High Bone Mass in Mice Lacking Cx37 Because of Defective Osteoclast Differentiation*

Background: Connexin proteins are essential for cell differentiation, function, and survival. Results: Global deletion of Cx37 results in increased bone mass caused by reduced osteoclast maturation. Conclusion: Our findings demonstrate a previously unrecognized role of Cx37 in bone homeostasis in vivo. Significance: Therapeutic approaches to increase bone mass might be developed by interfering with Cx37 function. Connexin (Cx) proteins are essential for cell differentiation, function, and survival in all tissues with Cx43 being the most studied in bone. We now report that Cx37, another member of the connexin family of proteins, is expressed in osteoclasts, osteoblasts, and osteocytes. Mice with global deletion of Cx37 (Cx37−/−) exhibit higher bone mineral density, cancellous bone volume, and mechanical strength compared with wild type littermates. Osteoclast number and surface are significantly lower in bone of Cx37−/− mice. In contrast, osteoblast number and surface and bone formation rate in bones from Cx37−/− mice are unchanged. Moreover, markers of osteoblast activity ex vivo and in vivo are similar to those of Cx37+/+ littermates. sRANKL/M-CSF treatment of nonadherent Cx37−/− bone marrow cells rendered a 5-fold lower level of osteoclast differentiation compared with Cx37+/+ cell cultures. Further, Cx37−/− osteoclasts are smaller and have fewer nuclei per cell. Expression of RANK, TRAP, cathepsin K, calcitonin receptor, matrix metalloproteinase 9, NFATc1, DC-STAMP, ATP6v0d1, and CD44, markers of osteoclast number, fusion, or activity, is lower in Cx37−/− osteoclasts compared with controls. In addition, nonadherent bone marrow cells from Cx37−/− mice exhibit higher levels of markers for osteoclast precursors, suggesting altered osteoclast differentiation. The reduction of osteoclast differentiation is associated with activation of Notch signaling. We conclude that Cx37 is required for osteoclast differentiation and fusion, and its absence leads to arrested osteoclast maturation and high bone mass in mice. These findings demonstrate a previously unrecognized role of Cx37 in bone homeostasis that is not compensated for by Cx43 in vivo.


Connexins (Cx)
are transmembrane proteins expressed in osteoblasts, osteocytes, and osteoclasts (1). Six connexin proteins arrange in the cell membrane to form hemichannels or connexons. Connexins mediate intercellular communication through gap junction channels formed by two connexons that connect adjacent cells or as unopposed hemichannels, via the release of small molecules to the extracellular medium (2). The C terminus of connexin proteins faces the cytoplasm and through interactions with kinases and structural molecules participate in the regulation of intracellular signaling independently of channel activity (3).
Most cells express multiple connexins that form hemichannels composed of one or two different connexins (4). Cx43, for example, can form hemichannels with Cx37, Cx40, or Cx46. In addition, hemichannels composed by one connexin protein in one cell can dock with hemichannels formed by another connexin in a neighboring cell to form heterotypic gap junction channels, which exhibit unique properties, different from channels composed by only one connexin (5).
In bone, Cx43 is the main connexin expressed in osteoblasts, osteocytes, and osteoclasts, and it has been the focus of research on the role of this family of proteins in the skeleton (1). A recent gene array study showed that Cx37 (GJA4) is also expressed in bone cells, and it is five times more abundant in osteocytes than in osteoblasts (6). However, Cx37 expression in osteoclasts and their precursors has not been examined. In addition, the relative expression level of Cx37 compared with Cx43 in bone and the role of Cx37 on bone cell function remain unknown.
Cx37 is expressed in monocytes/macrophages, platelets, kidney, and ovaries, but it is more abundant in endothelial cells in the blood vessels (7)(8)(9). However, Cx37 deletion does not impair vascular function because Cx40, which is also expressed in endothelial cells, can compensate for the absence of Cx37 (10,11). Indeed, double knock-out mice lacking both Cx37 and Cx40 die at birth as result of vascular abnormalities. The main phenotypic characteristic of Cx37 null mice is female sterility caused by lack of terminal oocyte maturation, absence of ovulation, and accumulation of abnormal corpora lutea. In contrast, male homozygous Cx37 Ϫ/Ϫ and heterozygous Cx37 ϩ/Ϫ mice breed normally.
