Modulation of spine fusion with BMP-2, MEK inhibitor (PD0325901), and zoledronic acid in a murine model of NF1 double inactivation
Justin D. Bobyn a, b, Nikita Deo a, b, David G. Little a, b, Aaron Schindeler a, b, *
aOrthopaedic Research & Biotechnology Unit, The Children’s Hospital at Westmead, Sydney, NSW, Australia
bDiscipline of Paediatrics and Child Health, Faculty of Medicine, University of Sydney, Sydney, NSW, Australia

a r t i c l e i n f o

Article history:
Received 30 March 2020 Received in revised form 5 May 2020
Accepted 24 May 2020 Available online xxx
a b s t r a c t

Background: Spine fusion is a common procedure for the treatment of severe scoliosis, a frequent and challenging deformity associated with Neurofi bromatosis type 1 (NF1). Moreover, defi ciencies in NF1- Ras-MEK signaling affect bone formation and resorption that in turn impacts on spine fusion outcomes. Methods: In this study we describe a new model for AdCre virus induction of Nf1 defi ciency in the spines
flox/flox
of Nf1 mice. The virus is delivered locally to the mouse spine in a fusion procedure induced using BMP-2. Systemic adjunctive treatment with the MEK inhibitor (MEKi) PD0325901 and the bisphosph- onate zoledronic acid (ZA) were next trialed in this model.
null
Results: AdCre nulldelivery resulted in abundant fi brous tissue (Nf1 þ393%, P < 0.001) and decreased
marrow space (Nf1 ti 67%, P < 0.001) nullcompared to controls. While this did not signifi cantly impact on
the bone volume of the fusion mass (Nf1 ti 14%, P ¼ 0.999 n.s.), the presence of fi brous tissue was anticipated to impact on the quality of spine fusion. Multinucleated TRAP þ cells were observed in the fi brous tissues seen in Nf1null spines. In Nf1null spines, MEKi increased bone volume (þ194%, P < 0.001) whereas ZA increased bone density (þ10%, P < 0.002) versus BMP-2 alone. Both MEKi and ZA decreased TRAP þ cells in the fi brous tissue (MEKi ti 62%, P < 0.01; ZA ti 43%, P ¼ 0.054). No adverse effects were seen with either MEKi or ZA treatment including weight loss or signs of illness or distress that led to premature euthanasia.
Conclusions: These data not only support the utility of an AdCre-virus induced knockout spine model, but also support further investigation of MEKi and ZA as adjunctive therapies for improving BMP-2 induced spine fusion in the context of NF1.
Crown Copyright © 2020 Published by Elsevier B.V. on behalf of The Japanese Orthopaedic Association.
All rights reserved.

1.Introduction

Neurofi bromatosis type 1 (NF1) is a common autosomal domi- nant genetic disorder with an estimated incidence of 1 in 2500 live births [1]. The NF1 gene is expressed in numerous tissues and can thus affect a wide range of clinical features, each with variable penetrance. Individuals with NF1 are prone to tumor development, including cutaneous and subcutaneous neurofi bromas, plexiform neurofi bromas, optic gliomas, and peripheral malignant nerve sheath tumors. A subset of NF1 individuals are born with or develop osseous manifestations, with scoliosis and tibial pseudarthrosis being the most clinically challenging to treat [1].

The reported incidence of scoliosis in NF1 varies, but has been reported to be between 11% [2] and 64% [3], and may be classified into non-dystrophic or dystrophic types. Dystrophic scoliosis is associated with short, acute curves [4], and present earlier and with more severe deformity [5]. Intervention for scoliosis can be further complicated by intraspinal and paraspinal neurofibromas, that can be present in up to 50% of individuals with NF1-associated dystrophic scoliosis. Paraspinal neurofi bromas are often found to the concavity of the curve and may present as massive soft tissue tumors which may invade the muscle and compress surrounding structures [6e10]. Similarly, intraspinal tumors may represent a source of significant morbidity as they compress neurological structures. Dural ectasia is a widening of the thecal sack which may erode bony and soft paraspinal tissues and is seen as vertebral scalloping on plain radiographs. In the presence of severe angula-

* Corresponding author. Orthopaedic Research & Biotechnology Research Build- ing, The Children's Hospital at Westmead, Locked Bag 4001, Westmead, NSW, 2145, Australia. Fax: þ61 2 98453180
E-mail address: [email protected] (A. Schindeler).
tion, it may serve a protective function by limiting cord compres- sion [7,11].

https://doi.org/10.1016/j.jos.2020.05.016

0949-2658/Crown Copyright © 2020 Published by Elsevier B.V. on behalf of The Japanese Orthopaedic Association. All rights reserved.

