately resulted in delayed bone ossification. To better illustrate the role of Ddr1 in mediating osteoblast functions during development, we performed gain-and loss-of-function assays in preosteoblast MC3T3-E1 cells transduced with different Ddr1 lentiviruses. Mineralization capacity was decreased with Ddr1 knockdown in MC3T3-E1 cells as indicated by Alizarin red staining ( Figure 6A), and ALP activity was decreased in MC3T3-E1 cells with Ddr1 knockdown ( Figure 6B). On the other hand, overexpression of Ddr1 in MC3T3-E1 cells increased both osteoblast mineralization ( Figure 6C) and ALP activity ( Figure 6D). These results indicated that Ddr1 positively regulates the mineralization capacity in preosteoblast MC3T3-E1 cells. We also detected the protein levels of osteogenesis-related markers and revealed that the expression of Runx2, BMP2, ALP, Col-I, and OC was decreased in MC3T3-E1 cells with Ddr1 knockdown ( Figure 6E) and increased in MC3T3-E1 cells overexpressing Ddr1 ( Figure 6F). These results indicated that Ddr1 regulates Runx2, BMP2, ALP, Col-I, and OC expression to mediate osteoblast differentiation and to promote mineralization in MC3T3-E1 cells.

DDR1 Regulated Osteoblast Differentiation through Phosphorylation of p38.
Previous reports have shown that upon induction with BMP2, Runx2 transcriptional activity is regulated by ERK1/2 rather than by p38 MAP kinase in MC3T3-E1 cells [18]. We further investigated whether DDR1 regulates osteoblast differentiation via phosphorylation of p38 or ERK1/2 activity, and we detected the levels of phosphorylated p38 in MC3T3-E1 cells. The results showed that Ddr1 knockdown decreased p38 phosphorylation but did not affect ERK1/2 phosphorylation in MC3T3-E1 cells ( Figure 7A). Additionally, inhibition of p38 phosphorylation by SB203580 decreased the expression of markers related to osteoblast differentiation, such as OC, ALP, and collagen, in MC3T3-E1 cells [19]. DDR2 regulates osteoblast differentiation by regulating p38 MAPK and inducing activation of Runx2 and the extracellular matrix protein OCN [20]. To determine whether phosphorylated p38 mediated Runx2, OC, ALP, and Col-I expression, we conducted gene expression and Western blotting and showed that inhibition of p38 phosphorylation significantly decreased the

DDR1 Regulated Osteoblast Differentiation through Phosphorylation of p38
Previous reports have shown that upon induction with BMP2, Runx2 transcriptional activity is regulated by ERK1/2 rather than by p38 MAP kinase in MC3T3-E1 cells [18]. We further investigated whether DDR1 regulates osteoblast differentiation via phosphorylation of p38 or ERK1/2 activity, and we detected the levels of phosphorylated p38 in MC3T3-E1 cells. The results showed that Ddr1 knockdown decreased p38 phosphorylation but did not affect ERK1/2 phosphorylation in MC3T3-E1 cells ( Figure 7A). Additionally, inhibition of p38 phosphorylation by SB203580 decreased the expression of markers related to osteoblast differentiation, such as OC, ALP, and collagen, in MC3T3-E1 cells [19]. DDR2 regulates osteoblast differentiation by regulating p38 MAPK and inducing activation of Runx2 and the extracellular matrix protein OCN [20]. To determine whether phosphorylated p38 mediated Runx2, OC, ALP, and Col-I expression, we conducted gene expression and Western blotting and showed that inhibition of p38 phosphorylation significantly decreased the gene expression of Runx2, ALP, Col-I and OC, ( Figure 7B) and also the decrease in the protein levels of Runx2, ALP, Col-I and OC in MC3T3-E1 cells ( Figure 7C). To confirm that Ddr1 regulated osteogenesis of osteoblasts via phosphorylation of p38, we overexpressed Ddr1 in MC3T3-E1 cells treated with the p38 inhibitor SB203580. The results showed that inhibiting p38 phosphorylation in MC3T3-E1 cells overexpressing Ddr1 decreased the protein level of Runx2 ( Figure 7D). These results indicated that Ddr1 mediates Runx2 via p38 MAPK to regulate osteoblast differentiation and results in bone ossification. gene expression of Runx2, ALP, Col-I and OC, ( Figure 7B) and also the decrease in the protein levels of Runx2, ALP, Col-I and OC in MC3T3-E1 cells ( Figure 7C). To confirm that Ddr1 regulated osteogenesis of osteoblasts via phosphorylation of p38, we overexpressed Ddr1 in MC3T3-E1 cells treated with the p38 inhibitor SB203580. The results showed that inhibiting p38 phosphorylation in MC3T3-E1 cells overexpressing Ddr1 decreased the protein level of Runx2 ( Figure 7D). These results indicated that Ddr1 mediates Runx2 via p38 MAPK to regulate osteoblast differentiation and results in bone ossification.

