- Email: [email protected]
Multiple osteoporosis susceptibility genes on chromosome 1p36 in Chinese
Bone 44 (2009) 984–988
Contents lists available at ScienceDirect
Bone j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / b o n e
Multiple osteoporosis susceptibility genes on chromosome 1p36 in Chinese Qing-Yang Huang a,b,⁎, Gloria H.Y. Li a, Annie W.C. Kung a,b,⁎ a b
Department of Medicine, The University of Hong Kong, Hong Kong The Center of Heart, Brain, Hormone and Health Ageing, The University of Hong Kong, Hong Kong
a r t i c l e
i n f o
Article history: Received 29 January 2008 Revised 14 January 2009 Accepted 15 January 2009 Available online 30 January 2009 Edited by: B. Olsen Keywords: Osteoporosis BMD Association Candidate gene Interaction
a b s t r a c t Introduction: Chromosome 1p36 is a region that has previously shown good evidence of linkage to bone mineral density (BMD) in multiple studies, but the genes that are responsible for the linkage signals are unknown. Materials and methods: We performed a gene-wide and tag SNP-based association study of four positional and functional candidate genes (TNFRSF1B, PLOD, CNR2, and MTHFR) at 1p36 in 1, 243 case–control Chinese subjects. Twenty-three tag SNPs were selected and genotyped using the high-throughput Sequenom genotyping platform. Binary logistic regression analyses were performed to test for genotype associations between each SNP and BMD. Allelic and haplotype association analyses were conducted by Haploview. Gene– gene interactions were investigated using multifactor dimensionality reduction method. Results: The PLOD rs7529452 (C385T; F98F) and MTHFR rs1801133 (C677T; A429E) showed significant genotypic/allelic associations with BMDs at all sites measured (P = 0.08–0.001), and a promising two-locus gene–gene interaction for femoral neck BMD. The CNR2 rs2501431 (A592G; G155G) showed nominally significant allelic associations with trochanter and hip BMD. The TNFRSF1B rs976881 showed genotypic associations with BMDs (P = 0.08–0.04). Conclusions: Our results suggest that multiple genes at 1p36, individually or in different combinations, contribute to osteoporosis susceptibility in Chinese. © 2009 Elsevier Inc. All rights reserved.
Introduction Osteoporosis is a major public health problem worldwide and is highly heritable: there is an increased rate of concordance in MZ versus DZ twins and a substantially increased incidence in individuals with a positive family history. More than 20 genome-wide linkage scans have been launched to search for quantitative trait loci that underlie BMD variation. Several genomic regions, such as 1p36, 1q21–25, 2p22–24, 3p14–25, 4q25–34, 6p21, 7p14–21, 11q14–25, 12q23–24, 13q14–34 and 20p12, have been well validated [1]. Evidence for replication of linkage across populations seems to be strongest for a quantitative trait locus on chromosome 1p36. Devoto et al. [2] reported a genome-wide scan in 149 members of seven large pedigrees with the strongest evidence of linkage for hip BMD on chromosome 1p36 (LOD = 3.51). This finding was confirmed and extended in an expanded sample of 42 families [3]. Wilson et al. [4] performed a genome-wide screen of 1097 unselected female UK twin pairs, and observed suggestive linkage for whole-body BMD on chromosome 1 at 17 cM (LOD= 2.38). Xiao et al. [5] recently found suggestive evidence of linkage to wrist BMD on 1p36 in men (LOD = 2.81). Streeten et al. [6] reported suggestive linkage of total hip BMD to a quantitative trait locus on chromosome 1p36 at D1S1597 in ⁎ Corresponding authors. Fax: +852 28162187. E-mail addresses: [email protected] (Q.-Y. Huang), [email protected] (A.W.C. Kung). 8756-3282/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.bone.2009.01.368
