Strong association between polymorphisms in ANKH locus and skeletal size traits
- First Online:
- Cite this article as:
- Malkin, I., Ermakov, S., Kobyliansky, E. et al. Hum Genet (2006) 120: 42. doi:10.1007/s00439-006-0173-6
- 57 Views
Loss of bone strength is the main determinant of bone fragility. Bone strength is directly dependent on bone size (BS). A substantial portion of BS variation is attributable to genetic effects. However, the list of genes and allelic variants involved in the determination of BS variation is far from being complete. Polymorphisms in the ANKH gene have been shown to be associated with radiographic hand BS-related phenotypes. The present study examined the possible association of the ANKH gene with skeletal size and shape in order to test the universality of the ANKH effect on BS traits. Our sample consisted of a total of 212 ethnically homogeneous nuclear families (743 individuals) of European origin. Each individual was measured for body height, weight, and several other anthropometrical measurements, and genotyped for nine polymorphic markers (the average heterozygosity level was 0.4). We observed significant associations with practically all the anthropometrical phenotypes studied. More specifically, we found associations with body weight and height, limb length (P≤0.001; promoter region). After adjustment for body height, we demonstrated the substantial association increase for biacromial breadth (P=0.0012; some 140 kb downstream from ANKH) and vertebral column length (P=0.0008; exons 2–7 region). The majority of the observed associations persisted even after correction for multiple testing. For the first time the reliable evidence in support of universality of ANKH gene polymorphisms effect on bone size was provided.
Loss of bone strength is the main determinant of bone fragility (Hayes et al. 1996). Many studies have suggested that bone strength is directly dependent on bone size (BS) and bone geometry (BG) (McCreadie and Goldstein 2000; Gatti et al. 2001; Felsenberg and Boonen 2005). It has been shown that a substantial portion of BS and BG variation is attributable to genetic effects (Peacock et al. 2002; Deng et al. 2002; Livshits et al. 2003). However, the list of genes and allelic variants involved in the determination of BS and BG variation is far from being complete. Several genome-wide scans have been conducted to identify susceptibility loci for bone size (e.g., Koller et al. 2001; Deng et al. 2003b) and adult height variation (e.g., Mukhopadhyay et al. 2003; Willemsen et al. 2004). With a few exceptions the results of the linkage analyses are so far largely inconsistent.
Numerous factors acting in a complex and integrated network tightly control skeletal development and growth (e.g., see a review by Provot and Schipani 2005). Allelic variants of several genes were reported to be associated with BS and BG-related phenotypes, e.g., TNF (Moffett et al. 2005), COL1A1 (Qureshi et al. 2001; Long et al. 2004) and COL1A2 (Deng et al. 2003a; Lei et al. 2005), ESR1 (van Meurs et al. 2003; Schuit et al. 2004) and VDR (Minamitani et al. 1998; Lorentzon et al. 2000), IGF1 (Rivadeneira et al. 2004), CYP17A1 (Zmuda et al. 2001), and others.
Recently, we reported that the genetic variants in human orthologue of the mouse progressive ankylosis gene (ANKH) locus were strongly associated with bone quantitative traits assessed from plain hand radiographs (Malkin et al. 2005). More specifically, we performed a family-based association study with three STR and six SNP markers covering the entire ANKH locus and consistently found significant associations between several marker alleles and a number of hand bone traits such as the metacarpal cortical index, diameter/length ratio, and the bone-breaking resistance index.
The ANKH gene is one of the key genetic factors involved in the regulation of hydroxipatite deposition and bone mineralization (Harmey et al. 2004). It may therefore have a universal influence on the size and shape variation of different bones in the human skeleton. Correspondingly, the purpose of the present research was to further test the ANKH locus for a possible association with BS traits by using additional informative SNP markers to narrow down the linkage area, and by studying the phenotypes characterizing skeletal size and shape in order to test the universality of the ANKH effect on BS traits.
Materials and methods
In order to increase statistical power of the study, we added 43 new pedigrees (169 individuals) to our previous sample, in which the association of ANKH polymorphisms with radiographic hand phenotypes was tested (Malkin et al. 2005). The resultant new sample consisted of 373 males (the mean age was 45.9 years, ranging from 18 to 89 years) and 370 females (the mean age was 45.3 years, ranging from 17 to 90 years), which were recruited and enrolled randomly, i.e., regardless of the outcome of any of the measured variables. A total of 743 individuals formed 212 nuclear families (the mean number of offspring in a family was 1.9, ranging from 1 to 4 offspring). DNA samples were available for 705 individuals. All studied individuals belonged to the ethnic group of Chuvashes living in numerous small villages along the Volga river in the Chuvasha and Bashkortostan Autonomies, Russian Federation. Further details regarding the present sample can be found elsewhere (e.g., Livshits et al. 2002). All subjects who agreed to participate in the study signed an informed consent form, and the Tel Aviv University ethics committee approved the project.
