Of the ephrin (Eph) receptors, mature osteoclasts express predominantly EphA4. This study sought to determine if EphA4 has a regulatory role in osteoclasts. Treatment of RAW/C4 cells with Epha4 small interfering RNAs (siRNAs) increased average size, Ctsk mRNA expression level, and bone resorption activity of the derived osteoclast-like cells. Activation of the EphA4 signaling in osteoclast precursors with EfnA4-fc chimeric protein reduced cell size and resorption activity of the derived osteoclasts. Homozygous Epha4 null mice had substantially less trabecular bone in femur and vertebra compared to wild-type controls. The bone loss was due to a decrease in trabecular number and an increase in trabecular spacing, but not to an increase in osteoclast-lined bone surface or an increase in the number of osteoclasts on bone surface. Dynamic histomorphometry and serum biomarker analyses indicate that bone formation in Epha4 null mice was reduced slightly but not significantly. Osteoclasts of Epha4 null mice were also larger, expressed higher levels of Mmp3 and Mmp9 mRNAs, and exhibited greater bone resorption activity than wild-type osteoclasts in vitro. Deficient Epha4 expression had no effects on the total number of osteoclast formed in response to receptor activator of NF-κB ligand nor on apoptosis of osteoclasts in vitro. It also did not affect the protein-tyrosine phosphorylation status of its ligands, EfnB2, EfnA2, and EfnA4, in osteoclasts. Deficient Epha4 expression in Epha4 null osteoclasts activated the β3-integrin signaling through reduced phosphorylation of the tyr-747 residue, which led to increased binding of the stimulatory talin and reduced binding of the inhibitory Dok1 to β3-integrin. This in turn activated Vav3 and the bone resorption activity of osteoclasts. In conclusion, we demonstrate for the first time that EphA4 is a potent negative regulator of osteoclastic activity, mediated in part through increased Dok1 binding to β3-integrin via an increase in EphA4-dependent tyr-747 phosphorylation. © 2014 American Society for Bone and Mineral Research.
Ephrin (Eph) receptors, the largest family of receptor protein-tyrosine kinases (PTKs), are categorized into two multigene families: EphA (A1–A10) and EphB (B1–B6). Eph receptors are activated by membrane bound ligands, Efn, of neighboring cells. There are also two structurally distinct Efn families: EfnA (A1–A5), which attach to plasma membrane through a glycophosphatidylinositol anchor, and EfnB (B1–B3), which are transmembrane proteins. With the exception of EphB4 that interacts exclusively with EfnB2, each Eph receptor can bind more than one Efn ligand and vice versa. Also with exception of EphA4 and EphB2,[1, 3] EphA receptors preferentially bind EfnA ligands, and EphB receptors bind selectively EfnB ligands. Binding of Efn induces conformational change in the cytoplasmic portion of the Eph receptor, leading to autophosphorylation of two juxtamembrane tyrosines, which provide binding sites for SH2-containing signaling proteins and the relief of inhibition of the PTK domain. Subsequent phosphorylation of additional tyrosines helps to stabilize its active conformation. Efns not only function as ligands, but are themselves capable of receptor-like signaling transduction. Interaction with EphB receptors induces protein-tyrosine (pY) phosphorylation of EfnB ligands, which then recruits and activates SH2-containing or PDZ-containing proteins to transmit the signal. Interaction with EphAs causes clustering of EfnAs and recruitment of signaling proteins to the lipid raft to initiate signal transduction. Thus, the Efn-Eph interaction offers a unique bidirectional signaling mechanism. The Eph-mediated signal is known as the forward signaling and the Efn-triggered signal is termed the reverse signaling.
Eph receptors have been extensively studied for their roles in the developing and adult central nervous system (CNS) and in the developing cardiovascular system. The Efn-Eph bidirectional signaling is implicated in many physiological and pathological processes, including the regulation of insulin secretion, bone homeostasis, immune function, blood clotting, pathological forms of angiogenesis, stem cell differentiation, and cancer.[8-10] An important regulatory role of the Efn-Eph bidirectional signaling in bone biology has recently emerged. Accordingly, the EfnB-EphB signaling is shown to be essential for skeletal patterning. Deficient Efnb1 expression in mice caused perinatal lethality and severe skeletal defects.[12, 13] Conditional disruption of Efnb1 in osteoblasts reduced bone size and peak bone mass in vivo. Efnb1 regulates marrow stromal cell differentiation by influencing TAZ transactivation via complex formation with Nherf1. Mutations in the Efnb1 gene are associated with craniofrontonasal syndrome in humans. EfnB2 promotes osteoblast differentiation in vitro and bone formation in vivo through its action on EphB4, and the forward signaling of EphB4 may play a role in the local anabolic action of parathyroid hormone (PTH)/PTH-related protein (PTHrP) in bone.
Recent studies have implicated an important regulatory role for the reverse signaling of Efns in osteoclasts and resorption. The EphB4-mediated activation of the reverse signaling of EfnB2 in osteoclast precursors has been shown to suppress osteoclast differentiation through an inhibition of the Fos-Nfatc1 cascade. Similar suppressive action of the reverse signaling of EfnB1 on osteoclast differentiation has been reported. EphA2-mediated activation of the reverse signaling of EfnA2 in osteoclast precursors also enhanced osteoclast differentiation and bone resorption. Although overexpression of Epha2 in osteoclast precursors impaired osteoclast differentiation, this effect was not prevented by overexpression of a dominant negative Epha2 mutant, indicating that the reverse signaling of EfnA2, and not the forward signaling of EphA2, is responsible for the enhanced osteoclast differentiation.
The regulatory role of the forward signaling of Eph receptors in osteoclasts is unclear. Unlike osteoblasts that express multiple EphA and EphB receptors, mature osteoclasts express predominantly EphA4.[17, 19] Because Epha4 null mice have shown a significant skeletal phenotype, ie, craniosynostosis that may be associated with defective bone remodeling, we sought to investigate the potential regulatory role of EphA4 in osteoclasts and bone resorption. In this report, we present compelling evidence that deficient Epha4 expression in osteoclastic cells yielded large and functionally more active osteoclasts in vitro and that disruption of the Epha4 gene in mice resulted in marked increases in bone loss and osteoclastic resorption due to an increase in osteoclastic activity and not to an increase in the osteoclast number in vivo. This study presents strong in vivo and in vitro evidence that the forward signaling of EphA4 is a potent negative regulatory mechanism of the osteoclast activity.
Subjects and Methods
RAW264.7 cells (RAW/C4 clone) were originally obtained from Dr. A. Ian Cassady of the Griffith University (Gold Coast, Queensland, Australia). Tissue culture media and fetal bovine serum (FBS) were purchased from Life Technologies (Grand Island, NY, USA) and Atlanta Biologicals (Lawrenceville, GA, USA), respectively. The enhanced chemiluminescence (ECL) detection kit was obtained from Millipore (Billerica, MA, USA). Nitrocellulose transfer membrane (Transblot Nitrocellulose) was a product of BioRad (Hercules, CA, USA). Recombinant soluble receptor activator of NF-κB ligand (sRANKL) and macrophage colony stimulating factor (m-CSF) were from Santa Cruz Biotechnology (Santa Cruz, CA, USA) and CalBiochem (San Diego, CA, USA), respectively. EphA4-fc and EfnA4-fc chimeric proteins were obtained from R&D Systems (Minneapolis, MN, USA). An ELISA-based apoptosis assay kit (Cell Death Detection ELISA) was a product of Roche Diagnostics GmbH (Manneheim, Germany).
