Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Pathway-Based Association Analyses Identified TRAIL Pathway for Osteoporotic Fractures

  • Yin-Ping Zhang ,

    cathyzh@mail.xjtu.edu.cn (Y-PZ); hdeng2@tulane.edu (H-WD)

    Affiliation The Key Laboratory of Environment and Genes Related to Diseases, Xi'an Jiaotong University College of Medicine, Ministry of Education, Xi'an, Shaanxi, People's Republic of China

  • Yao-Zhong Liu,

    Affiliation Department of Biostatistics, Tulane University School of Public Health and Tropical Medicine, New Orleans, Louisiana, United States of America

  • Yan Guo,

    Affiliation The Key Laboratory of Biomedical Information Engineering, Xi'an Jiaotong University School of Life Science and Technology, Ministry of Education and Institute of Molecular Genetics, Xi'an, Shaanxi, People's Republic of China

  • Xiao-Gang Liu,

    Affiliation The Key Laboratory of Biomedical Information Engineering, Xi'an Jiaotong University School of Life Science and Technology, Ministry of Education and Institute of Molecular Genetics, Xi'an, Shaanxi, People's Republic of China

  • Xiang-Hong Xu,

    Affiliation The Key Laboratory of Biomedical Information Engineering, Xi'an Jiaotong University School of Life Science and Technology, Ministry of Education and Institute of Molecular Genetics, Xi'an, Shaanxi, People's Republic of China

  • Yan-Fang Guo,

    Affiliation The Key Laboratory of Biomedical Information Engineering, Xi'an Jiaotong University School of Life Science and Technology, Ministry of Education and Institute of Molecular Genetics, Xi'an, Shaanxi, People's Republic of China

  • Yuan Chen,

    Affiliation The Key Laboratory of Biomedical Information Engineering, Xi'an Jiaotong University School of Life Science and Technology, Ministry of Education and Institute of Molecular Genetics, Xi'an, Shaanxi, People's Republic of China

  • Feng Zhang,

    Affiliation The Key Laboratory of Environment and Genes Related to Diseases, Xi'an Jiaotong University College of Medicine, Ministry of Education, Xi'an, Shaanxi, People's Republic of China

  • Feng Pan,

    Affiliation The Key Laboratory of Biomedical Information Engineering, Xi'an Jiaotong University School of Life Science and Technology, Ministry of Education and Institute of Molecular Genetics, Xi'an, Shaanxi, People's Republic of China

  • Xue-Zhen Zhu,

    Affiliation University of Shanghai for Science and Technology, Shanghai, People's Republic of China

  • Hong-Wen Deng

    cathyzh@mail.xjtu.edu.cn (Y-PZ); hdeng2@tulane.edu (H-WD)

    Affiliations Department of Biostatistics, Tulane University School of Public Health and Tropical Medicine, New Orleans, Louisiana, United States of America, The Key Laboratory of Biomedical Information Engineering, Xi'an Jiaotong University School of Life Science and Technology, Ministry of Education and Institute of Molecular Genetics, Xi'an, Shaanxi, People's Republic of China, University of Shanghai for Science and Technology, Shanghai, People's Republic of China

Abstract

Introduction

Hip OF carries the highest morbidity and mortality. Previous studies revealed that individual genes/loci in the Tumor Necrosis Factor (TNF) -Related Apoptosis-Inducing Ligand (TRAIL) pathway were associated with bone metabolism. This study aims to verify the potential association between hip OF and TRAIL pathway.

Methods

Using genome-wide genotype data from Affymetrix 500 K SNP arrays, we performed novel pathway-based association analyses for hip OF in 700 elderly Chinese Han subjects (350 with hip OF and 350 healthy matched controls).

