Connexin43 and Runx2 Interact to Affect Cortical Bone Geometry, Skeletal Development, and Osteoblast and Osteoclast Function

Abstract

The coupling of osteoblasts and osteocytes by connexin43 (Cx43) gap junctions permits the sharing of second messengers that coordinate bone cell function and cortical bone acquisition. However, details of how Cx43 converts shared second messengers into signals that converge onto essential osteogenic processes are incomplete. Here, we use in vitro and in vivo methods to show that Cx43 and Runx2 functionally interact to regulate osteoblast gene expression and proliferation, ultimately affecting cortical bone properties. Using compound hemizygous mice for the Gja1 (Cx43) and Runx2 genes, we observed a skeletal phenotype not visible in wild-type or singly hemizygous animals. Cortical bone analysis by microCT revealed that 8-week-old male, compound Gja1^+/− Runx2^+/− mice have a marked increase in cross-sectional area, endosteal and periosteal bone perimeter, and an increase in porosity compared to controls. These compound Gja1^+/− Runx2^+/− mice closely approximate the cortical bone phenotypes seen in osteoblast-specific Gja1-conditional knockout models. Furthermore, microCT analysis of skulls revealed an altered interparietal bone geometry in compound hemizygotes. Consistent with this finding, Alizarin red/ Alcian blue staining of 2 day-old Gja1^+/− Runx2^+/− neonates showed a hypomorphic interparietal bone, an exacerbation of the open fontanelles, and a further reduction in the hypoplastic clavicles compared to Runx2^+/− neonates. Expression of osteoblast genes, including osteocalcin, osterix, periostin, and Hsp47, was markedly reduced in tibial RNA extracts from compound hemizygous mice, and osteoblasts from compound hemizygous mice exhibited increased proliferative capacity. Further, the reduced osteocalcin expression and hyperproliferative nature of osteoblasts from Cx43 deficient mice was rescued by Runx2 expression. In summary, these findings provide evidence that Cx43 and Runx2 functionally intersect in vivo to regulate cortical bone properties and affect osteoblast differentiation and proliferation, and likely contributes to aspects of the skeletal phenotype of Cx43 conditional knockout mice.

INTRODUCTION

Gap junctions are specialized transmembrane channels that permit cell-to-cell communication via the bi-directional exchange of small molecules, ions, and second messengers between the cytosols of adjacent cells. This sharing of signals is of particular importance in bone wherein the action of osteoblasts, osteocytes, and osteoclasts must be tightly coordinated in order to ensure normal bone development and homeostasis [1]. Indeed, the formation of new bone requires the synchronized action of hundreds of gap junction-coupled osteoblasts working together to facilitate the deposition and mineralization of an organized collagen-rich matrix. Likewise, osteocytes embedded in bone relay hormonal-, growth factor-, and mechanical load-derived signals that regulate both osteoblast and osteoclast activation to the cells at the surface of bone via gap junction-dependent mechanisms [2]. This resulting syncytium of interconnected cells ultimately serves to coordinate the rapid communication of structural and functional changes that enable bone tissue to adapt in response to environmental (contextual) cues. However, molecular details of how gap junctions regulate bone development and homeostasis are only partially defined.

Connexin43 (Cx43) is the predominantly expressed connexin protein in bone, and osteoblasts and osteocytes are extensively coupled by gap junctions composed of Cx43 monomers [3]. Cx43 acts through both direct gap junctional communication and by unopposed Cx43 hemichannels to regulate osteoblast differentiation, bone mass accrual, and bone remodeling [4, 5]. Mutations in the Cx43 gene GJA1 have been linked to the clinical hereditary disorders oculodentodigital dysplasia (ODDD) and craniometaphyseal dysplasia, both of which manifest with distinct skeletal abnormalities [6, 7]. Germline or osteoblast-lineage specific conditional deletion of Cx43 results in osteopenia, alterations in cortical bone geometry, poor bone quality and an attendant osteoblast differentiation defect (reviewed in Stains et al [1] and Buo et al [8]). Numerous studies have also detailed the involvement of Cx43-comprised gap junctions and hemichannels in various bone physiologic processes, including mechanotransduction [9–11], sensitivity of bone to hormonal and growth factor cues [12, 13], osteocytes viability [14, 15], and control of the balance of bone resorption and bone formation [15, 16]. While the impact of Cx43 on bone is well documented, the molecular details underlying how Cx43 is able to regulate these diverse skeletal processes are largely lacking. Questions remain regarding the downstream factors that are regulated by Cx43 to induce gene expression changes that ultimately impact bone quality.

Several studies have shown that Cx43 deletion in cells of the osteoblast lineage affects Runx2 target genes, including bone sialoprotein, osteopontin, osteocalcin and osterix ([12, 16–19]. Accordingly, our lab has examined the role of Runx2 as a downstream target of Cx43 signaling. Runx2 is a master transcriptional regulator of osteogenesis and is required for osteoblast differentiation and the development of mineralized bone [20]. Work from others has shown that the induction of Runx2 transactivation by fibroblast growth factor-2 (FGF-2) is mediated by ERK and PKCδ signaling [21–23]. We have shown that Cx43 enhances Runx2 transcriptional activity in MC3T3 cells by enhancing ERK and PKCδ signaling, and that treating Cx43-overexpressing cells with FGF-2 synergistically enhanced Runx2 transcriptional activity [13, 24, 25]. Further, we determined that these effects required gap junctional channel function as well as an intact Cx43 C-terminal domain, by which signaling via ERK and PKCδ is efficiently activated by the communication of second messengers [26, 27]. From these in vitro findings, we concluded that Cx43 modulates Runx2 activity, impacting osteoblast differentiation by sharing signals that amplify the activation of effector molecules like ERK and PKCδ. While these findings strongly suggest that a portion of Cx43’s effect on bone is attributable to the coordinated and amplified induction of Runx2 activity among osteoblast lineage cells, evidence of a Cx43-Runx2 intersection of function is lacking in vivo. Confirmation of this intersection of function between Cx43 and Runx2 in vivo would provide convincing evidence for the role of Runx2 as a downstream target and mediator of the Cx43-dependent effects on bone.

