Eur J Endocrinol
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

DOI: 10.1530/EJE-08-0021
European Journal of Endocrinology, Vol 159, Issue 2, 187-196
Copyright © 2008 by European Society of Endocrinology
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
EJE-08-0021v1
159/2/187    most recent
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lazáry, A.
Right arrow Articles by Lakatos, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Lazáry, A.
Right arrow Articles by Lakatos, P.
CLINICAL STUDIES

Single nucleotide polymorphisms in new candidate genes are associated with bone mineral density and fracture risk

Áron Lazáry1,2,*, János P Kósa1,*, Bálint Tóbiás1, Judit Lazáry3, Bernadett Balla1, Krisztián Bácsi1, István Takács1, Zsolt Nagy1, Tibor Mezo1, Gábor Speer1 and Péter Lakatos1

1 1st Department of Internal Medicine, Semmelweis University, Koranyi S. u. 2/a., Budapest H-1083, Hungary2 National Center for Spinal Disorders, Buda Health Center, Budapest H-1126, Hungary3 Department of Pharmacology and Pharmacotherapy, Semmelweis University, Budapest H-1089, Hungary

(Correspondence should be addressed to Á Lazáry; Email: lazaron{at}bel1.sote.hu)


   Abstract
Top
 Abstract
Introduction
 Materials and methods
 Results
 Discussion
 Footnotes
 References
 
Objective: Osteoporosis (OP) is a multifactorial disease with high heritability but its exact genetic background is still poorly understood. We examined the effect of 24 single nucleotide polymorphisms (SNPs) located in five genes – alkaline phosphatase, matrix metalloproteinase-2, tissue inhibitor of metalloproteases-2 (TIMP2), fibroblast growth factor receptor-1 (FGFR1), and fatty acid-binding protein-3 (FABP3) – previously not associated with OP.

Design: We performed a genotype–phenotype association study at a university hospital.

Methods: A total of 360 Hungarian postmenopausal women were involved in the study. Bone mineral density (BMD) was determined at spine, hip, and distal radius. Genomic DNA was extracted from venous blood samples and a high-throughput genotyping method based on single-based primer extension was applied for allelic discrimination. Robust statistical tools were utilized for multiplex data analysis.

Results: SNP rs6996321 in FGFR1 was significantly related to spine BMD (P=0.002) and rs10914367 in FABP3 was associated with hip BMD (P=0.028). Non-vertebral fracture risk was significantly increased in carriers of ‘A’ allele of rs9900972 in TIMP2 (odds ratio=2.06, P=0.0187). We could also identify validated gene–gene interactions significantly affecting BMD and fracture risk.

Conclusions: We identified two previously unreported SNPs in FGFR1 and FABP3 associated with BMD and a third SNP in TIMP2 related to risk for non-vertebral osteoporotic fractures. This is the first report about the association between these allelic variants and the phenotypes of postmenopausal OP. Further studies need to clarify the role of these genes and their polymorphisms in the process of bone loss.


   Introduction
Top
 Abstract
 Introduction
Materials and methods
 Results
 Discussion
 Footnotes
 References
 
Osteoporosis (OP) is a common, multifactorial disease with high heritability (1). Linkage studies have identified several loci associated with OP across every human chromosome. More than 200 candidate genes influencing the disease have already been reported (2, 3), but the exact genetic background and the interactions among suspected genes as well as genes and environmental or lifestyle factors (4) are still poorly understood (5). Previous studies have also shown that genetic influences can differ among distinct populations and between genders (6, 7, 8). Bone mineral density (BMD) and risk of osteoporotic fracture, predominant study phenotypes in genetic studies, are affected by several genes and their polymorphisms. Previously reported single nucleotide polymorphisms (SNPs) have small individual effects and the genetic correlation with the phenotypes is low (9). Therefore, more haplotype and gene–gene interaction cohort studies should be performed for the investigation of the genetic background of OP (5, 10, 11).

Previously, less investigated genes came to our attention in our previous studies regarding the relationship between deer antler development and human OP (12). Based on literature data and on our pilot studies, we selected five genes with potential roles in human OP: alkaline phosphatase (ALPL), matrix metalloproteinase-2 (MMP2), tissue inhibitor of metalloproteases-2 (TIMP2), fibroblast growth factor receptor-1 (FGFR1), and fatty acid-binding protein-3 (FABP3). In a previous human gene expression study (13), significant differences in gene expression of ALPL and MMP2 were determined between the healthy and the porotic human bone tissue. The change was in trend in the case of FGFR1 gene in this study. We supposed that different gene expression patterns determined in healthy and porotic bone tissue could be related to the allelic variants of these genes. In this work, we aimed to investigate the effect of multiple SNPs in ALPL, FABP3, FGFR1, MMP2, and TIMP2 on BMD and osteoporotic fracture risk in Hungarian postmenopausal women using a high-throughput genotyping method.