Heterozygous Epha4 null breeder mice (in C57BL6/J genetic background), originally created in the laboratory of Prof. Andrew Boyd of the Queensland Institute for Medical Research in Australia, were generously provided to us by Prof. Elena Pasquale of the Sanford-Burnham Medical Research Institute (La Jolla, CA, USA). Initial crosses among the heterozygous mice on normal laboratory mouse chows yielded no viable homozygous Epha4 null pups. Upon the recommendation of Prof. Andrew Boyd, the breeders were put on high fat-containing breeder's diet for more than two generations. Viable homozygous Epha4 null mice were then obtained. All pups were subsequently fed the high-fat–containing breeder's diet. The number of homozygous Epha4 null mutant mice in litters born from crossing heterozygotes showed a subnormal Mendelian ratio (<10%). Transferring the genetic background back to 129J slightly increased the ratio to ∼15%. The bone phenotype of homozygous Epha4 null mutants in 129J genetic background was very similar to that of homozygotes in C57BL/6J genetic background (not shown). Most experiments used mice of the C57BL/6J genetic background. Although the number of viable homozygous Epha4 null mice in each litter was below the expected Mendelian ratio, we noted no noticeable health problems in viable homozygous Epha4 null mice other than the well-known neurological abnormalities. The mutant mice have a normal lifespan, and show no evidence for a compromised immunity. No special care, housing requirement, or treatments are needed for handling these mice. EphA4 is known to have essential functions in the development, maturation, and functions of neurons and the CNS.[23-26] We surmise that a likely cause for the subnormal Mendelian ratio in our homozygous Epha4 null mice is that embryos with severe underdeveloped CNS due to Epha4 deficiency did not survive during early gestation periods. The fact that viable homozygous Epha4 null pups was obtained only after breeder mice were fed a high-fat–containing diet for several generations supports this possibility, as high maternal intakes of fats is essential for and supports embryonic brain development. Moreover, there were no more perinatal deaths with the homozygous Epha4 null mice after the colony was put on a high-fat diet for several generations, albeit the yield of homozygous null mice was still below the expected Mendelian ratio.
All animal protocols were reviewed and approved by the Animal Care and Use Committee of the Pettis Memorial VA Medical Center.
Tail vertebrate tissue (2–3 mm) was taken from each pup at weaning and digested overnight using the DNeasy kit from Qiagen (San Diego, CA, USA). The quality and quantity of genomic DNA were analyzed by measuring the absorbance ratio of 260 nm/280 nm. A PCR-based genotyping assay was used to identify EphA4 null mice using an extra neo primer corresponding to a unique region downstream to the 5′ end of the neo gene. The sequences of primers (synthesized by Integrated DNA Technologies, San Diego, CA, USA) were as follows: (1) the forward primer corresponds to the exon 3 of Epha4 (5′-ACC GTT CGA CTA GCC CAG-3′); (2) the reverse primer corresponds to the intronic sequence 3′ to exon 3 (5′-AGC CTT GCC ACC TGG AGC-3′); and (3) the forward primer corresponds to a unique region of the neo cassette (5′-TCC TCG TGC TTT ACG GTA TC-3′). The PCR reaction included a hot start at 95°C with Platinum Taq polymerase (Invitrogen, Carlsbad, CA, USA): 95°C for 5 minutes followed by 35 amplification cycles of 94°C for 30 seconds, 56°C for 30 seconds, and 68°C for 1 minute. These cycles were followed by 68°C for 1 minute and ended at 4°C. The PCR product was analyzed by 1.5% agarose gel electrophoresis (ThermoFisher Scientific, Los Angeles, CA, USA). The wild-type (WT) littermates (with the exon3/forward and exon3/reverse primer set) visualized a single 386-bp band, the homozygote (with the neo/forward and exon3/reverse primer set) yielded a single 600-bp band, whereas the heterozygote yielded both the 386-bp and the 600-bp band of the WT and mutant strand, respectively. The body phenotype of heterozygous Epha4 null mice was not different from that of WT littermates (Table 1), but only WT littermates (albeit not necessarily from the same litter) were used as control mice in this study.
|Parameters||WT littermates (n = 17)||HT Epha4 null mice (n = 15)||HM Epha4 null mice (n = 7–8)|
|Body weight (g)||25.82 ± 0.78||25.59 ± 0.77||18.14 ± 0.67*,***|
|Femur length (mm)||15.12 ± 0.17||15.16 ± 0.12||13.99 ± 0.14*,***|
|Bone area (mm2)||2.35 ± 0.09||2.23 ± 0.08||1.68 ± 0.07*,***|
|Cortical thickness (mm)||0.24 ± 0.00||0.25 ± 0.01||0.21 ± 0.01*,***|
|Periosteal circumference (mm)||4.83 ± 0.09||4.75 ± 0.10||4.17 ± 0.06*,***|
|Endosteal circumference (mm)||3.24 ± 0.07||3.19 ± 0.08||2.80 ± 0.04*,***|
|Total BMC (g)||1.79 ± 0.07||1.71 ± 0.07||1.25 ± 0.06*,***|
|Cortical BMC (mg)||1.31 ± 0.04||1.35 ± 0.06||1.00 ± 0.05*,***|
|Trabecular BMC (mg)||0.57 ± 0.05||0.50 ± 0.06||0.30 ± 0.03*,***|
|Total BMD (mg/mm3)||763.10 ± 7.22||767.27 ± 0.88||748.19 ± 24.51|
|Cortical BMD (mg/mm3)||1153.3 ± 11.9||1181.6 ± 9.6||1099.0 ± 19.8*,***|
|Trabecular BMD (mg/mm3)||261.63 ± 18.88||237.67 ± 21.62||189.83 ± 9.97**,****|
RAW/C4 cells were maintained in alpha modified Eagle's medium (α-MEM) supplemented with 10% FBS. For osteoclastic differentiation, RAW/C4 cells were plated in α-MEM containing 10% FBS, and sRANKL (at 200 ng/mL) was added to initiate osteoclastic differentiation into tartrate-resistant acid phosphatase (TRACP)-positive, multinucleated (two or more nuclei) “osteoclast-like” cells (OCLs). At day 3, the culture medium was replaced with a fresh medium containing only 66 ng/mL sRANKL. For generation of primary marrow-derived osteoclasts, bone marrow cells were flushed out of long bones of 10-week-old to 12-week-old Epha4 null mice or WT littermates. The bone marrow cells were cultured in α-MEM containing 10% FBS for 24 hours. Nonadherent marrow cells containing osteoclast precursors were collected and plated at a density of 6 × 105 cells/cm2 (in 24-well plates) or 4 × 106 cells/cm2 (in six-well plates) in α-MEM with 10% FBS in the presence of sRANKL (66 ng/mL) and m-CSF (50 ng/mL). Culture medium was changed at day 3 with fresh α-MEM containing 10% FBS and sRANKL/m-CSF. After 3 or 6 days, the total number of OCLs in 24-well culture wells was counted. Endpoint measurements were performed on day 6 of the sRANKL treatment unless otherwise indicated.
To count the number of nuclei per osteoclasts, marrow osteoclasts precursors were first treated with sRANKL/m-CSF for 5 days to induce formation of osteoclasts. At the end of the treatment, cells were formalin-fixed, and stained for 3 minutes with the CAT hematoxylin stain (BioCare Medical, Concord, CA, USA). The number of nuclei of >200 randomly selected osteoclasts per culture well was counted. The average number of nuclei per osteoclasts and the distribution profile of osteoclasts with different numbers of nuclei were obtained.
Mineralized nodule formation assay
Briefly, bone marrow cells of 10-week-old to 12-week-old Epha4 null mice or WT littermates were flushed out of long bones and were cultured in α-MEM containing 10% FBS for 24 hours. The nonadherent cells were removed, and the adherent cells were further cultured for 48 to 96 hours until confluent. The attached bone marrow stromal cells were isolated by trypsin and cells at passage 2 were plated at 1 × 105 cells per well in six-well plates and incubated with α-MEM containing 10% FBS, 10 mM of β-glycerophosphate, and 50 µg/mL of ascorbic acid for 7 days. The number of mineralized nodules were stained with Alizarin red and the number of mineralized nodules per well were counted.