In this study, we investigate whether Cx43-dependent effects on bone quality converge on the activity of Runx2 in vivo by generating mice that are compound hemizygous for the Cx43 gene, Gja1, and Runx2 and assessing the severity of the skeletal phenotype of these animals in comparison to their wildtype and singly hemizygous controls. We demonstrate that these compound animals have phenotypic features that are not seen in the single hemizygous controls and that overlap those seen in osteoblast-specific Cx43 conditional knockout models. We provide underlying histological and gene expression data. Furthermore, we provide for a cell autonomous defect in proliferation in these animals that potentially underscores the importance of the functional intersection of Cx43-Runx2 in vivo.

MATERIALS AND METHODS

Animals

All animal studies and procedures were performed with approval by the Animal Care and Use Committee at the University of Maryland School of Medicine. The Gja1flox/flox (Gja1-floxed) mice (C57BL/6J background) were from the Jackson Labs, and a breeding colony maintained at the University of Maryland School of Medicine. The Gja1^+/− mice and the Runx2^+/− mice were generously provided by Roberto Civitelli (Washington University, St Louis) and Jane Lian (University of Vermont), respectively, and described previously [28, 29]. The compound Gja1^+/− Runx2^+/− mice were generated by mating Gja1^+/− animals with and Runx2^+/− animals to produce dually hemizygous offspring as well as their wild-type and singly hemizygous littermates (WT, Gja1^+/− and Runx2^+/−). For genotyping, pups were tailsnipped at < 3 weeks of age, and genomic DNA extracted using the Extracta DNA Prep kit (Quanta). Genotyping PCR was subsequently performed using the Accustart II Mouse Genotyping kit (Quanta), using specific primers to determine the Gja1 [30] and Runx2 zygosity [29]. All four of the expected genotypes appeared at expected Mendelian ratios. Mice were group housed in micro-isolator cages, and food (standard rodent chow) and water were available ad libitum.

Cell cultures

The hTERT-immortalized wildtype mouse calvaria osteoblast cell line (MOB) and the Cx43-null variant (43KO-MOB) were provided by Mia Thi [31]. Primary bone marrow stromal cells (BMSCs) were isolated from mouse hind-limb long bones using methods previously described [17, 32, 33]. BMSCs were cultured in Dulbecco’s Modified Eagle Medium (DMEM), 10% FBS, 1% pen/strep, 50 μM gentamicin, supplemented with 50 μM ascorbic acid 2-phosphate. Primary osteoblasts were isolated by dissecting out the calvaria from euthanized neonates (2–3 days old), removing the sutures, and digesting with 1 mg/mL collagenase A in serum-free α-MEM for 30-minute intervals. The first fraction was discarded. Fractions 2 & 3 were pooled together and cultured in Minimal Essential Medium Eagle, Alpha modification (α-MEM) containing 10% FBS (Gemini), 1% penicillin/streptomycin (Corning), and 50 μM gentamicin (Gibco). To induce mineralization, cells were treated with α-MEM culture medium supplemented with 50 μM ascorbic acid 2-phosphate and 7.5 mM glycerol 2-phosphate as described [34].

Protein Extraction and Western Blotting

Cells were lysed using a modified RIPA buffer (50 mM Tris (pH 8.0), 150 mM NaCl, 1.0% NP-40, 0.5% Sodium deoxycholate, 0.1% SDS, 10mM Na₄P₂O₇, 10 mM β-glycerolphosphate, 10mM NaF, 10mM EDTA, 1mM EGTA) with 1X HALT phosphatase/protease inhibitor cocktail (Thermo) added prior to lysis as described [35]. The following antibodies were used: Sigma, anti-Cx43 (C6219); Cell Signaling Technology, total ERK (9102L), phospho-ERK (9101L), total PKCδ (2058S), phospho-PKCδ Thr505 (9374S), PCNA (2586S), anti-Rabbit HRP (7074S), anti mouse HRP (7076S); Millipore, GAPDH (MAB374) and anti-osterix (AB3743). Representative blots are shown and were repeated a minimum of two times.

Specimen acquisition and X-ray

Offspring from the Gja1^+/− x Runx2^+/− crosses were euthanized at 8 weeks of age. Body weight and length (nose to tail) of littermates were recorded. Gross skeletal morphology was assessed using a Faxitron digital x-ray system to obtain lateral and AP radiographs (35 kV, 10 sec), as described [34]. Femurs and skulls were dissected from male and female offspring, fixed in 4% PBS-buffered paraformaldehyde (PFA) for up to five days, and transferred to 70% ethanol. Tibiae (epiphyses removed and flushed of marrow) were dissected from each animal and placed in Tripure reagent for RNA isolation or in RIPA buffer for protein.

Microcomputed Tomography (microCT)

Three-dimensional microCT analysis was performed on fixed femurs and skulls from each of the four genotypes using a Bruker SkyScan 1172 system as described [34]. Femurs were scanned at high resolution (2000 × 1666 pixels) with a voxel size of 10 μm and at 60 kV, 167 μA (0.5 mm Al filter). Microstructural properties of femoral cortical and trabecular bone were assessed at the distal metaphysis for trabecular parameters and the mid-diaphysis for cortical parameters, following the nomenclature guidelines outlined by Bouxsein and colleagues [36]. Trabecular bone was delineated via manual tracing and interpolation of trabeculae in a region of interest 0.2 mm to 2.0 mm proximal to the distal femoral growth plate, as described [34]. For cortical bone parameters, analysis was performed at a volume of interest beginning at the mid-diaphysis of the femur and extending 0.6 mm distally. Skulls were scanned at low resolution (1000 × 666 pixels), with a fixed voxel size of 20 μm at 90 kV and 112 μA. Reconstructed skulls were subjected to volumetric 3D modeling using CTVol software (Bruker).