Eight anthropometrical phenotypes were chosen for this study including body height (stature) and weight, vertebral column length (VCL), biacromial breadth (BIAC), humeral (HUML) and forearm (FARML) lengths, and femoral (FEML) and tibial (TIBL) lengths. All original phenotypes were assessed using a standard anthropometric technique (e.g., Lohman et al. 1988).
DNA was prepared from peripheral blood lymphocytes by standard techniques, using Nucleon™ BACC Genomic DNA Extraction Kits (Amersham International plc, UK) according to the manufacturer’s protocol. A total of 705 individuals were genotyped for nine SNP markers in ANKH gene: rs835141 (A), rs835154 (B), hCV3191922 (C), hCV11658675 (D), rs258360 (E), rs258215 (F), rs875525 (G), rs39968 (H), rs152628 (I). Information about SNP markers beginning with rs (A, B, and E–I) is available online at http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?CMD=search&DB=snp. Markers C and D were chosen using SNPbrowser™ Software v.2.0.8 and are referred to according to their ID numbers (for more details see: http://www.marketing.appliedbiosystems.com/mk/get/snpb_landing?isource=fr_E_RD_www_allsnps_com_snpbrowser#). Genotyping was performed using TaqMan® SNP Genotyping Assays (http://www.appliedbiosystems.com) on the ABI PRISM® 7900HT Sequence Detection System. The protocol of the genotyping procedure can be found at http://www.docs.appliedbiosystems.com/pebiodocs/04332856.pdf. The genotyping was carried out under the supervision of Dr. M. Korner at The Center for Genomic Technologies, The Alexander Silberman Institute of Life Sciences, Hebrew University of Jerusalem, Israel. Genotyped data were checked for Mendelian errors by means of the MAN-6 package for Windows (Malkin and Ginsburg 2003).
In order to validate the genotyping quality and the sample’s informativity, we tested the studied SNPs pairwise for the extent of linkage and linkage disequilibrium (LD) between them. The full parameter marker–marker maximum likelihood linkage procedure was carried out for each pair of markers to estimate the respective LD parameter D′ proposed by Lewontin (1974), using the MAN-6 package for Windows (Malkin and Ginsburg 2003). Additionally, we used Haploview 3.2 Software (http://www./broad/mit/edu/mpg/haploview/) to calculate the same D′ estimates for all possible pairwise combinations of the studied SNP markers. The data presented at http://www.hapmap.org/cgi-perl/gbrowse/gbrowse/ was used as reference for ANKH haplotype structure. With MAN-6 software we generated haplotypes for adjacent SNP pairs. Then we considered each pair of markers as four-allelic locus with each allele corresponding to the respective haplotype variant.
Statistical and genetic analyses
Preliminary statistical analyses were conducted using the STATISTICA 5.5 package for Windows (Statsoft Inc., USA). First, descriptive statistics of each of the original traits for each sex were obtained separately. To reduce the number of the inter-correlated quantitative phenotypes, we carried out a principle component analysis for limb bone length (LBL) in order to construct a synthetic index as the first factor score of HUML, FARML, FEML, and TIBL. In further genetic analyses, we examined stature, weight, horizontal (BIAC) and vertical (VCL) body sizes, and the synthetic index LBL mentioned above. The studied phenotypes were adjusted for age effects according to sex, implementing the interval piecewise aging models as described in Malkin et al. (2002b). The most parsimonious age-dependence model for each trait was selected, using a maximum likelihood ratio test as a model-fitting technique. After adjustment for age, we estimated familial correlations (rsp for spouses, rpo for parent–offspring, and rsib for siblings) and calculated the heritability for each trait as h2=(rsib+rpo)(1+rsp)/(1+rsp+2rsprpo) (Rice et al. 1997). In the next stage we conducted the joint quantitative trait—DNA marker association analysis by means of transmission disequilibrium tests (TDTs). Specifically, we used four tests: (1) the family-based association test proposed by Horvath et al. (2001) and implemented in the FBAT program; (2) the extreme offspring design t test (EOT) proposed by Malkin et al. (2002a) and implemented in the MAN-6 package; (3) and (4) two different versions of the orthogonal test proposed by Abecasis et al. (2000), with adjustment for the parent phenotypes (OTP) and without it (OT). Both tests are included in the QTDT program. We recently discussed the statistical features of these tests elsewhere (Malkin et al. 2005).
