Neurofibromatosis type 1 (NF1) is a common genetic condition characterized by cutaneous and subcutaneous neurofibromas, café au lait macules, and musculoskeletal lesions.1 Genetic linkage studies have identified the NF1 gene as being responsible for this condition. NF1 encodes for the neurofibromin protein, a key negative regulator of the small GTPase Ras.2, 3 Affected patients have been shown to be heterozygous for NF1; however, some disease phenotypes have been attributed to sporadic somatic mutations within defined cell types.4, 5 This not only includes neurofibromas and other NF1-associated tumors, but also more recently skeletal lesions associated with focal bone defects.4, 5 Nevertheless, the pathobiological importance of heterozygous versus homozygous NF1 cells as well as crosstalk between these cell types remains unclear, particularly in the context of bone.6–8
The bone and orthopedic manifestations of NF1 can have a significant negative impact on patient quality of life. Generalized osteopenia and/or osteoporosis, scoliosis, compromised bone repair and pseudarthrosis,9 and other bony manifestations can affect up to 50% of individuals with NF1. In particular, the orthopedic complications have a poor prognosis and can be recalcitrant to current medical interventions. Genetic mouse models have been quintessential to the improved understanding of the role of NF1 in bone. Nf1+/− primary osteoblasts showed a decreased osteogenic potential in culture.10 Heterozygous Nf1+/− deficient mice exhibited an impaired response in BMP-7 in an ectopic bone formation assay11 as well as reduced bone healing in a distal tibial fracture model.12 More advanced models have been subsequently generated in an attempt to model the double inactivation of NF1 seen in some human bone defects. Nf1 mice lacking Nf1 in the calvaria and limb mesenchyme showed runting, congenitally bowed tibias, and impaired bone healing in a drill hole defect.13, 14 Osteoblast-specific knockout mice (Nf1) showed an increased bone phenotype and impaired bone healing, albeit with increased callus size.15 Whereas these models indicate double inactivation of Nf1 can have profound and negative effects on bone, these models do not optimally model the local focal defects hypothesized to be causal of congenital tibial dysplasia. In this study we report a new genetic and surgical model that utilizes a Cre-expressing adenovirus (Ad-Cre) to locally knock out Nf1 in healing fractures. Midshaft fractures were performed and bony union was assessed after 3 weeks. The effects of double inactivation of Nf1 in committed and inducible osteoblasts were also examined in vitro. In cell culture we investigated the effects of early and late double inactivation of Nf1 in cultured calvarial osteoblasts to explore the hypothesis that NF1 is important for early osteoblast differentiation. This was further explored in an alternative model for examining recombinant bone morphogenetic protein-2 (rhBMP-2)-induced osteogenesis in heterozygous or homozygous muscle-derived cells that can function as inducible osteoprogenitors.16, 17
In addition to an osteoblastic deficiency, Nf1+/− osteoclasts have been implicated in the NF1 bone phenotype.18 Interactions between Nf1null osteoblasts and Nf1+/− preosteoclasts may also be significant. To explore this concept, fracture models were performed in both Nf1flox/flox and Nf1flox/− mice. The former genotype has wild type osteoclast precursors whereas the latter has functionally Nf1+/− osteoclast precursors. It was speculated that the presence of Nf1+/− osteoclasts precursors may lead to a more severely affected repair phenotype.
Materials and Methods
Nf1flox/flox mice originally generated by Prof L. Parada were sourced from the National Cancer Institute (NCI) mouse repository (Bethesda, MD, USA).19Nf1+/− mice were a generous gift from Prof L. Parada (University of Texas Southwestern Medical Center, Dallas, TX, USA).20 We maintained both colonies on a C57/Bl6 background. We bred Nf1flox/flox mice with Nf1+/− mice for two generations to generate experimental mice of Nf1flox/flox and Nf1flox/− genotypes. MyoD-cre transgenic mice were a gift from Prof David Goldhamer (University of Connecticut, Storrs, CT, USA)21; we crossed them in house with the Nf1flox/flox line to generate Nf1 mice. Z/AP transgenic reporter mice were a gift from Prof Patrick Tam (Children's Medical Research Institute, The University of Sydney, Sydney, Australia) with the permission of Prof Andreas Nagy,22 we backcrossed this strain onto a C57Bl/6 background in house. We used female mice for fracture experiments. Mice were housed in autoclaved polypropylene solid boxes with stainless steel lids and polycarbonate water bottles with food and mouse chow supplied ad libitum. Animal experiments were performed with approval from the South Western Area Health Service (SWAHS) Animal Ethics Committee.