Bone Histomorphometry and Histology

For static histomorphometric analysis, fixed femurs were decalcified in 14% EDTA (pH 8.0), embedded in paraffin, and cut in 5-μm serial longitudinal sections. Tartrate-resistant acid phosphatase staining was performed using an Acid Phosphatase, Leukocyte (TRAP) kit (Sigma-Aldrich) as done previously [34]. Slides were imaged using a microscope (Eclipse 50i; Nikon) outfitted with a 20x objective, a camera (DS-Fi2, Nikon) and imaging software (NIS Elements; Nikon). Images were analyzed with Bioquant Osteo 2016 histology software to quantify osteoblasts and osteoclasts in both the endocortical and trabecular compartments, as indicated, using methods described previously [34]. Collagen fiber orientation was assessed by fixing paraffin sections in 4% PFA, staining with a 0.1% (w/v) picrosirius red solution, and destaining with 0.5% (w/v) acetic acid solution. Sections were visualized and imaged using a polarizing light microscopy filter. Quantification of fiber thickness and orientation was completed using a hue range method [37].

Immunohistochemistry

In order to assess expression of proliferating cell nuclear antigen (PCNA) in vivo, decalcified, paraffin-embedded femur sections were subjected to immunohistochemical analysis, as previously described [38]. Briefly, slides were deparrafinized in xylenes, rehydrated in sequential alcohol baths, and incubated in 10 mM sodium citrate buffer for 10 minutes at 37°C and 30 minutes at room temperature to unmask antigens. Sections were blocked with SuperBlock (Thermo Scientific) for 1 hour and incubated overnight at 4°C in rabbit anti-PCNA antibody (1:1000) in SignalStain Ab Diluent (Cell Signaling Technology). The following day, slides were incubated with biotinylated secondary antibody and the Vectastain ABC reagent (Vector Laboratories) according to manufacturer’s instructions. Slides were stained with Vector NovaRED substrate (Vector Laboratories) and counterstained with Fast Green (Electron Microscopy Sciences). Slides were rapidly dehydrated through alcohols and xylenes prior to mounting with Permount and coverslips.

Whole-mount skeletal preparations

Newborn pups (2 days of age) were subjected to whole-mount staining with Alcian blue/Alizarin red following a procedure outlined by Wallin et al. [39] with minor modifications. Specimens were stored in 80% glycerol and imaged using a high-resolution flat-bed scanner. The length of the mineralized clavicle was measured using a calibrated image in ImageJ. To quantify the fontanelle area, the area of the fontanelle was traced using the freeform selection tool in ImageJ and the area of the selection was automatically computed based on the calibration.

Adenoviral constructs and amplification

The eGFP adenovirus (Ad-GFP) and the GFP-tagged Cre recombinase adenovirus (Ad-Cre), were purchased from Vector Biolabs at a viral titer of 1×10¹⁰ plaque-forming units/ml (PFU/ml). The Runx2-expressing adenovirus (Ad-Runx2) was created using the Gateway Cloning Technology system (Invitrogen). The adenoviral construct was made by amplifying the murine Runx2 gene (MASN-variant) with PCR primers that included attP recombinase sites, and then cloning the resulting PCR product into the donor vector pDONR/Zeo (Invitrogen). Upon verification of the presence of Runx2 by sequencing, the Runx2 insert was then swapped from the donor plasmid into the adenoviral vector pAd/CMV/V5-DEST to create the destination clone pAd-Runx2. Expression and transcriptional activity of the Runx2 expression construct was validated before use (Supplemental Fig S1). To generate the virus, Pac I-digested pAd-Runx2 was submitted to the University of Maryland Recombinant Virus Core for transfection into 293a packaging cells for amplification and titered via agarose plaque assays as described [17]. Both Ad-GFP and Ad-Cre had a viral titer of 1.1×10⁹ pfu/mL. Ad-Runx2 had a viral titer of 1.1×10¹² pfu/mL.

Adenoviral Transduction

Adenoviral transduction experiments were performed on primary calvarial osteoblasts or BMSC cultures isolated from Gja1-floxed animals using a slightly modified procedure reported previously [17]. Briefly, prior to the addition of virus, a 0.5 mg/mL concentration of cell culture-grade poly-L-lysine (Sigma) was added to the serum-free MEMα at a 1:1000 dilution (0.5 μg/mL final concentration) in order to improve transduction efficiency [17]. Amplified virus was then added to individual tubes using a volume required to obtain an MOI of 60 for AD-GFP and AD-Cre, and an MOI of 600 for AD-Runx2.

For Runx2 rescue experiments, four p100 dishes containing sub-confluent calvarial osteoblasts were prepared, with 2 dishes being transduced with AD-GFP and the other 2 being transduced with AD-Cre to delete Cx43. The next day, transduction efficiency for Cx43 deletion was confirmed by fluorescence microscopy for green fluorescence of the Cre adenoviral construct, and each pair of AD-GFP- and AD-Cre-treated dishes was subsequently transduced with AD-GFP or AD-Runx2 to create 4 expression groups of cells (GFP/GFP, Cre/GFP, GFP/Runx2, and Cre/Runx2). The day after, cells were detached with Accutase, counted and seeded appropriately into multi-well plates for subsequent experiments.