Variants of the Cx37 gene are associated with several human diseases. Thus, a Cx37 gene polymorphism resulting from replacement of serine 319 (319S in designations below) by proline (319P in designations below) in the regulatory cytoplasmic tail is a diagnostic marker for atherosclerosis in humans; with higher risk for myocardial infarction and reduced risk of coronary artery disease in patients carrying the Cx37-319S allele. The increased permeability exhibited by Cx37-319P hemichannels may mediate the differential effects of the two Cx37 polymorphic alleles. In addition, Cx37-319P, but not Cx37-319S, decreases proliferation when transfected into HeLa or SK-HEP-1 cells (12). The proline variant can be phosphorylated by GSK3␤, an event that leads to reduced channel activity and decreased proliferation in HeLa cells. Interestingly, the same human gene polymorphism is associated with bone mass in a Japanese population, with men carrying the Cx37-319S allele exhibiting lower total body, lumbar spine, femoral neck, and trochanter bone mineral density (BMD) (13).
We examined in the current study the skeletal phenotype of mice with global deletion of Cx37. We found that Cx37 is expressed in osteoblasts, osteocytes, and osteoclasts, albeit at lower levels than Cx43. Nevertheless, lack of Cx37 impairs osteoclastogenesis, without affecting osteoblast differentiation or function. Our data indicate that reduced bone resorption in Cx37 null mice leads to increased bone mass and bone volume, in particular in cancellous bone of the vertebrae.

EXPERIMENTAL PROCEDURES
Animals-Mice with global deletion of Cx37, generated by A. Simon (14) and provided by J. M. Burt (University of Arizona, Tucson, AZ), were maintained in a C57BL/6 background. Mice expressing green fluorescent protein under the control of the dentin matrix protein 1 promoter (DMP1-GFP) in a C57BL/6 background were previously described (15). All mice were fed a regular diet and water ad libitum and maintained on a 12-h light/dark cycle. All protocols involving mice were approved by the Institutional Animal Care and Use Committee of Indiana University School of Medicine.
Mice were genotyped by PCR using specific primer sets as follows: wild type allele was detected using 5Ј-TGC TAG ACC  AGG TCC AGG AAC-3Ј and 5Ј-GTC CCT TCG TGC CTT  TAT CTC-3Ј, and the mutant allele was detected by 5Ј-GAT  CTC TCG TGG GAT CAT TG-3Ј and 5Ј-TGC TAG ACC  AGG TCC AGG AAC-3Ј. An amplified fragment of 750 bp corresponds to the wild type allele and of 233 bp to the mutant allele. The mice received intraperitoneal injections of calcein (20 mg/kg; Sigma) and alizarin red (20 mg/kg; Sigma) 7 and 2 days before sacrifice, respectively, to allow for dynamic histomorphometric measurements (16).
Bone marrow cells were isolated from Cx37 ϩ/ϩ and Cx37 Ϫ/Ϫ by flushing the bone marrow out with ␣-minimal essential medium supplemented with 15% FBS and 1% penicillin/streptomycin and then plated at a density of 4 ϫ 10 5 /cm in 24-well plates for 48 h. Nonadherent cells were collected and either plated for osteoclast formation or digested to obtain mRNA. For the osteoclastogenesis assay, 2 ϫ 10 5 nonadherent cells/ cm 2 were seeded and cultured with 20 ng/ml recombinant murine M-CSF (PeproTech Inc., Rocky Hill, NJ) for 24 h. Subsequently, 80 ng/ml recombinant murine sRANKL (PeproTech Inc.) was added. Medium was changed every 2 days for 5 days. Osteoclasts were enumerated after staining for TRAPase using a commercial kit (Sigma-Aldrich). Images were acquired using a Zeiss Axiovert 35 microscope equipped with a digital camera. mRNA was isolated from parallel cultures, and gene expression was measured by quantitative RT-PCR (Applied Biosystems, Foster City, CA).