To combine the results of the four separate association tests, we formulated a multiple comparison procedure (MCP) and computed the combined P-values, the probability of erroneous rejection of the general null hypothesis of no linkage disequilibrium, which unites all certain null hypotheses. MCP was constructed based on the joint simulated null-distribution for four separate tests. We generated 20,000 simulation replicates using the pedigree structure of our sample. The trait pattern of inheritance was simulated using the inheritance model exhibiting the observed familial correlations, but assuming that the tested marker had no effect on the trait variation. Further details are given in the appendix.
The array of computed MCP P-values was further tested for significance using the false discovery rate (FDR) approach proposed by Benjamini and Yekutieli (2001) for multiple testing under dependency. This approach is used when there are a number of different null-hypotheses tested, some of which are probably true and some are false. The total number of tested null-hypotheses was computed as follows. Each trait-SNP pair produced only one separate null-hypothesis. The marker haplotype produced from two adjacent diallelic SNPs generally has four alleles, except situations when this number is reduced to three, as a consequence of strong linkage disequilibrium. For these cases the number of tested dichotomous schemes employed was three: each allele against two others. For the four allelic case we performed five dichotomous schemes: four 1-to-3 factorizations and additionally, one 2-to-2 factorization, which cannot be reduced to the testing of initial SNPs. Thus, for each haplotype marker the number of null-hypotheses corresponds to the number of different dichotomous factorizations employed for a trait tested (3 or 5).
To estimate the association between marker alleles and phenotypes in the whole sample including parent phenotypes and families not informative for TDT, we also applied the pedigree disequilibrium test (PDT). The PDT estimates the genetic-environmental trait inheritance model as described by Ginsburg and Livshits (1999) under the assumption that the gene included in the model is the marker locus itself. The likelihood ratio test was used to reject the null hypothesis that all marker genotypes exhibit the same mean trait value. The application of this test was recently reported elsewhere (Suk et al. 2005).
Because most of the investigated traits showed a similar linkage disequilibrium pattern in the chromosome region close to the gene promoter, we tried to separate the shared factor of four of them exhibiting the most significant P-values (stature, weight, VCL, and BIAC). The association of the first factor score of these age-adjusted variables (SIZE_PC) with the DNA markers was also examined by TDTs. Finally, to test for an association of the anthropometrical traits with the selected DNA polymorphisms, independent of body height, each of these traits was adjusted for stature.
Descriptive statistics and analyses of familial correlations
Descriptive statistics of the studied sample, summaries of multiple regression analysis of anthropometric bone traits, and heritability estimates of the traits adjusted for sex and age
Multiple R2 (sex, age)
Vertebral column length (mm)
Biacromial breadth (mm)
Humeral length (mm)
Forearm length (mm)
Femoral length (mm)
Tibial length (mm)
Family-based association study with DNA markers
Pedigree-based disequilibrium test (P-values)
Weight adjusted for stature
BIAC adjusted for stature
VCL adjusted for stature
To correct our TDT results for multiple testing, we allowed for the false discovery rate level 0.05. The number of different null-hypotheses for each trait was 43 (9 SNPs and 34 dichotomous factorizations for 8 haplotypes, see Materials and methods). Thus, the total number of tests for 10 analyzed traits was 430. Ranging MCP P-values for trait-marker pairs (Benjamini and Yekutieli 2001), we found that with FDR=0.05 we can reject 17 general null-hypotheses; the rejection threshold was P<0.002. The corresponding P-values are highlighted with gray in Tables 3 and 4.
