During fracture repair, endochondral ossification occurs in a spatial and temporal manner, progressing longitudinally across the fracture line. The fracture repair process begins with the infiltration of inflammatory cells followed by the proliferation of fibrous/mesenchymal cells at the fracture site.1, 2 Chondrocytes, derived from mesenchymal stem cell (MSC) progenitors, synthesize a cartilaginous matrix to form the early bridging “soft” callus that provides the initial stability to the fractured cortical bone. This cartilage template is subsequently mineralized1 and is then resorbed by osteo/chondroclasts and replaced by a disorganized woven bone matrix, synthesized by infiltrating MSC-derived osteoblasts as early as 2 weeks postfracture.2 The irregular woven bone is then remodeled by the coordinated activity of osteoclasts and osteoblasts to form mechanically stable lamellar bone, with the same structural and functional capacity as prefractured bone, between 6 and 12 weeks postfracture.1 To date, various molecules secreted by MSC and osteogenic cells have been implicated in facilitating the bone remodeling process. These include colony-stimulating factor (M-CSF), stromal derived factor-1 (SDF-1), receptor activator of nuclear factor kappa-B ligand (RANKL), and the RANKL decoy receptor, osteoprotegerin (OPG). These molecules regulate the recruitment and maturation of RANK-expressing osteoclasts, which resorb the damaged bone. However, the precise mechanisms that mediate the interplay between the osteogenic and osteoclast cell populations still remain to be determined.
Recent studies by Zhao and colleagues demonstrated, for the first time, the importance of EphB4 and ephrinB2 during the “coupling” of osteoblast and osteoclast function in the bone remodeling process.3 The Eph molecules are a large family of receptor tyrosine kinases divided into A and B subclasses that bind to ephrin ligands. They are well known for their function as contact-dependent repellent molecules, playing diverse roles during development and in the postnatal organism, including boundary formation, axon guidance, angiogenesis and skeletal formation, and bone homeostasis.4 During bone homeostasis, EphB4 expressed by osteoblasts inhibits osteoclast differentiation by the activation of ephrinB2, expressed by osteoclast precursors,3 which is thought to be mediated by dishevelled 2 binding to the PDZ domain of ephrinB2.5 In addition, we have previously demonstrated that EphB4 interactions are important in the in vitro osteogenic differentiation and migration of human bone marrow–derived MSC.6 Other studies have shown that ephrinB2 expression by osteoblasts is stimulated by parathyroid hormone receptor 1 (PTH) and PTH-related protein (PTHrP) in a dose-dependent manner, and that ephrinB2:EphB4 signaling promotes osteoblastic mineralization in vitro.7
During development, the complete ablation of ephrinB28, 9 or EphB410 confers embryonic lethality at day E11.5 and E10, respectively, both of which precede embryonic skeletal development at E14.5. As such, little is known about the possible role for ephrinB2 or EphB4 in intramembranous or endochondral ossification. Because skeletal development includes the same process of endochondral ossification as seen during bone fracture repair, the present study examined the role of EphB4 during endochondral ossification using a model of fracture repair in transgenic mice that overexpress EphB4 under the control of the collagen type I promoter (Col1-EphB4).3
Materials and Methods
The Col1-EphB4 transgenic mice overexpressing EphB4 (equal numbers of 7-week-old male and female mice) under the control of the collagen type 1 promoter were backcrossed for seven generations.3 The use of these mice for the fracture studies was approved by the SA Pathology Animal Ethics Committee (# 56/09). The surgical procedure was adapted from Duvall and colleagues.11 Briefly, the animals were anesthetized by an intraperitoneal injection of 80 mg/kg ketamine and 10 mg/kg xylazine. An incision was made through the skin parallel to the femur, followed by a blunt end dissection separating the muscles covering the femur, which was fractured in the center using bone scissors. A carbon rod was inserted into the intramedullary cavity to stabilize the femur. The muscle covering the fracture was sutured, and auto-wound clips were used to close the skin. An X-ray of the fractured femur was taken immediately after surgery to determine any displacement of the femur.