Discussion
To investigate the role of Ddr1 in osteoblasts, we created an inducible osteoblast-specific Ddr1 knockout (OKOΔDdr1) mouse model and evaluated the roles of DDR1 during osteogenesis in vivo. Although osteoblasts with Ddr1 knockout had no effect on mouse survival, they did show markedly reduced postnatal bone formation. Bone ossification was observed to be impaired in 2-and 4-weekold OKOΔDdr1 mice. Knocking out of DDR1 in osteoblasts reduced the expression of osteogenicrelated markers, including Runx2, BMP2, ALP, Col-I, and OC; these results showed reduction effects in bone mass. By using overexpression/knockdown Ddr1 in osteoblasts in vitro, we also confirmed that Ddr1 is required for the phosphorylation of p38 to regulate the expression of the osteogenesis-

Discussion
To investigate the role of Ddr1 in osteoblasts, we created an inducible osteoblast-specific Ddr1 knockout (OKO∆Ddr1) mouse model and evaluated the roles of DDR1 during osteogenesis in vivo. Although osteoblasts with Ddr1 knockout had no effect on mouse survival, they did show markedly reduced postnatal bone formation. Bone ossification was observed to be impaired in 2-and 4-week-old OKO∆Ddr1 mice. Knocking out of DDR1 in osteoblasts reduced the expression of osteogenic-related markers, including Runx2, BMP2, ALP, Col-I, and OC; these results showed reduction effects in bone mass. By using overexpression/knockdown Ddr1 in osteoblasts in vitro, we also confirmed that Ddr1 is required for the phosphorylation of p38 to regulate the expression of the osteogenesis-related markers Runx2, BMP2, ALP, Col-I, and OC in MC3T3-E1 cells. These results indicated that DDR1 is critical for the regulation of osteoblasts during postnatal bone formation in mice.
Osteoprogenitor cells that commit to osteoblasts eventually become mature osteocytes. Runx2 is an essential transcriptional regulator during osteoblast differentiation, and MSCs differentiate into osteoprogenitor cells with high Runx2 expression [3,21]. Cell proliferation was decreased in calvarial osteoblasts in Runx2 ∆C/∆C mice [22], and in Runx2-null mice, chondrocyte and osteoblast maturation is delayed, resulting in decreased bone formation [14]. The specific deletion of exon 8 of Runx2 in osteoblasts decreased bone formation and caused osteopenia in mice [23], and DDR1 deletion reduced Runx2 activity and decreased vascular calcification [24]. p38 MAP kinase has been established as a critical contributor to early osteoblast differentiation [25]. Mice with global p38alpha knockout die in utero due to defective placental development [26]. Another study reported that induction of BMP2/7 through the ERK pathway rather than the p38 pathway increased Runx2 phosphorylation and transcriptional activity in MC3T3-E1 cells [18]. The MKK3/6-p38 pathway induces phosphorylation of Runx2 to promote osteogenesis of osteoblasts [25]. DDR1-null mice showed upregulation of p42/44 ERK and p38 to increase mesangial cell proliferation [27]. However, the role of Ddr1 in osteoblasts remains undefined. In this study, we demonstrated that knocking out Ddr1 in osteoblasts in vivo and knocking down Ddr1 in MC3T3-E1 preosteoblasts in vitro decreased the expression of osteogenesis markers, such as Runx2, BMP2, ALP, Col-I, and OC. Overexpression experiments showed that DDR1 promoted osteogenesis of osteoblasts via p38 to induce Runx2 protein expression. First, we demonstrated that DDR1 regulates p38 in osteoblasts by knocking down Ddr1 in MC3T3-E1 cells and observing no effect on ERK phosphorylation but downregulation of p38 activity. These results indicated that DDR1 regulates osteogenesis through the p38 MAPK pathway but not the ERK pathway. However, the mechanism by which Ddr1 acts through p38 to regulate Runx2 protein levels in osteoblasts is still unknown. GATA4 (a zinc finger transcription factor) promotes Runx2 gene expression to stimulate bone mineralization and regulates ALP activity during osteoblast differentiation [28]. Knocking down GATA4 resulted in decreased levels of phosphorylated p38 and disruptions in osteoblast differentiation and bone remodeling [29]. Taken together, these data have led us to propose that Ddr1 may act through GATA4 via p38 to regulate Runx2 activity and induce the expression of downstream markers related to osteogenesis during osteoblast differentiation and mineralization.
In the intramembranous ossification process, the bone collar from the perichondrium becomes the periosteum, which differentiates into osteoblasts. In the calcification zone in the growth plate, osteoprogenitor cells enter the area and secrete bone matrix as a scaffold to form trabecular bone. Ddr1 is expressed at the periosteal collar of the clavicle and is especially dense in ribs in E18.5 mice [10]. In this study, the cortical and trabecular bones of OKO∆Ddr1 mice were significantly smaller than those of Ddr1 f/f-4OHT mice at 4 weeks of age. Additionally, the cell numbers of osteoblasts were reduced in cortical bone from OKO∆Ddr1 mice. The difference in cortical bone thickness between Ddr1 f/f-4OHT and OKO∆Ddr1 mice was due to a decrease in osteoblast activity, which caused a delay in ossification in the cortical and trabecular bone. Although the BMD had no significant differences between control and OKO mice, the thickness of Ddr1 f/f-4OHT was thicker than in OKO∆Ddr1 mice, and the BMC was significantly decreased in OKO∆Ddr1 mice. It showed that although BMD had not altered in OKO∆Ddr1 mice, the mouse size was smaller than the wild-type which caused the BMC decrease in OKO∆Ddr1 mice. Our results showed that the expression levels of chondrocyte terminal differentiation markers such as Col-X and MMP13 in the hypertrophic zone of the growth plate were not significantly different in the OKO∆Ddr1 mice compared to the Ddr1 f/f-4OHT mice, which indicated that knocking out Ddr1 in osteoblasts had no influence on the function of the chondrocytes.
It was reported that the osteoblasts had a specific enhancer of the Col-I gene in mice [30,31], and they generated a triple mutant with 2.3-kb. Col-I Cre promoter mice have time-and tissue-specific in osteoblasts and odontoblasts [32]. The Col-I Cre mRNA was firstly expressed at E13.5 in embryos which was the same with the expression of Col-I in the differentiation of osteoblasts and bone formation. The Cre recombinase specific was expressed at the cranial bones, facial bones, ribs, molars and limbs but not expressed at muscle, brain, chondrocyte in growth plate, skin, liver and kidney which also express Col-I. Thus Col-I cre mice have the most properties for our experiment.