   Materials and methods
Top
 Abstract
 Introduction
 Materials and methods
Results
 Discussion
 Footnotes
 References
 
Study population

Three hundred and sixty Hungarian unrelated postmenopausal women were recruited to the study through our clinic. Subjects visited the clinic for a standard dual energy X-ray absorptiometry (DXA) screening. All the subjects underwent physical examination and completed a detailed questionnaire on family and medical histories and lifestyle habits. Exclusion criteria were history of bone, metabolic, or endocrine disease; any chronic illness; hormone replacement, steroid therapy, or any medication known to influence bone metabolism; premature menopause (before 40 years of age); and alcohol consumption greater than two units per day. Subjects with biochemical abnormalities such as increased levels of serum ALPL, thyrotrophin, parathyroid hormone, or reduced level of 25-OH vitamin D (<30 ng/ml) were not included in the study. BMD values at the lumbar spine (L2–L4) and total hip were measured using a Lunar Prodigy DXA (GE Medical Systems, Diegem, Belgium). BMD at the distal radius was determined by a Norland pDEXA (CooperSurgical Inc., Trumbull, CT, USA) densitometer. Coefficient of variation was below 1% at both sites. OP was defined according to the WHO criteria, i.e., T-score less than –2.5 S.D. at any measured site. Detailed fracture history was obtained from each subject. Non-vertebral osteoporotic fracture was defined as low-trauma fractures after the age of 45 years excluding the fractures of the face, skull, fingers, toes, and spine. Vertebral compression fractures were not investigated in this study. The study was approved by the local ethical committee, and written informed consent was obtained from all participants. Characteristics of the study population are shown in Table 1.


View this table:
[in this window]
[in a new window]

  		Table 1 Characteristics of study population (n=353a).

 
Gene and SNP selection

Based on our previous study (13) and on online data mining (http://www.ncbi.nlm.nih.gov/OMIM/, http://www.hapmap.org), we chose five genes expressed in human bones that are likely to play an essential role in bone metabolism; ALPL, MMP2, TIMP2, FGFR1, and FABP3. We used the NCBI dbSNP database (http://www.ncbi.nlm.nih.gov/SNP/) and data from HapMap project (http://www.hapmap.org) to extract the available information on the SNPs of these genes. We chose 24 SNPs from the sites of the genes with optimal (40–65%) GC content using the following criteria: a) polymorph alleles are C/T or A/G (these two types of SNPs can be genotyped on the same plate at the same time by the high-throughput method used in our study); b) validation status in Caucasians is available at public databases; c) reported minimal allele frequency (MAF) is more than 5%; and d) SNP site (promoter, intron, exon, and 3'-UTR) and function (non-synonymous coding, synonymous coding, and splice site) according to general rules of SNP selection for association studies (14). The most of the genotyped SNPs are either haplotype-tagging SNPs identified by HapMap or strong linkage disequilibrium (LD) with tag SNPs. The full coverage could be realizable with genotyping at least 30, 5, 11, 9, and 20 tag SNPs across the ALPL, FABP3, FGFR1, MMP2, and TIMP2. The 24 genotyped SNPs in this study tagged with an r2>0.8 the 29, 80, 64, 41, and 55% of the all allelic variants located in the intragenic as well as upstream and downstream regions of ALPL, FABP3, FGFR1, MMP2, and TIMP2 respectively. Selected SNPs in candidate genes are shown in Fig. 1 and Table 2.


Figure 1
View larger version (33K):
[in this window]
[in a new window]

  		Figure 1 Position of genotyped SNPs and the pattern of linkage disequilibrium in selected genes. Pairwise linkage disequilibrium (LD) statistics in examined genes were calculated with Haploview. Squares are colored darker if the D' value is high (i.e., LD is strong), lighter shades indicate less LD. Empty dark squares mean D'=1 (i.e., complete LD between two SNPs).

 

View this table:
[in this window]
[in a new window]

  		Table 2 Descriptive statistics of genotyped single nucleotide polymorphisms (SNPs).

 
Genotyping

Blood samples were collected from each subjects and genomic DNA was extracted using High Pure PCR Template Purification kit (Roche Diagnostics, GmbH, Mannheim, Germany). DNA quality and quantity was determined with NanoDrop B-100 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA), and all the samples were diluted to a DNA concentration of 10 ng/µl. Large-scale genotyping process was performed at the SNP Core Facility of Semmelweis University using a GenomeLab SNPstream Genotyping System (Beckman Coulter, Fullerton, CA, USA). This automated ultra-high-throughput genotyping system utilizes multiplexed PCR in conjunction with tagged-array, multiplexed, single-base primer extension technology (15). Briefly, the workflow steps were as follows: 1) A multiplex PCR was applied to amplify the selected genes. 2) After a clean-up step (enzymatic degradation of unconsumed primers and dNTPs present in PCR products with a specialized formulation of two hydrolytic enzymes, exonuclease I and shrimp ALPL), a single-base primer extension was performed. 3) Due to the SNP primers with unique 5'-tagged sequence, extended products were hybridized to the glass-bottom SNPware Tag Array Plate. 4) Genotypes were detected with direct allele-specific, two-color fluorescent detection method using the SNPstream Imager. 5) Data analysis and genotype determination were performed with SNPstream software. All PCR primers and tagged SNP primers for primer extension were designed by the web-based primer design tool of Beckman Coulter (Autoprimer.com, http://www.autoprimer.com). Each step of genotyping was performed on 384-well plates and the pipetting procedures were automated using Biomek FX Dual Arm system (Beckman Coulter). Ten random samples were genotyped in three parallel runs for testing technical errors.