Harvested femurs were stored frozen at –20°C and wrapped in PBS-soaked gauze before use for mechanical testing. Four-point bending analyses to failure were conducted using a materials testing machine (Model 800LE4; TestResources Inc, Shakopee, MN, USA) based on previously published protocols.12, 13 The upper span (loading) width was 3.0 mm and the lower span (support) width was 7.0 mm. Both the upper and lower contact anvils had a radius of 1.0 mm. Femora were positioned posterior side down on the support anvils to cause bending about the medial-lateral axis (posterior side in tension). The bones were centered on the supports to ideally induce failure at the midpoint along the diaphysis on the tension (posterior) side. A preload of 1 N was applied and then a constant cross-head displacement rate of 0.017 mm/s for the bend tests to failure. Compliance in the load-line and bend fixtures was removed from the displacement measurement by using a correction factor obtained from tests with an aluminium calibration specimen. The bend tests were conducted in air (22°C to 24°C) and, before testing, the bones were kept in PBS-soaked gauze at room temperature for 1 hour. Load-displacement curves were analyzed using standard Euler-Bernoulli theory for linear-elastic beams to obtain the bending stiffness and the ultimate bending moment. The bending stiffness, or flexural rigidity, was calculated by multiplying the measured displacement by 6/(3aL-4a2), where L is the lower span width and a is half the difference between the lower and upper span widths. The bending moment was calculated by multiplying the ultimate force by the value a. No correction for shear deformation was made.
Micro-computed tomography (µCT)
Analysis was conducted as previously described.14 3D microarchitecture of the fractured and nonfractured femora was evaluated using µCT (Skyscan 1174 X-ray Microtomograph, SkyScan, Kontich, Belgium). All bone samples were scanned at 74kV/100mA with an isometric resolution of 8.7 µm per pixel using a 0.25-mm aluminium filter and two-frame averaging. Reconstruction of the original scan data was performed using NRecon (SkyScan). Analysis of microarchitectural parameters was performed using CTAn (SkyScan). For the fractured femora, the region of interest was defined as 450 slices above and below the fracture site, whereas the nonfractured femora was defined as midpoint of the cortical bone and then the region of interest was set 450 slices above and below this point. For tissue volume and bone volume analysis, CTAn was used. To determine the actual increase in tissue and bone volume, the value obtained for the contralateral femur was subtracted from the fractured femur.
Colony-forming unit fibroblast (CFU-F) assays
Mouse bone marrow stromal progenitors were isolated from flushed fractured or nonfractured femora as previous described.14 Briefly, the femora were flushed with ice-cold growth media and the compact bone was crushed then incubated with collagenase (3 mg/mL) and DNAseI (50U/mL) for 45 minutes at 37°C. The bone chips were then filtered and added to the flushed marrow, samples were washed twice with ice-cold Hank's buffered salt solution and resuspended in growth media. For CFU-F assays, cells were plated in duplicate at 9 × 105, 1.8 × 106, and 3.6 × 106 cells/well in six-well plates, incubated for 45 minutes, and the nonadherent cells were transferred to another six-well plate. Samples were then cultured for 14 days in alpha modified essential medium (αMEM), supplemented with 20% (v/v) fetal bovine serum, 2 mM L-glutamine, 1 mM sodium pyruvate, 100 µM L-ascorbate-2-phosphate, penicillin (50 iu/mL), and streptomycin sulphate (50 µg/mL), at 37°C humidified atmosphere in the presence of 5% CO2 and then stained in 1% toluidine blue and 0.5% paraformaldehyde solution, rinsed in PBS, air-dried, and colonies (aggregates ≥50 cells) scored. The remaining cells and bone chips were cultured in growth media, expanded, and used for assays. Mineralization analysis in vitro was conducted as previously described.6, 15 Glycosaminoglycan synthesis was measured in replicate high density (5 × 104 cells per well in 96-well plates) in chondrogenic inductive media6 supplemented with TGFβ1 10 ng/mL cultures by 35SO4 (Perkin Elmer Life and Analytical Sciences, Downers Grove, IL, USA) incorporation using a TopCount NXT Microplate Scintillation & Luminescence counter (Perkin Elmer Life and Analytical Sciences). Values were normalized to DNA content per well. Cell proliferation was assessed in replicate cultures (seeded at 5 × 103 cells per cm2) by the measurement of 3H-thymidine (GE Healthcare, Buckinghamshire, UK) incorporation using a TopCount NXT Microplate Scintillation & Luminescence counter.