Quantitative PCR

Total RNA was isolated from tibia (flushed of marrow) using Tripure reagent (Roche), as described [34]. Reverse transcription-quantitative PCR was carried out as described [35]. The data are simultaneously normalized to the expression of Gapdh, Rpl13, and Hprt using geNorm v3.5 software (Ghent University Hospital Ghent, Belgium), as described [25]. The primer sets for Gapdh, Rpl13, Hprt, Gja1/Cx43, Runx2, Sp7/osterix Bglap/osteocalcin, Ibsp/bone sialoprotein, Alpl/alkaline phosphatase, Postn/periostin, and Serpinh1/heat shock protein 47 are all described previously [25, 27]. Experiments were performed in triplicate wells and repeated at least three times.

Cell Proliferation

Cell proliferation was assessed with the Cell Counting Kit-8 (CCK-8) colorimetric assay according to the manufacturer’s instructions (Dojindo). Experiments were performed in triplicate wells and repeated at least three times.

Statistical Analysis

Data are presented as mean ± SD, unless indicated otherwise. For comparisons between the four different genotypes, a two-way ANOVA was performed to assess data for statistical significance followed by a Fisher’s LSD post-hoc test using GraphPad Prism (v6). A t-tests was performed to compare two groups. To validate whether the outcomes are indicative of an interaction between Cx43 and Runx2, we indicate the statistically significant interaction of the variables (‡, p-value < 0.05 unless indicated otherwise) as determined by the two-way ANOVA. For histology and microCT analysis the reviewer was blinded to the genotype of the mice.

RESULTS

Cx43 modulates signaling pathways converging on Runx2 in osteogenic cells

Given our prior work in MC3T3 cells, which implicated PKCδ and ERK in the modulation of Runx2 activity downstream of Cx43 [13, 24], we confirmed the importance of Cx43 as a regulator of signaling in osteoblasts, using a Cx43-deficient hTERT-immortalized primary mouse osteoblast model generated by Mia Thi and colleagues [31] to compare signaling cascades between wild type (MOB) and Cx43-deficient (43KO-MOB) cells. Western blotting of whole-cell extracts showed that the 43KO-MOB cells have reduced phosphorylation of ERK and PKCδ in comparison to the wild type MOB cells (Fig. 1A). These findings are consistent with our published gain and loss of function data for Cx43 that ERK and PKCδ activity are affected by Cx43 status [13, 24, 25, 27]. Next, we examined the protein levels of osterix, a transcription factor obligated for osteoblast differentiation and whose expression is downstream of Runx2 [40]. As predicted by published studies showing decreased Sp7/osterix mRNA following loss of Cx43 [16, 17, 31], osterix protein was reduced in 43KO-MOB cells (Fig. 1A). These results were confirmed in primary BMSCs isolated from Gja1-floxed mice and transduced them with either AD-GFP or AD-Cre, to generate a population of Cx43-expressing cells and a population Cx43-deficient cells. In these cells, deletion of Cx43 also reduced phosphorylation of ERK and PKCδ and reduced osterix protein levels (Fig. 1A). Consistent with our work in MC3T3 cells [13, 24], loss of Cx43 in 43KO-MOB cells also resulted in a blunted induction of ERK and PKCδ phosphorylation following FGF2 treatment, indicating a reduced sensitivity to growth factor signaling (Fig. 1B).

Figure 1. Overexpression of Runx2 in Cx43-null primary osteogenic cells, which exhibit attenuated signaling, is able to restore osteocalcin gene expression.

(A) Western blots probing for phospho-ERK, phospho-PKCδ and osterix on protein extracts from hTert-immortalized primary mouse osteoblasts (Cx43MOB or 43KO-MOB) and Gja1-floxed primary bone marrow stromal cells (BMSCs) treated with AD-GFP or AD-Cre. Cx43 was probed to verify status of Cx43 expression. Densitometric quantitation of the western blots is shown. (B) Western blots on nuclear extracts from Cx43MOB and 43KO-MOB cells treated with Fibroblast growth factor-2 (FGF-2) for either 0 (no FGF2), 5, or 30 minutes, showing the Cx43-dependent effects on FGF2-induced ERK and PKCδ phosphorylation. (C) RT-qPCR from RNA isolated from Gja1-floxed primary osteoblasts treated with AD-GFP or AD-Cre showing Cx43 deletion reduces osteocalcin gene expression. (D) RT-qPCR showing adenoviral overexpression of Runx2 (AD-Runx2) in these Cx43-deleted cells rescues osteocalcin gene expression. * = p<0.05, ** = p<0.01.

Next, the effects of Cx43 deletion on the expression of osteocalcin, a downstream target of both Cx43 and Runx2 [16, 17, 41–43] was examined in primary osteoblasts isolated from Gja1-floxed mice. Consistent with published work, transduction of these cells with Cre recombinase-expressing adenovirus resulted in a more than two-fold reduction of osteocalcin gene expression (Fig. 1C). To verify that this effect was due to the Cx43-dependent regulation of Runx2, we ablated Cx43 in primary Gja1-floxed osteoblasts and subsequently transduced the cells with a Runx2-expressing adenovirus (Ad-Runx2) or a GFP-expressing control (Ad-GFP) to determine if overexpression of Runx2 could rescue osteocalcin gene expression in these cells. We observed a Cx43-dependent downregulation of osteocalcin in Cre-treated control cells, an effect that could be fully rescued when Runx2 was overexpressed (Fig. 1D). These data confirm that Cx43 influences signaling pathways in osteoblasts that affect Runx2 activity.