Data pre-processing and statistical analysis

Samples with more than 15% missing genotypes (less than 20 successfully genotyped polymorphic sites) were removed from the database to minimize the false results due to genotyping error. A 10-nearest neighbor method was used to assign the missing genotypes in the final database (16). We used Haploview 3.0 software (Broad Institute of MIT and Harvard, Cambridge, MA, USA) (17) for descriptive analyses (Table 2) and its ‘Tagger’ option for tagging process. SNPs with MAFs less than 5% and SNPs in Hardy–Weinberg disequilibrium (P value of {chi}2 test <0.05) were excluded from further analyses. Pairwise LD between genotyped SNPs was determined with Haploview.

Kolmogorov–Smirnov test was applied to analyze the distribution of the study phenotypes. Stepwise regression analyses were applied to spine, total hip, radius BMD (linear regression), and non-vertebral osteoporotic fracture (logistic regression), as phenotypes; and age, menopausal age, body mass index (BMI), smoking status, and alcohol consumption, as covariates that were used to identify significant covariates with phenotypes. Spine, total hip, and radius BMD values were also entered into logistic regression model to determine their influence on fracture risk. Computed regression residuals, representing covariate-adjusted study phenotypes, were used in subsequent analyses (8). Association analyses were performed using the program PedGenie (18), statistical software developed for analyses of pedigrees and independent individuals, computing a statistic equivalent to ANOVA under three genetic models (additive, dominant, and recessive). We performed the ‘quantitative’ and the ‘odds ratio (OR)’ statistics built into the software in this study. To assess the reliability of the results, permutation procedures were performed to generate empirical P values. P values reported in individual SNP analyses are empirical P values after 10 000 Monte Carlo permutations. Permuted global P<0.05 was considered significant. The statistical power of the study was calculated with the Quanto 1.1 software (University of Southern California, Los Angeles, CA, USA) (19).

We conducted haplotype analyses using the ‘haplo.stats’ software in the statistical environment ‘R’ (http://cran.r-project.org/). Sliding windows of 2, 3, and 4 loci were used to construct haplotypes in each gene and score statistics were applied to test the association between haplotypes and study phenotypes (20). This method computes permuted P values of the associations. Effect of haplotypes on covariate-adjusted study phenotypes was calculated in a regression model using the ‘haplo.glm’ function built in the software.

Gene–gene interactions (epistasis) were tested using ‘SNPassoc’ software published by Gonzalez et al. (21). This ‘R’ package, developed for genetic studies, determines epistasis effects performing log-likelihood ratio tests (LRTs). The graphical output of this command visualizes the P values of the interactions between each pair of SNPs and the examined trait. To avoid false-positive results, only highly suggestive interactions characterized by a P value less than 0.001 were selected and gene–gene interactions were validated under conditional regression models. Interactions with a P<0.01 in the regression model were considered significant. The regression analyses were performed using SPSS 15.0 for Windows (SPSS Inc., Chicago, IL, USA).

Sequence analysis

Genomic sequences of the regions next to the SNPs were obtained from the NCBI dbSNP database (http://www.ncbi.nlm.nih.gov/SNP/). Two sequences containing the two allelic variants of each SNP were used to search for putative transcription factor binding sites in the TRANSFAC database, using the PROMO web-based software (http://alggen.lsi.upc.es/) (22).


   Results
Top
 Abstract
 Introduction
 Materials and methods
 Results
Discussion
 Footnotes
 References
 
Descriptive analyses

The reproducibility of the high-throughput genotyping method called SNPstream was 100% in our study (genotypes were the same in each parallel run). We excluded seven subjects from the statistical analyses, because their genotyping success rate was lower than 85%. MAF was lower than 5% in the case of rs11542001 in MMP2 and two SNPs were not in Hardy–Weinberg equilibrium (rs11578034 in FABP3 and rs243843 in MMP2; Table 2). These markers were excluded from further evaluations. The final statistical study cohort consisted of 21 SNPs and 353 subjects. In this dataset, 54 genotypes were imputed. Age, menopausal age, and BMI proved to be significant covariates in the regression model fitting of BMDs. Fracture risk was significantly influenced by age, BMI, and BMDs. Regression residuals representing covariate-adjusted study phenotypes were used in all genetic analyses.

Association analyses

The power of each significant association was also computed and only statistically significant associations with a power of study higher than 80% (0.80) are discussed in this paper. For spine BMD (mean±S.D.=0.887±0.176) at a significance level of 0.05, MAF of 0.322, and 353 individuals, there would be a power of 83% to detect differences in BMD of 0.055 g/cm2 under a dominant model. For hip BMD (mean±S.D.=0.785±0.168), there would be a power of 82% to detect difference of 0.157 g/cm2 under the recessive model (MAF=0.166). For the fracture risk, the sample size of this study (76 cases and 277 controls) with a MAF of 0.166 and an OR of 2.06 would give a power of 86%.