Methacrylate processing and embedding of tissues and staining of sections (5 µm) were conducted as previously described.14, 16, 17 Briefly, sections were treated with acetone for 15 minutes, rinsed in RO water, and then processed for the appropriate stain. Toluidine blue sections were stained in 2% toluidine blue solution for 20 minutes. Safranin O sections were stained in 0.2% Fast Green (Sigma, St. Louis, MO, USA) for 5 minutes and then in 0.1% Safranin O Stain. Von Kossa sections were stained in 1% aqueous silver nitrate (AgNO3) for 60 minutes under UV light, washed in dH2O, and subsequently bathed in 2.5% sodium thiosulphate for 5 minutes. Sections were washed in dH2O and then counterstained in hematoxylin and eosin. TRAP slides were incubated at 37°C for 30 minutes in AS-BI phosphate (0.4 mg/mL) in acetate-tartrate buffer (200 mM sodium acetate, 100 mM potassium sodium tartrate, pH 5.2). The samples were then transferred to sodium nitrite solution (1:1 v/v), in prewarmed tartrate-acetate buffer and incubated for 30 to 60 minutes at 37°C. Sections were rinsed twice in tap water and then counterstained with 0.05% methyl green solution. Sections were imaged using NanoZoomer imaging (Hamamatsu Photonics, Hamamatsu City, Shizuoka, Pref., Japan) and viewed using NDP.view software (Hamamatsu Photonics). Quantitation of images was conducted using either NDP.view or Image J software (National Institutes of Health, Bethesda, MD, USA). Immunohistochemistry was performed on 5-µm decalcified paraffin-embedded sections to detect the expression of either EphB4 or alpha-smooth muscle actin, using either rabbit anti-EphB4 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), mouse anti-human SMA (DAKO, Glostrup, Denmark) or corresponding negative controls, rabbit IgG or mouse IgG1 isotype control antibodies (Caltag Laboratories, Burlingame, CA, USA) at 2 µg/µL, respectively. Samples were then incubated with either rabbit anti-mouse biotin or goat anti-rabbit biotin (1/200), followed by Streptavidin-HRP, then visualized with DAB substrate (DAKO). Sections were counterstained with hematoxylin and mounted with DePix (BDH, Poole, UK).
All statistical analysis was conducted using GraphPad Prism 5 software. All samples were analyzed using either paired or unpaired Student's t test parametric analysis.
EphB4 gene expression is elevated in the early stages of bone fracture repair
Initial experiments assessed the EphB4 transcript and protein expression in bone samples harvested from fractured femora of wild-type mice, using a femoral fracture model stabilized with an intramedullary carbon rod. EphB4 gene expression was significantly elevated in the first 2 weeks of fracture repair compared with the corresponding contralateral nonfractured limbs (Fig. 1A). Immunohistochemical analysis revealed positive staining for EphB4 protein by chondrocytes in tissue sections of developing callus at 1 week postfracture and by osteoblasts and osteocytes during new bone formation at 2 weeks postfracture (Fig. 1B). Blood vessels were also positive for EphB4 expression at 2 weeks postfracture.
Transgenic EphB4 mice exhibit enhanced bone rigidity and strength
Transgenic mice that overexpress EphB4 under the control of the collagen type 1 promoter (Col1-EphB4), and their wild-type controls,3 were used to examine the functional relevance of osteoblast-specific expression of EphB4 during the endochondral bone repair process. Preliminary analyses found no statistical differences in bone volume prefracture between 7-week-old Col1-EphB4 and wild-type mice (data not shown). At 8 weeks after surgery, the fractured and nonfractured femora from Col1-EphB4 and wild-type mice were processed for four-point biomechanical testing. This time period has been previously shown to coincide with a period of extensive endochondral bone repair.1 Fractured femora from Col1-EphB4 were found to be significantly more rigid and were able to sustain a significantly greater ultimate load than the corresponding contralateral, nonfractured femora (Fig. 2A, B). In addition, at 8 weeks postfracture, the difference in flexural rigidity between fractured and nonfractured femora was significantly greater with Col1-EphB4 mice than their wild-type controls (Fig. 2A), in contrast to ultimate load capacity, which showed no difference between Col1-EphB4 and wild-type fractured limbs at the week 8 time point (Fig. 2B).
