Osteoprotegerin (OPG) is a soluble receptor for RANKL and therefore a competitive inhibitor of osteoclast differentiation and activity. With this key role in the control of resorptive activity, we found that OPG is a candidate gene for genetic control of bone mass. We examined the promoter and the five exons with surrounding intron sequences of the OPG gene for polymorphisms in 50 normal patients and 50 patients with osteoporosis. We found 12 polymorphisms. Two sets of four and five polymorphisms, respectively, were in complete linkage. Subsequently, we examined the effect of the informative polymorphisms A163-G (promoter), T245-G (promoter), T950-C (promoter), G1181-C (exon 1), and A6890-C (intron 4) on the prevalence of osteoporotic fractures, bone mass, and bone turnover in 268 osteoporotic patients and 327 normal controls. In A163-G the variant allele G was more common among fracture patients: 34.0% versus 26.3% in normal controls (p < 0.05) and the odds ratio (OR) for a vertebral fracture, if an individual has the G allele, was 1.44 (1.00–2.08). In T245-G the variant allele G was more common in osteoporotic patients: 12.4% versus 6.5% (p < 0.02) and the OR for vertebral fracture, if an individual has the G-allele, was 2.00 (1.10–3.62). G1181-C is located in the first exon and causes a shift in the third amino acid from lysine to asparagine. The CC genotype was less common among fracture patients: 26.3% versus 36.7% in the normal controls (p < 0.01). T950-C and A6890-C were not distributed differently among patients with osteoporosis and normal controls. None of the polymorphisms affected bone mineral density (BMD) or biochemical markers of bone turnover in the normal controls. In conclusion, we have examined the human OPG gene for polymorphisms and found 12. The rare alleles of the A163-G and T245-G were significantly more common among patients with vertebral fractures.
OSTEOPOROSIS IS characterized by a combination of low bone mass and deteriorated microarchitecture of the bone. Bone mass is determined by interaction of genetic, metabolic, and environmental factors. Genetic factors have been shown to be responsible for 40-75% of the interindividual variation.(1) In an attempt to identify the genes involved in the pathogenesis of low bone mass, and fractures, many candidate genes have been examined, transforming growth factor (TGF) β1,(2) collagen Iα1,(3) vitamin D receptor,(4) interleukin (IL)-6,(5) IL-1 receptor antagonist,(6,7) and calcitonin receptor.(8) However, polymorphisms or sequence variations in these genes have been shown only to have, at best, a modest effect on bone mass and fracture risk.
RANKL is a transmembrane ligand expressed on osteoblasts that binds to RANK, a transmembrane receptor on hemopoietic osteoclast precursor cells as well as mature osteoclast. Interaction between RANK and RANKL induces differentiation, increases activity, and prevents apoptosis of the osteoclasts.(9–18) Osteoprotegerin (OPG) is a decoy receptor for RANKL and a competitive inhibitor of osteoclast recruitment and activity.(16,19–22) Many of the calciotropic hormones (PTH, vitamin D, and glucocorticoids) and cytokines (IL-11, prostaglandin E2 [PGE2], basic FGF [bFGF], IL-1α, IL-1β, and TNFα) appear to stimulate bone resorption through inhibition of OPG production and stimulation of RANKL production.(17,23–36) However, TGF-β and estrogen prevent bone resorption by increasing OPG production.(28,37,38) It has been indicated in mice that administration of OPG exerts a hypocalcemic effect and can prevent stimulation of bone resorption by parathyroid hormone (PTH), vitamin D, IL-1β, and TNF-α.(11,39) The in vitro production of OPG mRNA by the osteoblasts has been shown to decrease with increasing age.(40) Yano et al. showed that OPG in serum increased with increasing age in both men and women. Furthermore, serum levels of OPG were higher in osteoporotic postmenopausal women compared with normal controls.(41) Therefore, OPG has been suggested as a mediator of the increased bone resorption with age or after menopause. However, serum levels of OPG also have been found to be correlated with serum levels of bone-specific alkaline phosphatase and urinary excretion of pyridinoline and deoxypyridinoline. These findings suggest that OPG in serum merely reflects bone turnover and is not a marker of the degree of inhibition of bone resorption.