The rs6996321 in FGFR1 was associated with BMD at lumbar spine (unadjusted mean BMD±S.E.M. was 0.858±0.013, 0.912±0.015, and 0.916±0.029 in subjects with G/G, A/G, and A/A genotypes respectively). Adjusted BMD significantly differed between genotypes in the dominant genetic model (mean of adjusted spine BMD±S.E.M. was –0.031±0.011 and 0.031±0.011 in groups of G/G and A/G+A/A respectively, P=0.002; Fig. 2A). The homozygous recessive genotype of rs10914367 in FABP3 was related to higher BMD at the total hip (mean BMD±S.E.M. was 0.783±0.009, 0.776±0.028, and 0.945±0.029 in subjects with G/G, A/G, and A/A genotypes respectively; mean of covariates adjusted BMD±S.E.M. in the recessive model was –0.004±0.008 and 0.143±0.037 in G/G+A/G and A/A groups respectively, P=0.028; Fig. 2B). Table 3 shows the OR of the non-vertebral fracture risk in relation to rs9900912 (TIMP2) genotypes. The risk for fractures was twofold higher in carriers of the recessive ‘A’ allele of this SNP (P=0.018). None of the SNPs significant for OP phenotypes were associated with body weight or BMI.


Figure 2
View larger version (10K):
[in this window]
[in a new window]

  		Figure 2 (A) Association of rs6996321 in FGFR1 with spine BMD, (B) Association of rs10914367 in FABP3 with hip BMD. Genotype-specific BMD means with standard errors of means (mean±S.E.M.) are represented. The data were adjusted for significant covariates.

 

View this table:
[in this window]
[in a new window]

  		Table 3 The association of tissue inhibitor of metalloproteases-2 (TIMP2) rs9900972 with risk for non-vertebral osteoporotic fractures in the recessive model.

 
Thirty four haplotypes across the genes were constructed and tested in ‘haplo.stats’. We identified a four loci haplotype in FGFR1 constructed by rs13317, rs3925, rs2280846, and rs6996321 that was significantly related to spine BMD in score tests (permuted P=0.007 of global association). The common haplotype (TCGG) was associated with significantly decreased spine BMD while a less common haplotype (CTGA) was related to higher spine BMD (Table 4). rs6996321 was also significantly associated with spine BMD in individual analyses. Therefore, we also performed LRTs to analyze whether the rs6996321 is associated with the spine BMD in the different haplotype structures. The effect of this SNP proved to be constant across the different haplotypic backgrounds in which it is carried. The effect of rs6996321 on the TCG background (Ht1-Ht2) did not significantly differ from its effect on the CTG background (Ht4-Ht3) ({chi}2=2.38, df=2, P>0.25). We also compared the effects of the two similar haplotypes with the two different rs6996321 alleles (underlined). The effect of Ht1 (TCGG) was not different from the effect of Ht2 (TCGA) ({chi}2=1.48, df=2, P>0.25). Similar results were observed comparing the effect of Ht4 (CTGG) with the effect of Ht3 (CTGA) ({chi}2=5.6, df=2, P>0.05). These tests showed that the significant association between the spine BMD and the FGFR1 haplotypes was not driven by the effect of the rs6996321.


View this table:
[in this window]
[in a new window]

  		Table 4 Haplotypes built by rs13317, rs3925, rs2280846, and rs6996321 in fibroblast growth factor receptor-1 (FGFR1) and their association to spine bone mineral density (BMD).

 
Analyses of gene–gene interactions

We found that some gene–gene interactions associated with BMD or fracture risk where the P value of LRT analyses was lower than 0.001 (P1). These interactions were indicated with black symbols on the graphical output of SNPassoc (Fig. 3). We validated all of these strongly significant gene–gene interactions using conditional regression analyses, and we confirmed interactions only in cases where regression analysis was also strongly significant (P2<0.01). Table 5 shows highly suggestive gene–gene interactions and the mean BMDs and OR for non-vertebral fracture risk of each genotype combinations. We identified a gene–gene interaction between FABP3 and MMP2 (rs10914367 and rs1030868) highly suggestive for spine BMD, and an interaction between ALPL and TIMP2 (rs3738099 and rs9894295) significantly associated with hip BMD. Interaction of ALPL and TIMP2 (rs871132 and rs9894295) was also related to radius BMD. Interaction between MMP2 and TIMP2 (rs243847 and rs931227) was highly suggestive for the risk of non-vertebral osteoporotic fracture.


Figure 3
View larger version (42K):
[in this window]
[in a new window]

  		Figure 3 Gene–gene interaction plot generated by SNPassoc. This representative plot shows the significance levels of gene–gene interactions and spine BMD in the codominant model. Each plot contains the P values obtained from different likelihood ratio tests. Different colors indicate different statistical significance levels. The diagonal line contains the P values from likelihood ratio test for the crude effect of each SNP. The upper triangle in the matrix contains the P values for the interaction (epistasis) log-likelihood ratio test. Finally, the lower triangle contains the P values from likelihood ratio test comparing the two-SNP additive likelihood to the best of the single-SNP models.

 

View this table:
[in this window]
[in a new window]

  		Table 5 Validated gene–gene interactions associated with (A) bone mineral density, (B) non-vertebral fracture risk.