Cortical and trabecular bone phenotypes are manifested in the long bones of male mice with compound Gja1 and Runx2 hemizygosity

To determine if the functional intersection of Cx43 and Runx2 activity in vitro also extends in vivo to affect bone, we implemented a breeding strategy that involved crossing Gja1^+/− mice with Runx2^+/− mice in order to produce offspring that are hemizygous for both Gja1 and Runx2. Similar approaches have been used to detect interactions of genes controlling bone remodeling [44, 45]. These compound Gja1^+/− Runx2^+/− animals (referred to as Cmpd) are viable and thus provide a way to assess the effects of Cx43 and Runx2 deficiency on bone while avoiding the perinatal lethality caused by homozygous null mutations of either gene individually [20, 28]. If Cx43 and Runx2 do functionally intersect, then we hypothesized that there would be a unique (or more than additive) skeletal phenotype in the skeletal properties of the Cmpd animals not seen in singly hemizygous (Gja1^+/− or Runx2^+/−) or wildtype (WT) littermate controls. Unlike the homozygous null mutations of Gja1 and Runx2, both Gja1^+/− and Runx2^+/− mice are viable and fertile, and several aspects of their skeletal phenotype have been characterized. Briefly, Gja1^+/− animals do not display any overt skeletal phenotype in comparison to WT controls [46], whereas Runx2^+/− animals present with a cleidocranial dysplasia-like phenotype of the skulls and hypoplastic clavicles [47, 48].

At 8 weeks of age, the Cmpd mice were overtly indistinguishable from littermate WT, Gja1^+/− and Runx2^+/− mice. No significant differences in body weight or body length were observed between the four genotypes, and gross skeletal morphology, with the exceptions noted in the skull and clavicle as expanded upon below, appeared radiographically similar between groups (Supplemental Fig. S2). MicroCT analysis of femurs from 8-week-old males revealed several differences in bone microarchitecture unique to the Cmpd animals (Fig. 2A–B). In the mid-diaphyseal cortical compartment, Cmpd animals possessed statistically significant increases in total cross-sectional tissue area, marrow area, and cortical porosity in comparison to each of the other genotypes. (Fig. 2C). These geometric changes also accounted for a marked increase in the calculated mean polar moment of inertia (pMMI) in Cmpd mice (Fig. 2C). These changes showed a statistically significant interaction effect for hemizygosity of the two alleles supporting a functional interaction between Gja1 and Runx2 in the cortical bone phenotype. This phenotype is strikingly similar to the characteristic changes in cortical bone geometry previously reported in osteoblast and/or osteocyte-specific Cx43 conditional knockout models [12, 14, 16, 49].

Figure 2. MicroCT scans of femurs from 8-week male mice show that compound Gja1^+/−Runx2^+/− hemizygosity affects both cortical and trabecular bone microstructure.

(A) Transverse (top) and axial (bottom) cross sections of reconstructed femurs indicating anatomical locations for cortical and trabecular analyses. (B) 3D reconstructions showing trabeculae for each genotype. (C) Quantification of cortical parameters at the femoral mid-diaphysis; T.Ar = cross-sectional tissue area, Co.Po = cortical porosity, Ma.Ar = marrow area, pMMI = polar moment of inertia, Cs.Th = cross sectional thickness, BV = total bone volume. (D) Quantification of trabecular phenotype at the distal femoral metaphysis; Tb.N = trabecular number, Tb.Th = trabecular thickness, BV/TV = % bone volume, Tb.Sp = trabecular separation. Number of mice per genotype is at least 5. Graphs depict mean ± SD. * = p<0.05 when compared to WT and Gja1. “‡” signifies a statistical interaction effect between Gja1 and Runx2 at p < 0.05.

Trabecular bone analysis revealed that there were no significant additive changes in trabecular parameters that were attributable to interaction of the two hemizygous alleles in the Cmpd mice (Fig. 2D). Rather, changes in trabecular bone quality were dominated by Runx2 hemizygosity, which included significant decreases in trabecular bone volume fraction (BV/TV) and trabecular number, and a significant increase in trabecular separation when compared to WT and Gja1^+/− specimens. Gja1 hemizygosity did not exacerbate the changes caused by Runx2 hemizygosity in Cmpd mice; rather Gja1 hemizygosity surprisingly rescued the trabecular phenotype of the Runx2^+/− mice to control levels (Fig. 2D). Considering that a lack of a trabecular bone phenotype has also been reported in most of the osteoblast-lineage-specific Cx43 conditional knockout models [12, 16], these findings are consistent with a predominant role for Cx43 in cortical bone but also hint at an underappreciated and complex role for Cx43 in the trabecular compartment.

While female mice had trends towards a statistical interaction of Gja1 and Runx2 on the cortical phenotype for cross–sectional area and marrow area, these did not reach statistical significance (Supplemental Fig. S3). Unlike male mice, cortical porosity and pMMI were unaffected in Cmpd female mice. Likewise, the effects of Gja1 and Runx2 haploinsufficiency or compound heterozygosity on trabecular bone were blunted in female mice relative to the male mice (Supplemental Fig. S3).

In total, these findings demonstrate that Gja1^+/− Runx2^+/− compound hemizygous mice possess a unique cortical bone phenotype, not observed in the singly hemizygous controls, and provide evidence that the regulation of bone structure by Cx43 may be due to a functional convergence on Runx2 activity.

Altered osteogenesis, osteoblast function and osteoblast number in situ

To begin to understand the underlying mechanisms contributing to the phenotypic observations in the Cmpd mice, gene expression analysis was performed. Reverse transcription-quantitative PCR analysis of RNA isolated from 8-week-old dissected male tibias showed that the expression of several osteoblast genes was reduced in the Cmpd animals compared to controls, including osterix, alkaline phosphatase, periostin and SerpinH1, which encodes the collagen-binding chaperone heat shock protein 47 (HSP47) (Fig. 3A). Based on previous findings of altered collagen processing in Cx43 conditional knockout models [16, 50, 51], we examined collagen fibril orientation via polarizing light microscopy of picrosirius red-stained of longitudinal femoral sections from these 8 month-old mice. Although collagen orientation and thickness at the mid-diaphyseal cortical bone was disrupted in our Cmpd samples relative to WT mice, this effect was not different than what was observed in Gja1^+/− hemizygous mice (Supplemental Fig. S4). These data confirm a role for Cx43 in collagen processing and organization, but suggest that these effects are not exclusively downstream of any effects Cx43 may have on Runx2.