Transgenic mice without OPG develop severe osteoporosis,(22,42) whereas mice overexpressing OPG develop an osteopetrotic phenotype.(19) Mice without RANK or RANKL lack osteoclasts and have severe osteopetrosis.(43,44) Therefore, it has been suggested that it is the ratio RANKL/OPG that controls osteoclast activity and bone resorption.(45,46) The OPG gene consists of five exons and spans 29 kb.(47)
Because of the potential role of OPG in controlling bone resorption and subsequently bone turnover, we find that the OPG gene is a candidate for mediating genetic influence on bone mass and risk of fractures. We therefore wanted to examine the entire OPG gene for polymorphisms and if polymorphisms were found, investigate the effect of the polymorphisms on the risk of osteoporotic fractures in men and women.
MATERIALS AND METHODS
The study was a case control study. The osteoporotic group consisted of 217 women and 51 men with primary spinal osteoporosis defined by the presence of at least one nontraumatic fracture of the spine, referred to the Department of Endocrinology, Aarhus University Hospital. The diagnosis of primary osteoporosis was made after extensive examination for secondary causes. Spinal fracture was defined as a 20% or more reduction of the anterior, central, or posterior height of a vertebra. The normal control group comprised 255 normal women and 72 normal men without diseases or medications that could influence bone mass and turnover. The normal controls were recruited from the local community by invitations posted at places of work, senior citizens clubs, schools, educational institutions, hospitals, and at general practitioners. Characteristics of the osteoporotic patients and normal controls are presented in Table 1. For identification of polymorphisms in the OPG gene, we randomly selected 50 patients with osteoporosis and 50 normal controls. For comparison of genotype frequencies, we selected an age-matched subgroup of normal women (217 normal women; mean age, 62.6 ± 10.3 years; range, 41-82 years). The study was approved by the local ethical committee and was conducted according to the Helsinki Declaration II.
Table Table 1.. Characteristics of the Patients With Osteoporosis and the Normal Controls
Bone mass measurements
Bone mineral density (BMD) of the lumbar spine and the standard sites at the hip, femoral neck, trochanter, intertrochanteric region, and Wards triangle were assessed using DXA on a Hologic, Inc. 1000 (Hologic Europe, Zaventum, Belgium) or a Norland Medical Systems (Norland Corp., Vaerlose, Denmark) bone densitometer. Results obtained on the Norland Medical Systems densitometer were corrected for the difference between the two densitometers using the correction formulas suggested by Genant et al.(48) All BMD values are corrected for age and gender.
Biochemical markers of bone turnover
Serum samples were collected in the morning after an 8-h fasting period. All samples from patients with osteoporosis were collected before institution of antiosteoporotic treatment.
Serum cross-linked carboxy-terminal telopeptide of type I collagen (s-ICTP) was measured by an equilibrium radioimmunoassay, the intra-assay CV was 5% and the interassay CV was 6%.(49)S-osteocalcin was determined by a radioimmunoassay using rabbit antiserum against bovine bone gla protein.(50) The intra-assay CV was 5% and the interassay CV was 10%.
Sequencing of the OPG gene
DNA was isolated from whole blood leukocytes as described by Kunkel et al.(51) Eleven pairs of upstream and downstream primers were constructed to cover the promoter region (1105 bp) and the five exons of the OPG gene (Table 2). Primers for the exons were located in the intron sequences ∼50 bp apart from the exon. Semiautomated solid-phase sequencing was performed using AmpliTaq FS and four fluorescent dideoxynucleotides on an automated ABI 377 sequencer as previously described.(2) Sequencing was performed in a random sample of 50 patients with osteoporosis and 50 normal controls. The numbers of bases and amino acids are in accordance with the published sequence by Morinaga et al.(47)
Table Table 2.. Primers for Sequencing of the Promoter and Exons of the OPG Gene
Restriction site polymorphisms
Five restriction site fragment length polymorphism assays were constructed for the informative polymorphisms: A163-G, T245-G, T950-C, G1181-C, and A6890-C. In short, polymerase chain reaction (PCR) were performed in a final volume of 25 μl containing 100 ng of genomic DNA, 125 ng of each primer, and AmpliTaqGold DNA polymerase (Perkin Elmer, Allerod, Denmark) using standard conditions on a Perkin Elmer Termocycler 2400 (Table 3). All restriction enzymes were purchased from New England Biolabs (Hellerup, Denmark).