 

   Discussion
Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
Footnotes
 References
 
In this report, we investigated the association of postmenopausal OP and multiple allelic variants of five potential candidate genes. Polymorphisms of four of the selected genes were not previously reported in this field. Three individual SNPs associated with osteoporotic phenotypes have been identified in this study. FGFR1 (member of receptor tyrosine kinases) is expressed on both osteoclasts and osteoblasts, and influences osteogenic differentiation of stem cells (23, 24, 25). Its mutations and haplotypes are associated with skeletal deformities seen in Pfeiffer and Kallmann syndromes, human tooth agenesis (26), and normal variations in craniofacial shape (27). In a recent study, decreased expression of FGFR1 in osteoporotic bone tissue was demonstrated (13). rs6996321 is a tag SNP located in the first intron of FGFR1 tagging the whole promoter region, the first exon and the first intron with an r2>0.8, suggesting that this polymorphism may be an important marker rather than a direct contributor to the genetic functions. It is also in strong LD with the rs2568231, which is located in a splice site of the gene, 7 bp from the first exon. In our study, this marker was associated with lumbar BMD at a high level of significance. Haplotype of FGFR1 constructed by four loci across the gene was also related to the variance in BMD at the lumbar spine suggesting a major role for this gene in bone metabolism.

Homozygous recessive genotype of rs10914367 in the promoter region of FABP3 was significantly correlated with increased hip BMD. This polymorphism tags the promoter region, the first exon and the first intron of the gene. In sequence analyses, we could identify putative transcription binding sites containing this locus. In the case of the ‘A’ allele, four transcription factors can bind to this sequence with a similarity of more than 90%. These transcription factors with their binding sites (rs10914367 is underlined) are as follows: forkhead box protein P3, (GCCAAC); C/EBP-β, CCAAT–enhancer-binding protein-β, (CCAA); and PR-A and PR-B, progesterone receptor isoforms A and B, (AACACCA). Interestingly, these physical transcription factor binding sites seem to disappear in the case of the ‘G’ allele, suggesting a transcription-regulating role of this polymorphism. FABP3 participates in fatty acid metabolism, particularly in the regulation of fatty acid uptake and intracellular transport (28) and it is also expressed by bone-derived mesenchymal stem cells (29). FABP3 has been previously reported to have an impact on the development of diabetes mellitus and obesity (28, 30). Expression of this gene is also modulated by the peroxisome proliferator-activated receptor (PPAR)-{gamma}2 transcription factor that has been shown to interfere with age-related bone loss via the activation of adipogenic and suppression of osteogenic programs of mesenchymal stem cells (31). Our results strengthen the putative connection between bone and lipid metabolism (32, 33, 34); however, further studies are needed to clarify the role of FABP3 and its genetic variations in the pathomechanism of bone loss.

We have identified a SNP in the TIMP2 gene (rs9900972) that has shown a significant correlation with non-vertebral osteoporotic fracture risk. Subjects carrying the ‘A’ allele of this polymorphism have twofold risk for fragility fracture compared with non-carriers. This intronic polymorphism is in high LD with the middle region of the gene, suggesting a marker role for fracture risk. TIMP2 is a specific inhibitor of MMPs, matrix degradation enzymes liable for normal (35) and pathological metabolism (36, 37) of the organic compounds of bone extracellular matrix (ECM). ECM produced by osteoblasts has a crucial role in the determination of quality and fracture susceptibility of the bone tissue. Allelic variants in members of the MMP-cascade system were previously related to bone metabolism. SNPs in MMP1 (38) and MMP9 (39) have been suggested to be associated with BMD. Polymorphisms in MMP2 have been described as predictors for multicentric osteolysis and arthritis syndrome (36). This enzyme may have an essential function in malignant diseases (40) as well as an impact on bone fracture healing (41). In our study, MMP2 alone was not linked with BMD or fracture risk in individual SNP analyses; however, it appeared to contribute to gene–gene interactions significantly (two highly suggestive interactions out of four). A highly suggestive interaction between MMP2 and its inhibitor, TIMP2, was found to influence fracture risk in our study population. The allelic variants of TIMP2 might alter the inhibition process of MMPs. The decrease in the inhibition of MMPs could result in accelerated degradation of the ECM and consequently, increased fragility of the bone.

TIMP2 also appeared to have a gene–gene interaction with ALPL affecting BMD at the hip and radius. ALPL is essential for the mineralization of the osteoid matrix, and its serum level is related to the speed of bone turnover. ALPL expression is decreased in the subjects characterized by low bone mass in mice and human studies (42, 13). Allelic variants of this gene have been associated with various forms of hypophosphatasia (43) and a previous study reported two SNPs of these genes related to BMD in the Japanese population (44). Our results may suggest a putative role of ALPL allelic variants in the pathogenesis of postmenopausal OP, but further studies need to clarify it.