EphA4 small-interfering RNA transfection
A set of four EphA4 small-interfering RNAs (siRNAs) (FlexiTube Gene Solution GS13838), a control siRNA (FlexiTube Gene Solution S103650325), and the HiPerfect HTS transfection reagent used for siRNA transfection were purchased from Qiagen. The transfection reaction was carried out overnight in 24-well cultures of RAW/C4 cells at the cell density of 1 × 104 cells/cm2 in αMEM supplemented with penicillin/streptomycin (100 µg/mL each) and 10% FBS. Each test siRNA (5 nM) and 200 ng/mL of sRANKL was added to RAW/C4 cells 1 day after plating. After 3 days, the medium was replaced with fresh medium containing 66 ng/mL sRANKL but without siRNA. A large number of multinucleated, TRACP-positive OCLs began to form on day 5. Although siRNA may last only for ∼24 hours in culture media, this experimental strategy was effective in suppressing Epha4 expression even after 6 days of culture (Fig. 2A). For primary bone marrow-derived osteoclasts, the same procedure was used, except that the cells were plated at a density of 1.2 × 106 cells/cm2 and that 66 ng/mL of sRANKL and 50 ng/mL of m-CSF were used to stimulate osteoclast differentiation.
Resorption pit formation assay
The resorption pit formation assay was performed as described. The sizes of individual pits were measured with the OsteoMeasure system. The average pit area per pit was reported as an index of the average bone resorption activity per osteoclast.
Quantative polymerase chain reaction assays
Quantative polymerase chain reaction (qPCR) was carried out with the SYBR Green method on the MJ Research DNA Engine Opticon 2 System (Waltham, MA, USA). Total RNA was extracted with the Qiagen RNeasy kit. The purified RNA was used as template for synthesizing cDNA by reverse transcription using oligo dT(20) primers and ThermoScript RT-PCR kit (Invitrogen, Carlsbad, CA, USA). An aliquot of the cDNA was subjected to qPCR amplification using gene-specific primer sets listed in Supplemental Table 1. The reaction mixture (25 µL) in each assay consisted of 12.5 µL of 2× GoTag qPCR master mix (Promega, Madison, WI, USA), which contained the Hot Start Taq polymerase, 3 µM of primers, and 3 µL of cDNA template. The PCR amplification condition consisted of an initial 10-minute hot start at 95°C, followed by 40 cycles of denaturation at 95°C for 30 seconds, annealing and extension for 30 seconds at an appropriate temperature (see Supplemental Table 1), and a final step of melting curve analysis from 60°C to 72°C. The data (normalized against cyclophilin [Ppia] mRNA) were analyzed with Opticon Monitor Software 2.0. Data and the relative fold change was calculated by the threshold cycle (ΔCT) method.
Western immunoblot assays
Relative levels of EphA4 in osteoclastic cells were determined by Western immunoblots with an anti-EphA4 polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA), and quantified with the FluorChem Q imaging system (Proteinsimple, Santa Clara, CA, USA). Antibodies against other signaling proteins were obtained from Santa Cruz Biotechnology, BioSource International (Camarillo, CA, USA), UpState Biotechnology (Lake Placid, NY, USA), R&D Systems (Minneapolis, MN, USA), or BD Transduction Laboratories (San Diego, CA, USA).
pY-levels of EfnB2, EfnA2, or EfnA4
Phosphorylation of key intracellular tyrosines is one of the earliest event of an activation of the reverse signaling of Efn ligands. Thus, we measured the relative pY-EfnB2, pY-EfnA2, pY-EfnA4 levels as an index of activation of the reverse signaling of respective Efn ligand. The pY316-EfnB2 level was analyzed by the Western immunoblot assay as described above. The relative levels pY-EfnA2 and pY-EfnA4 were determined by the immunoprecipitation/immunoblotting approach described as follows: Briefly, RAW/C4 cells and primary marrow-derived osteoclast precursors in six-well plates were treated with 200 ng/mL of sRANKL or 66 ng/mL of sRANKL along with 50 ng/mL of m-CSF, respectively, for 6 days as described in the Cell cultures section. At the end of the sRANKL treatment, cell layer proteins were extracted with 0.65 mL of radioimmunoprecipitation assay (RIPA) buffer containing various protease inhibitors and 2 mM sodium orthovanadate for an hour. Total cellular pY-containing proteins were first isolated from pooled cell lysates by immunoprecipitation using an anti-pY (4G10) antibody. Because large amounts of proteins are required for immunoprecipitation, cellular proteins from three separate wells of RAW/C4-derived OCLs or osteoclasts derived from three Epha4 null or WT mice were pooled. A total of 720 µg of the pooled cell extract proteins were incubated overnight with 4 µg of anti-pY (4G10) antibody. The 50% slurry of Protein A/G beads in RIPA (50 µL) was added to each immunoprecipitation mixture and incubated for an additional hour. The pY-proteins were collected by centrifugation, washed three times with RIPA, resuspended in the SDS-PAGE loading buffer, and boiled for 5 minutes. The immunoprecipitated pY proteins were then separated on 10% SDS-PAGE. The pY-EfnA2 and pY-EfnA4 were identified by Western blots using an anti-EfnA2 or an anti-EfnA4 antibody, respectively.
Peripheral quantitative computed tomography bone parameter measurements
Peripheral quantitative computed tomography (pQCT) scanning was performed on the femur of 10-week-old mice as described, using a Stratec XCT 960 M pQCT (Norland Medical Systems, Madison, WI, USA). Briefly, femurs were fixed in 10% formalin for 24 hours and stored in ice-cold phosphate-buffered saline (PBS) supplemented with 0.1% NaN3 until measurement. The length of each femur was determined with a digital caliper. The distance between individual scanning slices was calculated by the expression, (femur length × 9)/100. The entire femur was scanned to generate values for nine consecutive slices. The derived values of slice 2 or slice 5 represented measurements at the primary/secondary spongiosa or midshaft, respectively. Trabecular bone mineral content (BMC) and bone mineral density (BMD) were determined with the threshold setting of 230 to 630 mg/cm2. A threshold setting of 630 mg/cm2 was used to determine cortical bone parameters.
Micro–computed tomography bone parameter measurements
The bone phenotype was also assessed on the right femur by micro–computed tomography (µCT) using a Scanco vivaCT40 µ-CT scanner (Scanco Medical, Brüttisellen Switzerland) as described. Trabecular measurements were performed at the secondary spongiosa of distal femur (at a site that was 10% of the full length of the femur from the distal end). Accordingly, a region of 0.8 mm in thickness at 10% of the full length from the distal end of each femur was scanned. The trabecular masks were defined in a semiautomatic manner, starting from the outer mask of the femur and application of 15 erosion cycles to ensure that no cortex was included in the measurement. This approach would adjust for the 7.5% shorter femur in Epha4 null mutants. µCT measurements were also performed in the trabecular bone of the L5 vertebra.
Bone mass, and static or dynamic bone histomorphometric parameters were measured at the secondary spongiosa of the tibia as described. Specifically, bone histomorphometry was measured at a site that was 0.25 mm distal to the growth plate of the distal femur of WT mice. Because the femur length of homozygous Epha4 null mice was 7.5% shorter (13.99 versus 15.12 mm), bone histomorphometry in Epha4 null mice was measured at a site that was 0.23 mm distal to the growth plate of the distal femur to adjust for the difference in the femur length.
We were also interested in comparing the in vivo size of the mature osteoclasts on the bone surface of Epha4 null mice with that of WT littermates. Although it is feasible to trace the circumference of bone surface osteoclasts on the slide, such measurements of osteoclast size are misleading, because (1) bone histomorphometry measurements are two-dimensional, and the in vivo osteoclast size is three-dimensional, and (2) these measurements are dependent of and highly influenced by the orientation of the sectioning of the bone. Thus, we took an alternative approach to estimate the relative size of osteoclasts on the bone surface by dividing the total length of the osteoclast-occupied bone surface (Oc.Pm) by the total number of osteoclasts along the bone surface (N.Oc). This approach assumes that the average length of bone surface covered by each osteoclast is proportional to the average size of osteoclasts on the bone surface. We are aware that this approach can also be affected by the orientation of the bone sectioning, we feel that the influence of orientation on bone surface (which is two-dimensional) is less than that on direct measurements of cell size (which is three-dimensional).