Surgery remains the primary means of correcting and pre- venting further deformity for NF1 individuals with worsening scoliotic curves. Positive surgical outcomes are primarily depen- dent upon establishing robust fusion between adjacent vertebrae [12]. However, patients with NF1 can show impaired bone repair at the spine. Pseudarthrosis in the postoperative spine is seen in as many as 25% of cases [13], with as few as 7% of cases developing a clinically robust fusion [3].
NF1 is associated with increased Ras activity, which leads to downstream overactivity in pathways such as MEK. Nevertheless, to date the bone healing defi ciencies of NF1 have been largely approached using conventional bone anabolic agents (e.g. Bone Morphogenetic Proteins; BMPs) and anti-resorptive drugs (e.g. bisphosphonates). Indeed, positive outcomes were reported in a clinical tibial pseudarthrosis case series employing off-label BMPs and bisphosphonates [14]. In a murine spine fusion model, BMP and bisphosphonate were found to increase in bone volume and min- eral density of the fusion mass in Nf1þ/ti and wild type mice [15]. However, such approaches may poorly deal with the fi brosis seen in NF1 (which is normally aggressively resected in tibial pseudarth- rosis) and is a feature of localized Nf1null fracture mouse models [16].
In this study we further describe a new murine model that re- capitulates the challenging clinical nature of NF1 spine fusion. NF1 is inactivated in the spine of Nf1fl ox/fl ox mice by employing cre- expressing adenovirus (AdCre) system, similar to prior tibial non- union studies [16]. This fusion model shows a lack of bony union and extensive fi brosis. Next, interventions combining rhBMP-2 with the MEK inhibitor PD0325901 or the bisphosphonate Zole- dronic Acid (ZA) were tested. Outcome measures were radiographic and histological. We propose that this model could be used not only for therapeutic testing, but as an archetypical approach to gene deletion in the context of spine fusion.

2.Materials & methods

2.1.Animal colony

We sourced Nf1fl ox/fl ox mice (conditional Nf1 alleles knocked out by exposure to Cre recombinase), as originally generated by Prof Luis Parada, from the National Cancer Institute (NCI) mouse re- pository (Bethesda, MD, USA). The Nf1þ/ti strain (a heterozygous strain with one Nf1 knockout allele) was supplied directly by Prof Parada. Both colonies were maintained on a C57/Bl6 background. All animal experiments were approved by the local institution animal ethics committee.

2.2.Adenovirus production and purification

Human embryonic kidney cells, HEK293, were grown in T175 culture flasks to 80% confluence and transduced with AdCre (gift from Dr Thomas Gajewski) [17]. AdCre was purified from cells after 72 h after lysis and DNase treatment. The cell supernatant was ultracentrifuged at 35,000 rpm in a standard CsCl gradient in a Beckman SW-41Ti rotor. Purifi ed virus was dialyzed in a Slide-A- Lyzer Dialysis Cassette, 10 kDa cutoff (Thermo Scientific, USA) for up to three days in a 10 mM Tris (pH 8.0) solution. Virus was titered using the Adeno-X titration kit (Clontech, USA) and used or frozen at ti 80 ti C in 10% glycerol.
2.3.Surgical model

fl ox/fl ox
Female Nf1 mice aged 8e10 weeks were subjected to a surgical spine fusion procedure using a published model [18]. An- imals were randomly assigned to experimental groups (Fig. 1).

Surgery was performed by a single surgeon over two sessions with support from an experienced animal technician. 15 mm midline longitudinal incision was made centered at the level of the iliac crests. Mice were anaesthetized with an intraperitoneal injection of ketamine (35 mg/kg) and xylazine (5 mg/kg) with a 27G needle. Anesthesia was maintained with inhaled isofl uorane as required. Paraspinal muscles at the level of L4-L6 were swept laterally to expose the bony anatomy which was decorticated with a diamond tipped burr. A collagen sponge loaded with either saline, saline/
rhBMP-2 or AdCre virus solution/rhBMP-2 was inserted into this paraspinal pocket. The rhMBP-2 was delivered at a dose of 10 mg in each sponge. Animals were group caged and checked daily for wound integrity and animal wellbeing. Culling occurred at day 21 by CO2 asphyxiation and spines from L1 to lower sacrum, were harvested for radiographic and histological analysis.