Sample size might be a limitation in genetic studies. Our work is a medium size multiple association study. Considering the false-positive results due to sample size and multiple tests, we conducted robust statistical analyses with rigorous criteria. In the case of individual SNP analyses, where 21 markers were analyzed with four phenotypes, a permutation procedure was performed to calculate empirical P values. Significant results were discussed only if the power of study was more than 80%. In haplotype analyses, empirical P value was also obtained in score tests and a validation procedure was performed using generalized linear method. Gene–gene interactions were accepted as ‘highly suggestive’ if log-likelihood analyses showed high level of significance and the P value of the validation procedure using regression models was also less than 0.01. We were able to collect a vast array of clinical data on our subjects, thus exclusions and statistical adjustments were considerably effective. Another limitation of our study is that we could not provide full coverage of the selected genes with these 24 genotyped SNPs. Based on the results of this first report, further studies are required to clarify the relationship between these genes and the pathological processes leading to osteoporosis.

Two phenotypes of OP (low BMD and non-vertebral fractures) have been investigated in our study. Low BMD is an important risk factor for fragility fracture; however, the recent studies suggest that genetic background of the two phenotypes can be diverse (9). We did not find any individual SNP or gene–gene interaction affecting both BMD and fracture risk. Nevertheless, some of the investigated genes (TIMP2 and MMP2) participated in highly significant gene–gene interactions influencing BMD and fracture risk, as well. Our results suggest that phenotypes of OP can be affected by a number of different genomic patterns. These patterns (gene–gene interactions, SNP combinations) may result in similar phenotypes. Genetic background might be different in distinct populations and similar among people of the same sex and geographical region. From this point of view, age-related OP – despite its similar appearance – might be a genetically diverse disease. Large-scale studies with multiplex genomic analyses should be performed to draw the interaction map of the different genomic patterns related to bone loss. Furthermore, the role of protective SNPs or gene–gene interactions should be emphasized. The balance or imbalance between protecting and harmful genomic effects could be essential in the development of postmenopausal OP, and the knowledge of these mechanisms might be useful in the diagnosis and treatment of this disease.


   Acknowledgements
 
This work was supported by grants NKFP-1A/007/2004, NKFP-1A/002/2004 from National Research and Technological Office (NKTH) of Hungary, as well as by research grant ETT 022/2006 from the Ministry of Health, Hungary. The authors would like to acknowledge Prof. Andras Falus and Ildiko Ungvari for their invaluable help in high-throughput genotyping.


   Footnotes
Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Footnotes
References
 
* Á Lazáry and J P Kósa contributed equally to this work Back


   References
Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Footnotes
 References
 

    1. Deng HW, Mahaney MC, Williams JT, Li J, Conway T, Davies KM, Li JL, Deng H & Recker RR. Relevance of the genes for bone mass variation to susceptibility to osteoporotic fractures and its implications to gene search for complex human diseases. Genetic Epidemiology2002;22:12–25.[CrossRef][Web of Science][Medline]

    2. Huang QY, Recker RR & Deng HW. Searching for osteoporosis genes in the post-genome era: progress and challenges. Osteoporosis International2003;14:701–715.[CrossRef][Web of Science][Medline]

    3. Liu YJ, Shen H, Xiao P, Xiong DH, Li LH, Recker RR & Deng HW. Molecular genetic studies of gene identification for osteoporosis: a 2004 update. Journal of Bone and Mineral Research2006;21:1511–1535.[CrossRef][Web of Science][Medline]

    4. Kitamura I, Ando F, Koda M, Okura T & Shimokata H. Effects of the interaction between lean tissue mass and estrogen receptor alpha gene polymorphism on bone mineral density in middle-aged and elderly Japanese. Bone2007;40:1623–1629.[Web of Science][Medline]

    5. Xiong DH, Shen H, Zhao LJ, Xiao P, Yang TL, Guo Y, Wang W, Guo YF, Liu YJ, Recker RR & Deng HW. Robust and comprehensive analysis of 20 osteoporosis candidate genes by very high-density single-nucleotide polymorphism screen among 405 white nuclear families identified significant association and gene–gene interaction. Journal of Bone and Mineral Research2006;21:1678–1695.[Web of Science][Medline]

    6. Dvornyk V, Liu XH, Shen H, Lei SF, Zhao LJ, Huang QR, Qin YJ, Jiang DK, Long JR, Zhang YY, Gong G, Recker RR & Deng HW. Differentiation of Caucasians and Chinese at bone mass candidate genes: implication for ethnic difference of bone mass. Annals of Human Genetics2003;67:216–227.[CrossRef][Web of Science][Medline]

    7. Kung AW, Lai BM, Ng MY, Chan V & Sham PC. T-1213C polymorphism of estrogen receptor beta is associated with low bone mineral density and osteoporotic fractures. Bone2006;39:1097–1106.[Web of Science][Medline]

    8. Foroud T, Ichikawa S, Koller D, Lai D, Curry L, Xuei X, Edenberg HJ, Hui S, Peacock M & Econs MJ. Association studies of ALOX5 and bone mineral density in healthy adults. Osteoporosis International2007;19:637–643.[Web of Science][Medline]

    9. Lei SF, Jiang H, Deng FY & Deng HW. Searching for genes underlying susceptibility to osteoporotic fracture: current progress and future prospect. Osteoporosis International2007;18:1157–1175.[CrossRef][Web of Science][Medline]