Serum bone turnover biomarker assays
The serum bone resorption biomarker, c-telopeptide of type I collagen (CTx), was measured with a RatLaps EIA kit (Immunodiagnostic Systems [IDS], Inc., Fountain Hills, AZ, USA). The serum bone formation biomarker (pro-collagen I N-terminal peptide [P1NP]) was assayed with a mouse serum P1NP ELISA kit, also purchased from IDS.
Results are shown as mean ± SEM. Statistical significance was analyzed with two-tailed Student's t test. Statistical significance of studies with two or more variables was analyzed with two-way ANOVA using the Systat11 software (Systat Inc., Chicago, IL, USA). The difference was considered significant when p < 0.05.
Expression of EphA4 receptors and EfnA ligands in OCLs
OCLs derived from the RAW/C4 clone of RAW264.7 cells were used as the initial cell model system. This RAW/C4 clone was isolated through limiting dilution and screened for their ability to undergo sRANKL-induced osteoclastic differentiation, which showed threefold greater efficiency in sRANKL-induced differentiation into OCLs compared to the parental RAW264.7 cells. There were no apparent differences between RAW/C4 cells and the parental RAW264.7 cells in expression of osteoclastic genes, ability to resorb bone in vitro, and responsiveness to resorptive effectors. To confirm that OCLs derived from sRANKL-treated RAW/C4 cells express EphA4, we performed RT-PCR and Western immunoblot analyses for Epha4 mRNA and protein expression, respectively, in the derived OCLs. As in primary mouse osteoclast precursors, RAW/C4 cells expressed substantial levels of Epha2 mRNA that gradually decreased to a low level at the end of the 6-day sRANKL treatment (Fig. 1A), at which time a large number of multinucleated, TRACP-positive OCLs were formed. Conversely, Epha4 mRNA was very low in undifferentiated RAW/C4 cells, but the derived OCLs after 6 days of sRANKL treatment exhibited considerable levels of Epha4 mRNA. Neither undifferentiated nor differentiated RAW/C4 cells expressed Epha3 mRNA. Western immunoblots confirmed that primary mouse marrow-derived osteoclasts and RAW/C4 cell-derived OCLs (Fig. 1B) each expressed substantial amounts of EphA4 protein and that the relative amount of cellular EphA4 was significantly increased after 6 days of sRANKL treatment when a large majority of the cells were differentiated into osteoclasts or OCLs (Fig. 1B). However, the EphA4 protein level in osteoclasts appears to be less than that in osteoblastic cells, such as MC3T3-E1 cells.
EphA4 binds EfnB2, EfnB3, and all members of the EfnA family. Previous reports indicate that primary murine osteoclasts express Efnb2 and Efna2.[17, 19] The derived OCLs from RAW/C4 cells after 6 days of sRANKL treatment expressed ample amounts of mRNA for Efna4 and Efna2, but not Efna3 (Fig. 1A). To ascertain that primary osteoclasts express one or more ligands for EphA4, we measured the relative mRNA levels of various members of the EfnA and EfnB families by qRT-PCR and the results are reported as ΔCT (ie, CT[Efn] − CT[Ppia]). We found that mouse marrow-derived osteoclasts expressed substantial amounts of Efna1 (ΔCT, 3.35 ± 0.20, n = 9), Efna2 (4.74 ± 0.47, n = 7), Efna4 (5.97 ± 0.72, n = 8), and Efnb1 (3.60 ± 0.68, n = 8). These cells also expressed Efnb2 (7.12 ± 0.77, n = 9) and Efnb4 (8.62 ± 1.10, n = 4), mRNAs, but their levels were fourfold to eightfold less than Efna4 mRNA (ie, a two-cycle to three-cycle difference in ΔCT than Efna4).
Effects of suppression of Epha4 expression on the size and functional activity of OCLs
To evaluate the functional role of EphA4 in osteoclasts, we assessed the effects of knocking down Epha4 expression with a set of four specific siRNAs on differentiation and in vitro bone resorption activity of the derived OCLs. Each siRNA alone (at 5 nM) or all together (each at 1.25 nM) effectively suppressed Epha4 mRNA expression by 60% to 80% (Fig. 2A). For most subsequent studies, we used a mixture of all four siRNAs (each at 1.25 nM) to reduce potential off-target effects. Knocking down Epha4 expression yielded OCLs that were twice as large as those derived from control siRNA-treated cells (Fig. 2B), but it did not alter the total number of multinucleated, TRACP-positive OCLs formed per culture well after either 3 or 6 days of the sRANKL stimulation (Fig. 2C). Treatment of RAW/C4 cells with 10 ng/mL of soluble EfnA4-fc chimeric protein (to activate the EphA4 signaling) produced OCLs that were ∼50% smaller than control OCLs (Fig. 2D). OSCAR and OC-STAMP are recognized as markers of fusion of osteoclast precursors.[33, 34] Thus, we compared the relative levels of Oscar and OC-Stamp mRNAs between Epha4 siRNA-treated and control siRNA-treated cells to assess potential effects of Epha4 deficiency on fusion of osteoclast precursors. Figure 2E shows that the Oscar mRNA expression level in the control siRNA-treated cells was increased 92-fold and 255-fold of the basal level after 3 and 6 days of the sRANKL treatment, respectively. The Epha4 siRNA treatment did not affect the sRANKL-induced upregulation of Oscar expression after 3 days of the sRANKL treatment, but significantly (p < 0.05) increased (by 75%) the Oscar mRNA level at day 6. However, the siRNA-mediated suppression of Epha4 expression had no significant effect on the cellular levels of OC-Stamp mRNA after 3 or 6 days of the sRANKL treatment (Fig. 2F).
To determine if suppression of Epha4 expression would alter the functional activity of the derived OCLs, we measured the mRNA level of a marker gene of osteoclast activity (cathepsin K [Ctsk]) and the in vitro bone resorption activity of derived OCLs of Epha4 siRNA-treated cells. The Epha4 siRNA treatment upregulated Ctsk expression by twofold to threefold (Fig. 2G) and enhanced the in vitro bone resorption activity by ∼50% compared the control siRNA-treated cells (Fig. 2H). In contrast, an activation of the EphA4 forward signaling with EfnA4-fc fusion protein in primary osteoclasts suppressed the in vitro bone resorption activity by ∼50%.
Effects of deficient Epha4 expression on the bone in vivo
To evaluate functional role of EphA4 in the bone in vivo, we compared the bone phenotype of Epha4 null mice with that of WT littermates. Table 1 compares the body weight and pQCT bone parameters in femurs of 10-week-old male homozygous Epha4 null mice with those of age-matched and sex-matched heterozygotes or with those of WT littermates. Homozygous Epha4 null mice showed significant reductions in body weight (by 30%), femur length (by 7%), bone area (by 29%), cortical thickness (by 13%), periosteal (by 14%) and endosteal (by 14%) circumferences, total (by 30%), cortical (by 24%), and trabecular (by 47%) BMCs, as well as cortical (by 5%) and trabecular (by 27%) BMDs, when compared to WT littermates. Because the measurement site was adjusted for the difference in femur length, the observed reductions in bone mass or density were unlikely due to artifacts of more diaphysis being analyzed in the shorter Epha4 null femur. There were no significant differences in any of the test parameters between heterozygotes and WT mice. Thus, subsequent studies compared only homozygous null mice with corresponding WT littermates.