Adenovirus production and purification
HEK293 cells were grown to 80% confluence in T175 culture flasks and transduced with Cre-expressing adenovirus (AdCre), a gift from Dr Thomas Gajewski (University of Chicago, Chicago, IL, USA),23 or adenovirus containing green fluorescent protein (AdGFP) (AdEasy System, Microbix, Mississauga, Ontario, Canada). We harvested cells within 72 hours after transduction, then pelleted them and subjected them to cell lysis and DNase treatment. We ultracentrifuged the resulting supernatant for 1 hour in multiple tubes; viral fractions were then pooled and re-spun for 20 hours. Centrifugation was performed in a standard CsCl gradient at 35,000 rpm in a Beckman SW-41Ti rotor. We dialyzed purified virus in a Slide-A-Lyzer Dialysis Cassette, 10- kDa cutoff (Thermo Scientific, USA) for up to 3 days in a 10 mM Tris (pH 8.0) solution. Following dialysis, virus was titered using the Adeno-X titration kit (Clontech, USA) and used or frozen at −80°C in 10% glycerol.
We generated calvarial cultures from Nf1flox/flox neonatal mice harvested within 1 week of birth using published culture methods.24 In summary, mice were decapitated and the calvaria were washed in Dulbecco's PBS while stripping any remaining fibrous tissue. Calvaria were minced and digested three times in Collagenase D at 1 mg/mL (Roche) in α modified essential medium (α-MEM) (Invitrogen) at 37°C for 15 minutes per digestion. The resulting calvarial fragments were plated in 10% fetal bovine serum in α-MEM with penicillin/streptomycin and cells were allowed 3 to 5 days to grow onto the plate before splitting for experiments. Nf1flox/flox calvarial osteoblasts were treated with either AdCre23 or AdGFP via transduction with the Geneporter reagent (Stratagene, USA).25 Cre expression induces recombination and excision of exons 31 and 32 of the Nf1-flox allele, creating a nonfunctional protein product.19
We derived primary muscle cell cultures from Nf1 conditional knockout mice. Neonatal hind limb muscles were harvested from pups within a week of birth, washed in PBS, and then minced with surgical scissors. Tissue fragments were plated in 10% fetal bovine serum (FBS) in DMEM with penicillin/streptomycin, on a collagen coated dish for 3 to 5 days until cells were ready to be split for experiments.
Cultured cells underwent osteoblastic differentiation in α-MEM containing 10% FBS and penicillin/streptomycin, supplemented with 10 mM β-glycerophosphate, 50 µg/mL ascorbic acid, and 100 to 200 ng/mL of rhBMP-2 (Medronic Australasia).
Surgical methods and adenovirus delivery
We used female mice aged to 8 to 12 weeks in closed and open fracture surgery experiments. All fractures were generated at the tibial mid-shaft, near the fibular junction, by experienced staff. This mouse midshaft fracture model has been previously described.12 In summary, anesthesia was induced with ketamine (35 mg/kg) and xylazine (4.5 mg/kg) intraperitoneally (i.p.) and maintained using inhaled isoflurane. A small incision was made slightly below the knee and an entry point made using a 27G needle. An intramedullary rod (0.3 mm-diameter stainless steel insect pin) was surgically inserted inside the medullary canal of the tibia, and a second pin was inserted for stability. Fractures were generated manually by three-point bending using a modified set of surgical staple removers. For open fractures only, an incision was then made in mid-diaphyseal region and the periosteum was damaged, but not extensively stripped. Fractures were confirmed to be transverse, above the fibular junction, and without comminution via radiography using a digital X-ray machine (Faxitron X-ray Corp, USA).