    10. Bustamante M, Nogues X, Enjuanes A, Elosua R, Garcia-Giralt N, Perez-Edo L, Caceres E, Carreras R, Mellibovsky L, Balcells S, Diez-Perez A & Grinberg D. COL1A1, ESR1, VDR and TGFB1 polymorphisms and haplotypes in relation to BMD in Spanish postmenopausal women. Osteoporosis International2007;18:235–243.[CrossRef][Web of Science][Medline]

    11. Riancho JA, Valero C, Naranjo A, Morales DJ, Sanudo C & Zarrabeitia MT. Identification of an aromatase haplotype that is associated with gene expression and postmenopausal osteoporosis. Journal of Clinical Endocrinology and Metabolism2007;92:660–665.[Abstract/Free Full Text]

    12. Borsy A, Balla B, Gyurján I, Stéger V, Molnár A, Szabolcsi Z, Kósa J, Zomborszky Z, Papp P, Vellai T, Lakatos P & Orosz L. Red deer biology for biomedical research on human osteoporosis. 33rd European Symposium on Calcified Tissues. Calcified Tissue International2006;78:S74

    13. Balla B, Kósa JP, Kiss J, Borsy A, Podani J, Takács I, Lazáry A, Nagy Z, Bácsi K, Speer G, Orosz L & Lakatos P. Different gene expression patterns in the bone tissue of aging postmenopausal osteoporotic and non-osteoporotic women. Calcified Tissue International2008;82:12–26.[Web of Science][Medline]

    14. Wang L, Liu S, Niu T & Xu X. SNPHunter: a bioinformatic software for single nucleotide polymorphism data acquisition and management. BMC Bioinformatics2005;6:60[CrossRef][Medline]

    15. Bell PA, Chaturverdi S, Gelfand CA, Huang CY, Kochersperger M, Kopla R, Modica F, Pohl M, Varde S, Zhao R, Zhao X, Boyce-Jacino MT & Yassen A. SNPstream UHT: ultra-high throughput SNP genotyping for pharmacogenomics and drug discovery. Biotechniques32Suppl200270–77.

    16. Xie Q, Ratnasinghe LD, Hong H, Perkins R, Tang ZZ, Hu N, Taylor PR & Tong W. Decision forest analysis of 61 single nucleotide polymorphisms in a case–control study of esophageal cancer; a novel method. BMC Bioinformatics2005;6:S4

    17. Barrett JC, Fry B, Maller J & Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics2005;21:263–265.[Abstract/Free Full Text]

    18. Curtin K, Wong J, Allen-Brady K & Camp NJ. PedGenie: meta genetic association testing in mixed family and case–control designs. BMC Bioinformatics2007;8:448[Medline]

    19. Gauderman WJ & Morrison JM. QUANTO 1.1: a computer program for power and sample size calculations for genetic-epidemiology studies. http://hydra.usc.edu/gxe, 2006.

    20. Schaid DJ, Rowland CM, Tines DE, Jacobson RM & Poland GA. Score tests for association between traits and haplotypes when linkage phase is ambiguous. American Journal of Human Genetics2002;70:425–434.[CrossRef][Web of Science][Medline]

    21. Gonzalez JR, Armengol L, Sole X, Guino E, Mercader JM, Estivill X & Moreno V. SNPassoc: an R package to perform whole genome association studies. Bioinformatics2007;23:644–645.[Web of Science][Medline]

    22. Messeguer X, Escudero R, Farre D, Nunez O, Martinez J & Alba MM. PROMO: detection of known transcription regulatory elements using species-tailored searches. Bioinformatics2002;18:333–334.[Abstract/Free Full Text]

    23. Woei Ng K, Speicher T, Dombrowski C, Helledie T, Haupt LM, Nurcombe V & Cool SM. Osteogenic differentiation of murine embryonic stem cells is mediated by fibroblast growth factor receptors. Stem Cells and Development2007;16:305–318.[Web of Science][Medline]

    24. Aguirre JI, Leal ME, Rivera MF, Vanegas SM, Jorgensen M & Wronski TJ. Effects of basic fibroblast growth factor and a prostaglandin E2 receptor subtype 4 agonist on osteoblastogenesis and adipogenesis in aged ovariectomized rats. Journal of Bone and Mineral Research2007;22:877–888.[Web of Science][Medline]

    25. Kawaguchi H, Katagiri M & Chikazu D. Osteoclastic bone resorption through receptor tyrosine kinase and extracellular signal-regulated kinase signaling in mature osteoclasts. Modern Rheumatology2004;14:1–5.[Medline]

    26. Vieira AR, Modesto A, Meira R, Barbosa AR, Lidral AC & Murray JC. Interferon regulatory factor 6 (IRF6) and fibroblast growth factor receptor 1 (FGFR1) contribute to human tooth agenesis. American Journal of Medical Genetics. Part A2007;143:538–545.[Medline]

    27. Coussens AK & Van Daal A. Linkage disequilibrium analysis identifies an FGFR1 haplotype-tag SNP associated with normal variation in craniofacial shape. Genomics2005;85:563–573.[Web of Science][Medline]

    28. Chmurzynska A. The multigene family of fatty acid-binding proteins (FABPs): function, structure and polymorphism. Journal of Applied Genetics2006;47:39–48.[Web of Science][Medline]

    29. Xu Y, Mirmalek-Sani SH, Yang X, Zhang J & Oreffo RO. The use of small interfering RNAs to inhibit adipocyte differentiation in human preadipocytes and fetal-femur-derived mesenchymal cells. Experimental Cell Research2006;10:1856–1864.