2.4.Drug treatments

Oral drug delivery with PD0325901 was performed by oral dosing in strawberry jelly [19] Jelly was made with 0.8% DMSO (SigmaeAldrich, St. Louis, MO, USA), 16% Splenda (Splenda® Low Calorie Sweetener, JohnsoneJohnson Pacifi c Pty, NSW, Australia), 9.6% gelatine (Davis Gelatin, GELITA Australia Pty, NSW, Australia) and 7.9% fl avoring (QUEEN Flavoring Essence Imitation Strawberry, Queen Fine Foods Pty, QLD, Australia). Jelly was loaded with PD0325901 dissolved in DMSO to be delivered as a daily dose per animal of 10 mg/kg. Dosing began on day 1 post op and continued until cull at day 21.
Subjects in the bisphosphonate treatment group received a five doses of 0.02 mg/kg ZA (total dose 0.1 mg/kg) (Novartis AG) by biweekly subcutaneous injection commencing 3 days post- operatively, as previously published [20].

Fig. 1. (A) Study design showing group sizes and local treatments (Saline or AdCre virus ± BMP-2) and systemic adjunctive dosing with MEKi (PD0325901), ZA or no treatment (Nil). (B) Schematic showing posterolateral intertransverse spine fusion surgery where rhBMP-2 and/or AdCre are delivered to the lumbar spine via implan- tation of paired collagen sponges.

J.D. Bobyn et al. / Journal of Orthopaedic Science xxx (xxxx) xxx 3

2.5.MicroCT analysis

MicroCT analysis was using methods previous established for murine models of spine fusion [15] Spines and surrounding soft tissue were harvested at day 21 and fixed in a 4% paraformaldehyde solution for 24h at 4 ti C and then stored in 70% ethanol. Samples were scanned using a SkyScan 1174 compact microCT scanner (SkyScan, Kontich, Belgium). Samples were scanned in 70% ethanol at 21.3 um magnification, 0.5 mm aluminium fi lter, 50 kV X-ray tube voltage and 800 mA tube electric current. The dicom images were recon- structed using NRecon software, version 1.5.1.5 (SkyScan). The fusion mass was isolated from native bone by manually drawing regions of interest (ROI) and thresholds set using CTAn software version 1.10.03. A global threshold of 0.3 g/mm3 was set for bone tissue. Variables calculated by CTAn software included bone volume (BV, mm3) and bone mineral density (BMD, g/cm3). Representative 3- dimensional fractures were reconstructed with sagittal slices using CTVol Realistic Visualization software version 2.1.0.0 (SkyScan).
2.6.Tissue histology

Following microCT scanning, samples were decalcified in 0.34M EDTA (pH 8.0) solution at 4 ti C on an agitation tray for 30 days with solution changes every 2e3 days. Spines were cut axially at the level of the largest portion of the fusion mass, typically coinciding with a vertebral level of L5-6. Samples were embedded in paraffi n and sectioned in the axial plane at a thickness of 5 mm and were stained via Picro Sirius Red/Alcian Blue [16] to differentiate bone and cartilage. Adjacent sections were stained for tartrate-resistant acid phosphatase (TRAP) expression to highlight osteoclasts, and counterstained with light green. Histology was quantified using BioQuant Osteo software 2016 v16.1.60 (BioQuant Nashville, TN, USA). Quantifi cation was performed in representative regions of interest (ROI) with outcomes including bone volume (BV/ROI), fi brous tissue volume (FV/ROI), cartilage volume (CV/ROI), marrow volume (MV/ROI) and Osteoclast number (OcN/ROI).