Table Table 3.. Characteristics of the RFLP Assays
Differences in prevalence of the genotypes between patients with osteoporosis and age-matched normal controls were tested using the χ2 test. The effect of genotype and haplotypes (combined genotypes) on BMD and levels of biochemical markers was evaluated by ANOVA and post hoc Student's t-tests for unpaired data. Furthermore, the impact of the polymorphisms on fracture risk was evaluated using logistic regression using age, height, weight, sex, BMD, and the polymorphisms as independent variables. The impact of the polymorphisms on BMD also was investigated using linear regression with the same independent variables.
Linkage disequilibrium (D′) between the different polymorphisms was examined by Fischer's exact test of the distribution of haplotype frequencies using the 2by2 program (http://linkage.rockefeller.edu/software/utilities). The EH program of Terwillinger and Ott(52) was used to estimate haplotype frequencies and to test for associations between haplotypes and the presence of fracture. Several calculations were made, varying the disease allele frequency from 0.2 to 0.5 and the model of inheritance from autosomal dominant to codominant with varying degrees of penetrance 50-100%. The results were very similar and we chose to present the results obtained using a model with disease allele frequency = 0.3 and a codominant model of inheritance with penetrance = 50%. The level of significance was set at 0.05.
We found 12 polymorphisms in the OPG gene (Fig. 1). Five were located in the promoter: T149-C, A163-G, G209-A, T245-G, and T950-C. G1181-C is located in the first exon and causes a change in amino acid from lysine to asparagine. A C to T polymorphism C1217-T is located 15 bp downstream from the first exon. In the second intron, two C to T polymorphisms were found: C445-T, which is positioned 4 bp downstream from exon 2, and C4441-T, which is located 5 bp upstream from exon 3. A TC repeat comprising five repeats is located in the third intron 45 bp downstream from exon 3. A polymorphism comprising a deletion of one CT pair is found in this repeat (4690delCT). In exon 4, which encodes part of the mature protein, a conservative A to G polymorphism was found (A6833-G). Both the wild type and the variant encode a leucine. Finally, 8 bp downstream from exon 4 an A to C polymorphism (A6890-C) was found.
In our study population comprising 595 individuals, T149-C, G209-A, T245-G, C1217-T, and C4441-T are in complete linkage as are C445-T,4690delCT, A6833-G, and A6890-C. Therefore, for further analysis we chose one polymorphism from each of the two groups of polymorphisms in complete linkage T245-G and A6890-C. Five polymorphisms in the OPG gene were considered informative: A163-G, T245-G, T950-C, G1181-C, and A6890-C (Fig. 2). These five polymorphisms were all in Hardy-Weinberg equilibrium (χ2 = 0.00-1.56, NS). The five polymorphisms were in linkage disequilibrium (Table 4).
