Asfotase-α improves bone growth, mineralization and strength in mouse models of neurofibromatosis type-1.

Individuals with neurofibromatosis type-1 (NF1) can manifest focal skeletal dysplasias that remain extremely difficult to treat. NF1 is caused by mutations in the NF1 gene, which encodes the RAS GTPase-activating protein neurofibromin. We report here that ablation of Nf1 in bone-forming cells leads to supraphysiologic accumulation of pyrophosphate (PPi), a strong inhibitor of hydroxyapatite formation, and that a chronic extracellular signal-regulated kinase (ERK)-dependent increase in expression of genes promoting PPi synthesis and extracellular transport, namely Enpp1 and Ank, causes this phenotype. Nf1 ablation also prevents bone morphogenic protein-2-induced osteoprogenitor differentiation and, consequently, expression of alkaline phosphatase and PPi breakdown, further contributing to PPi accumulation. The short stature and impaired bone mineralization and strength in mice lacking Nf1 in osteochondroprogenitors or osteoblasts can be corrected by asfotase-α enzyme therapy aimed at reducing PPi concentration. These results establish neurofibromin as an essential regulator of bone mineralization. They also suggest that altered PPi homeostasis contributes to the skeletal dysplasias associated with NF1 and that some of the NF1 skeletal conditions could be prevented pharmacologically.

INTRODUCTION

Mutations in the NF1 gene cause neurofibromatosis type I (NF1), a genetic disorder with an incidence of 1/3500 worldwide. This condition is characterized by malignant and non–malignant pathologies, including skeletal manifestations[1-6]. Dystrophic scoliosis, tibia bowing, bone fragility, fracture and pseudarthrosis (non–union following fracture) are skeletal conditions associated with high morbidity in this population[7-10]. Despite recent progress in our understanding of the role of NF1 in skeletal tissues, it is still unclear why and how these bone pathologies arise, raising uncertainty regarding optimal treatment[2,3].

While NF1 individuals are typically born with heterozygous mutations in NF1, loss of heterozygosity has been detected in pseudarthrosis biopsies[11], suggesting that local somatic NF1 loss of function contributes to NF1 skeletal dysplasia. This point is further supported by the relative commonality of defects observed between NF1 pseudarthrosis biopsies and the skeleton of NF1 loss–of–function conditional mouse models, which tend to recapitulate, in their entire skeleton, the genetic and cellular consequences of local NF1 loss of function occurring in human NF1 pseudarthroses. Nf1 inactivation in osteochondroprogenitors, in Nf1f/f;Prx1–cre or Nf1f/f;Col2a1–cre mice (called herein Prx-Nf1 KO or Col2-Nf1 KO mice, respectively) indeed led to reduced stature, low bone mass, tibia bowing, diaphyseal ectopic blood vessel formation and hypomineralization associated with weakened bone mechanical properties, and indicated that neurofibromin is required for normal osteoblast differentiation and for the control of Rankl expression and osteoclastogenesis[12-17]. The coexistence of Nf1–deficient osteoblasts in a Nf1 heterozygote bone microenvironment was also shown to cause bone loss and delayed bone healing in Nf1f/f;Col1a1–cre mice (Col1-Nf1 KO) via activation of TGFβ signaling[18,19]. Importantly, each of these NF1 models, as well as bone biopsies from individuals with NF1 pseudarthrosis[20], are characterized by excessive unmineralized bone matrix (osteoid), despite normal serum phosphate and calcium concentration.

Bone matrix mineralization is a tightly regulated process and requires collagen, calcium and phosphate to form hydroxyapatite, as well as tissue–nonspecific alkaline phosphatase (ALP) activity to hydrolyze pyrophosphate (PPi, a potent inhibitor of mineralization) and generate inorganic phosphate[21]. Extracellular concentrations of PPi are determined by (1) its degradation via ALP, (2) synthesis catalyzed by the nucleoside triphosphate pyrophosphohydrolase PC–1/ENPP1 (called ENPP1 herein), and (3) its transport into the extracellular milieu through the PPi channel ANK[22]. Mineralization is also controlled by Phospho1, a phosphatase that provides intracellular inorganic phosphate to generate PPi[23], and by glycoproteins such as osteopontin, which inhibits crystal nucleation on collagen fibers in mineralizing vesicles[24,25]. Multiple growth factors such as TGFβ, activin A, BMP2, IGF1, FGF2 and FGF23 are involved in bone and/or cartilage mineralization[26-34]. A common signaling pathway engaged by these factors is the RAS/ERK pathway, which is constitutively activated in cells lacking neurofibromin, the RAS–GTPase Activating Protein (GAP) encoded by NF1[35]. We thus hypothesized that neurofibromin, via its inhibitory action on RAS/ERK signaling in bone–forming cells, could be an important regulator of bone matrix mineralization.

RESULTS

Uncontrolled PPi production in Nf1–deficient bone cells

To address if and how Nf1 regulates bone mineralization, we first asked whether Nf1 ablation in bone marrow stromal cells (BMSCs) affects extracellular PPi concentrations. BMSCs from Col2-Nf1 KO mice, lacking Nf1 in osteochondroprogenitor cells, were characterized by a 60–70% lower Nf1 expression compared to WT mice (Fig. 1a), consistent with the heterogeneous nature of these cultures[36]. This lower Nf1 expression level was accompanied by a significant 70% higher extracellular PPI concentration in the conditioned medium (CM) of undifferentiated BMSC cultures compared to WT controls (Fig. 1b). Addition of a recombinant form of ALP (sALP–FcD10, 0.5 μg.ml−1, see below) to induce PPi hydrolysis significantly reduced the amount of PPi detected in both genotypes, confirming the validity of the PPi measurements.