Results

The TRAIL pathway achieved a significant p value (p = 0.01) for association with hip OF. Among the 38 genes in the TRAIL pathway, seven genes achieved nominally significant association with hip OF (p<0.05); the TNFSF10 (TRAIL) gene obtained the most significant p value (p = 1.70×10−4). SNPs (rs719126, rs6533015, rs9594738, rs1805034, rs11160706) from five genes (CFLAR, NFKB1, TNFSF11, TNFRSF11A, TRAF3) of the pathway had minor alleles that appear to be protective to hip OF. SNPs (rs6445063 and rs4259415) from two genes (TNFSF10 and TNFRSF10B) of the pathway had minor alleles (A) that are associated with an increased risk of hip OF, with the ORs (odds ratios) of 16.51 (95%CI:3.83–71.24) and 1.37 (95%CI:1.08–1.74), respectively.

Conclusions

Our study supports the potential role of the TRAIL pathway in the pathogenesis of hip OF in Chinese Han population. Further functional study of this pathway will be pursued to determine the mechanism by which it confers risk to hip OF.

Introduction

The prevalence of osteoporotic fractures (OF), especially hip OF, has now become a significant public health burden due to the associated high morbidity, mortality and tremendous health care cost [1], [2] The incidence of hip OF will reach 6.27 million worldwide with a resultant cost of >$34 billion by the year 2050 [1]. The major demographic change of hip OF will occur in Asia. In 1990, 26% of hip OF worldwide occurred in Asia, and this rate could rise to 37% by the year 2025 and to 45% by the year 2050 [2]. Genetic factors play an important role in susceptibility to hip OF [3], [4]. However, to date, the genetic determination of hip OF is still largely unknown.

A pathway-based association analysis approach is based on gene set enrichment analysis. The method ranks genes/SNPs genome-wide by the significance of association with a disease phenotype to generate a gene list, then uses a statistic enrichment score () to examine whether a particular group of genes/SNPs within a certain functional pathway is enriched at the top (or bottom) of the list more than would be expected by chance [5]. This approach considers critical information about the interaction of a set of functionally related genes and their joint effects, and could potentially improve the power to detect genetic variants working in functional pathways. The method may play a major role in genetic analyses of complex diseases; for a pathway that contributes to a disease's risk, a single genetic variant within that pathway may have only limited contribution to the risk that might not be detectable (in regular genetic association analyses) if not otherwise be detected at the whole pathway level. Recently, through this new powerful approach, we have identified two novel pathways for wrist BMD and femur geometry in US whites [6], [7].

Previous studies revealed importance in the pathogenesis of bone metabolism for several individual genes/loci, or their transcripts and proteins in Tumor Necrosis Factor (TNF)-Related Apoptosis-Inducing Ligand (TRAIL) pathway, for example, TRAIL, TRAILR1, TRAILR2, CFLAR, RANK, and OPG [8], [9], [10]. However, it is unknown if the TRAIL pathway may contribute to osteoporotic fracture risk at the whole pathway level. We therefore undertook a novel, pathway-based association study to assess contribution of the pathway to hip OF risk.

Materials and Methods

Subjects

The study was approved by the institutional review board of Xi'an Jiaotong University. After signing an informed consent agreement, all subjects were assisted to complete a structured questionnaire including anthropometric variables, lifestyles, and medical history.

The sample consisted of 700 elderly Chinese Han subjects, including 350 with osteoporotic (low trauma) hip fractures (including 124 males and 226 females) and 350 controls (including 173 males and 177 females) (see Table 1 for the basic characteristics of these subjects). All the subjects were unrelated Chinese Han adults living in the city of Xi'an and its neighboring areas. Inclusion criteria for cases were (i) onset age >55 years, all female subjects were postmenopausal women; (ii) age <80 years to minimize the effect due to age, since previous studies showed that approximately half of females aged 80 years or older have fractures [11]; (iii) minimal or no trauma fractures, usually due to falls from standing height or less; (iv) fracture site at femoral neck or inter-trochanter regions; (v) hip fracture was identified/confirmed through diagnosis of orthopedic surgeons/radiologists according to radiological reports and X-rays. Patients with pathological fractures and high-impact fractures (such as due to motor vehicle accidents) were excluded.