Adherent cells were collected after the initial 48 h in culture, plated at a density of 4 ϫ 10 5 /cm 2 , and cultured until reaching 95-100% of confluence. Then differentiation medium (␣-minimal essential medium supplemented with 10% FBS, 50 g/ml ascorbic acid, and 10 mM ␤-glycerophosphate) was added. Medium was changed every third day. After 14 days in culture cells were stained with alizarin red for 10 min, as published (23). Images were acquired using a Zeiss Axiovert 35 microscope equipped with a digital camera. After imaging the cells, alizarin red was dissolved in 10% acetic acid for 30 min at room temperature. The content of the wells were transferred to microtubes and incubated at 85°C for 10 min, followed by centrifugation and neutralization of the supernatants with 10% ammonium hydroxide. Absorbance at 405 nm was measured to estimate the level of mineralization. Parallel cultures were digested to obtain mRNA and perform quantitative RT-PCR (Applied Biosystems).
Fluorescent-activated Cell Sorter-Freshly isolated bone marrow cells were incubated with anti-Mac-1-APC (CD11b) and anti-Gr1(granulocyte differentiation antigen 1)-FITC followed by sorting by FACS, as published (24). All antibodies were purchased from e-Biosciences. FACSCalibur and FlowJo were used for acquisition and analysis, respectively.
Whole Mount Skeletal Staining-Cartilage and mineralized tissue were analyzed in newborn mice using alizarin red/Alcian blue staining, as previously published (25). In brief, newborn mice were deskinned, eviscerated, and kept in 100% ethanol for 48 h. The carcasses were fixed in acetone for 24 h and then stained for 5 days in a solution containing 0.1% alizarin red, 0.3% Alcian blue, acetic acid, and 70% ethanol (1:1:1:17, v/v/v/ v). Carcasses were then transferred to a solution of 1% KOH overnight, to a solution of glycerol:KOH (2:8) overnight, and finally to 100% glycerol.
Bone Mineral Density and Microcomputed Tomography (CT) Analysis-Longitudinal study was performed monthly from 2 to 5 months of age by dual-energy x-ray absorptiometry (DEXA) using a PIXImus densitometer (G.E. Medical Systems, Lunar Division, Madison, WI) (20). BMD measurements included total BMD (excluding the head and tail), L1-L6 vertebra (spinal BMD), and entire femur (femoral BMD). For CT analysis, femora and lumbar vertebrae obtained from male mice at age 2.5 and 5 months were cleaned of adhering tissue and frozen until imaging at 6 m pixel resolution on a Skyscan 1172 (SkyScan, Kontich, Belgium) (20).
RNA Extraction and Quantitative RT-PCR (qPCR)-RNA was purified from cells or bones using TRIzol reagent (Invitrogen), as previously described (26). qPCR was performed using the housekeeping gene GAPDH, and the ⌬Ct method. Primers and probes were designed using the Assay Design Center (Roche Applied Science) or were commercially available (Applied Biosystems).
Bone Histomorphometry-Vertebrae (L6 and L2 for 2.5-and 5-month-old mice, respectively) were dissected, fixed, and  The symbols correspond to means Ϯ S.D., n ϭ 8 -17. *, p Ͻ 0.05 versus Cx37 ϩ/ϩ mice by t test. c, total, spinal, and femoral bone mineral density was assessed monthly by DEXA in Cx37 ϩ/ϩ and Cx37 Ϫ/Ϫ mice from 2 to 5 months of age. The symbols correspond to means Ϯ S.D., n ϭ 8 -17. *, p Ͻ 0.05 versus Cx37 ϩ/ϩ mice by t test. embedded in methyl methacrylate. Static historphometric analysis was performed on TRAP-stained sections, as previously described (20). Sequential sections were stained with von Kossa tetrachrome to visualize mineralized bone (16). Dynamic histomorphometric analysis of bone sections was performed on unstained sections, avoiding the primary spongiosa. All measurements were obtained using OsteoMeasure high resolution digital video system (OsteoMetrics Inc., Decatur, GA). The terminology and units used are those recommended by the Histomorphometry Nomenclature Committee of the American Society for Bone and Mineral Research (27).