women (LOD= 2.02). A recent genome-wide linkage scan in a single extended family also revealed a maximum LOD score of 3.07 on 1p36.3 (D1S468) for spine BMD [7]. Bin 1.1 (1pter-1p36.22) lies above the 95% confidence level (P =0.05) in a meta-analysis of published data from whole genome scans [8]. Obvious functional candidate genes situated within this region include tumor necrosis factor receptor superfamily member 1B (TNFRSF1B, 1p36.3-p36.2), lysyl hydroxylase (PLOD, 1p36.3p36.2), cannabinoid receptor 2 (CNR2, 1p36.11), and methylenetetrahydrofolate reductase (MTHFR, 1p36.3). To identify the quantitative trait locus genes responsible for the linkage signals found at chromosome 1p36, we performed a gene-wide and tag SNP-based association study of four positional candidate genes (TNFRSF1B, PLOD, CNR, and MTHFR) in a sample of 1243 cases and matched controls. Associations with SNPs within multiple genes and gene–gene interaction were observed. Materials and methods Subjects Samples consisted of 1243 case–control Chinese subjects, with 909 spine case–control subjects, 792 femoral neck case–control subjects, 706 trochanter case–control subjects and 760 total hip case–control subjects (Table 1). Cases were arbitrarily defined as subjects with low BMD (Z score ≤ −1.28, equivalent to the lowest 10th percentile of the population). Controls were age and sex-matched subjects with high
Q.-Y. Huang et al. / Bone 44 (2009) 984–988
985
Table 1 Characteristics of 1243 case–control subjects Spine Subject number Postmenopausal women Premenopausal women Men Age (years) Height (m) Weight (kg) BMD (g/cm2) Spine Femoral neck Trochanter Total hip Z-score Spine Femoral neck Trochanter Total hip
Femoral neck
Trochanter
Total hip
Case
Control
Case
Control
Case
Control
Case
Control
476 270 137 69 53.3 ± 14.4 1.54 ± 0.08 49.7 ± 8.2
433 214 153 66 53.7 ± 14.6 1.58 ± 0.07 63.2 ± 10.7
427 235 160 32 51.4 ± 14.9 1.53 ± 0.08 48.6 ± 7.7
365 166 162 37 51.0 ± 14.0 1.58 ± 0.07 64.1 ± 10.9
369 216 117 36 52.8 ± 15.5 1.53 ± 0.08 48.5 ± 7.8
337 142 154 41 50.8 ± 14.3 1.58 ± 0.07 64.8 ± 10.7
420 241 144 35 52.5 ± 15.1 1.53 ± 0.08 48.3 ± 7.3
340 148 152 40 51.0 ± 14.2 1.58 ± 0.07 64.5 ± 10.7
0.67 ± 0.10 0.56 ± 0.09 0.47 ± 0.08 0.64 ± 0.10
1.13 ± 0.11 0.80 ± 0.13 0.71 ± 0.11 0.92 ± 0.13
0.72 ± 0.14 0.52 ± 0.08 0.46 ± 0.08 0.61 ± 0.10
1.07 ± 0.13 0.89 ± 0.10 0.75 ± 0.10 0.97 ± 0.11
0.69 ± 0.13 0.52 ± 0.09 0.43 ± 0.07 0.59 ± 0.09
1.11 ± 0.12 0.87 ± 0.11 0.77 ± 0.09 0.99 ± 0.10
0.70 ± 0.13 0.53 ± 0.08 0.45 ± 0.08 0.59 ± 0.09
1.10 ± 0.12 0.88 ± 0.10 0.76 ± 0.09 1.00 ± 0.10
− 1.99 ± 0.48 − 1.42 ± 0.70 − 1.44 ± 0.72 − 1.61 ± 0.78
1.73 ± 0.63 0.87 ± 0.91 1.03 ± 0.90 1.01 ± 0.92
− 1.62 ± 0.79 − 1.80 ± 0.41 − 1.63 ± 0.64 − 1.89 ± 0.63
1.15 ± 0.94 1.60 ± 0.52 1.38 ± 0.83 1.49 ± 0.78
− 1.80 ± 0.78 − 1.72 ± 0.60 − 1.87 ± 0.45 − 2.01 ± 0.57
1.40 ± 0.93 1.40 ± 0.79 1.66 ± 0.61 1.59 ± 0.72
− 1.76 ± 0.74 − 1.72 ± 0.53 − 1.73 ± 0.57 − 2.00 ± 0.52
1.35 ± 0.92 1.49 ± 0.67 1.56 ± 0.69 1.70 ± 0.58
Data are expressed as mean ± SD.