Healthy control subjects were selected from our established large database. They were geographically matched to the cases. Inclusion criteria for controls were: (i) age at exam >55 years, without any fracture histories (all female controls were postmenopausal); and (ii) not suffering from chronic diseases and conditions that might potentially affect bone mass, structure, or metabolism. The criteria may ensure that controls are less likely to suffer OF during their remaining life span compared with the general populations. The diseases/conditions mentioned above included chronic disorders involving vital organs (heart, lung, liver, kidney, brain), serious metabolic diseases (diabetes, hypo- and hyper-parathyroidism, hyperthyroidism, etc.), other skeletal diseases (Paget disease, osteogenesis imperfecta, rheumatoid arthritis, etc.); chronic use of drugs affecting bone metabolism (e.g., hormone replacement therapy, corticosteroid therapy, anti-convulsant drugs), and malnutrition conditions (such as chronic diarrhea, chronic ulcerative colitis). In addition, subjects taking anti-resorptive or bone anabolic agents/drugs, such as bisphosphonates were also excluded.

Genotyping and quality control

Genomic DNA was extracted from peripheral blood leukocytes using a commercial isolation kit (Gentra systems, Minneapolis, MN, USA) following the standard protocols. SNP genotyping was performed using Affymetrix Human Mapping 500 K array set (Affymetrix, Santa Clara, CA, USA). Genotyping calls were determined from the fluorescent intensities using the dynamic model (DM) algorithm [12] as well as the B-RLMM (Bayesian Robust Linear Model with Mahalanobis Distance Classifier) algorithm [13]. DM calls were used for quality control of the genotyping experiment. BRLMM calls were used for the following pathway-based analyses. SNPs with a sample call rate <90%, with allele frequencies extremely deviating from Hardy-Weinberg equilibrium (p<10−7) and having a minor allele frequency (MAF) <0.01 in the total sample were discarded. The final SNP number for the analyses is 371,258. This SNP set covers 14,642 genes and yielded an average marker spacing of ∼7.9 kb throughout the human genome. Among these SNPs, 350 belong to the genes of the TRAIL pathway.

Statistical Analysis

Statistical analyses on individual SNPs were first conducted by the Wald test implemented in Plink (version 1.03) [14] with age and sex as covariates. The main procedures of pathway-based analysis [15] are briefly summarized as follows:

  1. Generation of gene-phenotype association rank and calculation of : Among all the SNPs of a given gene Gi, the SNP achieving the smallest p value in the single-marker association tests was used to represent the magnitude of association evidence of the gene. We ranked all genes by sorting their statistic values from the largest to smallest, denoted as a vector of the gene list L (r1,r2,…,rN). “N” equals to the total gene number contained in the GeneChip® Human Mapping 500 K set arrays. “r” represents the gene-phenotype association statistic value. Then, a weighted Kolmogorov-Smirnov–like running-sum enrichment score () was calculated in Equation 1. For the given pathway S composed of NS genes, by walking down the gene list L, we increased the running-sum statistic for the pathway when we encountered a gene in the S and decreased it when we encountered a gene not in the .(1)Where and parameter p (designated as 1 here) is designed to give higher weight to genes with extreme statistic values. The magnitude of the increment depends on the correlation of the gene with the phenotype. In short, is the maximum deviation from zero encountered in the random walk. It will be high if the association signal is enriched at the top of list , reflected by the significance level of observed (i.e. nominal p value).
  2. Permutation and nominal significance assessment: The phenotype data was shuffled and permutation () was done to compute a through repeating steps (1)∼(2) to estimate the nominal p value. A total of 1,000 permutations were done to create a histogram of the corresponding enrichment scores for the given pathway/gene set . The nominal p value was estimated as the percentage of permutations whose were greater than the observed .