Figure 1

Uncontrolled Ank, Enpp1, Opn expression and increased pyrophosphate production in Nf1–deficient osteoblasts

(a) Nf1 mRNA expression in BMSCs differentiated for 7, 14 and 21 days (n = 3). (b) Extracellular PPi concentration in the conditioned medium of undifferentiated BMSCs (n = 3). (c) Ank, Enpp1 and Opn mRNA expression in BMSCs treated with vehicle (DMSO) or U0126 for 24 h (n = 3). (d). Ank, Enpp1 and Opn mRNA expression in long bones, calvariae and epiphyses of 3 week–old WT (blue bars) and Col2-Nf1 KO mice (grey bars)(n = 6). (e, f) ENPP1 and ANK mRNA expression in bone marrow adherent cells from control (n = 6) and NF1 pseudarthrosis (PA, n = 9) biopsies. Blue bars: BMSCs from WT mice, grey bars: BMSCs from Col2-Nf1 KO mice, *:p < 0.05. ns: non–significant.

High extracellular PPi concentration can be generated by increased production of PPi by the ectonucleophosphatase ENPP1 and by increased cellular export through the transporter ANK. The expression of both Ank and Enpp1 mRNA (Fig. 1c, left and middle panels) and protein levels (Supplementary Fig. 1a) were higher in Nf1–deficient BMSCs compared to WT BMSCs. Osteopontin (Opn) expression was also higher in Nf1–deficient BMSCs (Fig. 1c, right panel), consistent with the reported stimulatory effect of PPi on Opn expression[25]. We obtained similar results when comparing Nf1flox/flox BMSCs cultures infected with a Cre–expressing adenovirus (Nf1–deficient cells) to control cultures infected by a GFP–expressing adenovirus (Supplementary Fig. 1b), confirming that the changes in gene expression measured in BMSCs from Col2-Nf1 KO mice were not caused by a reduced number of osteoprogenitors initially platted. Ank, Enpp1 and Opn expression was also significantly higher in long bones, calvariae and epiphyses (cartilage) from 3 week-old Col2-Nf1 KO versus WT mice (Fig. 1d), whereas Runx2 and Alpl expression was lower (Supplementary Fig. 1c). Lastly, MEK inhibition (U0126, 1 μM, 24 h) blunted the increase in Ank, Enpp1 and Opn expression observed in Nf1–deficient BMSCs (Fig. 1c and Supplementary Fig. 1b).

In an effort to assess whether these molecular findings were conserved from mice to humans, we obtained RNA from human bone adherent stromal cells prepared from 6 control and 9 NF1 tibial pseudarthrosis biopsies, and measured ENPP1 and ANK transcript levels by qPCR. Consistent with the mouse data, ENPP1 expression was significantly higher in cultured cells from NF1 pseudarthrosis tissues (Fig. 1e), despite the small number of available samples and the cell heterogeneity of these cultures. ANK expression, however, was variable between samples and not significantly different between cultures from normal and NF1 pseudarthrosis biopsies (Fig. 1f).

Mice lacking Nf1 in mature osteoblasts (Col1-Nf1 KO) have a uniform distribution of non–mineralized matrix throughout trabecular bone compartments[18], whereas mice lacking Nf1 in osteochondroprogenitors and chondrocytes are characterized by an osteoid preferentially distributed in the primary spongiosa, where osteoblasts and chondrocytes mineralize their matrix (Fig. 2a). Based on these observations and because neurofibromin is expressed in hypertrophic chondrocytes[37,38], we hypothesized that this RAS–GAP could also contribute to cartilage mineralization, which is a process important for bone growth and ossification during development and bone healing in adults. In support of this hypothesis, Col2-Nf1 KO chondrocyte high–density micromass cultures generated a typical Alcian blue–positive matrix but did not show signs of mineralization, in contrast to WT chondrocyte cultures (Fig. 2b). In addition, Ank, Enpp1 and Opn expression was significantly higher in Nf1–deficient micromass chondrocyte cultures versus WT cultures (Fig. 2c), in agreement with the data obtained from cartilaginous epiphyses, which contain a high proportion of chondrocytes (Fig. 1d). Accordingly, extracellular PPi concentration (Fig. 2d) and Enpp1 enzymatic activity (Fig. 2e) were significantly higher, whereas ALP activity was lower (Fig. 2f) in Nf1–deficient versus WT chondrocytes.

Figure 2

Altered pyrophosphate homeostasis in Nf1–deficient chondrocytes

(a) Hyperosteoidosis (pink, white arrow) in the primary spongiosa from Col2-Nf1 KO−/− mice (undecalcified sections stained by von Kossa/Van Gieson, bar: 150 μm). (b) High–density chondrocyte pellets prepared from WT and Col2-Nf1 KO pups. Proteoglycan production (top panels, Alcian blue staining) and matrix mineralization (bottom panels, von Kossa staining)(n = 3. bar: 100 μm). (c) Ank, Enpp1 and Opn mRNA expression in high–density chondrocyte pellets (n = 3). (d) Relative extracellular PPi concentration, (e) ENPP1 activity and (f) ALP activity in WT and Col2-Nf1 KO high–density chondrocytes pellets (n = 3). *:p < 0.05.