We injected approximately 1 × 108 infectious units of purified AdCre adenovirus at the fracture site and surrounding soft tissue in an isotonic solution (10 mM Tris, pH 8.0). All experimental mice received AdCre treatment, regardless of genotype. Pain relief was managed using buprenorphine (0.05 mg/kg subcutaneously postoperatively, then every 12 hours as required). Fracture repair was monitored by weekly X-ray and in the event that internal fixation had failed (due to pin slippage, bending, or breakage) the affected mouse was euthanized and excluded from subsequent analysis. Select animals received bromodeoxyuridine (BrdU) labeling reagent 2 hours prior to euthanasia at 21 days (Invitrogen). At the experimental endpoint, animals were euthanized using carbon dioxide and specimens collected postmortem for radiography and histology.
We removed fractured tibias with surrounding soft tissue, fixed them overnight in 4% paraformaldehyde, and then stored them in 70% alcohol at 4°C until ready for decalcification. Bone samples were decalcified in 0.34 M EDTA (pH 8.0) for 21 days at 4°C on shaker with changes every 2 to 3 days. Samples were embedded in paraffin and 5-µm-thick sections were stained with Picro Sirius Red/Alcian Blue to differentiate bone and cartilage. Adjacent sections were stained for tartrate-resistant acid phosphatase (TRAP) expression to highlight osteoclasts, and counterstained with 0.4% Light Green.11 Sections were also stained for BrdU using a commercial kit (Invitrogen). Stained sections were quantified and analyzed by a blinded assessor using the Bioquant Analysis System. Statistical analyses and graphing were done using GraphPad Prism, and a two-tailed t test or Fisher's exact test were used to determine statistical significance.
For cell tracking experiments to validate AdCre infection and recombination, we fixed Z/AP specimens and cut cryosections. Samples were heat treated for 40 minutes at 70°C to inactivate any endogenous alkaline phosphatases and then stained for the heat-resistant human placental alkaline phosphatase (hAP) using a nitro-blue tetrazolium and 5-bromo-4-chloro-3'-indolyphosphate (NBT/BCIP) solution (Roche Diagnostics; Cat #1161451001) diluted 1:50 in a staining buffer (0.1 M NaCl, 0.1 M Tris-HCl pH 9.4, 0.1 M MgCl2, 0.1% Tween-20).
PCR and quantitative PCR
We harvested DNA and RNA from cells using Trizol Reagent (Invitrogen), and cDNA using Superscript III (Invitrogen). qPCR was performed using a SYBR green master mix (Stratagene, USA) using the following primer sequences:
We examined bone repair using microcomputed tomography (µCT) using a SkyScan 1174 compact µCT scanner (SkyScan, Kontich, Belgium). Samples were scanned in 70% ethanol at 9.4-µm pixel resolution, 0.5-mm aluminum filter, 50-kV X-ray tube voltage, and 800-µA tube electric current. The images were reconstructed using NRecon, version 188.8.131.52 (SkyScan). A global threshold was used representing bone tissue. Representative three-dimensional fractures were reconstructed with sagittal slices using CTVol Realistic Visualization software version 184.108.40.206 (SkyScan).
We defined union based on 3-week X-ray radiographs of each fracture. X-rays from 1 week and 2 weeks were initially reviewed and any specimens showing a premature loss of fixation were excluded from subsequent analysis. Fractures were defined as not united when a complete lack of bridging of the outer neocortex was observed by X-ray. This was previously used to show reduced fracture healing in Nf1+/− mice.12 A lack of union at 3 weeks does not necessarily imply that fractures would proceed to non-union, only that normal healing was delayed. In this study, fractures were assessed by up to two blinded assessors, and statistical significance was determined using a Fisher's exact test.