To detect possible population stratification that may lead to spurious association results, we used Structure 2.2 software (http://pritch.bsd.uchicago.edu/software.html) to investigate the potential population substructure/stratification of the sample. The program uses a Markov chain Monte Carlo (MCMC) algorithm to cluster individuals into different cryptic sub-populations on the basis of multi-locus genotype data [16]. We performed the analysis assuming the number of population strata k = 2 and using 1,000 un-linked SNPs randomly selected genome-wide. To confirm the results achieved through Structure 2.2, we further tested population stratification in our sample using EIGENSOFT 2.0 software that uses both principal component analysis and a genomic control approach to correct possible statistical bias caused by ancestral differences [17], [18]. Reported p value was adjusted by the λ value estimated by genomic control method.

Results

Characteristics of study subjects

Basic characteristics of the 700 subjects are presented in Table 1. The STRUCTURE program revealed that all subjects in this Chinese sample were clustered together and could not be assigned into any subgroups, indicating that there was no significant population stratification within the sample. We also performed analysis using EIGENSOFT [17] and confirmed the homogeneity of the sample revealed by Structure; there is only one principal component that is significant in the principal component analysis (p<0.001), suggesting only one population ancestry existing for our sample. Finally, we used the genomic control method [18] to estimate the λ value, which equals to 1.02, suggesting again there is no population stratification in the sample.

Pathway-based association analysis

The TRAIL pathway, annotated by the Ambion GeneAssist Pathway Atlas (Table 2), achieved a high NES of 1.76 with a p value of 0.01. Among the 38 genes in the TRAIL pathway, the gene with the most significant p value is TNFSF10 (TRAIL, p = 1.70×10−4) (Table 3). Another six genes in the TRAIL pathway, which also contribute positively to the ES (i.e., the genes that ranked before or at the point of ES, also denoted as “leading edge genes”), are: CFLAR (c-FLIP, FLIP, p = 3.17×10−3), TNFSF11 (RANKL, p = 6.06×10−3), TNFRSF11A (RANK, p = 7.63×10−3), TNFRSF10B (TRAILR2, p = 1.07×10−2), TRAF3 (p = 1.67×10−2), and NFKB1 (NF-κB1, p = 2.04×10−2). Among the 38 genes in the TRAIL pathway, 4 genes, NFKBIE, IKBKG, RELB, DIABLO, were not covered by Affymetrix 500 k Array Set, and the SNPs of the XIAP and APAF1 genes were discarded for failure of passing the quality control criteria (Table 3).

thumbnail
Table 3. Association Results for the SNPs of Top Significance for Each Gene in the TRAIL Pathway.

https://doi.org/10.1371/journal.pone.0021835.t003

For the top SNP of each gene, the effect size of the minor allele is expressed by odds ratio (OR) derived from logistic regression analyses. The OR value is interpreted in the standard manner. A value of 1.0 indicates no effect. A value greater than 1.0 indicates that the minor allele is associated with an increased hip OF risk, and a value less than 1.0 implies that the minor allele is associated with a decreased risk, hence may be protective. Among the 7 SNPs representing the 7 leading edge genes, 5 SNPs had the protective minor allele. They are: rs719126 of the CFLAR gene (minor allele A, OR = 0.05, 95%CI: 0.01–0.36), rs6533015 of the NFKB1 gene (minor allele C,OR = 0.76, 95%CI: 0.60–0.96), rs9594738 of the TNFSF11 gene (minor allele C, OR = 0.55, 95%CI: 0.36–0.84), rs1805034 of the TNFRSF11A gene (minor allele C, OR = 0.72, 95%CI: 0.57–0.92), and rs11160706 of the TRAF3 gene (minor allele A, OR = 0.75, 95%CI: 0.60–0.95). The other two SNPs, rs6445063 in the TNFSF10 (TRAIL) gene and rs4259415 in the TNFRSF10B (TRAILR2) gene, had a minor allele “A” with OR values of 16.51 (95%CI: 3.83–71.24) and 1.37 (95%CI: 1.08–1.74), respectively. The minor alleles of these two SNPs are significantly associated with an increased risk of hip OF (P<0.05) (Table 3).