Lack of Nf1 in BMSCs impairs BMP2 osteogenic action

BMSCs isolated from Col2-Nf1 KO mice displayed, compared to BMSCs isolated from WT mice, a significantly lower differentiation potential, as measured by lower CFU–Ob colony number, TNSAP activity (Fig. 3a) and lower expression of osteoblast differentiation markers including Runx2, Alpl and Ocn (Fig. 3b). Similar results were obtained using Nf1flox/flox BMSCs infected with a cre–adenovirus (Supplementary Figs. 1d and e). In contrast to what was observed in the case of Ank and Enpp1 expression, however, MEK inhibition by U0126 (1 μM), Tremetinib or PD198306 (0.1 μM and 200 nM, respectively, data not shown) for 24 h did not correct the expression level of Runx2 or Alpl in Nf1–deficient BMSCs (Figs. 3c), indicating that the expression of these two genes is not directly controlled by neurofibromin. Extracellular PPi concentration, as well as Ank, Enpp1 and Opn expression, remained above or equal to WT controls throughout the differentiation period (Fig. 3d and e).

Figure 3

Blunted BMP2 response and osteoblast differentiation potential in Nf1–deficient osteoprogenitors

(a) BMSC differentiation analyzed by Alizarin red–S (differentiation/mineralization, CFU–Ob), crystal violet staining (cell number, CFU–F, left panel), soluble Alizarin redS/crystal violet optical density ratio (middle panel) and ALP activity/crystal violet ratio (right panel) (n = 6). (b) Runx2, Alpl and Ocn mRNA expression in BMSCs differentiated for 7, 14 and 21 days (n = 4). (c) Runx2 and Alpl mRNA expression in serum–starved BMSCs treated with vehicle (DMSO) or U0126 for 24 h (n = 6). (d) Extracellular PPi concentration/protein concentration in BMSCs differentiated for 7, 14 and 21 days (n = 4). (e) Normalized Ank, Enpp1 and Opn mRNA expression in BMSCs differentiated for 7, 14 and 21 days (n = 4). Blue bars: WT mice, grey bars: Col2-Nf1 KO mice, *:p < 0.05.

Bone morphogenic proteins (BMPs) are known for their ability to promote osteoprogenitor differentiation[39] but had limited effect on the differentiation of Nf1+/− osteoprogenitors and on bone union in Nf1+/− mice[40,41]. Recombinant hBMP2 (100 ng.ml−1) was unable to stimulate ALP activity, nor the formation of CFU–Ob in BMSC cultures from Col2-Nf1 KO mice, although it did, as expected, promote CFU–Ob formation and ALP activity in WT BMSC cultures, following 2 weeks of treatment (Fig. 4a). Smad1/5/8 phosphorylation in response to BMP2 treatment (100 ng.ml−1, 1 h) was not affected by Nf1 deficiency (Fig. 4b), indicating that the lack of stimulatory effect of BMP2 on Nf1–deficient BMSC differentiation is not caused by repression of BMP2 receptor expression, or by the production of factor(s) inhibiting canonical signaling. Treatment with rhBMP2 for 2 weeks also failed to increase the expression of Alpl, Runx2, and Col1a1 in BMSC cultures from Col2-Nf1 KO mice (Fig. 4c). However, it significantly increased the expression of Ank and Enpp1 (but not Opn) (Fig. 4d) and PPi extracellular concentration (Fig. 4e) in both WT and Nf1–deficient BMSCs. CFU–Ob formation, ALP activity (Figs. 4f and g) and the expression of Alpl and Col1a1 (Supplementary Figs. 2a and b) in Nf1–deficient BMSC cultures were higher following a 2 week–long combined treatment with the MEK inhibitor U0126 (1 μM) and BMP2 (100 ng.ml−1), but not with either of these treatments alone. This combination treatment also partially reduced the increased Ank, Enpp1 and PPi extracellular concentration detected in vehicle-treated Nf1–deficient BMSC cultures, possibly due to the antagonistic effect of these two drugs on Ank and Enpp1 expression (Supplementary Figs 2c and d).

Figure 4

BMP2 does not promote differentiation in Nf1–deficient BMSCs but exacerbates their mineralization deficit

(a) BMSC differentiation analyzed by Alizarin red–S (differentiation/mineralization, CFU–Ob) and crystal violet (cell number, CFU–F) staining (n = 3) and ALP activity (n = 3), following vehicle or BMP2 treatment. (b) Phospho–Smad1/5 induction in serum–starved BMSCs following BMP2 treatment for 1 h. Smad1/5 and β–actin served as loading control. (c and d) Alpl, Runx2, Col1a1, Ank, Enpp1 and Opn mRNA expression following BMP2 treatment for 2 weeks (n = 3). (e) Extracellular PPi relative concentration (normalized to protein concentration) in the conditioned medium of BMSCs treated with BMP2 for 24 h (n = 3). (f and g) BMSC differentiation analyzed by Alizarin red–S (differentiation/mineralization, CFU–Ob) and crystal violet (cell number, CFU–F) staining (f, n = 3) and ALP activity (g, n = 3) following treatment with vehicle or BMP2 or U0126 or both for 2 weeks. Blue bars: WT mice; grey bars: Col2-Nf1 KO mice. *:p < 0.05 versus WT in the same treatment group; #:p < 0.05 versus vehicle in the same genotype group.

