Main

Endochondral bone formation (EBF) and homeostasis require a regulated program of mesenchymal condensation, chondrocyte differentiation, vascular invasion, cartilage resorption, osteoprogenitor recruitment and differentiation and bone remodeling. These key processes are under the control of essential bone transcription factors (TFs), including RUNX2, SP7 and homeodomain protein families such as HOX, MSX and DLX.1, 2, 3 DLX proteins are important regulators of developmental and differentiation processes, including skeletal development.4, 5, 6 A missense mutation in DLX5 leads to split hand and foot malformation7 and DLX5 and DLX6 are positive transcriptional regulators of osteochondroblastic differentiation.6 DLX3 is defined as an osteogenic regulator, as human mutations in DLX3 lead to tricho-dento-osseous (TDO) syndrome, an ectodermal dysplasia that causes increased bone mineral density (BMD) in intramembranous and endochondral bones.8 In vitro, osteocalcin (Ocn), Runx29, 10 and osteoactivin9, 10, 11 are directly regulated by DLX3, and overexpression of DLX3 in osteoprogenitors stimulates transcription of osteogenic markers.9

Recently, we investigated the effects of neural crest deletion of Dlx3 in craniofacial bones.12 The gene signature of craniofacial bones from Wnt1cre:Dlx3 neonates predicted increased bone formation and mineralization. This was further supported by ex vivo assays on frontal bone osteoblasts, suggesting an inhibitory role for DLX3 in osteoblastic differentiation.12 Contrary to this prediction, adult mice exhibited decreased BMD and increased porosity in neural crest-derived craniofacial bones.12

A transgenic mouse model expressing the TDO DLX3 gene mutation driven by the osteoblast-specific 2.3 Col1A1 promoter was characterized by Choi et al.13 This model resulted in increased trabecular bone volume in young and adult mice; however, the diaphysis phenotype was not described. Despite bone marrow stromal cells (BMSCs) from mutants exhibiting enhanced osteoblastic differentiation and increased bone marker expression ex vivo, the dynamic bone formation rate was not increased in vivo and the trabecular phenotype was attributed to decreased osteoclast formation and bone resorption activity due to the increased serum levels of IFN-γ.13

Taken together, these data highlight an important role for DLX3 in bone; however, the studies above12, 13 provide contrasting results and emphasize that the function of DLX3 in the postnatal and adult skeleton has not yet been fully elucidated. To address the in vivo role of DLX3 in osteoblastogenesis, bone density, and remodeling in the appendicular skeleton, we generated conditional knockouts (cKOs) of Dlx3 in mesenchymal cells (Dlx3Prx1-cKO) and in osteogenic lineage cells (Dlx3OCN-cKO). In both models, Dlx3cKO mice experienced a significant increase in bone mass accrual throughout their lifespan associated with enhanced osteoblast activity. By combining in vivo gene profiling and ex vivo cellular analyses, we establish a newly defined role of DLX3 as a major regulator of bone apposition and homeostasis.

Results

Deletion of Dlx3 in osteogenic lineage cells leads to increased bone mass accrual

High levels of DLX3 were found in osteoprogenitor cells, bone-forming osteoblasts and matrix-embedded osteocytes in both endochondral developing bones and the postnatal skeleton (Figure 1). To address DLX3 function in osteoblastogenesis, Dlx3 was deleted in mesenchymal cells using Prx1-cre mice and in osteogenic lineage cells with OCN-cre mice, resulting in Dlx3Prx1-cKO and Dlx3OCN-cKO mice. We validated the temporal and tissue-specific expression of the Cre transgenes (Supplementary Figure S1). Deletion of Dlx3 was confirmed by Q-PCR and western blot (Figures 2a–c). Endochondral bones of Dlx3OCN-cKO neonates showed no obvious defects in developmental patterning (missing or transformed bone) or gross abnormalities in mineral deposition (P0.5 and P2.5; Supplementary Figure S2).

Figure 1
figure 1

Temporal and spatial Dlx3 expression during bone development. (A) DLX3 localization at E14.5 was shown by LacZ expression in Dlx3+/KinLacZ embryos using whole-X-gal staining (a). DLX3 expression in bone collar is shown in benzyl-benzoate-cleared X-gal-stained forelimb (b) and in longitudinal section of the radius (c). Asterisk: LacZ showed DLX3 was expressed in skin as has previously been reported.48 (B) DLX3 expression at E16.5 was shown by LacZ detection in longitudinal sections of Dlx3+/KinLacZ humeri (blue staining) (a, insert 1) coupled with toluidine blue staining to visualize the cartilage matrix (purple staining) (b). Immunohistochemistry was performed on E16.5 Dlx3+/+ tibias using DLX3 antibody (c). Hypertrophic chondrocytes in the growth plate and osteoblastic cells in perichondrium, primary spongiosa and cortical bone are shown in higher magnification in (c, inserts 2 and 3). (C) DLX3 localization in P1.5 Dlx3+/KinLacZ mouse shown by whole-X-gal staining in benzyl-benzoate-cleared calvaria (a), ribs (b), manus (c) and tibia (d). Longitudinal sections of the X-gal-stained tibia (e) coupled with toluidine blue staining (f). Hypertrophic chondrocytes in growth plate and osteoblastic cells in perichondrium, primary spongiosa and cortical bone are shown in higher magnification in (C:e, inserts 1 and 2). (D) DLX3 protein expression is detected at 5 wk by immunohistochemistry with DLX3 antibody on Dlx3+/+ tibia (a). Higher magnifications showed hypertrophic chondrocytes in the metaphysis (D, insert 1), active surface osteoblasts in the trabecular bone area (a, insert 2), endosteal (a, insert 3) and periosteal (a, insert 4) surfaces of the diaphysis, and osteocytes in the cortical bone (a, insert 4). Scale bars: 100 μm for the main images (letters), 20 μm for the inserts (numbers). Eosin was used as counterstaining in X-gal-stained sections and hematoxylin was used for immunochemistry