Biomechanical Testing-Femora and lumbar (L5) vertebrae obtained from 5-month-old mice were tested via three-point bending and compression, respectively, following previously published protocols with modifications (28). Briefly, bones were thawed to room temperature, hydrated in 0.9% saline, and loaded to failure at 2 mm/min with force versus displacement data collected at 10 Hz using a servo-hydraulic test system (Test Resources). Femora were loaded to failure in an anterior-posterior direction with the upper contact area at the mid-diaphysis (50% total bone length), and the bottom contact points centered around this point and separated by 8 mm. The posterior processes and end plates were removed prior to testing of the vertebra. Structural mechanical properties, ultimate load, stiffness, and energy to failure were determined from the load deformation curves using standard definitions, whereas material level estimations of ultimate stress, modulus, and toughness were calculated using standard equations. Cross-sectional moment of inertia and anterior-posterior diameter were determined by CT and were used to calculate material level properties, as previously described (29). . Global deletion of Cx37 results in high bone volume in cancellous bone. a, CT analysis of cancellous bone microarchitecture in L6 vertebrae in 2.5-month-old mice (n ϭ 3-4) and L4 vertebrae in 5-month-old mice (n ϭ 7-8). BV/TV: bone volume corrected by tissue volume; b, cancellous bone microarchitecture in the distal femur in 2.5-month-old mice (n ϭ 3-4) or 5-month-old mice (n ϭ 9 -11). c, cortical bone geometry in the femoral midshaft in 5-month-old mice (n ϭ 8 -13). The bars are means Ϯ S.D. *, p Ͻ 0.05 versus Cx37 ϩ/ϩ mice by t test. Representative images of vertebrae from wild type and Cx37 Ϫ/Ϫ littermate controls are shown in a. MARCH 21, 2014 • VOLUME 289 • NUMBER 12

Cx37 and Osteoclast Differentiation
Bone Turnover Markers-Plasma C-telopeptide fragments (RatLaps; Immunodiagnostic Systems Inc., Fountain Hill, AZ), osteocalcin (Biomedical Technologies, Soughton, MA), and sclerostin levels (ALPCO Immunoassays, Salem, NH) were measured as described by the manufacturer. Alkaline phosphatase activity was assessed by a standard automated method using a Randox Daytona chemical analyzer.
Histological Visualization of Ovaries-Ovaries were fixed in 3.7% PBS-buffered formaldehyde at room temperature for 48 h, followed by transfer to 70% ethanol. Specimens were subsequently dehydrated in ascending ethanol concentrations and embedded in paraffin. 5-m-thick sections were processed for hematoxilin and eosin staining. Images were acquired using an Olympus BX51 TRF microscope equipped with a digital camera.
Statistical Analysis-The data were analyzed by using Sigma-Plot (Systat Software Inc., San Jose, CA). Differences were evaluated by Student's t test and considered significant for p Ͻ 0.05. All values are reported as the means Ϯ S.D.

Cx37 Is Expressed in Osteoblasts, Osteocytes, and Osteoclasts-
Gene expression analysis shows that Cx37 is expressed in whole bone at approximately four times lower levels than Cx43 (Fig.  1a). Cx37 can also be detected at much lower levels than Cx43 in the OB-6 osteoblastic cells, in immortalized osteoblastic calvaria cells, and in MLO-Y4 osteocytic cells. In primary osteoblasts, the expression of Cx37 is ϳ300 times lower than that of Cx43, whereas in primary osteocytes it is ϳ60 times lower (Fig.  1b). In addition, Cx37 is detected in osteoclasts generated by treatment of nonadherent bone marrow cells with M-CSF and sRANKL. In these osteoclast cultures, the expression of Cx37 is also lower than Cx43 (ϳ120-fold).
Global Deletion of Cx37 in Mice Leads to High Bone Mass-Cx37-deficient mice did not exhibit gross abnormalities at birth and qualitatively appear to have a distribution of cartilage and mineralized bone similar to that of wild type littermates (Fig.  2a). Body weight was similar in male mice from either genotype from 2 to 5 months (Fig. 2b). Femoral length was not different between phenotypes at 5 months of age (15.19 Ϯ 0.27 versus 15.44 Ϯ 0.18 mm for wild type Cx37 ϩ/ϩ and Cx37 Ϫ/Ϫ mice, respectively). Female mice exhibit a slight increase in weight at 1 (not shown), 2, and 5 months of age (Fig. 2b).