2.7.Statistical analysis

Statistical analysis was performed using GraphPad Prism 7 (GraphPad, La Jolla, USA) using one-way ANOVA with Tukey's multiple comparisons test. Statistical signifi cant was calculated using a cut-off of P < 0.05.
3.Results

3.1.Fibrosis following spine fusion surgery is considerably increased in Nf1null spines

Spine fusion surgery was performed involving the implantation of collagen sponges containing rhBMP-2 to promote bone forma- tion. All mice were of the Nf1fl ox/fl ox genotype, with groups featuring Nf1null tissue having AdCre virus co-administered during surgery via the implanted sponges. No morbidity or mortality was noted in any of the test animals up until the cull date of 21 days. At the experimental endpoint, all of the animals possessed a palpable fusion mass (Fig. 2).
Fibrous tissue showed a significant increase in Nf1null spines compared to untreated controls, as quantifi ed on histological sec-
null
tions (Nf1 þ393%, P < 0.001) (Fig. 3B). Furthermore, fibrous tissue accounted for 48% of the fusion mass area in Nf1null spines, where it was only 10% of the fusion mass area in control spines. Nf1null spines showed a significant decrease in marrow space compared to con-
null
trols (Nf1 ti 67%, P < 0.001) (Fig. 3C).
Qualitatively, Nf1null spines exhibited fibrous tissue invading the remnants of the collagen sponge. This was rarely observed in control

Fig. 2. Representative microCT reconstructions of spines harvested at day 21 day (post-op) showing AdCre treatment results in a visually reduced fusion mass. Axial cross sections are taken from the midpoint of the fusion mass where most substantial.

spines, where this space was largely replaced by trabecular bone and marrow. Within the fibrous tissue of Nf1null spines, abundant multi- nucleated TRAP þ cells were seen (Fig. 5), comparable to prior models of pseudarthrosis [16] and clinical biopsies [21].

3.2.Bone volume and mineral density were increased by MEKi and ZA in Nf1null spines

Bone volume, as quantifi ed by microCT, showed no signifi cant
null decrease in Nf1null spines relative to controls (Nf1 ti 14%, P ¼ 0.999 n.s.) (Fig. 4A). This contrasts with prior findings in models of Nf1þ/ti spine fusion [15] and Nf1null pseudarthrosis [16].
However, both adjunctive pharmacotherapeutic interventions impacted on the fusion mass, as quantitatively assessed by microCT (Figs. 2 and 4). In Nf1null spines, MEKi dosing increased bone for- mation. Bone volume was increased by þ194% (P < 0.001) compared to saline controls. Similarly, ZA treatment increase bone density in Nf1null spines by þ10% (P < 0.002) relative to saline treated controls (Fig. 4B).
In wild type spines, MEKi dosing was also found to increase bone volume independent of the Nf1null genotype (þ301%, P < 0.001). Bone mineral density was also increased (þ13%, P < 0.05) (Fig. 4B).
The number of TRAP þ cells in the tissue surrounding the collagen sponge remnant was assessed for both treatments. MEKi signifi cantly reduced the number of TRAP þ cells found within the fibrous tissue adjacent to the collagen sponge remnant in wild type mice (ti62%, P < 0.03), with an identical decrease noted in Nf1null spines (ti 62%, P ¼ 0.001) (Fig. 6). Similarly, ZA reduced the number of TRAP þ cells surrounding the collagen sponge remnant (ti 43%, P ¼ 0.054) (Fig. 6).
3.3.MEKi and ZA treatments did not reduce fibrosis in Nf1null spines

Neither MEKi nor bisphosphonate treatment had any signifi cant effect on the amount of fibrous tissue (n.s.) (Fig. 3B).

Fig. 3. Histological analysis showing mean areas of (A) Bone, (B) Fibrous Tissue, (C) Marrow, and (D) Cartilage relative to equivalently sized regions of interest (ROI). *P < 0.05 **P < 0.01 ***P < 0.001.

Fig. 4. MicroCT analysis showing (A) Bone Volume and (B) Bone Mineral Density. *P < 0.05.

In wild type spines, a considerable increase was seen in the relative proportion of cartilage in the fusion mass (23-fold) (Fig. 3D). This is consistent with prior reports of retained cartilage with PD0325901 in mouse tibial fracture healing [22]. Notably, no change was seen in cartilage in the Nf1null spines.
4.Discussion

A variety of approaches have used historically to model the manifestations of NF1, and in this paper we describe a combination of genetic manipulation and surgery. While low mineral density are

J.D. Bobyn et al. / Journal of Orthopaedic Science xxx (xxxx) xxx 5

Fig. 5. Histological sections stained with Picro Sirius Red (bone) and Alcian Blue (cartilage) or and TRAP stain (osteoclasts) with Light Green counterstain. Scale bar ¼ 300 mm. (For interpretation of the references to colour in this fi gure legend, the reader is referred to the Web version of this article.)