The µCT analysis at secondary spongiosa of the distal femur (Fig. 3A) confirms that Epha4 null mutants had lower tissue volume (TV), bone volume (BV), and trabecular bone volume per tissue volume (BV/TV). They also had 28% lower connectivity density (Conn-Den), 28% lower trabecular number (Tb.N), 17% reduction in trabecular thickness (Tb.Th), and 50% increase in trabecular separation (Tb.Sp). The µCT measurement of cortical bone at mid-shaft reveals that cortical TV and BV of the Epha4 null mutants were each less than those of the WT littermates by 28%. Because BV and TV were reduced to the similar extent, the cortical BV/TV ratio was not different between the mutant mice and WT mice.
The µCT evaluation of the L5 vertebra of Epha4 null mice with that of WT littermates shows that the null mice had 27% and 31% less TV and BV, respectively, but not BV/TV (Fig. 3B). Tb.N and Tb.Th of Epha4 null mutants were also reduced by 10% and 16%, respectively, which was associated with 13% increase in Tb.Sp. The reductions in Tb.N and Tb.Th, and the increase in Tb.Sp raise the possibility that the trabecular bone loss in Epha4 null mice could be due to an increase in bone resorption.
Effects of deficient Epha4 expression on bone turnover in vivo
Histomorphometric analysis of trabecular bone at secondary spongiosa of distal femur of 8-week-old male Epha4 null mice and corresponding WT littermates confirms that Epha4 null mutants have a marked decrease in BV/TV (by −65%; Table 2). Increased bone turnover would increase the total active bone surface due to increased number of resorption lacunae. Thus, the large increase in the bone surface per bone volume (BS/BV, by +47%) suggests an increase in bone turnover. Consistent with the µCT results, the mutant mice had reduced Tb.Th (by −47%) and Tb.N (by −34%), and an increased Tb.Sp (by +67%). TRACP.Pm (TRACP stained bone surface perimeter) and osteoclast bone surface perimeter (OC/B.Pm) were not different significantly between the mutants and WT mice, suggesting that deficient Epha4 expression probably had only minimal effects on the number of TRACP-expressing osteoclasts and on active resorbing bone surface. The average bone surface length occupied by an osteoclast (osteoclast surface [Oc.Pm]/number of osteoclasts [N.Oc]) (which we used as an index for the relative osteoclast size on the resorbing bone surface) was also larger (by 14.5%) in 10-week-old male Epha4 null mice than corresponding WT littermates, albeit it did not reach the statistically significant level (Table 2). This finding suggests that the average size of Epha4 null osteoclasts on the bone surface may also be larger than that of WT osteoclasts in vivo. Dynamic bone formation histomorphometric parameters at secondary spongiosa of distal femurs of female young adult Epha4 null mice were each reduced slightly, but not significantly, than those of WT littermates (Table 3).
|Genotype||BV/TV (%)||TRAP.Pm (%)||Oc/B.Pm (#/mm)||Oc.Pm/N.Oc (µm)||BS/BV (mm2/mm3)||Tb.Th (µm)||Tb.N (#/mm)||Tb.Sp (µm)|
|WT||13.13 ± 1.41||24.36 ± 1.85||7.88 ± 0.66||28.9 ± 1.3||50.31 ± 4.09||41.49 ± 3.15||3.16 ± 0.23||286.9 ± 23.3|
|Epha4 null||5.40 ± 0.74||21.73 ± 2.02||7.70 ± 0.82||33.1 ± 2.7||87.83 ± 6.95||23.47 ± 1.54||2.24 ± 0.24||463.5 ± 69.6|
|Difference from WT||−58.9%||−10.8%||−2.3%||+14.5%||+74.6%||−43.4%||−29.1%||+61.6%|
|Genotype||TLS (mm)||TLS/BS (mm/mm)||MAR (µm/d)||BFR (mm2 × 10−3/d)||BFR/BS (mm2 × 10−3/d/mm2)|
|WT (n = 5)||1.18 ± 0.12||0.39 ± 0.04||1.98 ± 0.22||2.42 ± 0.18||0.77 ± 0.13|
|Epha4 null (n = 4)||1.18 ± 0.17||0.32 ± 0.02||1.60 ± 0.34||1.98 ± 0.64||0.52 ± 0.14|
|Difference from WT, %||0||−18||−19||−18||−32|
To confirm the effects of Epha4 deficiency on bone turnover, we measured serum biomarkers of bone formation (ie, P1NP) and resorption (ie, CTx) in 10-week-old male Epha4 null mice and age-matched male WT littermates. The serum CTx level of mutant mice was significantly greater (by ∼50%) than that of WT littermates (Fig. 4A). However, the serum P1NP level in mutant mice (Fig. 4B) was only slightly (∼10%), but not significantly, lower than that of WT littermates.
To assess whether Epha4 deficiency would have an effect on bone-forming capability of cells of osteoblastic lineage in vitro, we compared the in vitro mineralized bone nodule formation ability of marrow-derived stromal cells (Fig. 5A) of Epha4 null mice with those of WT mice. The number of mineralized bone nodules formed by Epha4-deficient marrow stromal cells in vitro was significantly less than that by corresponding WT marrow stromal cells, suggesting that deficient Epha4 expression may negatively affect the bone forming capability of stromal cell-derived osteoblasts. On the other hand, we cannot rule out the possibility that this could be an “ancillary” feedback downregulation of the bone forming capability of stromal cell-derived osteoblasts due to long-term Epha4 deficiency in Epha4 null mutant mice. Future work is needed to address this issue. In any event, because Epha4 null mice did not show significant reduction in any histomorphometric bone formation parameters (Table 3) and in the serum bone formation biomarker level (Fig. 4B), it is likely the in vivo direct effect of Epha4 deficiency on bone formation, if any, would probably be small and may not be physiologically significant.
Effects of Epha4 deficiency on the cell size, fusion, survival, and functional activity of osteoclasts
Marrow-derived osteoclasts of Epha4 null mice were twice as large as those of WT littermates in vitro (Fig. 5B). Treatment of WT osteoclast precursors with the soluble EphA4-fc chimeric protein (to block activation of the EphA4 signaling) yielded osteoclasts with cell size almost twice as large as those untreated control osteoclasts. As expected, the EphA4-fc treatment had little effect on the size of osteoclasts derived from marrow precursors of Epha4 null mice (Fig. 5B). Treatment of WT osteoclast precursors with EfnA4-fc chimeric protein (to activate the EphA4 forward signaling) reduced the relative size of the derived osteoclasts by >50% (Fig. 5C). We then determined the relative mRNA levels of several genes that are associated with osteoclast differentiation (ie, Acp5, Fos, Mitf, Nfatc1, and Rank), fusion (Oscar and OC-Stamp), or activity (ie, ATP6v0d2, Calcr, Clcn7, Mmp3, Mmp9, Itgav, and Itgb3) in osteoclasts of Epha4 null mice and WT littermates. Table 4 shows that disrupting Epha4 expression in osteoclasts significantly upregulated the mRNA levels of Mmp3 (by 740%, p < 0.005), Mmp9 (by 88%, p < 0.05), and Acp5 (by 82%, p < 0.05). The mRNA levels of the other test osteoclastic genes were not significantly different between Epha4 null and WT osteoclasts. We next compared the average size of the resorption pits formed by Epha4 null osteoclasts with that by WT osteoclasts in an in vitro resorption pit formation assay. Epha4 null osteoclasts produced resorption pits with an average size twice as large as those by WT osteoclasts (Fig. 5D).
