Evidence of Increased Bone Resorption in Neurofibromatosis Type 1 Using Urinary Pyridinium Crosslink Analysis

Abstract

Although neurofibromatosis type 1 (NF1) is a neuro-cutaneous disorder, skeletal abnormalities such as long-bone dysplasia, scoliosis, sphenoid wing dysplasia, and osteopenia are observed. To investigate the role of bone resorption as a mechanism for the bony abnormalities, we selected urinary pyridinium crosslinks (collagen degradation products excreted in urine) as a measure of bone resorption in NF1. Bone resorption was evaluated by quantitative assessment of the urinary excretion of pyridinium crosslinks [pyridinoline (Pyd) and deoxypyridinoline (Dpd)]. Total (free plus peptide-bound) pyridinium crosslinks from the first morning urines from 59 NF1 children (ages 5–19) were extracted and analyzed (17 children with a localized skeletal dysplasia, and 42 without). The data were compared with a healthy reference population without NF1 (n = 99). Multivariate analyses, controlling for age showed statistically significant increases for Dpd (p < 0.001) and the Dpd/Pyd ratio (p < 0.001) in NF1 individuals with and without a skeletal dysplasia. NF1 children have an increase in the urinary excretion of pyridinium crosslinks, reflecting increased bone resorption. The effects of NF1 haploinsufficiency likely contribute to abnormal bone remodeling, either directly or indirectly by aberrant Ras signaling, potentially predisposing NF1 individuals to localized skeletal defects.

Neurofibromatosis type 1 (NF1), a common autosomal dominant disorder affecting ≈1/3500 individuals worldwide, has variable expressivity. Clinical manifestations include café-au-lait macules, intertriginous freckling, Lisch nodules, neurofibromas, optic pathway tumors, and distinctive osseous lesions (1–3). The prototypical skeletal manifestations of NF1 are proportionate short stature, macrocephaly, long-bone dysplasia, progressive scoliosis, and sphenoid wing dysplasia. Long-bone dysplasia most often affects the tibia and presents with anterolateral bowing often leading to fracture and nonunion or pseudarthrosis (4,5). The long-bone dysplasia in NF1 is very distinctive, and the presence of tibial pseudarthrosis alone should raise the potential diagnosis of NF1, as 50–80% of individuals with pseudarthrosis have NF1 (6–8). Scoliosis is the most common orthopedic manifestation in NF1 with reports documenting between 10 and 33% of NF1 individuals having scoliosis (9). Other osseous manifestations of NF1 include bone cysts, spinal canal widening, vertebral body narrowing, rib-penciling, vertebral scalloping with dural ectasias, and decreased bone mineral density. In isolation, each skeletal abnormality associated with NF1 is rare, but, as a whole, the osseous defects are relatively frequent.

The NF1 gene, located on the long arm of chromosome 17, encodes the protein neurofibromin, which is a Ras-GAP protein (10). The “tumor suppressor” properties of this protein do not easily explain the mesodermally derived osseous manifestations observed in NF1. The frequent association of osseous dysplasias seen in NF1 suggests, however, that the neurofibromin-ras signal transduction pathway plays a role in the cellular processes of bone in NF1.

Several studies have reported decreased bone mineral density in NF1 individuals (11–16), thus pointing to an underlying disorder of skeletal homeostasis and mineralization in humans with NF1. One of these studies reported that all postmenopausal NF1 women in the cohort had either osteoporosis or osteopenia (13). Peripheral quantitative computed tomography modalities have also demonstrated that NF1 individuals display a different muscle mass, which may impact bone architecture (17). These generalized findings suggest that haploin-sufficiency of the NF1 gene contributes to abnormal regulation of bone remodeling.

Bone is a dynamic organ system with a delicate balance of turnover and regeneration. The strength of bone is due in part to the crosslinking of collagen molecules. Inferences to the dynamics of bone can be obtained through the analysis of markers of bone resorption and formation. Many markers are readily available and have shown good correlation between predicted and measured bone mass (18,19). Pyridinoline (Pyd) and deoxypyridinoline (Dpd), which are hydroxylysine-derived crosslinks of mature collagen, are excreted in urine, both in free and peptide-bound forms, as a result of collagen degradation (20). They have been shown to be sensitive markers of bone resorption and correlate well with other established markers of bone resorption (21–23). Large increases have been shown in patients with high bone turnover states including patients with osteoporosis and osteoporotic fractures, femur fractures, Paget disease, and hyperparathyroidism (20–23). A specific abnormal pattern of urinary excretion of pyridinium crosslinks has also been observed in Ehlers-Danlos syndrome type VI, an inherited disorder of collagen metabolism with kyphoscoliotic changes (24,25). Little is known about the pathogenesis of the osseous abnormalities in NF1; the constellation of decreased bone mineralization and poor fracture healing suggest a disruption in normal bone remodeling. Therefore, we selected urinary pyridinium crosslink excretion analysis to indirectly assess bone resorption in individuals with NF1.