BMD (Z score N +1) at the corresponding sites. For a common SNP (allele frequency = 0.2), this association study had N90% power to detect a 40% difference in allele frequencies between the case and control groups at the 5% level of significance [9]. All participants gave informed consent and the study was approved by the Ethics Committee of the University of Hong Kong and conducted according to the Declaration of Helsinki. Tag SNP selection and genotyping We selected SNPs on the basis of the following principal criteria: (1) tag SNPs (tSNP) identified using genotype data from the CHB panel (Han Chinese in Beijing) of the phase II HapMap Project. The criterion for tagging was set at r2 N 0.8 and minor allele frequency N0.2. (2) functional relevance and importance. (3) SNPs significantly associated with BMD in previous studies. A total of 23 SNPs of the TNFRSF1B, PLOD, CNR2, and MTHFR genes (9 in the TNFRSF1B, 5 in the PLOD, 3 in the MTHFR, and 6 in the CNR2) were selected (Table 2). The 23 tag SNPs captured 83% of common SNPs (minor allele frequency N0.2) in HapMap Chinese database at r2 N 0.8 (78% in the TNFRSF1B, 70% in the PLOD, 100% in the MTHFR, and 100% in the CNR2). SNPs were genotyped using the high-throughput Sequenom genotyping platform. DNA from case and control subjects was randomly assigned to the 96 well plates, and genotyping was performed blind to the status of the samples. Genotyping was repeated in 5% of samples for verification and quality control. Quality control testing revealed that genotype data had an error rate b 0.1%.
our association as a replication of previous studies. In this situation, a P value b 0.05 was considered statistically significant. The multifactor dimensionality reduction (MDR; http://www. epistasis.org) method was used to test for potential gene–gene interactions [10]. MDR first pooled multi-locus genotypes with high dimension into only one dimension. Cross-validation consistency and testing accuracy were calculated for each combination of a pool of genetic polymorphisms. The final best model was selected as the one with maximal cross-validation consistency and minimal prediction error. Statistical significance was assessed by comparing the average testing accuracy from the observed data with the distribution of average testing accuracy under the null hypothesis of no associations derived empirically from 1000 permutations. Two-locus interactions among the SNPs were tested using MDR. Conditional logistic regression was used to validate the reality of the interaction. If an interaction term is not significant (P N 0.05) by logistic regression, the corresponding significant MDR interaction model could be simply caused by the additive main effects from the component loci. Table 2 Tag SNP information Gene
TNFRSF1B
Statistical analyses The genotyping quality of each SNP was first checked for the call rate, minor allele frequency and Hardy–Weinberg equilibrium. SNPs that failed quality control checks were excluded from further consideration. The program Haploview (http://www.broad.mit.edu/ mpg/haploview/) was used to calculate allele frequencies, to verify that the genotype data were in Hardy–Weinberg equilibrium, to calculate pair-wise linkage disequilibrium statistics, to plot haplotype block structure, and to test allelic and haplotype associations. Haplotypes were estimated using an accelerated EM algorithm in the program Haploview. Binary logistic regression analyses were performed to test for associations between SNP genotype and BMD with adjustment for sex, height, weight and the menopause status. Three possible genetic models (additive, dominant and recessive) were tested for each SNP. All statistical analyses were performed using SPSS version 13.0 for Windows (SPSS Inc., Chicago, IL, USA). As our study is not the first report, we mainly evaluated the significance of