Discussion

In this study, the TRAIL pathway was shown to be significantly associated with hip OF (p = 0.01) according to the pathway-based association analysis. This is the first time that the TRAIL pathway is implicated as an underlying factor for hip OF in Chinese. The pathway may contribute to bone mass variation via regulating osteoclast metabolism; previous findings have shown that the TRAIL pathway play roles in modulating the differentiation, function, survival and/or apoptosis of osteoclasts [19], [20], which may in turn accelerate bone resorption and consequently the susceptibility to hip OF. The core process of TRAIL and the functional interactions among the genes in the TRAIL pathway are depicted in Figure 1.

thumbnail
Figure 1. TRAIL pathway and the gene-gene interaction.

The binding of TRAIL to TRAILR2 may induce the apoptosis of osteoclast. Increased c-FLIP levels may decrease the apoptosis of Osteoclast. The osteoclast apoptosis signal transduced by TRAIL/TRAILR2 was transformed to activate NF-κB. A key step in downstream signaling of RANKL/RANK is binding of TRAFs to RANK. Stimulation of RANK also results in strong NF-κB activation. NF-κB1(p50) is an important signal for osteoclast development and osteoclast function.

https://doi.org/10.1371/journal.pone.0021835.g001

TRAIL, also known as Apo2L, is a widely recognized member of the tumor necrosis factor (TNF) ligand family [21], [22]. TRAIL is a typical type II membrane protein, binds to aggregates type I transmembrane receptors with cytoplasmic death domains (DD), which ultimately activate a protease cascade leading to apoptosis [23]. Interacting with five different receptors of the TNF receptor superfamily, TRAILR1, TRAILR2, TRAILR3, TRAILR4 and OPG, the ligand is expressed constitutively in a wide range of tissues [22], [24], [25], including bone-related cells, such as osteoclast or osteoclast precursors. The binding of TRAIL to TRAILR2 (DR5), which contains a conserved death domain (DD), may induce the apoptosis of osteoclast [19]. This may play an important role in the etiology of hip OF. TRAIL induces apoptosis in human differentiated osteoclasts by means of TRAILR2, and activates an intracellular pathway involving caspase-8 and Bid cleavage in the active forms [19]. However, in this study, SNP rs6445063 from TNFSF10 (TRAIL) gene and SNP rs4259415 from TNFRSF10B (TRAILR2) gene of the pathway had minor alleles (A) that are associated with an increased risk of hip OF, with the ORs (odds ratios) of 16.51 (95%CI:3.83–71.24) and 1.37 (95%CI:1.08–1.74), respectively. SNP rs6445063 is located on chromosome 3q26.31 and has never been studied previously. SNP rs4259415 is located on chromosome 8p21.3 and functions as intronic enhancer. According to the FASTSNP program (http://fastsnp.ibms.sinica.edu.tw), a change of “G→A” at rs4259415 may lead to alter the binding sites for transcription factor from GATA-1 to CdxA and TATA, which may change the function of apoptosis related TRAILR2 gene and accordingly alter the apoptosis of osteoclast. Furthermore, the apoptosis induced by death receptors can be modulated at several levels. Intracellular anti-apoptotic molecules can block the apoptotic signaling pathway. Soluble receptors inhibit TRAIL-mediated apoptosis in part by increasing baseline levels of CFLAR (c-FLIP), which competes with caspase-8 for binding to FADD, and inhibits active caspases [26]. Increased c-FLIP levels following exposure to soluble factors may decrease the apoptosis of osteoclast. Intracellular anti-apoptotic molecules can also divert the apoptosis signaling into alternative responses, i.e., the proliferation/survival of osteoclast via activating NF-κB [20]. Although lack of bone phenotype in mice with gene deletion of TRAIL was observed by Sedger et al [27], Yen et al [28] showed that TRAIL can induce osteoclast formation via direct engagement with the TRAIL death receptor. Yen et al [28], [29] suggested that TRAIL-induced osteoclastogenesisis was dependent on activation of NF-κB, ERK, and p38 MAP kinase and TRAF6 dependent. However TRAF6 was not found to contribute positively to the ES as TRAF3 in this study, suggests that there may be other factors that also can function in this capacity.