MATERIALS AND METHODS

Subjects

Physical examinations and medical histories were obtained on a cohort of 59 individuals (ages 5–19 y) who fulfilled the diagnostic criteria for NF1 (1,2). NF1 individuals were recruited from an NF1 clinic at the University of Utah. NF1 individuals with conditions and circumstances known to influence bone health (e.g., recent fracture, illnesses requiring systemic steroids, anorexia, pregnancy, lactation, oral contraception, or hormone replacement therapy) were not included. Pyridinium crosslink data from a local regional cohort of healthy children (n = 99) without NF1 or known orthopedic conditions were used for comparison (ages 1–17 y).

NF1 individuals were classified into two groups: NF1 individuals with and without a skeletal dysplasia. A skeletal dysplasia in this study population was defined as the presence of long-bone dysplasia, scoliosis, and/or sphenoid wing dysplasia based on physical examination and review of medical records. All NF1 individuals were examined at the University of Utah General Clinical Research Center. Long-bone dysplasia was defined as anterolateral bowing of the tibia, fibula, radius, and/or ulna.

Institutional Review Board approval at the University of Utah was obtained. Written informed consent was obtained from the parents or guardians of the children who served as subjects and, when appropriate, assent or consent from the participants themselves.

Measurements

Urine from two first morning voids was obtained from all individuals for the extraction and analysis of the total free and the peptide-bound Pyd and Dpd by high performance liquid chromatography (HPLC). To release the peptide-bound crosslinks, the urine samples were hydrolyzed in 6 M HCl, under vacuum, for 16 h at 150°C. Pyridinium crosslinks were then extracted from the hydrolysate by fractionation through a cellulose column, according to established procedures (26). The eluates containing Pyd and Dpd were lyophilized, reconstituted in 1% heptafluorobutyric acid, and analyzed by reverse-phase HPLC, using a Waters 2695 Alliance HPLC System, equipped with a fluorescence detector (Waters 2475; emission = 297 nm, excitation = 395 nm) and controlled by a computerized unit (Empower Software). The reverse phase column used for the separation of pyridinium crosslinks was a Waters Nova-Pak C18 column (4 μm; 15 cm × 3.9 mm). Eluant A was 0.18% n-heptafluorobutyric acid in 12% acetonitrile. Eluant B was 0.18% heptafluorobutyric acid in 100% acetonitrile. The column was equilibrated in 95% A and 5% B for 3 h before injection of the samples. The samples were eluted for 15 min with the same isocratic solvent composition at a flow rate of 0.7 mL/min. Pyridinium crosslink concentration was calculated using a 4-levels calibration curve obtained with an external standard. Standards for Pyd and Dpd were purchased from Quidel Corporation, San Diego, CA. Urinary pyridinium crosslink concentration was normalized to urinary creatinine, measured by a Beckman Creatinine Analyzer 2 and calculated as umol/mol creatinine. Two first-morning urine samples were obtained to adjust for potential variations in a single sample.

Data analysis

Data for Pyd, Dpd, and Dpd/Pyd ratios, and age from the NF1 individuals with and without a skeletal dysplasia were compared with each other and with the control group using the unpaired t test for data with normal distributions, and with Mann-Whitney’s U for data with non-normal distributions. Multivariate linear regression modeling was then used to model the comparison of the two NF1 groups and the controls on the dependent variables Pyd, Dpd, and Dpd/Pyd with control for the confounding effect of age. The resulting multivariate linear regression models were then used to generate least square means for all three groups. Additionally, full regression models (three groups and age) were compared with reduced models (age) to generate estimates of variance accounted for by the three groups. Statistical analyses were conducted using Stata 9.2 (College Station, TX).

RESULTS

Average age of individuals with NF1 was 10.6 y (median, 10.5 y; range, 5–19 y). Seventeen of 59 individuals with NF1 had a skeletal dysplasia (1 with sphenoid wing dysplasia; 1 with long-bone dysplasia; 12 with scoliosis; and 3 with scoliosis and long-bone dysplasia). The remaining 42 NF1 individuals did not have a history of long-bone dysplasia, scoliosis, or sphenoid wing dysplasia. Average age of controls was 9.1 y (median, 9 y; range, 1–17 y).