Full size image
Figure 2
figure 2

Altered bone formation in Dlx3OCN-cKO mice. (a) Q-PCR of Dlx3 in long bones of P3.5 and P9.5, (b) and in metaphysis (Meta) and diaphysis (Dia) of 5 wk Dlx3OCN-cKO and Dlx3+/+mice. (c) Western blot of DLX3 in tibia, Dia and Meta of 5 wk mice. P2.5 skin was control for bone-specific deletion. Sagittal femur μCT (d), 3D trabecular reconstructions (e) and transverse scans at mid-diaphysis (f) in 5 wk and 6 mo Dlx3+/+ and Dlx3OCN-cKO males. (g) Calcein labeling of cortical (left) and trabecular (right) tibia of 5 wk Dlx3+/+ and Dlx3OCN-cKO males. Scale bars: 1 mm. (h) μCT parameters. Trabecular BMD (Tb. BMD), bone volume ratio (Tb. BV/TV), number (Tb. N), spacing (Tb. Sp) and thickness (Tb. Th) were calculated as were cortical bone mineral density (Cort. BMD), porosity and sub-periosteal and sub-endosteal areas. Data are presented as the mean±S.E.M. ns: non-significant, *P<0.05, **P<0.01, ***P<0.001, ****P<0.001. Scale bars: 1 mm

Full size image

With micro-computed tomography (μCT) analysis, we quantified structural parameters of femurs from 2 weeks (wk) to 6 months (mo) male mice. At 5 weeks, Dlx3OCN-cKO mice showed a striking phenotype of an overall increase in trabecular and cortical bone mass that continued throughout adult life (Figure 2, Table 1). Although no significant difference was observed in the length of femurs, 3D reconstruction showed an increased length of the trabecular bone area that extended into Dlx3OCN-cKO diaphysis (Figures 2d and e). Across age groups, as early as 2 weeks and maintained as late as 6 months, an increase of trabecular bone volume and number of trabeculae in Dlx3OCN-cKO mice was observed (Figure 2h, Table 1). The increased number of trabeculae and connectivity was complemented by decreased trabecular spacing. However, the thickness and BMD (a measurement of mineral content/unit area) of individual trabeculae remained similar to Dlx3+/+ mice, a reflection of normal mineralized bone matrix being produced by osteoblast lineage cells in Dlx3OCN-cKO mice. These results indicate that Dlx3OCN-cKO osteoprogenitors form more trabeculae that extend deeper into the medullary cavity.

Table 1 Summary table of histomorphometric bone parameters in femurs from Dlx3OCN-cKO compared with Dlx3+/+ males
Full size table

Cortical bone quantification showed increased thickness in Dlx3OCN-cKO mice due to a significant increase of the sub-periosteal area, a rich source of osteoprogenitor cells (Figures 2f–h, Table 1). Dynamic bone formation was assessed by calcein injections in 5 wk mice. The mineral apposition rate (MAR) was significantly increased on the periosteal surface of the Dlx3OCN-cKO tibias (3.67±0.18 μm/day and 2.2±0.12 μm/day, respectively (P<0.001)). Calcein labeling of the cortical surfaces further demonstrated an increased bone apposition rate (distance between two calcein bands) in Dlx3OCN-cKO mice, which is an osteoblast-mediated process (Figure 2g). An intense irregular calcein-labeling pattern was observed around trabeculae, indicative of increased metaphyseal bone remodeling in young mice (Figure 2g). However, Dlx3OCN-cKO cortical bone also showed increased porosity and decreased BMD (Figures 2f–h, Table 1). As predicted from Prx1-cre transgene expression (Supplementary Figure S1), Dlx3Prx1-cKO mice exhibited a strikingly similar phenotype to Dlx3OCN-cKO mice with increased trabecular bone formation, increased thickness, decreased BMD and increased porosity in cortical bone (Supplementary Figure S3).

In younger Dlx3OCN-cKO mice (2 and 5 wk), there is highly active bone formation and remodeling to convert woven bone to lamellar bone and this is reflected in the porosity of the cortical bone. The % porosity in the femur decreased from a fold-change (Dlx3OCN-cKO/wild type (WT)) of 1.6 to 1.4 to 1.2 at 2 wk, 5 wk and 4 mo, respectively (Table 1). Thus, the small marrow spaces within the cortical bone observed at 5 wk (Figure 3A) are there to support continued remodeling and viability of a larger mineralized bone area, as shown by hematoxylin and eosin (H&E) and tartrate-resistant acid phosphatase (TRAP) stainings (Figures 3 and 4). Therefore, the decreased porosity in Dlx3OCN-cKO mice with age suggests that cortical bone has matured into adulthood.