Nf1 deletion impairs the differentiation of committed osteoprogenitors but not more differentiated osteoblasts
To validate the capacity of AdCre to induce deletion of the Nf1-flox allele in vitro, we transduced cultured osteoblasts and extracted genomic DNA for further analysis. PCR experiments using primers specific for the excised Nf1 allele confirmed recombination in AdCre-treated test samples but not in AdGFP-treated controls (Fig. 1A).
To functionally evaluate the role of Nf1 in osteoblast differentiation, we generated Nf1null cells by treating Nf1flox/flox calvarial osteoblasts with AdCre and rhBMP-2 to promote osteogenic differentiation. Early treatment with AdCre of committed osteoprogenitors prior to rhBMP-2 treatment led to delayed mineralization at days 5 and 7 compared to AdGFP-treated cells (Fig. 1B). In contrast, late AdCre exposure of calvarial cells following 48 hours of rhBMP-2 treatment to induce osteoblastogenesis did not affect further Nf1null osteoblast differentiation or mineralization (Fig. 1C). These findings were further confirmed by assays for gene expression for alkaline phosphatase (Alp) and Runx2 after 5 days in culture (Fig. 1D). These data are consistent with NF1 influencing the differentiation of osteoprogenitors but not later stages of osteoblastic differentiation.
rhBMP-2–induced osteogenesis is impaired in Nf1null-inducible osteoprogenitors from muscle
We cultured neonatal myoblasts as an alternate type of inducible osteoprogenitor from Nf1 and Nf1 conditional knockout mice. Control cells were sourced from MyoD-Cre negative littermates. Nf1 excision was confirmed by PCR from explant muscle tissue (Fig. 2A). When treated for 4 days with rhBMP-2 to induce osteogenesis in vitro, control cells underwent complete matrix mineralization as seen by Alizarin Red S staining. In contrast, functionally Nf1 heterozygous and homozygous cells show a graded insensitivity to rhBMP-2 (Fig. 2B). Gene expression analysis confirmed that alkaline phosphatase expression was significantly decreased, and osteocalcin and Runx2 showed a downward trend in expression that did not reach significance. Myoblasts that were not treated with rhBMP-2 did not express any osteoblastic markers (Fig. 2C).
Validation of AdCre-induced gene deletion in fracture repair
Whereas it has been previously reported that Nf1+/− mice show delayed healing in the distal tibia,12 we generated a novel model of AdCre-mediated deletion of both Nf1 alleles. To confirm local recombination in mice, AdCre was injected directly into the fracture site at the time of fracture. PCR assays were performed on genomic DNA isolated from tissue isolated from the entire fracture callus at 5 days postsurgery. Positive bands indicating recombination of an Nf1flox allele were seen in AdCre-treated but not AdGFP-treated specimens (Fig. 3A).
As an alternate method of demonstrating AdCre-induced recombination, we injected Z/AP reporter mice with AdCre following fracture. The Z/AP reporter mouse responds to Cre expression by expressing a heat resistant human placental alkaline phosphatase (hAP) as described by Lobe and colleagues.22 In the Z/AP mice, Nf1 recombination was not induced. Specific hAP staining showed recombined cells persisting within the callus 3 weeks postfracture (Fig. 3B, arrows). No staining was seen in other osteogenic sites (such as the growth plate), confirming heat inactivation of endogenous AP.
Local double-inactivation of Nf1 impairs fracture repair
To identify the role of Nf1 in fracture healing, we used the AdCre virus to generate Nf1null fractures in Nf1flox/flox and Nf1flox/− mice. AdCre was also given to C57Bl/6 mice to control for any effects of virus dosing alone. Both closed fractures (intact skin and periosteum) and open fractures (open injury with periosteal stripping) were examined.
At 3 weeks postfracture, Nf1null fractures displayed poor union rates based on radiographic analyses. In the closed fracture model all control (wild type) fractures healed within 3 weeks. However, Nf1null fractures showed a significant impairment with union rates of 38% and 50% in the Nf1flox/flox and Nf1flox/− mouse lines, respectively (p < 0.05) (Fig. 4A). In the open fracture model, 77% of control fractures healed within 3 weeks. Again, less Nf1null fractures were bridged at this time, with 31% and 23% union in Nf1flox/flox and Nf1flox/− mice, respectively (p < 0.05) (Fig. 4B).