sALP–FcD10 improves bone growth and BMD in Col2-Nf1 KO mice

If excessive extracellular PPi levels cause the mineralization deficit observed in Col2-Nf1 KO mice, then reducing PPi concentration should have beneficial effects on matrix mineralization. This is experimentally possible by inhibiting PPi generation or increasing its catabolism. The latter approach was chosen because PPi is a substrate for ALP and a recombinant form of human ALP (sALP–FcD10 or Asfotase–α) is clinically available to treat ALPL–deficient subjects with hypophosphatasia[42,43]. We thus treated WT and Nf1–deficient BMSCs with vehicle or sALP–FcD10 (0.5 mg.ml−1) in osteogenic condition for 14 days and assessed matrix mineralization. As predicted, sALP–FcD10 increased matrix mineralization in both genotypes, although the relative increase was more pronounced in cultures from Col2-Nf1 KO versus WT mice (Fig. 5a), and despite the persistent differentiation deficit of Nf1–deficient BMSCs in the presence of sALP–FcD10 (Supplementary Fig. 3a). This treatment reduced Opn expression in Nf1–deficient BMSCs (Supplementary Fig. 3a), in agreement with the known stimulatory effect of PPi on Opn expression[25].

Figure 5

sALP–FcD10 improves bone growth and cortical bone parameters in growing Col2-Nf1 KO mice

(a) BMSC matrix mineralization (CFU–Ob) and number (CFU–F) analyzed by Alizarin red–S and crystal violet staining, respectively (n = 3) following vehicle or sALP–FcD10 treatment for 2 weeks. (b–f) Bone growth (b, naso–anal length), vertebral (c, bar: 250 μm) and tibial (d, bar: 250 μm) bone mineral density (X–rays), cortical thickness (e, Ct.Th, μCT), epiphyseal diameter (f, white arrow, bar: 45 μm, μCT) and hypertrophic zone von Kossa–positive calcified Bone Volume/Tissue Volume (hBV/TV, histology) in Col2-Nf1 KO newborn pups treated daily by sALP–FcD10 for 18 days (n > 8 mice/group). *:p < 0.05 versus WT; #:p < 0.05 versus vehicle in the same genotype group.

Based on these encouraging results, we treated Col2-Nf1 KO newborn mice daily by subcutaneous injections of sALP–FcD10 (8.2 mg/kg/day) for 18 days[44,45]. Col2-Nf1 KO mice exhibit short stature, low bone mass, mineralization, cortical thickness and mineral density (BMD), and high cortical porosity[37]. Following this short treatment (dictated by the relatively high lethality of these mice at weaning), a significant 73% increase in the size of mutant mice (Fig. 5b) and a clear increase in vertebral and tibial BMD were observed on radiographs (Fig. 5c and d). sALP–FcD10 significantly increased mid–diaphyseal cortical bone thickness, as measured by 3D–microcomputed tomography (μCT) (Fig. 5e), partially rescued the formation of secondary ossification centers, expanded tibia metaphyseal envelopes and increased the amount of calcified matrix in the growth plate hypertrophic zone of Col2-Nf1 KO mice (hBV/TV, Fig. 5f). Despite the seemingly pronounced effects of sALP–FcD10 observed by radiography and μCT, tibia cortical tissue mineral density and mineral–to–collagen ratio (Supplementary Fig. 3b and c) were not increased following treatment.

sALP–FcD10 increases bone mineralization in Osx-Nf1 KO mice

Because Col2-Nf1 KO mice manifest severe developmental phenotypes that limit their survival, and thus the duration of treatments, we generated mice in which Nf1 can be ablated postnatally in osteoprogenitors, using the inducible Tet-off–based Osx–cre transgenic mice[46] crossed to Nf1 mice[47]. This new mouse model makes it possible to dissect the mechanisms by which postnatal Nf1 ablation impairs bone homeostasis, without complications arising from developmental phenotypes. Osx–Nf1 mice had a size undistinguishable from WT littermates upon doxycycline administration (i.e. cre–recombinase repression) from conception to day 14 (Fig. 6a) and had normal phosphate, calcium and 25OH vitamin D serum concentrations (Supplementary Table 1). Osx–cre–mediated Nf1 ablation in osteoprogenitors at post–natal day 14 following doxycycline withdrawal, as seen in Col2-Nf1 KO mice, caused hyperosteoidosis (Fig. 6b), low bone mass (Fig. 6c), higher femoral diaphyseal cortical porosity (Fig. 6d), lower cortical thickness, mid–shaft moment of inertia and cortical TMD (Fig. 6e–g). Cortical mineral–to–collagen ratio measured by Raman spectroscopy (Fig. 6h) was also lower in Osx-Nf1 KO mice, and femurs from Osx-Nf1 KO mice were mechanically weaker than those from WT controls, as measured by a 3–point bending tests (Supplementary Table 2).