The present study provides compelling and unambiguous evidence of a significant association between ANKH polymorphisms and normal variability of all studied original and synthetic skeletal size-related phenotypes. In agreement with Malkin et al.’s (2005) findings, the significant results were obtained for markers and their haplotypes located upstream from the first exon, in the region of the second through the seventh exons, and for the haplotype downstream from the ANKH gene. Thus, variation in haplotypes of the A_B SNP pair, which is probably in LD with the promoter and the first exon region, was significantly associated with all anthropometrical traits studied. The association was significant regardless of the particular type of TDT, and partially persisted even after FDR correction. For example, the association of the aforementioned haplotype marker A_B with the complex trait SIZE_PC, which may be viewed as reflecting the overall skeletal development, was significant at P<0.0001 and still remained significant after FDR correction. Also of particular interest are the associations between this marker and those adjusted for stature LBL, VCL, and weight (P≤0.0007), all of which remained significant after FDR correction. Polymorphisms in the “central” ANKH region showed significant results with several phenotypes; the most significant association was observed for E_F with VCL adjusted for stature (P=0.0008). The region downstream from the ANKH gene, represented by the H_I haplotype marker, showed specific significant association with BIAC (P=0.0012). The evidence of an association between ANKH polymorphisms and anthropometric traits provided by means of TDTs is further strengthened by the results of the PDT analysis.
The estimated effect on the trait variance of markers showing significant PDT is about 2%. However, we do not believe that discovered LD sites in ANKH are themselves the functional variants. Consequently, the effect of putative QTL, accounting for non-perfect disequilibrium and allele frequency differences, may be much greater at least by factor 1/(D′)2 (here D′ is the assumed LD level between the marker and QTL; the distribution of D′ should be similar to that presented in Fig. 2). In summarizing the results of the present study and those reported by Malkin et al. (2005), we concluded that there are at least two putative quantitative trait loci (QTLs) in the ANKH gene: the first one located in the promoter region or the first exon region, and the second QTL being somewhere in the area of the second through the seventh exon. Both QTLs independently contribute to the normal variation of bone size and shape-related traits. Interestingly, marker H, located some 140 kb downstream from the ANKH gene, was also significantly associated with weight and BIAC as a single SNP and as part of the H_I haplotype marker.
There may be different explanations for this observation. For example, it is possible that despite its relatively distant location from the AHKH gene, marker H is actually in LD with some functional polymorphism in that gene. This hypothesis can be supported by the results of LD decay analysis, which revealed substantial LD between H and G, which is near exon 6 of ANKH. However, marker H may be in LD with polymorphism in some yet unknown regulatory element, located downstream from ANKH and may affect the transcription levels of that gene. Finally, it may turn out that this marker points to another QTL, possibly belonging either to hypothetical FLJ11127 or to the triple functional domain (PTPRF interacting) (TRIO) (NCBI Map Viewer; http://www.ncbi.nlm.nih.gov/mapview/). Although it is difficult to speculate on the functional relationship between the latter polymorphisms and bone size-related phenotypes at this stage, the link between ANKH polymorphisms and the studied anthropometric traits seems obvious. ANKH, encoding a multiple-pass transmembrane protein, is known to be a key player in the regulation of extracellular levels of PPi (Harmey et al. 2004). On the other hand, it is well established that the tight balance between PPi and Pi extracellular levels are crucial for the proper mineralization of growth plate cartilage, which is one of the central processes during endochondral bone formation (Wang et al. 2005). Moreover, Pi induces apoptosis in chondrocytes and osteoblasts, and regulates the expression of a variety of genes (Meleti et al. 2000; Mansfield et al. 2001; Beck et al. 2003). Thus, quantitative and qualitative changes in ANKH transcription that are due to variations at the putative QTLs are likely to affect the relative extracellular levels of PPi and Pi, and consequently, to influence the parameters of mineralization, as well as chondrocytes and osteoblast physiology. Certainly, further studies are needed for precise identification of the QTLs implied by our and Malkin et al.’s (2005) findings, and for a full functional explanation of the observed effects.
In conclusion, the present study provides reliable evidence supporting the association between polymorphisms in the ANKH locus and normal variability of skeletal size and shape-related phenotypes. The majority of the observed associations were significant regardless of the particular analysis implemented, and persisted even after multiple testing corrections. The studied polymorphisms explained about 2% of the variance of the chosen anthropometric traits. Our findings contribute to a better understanding of the ANKH involvement in bone development and normal skeletal size and shape variation. Further studies are required to confirm our findings and to provide a functional biological explanation of our results.
This study was performed in partial fulfillment of the doctoral degree requirements of Sergey Ermakov. We wish to thank Dr. Svetlana Trofimov (Department of Anatomy and Anthropology, Sackler Faculty of Medicine, Tel Aviv University) for help in DNA preparation, Dr. Mira Korner and her staff (The Center for Genomic Technologies, The Institute of Life Sciences, The Hebrew University of Jerusalem) for the genotyping of the samples, and Galit Schwartz (Applied Biosystems, Agentek-Israel) for her assistance. This study was supported by the Israel National Science Foundation (Grant No. 1042/04).