Figure 3. Compound mice display defective osteogenesis and osteoblast function.

(A) RNA was isolated from tibial extracts and RT-qPCR was performed to quantitate gene expression of several osteoblast markers, including: osterix, collagen 1a1, alkaline phosphatase, bone sialoprotein, periostin, and Hsp47. (B) Histomorphometric quantitation of trabecular osteoblasts from Goldner’s Trichrome-stained femur sections (N.Ob/BS). Graphs depict mean ± SD. n = 3/genotype. * = p<0.05 when compared to wildtype (WT).

Histomorphometric analysis of Goldner Trichrome-stained femoral sections showed a statistically significant interaction of the Gja1 and Runx2 alleles resulting in a reduction in mature osteoblast number in the Cmpd sections (Fig. 3B). When taken together, these findings allude to a disruption of osteoblastogenesis caused by Gja1 and Runx2 compound hemizygosity.

Increased endocortical TRAP-positive osteoclasts in compound animals

The characteristic expansion of the marrow cavity observed in many osteoblast-specific Cx43 knockout models is due to an indirect increase in osteoclast-mediated bone resorption, particularly along the endocortical surface [16, 52]. Hence, we were curious whether increased osteoclast recruitment was also responsible for the apparent phenotype in our Cmpd animals, which would be indicative of a convergence of Cx43 and Runx2 function in modulating osteoclastogenesis. The number of TRAP-positive osteoclasts in the trabecular compartment of femoral sections was not uniquely altered in Cmpd mice relative to the other genotypes, although trabecular osteoclast numbers were reduced in Gja1^+/−, Runx2^+/− and Cmpd mice (Fig. 4A). In contrast, osteoclast numbers along the endocortical surface of the femur were increased nearly 3-fold in Cmpd specimens, which were often displayed as linear aggregates of TRAP-positive osteoclasts along the cortical bone surface (Fig. 4B–C). Mechanistically, this increase in osteoclast-mediated endocortical resorption in Cx43 conditional knockout models has been reported to be due to a cell autonomous dysregulation of the ratio of RANKL to OPG mRNA expression in osteogenic lineage cells [14–16, 52, 53]. In support of those findings, BMSCs from Cmpd mice showed an increased RANKL/OPG mRNA ratio in comparison to the other genotypes (Fig. 4D). Interestingly, no statistical increase in endocortical osteoclasts was detected in female mice (data not shown), which is consistent with their lack of a cortical phenotype by microCT (Supplemental Fig. S3).

Figure 4. Endocortical, but not trabecular, osteoclast number is increased in compound animals.

(A–B) Static histomorphometry of osteoclast number per bone surface from the trabecular and endocortical regions (n=3/genotype). (C) TRAP-stained images at the endocortical surface representative for each genotype. (D) RT-qPCR of RNA isolated from primary BMSCs was performed to show the ratio of RANKL:OPG mRNA expression for the corresponding genotype (n=3/genotype). * = p<0.05 when compared to every other genotype. # = p<0.05 when compared to wildtype (Gja1^+/+ Runx2^+/+). “‡” signifies a statistical interaction effect at indicated p value.

Compound animals display developmental defects in skull bone morphology and reduced clavicle bone ossification

Next, we evaluated the skulls of our mice for additional evidence of the intersecting role of Cx43 and Runx2 on skeletal biology. Although Gja1 hemizygosity does not result in a cranial phenotype (Fig. 5A), studies have reported hypomineralized skulls in ODDD mutant mice and in osteoblast-specific Cx43 conditional knockout models [6, 16, 54]. Additionally, Runx2 deficiency results in several hallmark cranial defects, including hypomineralization, morphological defects, and incomplete closure of sutures/open fontanelles [20, 47, 48].

Figure 5. Compound animals have smaller interparietal bones, wider fontanelles at birth, and hypomineralized clavicles.

(A) MicroCT scans of skulls from 8-week old males from each genotype highlighting interparietal bone and fontanelles (black arrow). Skull bones are labeled as follows: F = frontal, P = parietal, I = interparietal, O = occipital. (B–C) Quantification of fontanelle and interparietal bone area for Runx2^+/− and Cmpd skulls. (D) Skulls from Alizarin red/Alcian blue wholemount skeletal preparations from 2-day old (P2) pups. (E–F) Quantification of the area of fontanelle and interparietal bone area for each genotype. (G) Excised clavicles from P2 wholemounts. (H) Measurement of clavicle length. Graphs depict mean ± SD. *= p<0.05 when compared to every other genotype. “‡” signifies a statistical interaction effect at the indicated p value.

MicroCT reconstructions of skulls from 8-week male mice showed that both Runx2^+/− and Cmpd animals displayed open fontanelles due to Runx2 haploinsufficiency, but the size of the fontanelles was not further increased in the Cmpd animals (Fig. 5A–B). Intriguingly, Cmpd animals frequently displayed a noticeable reduction in the size of the interparietal bone in comparison to WT, Gja1^+/− and particularly Runx2^+/− mice, thus providing potential evidence of intersecting function of Cx43 and Runx2 in cranial morphology (Fig. 5A, C).