Figure 3
figure 3

In vivo effects of Dlx3 deletion in osteogenic lineage cells on bone formation. Paraffin sections of decalcified tibias of 5 wk Dlx3+/+ and Dlx3OCN-cKO males stained with hematoxylin and eosin (A). Osteoblast localization and activity are shown with ALPL antibody (B) on decalcified tibias. Metaphysis and diaphysis are shown in high magnification (inserts 1, 3 and inserts 2, 4, respectively). Scale bars: 500 μm for the main figures (letters), 200 μm for the inserts (numbers)

Full size image
Figure 4
figure 4

Effects of Dlx3 deletion on osteoclastic bone resorption in 5-week-old male mice. (A) Serum levels of CTX-1 were not significantly affected in Dlx3OCN-cKO mice (n=5) when compared with Dlx3+/+ mice (n=3). Bottom panel, TRAP staining was performed on femur sections and computerized images of trabecular bone were used for histomorphometric analysis. TRAP-positive cells (TRAP(+)cells) and TRAP-positive surface (TRAP(+)S) in metaphysis and diaphysis areas were normalized against the matrix bone surface (BS). Both parameters showed no significant difference between Dlx3OCN-cKO mice (n=2) when compared with Dlx3+/+ littermates (n=2). (B) Osteoclasts were visualized using TRAP staining on sections from decalcified and paraffin-embedded 5 wk Dlx3+/+ and Dlx3OCN-cKO femurs. Metaphysis (a, b) and diaphysis (c, d) are shown in higher magnification (inserts). (C, D) M-BMMs (M-CSF-dependent bone marrow macrophages) were isolated from femur and tibia from Dlx3OCN-cKO and Dlx3+/+ 5 wk males and cultured in presence of 50 ng/ml M-CSF and various concentration of RANKL (0, 10, 30, 50 ng/ml). (C) TRAP activity staining was performed at D6. (D) mRNA expression of Trap was monitored at D6 by Q-PCR. mRNA levels have been normalized to the expression levels of the housekeeping gene beta-actin and are presented as fold-change, relative to gene expression in Dlx3+/+ M-BMMs with 0 ng/ml RANKL added. (E, F) Q-PCR shows mRNA-fold-change of osteoclastogenesis markers (Mcsf, Rankl, Tnfrsf11b (Opg)) (E) and Opg/Rankl ratio (F) in the metaphysis and diaphysis from femurs of 5 wk Dlx3+/+ and Dlx3OCN-cKO males. Data are presented as the mean±S.E.M. ns: non-significant, *P<0.05. Scale bars: 500 μm for the main figures (letters) and 100 μm for the magnification boxes

Full size image

Increased osteoblast activity and remodeling underlie the Dlx3OCN-cKO phenotype

H&E staining of tibia sections from 5 wk mice confirmed μCT results showing increased trabecular density and revealed blood vessels in thickened Dlx3OCN-cKO cortical bones (Figure 3A, insert 4). Primary spongiosa trabeculae (region in close proximity to the growth plate) and cortical bones were densely lined with alkaline phosphatase (ALPL)-positive cells, supporting that the increase in bone formation resulted from increased osteoblast activity (Figure 3B).

To investigate bone resorption, serum levels of carboxy-terminal type I collagen crosslink (CTX-1), a biomarker of osteoclastic activity were assessed and showed no significant difference between Dlx3+/+ and Dlx3OCN-cKO mice (Figure 4A). TRAP staining was examined to localize osteoclast cells in long bones (Figures 4A and B). TRAP-positive osteoclast counts and TRAP(+) surface normalized by bone matrix surface showed no significant difference between Dlx3+/+ and Dlx3OCN-cKO trabeculae (Figure 4A). In the diaphysis, however, while osteoclast count and TRAP(+) surface were not significantly affected in Dlx3OCN-cKO mice (Figure 4A), histological analysis revealed a variation in osteoclast localization as Dlx3OCN-cKO bone matured in the diaphysis region (Figure 4B). Although Dlx3+/+ mice showed resorptive pit areas along the length of the periosteal surface of the cortical bone, Dlx3OCN-cKO mice showed few TRAP-positive cells on the periosteal surface, whereas notable TRAP staining was detected on the endosteal surface and in vascular channels (Figure 4B). However, in cortical bone surrounding the metaphysis region (Figure 4B), TRAP-stained osteoclasts were evident on the periosteal and endosteal surfaces of both controls and mutants.

To determine whether the altered osteoclast localization in Dlx3OCN-cKO diaphysis was due to cell autonomous defects in osteoclastogenesis, ex vivo M-CSF-dependent mononuclear cells isolated from bone marrow of Dlx3OCN-cKO mice were cultured (Figures 4C and D). Absence of Dlx3 expression in osteoclasts corroborated previous reports.14 Furthermore, no visible differences in TRAP staining or Trap mRNA levels were observed between Dlx3OCN-cKO- and Dlx3+/+-derived cells differentiated into multinucleated osteoclasts (Figures 4C and D). Altogether, these findings were consistent with the CTX-1 and TRAP quantification results (Figure 4A), and therefore confirmed that alteration in osteoclast localization did not result from osteoclast-cell autonomous defects.

Expression of genes associated with signaling between osteoblasts and osteoclasts was evaluated by Q-PCR (Figures 4E and F). The metaphysis of femurs of 5 wk Dlx3OCN-cKO and Dlx3+/+ males showed no significant difference in mRNA expression of Mcsf (Csf1), Rankl (Tnfs11) and Opg (Tnfrsf11b) or in the ratio of Opg/Rankl expression (index of osteoclast remodeling). These results were in line with the metaphyseal osteoclast counts and indicated that the increased Dlx3OCN-cKO trabecular bone volume did not arise from decreased osteoclast number or activity. In contrast, Dlx3OCN-cKO diaphysis showed increased Mcsf mRNA, significantly increased Opg, and a high Opg/Rankl ratio when compared with controls (Figures 4E and F). These results indicated that the alteration in osteoclast localization was associated with the variation in expression of osteoclastogenesis markers in cortical bone and suggested that the decreased osteoclasts on the active bone-forming sub-periosteum region of Dlx3OCN-cKO diaphysis might result from enhanced secretion from osteoblasts of the osteoclast inhibitor osteoprotegerin (OPG).

RNA sequencing (RNA-Seq) and chromatin immunoprecipitation (ChIP)-Seq of DLX3-deficient bones identify DLX3 molecular targets that regulate bone mass

To understand the DLX3-dependent molecular mechanisms involved in EBF, a global gene profiling in metaphysis and diaphysis of Dlx3OCN-cKO femurs was performed using RNA-Seq and significantly affected genes (fold-change (FC) ±1.85, q<0.05) were analyzed (Figure 5; Supplementary Figure S4). Data sets have been added to the GEO database (GSE53105). Dlx3OCN-cKO metaphyses and diaphyses showed upregulation of Emilin3, a regulator of mesenchymal commitment and osteoprogenitor differentiation,15 as well as upregulation of late stage-osteoblast regulators of mineralization and calcium phosphate balance such as Vdr16 and Enpp1.17 Genes encoding enzymes that promote extracellular matrix turnover such as Adamts15 and 18 were also upregulated. The enhanced expression of these enzymes was consistent with remodeling of newly formed bone to mature bone.

Figure 5
figure 5

RNA-Seq and ChIP-Seq on DLX3-deficient bones and osteoblasts identify molecular targets of Dlx3 regulation of bone mass. RNA-Seq differential gene expression profiling was performed on femurs from 5 wk Dlx3+/+ and Dlx3OCN-cKO males. Selected genes differentially expressed in the metaphysis (a) and diaphysis (b) of Dlx3OCN-cKO compared with Dlx3+/+ mice genes are organized by genes involved in ECM and genes related to the osteoblastic differentiation and ossification processes. (c) ChIP-Seq analysis performed in SMAA-positive BMSCs shows DLX3 binding to bone-related genes differentially regulated in Dlx3OCN-cKO mice identified by RNA-Seq. Tracks for Dlx5, Dlx6, Sp7, Ibsp, Enpp1, Adamts18 and Tnfrsf11b (Opg) were visualized on UCSC genome browser with Refseq displayed for gene annotation and H3K4me3 track provided as reference

Full size image

In the metaphysis, Alpl and the SIBLING genes Ibsp and Spp1 that are associated with bone formation, matrix mineralization and remodeling18, 19 were upregulated (Figure 5a). Other upregulated genes included TFs crucial for the regulation of osteoblast differentiation and ossification, such as Runx2, Sp7, Dlx5 and Dlx6 (Figure 5a)1, 3 and for long-bone development and patterning such as the Hox gene family members HoxA3/B2/B3/B4 (Supplementary Figure S5A).20 In Dlx3OCN-cKO diaphysis, genes related to bone mass were upregulated, particularly Mepe, a SIBLING member and regulator of bone formation and mineralization.21 Finally, genes encoding proteins associated with key-signaling pathways for bone development, such as WNT, TGFb/BMP and NOTCH signaling were enriched in Dlx3OCN-cKO femurs (Supplementary Figure S5B).22, 23, 24, 25 To investigate if the aforementioned genes are targets of a DLX3-regulatory network during EBF and homeostasis, a ChIP-Seq profile of osteoblasts cultured from bone marrow was performed. Our results showed that DLX3 was bound to the promoter of Dlx5, Dlx6 and Sp7, and multiple regions of the Ibsp, Enpp1, Adamts18 and Opg genes, supporting direct regulation of these genes by DLX3 in osteoblasts (Figure 5c).

Ex vivo differentiation of osteoprogenitors supports that DLX3 functions as an attenuator of osteoblastogenesis

To confirm that DLX3 regulates osteoblasts in a stage-specific, osteoblast cell autonomous fashion, calvarial osteoblasts and BMSCs from long bones were isolated from Dlx3F/F neonates and adolescents, respectively, and infected with adenovirus (Adv) containing GFP (Adv-GFP) or Cre recombinase (Adv-Cre) (Figure 6; Supplementary Figure S6). Q-PCR and western Blot showed Dlx3 was efficiently excised in Adv-Cre cells (Figures 6a and b; Supplementary Figure S6A). DLX3 deletion in both cell models resulted in increased ALPL activity and, in calvarial cells resulted in higher cell autonomous mineralization capacity (Figure 6c; Supplementary Figure S6B). Expression of osteoblast differentiation and maturation markers Alpl and Ibsp was increased in Adv-Cre calvarial cells and BMSCs at later time points (Figure 6d; Supplementary Figure S6C). Dlx3 deletion in BMSCs was also associated with significantly increased expression of the TFs Runx2 and Dlx5 and the regulator of matrix mineralization Enpp1 (Supplementary Figure S6C). Our results show that Dlx3 deletion in osteoblasts results in an enhancement of osteoblast differentiation analogous to the in vivo Dlx3OCN-cKO phenotype.