|mRNA||ΔCT, WT||ΔCT, Epha4 null||Epha4 null/WT (fold)a||p*|
|ATP6v0d2||−2.88 ± 0.21 (n = 11)||−2.99 ± 0.32 (n = 11)||1.08||NS|
|Calcr||2.30 ± 0.23 (n = 10)||2.25 ± 0.34 (n = 9)||1.04||NS|
|Clcn7||0.081 ± 0.57 (n = 11)||−0.37 ± 0.52 (n = 9)||1.37||NS|
|Mmp3||9.68 ± 0.37 (n = 8)||6.61 ± 0.83 (n = 9)||8.40||0.004|
|Mmp9||−6.67 ± 0.15 (n = 10)||−7.58 ± 0.36 (n = 9)||1.88||0.024|
|Itgav||0.72 ± 2.11 (n = 11)||−0.05 ± 0.35 (n = 8)||1.71||NS|
|Itgb3||3.89 ± 0.23 (n = 9)||4.72 ± 0.36 (n = 10)||0.56||NS|
|Acp5||−6.05 ± 0.22 (n = 10)||−6.91 ± 0.38 (n = 8)||1.82||0.05|
|Oscar||−1.78 ± 0.20 (n = 3)||−2.63 ± 0.28 (n = 3)||1.84||NS|
|OC-Stamp||−2.30 ± 0.24 (n = 4)||−1.68 ± 0.28 (n = 4)||0.65||NS|
|Fos||−0.32 ± 0.25 (n = 11)||−0.35 ± 0.42 (n = 11)||1.02||NS|
|Mitf||2.94 ± 0.41 (n = 10)||2.50 ± 0.53 (n = 10)||1.36||NS|
|Nfatc1||−1.26 ± 0.24 (n = 10)||−1.60 ± 0.51 (n = 10)||1.27||NS|
|Rank||0.61 ± 0.29 (n = 10)||−0.12 ± 0.48 (n = 10)||1.66||NS|
To evaluate whether increased cell fusion could contribute to the larger cell size of Epha4 null osteoclasts, we assessed the time-dependent effect of Epha4 deficiency on the expression level of Oscar and OC-Stamp mRNAs. Figure 6A shows that the mRNA level of Oscar in Epha4 null osteoclasts after 6 days of the sRANKL/m-CSF treatment was upregulated (by 84%, n = 3 for each genotype, p = 0.067) compared to WT osteoclasts. However, the Oscar mRNA level in the Epha4 null and the WT osteoclast progenitor cells at day 0 (basal level) or at day 3 was not different. On the other hand, whereas sRANKL induced the time-dependent upregulation of OC-Stamp mRNA in both WT and Epha4 null osteoclasts, the expression level of OC-Stamp mRNA in the Epha4-deficient osteoclasts was ∼40% lower than that in WT osteoclasts at each time point (right panel of Fig. 6A), albeit the differences did not reach the statistically significant level (p > 0.05), as a result of the relatively large variation and small sample size (n = 4 per genotype). We also measured the average number of nuclei in osteoclasts derived from precursors of 3 Epha4 null mice and those of 3 WT littermates. The average number of nuclei per osteoclasts of Epha4 null osteoclasts was slightly (by ∼25%; 18.25 versus 14.53 nuclei/cell) but significantly (p < 0.03) more than that of WT osteoclasts (Fig. 6B). However, there was no significant difference (p > 0.05, two-way ANOVA) between the Epha4 null osteoclast population and the WT osteoclast population when the distribution profile of osteoclasts with different numbers of nuclei was compared (Fig. 6C).
If Epha4 deficiency increases fusion of osteoclast progenitors, the total number of multinucleated osteoclasts formed by Epha4 null osteoclast progenitor cultures in vitro should be significantly less than that formed by equal numbers of WT osteoclast progenitors. We found no significant difference in the total number of multinucleated TRACP-positive osteoclasts derived from the same numbers of Epha4 null and WT osteoclast progenitors in cultures after 3 or 6 days of the sRANKL/m-CSF treatment (Fig. 6D). Moreover, neither the EfnA4-fc-mediated activation nor the EphA4-fc-induced inhibition of the EphA4 forward signaling significantly altered the total number of osteoclasts formed from WT precursors after 6 days of sRANKL/m-CSF treatment. We also did not observe any significant differences in the relative apoptosis rate between WT and Epha4 null osteoclast precursors (Fig. 6E) or between WT and Epha4 null osteoclasts (Fig. 6F). Similarly, the EfnA4-fc or EphA4-fc treatment did not significantly affect the relative apoptosis of the derived osteoclasts (Fig. 6F).
Effects of Epha4 deficiency on the activation status of EphA4-binding Efn ligands in osteoclasts
The reverse signaling of at least two of the EphA4-binding Efn ligands (EfnB2 and EfnA2) has been shown to regulate osteoclast differentiation.[17, 19] To rule out the possibility that deficient Epha4 expression in Epha4 in osteoclastic cells may alter osteoclast activity (and bone resorption) in part through the negative regulation of the reverse signaling of these Efn ligands due to the lack of EphA4 binding to these ligands, we sought to determine the effect of deficient Epha4 expression on the steady state activation status of the reverse signaling of EfnB2, EfnA2, and EfnA4. (EfnA4 was included for comparison, because we used this ligand in this study.) Because increased phosphorylation of intracellular key tyrosines is one of the earliest steps in the activation process of the reverse signaling of these Efn ligands, we measured the pY-phosphorylation level of EfnB2, EfnA2, and EfnA4 in Epha4 null and WT osteoclasts and also in Epha4 siRNA-treated and control siRNA-treated RAW/C4-derived OCLs. There were no apparent differences in the pY316-EfnB2 level (Fig. 7A), or in the pY-EfnA2 or pY-EfnA4 levels (Fig. 7B) between Epha4 null and corresponding WT osteoclasts. There were also no differences between Epha4 siRNA-treated and control siRNA-treated RAW/C4 cell-derived OCLs. These findings suggest that deficient Epha4 expression in osteoclasts does not affect the activation status of the reverse signaling of these Efn ligands.
Regulation of the β3-integrin signaling pathway by Epha4 in osteoclasts
The Src and β3-integrin signaling are two well-known signaling mechanisms regulating the activity of osteoclasts.[35, 36] Activation of the β3-integrin signaling leads to increased phosphorylation and activation of Erk1/2. Conversely, the Src signaling is in a large part activated via the decreased pY527 phosphorylation. Thus, we compared the phosphorylation levels of Erk1/2 and pY527-Src in Epha4 null osteoclasts with WT osteoclasts. Figure 8A shows that there was no significant difference in the level of pY527-Src in Epha4 null osteoclasts when compared to that in WT osteoclasts. Neither the EphA4-fc nor the EfnA4-fc treatment altered the pY527-Src level in WT osteoclasts (Fig. 8B). The pErk1/2 level (normalized against total Erk1/2 level) was ∼50% higher in Epha4 null osteoclasts than in WT osteoclasts (Fig. 8C). Treatment with EphA4-fc increased pErk1/2 levels, and the EfnA4-fc treatment reduced pErk1/2 level in WT osteoclasts (Fig. 8D).
A well-known mechanism of β3-integrin to activate osteoclast activity is through increased phosphorylation and activation of Vav3, a member of the Rho family GTP exchange factors. Thus, we compared the pY173 level of Vav3 in Epha4 null osteoclasts with that in WT osteoclasts (Fig. 9A). The pY173-Vav3 level was greater (by >50%) in Epha4 null osteoclasts than in WT osteoclasts. We are mindful that activation of a number of cellular signaling pathways other than the integrin signaling occurs. These pathways include the signaling mechanism (eg, ITAM) of inflammatory cytokines and many growth factors also involve phosphorylation and activation of Erk1/2 and Vav3. The rationale for focusing on the β3-integrin signaling in this study was twofold: (1) there were an abundance of compelling evidence that the forward signaling of EphA4 in various cell types acts through the integrin signaling[38, 39]; and (2) increased phosphorylation and activity of Vav3 and Erk1/2 are key consequences of the β3-integrin signaling in osteoclasts. Nevertheless, the possibility that other signaling pathways may also be involved in the overall mechanism of EphA4 to regulate osteoclast activity cannot be overlooked.