As the soluble decoy receptor of TRAIL, OPG was originally identified as a decoy receptor for RANKL [30] later found to be able to bind TRAIL [31]. OPG acts as antagonist for osteoclast apoptosis induced by TRAIL. But when OPG binds to the RANKL on the surface of osteoblast/bone matrix, it prevents RANKL from binding to its receptor, RANK, which inhibits the formation, activation and survival of multinucleated osteoclasts. In this process, the RANKL/OPG ratio is an important determinant of bone mass and skeletal integrity [32]. Up-regulation of the RANKL gene increases the RANKL/OPG ratio and enhances bone loss [32]. Our study found a significant association of SNPs (rs9594738 and rs1805034) of RANKL and RANK with hip OF.

NF-κB1 (p50) is a subunit of NF-κB transcriptional factor complex. The balance of survival (anti-apoptotic) and death (apoptotic) signals through NF-κB activation cascades results in normal bone homeostasis and healthy remodeling [33]. NF-κB activation through RANK/RANKL signal pathway is closely related with osteoporosis and the bone resorbing activity of osteoclasts [34]. After RANKL binds to RANK, a key step in downstream signaling is binding of TRAFs to specific sites of the cytoplasmic domain of RANK [20], [35], [36]. Stimulation of RANK results in strong NF-κB activation. NF-κB1(p50) is an important signal for osteoclast development and osteoclast function. Moreover, NF-κB can be activated in many signal cascades related to bone metabolism, e.g., NF-κB activation through the Fas/FasL system leads to enhanced osteoclastogenesis and reduced apoptosis, and it may also cause apoptotic cell death in osteoblasts [37].

Population stratification is an important source of spurious association in genetic association studies [38], [39]. However, these factors are unlikely to exist in our sample to interfere with our pathway-based association results. Our cohort came from an apparently homogenous Chinese north-west Han ethnicity population, living in Xi'an city and its neighboring areas. More importantly, in the analyses using Structure 2.2 [40], all subjects used in our study consistently clustered together as a single group, suggesting no significant population substructure. In the analysis using EIGENSTRAT [17], the measure for population stratification, λ, was very close to 1, which also suggests no stratification in our cohort. For the above reasons, our association results are unlikely to be plagued by spurious associations due to population admixture/stratification.

In summary, we applied a pathway-based analysis method to identify functional gene sets associated with hip OF. The significant enrichment of the TRAIL pathway genes among the top ranking genes associated with hip OF, together with the pathway's functional relevance to bone metabolism, strongly supports the important role of TRAIL in human hip OF risk. Further detailed and specific functional studies of the TRAIL pathway will be pursued to provide new insights into the etiology of hip OF.

Acknowledgments

We thank all the study subjects for volunteering to participate in the study.

Author Contributions

Conceived and designed the experiments: Y-PZ H-WD. Performed the experiments: Y-PZ YG X-HX YC FP X-ZZ. Analyzed the data: X-GL Y-FG FZ. Wrote the paper: Y-PZ Y-ZL.