PLOD
MTHFR
CNR2
SNP ID
rs496888 rs976881 rs1201157 rs683240 rs653667 rs5746057 rs5746059 rs1061624 rs1061628 rs1208984 rs7529452 rs7551175 rs7514577 rs2273291 rs9651118 rs1801133 rs1801131 rs4649119 rs2229581 rs2229579 rs4649124 rs2501431 rs7530595
Genomic position (bp)
Genic position
Alleles (major/ minor)
MAFa
12167072 12168020 12183297 12184095 12186074 12196956 12197058 12201531 12202265 11929442 11944221 11944222 11958501 11959619 11796480 11790644 11788742 23916849 23946400 23946468 23946663 23946949 24026412
Intron 1 Intron 1 Intron 2 Intron 2 Intron 3 Intron 9 Intron 9 Exon 10 Exon 10 Intron 1 Exon 3 Exon 3 Exon 12 Intron13 Intron 2 Exon 5 Exon 8 3′ Exon 2 Exon 2 Exon 2 Exon 2 5′
A/G G/A C/T T/C A/C A/C A/G A/G C/T A/G C/T A/G C/T C/T T/C C/T A/C G/C C/G C/T A/G G/A C/T
0.18 0.19 0.27 0.15 0.34 0.01 0.15 0.15 0.31 0.22 0.39 0.48 0.40 0.28 0.46 0.24 0.23 0.45 0.45 0.22 0.46 0.46 0.33
P - values b 0.01 are highlighted in bold type. a MAF: minor allele frequency. b HWE: Hardy–Weinberg equilibrium.
HWEb P
Call rate
0.26 0.76 0.11 0.18 0.00 0.00 0.92 0.00 1.00 0.10 0.51 0.00 0.14 0.05 0.80 0.06 0.69 0.16 0.00 0.38 0.04 0.01 0.51
0.97 0.98 0.94 0.99 0.42 0.96 0.94 0.52 0.94 0.88 0.90 0.81 0.98 0.92 0.96 0.97 0.96 0.96 0.95 0.99 0.96 0.94 0.98
986
Q.-Y. Huang et al. / Bone 44 (2009) 984–988
Results The clinical characteristics of the study subjects are summarized in Table 1. Among the 23 HapMap tSNPs selected, genotyping call rate N90%, minor allele frequency N5%, duplicate error rate b 2% and Hardy–Weinberg equilibrium P values N 1% were achieved in 18 tSNPs. Five tSNPs failed quality control checks, with significant discrepancy with Hardy–Weinberg equilibrium, and were excluded from further analyses. The final coverage of the genes with 18 out of 23 original tag SNPs was 65%. Details for all tSNPs (including genomic and genic position, minor allele frequency, Hardy–Weinberg equilibrium, call rate) are shown in Table 2. Of these, 10 tSNPs were mapped within introns, and 11 were in exons, and 2 were in regulatory regions. Single marker analysis Results of the single marker association test are shown in Table 3. For the PLOD gene, SNP rs7529452 (C385T; F98F) showed consistent genotypic/allelic association with BMD at all sites measured
(P = 0.08–0.001). The SNPs rs2273291 showed significant allele association with spine BMD (P = 0.025), and rs7514577 showed significant genotypic associations with femoral neck BMD and trochanter BMD (P = 0.02). For the MTHFR gene, the SNP rs1801133 (C677T; A429E) showed nominally significant allelic associations with BMD at all sites measured (P = 0.02–0.05) and a weak genotype association trend. For the CNR2 gene, the SNP rs2501431 (A592G; G155G) showed nominally significant allelic associations with trochanter BMD and hip BMD (P = 0.02 and 0.04, respectively). For the TNFRSF1B gene, rs976881 showed genotypic associations with BMD (P = 0.04–0.08). For spine BMD, the genotype frequencies for CC, TC and TT of PLOD rs7529452 were 36.9%, 46.5%, 16.7% in case samples, and 40.8%, 49%, 10.2% in control samples. Frequencies of the TT homozygotes were significantly higher among case than control subjects (P = 0.008; OR = 1.6; 95% CI 1.14–2.35). For femoral neck BMD, the genotype frequencies for CC, TC and TT of MTHFR rs1801133 were 61.3%, 34.4%, 4.4% in case samples, and 52.2%, 42.4%, 5.3% in control samples. Individuals with CC genotype showed a significantly higher risk of
Table 3 Results of single marker association tests
Three possible genetic models (additive, dominant and recessive) were tested and the lowest P values with adjustment for sex, height, weight and the menopause status were shown for each genotype association. P - values V 0.05 are highlighted in bold type.