Transmembrane signal transduction through β3-integrin can be conducted in either direction, in what is referred to as outside-in or inside-out signaling. The binding of pY binding (PTB) domain-containing proteins to key pY residues at the cytoplasmic tail of β3-integrin is an important step in bidirectional signaling and thereby controls the activation state of integrins. Because the pY759 phosphorylation level of β3-integrin is essential for the outside-in signaling and plays a physiological role in multiple Vav/Rho-mediated cell functions, including spreading, migration, and adhesion, we compared the pY759 level of β3-integrin in osteoclasts of Epha4 null osteoclasts with that in WT osteoclasts. Deficient Epha4 expression not only did not increase, but significantly reduced, the pY759 level of β3-integrin in osteoclasts (Fig. 9B), indicating that the EphA4-mediated negative regulation of osteoclast activity does not involve the outside-in signaling of β3-integrin.
Binding of talin to the cytoplasmic tail of β3-integrin is the critical final step of integrin β3 activation. The cytoplasmic signaling protein, Dok1, which is a potent negative regulator of integrin signaling, is shown to compete with talin for binding to the same site on β3-integrin. The binding affinity of β3-integrin for talin versus Dok1 is regulated by the pY747 phosphorylation status of β3-integrin. Upon pY747 phosphorylation, Dok1 binds much more strongly to β3-integrin than talin and as a result, the β3-integrin signaling is inhibited. Conversely, when pY747 is dephosphorylated, the binding affinity of β3-integrin for talin is increased much more than that for Dok1, leading to activation of the β3-integrin signaling. Figure 9C shows that the pY747 level in Epha4-deficient osteoclasts was at least fourfold lower than that in WT osteoclasts. Consistent with a regulatory role of EphA4 signaling on pY174-Vav3 and pY747-β3-integrin in osteoclasts, treatment of WT osteoclasts with EphA4-fc increased, and treatment with EfnA4-fc reduced the pY173-Vav3 level (Fig. 9D). Conversely, the EphA4-fc treatment reduced, and the EfnA4-fc treatment increased the level of pY747-β3-integrin (Fig. 9E). To confirm that the reduction in pY747 level in β3-integrin in response to Epha4 deficiency shifted the binding affinity of β3-integrin for talin from that for Dok1, we compared the relative amounts of coimmunoprecipitated talin versus Dok1 with β3-integrin between WT and Epha4 null osteoclasts (Fig. 9F). There was significantly more Dok1 than talin coimmunoprecipitated with β3-integrin in WT osteoclasts. In contrast, much more talin than Dok1 was coimmunoprecipitated with β3-integrin in Epha4 null osteoclasts.
This study provides compelling in vitro and in vivo evidence that EphA4 in osteoclasts is a potent negative regulator of osteoclastic resorption. Accordingly, disruption of the Epha4 gene in mice yielded a small bone size phenotype characterized by low bone mass and density associated with an increase in bone resorption. This conclusion is supported by three key findings: (1) there were marked reductions in Tb.Th and Tb.N along with a large increase in Tb.Sp, but with only marginal reductions in dynamic bone formation histomorphometric parameters in Epha4 null mice; (2) the serum bone resorption biomarker (CTx) level was increased by >50%, but without a significant change in the serum bone formation biomarker (P1NP) level in Epha4 null mice; and (3) osteoclasts derived from Epha4-deficient progenitors in vitro were larger, contained more nuclei, and were functionally more active in resorbing mineralized matrix than those derived from WT precursors.
Our conclusion that the suppressive action of EphA4 on osteoclastic resorption is mediated through the forward signaling of EphA4 is based on the following four lines of strong circumstantial evidence: (1) activation of the EphA4 forward signaling with EfnA4-fc reduced the average size and the in vitro bone resorption activity of both RAW264.7 cell-derived OCLs and primary marrow-derived osteoclasts; (2) blocking activation of the forward signaling of EphA4 with soluble EphA4-fc mimicked the deficiency of Epha4 expression and yielded larger and functionally more active osteoclasts; (3) the addition of soluble EphA4-fc to Epha4 null osteoclasts did not “rescue” the phenotype as it did not reverse the enhancing effects of Epha4 deficiency on the cell size and in vitro bone resorption activity of Epha4 null marrow-derived osteoclasts; and (4) Epha4 deficiency had no apparent effects on the activation status of the reverse signaling of its Efn ligands (eg, EfnB2, EfnA2, and EfnA4) in osteoclasts (as reflected by the lack of changes in the steady state pY phosphorylation level of these Efn ligands).
Three noteworthy observations may offer mechanistic insights into how EphA4 acts to negatively regulate osteoclastic resorption.
First, deficient Epha4 expression in osteoclast progenitors had no apparent effect on the total number of osteoclasts formed in response to sRANKL in vitro. Activation of the forward signaling of EphA4 with soluble EfnA4-fc, similarly, altered the average cell size and resorption activity, but not the number, of the derived osteoclasts. The number of derived osteoclasts is determined by formation and survival of osteoclasts. In this regard, Epha4 deficiency had no apparent effect on the survival of either osteoclast progenitors or mature osteoclasts. It follows that Epha4 deficiency may also not have a significant effect on formation of mature osteoclasts. The fact that deficient Epha4 expression in Epha4 null mice increased bone resorption (reflected by the increase in serum CTx level) and bone loss in vivo without changes in the number of active osteoclasts per bone surface perimeter (OC/B.Pm) or in the total bone resorption surface length (TRACP.Pm) indicates that the increase in bone resorption in the Epha4 null mice was due to an increase in osteoclast activity and not to an increase in osteoclast differentiation. Consistent with this premise is the finding that deficient Epha4 expression increased the expression of osteoclastic genes associated with osteoclastic activity, such as Mmp3, Mmp9, and Ctsk, but not those associated with differentiation, such as Fos, Mtif, Nfatc1, and Rank, all of which are downstream responsive genes of the RANKL signaling in osteoclasts. Thus, it suggests that EphA4 is probably not involved in the RANKL-induced osteoclast differentiation. Accordingly, we conclude that unlike the reverse signaling of EfnB1, EfnB2, or EfnA2, which acts primarily through modulation of the RANKL-induced osteoclast differentiation, the forward signaling of EphA4 regulates mainly the functional activity of mature osteoclasts and does not affect the differentiation of osteoclast progenitor cells. That mature osteoclasts and not their precursors expressed high levels of EphA4 is consistent with the assumption that EphA4 is a regulator of mature osteoclasts and not of their precursors.
Second, the larger average cell size of Epha4-deficient osteoclasts, which can be a consequence of an increase in cell fusion and/or increased cell spreading/migration. On the basis of the following three observations, we conclude that an increased cell fusion of osteoclast progenitors may not be a major contributing factor to the larger osteoclasts in Epha4 null mice. (1) If the larger cell size were caused by an increased cell fusion, the total number of osteoclasts formed from the same number of precursors in vitro should be reduced significantly in Epha4 null cells, because two or more smaller osteoclasts are needed to be fused together to form a single larger osteoclast. However, we noted no significant differences in the total number of osteoclasts formed from the same number of Epha4 null osteoclast progenitors and those from WT progenitors in response to sRANKL. Thus, this finding is incompatible with the possibility of an increased cell fusion in Epha4-deficient osteoclast precursors. (2) An increase in cell fusion is expected to result in an increase in the average number of nuclei in Epha4 null osteoclasts. Although there was an approximately 25% increase in the average number of nuclei in Epha4 null osteoclasts, this increase is probably too small to account for the twofold increase in the cell size of Epha4 null osteoclasts. The lack of a significant difference in the distribution profile of osteoclasts with various numbers of nuclei between Epha4 null osteoclasts and WT osteoclasts also does not support the contention that an increased cell fusion is a key contributor to the larger Epha4 null osteoclasts. (3) Although Epha4 deficiency increased Oscar mRNA level at day 6 but not earlier, deficient-Epha4 expression in osteoclasts had no significant increase in OC-Stamp (another marker gene for osteoclast fusion). The mechanistic reason for the apparent discrepancy in the effect of Epha4 deficiency in Oscar and OC-Stamp mRNA levels in osteoclasts remains to be determined. However, although there is recent evidence indicating that OC-STAMP may involve in osteoclast fusion,[34, 46] evidence for OSCAR, an immunoglobulin G (IgG)-like receptor, involved in osteoclast fusion is less compelling. Much work is needed to sort out the apparently contradictory effects of EphA4 on Oscar versus OC-Stamp expression. Nevertheless, these three findings, together, suggest that the larger cell size of Epha4-deficient osteoclasts is probably due in a large part to an increase in the spreading and/or migration of osteoclasts rather than to an increased cell fusion. In this regard, a recent study with lung adenocarcinoma cells has concluded that EphA4 is an inhibitor of cell spreading, migration, and invasion. It has also previously been reported that the EfnA5-mediated EphA4 signaling functions as a key inhibitor of muscle precursor cell migration during development.