Figure 6

sALP–FcD10 improves trabecular bone mass, mineralization and bone structure in Osx-Nf1 KO mice

(a) Size of two month–old WT and Osx-Nf1 KO mice following doxycycline (Doxy) treatment from conception to P14. (b) Femoral hyperosteoidosis (pink stain following von Kossa/van Gieson staining), Osteoid Volume/Bone Volume ratio (OV/BV), Osteoid Surface/Bone Surface ratio (OS/BS) and Osteoid Thickness (O. Th) in WT and Osx-Nf1 KO mice and rescue by sALP–FcD10 administration for 6 weeks (histomorphometric analyses, bar: 150 μm). (c) Femoral Bone Volume/Tissue Volume (BV/TV) in WT and Osx-Nf1 KO mice and rescue by sALP–FcD10 administration (μCT). (d) Cortical porosity in Osx-Nf1 KO mice and partial beneficial effect of sALP–FcD10 administration (μCT). (e) Femoral cortical thickness in WT and Osx-Nf1 KO mice (μCT). (f) Moment of inertia in WT and Osx-Nf1 KO mice and rescue by sALP–FcD10 administration (μCT). (g) Cortical Tissue Mineral Density (TMD) in WT and Osx-Nf1 KO mice (μCT). (h) Mineral–to–Collagen ratio (ν1 phosphate/Proline) in WT and Osx-Nf1 KO mice and rescue by sALP–FcD10 administration (Raman spectroscopy). (n > 8 mice/group). *:p < 0.05 versus WT; #:p < 0.05 versus vehicle in the same genotype group.

To assess the effect of sALP–FcD10 on the skeleton of this mouse model, we administered sALP–FcD10 daily from 2 weeks of age (at the time of Nf1 ablation) and for 6 weeks. sALP–FcD10 significantly increased trabecular BV/TV and moment of inertia, assessed by μCT (Fig. 6c and f), as well as femoral stiffness, modulus and peak force, measured by 3–point–bending (Supplementary Table 2), and led to a non–significant trend for higher cortical femoral thickness (Fig. 6e). sALP–FcD10 improved bone mineralization in Osx-Nf1 KO mice, as measured by a drastic 73% reduction in osteoid volume per bone volume, a 65% reduction in osteoid surface per bone surface, a 53% decrease in osteoid thickness (Fig. 6b) and a 20% increase in mineral–to–collagen ratio (Fig. 6h).

Discussion

We show here that the RAS–GAP activity of neurofibromin in the bone mesenchymal lineage restrains the expression of Enpp1 and Ank, two main genes controlling PPi homeostasis, and that increasing pyrophosphate catabolism through enzyme therapy significantly improves bone mineralization and bone mechanical properties in mouse models of NF1 skeletal dysplasia. These results, along with suggestive evidence of conservation of function between mice and humans, support the causal role of increased PPi levels in the etiology of NF1 hyperosteoidosis, and position neurofibromin as a critical and obligatory regulator of cartilage and bone mineralization. They also provide pre–clinical evidence that some of the most clinically challenging NF1–related skeletal maladies may be amenable to prevention.

Hyperactive TGFβ signaling was proposed to cause bone loss and to delay bone healing in mice deficient for Nf1 in mature osteoblasts and heterozygote for Nf1+/− globally[19]. TGFβ is also known to stimulate ERK activity, Ank and Enpp1 expression and to increase PPi concentration in WT chondrocytes[48,49]. Therefore, NF1–deficient BMSCs may contribute cell–autonomously and/or in a hyperactive TGFβ paracrine fashion to the extraphysiological skeletal accumulation of PPi and to the impaired osteoblast differentiation and matrix mineralization observed in the setting of NF1. The beneficial effect of sALP–FcD10 on bone growth, mineralization and strength observed in this study suggests that PPi accumulation and abnormal mineralization are important components of NF1 bone dysplasia. However, further studies will be necessary to determine the evolution and contribution of all the cellular defects typical of Nf1–deficient bone cells on bone mass and strength over extended periods of treatment with sALP–FcD10, as this drug does not correct the differentiation phenotype of Nf1-deficient osteoblasts. Although TGFβ blockade might theoretically serve as a target to promote bone union in individuals with NF1 pseudarthrosis, the cancer–prone status of this pediatric population and the known tumor suppressor activity of TGFβ signaling limit this therapeutic approach[50]. Our results, on the other hand, suggest that stimulation of pyrophosphate catabolism through enzyme therapy could be applied on a more chronic basis prior to fracture to strengthen the NF1 dysplastic bones and prevent their mechanical failure.

The mineralization deficit of Nf1–deficient BMSCs could be detected in immature BMSCs, prior to their differentiation into osteoblasts. Therefore, this phenotype cannot be attributed to the reduced differentiation potential of Nf1–deficient BMSCs, although the latter certainly contributes to the low bone mass phenotype observed in the two NF1 mouse models used in this study. It is also worth noting that BMP2 treatment, without the need of ERK blockade, stimulated the expression of Ank and Enpp1, as well as extracellular PPi concentration, in Nf1–deficient BMSCs, as shown previously in WT cells[28]. This observation could explain why rhBMP2 alone did not improve bone healing in NF1 mouse models[40,41] and bone union in individuals with NF1 pseudarthrosis[51-53].

Our results indicate that Nf1–deficient BMSCs are not responsive to BMP2 with regard to their differentiation potential and suggest that this defect may in part underlie their inability to differentiate. In addition, the response of Nf1–deficient BMSCs to BMP2 with regard to Ank and Enpp1 expression suggests that neurofibromin is not the sole negative regulator of the RAS/ERK signaling pathway upstream of these two genes. These results also indicate that the stimulatory effect of BMP2 on osteoprogenitor differentiation requires controlled ERK signaling by neurofibromin.

It is unknown to what extent poor matrix mineralization contributes to the low BMD, tibia bowing, poor mechanical properties and possibly pseudarthrosis observed in children with NF1. Although local PPi concentration could not be quantified, the observed increase in the expression of ENPP1 measured in BMSCs extracted from NF1 pseudarthrosis biopsies, as well as the presence of thick osteoid seams on histological sections[20], supports conservation of function between mice and humans.