Q.-Y. Huang et al. / Bone 44 (2009) 984–988 Table 4 Best two-locus gene–gene effect models identified by MDR method
Locus combination
Femoral neck BMD Spine BMD Hip BMD
Trochanter BMD
rs7529452 rs1801133 0.0013 20/20
P value Cross-validation consistency Testing accuracy (%) 59.18
rs2273291 rs1208984 0.4119 15/20
rs2273291 rs7529452 0.0207 15/20
rs1201157 rs1801133 0.8684 10/20
50.23
53.90
49.92
P - values b 0.05 are highlighted in bold type.
osteoporosis compared with those with TC and TT genotypes (P = 0.012; OR = 1.17; 95% CI 1.03–1.33). The T allele of PLOD rs7529452 was associated with low femoral neck BMD (P = 0.04). For trochanter BMD, genotype associations for the PLOD rs7529452 and rs7514577, TNFRSF1B rs976881, and CNR2 rs7530595 and allelic associations for the PLOD rs7529452, MTHFR rs1801133 and CNR2 rs2501431 were observed (Table 3). For hip BMD, the PLOD rs7529452 and rs7514577 showed genotypic associations, and PLOD rs7529452, MTHFR rs1801133 and CNR2 rs2501431 showed marginally significant allelic associations (Table 3). Haplotype analysis Pair-wise linkage disequilibrium between the 18 SNPs is shown in Table 3. For the MTHFR gene, three SNPs were in strong linkage disequilibrium. We identified four haplotypes, each with frequency greater than 5%. Combined, these four haplotypes accounted for 99.6% of the chromosomes in this population. The haplotype ATT showed consistent association with high BMD at all sites measured (P b 0.05). For the CNR2 gene, we identified three common haplotypes, and found no association between any of them and BMD. For the TNFRSF1B gene, two linkage disequilibrium blocks were observed: no association was found between any of these TNFRSF1B haplotypes and BMD. For the PLOD gene, all SNPs were in low linkage disequilibrium. Gene–gene interaction analyses The best two-locus gene–gene effect models identified by MDR method are shown in Table 4. The most promising gene–gene interaction was rs7529452 (PLOD) and rs1801133 (MTHFR) for femoral neck BMD (P = 0.0013). These two SNPs showed significant consistent genotypic/allelic associations with femoral neck BMD in single marker association analyses. The other significant gene–gene interactions were rs2273291 (PLOD) and rs7529452 for hip BMD (P = 0.021). These significant interactions were supported by conditional logistic regression analyses. The P values for the interaction terms of the femoral neck BMD and hip BMD were 0.006 and b0.001, respectively. Discussion Previous linkage studies in multiple populations have shown well-replicated evidence of linkage to chromosome 1p36 for BMD [2–7]. To identify the quantitative trait locus gene underlying BMD variation in Chinese at chromosome 1p36, we targeted four positional and functional candidate genes (TNFRSF1B, PLOD, CNR2, and MTHFR), and performed a gene-wide and tSNP-based association study in a sample of 1243 cases and matched controls. The SNP rs7529452 of PLOD and the SNP rs1801133 of MTHFR showed consistent genotypic/allelic associations with BMD at all sites measured (P = 0.08–0.001). The CNR2 rs2501431 showed nominally significant allelic associations with trochanter BMD and hip BMD (P = 0.02 and 0.04, respectively). The TNFRSF1B rs976881 showed genotypic associations with BMDs (P = 0.08–0.04). In addition, we identified a promising two-locus gene–gene interaction between
987