Third, the osteoclastic resorption process is complex and consists of a number of key processes that include, but are not limited to, formation and maintenance of the sealing barrier needed for formation of a microenvironment between osteoclast and the underlying bone matrix; a process that involves the integrin signaling-dependent cytoskeletal reorganization, activation of the H+-ATPase protein pump and the ClCN7-mediated chloride channel, and production and secretion of lysosomal degradative enzymes, ie, Acp5, Ctsk, Mmp3, Mmp9, and related extracellular matrix proteases into the sealing zone to degrade matrix proteins and dissolve matrix minerals. Our finding that deficient Epha4 expression upregulated only a subgroup of genes that are involved primarily in bone matrix degradation (ie, Mmp3, Mmp9, Acp5, and Ctsk) and not other genes that are associated with osteoclast activity (ie, ATP6v0d2, Calcr, Clcn7, Mmp3, Mmp9, Itgav, and Itgb3) suggests that the regulatory effect of EphA4 on osteoclast activity may be restricted to production and secretion of lysosomal degradative enzymes by mature osteoclasts. However, we should note that many processes involved in the osteoclastic resorption, such as the integrin signaling, are regulated by posttranslational modification, such as phosphorylation. Because EphA4 is a PTK, we cannot rule out the likely possibility that the EphA4 signaling may also regulate other essential osteoclastic resorption processes, such as the integrin signaling, through protein phosphorylation rather than gene expression.
β3-Integrin signaling is a well known mediator of osteoclast spreading, migration, and activity,[36, 52] which acts through a canonical signaling complex to promote cell spreading and formation of actin rings. A critically important mediator of the β3-integrin signaling in osteoclasts is the Vav3, which is a Rho family GTP exchange factor (GEF). Activated Vav3 would activate Rac and subsequently Erk1/2 to enhance bone resorption activity of the osteoclast. The β3-integrin signaling in osteoclasts activates Vav3 through an increase in phosphorylation of its tyr-173 residue. Thus, our findings that Epha4 null osteoclasts had a significant elevated level of pY173-Vav3, and that inhibition of the EphA4 signaling with soluble EphA4-fc fusion protein significantly increased, whereas activation of the EphA4 signaling with soluble EfnA4-fc protein reduced, pY174-Vav3 level in WT osteoclasts, were consistent with the potential involvement of an EphA4-dependent modulation of the β3-integrin signaling. The β3-integrin signaling regulates not only cell functions related to cytoskeletal reorganization but also the release of degradative enzymes needed for the bone resorption process. That deficient Epha4 in osteoclasts also upregulated expression of Mmp3, Mmp9, and Ctsk is consistent with the premise that the suppressive action of EphA4 on osteoclast activity is mediated through negative modulation of the β3-integrin signaling. Previous studies in other cell types indicate that one of the major molecular mechanisms of EphA receptors (including EphA4) to regulate cell differentiation, adhesion, and migration involves differential regulation of the activities of various specific members of Rho family GEFs, such as ephexin in the brain[56, 57] and Vsm-RhoGEF in vascular smooth muscle, or a RhoGAP, such as SPAR in axon.[56, 59] That deficient Epha4 expression in osteoclasts increased pY173-Vav3 levels supports the assumption that EphA4 also regulates osteoclastic activity in part through suppression of the β3-integrin-mediated activation of Vav3 in osteoclasts.
Two findings of this study provide important clues to the molecular mechanism by which EphA4 suppresses the β3-integrin signaling in osteoclasts.
The first finding is that deficient Epha4 expression in osteoclasts significantly reduced the pY levels of two key tyrosines (i.e., pY747 and pY759) at the C-terminal tail of β3-integrin that are essential for the outside-in signaling of β3-integrin. The reduction in phosphorylation levels of these two pY residues in Epha4 null osteoclasts was EphA4-dependent, since the EphA4-fc treatment reduced, while the EfnA4-fc treatment increased, the pY747-β3-integrin level. There are two important implications: (1) it implies that the suppressive effect of EphA4 on osteoclastic activity is not mediated through an EphA4-dependent negative modulation of the outside-in signaling; and (2) these two tyrosine residues of β3-integrin could be potential cellular substrates for the EphA4 PTK. In any event, our results clearly indicate that EphA4 regulates the β3-integrin signaling by a mechanism that is independent of phosphorylation of these two tyrosines.
The second finding is the shift in the binding affinity of β3-integrin from one that favors Dok1 to one that favors talin in Epha4 null osteoclasts. Binding of talin to the PTB domain of the cytoplasmic tail of β3-integrin stabilizes the active, open conformation. Dok1 competes with talin for binding to this PTB domain. In contrast to the talin binding, the Dok1 binding negatively regulates the integrin signaling. The affinity of this PTB for talin versus Dok1 is determined by the pY level of tyr-747, in that increased pY747 phosphorylation favors Dok1 binding, while reduced phosphorylation prefers talin binding. Thus, the observed shift in the binding affinity of β3-integrin from one that favors Dok1 to the one that favors talin in Epha4 null osteoclasts is consistent with the observed reduction in pY747 level in Epha4 null osteoclasts. Based on these findings, we postulate that activation of EphA4 signaling in osteoclasts increases the pY747 level, leading to an increased binding of Dok1 over talin, which in turn results in suppression of the β3-integrin signaling and subsequent reduction in osteoclastic activity. Consistent with this hypothesis, Dok1 null mice developed severe osteopenia via an activation of the osteoclast activity and not through increased osteoclastogenesis. In addition, peptide-based arrays have identified Dok1 as one of the potential substrates for EphA4. The activated Dok1 also acts as negative regulators of the Ras-Erk1/2 pathway. Thus, our results are consistent with the interpretation that the absence of EphA4-mediated Dok1 phosphorylation relieves the EphA4-mediated suppression of the β3-integrin signaling in Epha4 null osteoclasts, which in part plays a role in the observed increase in their resorption activity.
In conclusion, this study demonstrates for the first time that the forward signaling of EphA4 is a novel and potent negative regulator of osteoclast activity and that the molecular mechanism of EphA4 to negatively regulate osteoclast activity is mediated in part by suppressing the β3-integrin signaling through an increased binding of Dok1 to β3-integrin that resulting in suppression of Vav3 and ras-Erk1/2 signaling. Understanding the functional role of EphA4 and its molecular mechanism in regulation of osteoclastic resorption would not only offer important insights into the pathophysiology of osteoporosis and related bone-wasting diseases but could provide novel targets for development of novel effective therapeutic modalities to treat these disorders.
All authors state that they have no conflicts of interest.
This work was supported by a Merit Review provided by the Office of Research and Development, Medical Research Service, Department of Veteran Affairs. All work was performed in facilities provided by the Department of Veterans Affairs. The authors wish to thank Prof. Elena Pasquale and Prof. Andrew Boyd for providing us with breeding pairs of Epha4 null mice and also their invaluable advice on how to breed viable homozygous Epha4 null mice.
Authors' roles: Study design: KL. Study conduct: KL. Data collection: VS, MA, and MS. Data analysis: KL, VS, and SM. Data interpretation: KL and SM. Drafting manuscript: VS and KL. Revising manuscript: VS, MA, MS, SM, and KL. Approving final version of manuscript: VA, AM. MS, SM, and KL. KL takes responsibility for the integrity of the data analysis.