Pseudarthrosis and possibly dystrophic scoliosis can currently be treated only by invasive, and often repetitive, surgical orthopedic interventions[2,3]. Most approaches to date are corrective in nature, and only bracing techniques are available to reduce the incidence and severity of these complications. Of major interest is the possibility that sALP–FcD10, if applied preventatively, might improve mineralization, growth, architecture and mechanical properties of dysplastic bones affected by NF1 and, thus, limit their likelihood of deformation and fracture. This latter point is particularly important, as the current standard for treatment is limited to avoidance of prophylactic surgery and early long–term bracing to prevent fracture until skeletal maturity is reached. It is worth emphasizing that sALP–FcD10 is bone–targeted and already successfully utilized in the clinic to treat children with hypophosphatasia[42]. Therefore, its potential use in the context of NF1 skeletal dysplasia could be accelerated compared to other drugs.

Methods

Animals and drugs

All procedures were approved by the Vanderbilt University Medical Center Institutional Animal Care and Use Committee (IACUC). WT and Col2-Nf1 KO mice were generated by crossing Nf1flox/flox mice and Nf1flox/+; α1(II) collagen-Cre breeders[54,55]. Nf1flox/flox mice and Nf1flox/flox mice; α1(II) collagen-Cre mice were used as WT and cKO, respectively. Osx-Nf1 KO mice were generated by breeding doxycycline-fed Osx-cre; Nf1flox/flox mice with Nf1flox/flox breeders[47]. All mice were on a C57BL/6 background. sALP-FcD10 (Asfotase Alfa, Alexion Pharmaceuticals) was described previously[56]. Briefly, mineral-targeting recombinant tissue nonspecific alkaline phosphatase (ALP, sALP-FcD10) was produced in CHO cells by modifying the coding sequence of human ALPL. The GPI anchor sequence of the hydrophobic C-terminal domain of human ALPL was removed to generate a soluble, secreted enzyme (sALP). Then the human ALPL ectodomain sequence was extended with the coding sequence encoding the Fc region of human IgG1 (Fc). Finally the C-terminus of the Fc region was extended with ten aspartic acid residues (D10). The dose of 8.2 mg.kg−1 per day was selected because it was previously shown to be efficacious in short-term (16 days) efficacy study in Akp2−/− mice[56]. The specific activity of the lot used in the present study was 878 Units.mg−1. sALP-FcD10 was administered subcutaneously for the periods of time indicated in the text.

Human subjects

The study was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center, the Rizzoli Orthopaedic Institute (Bologna, Italy) and Vanderbilt University. Bone tissues were obtained from 9 patients with NF1 and tibial pseudarthrosis (aged between 7 months and 18 years) and 6 controls from children without NF1 who underwent surgery for congenital dysplasia of the hip without any other coexisting pathology (n = 3)[57] or scoliosis (n = 3) (aged between 3.3 and 17 years). Diagnosis of pseudarthrosis was based on radiographic and clinical findings. Diagnosis of NF1 was performed according to the criteria presented at the National Institute of Health Consensus Development Conference on Neurofibromatosis [http://consensus.nih.gov/1987/1987Neurofibramatosis064html.htm].

Cell culture

Mouse BMSCs were extracted from long bones by spinning down diaphyses at 1500 rpm for 3 min. Cells were then counted, plated at a density of 106 cells/well (12w plates) or 2×106 cells/well (6 well plates), and grown for 7 days in αMEM supplemented with 10% FBS, 100 I.U./ml penicillin, 100 μg/ml streptomycin (Cellgro, Manassas, VA, USA). At day 7, differentiation and mineralization was induced by the addition of 50 μg/ml ascorbic acid and 10 mM β-glycerophosphate, and the media was refreshed every 2–3 days. BMSCs differentiation and mineralization were assessed by ALP activity and Alizarin red S staining, respectively, using standard protocols.

Primary chondrocytes were extracted from 4-day-old pup rib bones. The cartilaginous part of the rib was dissected and soft tissues removed, then digested by collagenase D (3mg/ml, Roche, USA) and 0.25% trypsin/ethylenediaminetetraacetic acid (EDTA) (Gibco, USA) in DMEM for 3 h. At confluence, 5×10 μl drops of concentrated cells (2×107 cells/ml) were plated in 6-wells. After 2 hours of incubation, 2 ml of complete cell culture medium was delicately added. Cells were differentiated in DMEM supplemented with 10% FBS, 100 IU.ml−1 penicillin, 100 μg.ml−1 streptomycin, 50 μg.ml−1 of ascorbic acid and 10 mM β-glycerophosphate.

Human cells extracted from bone marrow[57] or bone tissue were maintained in alpha MEM supplemented with 10% FBS, 100 U.ml−1 penicillin, 0.1 mg.ml−1 streptomycin at 37 °C in a 5% CO2-humidified atmosphere. Cells from bone tissues were digested overnight with collagenase before platting. After 4 days, non-adherent cells were removed and adherent bone cells were grown until confluence or passaged before RNA extraction.