    30. Shin HD, Kim LH, Park BL, Jung HS, Cho YM, Moon MK, Lee HK & Park KS. Polymorphisms in fatty acid-binding protein-3 (FABP3) – putative association with type 2 diabetes mellitus. Human Mutation2003;22:180[Medline]

    31. Moerman EJ, Teng K, Lipschitz DA & Lecka-Czernik B. Aging activates adipogenic and suppresses osteogenic programs in mesenchymal marrow stroma/stem cells: the role of PPAR-{gamma}2 transcription factor and TGF-β/BMP signaling pathways. Aging Cell2004;3:379–389.[CrossRef][Web of Science][Medline]

    32. Nelson-Dooley C, Della-Fera MA, Hamrick M & Baile CA. Novel treatments for obesity and osteoporosis: targeting apoptotic pathways in adipocytes. Current Medicinal Chemistry2005;12:2215[CrossRef][Web of Science][Medline]

    33. Jadhav SB & Jain GK. Statins and osteoporosis: new role for old drugs. Journal of Pharmacy and Pharmacology2006;58:3–18.[Web of Science][Medline]

    34. Nuttall ME & Gimble JM. Controlling the balance between osteoblastogenesis and adipogenesis and the consequent therapeutic implications. Current Opinion in Pharmacology2004;4:290–294.[CrossRef][Web of Science][Medline]

    35. Page-McCaw A, Ewald AJ & Werb Z. Matrix metalloproteinases and the regulation of tissue remodelling. Nature Reviews. Molecular Cell Biology2007;8:221–233.[CrossRef][Web of Science][Medline]

    36. Martignetti JA, Aqeel AA, Sewairi WA, Boumah CE, Kambouris M, Mayouf SA, Sheth KV, Eid WA, Dowling O, Harris J, Glucksman MJ, Bahabri S, Meyer BF & Desnick RJ. Mutation of the matrix metalloproteinase 2 gene (MMP2) causes a multicentric osteolysis and arthritis syndrome. Nature Genetics2001;28:261–265.[CrossRef][Web of Science][Medline]

    37. Guo LJ, Luo XH, Wu XP, Shan PF, Zhang H, Cao XZ, Xie H & Liao EY. Serum concentrations of MMP-1, MMP-2, and TIMP-1 in Chinese women: age-related changes, and the relationships with bone biochemical markers, bone mineral density. Clinica Chimica Acta2006;371:137–142.[Web of Science][Medline]

    38. Yamada Y, Ando F, Niino N & Shimokata H. Association of a polymorphism of the matrix metalloproteinase-1 gene with bone mineral density. Matrix Biology2002;21:389–392.[Web of Science][Medline]

    39. Yamada Y, Ando F, Niino N & Shimokata H. Association of a polymorphism of the matrix metalloproteinase-9 gene with bone mineral density in Japanese men. Metabolism2004;53:135–137.[CrossRef][Web of Science][Medline]

    40. Mendes O, Kim HT, Lungu G & Stoica G. MMP2 role in breast cancer brain metastasis development and its regulation by TIMP2 and ERK1/2. Clinical & Experimental Metastasis2007;24:341–351.[Web of Science][Medline]

    41. Wang K, Vishwanath P, Eichler GS, Al-Sebaei MO, Edgar CM, Einhorn TA, Smith TF & Gerstenfeld LC. Analysis of fracture healing by large-scale transcriptional profile identified temporal relationships between metalloproteinase and ADAMTS mRNA expression. Matrix Biology2006;25:271–281.[CrossRef][Web of Science][Medline]

    42. Cao J, Venton L, Sakata T & Halloran BP. Expression of RANKL and OPG correlates with age-related bone loss in male C57BL/6 mice. Journal of Bone and Mineral Research2003;18:270–277.[CrossRef][Web of Science][Medline]

    43. Taillandier A, Lia-Baldini AS, Mouchard M, Robin B, Muller F, Simon-Bouy B, Serre JL, Bera-Louville A, Bonduelle M, Eckhardt J, Gaillard D, Myhre AG, Kortge-Jung S, Larget-Piet L L, Malou E, Sillence D, Temple IK, Viot G & Mornet E. Twelve novel mutations in the tissue-nonspecific alkaline phosphatase gene (ALPL) in patients with various forms of hypophosphatasia. Human Mutation2001;18:83–84.[Medline]

    44. Goseki-Sone M, Sogabe N, Fukushi-Irie M, Mizoi L, Orimo H, Suzuki T, Nakamura H, Orimo H & Hosoi T. Functional analysis of the single nucleotid polymorphism (787T>C) in the tissu-nonspecific alkaline phophatase gene associated with BMD. Journal of Bone and Mineral Research2005;20:773–782.[Web of Science][Medline]

Received 30 April 2008
Accepted 7 May 2008





This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
EJE-08-0021v1
159/2/187    most recent
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lazáry, A.
Right arrow Articles by Lakatos, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Lazáry, A.
Right arrow Articles by Lakatos, P.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS