Loss of Osteoblast Runx3 Produces Severe Congenital Osteopenia

Abstract

Congenital osteopenia is a bone demineralization condition that is associated with elevated fracture risk in human infants. Here we show that Runx3, like Runx2, is expressed in precommitted embryonic osteoblasts and that Runx3-deficient mice develop severe congenital osteopenia. Runx3-deficient osteoblast-specific (Runx3^fl/fl/Col1α1-cre), but not chondrocyte-specific (Runx3^fl/fl/Col1α2-cre), mice are osteopenic. This demonstrates that an osteoblastic cell-autonomous function of Runx3 is required for proper osteogenesis. Bone histomorphometry revealed that decreased osteoblast numbers and reduced mineral deposition capacity in Runx3-deficient mice cause this bone formation deficiency. Neonatal bone and cultured primary osteoblast analyses revealed a Runx3-deficiency-associated decrease in the number of active osteoblasts resulting from diminished proliferation and not from enhanced osteoblast apoptosis. These findings are supported by Runx3-null culture transcriptome analyses showing significant decreases in the levels of osteoblastic markers and increases in the levels of Notch signaling components. Thus, while Runx2 is mandatory for the osteoblastic lineage commitment, Runx3 is nonredundantly required for the proliferation of these precommitted cells, to generate adequate numbers of active osteoblasts. Human RUNX3 resides on chromosome 1p36, a region that is associated with osteoporosis. Therefore, RUNX3 might also be involved in human bone mineralization.

INTRODUCTION

Long bones in vertebrates are formed by endochondral ossification via a cartilage intermediate. Other parts of the skeleton, e.g., the flat bones of the skull, are formed via intramembranous ossification, involving the direct differentiation of mesenchymal cells to osteoblasts (OBLs). Osteogenesis depends on the development of bone-forming OBLs (osteoblastogenesis) and is the endpoint in both these processes (1). Osteoblastogenesis proceeds through the commitment of progenitor cells, their differentiation into pre-OBLs, and the final maturation into active OBLs (2). OBLs then differentiate through three major stages: cellular proliferation (mainly of early mesenchymal stem cells), matrix maturation, and mineral deposition (3). In mice, the first OBLs appear in the bone collar (perichondrium) on embryonic day 14 (E14) to E14.5. These cells then ride the invading vasculature and reach the inner calcified matrix around the hypertrophic chondrocytes, where they form the primary spongiosa (1, 4). Primary spongiosa OBLs form the bone trabeculae (i.e., the cancellous part), while bone collar OBLs give rise to the bone cortex (5).

Osteoblastogenesis is tightly controlled by signal transduction and gene expression regulation (3, 6). The runt domain transcription factor (TF) Runx2 is the osteogenic master regulator that hubs the majority of these processes by integrating signals from various developmental cues (7, 8). Thus, Runx2-null mice are completely devoid of OBLs and bones (9–11), and transgenic expression of Runx2 (12) or its two isoforms, Runx2-I and Runx2-II (13), in OBLs results in severe osteopenia and bone fragility due to the inhibition of OBL maturation. Interestingly, recent findings suggest that Runx2 is not required for precommitted OBL lineage cell generation, even though it is mandatory for progenitor cell commitment to the osteogenic lineage (14, 15). This indicates that as-yet-unidentified participants are involved in osteoblastogenesis proper.

Runx3 is another member of the mammalian runt domain TF family, which includes Runx1, Runx2, and Runx3 (16). These TFs bind DNA as heterodimers with the non-DNA-binding subunit core-binding factor β (Cbf-β), which is essential for their proper function (17). More recently, Cbf-β was implicated in bone and skeletal development (18–20). During mouse embryogenesis, Runx3, the evolutionary founder of the Runx gene family (21–23), is expressed in the peripheral nervous system, hematopoietic cells, skin appendages, teeth, and skeleton (24, 25). Accordingly, Runx3-null mice develop sensory limb ataxia (26) as well as airway and colon inflammation (27–29). Although the three Runx genes are expressed in the limb bud mesenchyme, their role in the formation of the cartilaginous anlagen of the skeleton is unclear (30). In contrast, Runx3 along with Runx2 are well-established key regulators of chondrogenesis (30, 31). Ablation of Runx3, which regulates early and late chondrocyte differentiation (32), leads to delayed chondrocyte maturation and vascular invasion in E15.5 Runx3-null embryos (33). On the other hand, Runx3 overexpression in chondrocytes results in ectopic mineralization of the mouse rib cage (31). Likewise, Runx2-null mice develop a plethora of chondrocyte maturation abnormalities (33–35), and mice deficient in both Runx2 and Runx3 lack mature chondrocytes (33), attesting to their cooperative function in this lineage. Thus, the emergence and homeostasis of bone-building chondrocytes and OBLs that share a common mesenchymal osteochondroprogenitor cell (6, 36) are regulated by both Runx family members.

While the central role of Runx2 in osteoblastogenesis is well established (6, 37), the function of Runx3 in this process remains obscure. Given the common OBL-chondrocyte progenitor and Runx3/Runx2 cooperation in chondrogenesis, we hypothesized that Runx3 could also be involved in osteoblastogenesis. Here we report that Runx3 functions in committed OBLs of the developing bone and that its loss leads to reduced numbers of active OBLs, resulting in diminished bone formation and severe congenital osteopenia. Of note, the human RUNX3 gene resides on chromosome 1p36 (38), a region associated with low bone mineral density and osteoporosis (39–45). Thus, RUNX3 deficiency might also affect bone mineralization in humans and may constitute a risk factor for human osteopenia.

MATERIALS AND METHODS

Mice.

Mouse strains used in this study include the previously described Runx3 knockout (KO) (ICR background) (26), LoxP-flanked Runx3 (Runx3^LoxP/LoxP) and Runx3^fl/fl/Pgk-cre (46), and Runx2^+/− (11) mouse strains. Runx3^fl/fl/Col1α1-cre, Runx3^fl/fl/Col1α2-cre, and Runx3^fl/fl/Osx1-cre mice were generated by crossing Runx3^LoxP/LoxP mice to Col1α1-cre (47), Col1α2-cre (48), and Osx1-cre (49) transgenic mice, respectively. Mice used in the experiments were 23-day-old (D23) females, unless stated otherwise. Histopathology was used to verify the absence of colitis in Runx3 KO mice. The different cre-deleter mouse strains were of a mixed genetic background; hence, we used only gender-matching littermates as controls, to account for variation in bone parameters. Control mice were wild-type (WT) (Runx3^+/+) gender-matched littermate mice, unless indicated otherwise. All experiments were conducted according to protocols approved by the Institutional Animal Care and Use Committee of the Weizmann Institute of Science (permit numbers 01370311-2 and 00760108-2).

Alcian blue-alizarin red S staining.

Mouse cadavers were deskinned, eviscerated, ethanol-acetone fixed, and stained with alcian blue and alizarin red S, as previously described (50). Stained cadavers were immersed in a 1% potassium hydroxide solution (in 20% glycerol) until complete clearing of soft tissue. Incubation times were empirically adjusted to the mouse developmental stage (embryonic or postnatal).

Micro-computed tomography.

Femora were scanned with a high-resolution X-ray micro-computed tomography (μCT) system (eXplore Locus SP; General Electric, USA) with source settings at 80 kV and 80 μA. Sixteen-micrometer voxel-size images were analyzed with custom software (MicroView reconstruction version 5.2.2). Cortical and cancellous bone parameters were obtained from a fixed 0.5-mm (32 image slices) transverse section of the diaphysis (the long bone shaft; distally adjacent to the trochanter tertius) and the distal epiphysis (end part), respectively. A direct three-dimensional (3D) model was employed to determine the number, thickness, and separation of trabeculae as well as bone mineral density (BMD) and bone mineral content (BMC).

Whole-bone biomechanical testing.

A 3-point bending assay was performed with an Instron material testing machine (model 3345), as described previously (51). Loading proceeded at a constant rate (2 mm/min) until bone fracture. Force-displacement data were collected with the system software (BlueHill) at 10 Hz. Tibial stiffness (S) was deduced from the slope of the load displacement plot, and the effective Young's modulus value (E) was calculated from the equation E = S · L³/48 · I (where S is bone stiffness, L is support span, and I is the moment of inertia around the bending axis, determined by μCT at the area of fracture).

RNA in situ hybridization (RISH).

Radioactive in situ hybridization on paraffin sections was performed according to standard procedures (31), using previously reported ³⁵S-labeled riboprobes against Runx2, Col1α1 (collagen 1α1), and Bglap2 (osteocalcin) (9). To detect the expression of Runx3, we used an 846-bp probe that spanned the 3′ end of Runx3 (nucleotides 1201 to 2047 of the sequence reported under NCBI RefSeq accession number NM_019732). Hybridization was performed at 55°C (overnight), and sections were washed at 63°C. Autoradiography and Hoechst 33258 staining were performed as previously described (52).

Bone histomorphometry.

Static and dynamic histomorphometry was performed on tibiae and vertebrae of 3-month-old littermate mice. Three Runx3 KO and WT mice were used for the static analyses, and four mice of both groups were used for the dynamic analyses. For assessment of dynamic indices, mice were injected with a freshly prepared calcein solution (catalog number C0875; Sigma) (1 mg/40 g mouse, intraperitoneally [i.p.]) 6 and 2 days before sacrifice. Data were collected with the OsteoMeasure and Trabecular Analysis systems (Osteometrics, Atlanta, GA). Data are reported in accordance with standard nomenclature (53).

Proliferation and apoptosis assays.

To measure the OBL proliferation capacity, we used a bromodeoxyuridine (BrdU) incorporation assay on calvaria sections of 9-day-old Runx3 KO and WT littermate mice, as previously described (54). Briefly, calvariae were fixed (4% paraformaldehyde [PFA] for 24 h) 12 h after BrdU injection (Sigma) (50 μg/g mouse). Following decalcification (14% EDTA for 48 h) and sucrose immersion (20% for 24 h), calvariae were embedded in an optimal-cutting-temperature (OCT) compound (Tissue-Tek), and 5-μm-thick coronal sections were made, fixed in PFA (4% for 20 min), treated with HCl (2 N for 30 min), and blocked (5% horse serum for 30 min). Sections were incubated with a rat anti-BrdU antibody (catalog number MCA2060; Serotec), followed by incubation with secondary Cy3-conjugated anti-mouse IgG (Jackson ImmunoResearch) and mounting (4′,6-diamidino-2-phenylindole [DAPI]-containing medium). Analysis was performed by counting BrdU-positive/DAPI-positive cells in ≥6 calvaria sections from each mouse.

To determine OBL apoptosis, we performed a transferase dUTP nick end labeling (TUNEL) assay on calvaria sections of 14-day-old Runx3 KO and WT littermate mice, as described previously (54). Briefly, 5-μm-thick sections were prepared as described above for the BrdU procedure, and the TUNEL ApopTag kit (catalog number S7165; Millipore) was used according to the manufacturer's instructions. Counterstaining was performed with toluidine blue. Analysis was performed by counting TUNEL-positive cells in ≥5 calvaria sections from each mouse.

Culture and analysis of primary OBLs.