Figure 6
figure 6

Dlx3 excision in calvarial osteoblasts increases expression of osteoblast markers. (a) Q-PCR of Dlx3 mRNA expression represented as fold-change in Adv-GFP and Adv-Cre-infected calvarial osteoblasts from Dlx3F/F neonates during proliferating (D4) and matrix maturation stages (D12). (b) Western blot of DLX3 and Lamin B in virus-infected calvarial osteoblasts at D12. (c) ALPL and Von Kossa stainings of virus-infected calvarial osteoblasts demonstrated increased ALPL activity and mineralization capacity in Adv-Cre-infected cells. (d) Q-PCR of mRNA expression represented as fold-change of osteoblast-related markers (Alpl, Ibsp, Ocn) and bone transcription factors (Runx2, Msx2, Dlx5). Data are presented as the mean±S.E.M. *P<0.05, **P<0.01, ***P<0.001

Full size image

DLX3 alters osteoblast differentiation by modulating the binding of bone-activator TFs on regulatory elements on the OCN promoter

The Ocn gene is repressed in proliferating osteoprogenitor cells and highly induced in mature osteoblasts, producing a mineralized matrix. DLX3 is primarily bound to the Ocn promoter at the onset of osteoblastogenesis, whereas DLX5 associates with Ocn during mineralized tissue formation.9 We examined by ChIP the proximal Ocn promoter containing the OC box (Figure 7), and observed that DLX3-deficient calvarial osteoblasts had a significant increase in the amount of RUNX2 occupying the Ocn promoter in both the proliferative stage (Day 4, 5) as well as matrix maturation stages (Day 9, 12, 18) compared with Adv-GFP cells (Figure 7b). Because other homeobox proteins such as DLX5 have been demonstrated to bind to the same regulatory region and regulate Ocn gene expression,9 we investigated if the promoter occupancy of DLX5 was altered in cells lacking DLX3. In proliferating (Day 4, 5) DLX3-deficient osteoprogenitors, DLX5 occupancy of the Ocn proximal promoter was significantly increased (Figure 7b) but did not show an appreciable difference during the differentiation/matrix maturation stages (Day 9–18) that corresponds to an exponential increase in Ocn mRNA transcription. Taken together, these data suggest that Ad-Cre excision of Dlx3 in calvarial osteoblasts resulted in increased expression of osteoblast-related markers and an increased occupancy of DLX5 at the Ocn homeodomain element as well as the increased and earlier occupancy of RUNX2.

Figure 7
figure 7

Increased occupancy of RUNX2 on the osteocalcin promoter in DLX3-deficient calvarial osteoblasts. (a) Diagram of the mouse osteocalcin promoter displaying relative binding sites and primer sites used for chromatin immunoprecipitation analysis. (b) Calvarial osteoblasts isolated from Dlx3F/F mice were infected with Adv-Cre or Adv-GFP and cultured in osteogenic media for 18 days. ChIP was then preformed on cleared cell lysates on the indicated day using ∼5 μg of RUNX2, DLX3, DLX5, or non-specific IgG antibody. Recovered DNA was then quantified by Q-PCR and normalized to input. Data are presented as the mean of three experiments±S.E.M. *P<0.05, **P<0.01, ***P<0.001

Full size image

Discussion

Herein we demonstrate a distinct role for DLX3 as an essential regulator of three integral processes of EBF and maintenance: osteoblastogenesis, matrix deposition and bone homeostasis. In Dlx3cKO mice, the coordinated process of EBF at the growth plate was retained, although the amount of bone tissue deposited was significantly enhanced. Trabecular bone extended into the medullary cavity; ALPL staining was enhanced, and bone volume ratio and trabecula number was increased. Cortical bone was also notably thicker with an increased MAR but also featured increased porosity and decreased BMD.

We previously reported the function of DLX3 in intramembranous bones of neural crest origin in the craniofacial area. Similar to the present study, Wnt1cre:Dlx3 mice exhibited decreased BMD in mandibular and calvarial bones and increased porosity in mandible.12 Furthermore, here we show that adenoviral-Cre-infected calvarial cells and long-bone BMSCs isolated from Dlx3F/F neonates and 5 wk mice, respectively, showed increased differentiation capacity as demonstrated by ALPL staining, which is consistent with calvarial cell differentiation from Wnt1cre:Dlx3 neonates.12 However, craniofacial bones from Wnt1cre:Dlx3 neonates and femurs from Dlx3OCN-cKO 5 wk mice had a particular gene signature where few genes including Alpl, Ibsp, Mepe, Ihh and Adamst18 were commonly affected. We presently show that although expression of Runx2 and Dlx5 were significantly upregulated in Dlx3-deleted BMSCs from long bones of 5 wk Dlx3F/F, these genes were not affected in Dlx3-deleted calvarial cells isolated from Dlx3F/F neonates. This indicates that Dlx3 regulates osteoblast differentiation in intramembranous and endochondral bones via tissue- or stage-specific molecular mechanisms. This hypothesis is further supported by our transcriptome studies which reveal that in vivo Dlx3 deletion in osteogenic cells affects a wide range of developmental signals including Wnt, Tgf-β/Bmp and Notch pathways, but only few genes encoding signaling molecules are commonly affected in the Wnt1cre:Dlx3 and Dlx3OCN-cKO models.