EphB4 augments callus formation
To evaluate the actual contribution of EphB4 to callus and bone volume during the bone repair process, the difference in total tissue or mineralized tissue volume between the fractured and the contralateral nonfractured femora was analyzed by µCT at (1, 2, 4, 8, and 12 weeks postfracture) (Fig. 3A, B). No significant difference in callus size was observed 1 week postfracture between transgenic and wild-type mice, which coincides with hematoma formation or the acute inflammatory response. However, at 2 weeks postfracture, the callus size, as assessed by total tissue volume, was significantly greater in Col1-EphB4 mice when compared with that of the wild-type control mice (Fig. 3A–C). Similarly, mineralized tissue volume was significantly greater in Col1-EphB4 mice at the 2-week time point (Fig. 3B). This time frame is consistent with callus formation and endochondral ossification. The remaining stages of bone repair, 4 to 12 weeks postfracture, did not reveal a significant difference in tissue or mineralized tissue volume between transgenic and wild-type mice; however, there was a consistent trend toward elevated tissue and bone volume in Col1-EphB4 mice (Fig. 3A, B). Histomorphometric analysis largely confirmed the µCT findings for the 4- and 8-week time points in relation to cortical thickness but found that there was a significant increase in the bone area in relation to total tissue area at week 4 postfracture (Fig. 3D), which was not evident by the week 8 time point (Fig. 3E).
EphB4 contributes to all stages of endochondral ossification during bone repair
The second stage of bone repair, endochondral ossification, commences once the inflammatory response subsides, approximately 1 to 2 weeks postfracture. This process continues in a spatial and temporal manner from the outside of the callus toward the fracture gap or bone bridge until the woven bone is replaced with lamellar bone.2 Considering that EphB4 transgenic mice displayed a significantly greater callus volume at 2 weeks postfracture, we investigated the formation of fibrous connective tissue matrix, cartilaginous matrix, and new bone matrix at the bone bridge and within the callus at this time point (Fig. 4). Because of the spatial and temporal endochondral ossification of the callus from the outside in, each half of the callus was segmented into thirds and referred to lateral, intermediate, or medial to the cortical bone. This analysis indicated that the fractured femora of Col1-EphB4 mice displayed a significantly larger bone bridge composed of fibrous tissue matrix proximal to the intramedullary cortical bone and cartilaginous tissue matrix in intermediate and distal regions, when compared with wild-type controls (Fig. 4A). Furthermore, Col1-EphB4 mice produced significantly more cartilage (Fig. 4B) within the callus and displayed significantly more mineral deposition when compared with the wild-type controls (Fig. 4C).
To determine whether EphB4 affected bone resorption during the fracture repair process, sections from fractured femora 2 weeks postfracture were evaluated for TRAP-positive cells with three or more nuclei14 to identify mature chondro/osteoclasts (Fig. 4D). The data showed that there were significantly fewer mature chondro/osteoclasts observed in the callus of EphB4 transgenic mice at 2 weeks postfracture compared with the wild-type littermate controls enumerated over the total callus area. However, histomorphometric examination of weeks 4 and 8 times points found no significant difference in the number of TRAP-positive multinuclear osteoclasts over the bone surfaces at the callus sites (data not shown).
EphB4 increases numbers of osteogenic progenitors and functional potential
To determine whether EphB4 influences the number of mesenchymal stem cells/progenitors, compact bone and marrow from prefractured femora were isolated from Col1-EphB4 and wild-type mice and assessed for the incidence of CFU-F. The number of clonogenic progenitors (CFU-F) isolated from transgenic EphB4 mice was significantly higher than those of the wild-type littermate controls (Fig. 5A). Moreover, Col1-EphB4 mice exhibited significantly greater numbers of clonogenic progenitors in the fractured femora, when normalized to that of the contralateral nonfractured femora, than wild-type controls at 2 and 8 weeks postfracture (Fig. 5B). Immunohistochemical staining of 2-week-old fractures showed elevated levels of α-smooth muscle actin-positive cells in the callus and fracture sites of wild-type and Col1-EphB4 mice (Fig. 5C). Interestingly, we found higher levels of α-smooth muscle actin staining at the proximal and distal metaphysis of the fractured femora within the bone marrow spaces adjacent to the trabeculae for the Col1-EphB4 mice in comparison to the wild-type mice.