Multivariate linear regression analysis adjusting for age showed statistically significant differences (p < 0.001) in Dpd and the Dpd/Pyd ratio between groups (n = 17 for individuals with NF1 and a skeletal dysplasia; n = 42 for individuals with NF1 and no skeletal dysplasia; n = 99 for controls without NF1). Least square means, reported in Table 1, showed moderate increases of Pyd in NF1 individuals with and without a skeletal dysplasia compared with controls, but did not reach statistical significance when comparing all groups. Similar analysis of Dpd showed a more dramatic increase in NF1 individuals with and without a skeletal dysplasia compared with controls and an increase of Dpd in NF1 individuals with a skeletal dysplasia compared with NF1 individuals without a skeletal dysplasia (Table 1). The Dpd/Pyd ratio was increased by approximately 19% in individuals with NF1 without a skeletal dysplasia compared with the control group, and approximately 38% in individuals with NF1 with a skeletal dysplasia compared with the control group. These elevations in the Dpd/Pyd ratio were statistically significant (p < 0.001) and consistent with increased bone resorption in the NF1 patients (Table 1).

Table 1.

Pyridinium crosslink analysis

Dependent variable	Individuals with NF1 with skeletal dysplasia			Individuals with NF1 without skeletal dysplasia			Individuals without NF1			p
	LS mean	SE	N	LS mean	SE	N	LS mean	SE	N
Pyd (umol/mol creatinine)	232.6	16.4	17	220.9	8.3	42	208.3	8.1	99	NS
Dpd (umol/mol creatinine)	71.0	4.2	17	58.9	2.1	42	46.8	2.1	99	<0.001
Dpd/Pyd ratio	0.31	0.009	17	0.27	0.005	42	0.23	0.005	99	<0.001

Comparison of urinary pyridinoline (Pyd) and deoxypyridinoline (Dpd) between individuals with neurofibromatosis type 1 (NF1) with skeletal dysplasia, individuals with NF1 without skeletal dysplasia, and individuals without NF1. p-values generated via multivariate linear regression controlling for age.

LS, least squares; SE, standard error; NS, not significant.

The comparisons of R² values between the reduced and full multivariate model are shown in Table 2. The differences seen in the models for the dependent variables, Dpd and Dpd/Pyd ratio, show substantial increases in R² values, indicating that the variable representing the grouping of NF1 individuals and the controls accounts for a significant portion of the variation present in the dependent variables.

Table 2.

R² for reduced and full multivariate models

Dependent variable	R2 reduced model	R2 full model	Difference
Pyd (umol/mol creatinine)	0.2834	0.2828	−0.0006
Dpd (umol/mol creatinine)	0.2151	0.3172	0.1021
Dpd/Pyd ratio	0.0030	0.2512	0.2482

Pyd, pyridinoline; Dpd, deoxypyridinoline.

DISCUSSION

Children and adolescents with NF1 had statistically significant increases in Dpd reflecting increased collagen degradation in children with NF1. In addition, the Dpd/Pyd ratio was elevated in the NF1 individuals consistent with increased bone turnover. Interestingly, individuals with Ehlers-Danlos syndrome type VI, who have normal amounts of pyridinium crosslinks but have a significantly increased Dpd/Pyd ratio (24), are in part differentiated from other types of Ehlers-Danlos syndrome based on the presence of skeletal abnormalities characterized by a progressive kyphoscoliosis usually present in the first year of life, osteoporosis, and clubfeet (27). Pyd is found in many tissues including both bone and cartilage, but Dpd is most abundant in bone and dentine and reflects more the status of bone (24,25,28,29). The observation that the Dpd/Pyd ratio is elevated in NF1 children and adolescents indicates a preferential increase in bone resorption rather than a generalized collagen breakdown.

Although decreased bone mineral density has been reported in individuals with NF1 (11–16), we are not aware that they display a substantially increased number of fractures compared with the general population, although long-bone fractures with poor bone healing is observed in patients with localized long-bone dysplasia. Potentially, decreased bone mineral density could predispose individuals with NF1 to clinically undetected microfractures. Microfractures are resorbed by osteoclasts and can alter trabecular architecture (30–32). The observed increase in the excretion of urinary pyridinium crosslinks in our cohort of children and adolescents with NF1 may be the result of microfractures leading to an abnormal microarchitecture.