Appendicular skeleton and craniofacial bones display differential molecular fingerprints,26 one notable difference being that in the appendicular skeleton skeletal patterning and osteoblastogenesis are regulated by HOX TFs,20 whereas most of the craniofacial bones are Hox free. This Hox status would be determinant in osteogenic potential and cellular plasticity.27 In the present study, Dlx3 deletion resulted in upregulation of HoxA3/B2/B3/B4. Our results support that the transcriptional network required for proper EBF involves a direct or indirect inhibition by DLX3 of the Hox gene family and their downstream Hox-specific gene regulatory network. In line with this hypothesis, Hassan et al.28 previously demonstrated the selective association of HOXA10 and DLX3 to the Runx2 and Ocn regulatory promoter regions, an association correlated with the stages of osteoblast maturation. We hypothesize that this difference in HOX regulation, along with aforementioned differentially regulated TFs and osteoblastogenic markers contribute to the tissue-specific differences in DLX3-dependent network regulation between intramembranous and endochondral bones.

In the TDO model, transgenic mice lacked an enhanced dynamic bone formation rate and the increased bone volume and BMD trabecular phenotype was attributed to decreased osteoclast bone resorption activity due to increased IFN-γ.13 In contrast, our analyses in Dlx3OCN-cKO mice support that increased trabecular bone mass does not arise from impaired osteoclastic activity but from direct enhancement of the bone-forming osteoblast activity. This leads to accelerated bone formation, thereby inducing an imbalance in homeostasis in favor of bone apposition.

In support of increased osteoblast activity due to Dlx3 deletion, RNA-Seq demonstrated that in Dlx3OCN-cKO metaphysis, positive TFs of osteoblastogenesis were upregulated, including Runx229 and its downstream osteoblast-specific target Sp7.30 Dlx5 and Dlx6, two biologically equivalent positive regulators of chondrocyte and osteoblast differentiation6, 31, 32, 33, 34 were also upregulated in Dlx3OCN-cKO metaphysis. Given that all four networked genes have been characterized as stimulators of osteoblast differentiation and activity, we hypothesize that increased Dlx5, Dlx6, Runx2 and Sp7 expression in Dlx3OCN-cKO mice contributes to the significantly increased trabecular bone volume and number of trabeculae. This hypothesis is further supported by our in vitro model; Dlx3-deleted BMSCs showed an increased osteoblast differentiation and bone-forming activity associated with increased gene expression of Runx2 and Dlx5. Both our in vivo and in vitro data, together with previous studies reporting in vivo loss of Dlx3 expression in Dlx5/Dlx6−/−mutants35 suggest a mutually negative regulatory loop between DLX5/DLX6 and DLX3 during bone formation and homeostasis. ChIP analysis on BMSCs demonstrated that DLX3 binds to the promoters of Sp7, Dlx5 and Dlx6. These data together with previous in vitro studies showing that DLX3 can bind to the Runx2 promoter and directly modulate its transcription10 lead us to hypothesize that in long bones, DLX3 could attenuate osteoblastogenesis by reducing Runx2, Sp7, Dlx5 and Dlx6 gene expression.

Upregulation of several bone ECM genes, including Ibsp and Spp1 and local regulators of bone mineralization such as Alpl and Enpp1 was also observed in Dlx3OCN-cKO metaphysis. ChIP analysis of long-bone cells demonstrated potential DLX3 binding to Enpp1 and Ibsp promoters. Enpp1 is necessary for osteoblast differentiation and inhibits bone mineralization.36 Furthermore, Enpp1−/− mice displayed trabecular bone loss and hyper-mineralization in long bones.17 Ibsp−/− mice displayed impaired bone growth and mineralization and dramatically reduced bone formation.37 We postulate that increased Enpp1 and Ibsp expression in Dlx3OCN-cKO metaphysis could contribute to the increased trabeculae phenotype. Collectively these data support that DLX3 regulation of ECM genes and major regulators of bone mineralization is necessary for conventional matrix formation.

Different stages of osteoblast differentiation – proliferation, maturation and mineralization – are governed by coordinated TFs binding on promoters of bone-related genes to regulate their temporal expression.28 In vitro studies have shown that DLX3 and DLX5 associate with the Ocn promoter at the onset of transcriptional activation, concomitant with RUNX2 occupancy, and whereas DLX3 occupancy on the Ocn promoter decreased from osteoblast maturation to mineralization, DLX5 occupancy was maximal during the mineralization stage.9 In the present study, we demonstrated that Dlx3 deletion in calvarial osteoblasts results in increased occupancy of DLX5 and increased and premature occupancy of RUNX2 on regulatory elements on the Ocn promoter. These data suggest that Dlx3 deletion could directly affect the network of TF molecular switches and promote osteoblastic differentiation and bone-forming activity via an enhancement of the occupancy of bone-activator TFs such as DLX5 and RUNX2. Among the genes differentially expressed in Dlx3OCN-cKO metaphysis, DLX5 has been shown to regulate and bind to the promoter regions of Runx2,10 Hox family members,28 Sp7,38 Alpl39 and Ibsp.40, 41 These data together with our ChIP results demonstrating DLX3 binding to some of these genes lead us to speculate that DLX3 and DLX5 may have a coordinated role in the transcription of bone-related genes via molecular switches at their promoter regions.

We also performed in-depth characterization of the Dlx3OCN-cKO cortical bone and found increased thickness associated with increased periosteum MAR, decreased BMD and increased porosity; also, these porous regions were vascularized. It remains to be determined whether this increased vascularization and decreased BMD result from the high bone mass phenotype of Dlx3cKO mice or from the upregulation of modulators of bone mineralization and porosity such as Enpp1.17 In cortical bone, RNA-Seq showed upregulation of Opg and Mepe, two major inhibitors of osteoclastogenesis strongly expressed by late-differentiated osteoblasts and osteocytes.21, 42, 43, 44 ChIP analysis revealed that DLX3 was bound to multiple regions of Opg. This supports a model in which Dlx3 deletion in osteoblasts induces increased levels of Opg and Mepe in cortical bone, and thereby deregulates normal homeostasis in favor of a positive bone balance. Samee et al.33 reported that Dlx5−/− mice had an increased number of osteoclasts and Dlx5−/− osteoblasts exhibited decreased Opg expression resulting in a higher Rankl/Opg ratio. This suggests that, as previously shown for DLX5,33 DLX3 is a central regulator of osteoblast/osteoclast coupling. However, conversely to DLX5, DLX3 would inhibit osteoblast bone-forming activity via a direct or indirect negative transcriptional control of genes responsible for bone formation while simultaneously activating bone resorption through regulation of osteoblastic modulators of osteoclastogenesis.

Collectively, our studies establish that DLX3 has an important role in regulating endochondral bone mass throughout the lifespan. We propose that DLX3 acts as a negative modulator that regulates expression of crucial bone-related genes throughout osteoblastogenesis, bone matrix synthesis and skeleton homeostasis. These results suggest that DLX3 is an attractive translational target for stimulating bone formation in skeletal disorders.

Materials and Methods

Mice breeding

Cre recombinase activity was traced by mating Prx1-cre and OCN-cre mice with Rosa LacZ or YFP mice. For details, see Supplementary information. All animal work was approved by the NIAMS Animal Care and Use Committee.

Histological and immunohistochemical (IHC) analyses

Samples were fixed, dehydrated, decalcified, embedded and cut in 10-μm sections. Toluidine blue stained for bone cartilage. ALPL (R&D Systems, Minneapolis, MN, USA) antibody was used 1 : 200 with secondary anti-rat antibody 1 : 400 (Vector Laboratories, Burlingame, CA, USA). For detailed IHC methods, see Supplementary information.

Bone tissue collection and RNA extraction

Long bones were harvested at P3.5, P9.5 and 5 wk. In 5-wk-male mice, cartilage caps from femur were removed and the metaphysis was harvested. Bone marrow cells were flushed with 1 × PBS using a 27-gauge needle. The remaining cortical bone was harvested as the diaphysis. For RNA extraction details, see Supplementary information.

Q-PCR

cDNA was prepared from RNA and Q-PCR analysis was performed for target genes, run in triplicate for each sample. Gene expression levels were normalized to the expression of beta-actin and presented as FC, relative to WT expression. Oligonucleotides used are summarized in Supplementary Table S1 and detailed Q-PCR methods are in Supplementary information.

Western blot

Primary antibodies used were anti-DLX3 (1 : 2000) and anti-GAPDH (1 : 5000; Abcam, Cambridge, MA, USA). Secondary antibodies used were goat anti-rabbit HRP for DLX3 (1 : 3000) and goat anti-mouse HRP for GAPDH (1 : 5000) (Bio-Rad, Hercules, CA, USA).

μCT analysis

Fixed femurs and tibias from 2 wk to 6 mo Dlx3+/+, Dlx3OCN-cKO and Dlx3Prx1-cKO mice (n=2 mice minimum per group) were scanned at 10 μm voxel resolution (μCT 40; Scanco Medical AG, Brutisellen, Switzerland) and images were reconstructed using Scanco software v5.0. Trabecular and cortical bone areas were selected and analyzed as previously described.45 Parameters were obtained using thresholds range 220–1000 and a density range superior to 500 mg of HA/cm3.

Fluorochrome measurement of bone formation rates

Calcein labeling was performed on 5 wk mice as previously described.45 Distances between two calcein labels were measured at 4–6 points along the periosteal surface of the cortical bone in n=2 Dlx3+/+ and Dlx3OCN-cKO mice. The points were added together and the mean±S.E.M. was plotted for each mouse.

RNA-Seq

mRNA isolated from 5 wk metaphysis and diaphysis was processed with Illumina TruSeq RNA Sample Preparation kit (Illumina, San Diego, CA, USA), and data were generated with an Illumina HiSeq 2000 sequencing system. Criteria for significant gene selection included q-values≤0.05 (for multi-test adjustment), FC≥1.85 or ≤−1.85, and mean RPKM≥1 for the mutant group in selecting upregulated genes, or for the control group in selecting downregulated genes. For detailed RNA-Seq methods, see Supplementary information.

Measurement of serum CTX-1 level

The CTX-1 Ratlaps ELISA was performed with serum from 5 wk Dlx3+/+ and Dlx3OCN-cKO mice according to manufacturer protocol (Immunodiagnostic Systems Inc., Scottsdale, AZ, USA).

TRAP activity assay

A TRAP activity assay was performed as previously described.46 For detailed quantification methods, see Supplementary information.

M-CSF-dependent bone marrow macrophages (M-BMM) isolation and culture

For osteoclast experiments, BMSCs isolated from femurs and tibias of 5 wk Dlx3+/+ and Dlx3OCN-cKO males were treated with 50 ng/ml of MCSF (R&D Systems) and various concentrations (0, 10, 30 and 50 ng/ml) of RANKL (R&D Systems). For BMSCs and M-BMMs culturing and TRAP staining methods, see Supplementary information.

ChIP and high-throughput sequencing of DLX3-associated DNA in BMSCs

For ChIP-Seq studies (Figure 5), BMSCs were isolated from femurs and tibias of 6-wk-mice expressing mCherry aSMA promoter.47 Cells were initially FACS sorted to enrich for a homogenous population of progenitor cells for osteoblast differentiation. 1 × 108 BMSCs were washed with PBS and then fixed with 1% formaldehyde for 10 min to crosslink DNA-protein complexes. Isolated chromatin was obtained as previously described9 and used for immunoprecipitation with anti-DLX3 antibody (ab66390, Abcam) or IgG as a control followed by Protein-G Dynabeads (Invitrogen, Carlsbad, CA, USA). DNA was amplified using the Illumina system (Illumina) and sequenced on an Illumina Genome Analyzer II. For BMSCs isolation and culturing methods and detailed ChIP data analysis, see Supplementary information.

Calvarial osteoblast culture, excision of Dlx3 allele by adenoviral-Cre recombinase infection

Primary mouse calvarial osteoblasts were isolated from neonates containing two floxed Dlx3 alleles (Dlx3F/F) using sequential collagenase.9 Isolated cells were infected with Cre recombinase or GFP-containing Adv (University of Iowa Gene Transfer Vector Core) at an m.o.i. of 100 for 6 h. Cells were assessed for Dlx3 excision by Q-PCR. For osteogenic differentiation, aMEM was further supplemented with 280 μM ascorbic acid and 10 mM b-glycerophosphate. Whole-cell lysates from Adv-Cre or Adv-GFP infected primary calvarial cells (∼1 × 107 cells) were analyzed via western blot as previously described.9 Primary antibodies anti-DLX3 (C-20, Santa Cruz Biotechnologies, Santa Cruz, CA, USA) or anti-Lamin B1 (1:1000; Invitrogen) were used. Secondary antibodies were species-specific HRP-conjugated antibodies (1 : 10 000; Santa Cruz Biotechnologies). ALPL activity and Von Kossa staining assay were performed as previously described.12 For Q-PCR, cDNA was made and levels were normalized to the expression of beta-actin represented as FC, relative to target gene expression in Adv-GFP calvarial osteoblasts at D4.

ChIP assay in DLX3-deficient calvarial osteoblasts

ChIP assays were performed as previously described.9 Briefly, calvarial osteoblasts infected with Cre recombinase or GFP-containing Adv were collected from Dlx3F/F neonates, crosslinked, lysed and genomic DNA sheared by sonication. Diluted lysates were then immunoprecipitated with protein-specific antibodies; anti-RUNX2 (PEBP2aa) (M-70; sc-10758), anti-DLX3 (C-20; sc-18143), anti-DLX5 (C-20; sc-18152) or IgG as a non-specific binding control (Santa Cruz Biotechnologies). Aliquots of ChIP samples were assayed by Q-PCR to detect the presence of specific DNA fragments using oligos from the mouse OCN (Ocn also known as Bglap2) proximal promoter spanning bp −198 to −28; forward, 5′-GGC AGC CTC TGA TTG TGT CC-3′; reverse, 5′-TAT ATC CAC TGC CTG AGC GG-3′. For each immunoprecipitation, DNA levels were normalized to input. For calvarial cell isolation, culturing and infection, see Supplementary information.

Statistical analyses

FC is defined as (nmutant/nWT) if this ratio is ≥1 or –(nWT/nmutant) if the ratio is <1. Statistical analyses of the Q-PCR data, CTX-1 data and TRAP-related parameters were performed on Prism 5 software (GraphPad Software, La Jolla, CA, USA), using the Mann–Whitney test. μCT data were analyzed with Partek GS 6.6 using two-way ANOVA. A random ‘litter’ factor is included to account for the litter effect. For all tests, the significant P-value limit is 0.05. All quantitative experiments were performed on at least two control and two mutant animals and for cell culture, at least two independent experiments were run. Data are expressed as mean±S.E.M. Statistical analyses for RNA-Seq and ChIP data are described in related sections.