In functional studies, ex vivo expanded osteogenic progenitors isolated from fractured femora of Col1-EphB4 and wild-type mice were cultured under osteogenic inductive conditions to determine whether elevated levels of EphB4 affected mineral formation in vitro. Osteogenic progenitors isolated from transgenic EphB4 mice were found to produce significantly more mineralized matrix, as reflected by calcium quantitation and alizarin red staining, when compared with those of the wild-type littermates (Fig. 5D, E). However, histomorphometric analysis at 4 and 8 weeks postfracture failed to find any statistical significant differences in the number of osteoblasts over the bone surfaces at the callus sites (data not shown). To provide a potential mechanism of the observed increase in cartilage formation, we showed that mouse chondrocytes, known to express ephrinB2, exhibited an increase in the proliferation rate (Fig. 5F) and glycosaminoglycan synthesis potential (Fig. 5G) after stimulation with clustered EphB4-Fc fusion molecules compared with human Ig alone.
Bone remodeling is not only important during bone repair after injury but is also continuously active throughout life helping to maintain the mechanical and structural integrity of the skeleton. In this study, we demonstrate, for the first time, the role of EphB4 in endochondral ossification during bone repair after fracture. Our studies showed that upregulation of EphB4 gene and protein expression during the first 2 weeks after fracture in the chondro/osteogenic cell lineages resulted in aberrant repair that led to stiffer and more rigid bones 8 weeks postfracture when compared with the contralateral unfractured limbs. Moreover, this response correlated with the enhancement of a number of processes associated with endochondral ossification, including callus formation, cartilaginous matrix formation, and bone mineral deposition, which was associated with an impairment in bone remodeling.
Examination of the stromal compartment showed an enhanced mineral-forming capacity of MSC isolated from bone collected at prefracture and at 2 and 8 weeks postfracture, which was not associated with increases in the number of osteoblasts at week 8. These findings are consistent with previous reports that demonstrated that higher levels of activated EphB4 facilitated increased levels of mineralization by cultured murine calvaria cells.3 Supportive studies demonstrated that inhibition of ephrinB2:EphB4 signaling using a blocking peptide (TNYL) or a decoy soluble form of EphB4 (sEphB4) caused a decrease in the capacity of human MSC6 and mouse bone cells7, 18 to form mineral in vitro, which correlated with an inhibition in the gene expression levels of osteoblastic bone-associated genes associated with mineralization. However, those studies could not discern whether the effect of these inhibitors was because of inhibiting ephrinB2 signaling in osteoblasts, or inhibiting EphB4 signaling. Moreover, our studies have also shown that EphB4 transgenic mice–derived MSC have a decreased capacity to undergo adipogenesis compared with wild-type MSC (data not shown). Collectively, these studies implicate a role for EphB4 in osteogenic cell fate determination.
Another potential mechanism for the observed enlarged callus and increase in ossification in EphB4 transgenic mice may be associated with the increase in clonogenic osteoprogenitor cell numbers derived from bone harvested at prefracture and at 2 and 8 weeks postfracture. These data imply that EphB/ephrin-B interactions may be involved in the recruitment of MSC from their perivascular niche toward areas of tissue damage after injury.6 In support of this notion, we found elevated levels of α-smooth muscle actin–positive progenitors within the bone marrow spaces of EphB4 transgenic mice 2 weeks after fracture. Previous studies have shown that MSC-like populations reside in perivascular sites within dental pulp tissue, which mobilize to areas of dentine injury through changes in Eph/ephrin expression in the surrounding fibrous tissue.19 In other systems, EphB4/ephrinB2 interactions stimulate sinusoidal endothelial recruitment after chemical-induced injury in the liver.20 In the present study, it appears that EphB/ephrinB molecules may act on mesenchymal cells at multiple levels to promote the self-renewal of MSC, stimulate the migratory capacity of osteogenic progenitor populations toward sites of bone injury, and stimulate osteoblast differentiation and activity for enhanced bone formation.