site of treatment and aberrant multinucleated TRAP þ cells within the fi brous tissues.
The fi brous tissue proliferation was a major complication in the Nf1null spine model, with 47% of the mean fusion mass identifi ed as mesenchymal or fi brous. Nevertheless, AdCre virus treatment did not significantly decrease bone volume in the fusion mass compared to no virus controls. Unlike Nf1þ/ti and Nf1null fracture models [22], the spine fusion procedure employs a substantive dose of rhBMP-2 that stimulates considerable bone formation. We hy- pothesize that this dose of rhBMP-2 in the mouse is sufficient to overcome the limited bone anabolism associated with NF1, similar to that seen with the Nf1þ/ti heterozygous line [15]. However, in a clinical setting, rhBMP-2 doses may be insufficient to yield a robust bony fusion mass, and the high degree of fibrosis seen in this model may similarly impede successful union.
Considering the role of the Ras-MEK-ERK cascade in bone ho- meostasis, interruption of this pathway was expected to shift the equilibrium in favor of bone formation. In tibial fracture models, MEK inhibitor PD0325901 increased the volume of mineralized

Fig. 6. Quantifi cation of osteoclast number at the perimeter of collagen sponge or sponge remnant. The center of the fusion mass was chosen as the ROI in these animals. MEK inhibitor PD0325901 significantly decreased osteoclast number. *P < 0.05.

reported in adults and children with NF1 [23], focal orthopedic problems such as tibial pseudarthrosis has been linked to localized double inactivation of NF1 [24]. Moreover, a recent study suggests that the contribution of somatic mutations may contribute to the pathology of dystrophic scoliosis in NF1 in some cases [25], which prompted our trialing of a model featuring local double inactiva- tion in the spine. This was achieved using a Cre recombinase expressing adenovirus to drive deletion of Nf1-flox alleles, analo- gous to a prior technique used in the tibia to model NF1 pseu- darthrosis [22]. In this prior study, genetic recombination was validated by PCR and by fl uorescent reporter. While this study does not explicitly demonstrate recombination, phenotypically the outcomes were comparable, with increased fi brous tissue at the
bone in a fracture callus in wild type and Nf1null fractures [22]. In the spine, PD0325901 increased the bone volume in both wild type and Nf1null fusion masses, and by a similar amount. MEKi treatment did not improve on fusion rate as it was complete in all groups receiving rhBMP-2. The complete fusion rate seen with 10 mg rhBMP-2, even in Nf1null fusion masses, could be considered a limitation of the model. In the case of wild type fractures, PD0325901 led to cartilage retention at the fracture site, possibly due to suppression of matrix metalloproteinase activity [16]. In the context of spine fusion, rhBMP-2 induced fusion mass were thought to form with minimal cartilage, yet increased cartilage retention was seen in MEK-treated in wild type spine fusions. This was not a complication seen in the Nf1null fusion setting. Notably, PD0325901 was found to decrease the number of osteoclasts (TRAPþ) cells in the fusion mass of all mice, which suggest it has anti-resorptive effects.
The bisphosphonate ZA was used as a positive control in this system as it has reliably increased rhBMP-2 bone masses in Nf1-

deficient fracture [26,27] and spine fusion [15] models. The effect of ZA was chiefl y to increase bone mineral density, rather than bone volume; this was the inverse of that seen with PD0325901 treat- ment. These data can be interpreted to show that PD0325901 led to larger fusion masses with similar density to control BMP-2 induced fusion masses, whereas ZA created smaller but denser masses. Mechanistically, this could be attributed to MEKi promoting a wider zone of osteoblastic bone formation, whereas ZA decreased pre- mature osteoclastic resorption of the fusion mass resulting in higher bone density. There is scope in future studies for the adjunctive treatments of MEKi and ZA to be combined.
In terms of clinical translation, both PD0325901 and ZA both have potential utility in augmenting rhBMP-2 induced spine fusion. ZA is an established and well tolerated therapy for improving bone mineral density that has been used off-label for improving BMP-induced NF1 pseudarthrosis repair [14]. PD0325901 is currently undergoing clin- ical trials for treating NF1 tumors, however has not been applied in an orthopedic scenario. MEK inhibitors are tolerated in a cancer setting [28], however their side effect profile is significant with many pa- tients exhibiting hematological dyscrasias (leucopenia, thrombocy- topenia, and anemia) and symptoms such as rash, fatigue, diarrhea, epistaxis, and blurred vision. Nevertheless, brief and strategic dosing centered around operative windows has not yet been trialed. Such trials will need to still consider risks associated with associated tu- mors, such as paraspinal plexiform neurofibromas, which may be stimulated by osteoinductors such as BMPs [29].
5.Conclusions