Table Table 4.. Linkage Disequilibrium Between the Five Informative Polymorphisms in the OPG Gene
In position 875 in the promoter, a nonpolymorphic change from adenosine in the published sequence(47) to cytosine was found in our population
The variant allele G in position 163 in the promoter was more common in patients with osteoporosis (33.9%) than in normal controls (26.3%; χ2 = 3.88; p < 0.05; Table 5). Homozygosity for the G allele was rare but still almost twice as common in patients with osteoporosis (5.2%) as in normal controls (2.8%). This difference in allele and genotype frequency between patients with osteoporosis and normal controls also were found in women and men, although the difference did not reach significance. The odds ratio (OR) for having an osteoporotic fracture if an individual has the G allele was 1.44 (1.00-2.08). The percentage of patients with osteoporosis that can be explained by the A163-G polymorphism was 11.7%. There was no significant effect of this polymorphism on BMD in the patients with osteoporosis or in the normal controls (Table 6). When patients with osteoporosis and normal controls were combined, BMD was lower in individuals with the variant G allele at the lumbar spine (0.852 ± 0.171 g/cm2 vs. 0.884 ± 0.185 g/cm2, p = 0.052), at the femoral neck (0.693 ± 0.123 g/cm2 vs. 0.717 ± 0.119 g/cm2, p < 0.05), and at the total hip (0.836 ± 0.138 g/cm2 vs. 0.845 ± 0.146 g/cm2, NS).
Table Table 5.. Distribution of A163-G, T245-G, T950-C, G1181-C and A6890-C Genotypes in Patients With Osteoporosis and Normal Controls
Table Table 6.. BMD of the Lumbar Spine, Femoral Neck, and the Total Hip in Patients With Osteoporosis and Normal Controls With Different Genotypes
In position 245 in the promoter, the GG genotype was only found in 2 female patients with osteoporosis (Table 5). The genotype distribution was significantly different between patients with osteoporosis and normal controls (χ2 = 6.71, p < 0.05). Furthermore, the G allele was significantly more frequent among patients with osteoporosis (12.3%) than in normal individuals (6.5%; χ2 = 5.60, p < 0.05). This difference in genotype distribution also was found among women (χ2 = 6.24, p < 0.05). The difference was less pronounced among men (NS). Because only 2 individuals with osteoporosis had the GG genotype, GG and TG genotypes were combined for further analyses. The OR for having an osteoporotic fracture if an individual has the G allele was 2.00 (1.10-3.62). The percentage of patients with osteoporosis that can be explained by the T245-G polymorphism is 8.6%. No significant differences could be shown in BMD between individuals with the TT and TG/GG genotypes at any site (Table 6).
The T950-C polymorphism in the promoter was not distributed differently among patients with osteoporosis and normal controls (χ2 = 2.60; NS; Table 5). Furthermore, the genotype distribution also was similar in the female and male subgroups. At the lumbar spine, BMD in individuals with the CC genotype was 0.903 ± 0.172 g/cm2; BMD was 0.870 ± 0.180 g/cm2 in individuals with the TC genotype; and BMD was 0.856 ± 0.193 g/cm2 in individuals with the TT genotype (p = 0.07; ANOVA; Table 6). BMD was significantly lower in individuals with the TT genotype compared with individuals with the CC genotype (p < 0.05). No significant differences were found at the femoral neck or the total hip. These differences were also found when patients with osteoporosis were examined separately.
The polymorphism in the first exon G1181-C was distributed highly significantly different among patients with osteoporosis and normal controls (χ2 = 10.45; p < 0.01; Table 5). This difference, which was found also in the female and male subgroup, was based mainly on a higher frequency of CC homozygotes (36.4%) among normal individuals compared with patients with osteoporosis (26.1%). Surprisingly, the GG genotype was not more common among patients with osteoporosis than in normal controls; however, heterozygosity was more frequent in the patients with osteoporosis (56.0%) than in normal individuals (42.6%). The G allele was significantly more common in patients with osteoporosis (73.9%) than in normal controls (63.6%; χ2 = 6.87; p < 0.01). In the patients with osteoporosis and in the normal controls, no significant effect of G1181-C on BMD could be shown (Table 6). However, in the whole study population, BMD at the lumbar spine was different between genotypes: 0.876 ± 0.188 g/cm2 in individuals with the GG genotype, 0.858 ± 0.180 g/cm2 in individuals with the GC genotype, and 0.901 ± 0.180 g/cm2 in individuals with the CC genotype (p < 0.05; ANOVA). The difference between BMD in individuals with GC or CC genotypes was significant (p < 0.05). Furthermore, individuals with the G allele (GG and GC genotypes) had lower BMD at the lumbar spine: 0.863 ± 0.182 g/cm2 versus 0.901 ± 0.180 g/cm2 (p < 0.05). No effect of this polymorphism on BMD was shown at the femoral neck or at the total hip.