Double Staining Analysis for Skeleton
Two-and 4-week-old Ddr1 f/f-4OHT and OKO∆Ddr1 mice were harvested, and their skeletons were fixed in 4% paraformaldehyde for 24 h followed by soaking in freshly prepared 2% Alcian blue 8GX (A5268, Sigma-Aldrich) in 1% acetic acid for 2 days. Then, the skeletons were washed in 0.5% KOH (60377, Sigma-Aldrich) in 1X PBS until the muscles became transparent. Finally, the skeletons were stained with freshly prepared 1% Alizarin red S (A5533, Sigma-Aldrich) and for 10 min and washed with 0.5% KOH to removed excess dye. The skeletons were observed under a Leica-DSM1000 microscope (Leica Microsystems, Wetzlar, Germany).

Microcomputed Tomography (Micro-CT)
High-resolution microcomputed tomography (micro-CT) analysis (SkyScan 1076; SkyScan NV, Kontich, Belgium) was used to render 3-D reconstructions of 4-week-old mouse tibias with the following conditions: isotropic voxel resolution, 9 µm; 0.5 mm aluminum filter; tube voltage, 50 kV; tube current, 200 µA; and exposure time, 1100 ms. A scale of 0-0.09 (NRecon version 1.6.1.7; SkyScan NV) was used to reconstruct the 3-D morphometric parameters of 5 mm of the tibia. For the trabecular analysis, the inner circle of the ROI (2 mm circle; 120 cuts) beneath the growth plate from 0.5 mm to 2.5 mm (total length of 2 mm) was selected for analysis of the trabecular bone mineral density (BMD, g/cm 3 ), percent bone volume (BV/TV, %), trabecular thickness (Tb.Th, µm), trabecular number (Tb.N, mm) and trabecular separation (Tb.Sp, µm). A total length 2 mm outer circle of the ROI at the middle of the tibia was selected for cortical analysis of the cortical bone volume (BV, mm 3 ), bone mineral content (BMC, mg/mm), and bone mineral density (BMD g/cm 3 ).