There is evidence that somatic loss of the nonmutant NF1 allele in NF1 heterozygous individuals leads to localized skeletal abnormalities as double inactivation of NF1 was observed in pseudarthrosis tissue from patients with NF1 with tibial bowing and fracture with nonunion (33). In this cohort, 29% of individuals with NF1 had scoliosis, long-bone dysplasia, and/or sphenoid wing dysplasia. Although active fractures were not present, the contribution of scoliosis, long-bone dysplasia, and/or sphenoid wing dysplasia on the excretion of pyridinium crosslinks is unknown. Therefore, in this study NF1 individuals were divided into those with and without scoliosis, long-bone dysplasia, and/or sphenoid wing dysplasia. The NF1 individuals without a skeletal dysplasia still had increased Dpd and Dpd/Pyd ratios compared with controls, suggesting that NF1 haploinsufficiency contributes to increased bone resorption and represents a generalized abnormality of bone remodeling in the background of NF1. However, the precise role neurofibromin plays in the growth and development of bone in individuals with NF1 is not well understood.

Given the paucity of studies on bone remodeling in the NF1 human model, insights into the role of neurofibromin in osteoclast and osteoblast functioning must be taken from the murine model. It is known that neurofibromin directly impacts the Ras-signaling pathway, which interacts with multiple signaling pathways, several of which are important in bone. For example, transforming growth factor-beta increases neurofibromin mRNA (34), murine Nf1^± mast cells secrete elevated concentrations of transforming growth factor-beta (35), fibroblast growth factors activate the Ras/mitogen-activated phosphorylation kinase pathway (36), and inactivation of the SHP2-Ras-mitogen-activated phosphorylation kinase pathway in mice results in enhanced bone formation after an increase in osteoclast activity suggesting a dissociation of the intercellular communications between osteoclasts and osteoblasts (37,38). Specific investigations using the Nf1 haploinsufficient transgenic mouse model has shown abnormalities in myeloid cells. Yang et al. (39) showed that monocytes from Nf1^± mice have an increased potential to mature into multinucleated osteoclasts, and that the osteoclasts have increased adhesive and lytic properties, which was also observed in a small cohort of two individuals with NF1. This increase in osteoclast formation and lytic activity in the murine model and the two NF1 individuals (39) is in concert with our findings of increase bone resorption in children and adolescents with NF1.

Because of the complexities of bone remodeling, the observed increase in bone resorption in our cohort and the previously described abnormalities in the myeloid lineages do not fully explain the pathophysiology of the skeletal defects of NF1. Additional animal studies have shown that deficiency of neurofibromin also impacts differentiation of mesenchymal progenitor cells. Yu et al. (40) reported that Nf1^± committed osteoprogenitors exhibited premature apoptosis and higher proliferation, and Wu et al. (41) showed impaired osteoblast differentiation in Nf1^± mesenchymal stem/progenitor cells. The effects of neurofibromin deficiency on skeletal morphogenesis and remodeling likely vary depending on timing of expression, potentially impacted by early somatic inactivation of NF1 in the normal allele in a subset of mesenchymal cell lineages. This is evidenced by the discordant phenotypes observed in various Nf1 transgenic mouse models. For example, mice lacking neurofibromin in osteoblasts (Nf1_ob^−/−) show increased bone formation without long-bone bowing (42), whereas mice lacking neurofibromin in undifferentiated mesenchymal cells of the developing limb (Nf1^Prx1) display tibial bowing with a high degree of porosity (36). The Nf1_ob^−/− mice also showed increased urinary excretion of pyridinium crosslinks (42), which is consistent with our results in children.

Investigation of the effects of downstream targets of the neurofibromin-Ras signal transduction pathway on osteoprogenitor cells will be important in selecting targeted therapies for the localized skeletal defects of NF1. With the advancement of transgenic mouse models that partially recapitulates the human NF1 skeletal phenotype such as the long-bone bowing in the Nf1^Prx1 mouse model described by Kolanczyk et al. (36), human clinical trials with therapeutic agents identified through animal studies are likely to develop. Given that decreased bone mineral density has been reported previously in individuals with NF1, and the herein reported data show evidence of increased bone resoprtion, currently used therapeutic agents within the general population such as bisphosphonates are potential candidates. Schindeler et al. (43) used bone morphogenetic protein and bisphosphonate combination therapy in the mouse model showing an increase of net bone production in vivo in Nf1^± mice. Therefore, single agent therapy may not be the most appropriate treatment strategy, and further studies will be necessary to characterize the human NF1 skeletal phenotype more accurately, understand the effects of increased Ras signaling on bone remodeling, and understand the safety efficacy of each candidate agent before proceeding with clinical trials.

Pyridinium crosslinks may prove to be a good surrogate marker for future clinical trials. In addition, the analysis of pyridinium crosslinks may prove useful to identify individuals with NF1 who are at risk for clinical osseous complications. However, prospective studies measuring urinary pyridinium crosslinks before the development of an osseous complication will be needed to assess their clinical utility in individuals with NF1.

Abbreviations

Dpd

deoxypyridinoline

HPLC

high performance liquid chromatography

NF1

neurofibromatosis type 1

Pyd

pyridinoline