References

  1. 1. Cooper C, Campion G, Melton LJ 3rd (1992) Hip fractures in the elderly: a world-wide projection. Osteoporosis international : a journal established as result of cooperation between the European Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA 2: 285–289.
  2. 2. Gullberg B, Johnell O, Kanis JA (1997) World-wide projections for hip fracture. Osteoporosis international : a journal established as result of cooperation between the European Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA 7: 407–413.
  3. 3. Tranah GJ, Taylor BC, Lui LY, Zmuda JM, Cauley JA, et al. (2008) Genetic variation in candidate osteoporosis genes, bone mineral density, and fracture risk: the study of osteoporotic fractures. Calcified tissue international 83: 155–166.
  4. 4. Styrkarsdottir U, Halldorsson BV, Gretarsdottir S, Gudbjartsson DF, Walters GB, et al. (2008) Multiple genetic loci for bone mineral density and fractures. The New England journal of medicine 358: 2355–2365.
  5. 5. Curtis RK, Oresic M, Vidal-Puig A (2005) Pathways to the analysis of microarray data. Trends in biotechnology 23: 429–435.
  6. 6. Zhang L, Guo YF, Liu YZ, Liu YJ, Xiong DH, et al. (2010) Pathway-based genome-wide association analysis identified the importance of regulation-of-autophagy pathway for ultradistal radius BMD. J Bone Miner Res 25(7): 1572–80.
  7. 7. Chen Y, Xiong DH, Guo YF, Pan F, Zhou Q, et al. (2010) Pathway-based genome-wide association analysis identified the importance of EphrinA-EphR pathway for femoral neck bone geometry. Bone 46: 129–136.
  8. 8. Vaira S, Alhawagri M, Anwisye I, Kitaura H, Faccio R, et al. (2008) RelA/p65 promotes osteoclast differentiation by blocking a RANKL-induced apoptotic JNK pathway in mice. The Journal of clinical investigation 118: 2088–2097.
  9. 9. Sanjay A, Houghton A, Neff L, DiDomenico E, Bardelay C, et al. (2001) Cbl associates with Pyk2 and Src to regulate Src kinase activity, alpha(v)beta(3) integrin-mediated signaling, cell adhesion, and osteoclast motility. The Journal of cell biology 152: 181–195.
  10. 10. Yao GQ, Sun BH, Insogna KL, Weir EC (2000) Nuclear factor-kappaB p50 is required for tumor necrosis factor-alpha-induced colony-stimulating factor-1 gene expression in osteoblasts. Endocrinology 141: 2914–2922.
  11. 11. Ross PD (1998) Risk factors for osteoporotic fracture. Endocrinol Metab Clin North Am 27: 289–301.
  12. 12. Di X, Matsuzaki H, Webster TA, Hubbell E, Liu G, et al. (2005) Dynamic model based algorithms for screening and genotyping over 100 K SNPs on oligonucleotide microarrays. Bioinformatics (Oxford, England) 21: 1958–1963.
  13. 13. Rabbee N, Speed TP (2006) A genotype calling algorithm for affymetrix SNP arrays. Bioinformatics (Oxford, England) 22: 7–12.
  14. 14. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, et al. (2007) PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81: 559–575.
  15. 15. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, et al. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences of the United States of America 102: 15545–15550.
  16. 16. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155: 945–959.
  17. 17. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, et al. (2006) Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet 38: 904–909.
  18. 18. Devlin B, Roeder K (1999) Genomic control for association studies. Biometrics 55: 997–1004.
  19. 19. Colucci S, Brunetti G, Cantatore FP, Oranger A, Mori G, et al. (2007) The death receptor DR5 is involved in TRAIL-mediated human osteoclast apoptosis. Apoptosis 12: 1623–1632.
  20. 20. Feng X (2005) Regulatory roles and molecular signaling of TNF family members in osteoclasts. Gene 350: 1–13.
  21. 21. Pitti RM, Marsters SA, Ruppert S, Donahue CJ, Moore A, et al. (1996) Induction of apoptosis by Apo-2 ligand, a new member of the tumor necrosis factor cytokine family. The Journal of biological chemistry 271: 12687–12690.