Cx37 Ϫ/Ϫ Mice Exhibit Increased Cancellous Bone Volume and Increased Mechanical Strength in the Femur-CT analysis showed higher bone volume and trabecular number in the cancellous bone of the lumbar vertebra at 2.5 and 5 months of age (Fig. 3a). This was associated with increased trabecular thickness and a tendency toward higher trabecular number in the young animals. In the older mice, only trabecular number was higher. Similar effects of Cx37 deletion were observed in the cancellous bone in the distal femur at 2.5 months of age (Fig. 3b). However, there were no differences between genotypes in 5-month-old mice. Deletion of Cx37 does not result in major changes in cortical architecture or bone material density in the femoral midshaft compared with wild type littermates, with only a slight but significant reduction in cortical thickness (Fig. 3c).
Lumbar vertebrae from Cx37 null mice exhibited higher energy to ultimate load, an indication of the capacity of the bone to resist compression forces (Fig. 4a). Ultimate force and stiffness showed a tendency toward higher values in mice lacking Cx37, but the difference did not reach significance. Deletion of Cx37 resulted in higher cortical bone strength as evidenced by increased ultimate force and mechanical stiffness, compared with wild type littermates in the cortical bone of the femoral midshaft (Fig. 4b). Estimations of material properties, on the other hand, showed no difference between genotypes.
Static histomorphometry revealed a lower number of osteoclasts per bone surface and reduced bone surface covered by osteoclasts and eroded surface/BS on the vertebral cancellous surface of Cx37 Ϫ/Ϫ mice compared with controls at 2.5 and 5 months of age (Fig. 5, a and b). The number of osteoblasts/BS and the bone surface covered by osteoblasts, on the other hand, was not different between genotypes at either age. The lower osteoclast parameters combined with the lack of change on osteoblasts resulted in an increase in quiescent bone surface in Cx37 Ϫ/Ϫ mice. Dynamic histomorphometry showed no differences in mineralizing surface/BS, mineral apposition rate, and bone formation rate/BS, parameters that reflect osteoblast number and activity in 5-month-old mice (Fig.  5c). Differences in circulating markers of bone resorption (C-telopeptide fragments) or formation (osteocalcin and alkaline phosphatase) or in sclerostin, the inhibitor of Wnt signaling, were not detected in either male or female mice at 2, 3, or 5 months of age (Fig. 6).  MARCH 21, 2014 • VOLUME 289 • NUMBER 12

Deletion of Cx37 Decreases Osteoclast Fusion and Differentiation, but Does Not Affect Osteoblast Differentiation or
Function-To test the cellular basis for the increased bone mass in the absence of Cx37, we examined osteoclast-specific gene expression in whole bone preparations. As expected, Cx37 expression was undetectable in lumbar vertebrae from Cx37 Ϫ/Ϫ mice, whereas Cx43 expression was not different from Cx37 ϩ/ϩ mice (Fig. 7a). Although the means of the osteoclastspecific genes TRAP and cathepsin K were approximately half in Cx37 Ϫ/Ϫ bones compared with Cx37 ϩ/ϩ , the difference did not reach significance. The ability of Cx37-deficient osteoclast precursors to generate mature cells ex vivo in the presence of M-CSF and sRANKL was impaired, as evidenced by a lower osteoclast number and size and number of nuclei/osteoclast (Fig. 7b).
Although the percentage of CD11b-positive cells (expressing or not the myeloid cell marker Gr1) present in the bone marrow of Cx37 Ϫ/Ϫ mice was similar to Cx37 ϩ/ϩ mice, as evidenced by FACS, nonadherent bone marrow cell preparations isolated from Cx37 Ϫ/Ϫ mice exhibited higher expression of osteoclast precursors markers CD11b, CD14, and RANK, compared with cells isolated from wild type littermates (Fig. 7c). These findings suggest that in the absence of Cx37, the expression of osteoclast progenitor-associated genes is increased without a concomitant increase in the abundance of osteoclast precursor cells.