To determine whether the cranial features observed in adult mice are more pronounced during development, we performed whole-mount analysis on two-day old (P2) pups stained with Alizarin red/Alcian blue. We consistently observed a conspicuous, hypomorphic interparietal bone in skulls from Cmpd animals when compared to skulls from the littermate controls (Fig. 5D, E). Furthermore, both Runx2^+/− and Cmpd mice at P2 displayed prominent open fontanelles as anticipated, but unlike the adult mice, the size of the fontanelles was significantly larger in Cmpd pups compared to the WT, Runx2^+/− and Gja1^+/− mice (Fig. 5D, F). Additionally, insufficient Runx2 function characteristically manifests in hypoplasia of clavicles in mice [20]. We observed that Cmpd mice had significantly hypomineralized clavicles when compared to littermate controls (Fig. 5G–H). Taken together, these data provide compelling evidence for the functional intersection of Cx43 and Runx2 in regulating cranial development and ossification.

Primary cells isolated from Compound animals exhibit enhanced proliferative capacity, a defect which is rescuable by Runx2 overexpression*

Growing evidence has suggested that Cx43 deletion can result in an expansion of the pre-osteoblast population and that this effect might lead to the precocious periosteal mineral apposition observed in Cx43 osteoblast-lineage specific conditional knockout models, despite impaired osteoblast differentiation [16]. Both Cx43 and Runx2 can impact cell cycle progression and regulate cell proliferation [55–57]. Quantification of primary osteoblast proliferation revealed a nearly three-fold increase in proliferation in cells from the Cmpd mice, an effect that was enhanced beyond the effects of Runx2^+/− hemizygosity (Fig. 6A). We verified that this increase in primary cell proliferation was also reflected in vivo by examining tibial protein extracts and histological sections from these animals for cell cycle progression marker PCNA, which is elevated during the S phase. Notably, Cx43 and Runx2 have been reported to impede cell cycle progression at the G1/S phase transition [55, 56]. While PCNA protein abundance increased in tibial extracts from Gja1^+/− and Runx2^+/− mice, it was highest in Cmpd mice (Fig. 6B). This is consistent with increased labeling of PCNA-positive cells along the periosteal surface in femoral sections from Cmpd animals (Fig. 6C). These data suggest that the periosteal expansion observed in the Cmpd animals may be due in part to a hyperproliferative phenotype of osteoblast lineage cells.

Figure 6. Cx43 and Runx2 functionally intersect to regulate osteoblast proliferation.

(A) CCK-8 proliferation assays on primary osteoblasts from mice of each genotype after 72 hours in culture (n=3 wells/genotype). (B) Western blots performed on diaphyseal tibial extracts (flushed of marrow) of mice of the indicated genotypes and probed with anti-PCNA and anti-actin antibodies. Densitometric quantitation of the bands is shown. (C) Immunohistochemistry for PCNA (red staining) on sections of cortical bone from the indicated genotypes. c, cortical bone; p, periosteum. Scale bar equals 100 μm. (D) Proliferation of Gja1-floxed primary osteoblasts that were transduced with Cre recombinase in order to delete Cx43 and then subsequently transduced with adenoviral Runx2 (n=3 wells per condition). Graphs depict mean ± SD. ** = p<0.001. “‡” signifies a statistical interaction effect between Gja1 and Runx2 at p < 0.05.

To validate that this effect on proliferation was caused by the convergence of Cx43 on Runx2, we overexpressed Runx2 in primary osteoblasts from Gja1-floxed mice. When Cx43 is deleted in these cells, cell proliferation was increased, consistent with published data [16]. This effect of Cx43 on cell proliferation was rescued by overexpression of Runx2 (Fig. 6D), indicating the intersection of Cx43 and Runx2 on the expansion of osteogenic cells.

DISCUSSION

In this study, we examine the intersection of Cx43 and Runx2 function on bone. We demonstrate that deletion of Cx43 impairs ERK and PKCδ signaling that converge on Runx2, affecting Runx2-dependent osteoblast genes, like osteocalcin. We show that overexpression of Runx2 in cells devoid of Cx43 expression can restore osteocalcin gene expression. Further, we describe the skeletal phenotype of mice that are compound hemizygous for the Cx43 gene Gja1 and Runx2 in an attempt to determine whether the convergence of Cx43 and Runx2 function is pertinent to post-natal bone quality. The study presented here is the first to directly assess the impact of this potential interaction between Cx43 and Runx2 and provide further evidence of the existence of a functional interplay between Cx43 and Runx2 that extends to in vivo bone physiology. Our findings show that compound Gja1^+/− Runx2^+/− mice manifest unique skeletal changes in cortical bone and cranial elements not seen in wild-type or singly hemizygous animals, indicating that the haploinsufficiency of both factors compromises a shared signaling axis responsible for normal skeletal formation. While this genetic approach does not establish the hierarchical intersection of this interaction, it does definitively establish the relevance of Cx43-Runx2 axis to the skeleton in vivo. This work is consistent with our prior in vitro work in osteoblasts [13, 24, 25, 27] and the work of others in other cell types [58, 59], indicating an interplay between Cx43 and Runx2.