Primary OBL cultures from bone marrow (BM) stromal cells (medullary cavities of humeri, femora, and tibiae) or calvariae were established from 2-month-old or 3- to 4-day-old mice (n = 4 or 5), respectively, as described previously (55, 56), and maintained for 21 to 28 days. Runx3 expression in calvaria-derived OBL cultures was analyzed by Western blotting (see Fig. S1 in the supplemental material). For BM stromal cultures, immunodepletion of adherent CD11b⁺ cells by fluorescence-activated cell sorter (FACS) sorting was performed (FACSAria sorter; BD) 14 days after initial plating (57). To induce osteogenesis, β-glycerophosphate (10 mM) and ascorbic acid (50 μg/ml) were added to culture medium 48 h after the start of culture. To assay bone mineralization, equal numbers of nucleated cells were plated, cultured, and stained with von Kossa stain (58) at different time points during culture growth. To quantify the relative mineralized area, digital images of whole plates were analyzed using ImagePro software (Media Cybernetics, MD). A CFU assay was performed on primary OBL cultures using an alkaline phosphatase (ALP) staining kit (catalog number 00-0009; Stemgent, CA), as described previously (59). ALP-positive colonies were counted in 10 random microscope fields (at a ×100 magnification) to determine osteoprogenitor cell numbers, as described previously(60).

Gene expression analyses.

Total RNA was extracted from calvaria-derived primary OBL cultures of Runx3OBL and WT control mice using the RNeasy minikit (Qiagen, USA). The quality and quantity of RNA were determined using the Agilent 2100 bioanalyzer. RNA was then reverse transcribed, amplified, and labeled with an Affymetrix GeneChip whole-transcript sense target labeling kit. Gene expression data (GeneChip Mouse Gene 2.0 ST array system) were generated according to the manufacturer's protocols (Affymetrix, Santa Clara, CA). Hybridized arrays were scanned using the Affymetrix GeneChip 3000 7G scanner. Custom software (GeneChip analysis suite) was used to analyze the intensity data and calculate a set of absolute metrics. Microarray data were normalized by the Partek genomic suite software. Data in CEL files, containing raw expression measurements, were preprocessed and normalized using the robust multichip average (RMA) algorithm (61). Over- and underexpression were defined as significant changes of ≥1.5-fold with a false discovery rate (FDR) q value of <0.1, applying analysis of variance (ANOVA). Selective validation of microarray expression results was conducted by reverse transcription-quantitative PCR (RT-qPCR), using the primers listed in Table S2 in the supplemental material.

RT-qPCR analysis.

cDNA was synthesized with the QuantiTect reverse transcription kit (Qiagen, USA), using 1 μg of purified RNA, and analyzed by SYBR green qPCR using the miScript SYBR green PCR kit (Qiagen, USA) and the LightCycler 480 system (Roche, USA). Each RT-qPCR experiment consisted of >3 biological repeats using two independently prepared cDNA samples.

Statistical analysis.

Data are presented as means and standard deviations (SD). Statistical analyses were performed with two-tailed Student's t test. Values were considered statistically significant at a P value of <0.05.

Microarray data accession number.

Data sets were submitted to the Gene Expression Omnibus database (www.ncbi.nlm.nih.gov/geo) under accession number GSE57195.

RESULTS

Runx3-null mice show delayed skeletal ossification.

To determine if Runx3 has a role in embryonic bone formation, we stained Runx3-null (Runx3 KO) and wild-type (WT) littermate E18.5 skeletons with alizarin red S and alcian blue (50). Besides having apparently smaller skeletons, Runx3 KO embryos showed delayed ossification of the hind- and forelimb distal bones (Fig. 1A, left) as well as of the mandible (angular process) and skull (Fig. 1A, right). Moreover, the intramembranous skull bones and clavicles of Runx3 KO mice displayed significantly fainter alizarin red A (bone) staining and were apparently softer than those of WT littermate mice (see Fig. S2A in the supplemental material). These findings were also evident by comparative X-ray imaging of 23-day-old (D23) calvariae, demonstrating significant diminution of bone tissue in Runx3 KO skulls (see Fig. S2B in the supplemental material). Comparative analyses of skeletons at earlier embryonic time points demonstrated that Runx3 KO delayed bone ossification could be detected in the limb bone extremities and mandible as of E16.5 using alizarin red S and alcian blue staining (see Fig. S2C in the supplemental material). Notably, while still small, the skeletons of D7 Runx3 KO mice displayed a WT-like ossification pattern (Fig. 1B), as previously reported (33). To further elucidate the role of Runx3 in skeletal development, we analyzed ossification kinetics in E18.5 and newborn (D0.5) heterozygote Runx2^+/− mice with various Runx3 gene dosages (+/+, +/−, or −/− genotype). Interestingly, although E18.5 Runx2^+/− Runx3^−/− embryos were generated at the expected Mendelian ratio, these mice showed considerable mortality, with only a few mice surviving to 1 week of age. Analysis revealed a gene dose dependence between Runx3 and the extent of distal limb bone ossification (Fig. 1C). Specifically, while Runx2^+/− mice showed progressive ossification of the palmar and plantar bones, digital bones of D0.5 Runx2^+/− Runx3^−/− mice remained grossly cartilaginous, with Runx2^+/− Runx3^+/− bones displaying an intermediate phenotype (Fig. 1C). Complementary analyses of E14.5 and E16.5 Runx3^+/− Runx2^+/− and Runx3^+/+ Runx2^+/− developing skeletons (see Fig. S2D in the supplemental material) did not reveal differences in ossification patterns between these two genotypes. Taken together, these findings suggest that Runx3 might serve a nonredundant function in the ossification of the distal limb bones.

FIG 1.