Mature osteoclasts derived from wild type or Cx37 Ϫ/Ϫ mice expressed similar levels of Cx43 mRNA but expressed reduced levels of osteoclast-specific genes (Fig. 7d). Thus, the lack of Cx37 led to a reduction in RANK levels of ϳ30%, whereas the markers of mature osteoclasts TRAP, cathepsin K, and calcitonin receptor were all ϳ50% lower. Moreover, NFATc1 (nuclear factor of activated T-cells), a transcription factor essential for osteoclast differentiation induced by RANKL (30), was also lower in Cx37-deficient osteoclasts. Similarly, the expression of DC-STAMP (dendritic cell-specific transmembrane protein) and Atp6v0d2 (d2 isoform of vacuolar ATPase V o domain), NFATc1 targets that participate in osteoclast fusion, was also reduced. Matrix metalloproteinase 9 and CD44, which are involved in osteoclast migration and fusion (31)(32)(33), were also decreased in Cx37-deficient cultures. On the other hand, either other molecules associated with osteoclast function were not changed (osteoclast-associated receptor, OSCAR), or their decrease did not reach significance (integrin ␤3 and E-cadherin).
Recent evidence indicates that the Notch signaling pathway negatively modulates osteoclast differentiation (34,35). We therefore investigated whether deleting Cx37 altered regulation of Notch signaling. Freshly isolated Cx37-deficient nonadherent bone marrow cells expressed higher levels of the Notch1-3 receptors compared with wild type cells, whereas the expression of the Notch target gene Hey-1 was unchanged (Fig. 8a). Moreover, the number of viable cells in cultures treated with M-CSF for 24 h was significantly higher in Cx37 Ϫ/Ϫ cells, and this difference disappeared when the cultures were treated with a Notch inhibitor (Fig.  8b). Increased expression of the Notch target Hey-1 was detected in mature osteoclasts and in whole bone lysates, whereas the levels of the receptors Notch1-3 were either unchanged or slightly increased. This evidence suggests that elevation of Notch signaling pathway in the absence of Cx37 contributes to the inhibition of osteoclastogenesis in Cx37 Ϫ/Ϫ mice.
Ex vivo cultures of adherent bone marrow cells maintained in the presence of ascorbic acid and ␤-glycerophophate to induce osteoblast differentiation showed no difference in calcium deposition, a marker for mineralizing activity, between cells isolated from Cx37 Ϫ/Ϫ and Cx37 ϩ/ϩ controls (Fig. 9a). Moreover, the expression of the osteoblastic genes osteocalcin, alkaline phosphatase, collagen type 1a1, and bone sialoprotein was not affected by Cx37 deletion in the osteoblastic cells derived from adherent bone marrow cells (Fig. 9a) or in whole lumbar vertebra lysates (Fig. 9b). Similarly, no differences were seen in the level of mineralization on the vertebral bone in 2.5-and 5-month-old mice.

DISCUSSION
In this report, we describe the skeletal phenotype of mice with global deletion of Cx37, a member of the connexin family of proteins with previously unrecognized functions in bone cells. Cx37 is expressed in whole bone at levels ϳ4-fold lower than Cx43. Osteoblasts, osteocytes, and osteoclasts all express Cx37, albeit at 60 -400-fold lower levels than Cx43. We found that the skeletal phenotype of mice with global deletion of the gene results from an osteoclast cell autonomous defect. Although nonadherent bone marrow cells express higher levels of markers of osteoclast precursors, their ability to become mature oste-oclasts is impaired. This evidence, together with the decrease in osteoclast parameters in vivo, suggests that the increased bone mass and volume observed in Cx37 Ϫ/Ϫ mice is due to defective osteoclast maturation and the consequent decreased osteoclast bone resorbing activity. Whether decreased osteoclastogenesis in the absence of Cx37 is due to lack of cell to cell communication, reduced communication with the extracellular medium The numbers indicate percentages of cells positive for the marker of osteoclast precursors CD11b. Expression levels of CD11b, CD14, and RANK mRNA were measured in nonadherent bone marrow cells. mRNA relative levels were measured by qPCR and corrected by GAPDH. The bars are means Ϯ S.D., n ϭ 4 mice/genotype. *, p Ͻ 0.05 versus Cx37 ϩ/ϩ mice by t