PLOD and MTHFR. These findings strengthen the potential importance of chromosome 1p36 in contributing to BMD variation. It is likely that the 1p36 locus harbors multiple osteoporosis susceptibility genes that affect BMD. Multiple testing is a major issue in association studies. Permutation testing, Bonferroni correction, and correction by false-discovery rate are effective strategies to address the problems of multiple comparisons. Replication may nonetheless be the best alternative to establish a valid genotype–phenotype association. Although smaller P-values generally provide greater support for a true association, it is the consistency and strength of the association across one or more replication studies rather than the strength of the P value in a single study that is critical to filter out a false-positive association. Instead of a purely positional-cloning approach, we opted to prioritize our association study by focusing on genes with known or putative functions involved in bone metabolism. The associations between these four genes and BMD have been replicated in other populations. Analysis of all genes in chromosome 1p36 can only confirm and will not diminish our findings. Polymorphisms in the PLOD gene and BMD The PLOD gene encodes the enzyme procollagen-lysine, 2oxoglutarate 5-dioxygenase, that catalyses the hydroxylation of lysine residues during the posttranslational modification of type 1 collagen, the major protein of bone matrix. The rs7551175 of PLOD was associated with spine BMD in Scottish women [11] and femoral neck and trochanter BMD in Japanese women [12]. Spotila et al. [13] detected an association between spine BMD and a T→G polymorphism in intron 6 of PLOD which was in complete linkage disequilibrium with rs7551175. In our case or control subjects, the distribution of rs7551175 genotypes was not in Hardy–Weinberg equilibrium (P b 0.001), and did not allow us to test its association with BMD although it is nominally significantly associated with spine BMD (P = 0.04). The SNP rs7529452, which is 1 bp from SNP rs7551175, nonetheless showed consistent genotypic/allelic association with BMD at all sites measured (P = 0.08–0.001). Tasker et al. [11] also observed a weak association between rs7529452 and spine BMD, which was primarily driven by low BMD values in heterozygotes. Thus our results, together with previous observations, support PLOD as a susceptibility gene for osteoporosis. Polymorphisms in the MTHFR gene and BMD MTHFR catalyzes the conversion of 5, 10-methylenetetrahydrofolate to 5-methylenetetrahydrofolate. The rs1801133 polymorphism of MTHFR has previously been associated with bone status in some studies, but the results have been mixed. A recent meta-analysis of studies about the association of the rs1801133 polymorphism and BMD indicated that individuals with TT genotype showed a small but significantly reduced BMD compared with those with TC and CC genotypes [14]. In the current study, the SNP rs1801133 of the MTHFR gene showed weak and consistent genotypic/allelic associations with BMD at all sites measured in single marker and haplotype analyses. Our present results and the previous association studies thus suggest that MTHFR is a susceptibility gene for osteoporosis in Chinese. Polymorphisms in the CNR2 gene and BMD CNR2 knockout mice have decreased bone mass that resembles human osteoporosis [15]. Karsak et al. [16] detected an association of multiple SNPs and haplotypes of CNR2 with osteoporosis, rs4649124 and rs2501431 showing most significant P-value for allele (0.0026 and 0.0014, respectively) and genotype (0.0023 and 0.00073, respectively). The SNP rs2501431 has recently been associated with BMD in Japanese, with the G allele related to reduced BMD [12]. In this study,
988
Q.-Y. Huang et al. / Bone 44 (2009) 984–988
we selected 6 tag SNPs of CNR2 from HaMap to test associations with osteoporosis. CNR2 is encoded by a single exon (exon 2). These SNPs are in tight linkage disequilibrium. The rs2501431 showed nominally significant allelic associations with trochanter and hip BMD. rs4649124 nonetheless showed no evidence of association with BMD in our samples. We also found no association between any of the CNR2 haplotypes and BMD. The role of the CNR2 gene in the etiology of Chinese osteoporosis thus requires further study in larger samples.