We processed specimens for histological analysis to investigate the potential cellular mechanism underlying healing delay. Picosirius Red/Alcian Blue staining showed that Nf1null fractures were characterized by a large amount of mesenchymal and fibrotic tissue in the fracture site. In both closed and open Nf1null fractures this nonosseous tissue represented an average of 15% to 20% of the callus area whereas only 1% was observed in control fractures (Fig. 5A,B). These elements were primarily fibrotic, with approximately 75% of the nonosseous tissue being mesenchymal or fibrotic. The predominance of fibrous tissue over cartilage normally seen in endochondral bone repair indicates a potential pathologically process leading to non-union.
Increased cellular proliferation is seen in Nf1null closed fractures
To examine proliferation within the Nf1null fracture calluses, we labeled mice with BrdU. Following staining, quantification was performed on cells lining the bone surface or within the callus area (including any nonosseous tissue but excluding hematopoietic cells within the marrow cavity). In Nf1null closed fractures a significant increase was seen in the number of proliferating cells at 3 weeks, compared to wild type control fractures that had largely healed (Fig. 6A). Proliferating cells were present in the fibrous tissue, as shown in the representative section, but were also generalized in the callus, suggesting an overall proliferative callus in Nf1null closed fractures.
In contrast, open fractures were not as advanced in healing and thus showed high levels of proliferating cells in both Nf1null and control fractures such that no significant difference was seen.
TRAP+ cells are present in Nf1null fibrous tissue not adjacent to bone surfaces
To examine whether bone resorption was increased in Nf1null fractures, we performed TRAP staining on fracture sections. Osteoclast number and surface relative to bone surface were not significantly increased in Nf1null fractures compared to controls (data not shown), but a large number of TRAP-positive osteoclast-like cells were observed within the fibrous tissue, and not adjacent to a bone surface (Fig. 7). Quantification revealed that these cells were more prevalent in open fractures, where they were observed in up to 80% of Nf1null animals (Fig. 7B). A comparable phenomenon was recently reported in NF1 patient samples.26
The findings in this study support the concept that Nf1null cells in the fracture site can significantly impair bone healing. In this mouse model, the vast majority of Nf1null fractures not only failed to heal, but the histological appearance was that of a fibrous non-union. For midshaft fractures, no impairment was seen with Nf1 deficiency in the previous Nf1+/− mouse model.12 In contrast, double inactivation of Nf1 in a midshaft fracture was able to induce a fibrous non-union. The presence of fibrous tissue is a key complication in congenital pseudarthrosis of the tibia; cultured cells from this tissue is poorly responsive to osteogenic signals and also secrete high levels of receptor activator of NF-κB ligand (RANKL) and low levels of osteoprotegerin (OPG), inducing osteoclastogenesis.27 Fibrous tissue invasion prevents ingrowth of functional bone tissue and provides inadequate biomechanical support. The generalized clinical approach to management requires the excision of this fibrous tissue. Fibrosis has been previously reported in an alternative model of Nf1 double inactivation in mice. In the Nf1 calvarial and limb specific knockout mouse, fibrosis and impaired healing were seen in a drill hole defect model.14 The precise source of fibrous tissue seen in our fracture model has yet to be fully characterized; however, we hypothesize that it results primarily from Nf1null osteoprogenitors that fail to commit to an osteogenic lineage and proliferate at the fracture site. Consistent with this concept, BrdU labeling showed high cell proliferation in the callus and in closed fractures at 3 weeks; it remained high in Nf1null fractures but was significantly lower in wild type controls. No difference was seen in open fractures with all samples showing high proliferation. This suggests that cell proliferation itself may not be the sole factor responsible for instigating a fibrous non-union.