test. d, gene expression was assessed in mature (differentiated) osteoclasts generated by treatment of bone marrow nonadherent cells with M-CSF/sRANKL. Genes associated with osteoclast maturation, function, and fusion are shown. mRNA relative levels were measured by qPCR and corrected by GAPDH. The bars are means Ϯ S.D., n ϭ 4 mice/genotype. *, p Ͻ 0.05 versus Cx37 ϩ/ϩ mice by t test. MARCH 21, 2014 • VOLUME 289 • NUMBER 12 through Cx37 hemichannels, or defective Cx37-protein association independent of channel function will be investigated in future studies. Similar to our in vivo results, treatment of osteoclast precursors with agents that disassemble connexin channels leads to decreased osteoclast differentiation in vitro, with reduced fusion of osteoclast precursors compared with vehicle treated cultures (36 -38). GAP27, a peptide that inhibits the function of the channel/hemichannel formed by Cx43 or Cx37, has been shown to block osteoclast differentiation (39,40). This decrease in osteoclastogenesis was attributed to reduced Cx43 function, because this was the only connexin protein described in osteoclastic cells. Our data, however, show that Cx37 expression is required for osteoclast differentiation and the regulation of bone mass. A preliminary study presented in abstract form indicates that osteoclast-specific deletion of Cx43 has a similar impact on osteoclastogenesis as the absence of Cx37 (41). This raises the possibility that channels formed by both connexins (heteromeric) or docking of different channels each formed by one connexin are required for osteoclast maturation. The co-localization of Cx43 and Cx37 in heteromeric channels may not only alter channel properties but also affect the interaction of the connexins with intracellular signaling molecules. On the other hand, expression of Cx43, but not Cx37, in osteoblast precursors is required for full osteoblast differentiation and for normal ossification of the developing embryos (42), suggesting that in osteoblastic cells the two connexins do not interact and that Cx43 is the main connexin active in these cells.

Cx37 and Osteoclast Differentiation
Preliminary mechanistic studies show that cells lacking Cx37 exhibit increased Notch signaling activation in response to proosteoclastogenic stimuli, whereas in mature osteoclasts, Notch1-3 mRNA levels were normal. Furthermore, the number of viable cells is higher in nonadherent bone marrow cells treated with M-CSF but returns to wild type control levels by inhibition of Notch signaling. The increase in cell number could be due to higher proliferation or reduced apoptosis. Enhanced proliferation of Cx37 Ϫ/Ϫ osteoclast progenitor cells is supported by evidence showing that Cx37 suppresses the proliferation of rat insulinoma cells (43,44). Our data, however, contrast with evidence showing that deletion of Notch1-3 in osteoclast precursors using LysM-Cre (35) leads to an increase in proliferation. This discrepancy could be due to changes in other genes in the absence of Cx37 or to the temporal preosteoclast-specific elevation of Notch1-3 signaling in our studies. The expression of the Notch signaling target Hey-1 is highly increased in mature osteoclasts from Cx37-deficient mice, consistent with evidence that the Notch/RBPJk signaling pathway inhibits osteoclast differentiation (34,35). Further studies are required to establish the mechanism by which Cx37 modulates cell number and osteoclast differentiation through the Notch signaling pathway.
The absence of Cx37 in osteoblasts does not appear to alter the ability of these cells to synthesize bone matrix. However, our preliminary studies show that MLO-Y4 osteocytic cells in which the expression of Cx37 has been silenced using shRNA exhibit reduced RANKL/OPG ratio (not shown), which could contribute to reduced osteoclast formation. This contrasts with the effect of deletion of Cx43 from osteocytic cells, which results in higher RANKL/OPG ratio and increased osteoclastic bone resorption on endocortical bone surfaces (20,45). Cellspecific deletion of the Cx37 will be required to determine whether changes in osteoclastogenic cytokines expressed by osteocytes contribute to increased bone mass in Cx37 Ϫ/Ϫ mice.
