Synopsis and meta-analysis of genetic association studies in osteoporosis for the focal adhesion family genes: the CUMAGAS-OSTEOporosis information system

Osteoporosis is a common skeletal disease characterized by generalized reduction in bone mineral density and microarchitectural deterioration of bone tissue, leading to impaired skeletal strength and increased susceptibility to fracture [1]. Genetic factors have long been recognized as playing an important role in osteoporosis [2]. Major efforts are currently underway to identify the specific genes and allelic variants predisposing patients to this disease. The identification of genes is achieved by conducting hypothesis-free, genome-wide association studies (GWASs) and candidate gene association studies (GASs) [3]. Candidate genes are typically chosen on the basis of having biological effects on bone metabolism or bone cell activity and whether they contribute to the risk of osteoporosis [3].

The focal adhesion (FA) gene family has emerged as a logical candidate for osteoporosis. Focal adhesions are specialized structures at the cellular-extracellular matrix contact points, where bundles of actin filaments are anchored to transmembrane receptors of the integrin family through a multimolecular complex of junctional plaque proteins. Some of the constituents of FA genes participate in the structural link between membrane receptors and the actin cytoskeleton, while others are signaling molecules [4, 5]. Although there are a number of genes constituting the pathway, only a small number of variants of these genes have been studied in GASs of osteoporosis. More recently, haplotype-based approaches and genome-wide genotyping platforms have enabled more comprehensive capture of genetic variation in these genes [6]. The most studied gene is the collagen, type I, α₁ (COL1A1) gene, especially its variant G2046T. Other genes studied in the FA family include the genes insulin-like growth factor I (IGF-I), integrin β-chain β₃ (ITGB3), α-actinin-3 (ACTN3), COL1A2 and type 1 insulin-like growth factor receptor (IGF-IR). However, the results of the GASs involving genes of the FA pathway and osteoporosis are controversial and inconclusive, possibly because of methodological limitations, including inadequate sample size, patient selection, ethnicity of the populations studied and lack of adjustments for confounders [7].

To explore the involvement of FA family gene polymorphisms in osteoporosis susceptibility, we systematically searched for all available GASs of FA family genes and osteoporosis (as a binary phenotype) and created the Cumulative Meta-Analysis of Genetic Association Studies-OSTEOporosis (CUMAGAS-OSTEOporosis) information system. Then we catalogued all retrieved articles and estimated the risk effects of all individually investigated variants. Finally, the available data were synthesized using meta-analysis techniques to increase the power for detecting significant results and to decrease the uncertainty of the estimated genetic risks [8].

In this project, the currently available data from GASs on human FA family genes in osteoporosis were catalogued. Then the data were synthesized, and the involvement of FA gene variants in disease susceptibility was assessed comprehensively. The eligible GASs were catalogued in a publicly available web-based database and information system called CUMAGAS-OSTEOporosis [9]. In total, eight variants were meta-analyzed. Significant results were shown for the variants COL1A1 G2046T, COL1A1 G-1997T and ITGB3 T176C. Type I collagen is the major protein constituent of bone and is therefore a strong and plausible candidate gene for osteoporosis. The COL1A1 G2046T polymorphism is a single base pair substitution (G → T) within the regulatory region of the COL1A1 gene [30]. The COL1A1 G-1997T polymorphism has been identified in the proximal promoter of COL1A1 at position -1997 and is in linkage disequilibrium with the COL1A1 G2046T polymorphism [31]. There is evidence that the promoter polymorphisms are functional and have effects on DNA binding and gene transcription [32], but it is unclear to what extent this polymorphism is associated with the biomechanical properties of bone or susceptibility to fracture. Taking into consideration the abundance of COL1A1 in the bones and the fact that these variants seem to play a major role in the function of the gene, the significant association found in our results may encourage intensive research in this area. The ITGB3 T176C polymorphism changes the conformational structure of the β₃-subunit of integrin [33]. The integrin β₃-subunit is known to play a key role in the resorptive function of osteoclasts, as shown in vitro and in transgenic animal studies [34]. Genetic variation in integrin β₃ may influence bone remodeling and subsequent bone loss and risk of osteoporotic fractures.

CUMAGAS-OSTEOporosis is an evidence-based information system for systematically searching, reviewing and synthesizing data for GASs of osteoporosis, with the capacity for continuous updating. CUMAGAS is being expanded to additional complex diseases such chronic lymphocytic leukemia, peripheral arterial disease, hypertension and osteoarthritis [10, 9, 35, 36]. CUMAGAS also has the capacity to incorporate data from GWASs subject to their public availability.

The GWASs of osteoporosis have not highlighted a significant role for the FA family of genes. However, the commercial genotyping platforms [15] may underrepresent the variants of the FA pathway and, of course, the variants included in the meta-analysis. In addition, the variants identified to date from the GWAS approach explain only a fraction of the disease heritability, and therefore the potential role for the variants of FA pathway may not be excluded [36]. Furthermore, the analyses of GWASs have missed associations of multilocus variants involved in pathways with pathophysiological relevance to disease mechanisms [3, 36].