In the fourth exon, the A6890-C polymorphism was distributed similarly among patients with osteoporosis and normal controls (χ2 = 4.10, NS; Table 5). Furthermore, no differences in BMD were found between the genotypes (Table 6).
Logistic regression revealed that the risk of having an osteoporotic fracture was determined by sex, height, and age-corrected BMD of the lumbar spine; none of the polymorphisms influenced the risk of fractures significantly. However, if the polymorphisms were included in the analysis separately, A163-G revealed a trend for predicting osteoporotic fractures (p = 0.06).
Linear regression analyses revealed that BMD of the lumbar spine was determined by sex, age, and weight. If all polymorphisms were included in the analysis, none significantly affected BMD; however, if they were included one by one, A163-G significantly influenced BMD (p < 0.05).
We used the EH program to analyze haplotype frequencies of the A163-G and T245-G polymorphisms in patients and controls separately. Comparing the haplotype frequencies revealed that the haplotypes were distributed significantly differently in patients and controls (χ2 = 6.88, p < 0.05), with overrepresentation of the GG haplotype among fracture patients (Table 7). Grouping the haplotypes by alleles at the individual marker showed significant associations with the A163-G and T245-G alleles and fracture (χ2 = 4.30, p < 0.05, and χ2 = 5.88, p < 0.02, respectively).
Table Table 7.. Haplotype Analysis of A163-G and T245-G Polymorphisms in Patients With Osteoporotic Fractures and Normal Controls (Upper Panel) and Distribution of the A163-G and T245-G Combined Genotypes in Patients With Osteoporosis and Normal Individuals
Because only the A163-G and T245-G polymorphisms were found to be predictors of fractures and bone mass, we also examined the effect of the combined genotypes on bone mass and fracture risk. Of the nine possible combined genotypes only six were found in our population (Table 7). The combinations, comprising individuals with two or more of the alleles that we found were associated with osteoporotic fracture, were rare and therefore combined. The combined genotypes were distributed differently among patients with osteoporosis and normal controls (χ2 = 6.07, p < 0.05). BMD of all measured sites were higher in individuals with the AATT haplotype, although the differences failed to reach significance (Fig. 3).
None of the examined polymorphisms significantly affected the biochemical markers of bone resorption and formation (s-ICTP and s-osteocalcin [BGP], respectively; data not shown).
In this study, we found 12 polymorphisms in the OPG gene. Two have been described previously.(47,53) Because two groups of five and four polymorphisms were in complete linkage, the remaining three and one from each group of linked polymorphisms were considered informative. Of these, three polymorphisms (A163-G, T245-G, and C1181-G) were found to affect fracture risk.
A163-G is located in the promoter and does not affect any recognition sites for known transcription factors or other gene-regulating sites. T245-G is in complete linkage with T149-C, G209-A, C1217-T, and C4441-T, and the effect seen of T245-G on fracture risk therefore could be caused by any of these five polymorphisms. The three first polymorphisms are located in the promoter region. They are not located in binding sites for known transcription factors. C1217-T and C4441-T are intronic polymorphisms located 15 bp downstream of exon 1 and 5 bp upstream from exon 3, respectively. From a theoretical point of view, C4441-T could affect splicing.(54) G1181-C is located in the first exon and causes a change in the 3rd of 10 amino acids in the signal peptide from lysine to asparagine. Both are charged polar amino acids carrying either a NH3+ or a NH2+ group, respectively. Therefore, from a theoretical point of view, it is not likely that the change would have a physiologically important effect on the characteristics of the peptide. The representative of the other group of completely linked polymorphisms A6890-C is an anonymous polymorphism in exon 4 that does not have any effect on fracture risk or bone mass.