Adenovirus Infection of BMSCs

BMSCs were isolated from Nf1flox/flox mice and seeded at a density of 106 cells/well in 12w plates. At 40% confluence, cells were incubated in complete culture medium (α-MEM, 10% FBS and 100 IU.ml−1 penicillin) containing either Ad5-CMV-GFP or Ad5-CMV-cre (Vector development lab, Baylor College of Medicine) at 2.5×109 PFUs. After 2 days of incubation, the medium was refreshed with complete culture medium. Nf1 recombination efficiency was determined according to Wang et al[37].

Serum vitamin D, calcium and phosphate assays

Blood samples were collected from WT and Osx-Nf1 KO mice at sacrifice. Vitamin D, phosphate and calcium concentration in mouse serum was determined using a 25OH-Vitamin-D ELISA Assay kit (Eagle Biosciences, cat# VID31-K01), a Phosphate Assay kit (BioVision, cat # k410–500) and a Calcium Assay kit (BioVision, cat# k380–250), respectively, according to the manufacturer’s instructions.

PPi and PC-1 assays

PPi release in cell-conditioned media (ePPi) was measured radiometrically using differential adsorption on activated charcoal of Uridine-diphospho-D-glucose [6-3H] (Cat #NET1163250UC, Perkin Elmer) as previously described[49,58,59]. Forty microliters of conditioned medium (or blank control) and 120 μl of assay solution (57 nM of Tris acetate, pH7.6; 5.2 mM MgAc; 18.6 μM Glucose 1,6-diphosphate (G1,6DP); 9 μM Uridine-diphosphoglucose (UDPG); 4 μM β-Nicotinamide adenine dinucleotide (NAD+); 0.136 U. Uridine-diphosphoglucose pyrophosphorylase (UDPGPP); 0.5 U. phosphoglucomutase; 0.5 U. Glucose-6-phosphate dehydrogenase (G6PD); 0.02 μCi 3H-UDPG) were incubated at 37°C for 1h, then adsorbed on 200 μl of charcoal for 10 minutes on ice. After centrifugation at 14,000 rpm for 10 minutes, 100 μl of the supernatant was transferred into a vial containing 5 ml of Bio-safe II for radioactivity count. PPi levels were normalized by protein concentration in cell lysates in each well. Measurements were performed in triplicate and similar results were obtained from at least 3 independent experiments.

ENPP1 activity was determined using 1.5 mM of the synthetic chromogenic substrate thymidine 5′-monophosphate p-nitrophenyl ester in reaction buffer (100 mM Tris/HCl, pH 8.0, 130 mM NaCl, and 15 mM MgCl2) incubated at 37°C for 30 min. The reaction was terminated by the addition of 50 μl 4N NaOH. Product formation was monitored by measurement of absorbance at 405 nm. ENPP1 activity in each well was normalized by cell number. Measurements were performed in triplicate and from at least 3 independent experiments.

RT-qPCR and genomic PCR

Total RNA was extracted using TRIzol (Invitrogen, Grand Island, NY, USA) and cDNAs were synthesized from 1 μg of RNA following DNase I treatment using the high-capacity cDNA reverse-transcription kit (Applied Biosystems, USA). Quantitative PCR (qPCR) was performed by using TaqMan or SYBR green gene expression assays. The probe and primer sets for mouse Runx2 (Mm00501578_m1); Alpl (Mm00475834_m1); Ank (Mm00445047_m1); Enpp1 (Mm00501097_m1); Opn (Mm00436767_m1), Igf1 (Mm01228180_m1), human ANKH (Hs00219798_m1) and human ENPP1 (Hs01054040_m1) and the normalizers Hprt (Mm00446968_m1); human GAPDH (Hs99999905_m1) were obtained from Applied Biosystems (Foster City, CA, USA). The SYBR green primers were: Opn (forward; CTCCTTGCGCCACAGAATG, reverse; TGGGCAACAGGGATGACA), Nf1 (forward; GTATTGAATTGAAGCACCTTTGTTTGG, reverse; CTGCCCAAGGCTCCCCCAG); Ocn (forward; ACCCTGGCTGCGCTCTGTCTCT, reverse; GATGCGTTTGTAGGCGGTCTTCA) and Col1a1 (forward; GACATCCCTGAAGTCAGCTGC, reverse; TCCCTTGGGTCCCTCGAC). Specificity of amplification was verified by the presence of a single peak on the dissociation curve. Amplification conditions are available upon request. Measurements were performed in triplicate and from at least 3 independent experiments.

For genotyping, genomic DNA was isolated from tail tips by sodium hydroxide digestion, and PCR was performed using primers P1, P2 and P4, as described by Zhu et al [54]. The Col2a1 collagen-Cre transgene was detected using the fwd: GAGTTGATAGCTGGCTGGTGGCAGATG and reverse: TCCTCCTGCTCCTAGGGCCTCCTGCAT primers.

Western blot analyses

Whole cell lysates were separated by SDS-PAGE electrophoresis according to standard protocols. Nitrocellulose membranes were probed with the indicated antibody using standard protocols (monoclonal anti-β-actin antibody (Sigma cat# AC-74, dilution 1: 5000), anti-Phospho-Smad1/5 antibody (Cell Signaling cat#9516S, dilution 1:1000), anti-Smad1+Smad5 antibody (Abcam cat# ab75273, dilution 1:1000), anti ENPP1/PC-1 (Aviva Systems Biology, cat# OAEB02445, dilution 1:500) and anti-ANK (Origen, cat# TA325111, dilution 1:1000).