  22. 22. Wiley SR, Schooley K, Smolak PJ, Din WS, Huang CP, et al. (1995) Identification and characterization of a new member of the TNF family that induces apoptosis. Immunity 3: 673–682.
  23. 23. Pan G, O'Rourke K, Chinnaiyan AM, Gentz R, Ebner R, et al. (1997) The receptor for the cytotoxic ligand TRAIL. Science (New York, NY) 276: 111–113.
  24. 24. Chamoux E, Houde N, L'Eriger K, Roux S (2008) Osteoprotegerin decreases human osteoclast apoptosis by inhibiting the TRAIL pathway. J Cell Physiol 216: 536–542.
  25. 25. Zauli G, Rimondi E, Nicolin V, Melloni E, Celeghini C, et al. (2004) TNF-related apoptosis-inducing ligand (TRAIL) blocks osteoclastic differentiation induced by RANKL plus M-CSF. Blood 104: 2044–2050.
  26. 26. Tschopp J, Irmler M, Thome M (1998) Inhibition of Fas death signals by FLIPs. Current Opinion in Immunology 10: 552–558.
  27. 27. Sedger LM, Glaccum MB, Schuh JCL, Kanaly ST, Williamson E, et al. (2002) Characterization of the in vivo function of TNF-alpha-related apoptosis-inducing ligand, TRAIL/Apo2L, using TRAIL/Apo2L gene-deficient mice. European Journal of Immunology 32: 2246–2254.
  28. 28. Yen ML, Tsai HF, Wu YY, Hwa HL, Lee BH, et al. (2008) TNF-related apoptosis-inducing ligand (TRAIL) induces osteoclast differentiation from monocyte/macrophage lineage precursor cells. Molecular Immunology 45: 2205–2213.
  29. 29. Hsu PN, Huang SC (2009) TNF-related apoptosis-inducing ligand (TRAIL) induces osteoclast differentiation from monocyte/macrophage lineage precursor cells via TRAF-6 dependent pathway. The Journal of Immunology 182: 14.
  30. 30. Lacey DL, Timms E, Tan HL, Kelley MJ, Dunstan CR, et al. (1998) Osteoprotegerin ligand is a cytokine that regulates osteoclast differentiation and activation. Cell 93: 165–176.
  31. 31. Emery JG, McDonnell P, Burke MB, Deen KC, Lyn S, et al. (1998) Osteoprotegerin is a receptor for the cytotoxic ligand TRAIL. The Journal of biological chemistry 273: 14363–14367.
  32. 32. Grimaud E, Soubigou L, Couillaud S, Coipeau P, Moreau A, et al. (2003) Receptor activator of nuclear factor kappaB ligand (RANKL)/osteoprotegerin (OPG) ratio is increased in severe osteolysis. Am J Pathol 163: 2021–2031.
  33. 33. Xu JK, Wu HF, Ang ESM, Yip K, Woloszyn M, et al. (2009) NF-kappa B modulators in osteolytic bone diseases. Cytokine & Growth Factor Reviews 20: 7–17.
  34. 34. Wakeyama H, Akiyama T, Takahashi K, Amano H, Kadono Y, et al. (2007) Negative feedback loop in the Bim-caspase-3 axis regulating apoptosis and activity of osteoclasts. Journal of bone and mineral research : the official journal of the American Society for Bone and Mineral Research 22: 1631–1639.
  35. 35. Hauer J, Pèuschner S, Ramakrishnan P, Simon U, Bongers M, et al. (2005) TNF receptor (TNFR)-associated factor (TRAF) 3 serves as an inhibitor of TRAF2/5-mediated activation of the noncanonical NF-kappaB pathway by TRAF-binding TNFRs. Proceedings of the National Academy of Sciences of the United States of America 102: 2874–2879.
  36. 36. Darnay BG, Ni J, Moore PA, Aggarwal BB (1999) Activation of NF-kappa B by RANK requires tumor necrosis factor receptor-associated factor (TRAF) 6 and NF-kappa B-inducing kinase - Identification of a novel TRAF6 interaction motif. Journal of Biological Chemistry 274: 7724–7731.
  37. 37. Mogi M, Ozeki N, Nakamura H, Togari A (2004) Dual roles for NF-kappa B activation in osteoblastic cells by serum deprivation: osteoblastic apoptosis and cell-cycle arrest. Bone 35: 507–516.
  38. 38. Thomas DC, Witte JS (2002) Point: population stratification: a problem for case-control studies of candidate-gene associations? Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology 11: 505–512.
  39. 39. Chanock SJ, Manolio T, Boehnke M, Boerwinkle E, Hunter DJ, et al. (2007) Replicating genotype-phenotype associations. Nature 447: 655–660.
  40. 40. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155: 945–959.