Phenotypic analysis of Runx3 KO mouse bones. (A) Short stature and delayed ossification of the distal limb and mandibular bones of Runx3 KO E18.5 embryos. Alizarin red S (bone) (purple)-alcian blue (cartilage) (blue) staining of forelimbs (left) and hindlimbs (middle) shows calcification of tarsal, metatarsal, and digital limb bones in a WT embryo (red arrows) and a lack of calcification in a Runx3 KO embryo. Similar deficits are seen by comparative μCT imaging of the palmar bones (bottom left). Alizarin red S-alcian blue staining shows a complete lack of ossification of the angular process of the mandibular bone (right, white arrow) in an E18.5 Runx3 KO embryo versus its apparent ossification in a WT littermate embryo. (B) The ossification pattern of a 7-day-old (D7) Runx3 KO skeleton is indistinguishable from that of a WT skeleton. Alizarin red S-alcian blue staining shows a WT-like ossification pattern of the Runx3 KO skeleton. Note the apparently smaller Runx3 KO skeleton. (Top left) Front limbs; (bottom left) hind limbs; (right) skulls. (C) Gene dosage effect of Runx3 on bone ossification in Runx2^+/− mice. Shown is alizarin red S-alcian blue staining of E18.5 and D0.5 palm bones of Runx2^+/− mice with either the Runx3^+/+, Runx3^+/−, or Runx3^−/− genotype. Note the relationship between the Runx3 dosage and the degree of ossification of the palmar and plantar bones at E18.5 and D0.5. (D) Short stature and kyphosis in adult Runx3 KO mice. Pictures of 5-month-old Runx3 KO and WT littermate mice show significantly smaller size and kyphosis in the Runx3 KO mouse. (E) Runx3 KO mice have short bones (left), thin cortices (top right), and underdeveloped trabecular cross sections (bottom right). Shown are representative 3D-reconstructed μCT images of femora of 23-day-old Runx3 KO and WT mice. Dashed lines mark the cortical (red) and trabecular (blue) regions of interest (see Materials and Methods). (F) Runx3 KO mice are severely osteopenic. Comparative μCT analyses of femora of 23-day-old Runx3 KO and WT mice were performed. (a to f) Whole-bone and bone cortex parameters (n = 6 Runx3 KO mice; n = 11 WT littermate mice); (g to i) cancellous bone parameters (n = 4 Runx3 KO mice; n = 6 WT littermate mice). Values are means and SD. Asterisks indicate a P value of <0.01.

Runx3 KO mice exhibit severe congenital osteopenia.

To complement the results of the perinatal analysis described above, we examined whole skeletons of adult Runx3 KO and WT littermate mice using alizarin-alcian staining. Apart from their lasting smaller size, Runx3 KO skeletons could not be distinguished from those of WT littermate mice (see Fig. S2E in the supplemental material). However, the apparent kyphosis, a skeletal condition clinically correlated with osteopenic bones (62) and prominent dwarfism of adult Runx3 KO mice (Fig. 1D), as well as the above-mentioned perinatal intramembranous ossification deficits implied a possible osteoblastic involvement in this phenotype and prompted us to further analyze their bone characteristics.

We then analyzed bone structure and architecture by micro-computed tomography (μCT) imaging of femora from female Runx3 KO and littermate WT mice (Fig. 1E and F). Because Runx3 KO mice manifest colitis at 4 weeks of age (27), we examined bone cortical and trabecular indices of 3-week-old (precolitic) mice, to avoid colitis-associated effects (56, 63). Comparative μCT imaging revealed that long bones of Runx3 KO mice were short and displayed significantly reduced cortical thickness and underdeveloped trabecular regions compared to those of WT mice (Fig. 1E). Quantitative analysis indicated that Runx3 KO mice had a 19% reduction in femoral length (P < 0.001) (Fig. 1Fa) and a pronounced reduction in femoral width, manifested by a 50% smaller bone cortex area than that of WT littermate mice (P < 0.001) (Fig. 1Fb). Consequently, compared to the WT, Runx3 KO bones displayed 41% and 34% decreases in thickness and total section area, respectively (P < 0.001) (Fig. 1Fc and d). Moreover, cortices of Runx3 KO mice displayed the hallmarks of osteopenia, including 14% and 58% reductions in bone mineral density (BMD) and bone mineral content (BMC), respectively, compared to the WT (P < 0.001) (Fig. 1Fe and f). This osteopenic bone phenotype was also apparent in the cancellous region of Runx3 KO femur, evidenced by significant decreases in trabecular thickness as well as in trabecular BMD and BMC (22%, 28%, and 59%, respectively) compared to the WT (P < 0.001) (Fig. 1Fg to i). Similar bone deficits were found in femora from male Runx3 KO mice or Runx3^fl/fl/Pgk-cre mice (see Fig. S3 in the supplemental material), substantiating the association between the observed osteopenic phenotype and the loss of Runx3.

To address whether the decreases in BMD and BMC affect Runx3 KO bone biomechanical properties, we measured bone rigidity of Runx3 KO and WT long bones (see Fig. S4 in the supplemental material). Applying the 3-point bending method (64) on whole tibiae, we found a 33% reduction in Young's modulus, a size-corrected measure of bone material rigidity, for Runx3 KO bone (70.12 N/m²) compared to that for WT littermate bone (104.63 N/m²). Because BMD correlates linearly with Young's modulus (65), the increased mechanical fragility of Runx3 KO long bones not only confirmed the osteopenic phenotype but also underscored its biomedical impact.

We then determined the temporal onset of the Runx3 KO bone defect by μCT analysis. Humeri of 1-day-old Runx3 KO mice were shorter by 15% and their cortical region had 11% and 38% lower BMD and BMC, respectively, than those of WT littermate mice (Fig. 2A). Qualitative imaging of E18.5 femora revealed a substantial difference in the structural organization of Runx3 KO long bones compared to that of long bones of WT and heterozygous littermates (Fig. 2B). Moreover, histological analysis of long bone mineralization at E14.5, the earliest time point of bone appearance, showed a complete lack of bone material in Runx3 KO embryos, as opposed to bulk calcification in the WT embryo bone collar and primary spongiosa regions (Fig. 2C). However, staining of E16.5 Runx3 KO bone sections showed mineral deposits in these regions (Fig. 2C), indicating that although delayed, osteogenesis in Runx3 KO mice commenced by ∼E16.5. Taken together, the delayed embryonic osteogenesis, the reduced cortical and cancellous BMD and BMC, and the decreased biomechanical strength demonstrate that Runx3 KO mice are congenitally osteopenic. In summary, while Runx3 KO femora appear outwardly normal, albeit significantly shorter than those of their WT counterparts, their cortical width axis and trabecular development are severely impaired. As these bone features are attained by OBL activity (5), these findings are consistent with the notion that loss of Runx3 is associated with osteoblastic dysfunction.

FIG 2.

Temporal onset of Runx3 KO osteopenia and Runx3 expression in developing bone OBLs. (A) Runx3 KO osteopenia is congenital. Comparative μCT analyses of humeri from D1 Runx3 KO and WT littermate mice were performed. Images showing congenital osteopenia of both cortical and cancellous regions in humeri of Runx3 KO versus WT mice (left) (the color bar indicates mineral content) and quantitative μCT structural analysis of these bones (right) are depicted. (B) Altered bone ultrastructure in E18.5 Runx3 KO mice. Shown are μCT images of 50-μm-thick reconstructed midhumeral cross sections. Runx3 KO mice (top) are compared to WT (middle) and Runx3^+/− (bottom) littermate control mice. (C) Delayed ossification in E14.5 Runx3 KO embryos. von Kossa-eosin staining of an E14.5 WT femoral section shows calcification in the bone collar (bc) and primary spongiosa (ps) regions (arrow), versus the complete lack of calcification in bone from an age-matched Runx3 KO mouse (top row) (magnifications, ×40 for images and ×100 for inset). By E16.5, the ossification patterns in WT and Runx3 KO femora were qualitatively indistinguishable (bottom row) (magnifications, ×40 for images and ×200 for insets). Bars, 500 μm (images), 125 μm (top inset), and 100 μm (bottom insets). (D) Runx3 expression in developing bone OBLs. RNA in situ hybridization of an E16.5 humeral section demonstrates intense Runx3 expression in primary spongiosa and bone collar OBLs. At this embryonic stage, Runx3 expression is also detected in prehypertrophic (ph), but not in hypertrophic (h), chondrocytes (c) in the developing growth plate. Runx3 signal intensity in OBLs is as high as that in prehypertrophic chondrocytes. Bar, 200 μm.

Runx3 is expressed in developing bone OBLs.

We then determined Runx3 expression in E16.5 embryonic long bones using ³⁵S-RISH. Osteoblastic Runx3 expression was detected in OBLs of the developing bone collar and primary spongiosa regions (Fig. 2D), as was also found previously by Soung et al. (32). Significantly, the intensity of Runx3 staining in OBLs was similar to that in prehypertrophic chondrocytes, an established Runx3-expressing cell population (24, 30), indicating a relatively high expression level of Runx3 in developing bone OBLs.

Mice bearing OBL-specific Runx3 inactivation recapitulate the osteopenic phenotype.

To further evaluate Runx3 function in bone formation, we crossed Runx3^LoxP/LoxP mice to transgenic Col1α1-cre mice, to generate mice that specifically lack Runx3 in OBLs (Runx3^fl/fl/Col1α1-cre; hereafter named Runx3OBL mice). Runx3OBL mice were born at the expected Mendelian ratio and had a 20 to 50% lower body mass than that of their WT littermates (Fig. 3A). Ossification deficits in Runx3OBL skeletons were similar to those found in Runx3 KO mice, including the short stature that persisted throughout life (Fig. 3B). μCT analyses of femora from 23-day-old Runx3OBL mice showed a 21.7% lower bone thickness, a 25.0% smaller cortical area, and a 27.1% BMC than those of littermate control mice (Fig. 3C to E). Conversely, heterozygote Runx3^fl/+/Col1α1-cre mice had no overt skeletal abnormalities (see Fig. S5A in the supplemental material). Of note, similar morphological findings of severe congenital dwarfism and retarded growth were observed in Runx3^fl/fl/Osx1-cre mice (see Fig. S5B in the supplemental material), functional analogs of Runx3OBL mice. However, due to low survival rates of Runx3^fl/fl/Osx1-cre mice, we used the Runx3OBL line for our extensive bone analyses. Furthermore, the generation of 23-day-old Runx3OBL mice on a Runx2^+/− background phenocopied the observed dwarfism of Runx3OBL mice (see Fig. S5C in the supplemental material), underscoring the unique function of Runx3 in OBL-derived skeletal elongation.

FIG 3.

The Runx3 KO osteopenic phenotype is recapitulated in Runx3OBL but not in Runx3CHN mice. (A) Runx3OBL mice recapitulate the Runx3 KO skeletal phenotype. The decreased body weight and short stature of Runx3OBL mice are indicated. Body weights of D3 mice (left) (n = 6 mice/genotype) and representative gross appearances at D23 (right) are shown. (B) Alizarin red S-alcian blue staining of D21 Runx3OBL and WT mice showing similar ossification patterns with distinct size differences. (Left) Front limb; (middle) hind limb; (right) skull. (C to E) Runx3OBL mice exhibit an osteopenic phenotype. μCT analysis of D23 Runx3OBL (n = 2) and control littermate (WT and Runx3^fl/+/Col1a1-cre) (n = 5) female mice was performed. (C) Bone thickness; (D) cortex area; (E) BMC. Values are means and SD. Asterisks indicate statistical significance (P < 0.01). (F) Runx3CHN mice are seemingly indistinguishable from WT littermate mice. Shown are 3-month-old mice. (G) Similar ossification patterns in Runx3CHN and WT mice. Shown are whole skeleton (left) and hind limbs (right) of alizarin red S-alcian blue-stained D0.5 mice. (H to J) Comparable bone parameters for Runx3CHN and WT littermate mice. μCT analysis of D23 Runx3CHN mice (n = 2) and WT littermate controls (n = 5) was performed. (H) Bone thickness; (I) cortex area; (J) BMC. Values are means and SD, and asterisks indicate statistical significance (P < 0.01).

To address the possible contribution of Runx3^−/− cartilage cells to the observed osteopenic phenotype of Runx3 KO mice, we generated mice bearing chondrocyte-specific Runx3 inactivation, by crossing Runx3^LoxP/LoxP mice to transgenic Col1α2-cre mice (Runx3^fl/fl/Col1α2-cre; hereafter named Runx3CHN mice). Of note, only 2/7 Runx3CHN mice had short stature, and when present, this phenotype was milder than that for Runx3OBL mice (Fig. 3F). Comparative bone staining (Fig. 3G) and μCT analyses (Fig. 3H to J) did not reveal significant differences in bone development or structural parameters between Runx3CHN mice and WT littermate controls.

Runx3 KO mouse osteopenia results from an OBL proliferation defect manifested by fewer active OBLs.

To assess whether the observed osteopenia of Runx3 KO mice was due to a dysfunction and/or a reduced number of OBL cells, we analyzed bone structural parameters by quantitative histomorphometry (53). Runx3 KO mice exhibited a significantly lower bone mass, evident by a 38% decrease in bone volume/total volume (BV/TV), than that of the WT controls (P < 0.01) (Fig. 4A). This bone mass reduction resulted from a diminution in OBL numbers, manifested by 38% and 34% reductions in the surface OBL/bone surface (Ob.S/BS; percentage of bone surface occupied by OBLs) and in the number of OBLs/bone perimeter (N.Ob/B.Pm), respectively (both P = 0.02) (Fig. 4B and C). Notably, no significant changes in the analogous osteoclast (OCL) parameters surface OCL/bone surface (Oc.S/BS) and number of OCLs/bone perimeter (N.Oc/B.Pm) were found (see Fig. S5D in the supplemental material), underscoring the OBL cell-autonomous nature of the observed Runx3 KO bone lesion.

FIG 4.

Comparative histomorphometry and OBL proliferation/apoptosis analyses. (A) Reduced bone volume in Runx3 KO mice. Shown are images of von Kossa-stained vertebral sections of 3-month-old Runx3 KO and WT littermate mice (left) (magnification, ×40; bar, 500 μm) and trabecular bone volume quantification (right) (n = 3 mice/genotype). Values are means and SD, and asterisks indicate statistical significance (P < 0.01). (B to F) Bone formation rates in 3-month-old Runx3 KO and WT littermate mice indicating a significant reduction in the number of active OBLs and a diminished bone formation rate in Runx3 KO mice. (B) Osteoblast surface (Ob.S). BS, bone surface. (C) Number of osteoblasts (N.Ob). B.Pm, bone perimeter. (D) Mineral apposition rate. (E) Mineralizing surface/bone surface. (F) Bone formation rate (n = 3 mice/genotype for panels B and C; n = 4 mice/genotype for panels D to F). Values are means and SD, and asterisks indicate statistical significance (P < 0.01). (G) Impaired OBL proliferation rate in Runx3OBL mice. Comparative BrdU incorporation assays using calvaria sections from 9-day-old Runx3OBL (n = 1) and WT (n = 4) littermate mice show significant reductions in proliferating OBLs in Runx3OBL mice. BrdU incorporation was determined for ≥6 calvaria sections/mouse. Values are means and SD, and asterisks indicate statistical significance (P < 0.01). (H) Comparable OBL apoptosis rates in Runx3OBL and WT littermate mice. OBL apoptosis was determined by a TUNEL assay using calvaria sections from 14-day-old mice (n = 2 mice/genotype; ≥6 calvaria sections per mouse). Values are means and SD.

Dynamic histomorphometric analyses of Runx3 KO bones revealed a 19% decrease in the mineral apposition rate (MAR) (Fig. 4D) as well as profound reductions of the mineralizing surface/bone surface (MS/BS) and the bone formation rate (BFR) of 59% and 68%, respectively, compared to values for the WT controls (P < 0.01) (Fig. 4E and F). These results show that while the loss of Runx3 mildly affects individual OBL activity, as reflected by the MAR, the marked reduction in the number of active OBLs (MS/BS) is the underlying cause for the largely diminished BFR and the resulting osteopenic phenotype.

To discern whether reduced OBL proliferation and/or enhanced apoptosis could explain the reduced number of OBLs in Runx3 KO bones, we conducted BrdU incorporation and TUNEL staining assays using calvarial sections from newborn mice. The loss of Runx3 severely decreased the proliferation capacity of calvarial OBL progenitors, as reflected by a >50% reduction in the number of BrdU-positive cells compared to the WT (Fig. 4G). On the other hand, WT and Runx3 KO calvarial sections had comparable numbers of TUNEL-positive nuclei (Fig. 4H). These results indicate that an impaired proliferation of Runx3 KO OBL progenitor cells and not enhanced apoptosis led to the lower number of active OBLs.

Cultured Runx3-deficient OBLs display impaired proliferation and bone matrix formation.

To directly evaluate whether a lack of Runx3 affects OBL formation and differentiation, we analyzed primary cultures of BM stromal cells. In situ ALP staining revealed similar numbers of ALP-positive osteoprogenitor-derived colonies in 3-week-old Runx3OBL and WT cultures (Fig. 5A), suggesting that the commitment to the osteoblastic lineage is independent of Runx3. However, the size of the ALP-positive colonies in Runx3OBL cultures was 75% smaller than that in WT cultures (Fig. 5A), indicating that their OBL lineage cell generation capacity was impaired. These in vitro results support our above-described histomorphometric measurements indicating that the reduced osteogenic activity in Runx3 KO mice was due to lower numbers of active OBLs.

FIG 5.