Histological examination of the fracture sites at 2 weeks postfracture demonstrated a significant decrease in the number of mature multinucleated TRAP-positive osteoclasts within the callus site, possibly delaying the early resorption of the cartilaginous tissue. However, further examination at 4 and 8 weeks postfracture failed to demonstrate any significant differences in the number of multinucleated TRAP-positive osteoclasts present on the bone surfaces at the callus sites. These findings are consistent with the findings of Zhao and colleagues, who reported that EphB4 transgenic mice displayed a significant decrease in osteoclast number, osteoclast surface, and eroded surface, which did not correlate to changes in RANKL, M-CSF, or OPG expression, but rather the activation of reverse signaling in osteoclast precursors expressing ephrinB2 and resulting in the inhibition of osteoclast maturation.3 Under pathological conditions, there is a decrease in EphB4 and ephrinB2 gene expression in cultured MSC derived from myeloma patients with osteolytic disease.21 Treatment of myelomatous SCID-hu mice with EphB4-Fc inhibited myeloma growth and osteoclastogenesis and stimulated bone formation by osteoblasts in vivo. Therefore, changes in the expression levels of EphB4 by committed osteogenic cells during bone repair may regulate osteoclast numbers or maturation, supporting the notion that EphB4 plays an important role during the bone remodeling process in coupling bone mineral formation to resorption.3
To date, EphB4/ephrinB2 interactions have not been reported to effect endochondral ossification during development, as complete ablation of either EphB4 or ephrinB2 results in embryonic lethality before skeletal development.8–10 However, aberrant EphB4 has been associated with skeletal diseases such as osteoarthritis,22, 23 where EphB4 was significantly upregulated in both the superficial and deep zones of human osteoarthritic chondrocytes and cartilage. Interestingly, treatment with soluble ephrinB2 blocked the action of several catabolic factors including IL-1β, IL-6, MMP-1, 9, and 13, and PAR-2, while significantly enhancing collagen type II expression.23 Therefore, it is plausible that ephrinB2 expressed by chondrocytes may activate EphB4 expressed by chondrocytes,23 which promotes their calcification and subsequent mineralization by recruited osteoblasts. Studies by Diercke and colleagues showed that ephrinB2 was significantly upregulated in periodontal ligament fibroblasts after mechanical tensile strain. This mechanical strain subsequently activated bone remodeling during tooth movement in the alveolar bone socket, which contributed to osteoblastic differentiation mediated through ephrinB2 activation of EphB4 expressing osteoblasts within the alveolar bone.24 In the present study, we found that mouse chondrocytes exhibited an increase in proliferation and glycosaminoglycan synthesis under chondrogenic conditions, after stimulation with soluble EphB4-Fc molecules.
In summary, the current study demonstrates that EphB4 increases the stiffness and strength of fractured bones by enhancing the processes of endochondral ossification. This is mediated by increased numbers of clonogenic MSC with high bone-forming potential that results in greater fibrous, cartilaginous, and mineralized tissue and delays the process of chondrocyte resorption at the early phases of callus formation. These findings lend support to the potential that developing EphB4/ephrinB2-based therapeutic strategies may enhance the fracture repair process. Current Eph-related drug therapies being investigated for cancer treatment25 involve administration of the EphB4 peptide (referred to as PEGylation of the TNYL-RAW peptide) that inhibits the EphB4 forward activation by ephrinB2.26 Alternatively, in vivo mouse studies examining osteolytic bone disease associated with multiple myeloma showed that administration of soluble Fc molecules bound to EphB4 or ephrinB2 could both stimulate osteoblastogenesis and bone formation. In addition, EphB4 also inhibited myeloma growth, osteoclastogenesis, and angiogenesis, whereas ephrinB2-Fc promoted angiogenesis.21 Utilizing these agents to stimulate or inhibit the function of EphB4 during different stages of fracture repair may be useful in enhancing the bone repair process after fracture.
All authors state that they have no conflicts of interest.
This work was supported by the NH&MRC project grants 565176 and 1023390.
Authors' roles: AA (conception and design, acquisition of data, data analysis, interpretation of data, manuscript preparation, manuscript editing); RAP (acquisition of data, data analysis, interpretation of data, manuscript editing); LC (acquisition of data, data analysis, interpretation of data, manuscript editing); DM (acquisition of data, data analysis, interpretation of data, manuscript editing); IHP (acquisition of data, data analysis, interpretation of data, manuscript editing); JDC (acquisition of data, data analysis, interpretation of data, manuscript editing); KV (acquisition of data, data analysis, interpretation of data, manuscript editing); ACWZ (conception and design, data analysis, interpretation of data, manuscript editing); SAK (conception and design, data analysis, interpretation of data, manuscript editing); NAS (interpretation of data, manuscript editing, provision of transgenic animals and reagents); KM (interpretation of data, provision of transgenic animals, manuscript editing); SG (corresponding author, conception and design, data analysis, interpretation of data, manuscript preparation, manuscript editing).