In this paper we describe a new model for spine fusion in in- dividuals with NF1. While this model has a number of signifi cant limitations, such as a lack of an initial scoliosis and an inability to produce a failure of fusion, key features such as extensive fi brotic and osteoclastic infiltration were recapitulated. Treatment with both ZA and the MEK inhibitor PD0325901 produced positive ef- fects on the model, indicating that symptomatic and pathway- focused approaches both have potential as therapeutic modalities. Based on its more widespread use in treating metabolic bone dis- orders and off-label use in orthopedic medicine, ZA is more likely to be adopted into patient care.

Author Contributions

JB performed surgery, was responsible for data collection and analysis, and drafted the initial manuscript. ND manufactured re- combinant virus. AS and DGL were responsible for conceptualizing and overseeing the study and editing the fi nal manuscript. All au- thors have read and approved the final manuscript.

Declaration of Competing Interest

AS reports non-fi nancial support from N8 Medical, outside the submitted work. DGL reports personal fees from OrthoPediatrics, outside the submitted work. JB and ND have no disclosures.

References

[1]Ferner RE. Neurofi bromatosis 1 and neurofi bromatosis 2: a twenty first cen- tury perspective. Lancet Neurol 2007 Apr;6(4):340e51.
[2]Akbarnia BA, Gabriel KR, Beckman E, Chalk D. Prevalence of scoliosis in neurofi bromatosis. Spine 1992 Aug;17(8 Suppl):S244e8.
[3]Crawford AH. Pitfalls of spinal deformities associated with neurofibromatosis in children. Clin Orthop Relat Res 1989 Aug;(245):29e42.