We examined impairment of osteogenic differentiation in cell culture models. Studies with the Col2.3Cre:Nf1flox/flox mouse (also known as the Nf1 mouse) reported increased bone formation in the spinal processes.28 This is again contrary to the clinical presentation of local osteopenia in congenital pseudarthrosis of the tibia (CPT) as well as systemic osteopenia in up to 30% of NF1 patients.29, 30 However, in this mouse strain the Nf1 gene is specifically inactivated in mature osteoblasts only, leading us to hypothesize differential roles for NF1 in osteoprogenitor differentiation versus mature osteoblast function. Consistent with this, we found that Nf1null osteoprogenitors were impaired in their ability to differentiate into mature bone cells in vitro. However, when Nf1flox/flox cells were pretreated with rhBMP-2 to induce osteogenic differentiation and then subsequently treated with AdCre to induce recombination, Nf1null mature osteoblasts were able to continue an osteogenic program. From a translational standpoint, these data may be interpreted to imply that targeted pathway-based therapies would only be required to convert Nf1 progenitors into osteoblasts and would not be required for maintenance of osteoblast function.
In this study we utilized both Nf1flox/flox mice and Nf1flox/− mice and both genotypes yielded similar impairments of repair with no significant worsening in the Nf1flox/− mice. This is inconsistent with previous studies suggesting a principal role for Nf1+/− osteoclasts in deficient bone healing.18 Rather, these data indicates that the major deficiency in CPT may involve poor Nf1null bone anabolism rather than increased responsiveness of Nf1+/− osteoclast precursors. The fact that increased TRAP+ cells were seen in fibrous tissue in Nf1flox/flox mice and Nf1flox/− mice indicates that increased bone catabolism may be present, but it is likely secondary and due to signaling from Nf1null cells, more so than a specific Nf1+/− sensitivity to these signals.27 TRAP+ cells are rarely present away from bone surfaces; however, such cells have been previously reported in human fibrotic pseudarthrosis tissue, yet their origin or function is unknown.26
The AdCre system is functionally effective in impairing bone healing, but does not appear to lead to widespread ablation of Nf1 at the fracture site. Based on PCR results showing some retention of the Nf1-flox allele (functionally wild-type) and Z/AP reporter mouse studies showing limited hAP staining with AdCre injection, it would appear that impairment can be brought about by a mix of Nf1null and normally functioning cells. Mechanistically, future studies will be required to address whether this is due to increased proliferation and physical impairment (fibrosis) by Nf1null cells or due to aberrant signaling from Nf1null cells to other cells in the fracture callus. Although this model adequately represents the attenuated response to fracture seen in congenital pseudarthrosis, further work is required to elucidate the cause of the tibial dysplasia and bowing that usually precedes fracture and pseudarthrosis.
In summary, we have generated a novel fracture model that accurately recapitulates critical characteristics of a human pseudarthrosis. Local Nf1 deletion in a subset of cells early in the fracture healing process is sufficient to induce a large fibrotic callus that is highly proliferative and contains large TRAP+ like cells. Complete inactivation of Nf1 in all cells within the fracture callus is not required, with a limited number of cells being sufficient to significantly impair healing. This model provides new insights into the pathobiology of congenital pseudarthrosis, and represents a key advancement as a resource for the screening of new pathway-based adjunctive drug therapies.
All authors state that they have no conflicts of interest.
This work was supported by funding from the Children's Tumor Foundation (CTF), the Neurofibromatosis Association of Australia (http://www.nfaa.org.au), and by the Australian National Health and Medical Research Council Project Grants (#512245 and #1003478). Jad El-Hoss receives postgraduate scholarship support from the University of Sydney. David Goldhamer, Patrick Tam, Thomas Gajewski, Andreas Nagy, and Luis Parada all supplied reagents for use in this study.
Authors' roles: DGL, AS, and JEH for study design and conceptualization. JEH, LP, and MK for study conduct. JEH, TC, and NYCY for data analysis. DGL, AS, KS, JDB, and JEH for data interpretation. JEH for drafting manuscript. DGL, AS, IEA, and PB for manuscript revision and approval.