Aspects of the phenotype of the Cmpd mice closely approximates the phenotype of osteoblast-lineage Cx43 conditional knockout models. Specifically, the cortical phenotype of the Cmpd mice closely resembles that observed in osteoblast-specific Cx43 conditional knockout mouse models, including an expansion of the cross-sectional and marrow areas, elevated osteoclast number, and increased cortical porosity [14, 16, 60]. Additionally, we see an increased proliferative capacity of osteoprogenitor cells, which is similar to that observed in the Dermo1-Cre driven deletion of Cx43 in osteoprogenitor cells [16]. Why these effects on cortical bone are mildly suppressed in female mice is not clear. We speculate that, since estrogens inhibit periosteal bone apposition and enhance endocortical bone apposition, these effects may partially blunt the action of Cx43 to stimulate periosteal apposition and endocortical resorption [61], thereby resulting in the lack of a statistical interaction. However, no studies have done a systematic comparative analysis of Cx43 action in both male and female mice. It would be interesting to examine these findings in estrogen depleted animals to test this hypothesis. Regardless, these overall findings highlight that, while cortical parameters remain relatively unchanged in specimens that only have one functional allele of either Gja1 or Runx2, the combination of having single copies of both Gja1 and Runx2 in the same animal results in a cortical phenotype. This is a convincing indication that some aspect of the impact of Cx43 and Runx2 on bone biology lies not on independent pathways but on a shared pathway. Because we are able to recapitulate the cortical phenotype seen in Cx43 conditional knockout animals, this indicates that part of the phenotype observed in the Cx43 conditional knockout mice is likely due to inefficient Runx2 activation.

Other aspects of the Cmpd skeletal phenotype seem to exacerbate the findings of Runx2^+/−. The importance of Runx2 in regulating craniofacial patterning is well documented. One of the hallmark cranial phenotypes that manifests in Runx2 deficiency is incomplete closure of the sutures resulting in persistent open fontanelles throughout adulthood and hypoplastic clavicles [48]. Likewise, deficiencies in Runx2 expression, as described in cases of cleidocranial dysplasia, have severe consequences on craniofacial patterning and mineralization of the clavicles [48]. The compound effect of Gja1 and Runx2 hemizygosity further exacerbates these defects, resulting in a significant reduction in interparietal mineralization, wider fontanelles and aplastic clavicles in 2-day old neonates, further underscoring the dependent functional intersection of Cx43 and Runx2.

Some aspects of the Cmpd skeletal phenotype are unique or at least unexpected based on either individual model. Consistent with osteoblast-specific conditional knockout models, we did not observe an exacerbated trabecular bone phenotype in our compound Gja1^+/− Runx2^+/− mice. Rather, changes in trabecular bone were exclusive to Runx2^+/− mice, which displayed significant reductions in BV/TV and trabecular number (Tb.N), and a significant increase in trabecular separation (Tb.Sp) when compared to all other genotypes. These findings seemingly suggest that Runx2 plays a more integral role than Cx43 in regulating trabecular bone formation; however, for this to be entirely true, we would expect that Runx2^+/− and Gja1^+/− Runx2^+/− mice would, at minimum, have the same deficits in trabecular bone volume. However, much to our surprise, we observed in Cmpd mice that removal of Cx43 “rescued” the trabecular changes observed in Runx2^+/− mice, indicating that Cx43 may oppose the anabolic function of Runx2 in trabecular bone. An analogous trabecular bone phenomenon was reported in studies by Lloyd and colleagues, which show that conditional deletion of Cx43 attenuates loss of trabecular bone during mouse hindlimb suspension [15, 60]. They concluded that loss of Cx43 prevented the sharing of catabolic unloading signals among osteocytes to produce a low bone mass phenotype. Therefore, a similar explanation could be applied to the rescue seen in these compound hemizygotes, in that when osteoblasts are not able to efficiently progress through differentiation, either genetically through Runx2 reduction or through lack of external stimuli, then Cx43 appears to be antagonistic to trabecular bone formation via the exchange of predominating catabolic signals. Regardless of the explanation, this unexpected “rescue” of the Runx2 trabecular compartment phenotype by Cx43 deficiency is still in accordance with our hypothesis. At some level, Cx43 and Runx2 functionally converge to regulate trabecular bone even if they appear to function antagonistically/inversely in this compartment.

Another unexpected finding was the altered shape of the interparietal bone in 8-week-old male mice. Analogous results have not been reported in any of the Cx43 deletion models, though impacts on cranial shape have been reported in human and animal models of point mutations in Gja1 resulting in oculodentodigital dysplasia [6, 54]. Since Cx43 deficiency has been shown to affect neural crest cell migration to cardiac tissues [62, 63] and neural crest cells contribute to the intraparietal bone development [64], it is possible that an underlying defect in neural crest cell migration could influence the calvarial phenotype of the Cmpd hemizygous mice. Yet, in homozygous Gja1^−/− mice, this defect in neural crest cell migration results in a cardiovascular defect and perinatal lethality [30]. In contrast, Gja1^+/− mice are phenotypically normal and the calvarial phenotype is only unmasked in the context of concurrent Runx2^+/−, implying that the defect is in osteoblastogenesis, rather than neural crest migration. However, future studies will be needed to understand the specific defect occurring to drive this unanticipated phenotype.

Importantly, we do not think that the Cx43 signaling through Runx2 is the only regulator of the action of Cx43 in bone. The role of Cx43 in the skeleton is multifactorial, including osteocyte survival and differential responsiveness to various cues including growth factors, hormones, aging and mechanical load [65–69]. Our results do not exclude other pathways that may converge on osteoprogenitors or that Cx43 regulates signals originating from osteocytes to affect bone. Indeed, deletion of Cx43 throughout the osteoblast lineage, including osteocytes results in a similar skeletal phenotype [1]. However, the Cx43 knockout phenotype does get worse the earlier in the lineage that Cx43 is deleted [1], underscoring that Cx43 plays an important role, not only in osteocytes but also in osteoblasts and osteo-progenitors, and that the Cx43/Runx2 axis may be a fundamentally important point of Cx43’s function in skeletal tissue.

In summary, genetic models reveal a functional intersection between Cx43 and Runx2 that contributes to a skeletal phenotype that largely mimics Cx43 deletion. These data establish this interaction as an important aspect of the molecular basis by which Cx43 influences the skeleton. Further, these data hint at disparate functions of Cx43 in the trabecular and cortical compartment and support the fundamental role of Cx43 in the early osteoblast lineage.