[4]Burk Jr DL, Brunberg JA, Kanal E, Latchaw RE, Wolf GL. Spinal and paraspinal neurofi bromatosis: surface coil MR imaging at 1.5 T1. Radiology 1987;162(3): 797e801.
[5]Koptan W, ElMiligui Y. Surgical correction of severe dystrophic neurofi bro- matosis scoliosis: an experience of 32 cases. Eur Spine J 2010 Sep 1;19(9): 1569e75.
[6]Hu Z, Liu Z, Qiu Y, Xu L, Yan H, Zhu Z. Morphological differences in the vertebrae of scoliosis secondary to neurofi bromatosis type 1 with and without paraspinal neurofi bromas. Spine 2016 Apr;41(7):598e602.
[7]Tsirikos AI, Saifuddin A, Noordeen MH. Spinal deformity in neurofi bromatosis type-1: diagnosis and treatment. Eur Spine J 2005 June 1;14(5):427e39.
[8]Khong PL, Goh WH, Wong VC, Fung CW, Ooi GC. MR imaging of spinal tumors in children with neurofi bromatosis 1. AJR Am J Roentgenol 2003 Feb;180(2): 413e7.
[9]Egelhoff JC, Bates DJ, Ross JS, Rothner AD, Cohen BH. Spinal MR fi ndings in neurofi bromatosis types 1 and 2. Am J Neuroradiol 1992;13(4):1071e7.
[10]Thakkar SD, Feigen U, Mautner VF. Spinal tumours in neurofi bromatosis type 1: an MRI study of frequency, multiplicity and variety. Neuroradiology 1999 Sep;41(9):625e9.
[11]Calvert P, Edgar M, Webb P. Scoliosis in neurofi bromatosis. The natural history with and without operation. J Bone Joint Surg Br 1989 March 1;71-B(2): 246e51. 1989.
[12]Crawford AH, Parikh S, Schorry EK, Von Stein D. The immature spine in type-1 neurofi bromatosis. J Bone Joint Surg Am 2007 Feb;89(Suppl 1):123e42.
[13]McMaster MJ, James JI. Pseudoarthrosis after spinal fusion for scoliosis. J Bone Joint Surg Br 1976 Aug;58(3):305e12.
[14]Birke O, Schindeler A, Ramachandran M, Cowell CT, Munns CF, Bellemore M, Little DG. Preliminary experience with the combined use of recombinant bone morphogenetic protein and bisphosphonates in the treatment of congenital pseudarthrosis of the tibia. J Child Orthop 2010 Dec;4(6):507e17.
[15]Bobyn J, Rasch A, Kathy M, Little DG, Schindeler A. Maximizing bone forma- tion in posterior spine fusion using rhBMP-2 and zoledronic acid in wild type and NF1 defi cient mice. J Orthop Res 2014 Aug;32(8):1090e4.
[16]El-Hoss J, Sullivan K, Cheng T, Yu NY, Bobyn JD, Peacock L, Mikulec K, Baldock P, Alexander IE, Schindeler A, Little DG. A murine model of neurofi- bromatosis type 1 tibial pseudarthrosis featuring proliferative fi brous tissue and osteoclast-like cells. J Bone Miner Res 2012 Jan;27(1):68e78.
[17]Zha Y, Shah R, Locke F, Wong A, Gajewski TF. Use of Cre-adenovirus and CAR transgenic mice for effi cient deletion of genes in post-thymic T cells. J Immunol Methods 2008 Feb 29;331(1):94e102.
[18]Bobyn J, Rasch A, Little DG, Schindeler A. Posterolateral inter-transverse lumbar fusion in a mouse model. J Orthop Surg Res 2013 Jan 23;8:2.
[19]Sullivan K, El-Hoss J, Little DG, Schindeler A. JNK inhibitors increase osteo- genesis in Nf1-defi cient cells. Bone 2011 Dec;49(6):1311e6.
[20]Schindeler A, Ramachandran M, Godfrey C, Morse A, McDonald M, Mikulec K, Little DG. Modeling bone morphogenetic protein and bisphosphonate com- bination therapy in wild-type and Nf1 haploinsuffi cient mice. J Orthop Res 2008;26(1):65e74.
[21]Heerv€a E, Alanne MH, Peltonen S, Kuorilehto T, Hentunen T, V€a€an€anen K, Peltonen J. Osteoclasts in neurofi bromatosis type 1 display enhanced resorption capacity, aberrant morphology, and resistance to serum depriva- tion. Bone 2010;47(3):583e90.
[22]El-Hoss J, Kolind M, Jackson MT, Deo N, Mikulec K, McDonald MM, Little CB, Little DG, Schindeler A. Modulation of endochondral ossifi cation by MEK in- hibitors PD0325901 and AZD6244 (Selumetinib). Bone 2014 Feb;59:151e61.
[23]Lammert M, Kappler M, Mautner VF, Lammert K, Storkel S, Friedman JM, Atkins D. Decreased bone mineral density in patients with neurofi bromatosis 1. Osteoporos Int 2005 Sep;16(9):1161e6.
[24]Stevenson DA, Moyer-Mileur LJ, Murray M, Slater H, Sheng X, Carey JC, Dube B, Viskochil DH. Bone mineral density in children and adolescents with neurofi bromatosis type 1. J Paediatrics 2007;150(1):83e8.
[25]Margraf RL, VanSant-Webb C, Mao R, Viskochil DH, Carey J, Hanson H, D’Astous J, Grossmann A, Stevenson DA. NF1 somatic mutation in dystrophic scoliosis. J Mol Neurosci 2019 May;68(1):11e8.
[26]Schindeler A, McDonald MM, Bokko P, Little DG. Bone remodeling during fracture repair: the cellular picture. Semin Cell Dev Biol 2008;19(5):459e66.
[27]Deo N, Cheng TL, Mikulec K, Peacock L, Little DG, Schindeler A. Improved union and bone strength in a mouse model of NF1 pseudarthrosis treated with recombinant human bone morphogenetic protein-2 and zoledronic acid. J Orthop Res 2018 Mar;36(3):930e6.
[28]Shapiro GI, Hilton J, Gandi L, Chau N, Cleary J, Wolanski A, Anderson A, Beardslee B, Hassinger F, Bhushan K, Downey E, Gibson J, Pruitt-Thompson S, Muzikansky A, Barry S, Feeney N, Paweletz C, Oxnard G, Supko J, J€anne P, Wong K-K, Johnson B. Abstract CT046: phase I dose escalation study of the CDK4/6 inhibitor palbociclib in combination with the MEK inhibitor PD- 0325901 in patients with RAS mutant solid tumors. Can Res 2017;77(Supp 13):CT046e.
[29]Steib JP, Bouchaib J, Walter A, Schuller S, Charles YP. Could an osteoinductor result in degeneration of a neurofi broma in NF1? Eur Spine J 2010 Jul;19(- Suppl 2):S220e5.