Growth rate and mineral deposition analysis of cultured Runx3-null OBLs. (A) Reduced number of active OBLs in Runx3OBL BM stromal cultures. Total BM cells were cultured to confluence for 21 days (21d) and stained with ALP. (Left) Whole-plate and magnified (×200) images. Bar, 200 μm. (Middle) Osteoprogenitor cell numbers were determined by counting ALP-positive colonies (>10 cells) in 10 microscope fields at a ×100 magnification. (Right) The number of OBL lineage cells generated from these osteoprogenitor colonies was deduced by determining the cumulative area of ALP-positive colonies per whole-plate area. Data shown represent data from three independent experiments with similar findings. Values are means and SD, and asterisks indicate statistical significance (P < 0.01). (B) Impaired mineralization capacity of cultured Runx3 KO BM OBLs. Following plating, the mineral deposition capacity of Runx3 KO and WT BM-derived OBL cultures was determined at 7-day intervals. Shown are whole-plate images (von Kossa staining) (left) and digital quantification of mineralized nodules (black) at 14 and 21 days of culture (right). Data shown represent data from two independent experiments with similar findings. (C) Cultured BM OBLs from Runx3OBL mice display a markedly reduced mineral deposition capacity. Shown are whole-plate images of 28-day cultures (von Kossa staining) (left) and digital enumeration (right). Data shown represent data from two independent experiments with similar findings. (D) Calvaria-derived OBL cultures exhibit a diminished mineralization capacity. Mineral staining (von Kossa) of 21-day-old confluent OBL cultures isolated from calvariae of WT and Runx3OBL mice was performed. Shown are images of whole plates (left), a microscopic view of an enlarged plate area (middle) (magnification, ×100; bar, 400 μm), and whole-plate digital quantification of mineral deposition (right). Data shown represent data from two independent experiments with similar findings. Values are means and SD. Note that both Runx3OBL and WT culture images were taken under the same settings. The difference in clarity is due to the fact that in the WT culture plate, the brown mineralized (∼35% of the plate) nodules cast shadows that enhanced the contrast of neighboring cells. This effect does not exist in the KO plate, where mineralization was significantly reduced (∼9% of the plate).

We then assessed OBL function by recording the mineralization capacities of stromal cultures derived from Runx3 KO, Runx3OBL, and WT mice. At 21 days, Runx3 KO BM stromal cultures yielded a 52% smaller mineralized area than that of cultures derived from WT mice (Fig. 5B). Similarly, at 28 days, Runx3OBL BM stromal cultures showed a 59% decrease in the formation of mineralized nodules compared to WT cultures (Fig. 5C). Significantly, an even more striking reduction in bone matrix formation was noted when OBLs from calvariae of newborn Runx3OBL mice were cultured for 21 days (Fig. 5D), yielding almost no mineralized nodules, compared to cultures from newborn WT mice. Collectively, these in vitro experiments demonstrate that the proliferation capacity of Runx3-deficient OBL progenitors is impaired, leading to a substantial decrease in the number of active OBLs, which results in reduced bone matrix formation.

Expression of major OBL markers and transcriptome analysis.

Using RISH, we determined the in vivo expression levels of selected OBL markers in femoral sections of 3-day-old Runx3 KO and WT mice. Compared to WT mice, a substantial decrease in expression levels was noted in bone regions where mRNAs of the early OBL markers Col1α1 and Runx2 and the mature OBL marker Bglap2 (osteocalcin gene) were expressed (37, 62) (Fig. 6A). These reduced expression levels in OBL-occupied areas correspond to our above-described findings of diminished OBL numbers and consequent low bone-forming ability in Runx3 KO mice.

FIG 6.

OBL markers and gene expression analysis. (A) Comparative expression analysis of OBL markers. Femoral sections from D3 Runx3 KO and WT mice were used for ³⁵S-RISH (magnification, ×40; bar, 100 μm). Decreased bone areas where marker genes are expressed in Runx3 KO bone sections reflect the lower numbers of OBLs in these mice. (B) Scatter plot of differentially expressed genes in calvaria-derived OBLs from Runx3OBL versus WT mice. Shown are gene expression levels (log₂ scale) in Runx3OBL versus WT OBLs. Significantly increased and decreased expression levels of genes are indicated in red and green, respectively. Enlarged circles indicate Runx3-responsive genes that are known to participate in osteoblastogenesis. (C) Validation of several differentially expressed OBL genes. RT-qPCR analyses of RNA isolated from WT and Runx3OBL calvaria cultures showed gene expression trends similar to those observed by gene expression chip analysis.

We next sought to gain insight into the transcriptional program involved in precipitating the Runx3-deficient osteopenic phenotype. Consequently, we compared gene expression profiles in primary calvarial OBLs from Runx3OBL and littermate WT mice. Using a >1.5-fold change and an FDR q value of <0.1, relative to the control, as criteria for significance, we identified 351 differentially expressed genes, of which 202 were upregulated and 149 were downregulated, in cultured OBLs from Runx3OBL mice (Fig. 6B; see also Table S1 in the supplemental material). Of note, a pronounced decrease in the levels of marker genes characteristic of active OBLs was noted (see Table S1 in the supplemental material). For example, there was a 5.2-fold reduction in the level of the matrix metallopeptidase 3 gene Mmp3 (66, 67), a 3.2-fold reduction in the level of the tyrosine phosphatase receptor type Z polypeptide 1 gene Ptprz1 (68), and a 3.1-fold reduction in the level of the dentin matrix protein 1 gene Dmp1 (69). In addition, substantial increases in the levels of key components of the Notch signaling pathway were found, including 3.3- and 5.9-fold increases in the levels of the TF genes Hes1 (Hairy enhancer of split 1) and Hey1 (Hes-related with YRPW motif), respectively (Fig. 6B; see also Table S1 in the supplemental material). Of note, transgenic overexpression of Hes1 results in osteopenia due to decreased OBL activity (70), while inhibition of canonical Notch signaling increased osteoblast numbers and BMD (71). The differential expression of several of these osteoblastogenesis signature markers was further validated by RT-qPCR (Fig. 6C). Together, the in situ hybridization results and gene expression data support the notion that the lack of Runx3 impairs OBL generation, leading to the observed osteopenic phenotype.