Most of the published GASs are underpowered for detecting the minor contribution of common alleles. For example, a sample size of more than 10,000 patients is needed to achieve >80% power to detect a significant OR between 1.1 and 1.5 (modest effect) for a polymorphic locus in association with a complex disease [37]. Meta-analysis is a tool that allows for analysis with the potential for higher power by pooling the results of multiple studies [8]. However, there is no formal, established methodology for calculating the power of a meta-analysis. In addition, power analysis may not be applicable in meta-analysis, since it is a retrospective, all-inclusive synthesis of published studies [8, 38]. Nevertheless, type II errors are expected to be less common in a meta-analysis than in single studies [8, 39]. Currently, no single institution alone is able to provide a sufficient number of patients, and therefore the creation of large databases from consortia where researchers share their data are required. However, this need for data sharing has been pointed out by previous initiatives [40, 41].

Two problems in human genome epidemiology research are that negative studies are frequently unpublished and some studies do not provide extractable data, which lead to the well-known phenomenon of publication bias [42]. However, negative results should also have a venue for publication. Moreover, the inclusion of "negative" and unpublished data in meta-analyses of GASs as a means of reducing publication bias is commonly suggested and is believed to help in pointing out genetic effects [43]. Thus, the establishment of an electronic information system to aid in performing cumulative meta-analyses of (published and unpublished) GASs of osteoporosis and identifying significant genetic variants could be a valuable tool for ongoing research in the field. Furthermore, CUMAGAS-OSTEOporosis will support rapid progress in human genome epidemiology of osteoporosis by identifying valid and replicable associations and making the overall effect of each variant from published and unpublished studies rapidly available to researchers. Concerning the retrieval of unpublished studies, the authors of unpublished studies will be able to submit their data to the CUMAGAS-OSTEOporosis database. Since these data will not have undergone peer review, a sensitivity analysis may be carried out (that is, a meta-analysis that examines the effect of excluding these studies). Finally, CUMAGAS-OSTEOporosis is an open access system, and it may support efforts to prevent publication bias [44].

Publication bias was tested using the method proposed by Harbord et al. [21], which is a modification of Egger's test [45], and it is appropriate for small-study effects. The visual inspection of funnel plots was avoided, since their validity is questionable [46, 47]. However, the statistical tests used to evaluate studies for publication bias actually compare the differential magnitude of effects in large versus small studies [8, 48].

The significance of risk effects in the GASs was assessed using the OR metric for various genetic models (dominant, recessive, additive and codominant) by merging genotypes. These models are not independent, and there is no a priori biological justification for their choice. Consequently, the interpretation of the results can be problematic, especially when all genetic contrasts are significant as in the case of COL1A1 G2046T. In these cases, the introduction of the recently proposed generalized OR (OR_G) as an overall genetic risk effect may be a remedy [42]. The OR_G is a single statistic that summarizes the magnitude and significance of the association without considering the hash of possible contrasts, and thus the interpretation of the results is straightforward [49]. The meta-analysis methodology in CUMAGAS-OSTEOporosis is expanding to incorporate the OR_G metric and the continuous phenotype of osteoporosis (that is, bone mineral density).

In the meta-analysis, various genetic contrasts in different populations were explored, since there is no biological justification for choosing a specific contrast [49]. Thus, adjustment for multiple testing is not strictly required for such an exploratory study [8, 50]. In addition, the adjustment for multiple testing might not be necessary, since the data were synthesized with the objective of reducing the uncertainty of effect size, without a prespecified hypothesis [50‐52]. Furthermore, an appropriate multiple test adjustment might be difficult because the investigated contrasts are not independent and there is no clear structure in the multiple tests [49, 52]. Finally, an adjustment for multiple comparisons (for example, Bonferroni's correction) concerns a general null hypothesis that there is no association in all genetic contrasts simultaneously, which is not likely [53, 54].

The phenotypic heterogeneity of osteoporosis in the included studies in our synopsis makes the effort to combine the findings of GASs meaningfully in the complex field of osteoporosis a difficult task. Thus, in the presence of the large heterogeneity, the results should be interpreted with caution. There may be fundamental differences in the nature of genetic susceptibility to osteoporosis between postmenopausal and premenopausal women and between the two sexes or even among patients of different ethnicities. Furthermore, environmental factors, particularly nutrients, have to be accurately evaluated together with complex genotyping, to weigh their importance in revealing functional variants with respect to specific genetic background [55, 56]. Our analysis used the available study-level allele and genotype distributions, precluding adjusted analysis for potential gene-gene and gene-environment interactions, for which raw genotype data would be required. Failure to account for interactions may have reduced the power of our analysis but is unlikely to have inflated the number of positive results.

There is evidence implicating the activity of the FA family of genes in osteoporosis. Future studies designed to investigate epistatic and gene-environment interactions may help in deriving more conclusive claims about the role of these genes in osteoporosis. The CUMAGAS-OSTEOporosis information system can be a useful resource for reviewing and interpreting the findings of the accumulating genomic epidemiologic research in osteoporosis.