Three-Point Bending Test
The 4-week-old mouse femurs were subjected to 3-point bending and tested to failure by using a materials testing system (INSTRON 5943, Norwood, MA, USA). The femur was placed on 2 lower supports 2 mm apart. The load was placed at the middle of the femur at a displacement rate of 0.05 mm/s until failure. Quantitative analysis was performed with Bluehill software (Illinois Tool Works Inc., Glenview, IL, USA).

Cell Culture and Drug Treatment
The mouse preosteoblast cell line MC3T3-E1 (CRL-2593) (ATCC, Manassas, VA, USA) was grown in α-minimal essential (α-MEM) medium supplemented with 10% fetal bovine serum (FBS) and 100 units/mL penicillin and was maintained at 37 • C with 5% CO 2 . SB203580 (Ciliobrevin A, Selleckchem, Houston, TX, USA), an inhibitor of phosphorylated p38, was dissolved in DMSO as a stock solution, and the cells were treated with the inhibitor at a final concentration of 20 µM for 16 h, after which the cells were harvested for Western blot analysis. system (BIO-RAD Laboratories Inc., Hercules, CA, USA). GAPDH was used as a housekeeping gene, and expression was normalized to that in Ddr1 f/f-4OHT mice.

Mineralization Assay
To generate stable cell lines, cells were transduced with Ddr1 lentivirus particles, subcultured and seeded at a density of 1 × 10 4 cell/cm 2 in 48-well plates. Once cells reached 80% confluence, the medium was changed to induction medium (α-MEM with 10% FBS, 100 unit/mL penicillin/streptomycin, 50 µM ascorbic acid, and 5 mM β-glycerophosphate) containing 1 mg/mL puromycin. The selective induction medium was replaced every 2 days for 21 days. Next, the cells were washed in PBS, fixed with 10% formaldehyde for 10 min, washed with ddH2O at least 5 times and stained with 2% Alizarin red (A5533, Sigma-Aldrich) solution (pH 4.1) for 5 min. After staining, the cells were washed in ddH2O and imaged with a camera. Quantitation of the staining intensity was performed by adding 200 µL of 10% acetic acid per well and shaking the plates for 10 min, after which 100 uL of the solution was added to 96-well plates per well and the OD415 was measured with a TECAN Sunrise ELISA reader (Switzerland) (n = 6).

Alkaline Phosphatase Assay
MC3T3-E1 cells were transduced with Ddr1 lentivirus and seeded at a concentration of 1*10 4 in 48-well plates. Next, induction medium with 1 mg/mL puromycin was replaced every 2 days for 21 days. An ALP assay kit (86R, Sigma-Aldrich) was used according to the manufacturer's instructions. The culture medium was removed, and the cells were washed once in PBS. Lysate buffer was added to the suspension, which was then centrifuged at 12,000 rpm. The pellet was resuspended and transferred to 96-well plates, to which ALP substrate solution (50 uL/per well) was added. The plates were incubated at 37 • C with a 5% CO 2 humidified atmosphere after which the absorbance at 405 nm was detected by a TECAN Sunrise ELISA reader (Switzerland) (n = 6).

Statistical Analysis
All animal experiments are expressed as the mean (SEM) of 6 independent groups. Each condition was repeated at least three times, and the total number of mice per group was more than 6. Statistical significance was determined using Student's t-test, and multiple comparisons were performed using Scheffé's method. In this study, the groups were compared with the Ddr1 f/f-4OHT group. For the p values, (*) indicated p < 0.05, and (**) indicated p < 0.01, and (***) indicated p < 0.001.

Conclusions
In conclusion, by using overexpression and knockdown of Ddr1 in osteoblasts, we demonstrated that DDR1 in osteoblasts regulates osteoblast differentiation via phosphorylation of p38 to regulate the expression of Runx2 and downstream markers related to osteogenesis, such as Runx2, BMP2, ALP, Col-I, and OC. This mechanism results in the reduction in long bone ossification and weakened mechanical properties in OKO∆Ddr1 mice in vivo, which indicate that Ddr1 plays a crucial role in osteoblast function to regulate the differentiation of osteoblasts. Our findings may be applied to patients with osteoporosis or fracture as a basis for developing Ddr1-targeted therapy.