The phenotype of Cx37 Ϫ/Ϫ mice is more pronounced in males than in females. Although the reasons for this sexual dimorphic skeletal phenotype are not known, the milder phenotype observed in female mice could be due to altered sex steroid levels secondary to the deficient oocyte maturation and abnormal corpora lutea previously reported (14). We confirmed the absence of terminally differentiated oocytes in our female Cx37 Ϫ/Ϫ mice (Fig. 10). More osteoclasts resulting from lower circulating estrogens could counteract the decreased osteoclasts induced by Cx37 deficiency in female mice. Remarkably, a similar gender-specific effect has been recently found in a polymorphism of the Cx37 gene associ-ated with decreased bone mass in males but not in females in a Japanese population (13). This suggests that the presence of androgen, rather than changes in estrogen levels, is required for the effect of Cx37 deletion in bone. Future work will explore the potential interaction between estrogen and androgen with Cx37.  Deletion of Cx37 has a more profound effect at 2-2.5 months of age than in 5-month-old mice. This is evidenced by a higher bone volume and lower osteoclast and eroded surface in the vertebral bone of younger Cx37 Ϫ/Ϫ animals compared with wild type. Together with the fact that the difference in spinal BMD between the two genotypes is maintained throughout the study, this evidence suggests that deletion of Cx37 has a greater impact in the developing skeleton and is maintained into adulthood in the spine. The consistent response at the vertebra is contrasted with that at the long bones, where the phenotype existed across the age span when assessed by DEXA, but only in the younger animals when trabecular bone was measured using CT. The reason for the absence of effect in the distal femur (and the proximal tibia; not shown) at 5 months is not clear and warrants further investigation.
The higher effect of Cx37 removal on bone mineral density and bone volume in the cancellous bone and, in particular, in the axial skeleton contrasts with deletion of Cx43 from osteoblasts/osteocytes, in which the main effect is geometrical changes in cortical bone (20,45,46). Although Cx37 and Cx43 belong to the same ␣-family of connexins, their channels differ markedly in unitary conductance, pore diameter, specificity of ion permeability, and, potentially, ability to bind intracellular molecules (47). The two connexins are co-expressed in several tissues, but with different cellular specificity, expression levels, and function. Thus, the connexins exert opposing effects on rat insulinoma cell cycle progression; Cx37 increases the percentage of cells in G 0 /G 1 and decreases the percentage of cells in the S phase, whereas Cx43 decreases the percentage of cells in G 0 /G 1 and increases the percentage in the S phase (43). In addition, Cx43, but not Cx37, is able to confer responsiveness to bisphosphonates in connexin-deficient HeLa cells (48). In the female reproductive tissue, Cx43 is expressed in granulosa cells and Cx37 is mainly expressed in oocytes, and the expression of both connexins is required for full oocyte maturation. Interestingly, ectopic expression of Cx43 in oocytes can rescue the effect of Cx37 deletion, suggesting that Cx43 can exert the same functions than Cx37 in oocytes (49). However, this is not the case in osteoclasts, where Cx43 is highly expressed but cannot compensate for the lack of Cx37.
Femora from Cx37 Ϫ/Ϫ mice exhibit tendencies toward a decrease in total bone area and toward an increase in marrow cavity area. Although these differences do not reach significance, the combination of these changes results in a significant decreased cortical thickness. Reduced cortical thickness could lead to lower mechanical strength. However, femora from Cx37 Ϫ/Ϫ mice are stronger, as evidenced by higher ultimate force and stiffness. This suggests that deletion of Cx37 results in changes in the bone matrix composition, likely the organic phase because it primarily dictates toughness, whereas the mineral phase dictates modulus, leading to improved bone strength. On the other hand, deletion of Cx43 from mature osteoblasts/osteocytes or only from osteocytes does not change ultimate force or stiffness but results in reduced bone material properties (Young's modulus and ultimate stress) (50). Similarly, expression of a Cx43 mutant associated with occulodentodigital dysplasia results in defective material properties (51). This evidence suggests that Cx43 and Cx37 have different effects on mature osteoblast/osteocyte synthetic activity or the ability of the cells to maintain the extracellular matrix in cortical bone. Future studies are required to determine the nature of the changes induced by Cx37 deletion on the bone material.
In summary, our studies show that Cx37 expression in osteoclast precursors is required for osteoclast differentiation and full maturation. Lack of Cx37 leads to a sustained increased in bone mass, in particular in the axial skeleton. Our findings demonstrate that Cx37 modulates osteoclast differentiation and therefore could be targeted to reduce osteoclastic bone resorption. Inhibition of connexin function in vivo has been tested using mimetic peptides, which block connexin channels (52). This approach has been successful in preventing alterations in blood-brain barrier permeability and spinal cord injury (53,54), raising the possibility that these mimetic peptides could be used to target specifically osteoclast precursors leading to reduced bone resorption and maintenance of bone mass.