DISCUSSION

This study describes a novel function of Runx3 in murine osteoblastogenesis. Using mouse genetic models, we show that loss of Runx3 in precommitted OBLs diminishes the formation of active/mature OBLs and consequently leads to severe congenital osteopenia and dwarfism. Bones of Runx3-deficient mice are significantly underdeveloped, and their strength is profoundly compromised. Although the majority of cases of human congenital dwarfism are related to defects in growth plate chondrocytes (72), the skeletal phenotype presented here highlights a somewhat overlooked OBL function in normal bone elongation (73, 74).

The human RUNX3 gene resides on chromosome 1p36, a locus that is associated with low BMD and osteoporosis in humans. Initially recognized by a whole-genome linkage scan analysis (39), this genomic region was independently confirmed to contain a quantitative trait locus for bone-related defects by numerous other studies (40–45). Moreover, similar findings on the linkage of mouse chromosome 4q, syntenic to the human 1p region, to low BMD were also reported previously (75, 76). Ongoing screenings for susceptibility genes that reside at the 1p36 locus identified several candidates, including TNFRSF1B (77), NPPB (78), MTHFR (79), WNT4, and ZBTB40 (45), suggesting that multiple genes, individually or in combination, contribute to the 1p36-associated low-BMD phenotype (80). It is therefore tempting to speculate, based on the findings described in this study, that RUNX3 deficiency might also affect human bone mineralization and thus may constitute a risk factor for low BMD/osteopenia in humans.

Previous studies have shown delayed chondrocyte differentiation/maturation and vascular invasion in E15.5 Runx3 KO bones (32, 33). Probing the later phases of skeletogenesis, we observed transient mineralization delays for the distal limb bones and mandible in E16.5 Runx3 KO embryos as well as for the intramembranous skull bones and clavicles in E18.5 Runx3 KO embryos. The finding that this mineralization deficit recovered in postnatal Runx3 KO mice could be explained by Runx2/Runx3 redundancy in this process. This possibility is supported by the observation that the decline in palm bone ossification is associated with Runx2/Runx3 gene dosage. Furthermore, the specific ablation of Runx3 in chondrocytes (Runx3CHN mice), the main drivers of longitudinal bone growth (34), did not produce a notable defect in bone development or structure. This finding could be explained by the dominance of Runx2 in chondrocyte maturation (33), a notion that corresponds with perinatal lethality in chondrocyte-specific Runx2-deficient mice (14, 15).

Our preliminary observations of delayed ossification in the distal limb bones, clavicles, mandibles, and skull of Runx3 KO and compound Runx3/Runx2 developing skeletons were made using qualitative bone staining. Subsequently, these findings were supported by quantitative μCT and biomechanical analyses of long bones, revealing their severe congenital osteopenia. Although these data suggest that Runx3 is involved in the ossification of both endochondral and intramembranous bones throughout the developing skeleton, our initial finding of delayed palmar/plantar bone ossification in Runx3 KO embryos might point to a more prominent function of Runx3 in the bones of the limb extremities. Interestingly, Runx3 and Runx2 were previously shown to be coexpressed in the developing digits of E14.5 mice, and Runx3 was reported to be intensely expressed in the digits of E14.5 Runx2^−/− mice, implying its Runx2-independent regulation at this site (81). Thus, as opposed to the bone-defining functions of Runx2 (10), whose partial or complete deletion results in cleidocranial dysplasia or complete lack of bone, respectively, the absence of Runx3 seems to cause a less dramatic and yet persistent skeletal phenotype.

By employing immunohistochemistry on limb sections from E15.5 to E16.5 mice, Soung et al. reported the expression of Runx3 in bone collar and primary spongiosa OBLs (32), in agreement with our RISH findings. Several other reports describe osteoblastic Runx3 expression in murine embryonic perichondrium (81) and OBL lineage cells (3), as well as in human OBLs (82). However, to the best of our knowledge, this study is the first to assign a distinct function to Runx3 in these bone-forming cells.

OBL abundance is determined by several factors: the proliferation rate of OBL precursors, their capacity to differentiate toward mature OBLs, and the death rate of mature cells (3). Thus, elements that govern these events determine the active OBL pool size and play a major role in new bone formation. Three independent lines of evidence support the finding that osteoblastic loss of Runx3 significantly reduces the numbers of active OBLs: (i) the marked reduction in MS/BS, determined by ex vivo histomorphometry (with a mildly changed MAR); (ii) the diminished mineralization in long bone- and calvaria-derived cultured OBL progenitors; and (iii) the profound reductions in the levels of OBL maturation marker genes in transcriptome analyses of cultured OBLs. This diminished production of active OBLs stems from a proliferation defect, as evidenced by the reduced BrdU incorporation, rather than from a smaller mesenchymal progenitor pool size in the BM of Runx3-deficient mice. Furthermore, comparative examination of Runx3OBL culture phenotypes implied that while calvaria OBLs were profoundly dependent on Runx3 for their mineralizing function, its requirement by OBLs derived from long bone stromal progenitors was much less stringent. Notably, a similar spatial requirement was reported previously for Runx2 (11). In this case, haploinsufficiency severely affected intramembranous bones, leading to cleidocranial dysplasia in humans and mice, but mildly impacted endochondral bones.

In summary, our findings show that an OBL cell-autonomous function of Runx3 is required for mouse osteoblastogenesis. Runx3 function is required, along with and independent of Runx2, similar to their requirements in chondrogenesis. Thus, it appears that while Runx2 is mandatory for progenitor cell commitment to the osteoblastic lineage (14), Runx3 is nonredundantly required for the proliferation of these precommitted cells, to generate adequate numbers of active OBLs. Hence, in the absence of Runx3, OBL numbers are significantly reduced, while their individual cell function is only marginally affected. This low number of active OBLs impairs proper bone formation and causes severe congenital osteopenia and lifelong short stature in Runx3-deficient mice. Thus, whether RUNX3 deficiency also constitutes a risk factor for low BMD in humans remains to be addressed.