The CC genotype of the G1181-C polymorphism was less frequent in patients with osteoporosis and was associated with changes in bone mass at the lumbar spine. However, the effect on BMD was only found when patients with osteoporosis and normal controls were grouped (not in either of the groups separately) and there was no clear gene-dose effect on bone mass, individuals with the GG genotype have higher BMD at all sites than individuals with the GC genotype, although the difference did not reach significance. Furthermore, evaluated by linear and logistic regression analyses, G1181-C was not a significant predictor of either BMD at the lumbar spine or hip or of osteoporotic fractures. Taken together, these results did not provide strong evidence for a role of the G1181-C polymorphism in the pathogenesis of osteoporosis.
The A163-G and T245-G polymorphisms were in strong linkage disequilibrium. Both were distributed differently among patients with osteoporosis and normal controls, the less common alleles being significantly more prevalent in patients with osteoporosis than in normal controls. Regression analyses revealed that the A163-G polymorphism predicts bone mass and tends to predict osteoporotic fractures independently of BMD. Both the combined genotypes and the haplotypes formed by these two polymorphisms were associated with increased risk of vertebral fractures. Based on these results, it is not possible to conclude if the associations between the A163-G and T245-G polymorphisms and osteoporotic fractures are caused by one of them and merely reflected by the other one because of linkage disequilibrium or by a true effect of both polymorphisms. However, the A163-G polymorphism displayed more significant results and is responsible for a higher percentage of osteoporotic fractures than the T245-G polymorphism; this may suggest that the A163-G is the most important of the polymorphisms with respect to causing osteoporosis. Furthermore, combining the two polymorphisms into haplotypes did not reinforce the differences.
Previously, Brandstrom et al. have reported that the T950-C polymorphism was associated with bone mass in Swedish men.(53) The T950-C polymorphism is located in the promoter, 129 bp upstream from the TATA box. Brandstrom et al. reported that the rare genotype CC found in 7% of the men was associated with increased bone mass. The same authors were not able to find any association between this polymorphism and bone mass in a cohort of 1044 elderly Swedish women.(55) In this study, we found that the rare genotype CC was more frequent in the Danish population (25.6%) and was associated with increased bone mass at the lumbar spine in patients with osteoporosis. However, no association of the T950-C polymorphism with fracture risk was established.
Recently, Wuyts et al. have shown that the C445-T polymorphism is associated with Paget's disease of bone.(56) The more common allele C was found on 89% of the chromosomes in patients with Paget's disease but only on 75% of the chromosomes in normal controls. Paget's disease does not affect the entire skeleton but often only a single bone. The changes in bone architecture are initiated by an increased resorption mediated by an increased number of abnormal osteoclasts. This could, in theory, be caused by an imbalance between OPG and RANK or RANKL. If the association indicated is true and mediated by changes in OPG, it should be suspected that the C allele causes reduced production of or defect in OPG. The polymorphisms are located 4 bp downstream from exon 2 and could in theory affect splicing. C445-T is in complete linkage with the4690delCT, A6833-G, and A6890-C polymorphisms. From that group, we examined the A6890-C polymorphism and did not find any association with either prevalence of osteoporotic fractures or BMD. Our results do not support that the C445-T polymorphism or any other polymorphism in this group of completely linked polymorphisms cause a general increase in bone resorption.
In conclusion, we have examined the entire OPG gene for polymorphisms and found 12. Two of these (A163-G and T245-G) were associated with increased fracture risk. How these polymorphisms affect fracture risk is not known and further studies on the effect of these polymorphisms on OPG mRNA production and stability are needed to clarify the mechanisms and pathophysiology underlying the associations with fracture risk and bone mass found in this study.
The authors thank The Danish Research Council, The Novo Nordisk Fonden, The Institute of Experimental Clinical Research at University of Aarhus, and The Fonden til Lægevidenskabens Fremme for financial support.