Histology

Static histomorphometry measurements were performed as previously described in accordance with standard nomenclature[60], using the Bioquant Analysis System (Nashville, TN, USA) on 5 μm undecalcified methymethacrylate sections. Calcified cartilage BV/TV was measured in the growth plate hypertrophic region following von Kossa and van Gieson staining.

X-rays and μCT analyses

Radiographs were obtained using a digital cabinet X-ray system (LX-60, Faxitron X-Ray, USA). μCT analyses were performed using a Scanco μCT 40 system (Scanco Medical, Bassersdorf, Switzerland). Tomographic images were acquired at 55 kVp and 145 mA with an isotropic voxel size of 12 μm and at an integration time of 250 ms with 500 projections collected per 180° rotation.

Raman Spectroscopy

To ensure anatomic consistency of data collection site among bones of differing length, mid-shaft vessel perforations were used as landmarks. Spectra were obtained with 5 accumulations of 20 s exposures with 20 mW laser power at a spot size of 1.5 μm in diameter. Spectra were processed via least squares modified polynomial fit[61] and smoothed for noise using an 2nd order Savitsky-Golay filter[62]. System Raman shift calibration was accomplished using a neon lamp and a silicon standard. Silicon standard measurements before and after data acquisition ensured wavenumber calibration consistency. Spectral intensities for known Raman peaks and peak ratios were extracted using custom Matlab software (Mathworks, Natick, MA) to generate markers of bone composition for mineralization (v1 Phosphate/Proline) and crystallinity (crystal grain size and perfection, determined by the inverse full-width at half maximum intensity of the v1 Phosphate peak).

Biomechanical Testing

Hydrated samples were tested in three-point bending with a span of 8mm at a rate of 3 mm.min−1 as per[63]. Force and displacement was measured from a 100 N load cell, and from the linear variable displacement transformer of the material testing system (Dynamight 8841, Instron, Canton, OH). Structural properties were extracted from force-displacement curves by custom Matlab algorithms (Mathworks, Natick, MA). Material properties were calculated by accounting for structure by utilizing cross-sectional area and moment of inertia as measured by μCT.

Statistical analysis

Depending on whether data per group passed the Shapiro-Wilk normality test or whether standard deviations were not different among the groups (Bartlett’s test), one-way analysis of variance (ANOVA) or the Kruskal-Wallis Test (non-parametric) was used to determine whether differences existed in μCT -, Raman-, and biomechanical-derived properties among the experimental groups. When differences existed at p < 0.05, post-hoc, pair-wise comparisons were tested for significance in which the p-value was adjusted (padj < 0.05) by Holm-Sidak’s method or Dunn’s method (non-parametric). Statistical analysis was performed using GraphPad PRISM (v6.0a, La Jolla, CA). Data are provided as mean +/− SD.

Supplementary Figure 1: Mineralization and differentiation potential ofNf1–deficient BMSCs.

(a) ANK and ENPP1 expression in BMSCs isolated from WT and Col2-Nf1 KO mice. β–actin serves as loading control (n = 2 in triplicates). (b) Ank, Enpp1 and Opn mRNA expression in serum–starved BMSCs infected with GFP–(blue bars) or cre–adenoviruses (grey bars) and treated with vehicle (DMSO) or U0126 for 24h (n = 3). (c) Runx2 and Alpl mRNA expression in long bones, calvariae and tibial epiphyses from 3 week–old WT and Col2-Nf1 KO mice (n = 6). (d) BMSC differentiation analyzed by Alizarin red–S (differentiation/mineralization, CFU–Ob) and crystal violet (cell number, CFU–F) staining (n = 4) and ALP activity/OD570 (n = 4) following infection with GFP–(control) or cre–adenovirus. (e) Normalized expression of osteoblast marker genes and control genes assessed in the same cultures (n = 4). Blue bars: BMSCs from WT mice, grey bars: BMSCs from Col2-Nf1 KO mice, *:p < 0.05 versus WT.

Supplementary Figure 2: BMP2 and U0126 combination treatment increases the differentiation of Nf1–deficient BMSCs.

(a–d) mRNA expression and (e) extracellular PPi concentration in BMSCs supplemented with vehicle or BMP2 or U0126 or both for 14 days (n = 3). Blue bars: BMSCs from WT mice, grey bars: BMSCs from Col2-Nf1 KO mice. *:p < 0.05 versus WT and #:p < 0.05 versus vehicle in the same genotype group.

Supplementary Figure 3: Effect of sALP–FcD10 in Col2-Nf1 KO mice.

(a) mRNA expression in BMSCs treated with vehicle or sALP–FcD10 for 14 days in osteogenic condition (n = 3). Blue bars: BMSCs from WT mice, grey bars: BMSCs from Col2-Nf1 KO mice. (b and c) Cortical Tissue Mineral Density (b, Ct.TMD) and mineral–to–collagen ratio (c, v1PO4/Proline) following sALP–FcD10 administration to Col2-Nf1 KO newborn pups for 18 days (n > 8 mice/group). *:p < 0.05 versus WT, #:p < 0.05 versus vehicle in the same genotype group.

Supplementary Table 1. Serum phosphate, calcium and 25OH vitamin D3 in WT and in Osx-Nf1 KO mice

(n > 8 mice/group).

Supplementary Table 2. sALP–FcD10 improves bone mechanical properties inOsx-Nf1 KO mice.

Femoral mid–shaft mechanical properties quantified by 3 point–bending tests (n > 8 mice/group. *:p.<0.05 versus WT; #:p.<0.05 versus vehicle in the same genotype group.