Acknowledgments This project is supported by Hong Kong Research Grant Council and seed funding for basic research, The University of Hong Kong, the Bone Health Fund, Hong Kong University Foundation, Matching Grant and the Osteoporosis and Endocrine Research Fund, The University of Hong Kong. QY Huang is partially supported by The KC Wong Education Foundation. References
Polymorphisms in the TNFRSF1B gene and BMD TNFRSF1B, which encodes the 75Kd TNF receptor (TNFR2), regulates the effects of TNF on osteoclastogenesis. Several previous studies have shown evidence of an association between BMD and haplotypes defined by polymorphisms at positions 593, 598 and 620 in the TNFRSF1B 3′ untranslated region. In one study, an association was observed between spine BMD and the A593-G598-T620 (AGT) haplotype [16], whereas another showed an association between femoral neck BMD and the A593-T598-C620 (ATC) haplotype [13, 17–19]. In the present study, TNFRSF1B haplotypes were not associated with BMD, but rs976881 showed genotypic associations with BMDs (P = 0.08–0.04). This suggests that the TNFRSF1B gene may contribute to BMD variation. It has become widely recognized that individually, most susceptibility alleles of complex diseases such as osteoporosis have modest effects. Hence, the well-replicated linkage signals determined by genome wide linkage studies likely arise from multiple genes in the region. The presence of multiple susceptibility genes in the same region makes the region more likely to be detected by multiple linkage studies. In this study, we established that there are multiple osteoporosis susceptibility genes in the 1p36 linkage region. Most of our results were nonetheless not significant with use of a strict Bonferroni correction for the number of tests, although this is a very conservative evaluation of significance. Given the consistency of effects at several sites of BMD measurement, and compelling biological relationships with osteoporosis, the associations we observed are likely to be true. One possibility is that the observed linkage comes from the contribution of multiple alleles and, therefore, no single SNP can reach statistical significance. Another possibility is that there may be other, unidentified osteoporosis genes in this wellreplicated region that we had not evaluated in present study. Genome/regional wide association studies that exhaustively survey all variations in this important interval are needed to demonstrate this in the future. The combined contributions of three genes have recently been reported to explain a major portion of blood pressure variations attributed to the 1q linkage region [20]. Similarly, multiple regions within 8q24 contribute to prostate/colorectal cancer [21–24]. Clustering of multiple genes that affect the same phenotype might be evolutionarily advantageous, and not just an exception. In summary, we have tested the association of four positional and functional candidate genes at 1p36 using a tSNP approach. PLOD and MTHFR showed consistent genotypic/allelic associations with BMD at all sites measured. CNR2 showed nominally significant allelic associations with trochanter BMD and hip BMD and TNFRSF1B showed genotypic association with BMDs. Significant gene–gene interactions between PLOD and MTHFR were identified. Our results suggest that multiple genes at 1p36, individually or in combination, contribute to osteoporosis susceptibility.
[1] Huang QY, Kung AWC. Genetics of osteoporosis. Mol Genet Metab 2006;88: 295–306. [2] Devoto M, Shimoya K, Caminis J, et al. First-stage autosomal genome screen in extended pedigrees suggests genes predisposing to low bone mineral density on chromosomes 1p, 2p and 4q. Eur J Hum Genet 1998;6:151–7. [3] Devoto M, Specchia C, Li HH, et al. Variance component linkage analysis indicates a QTL for femoral neck bone mineral density on chromosomes 1p36. Hum Mol Genet 2001;10:2447–52. [4] Wilson SG, Reed PW, Bansal A. Comparison of genome screens for two independent cohorts provides replication of suggestive linkage of bone mineral density to 3p21 and 1p36. Am J Hum Genet 2003;72:144–55. [5] Xiao P, Shen H, Guo YF, et al. Genomic regions identified for bone mineral density in a large sample including epistatic interactions and gender-specific effects. J Bone Miner Res 2006;21:1536–44. [6] Streeten EA, McBride DJ, Pollin TI, et al. Quantitative trait loci for bone mineral density identified by autosome-wide linkage scan to chromosomes 7q and 21q in men from the Amish Family Osteoporosis Study. J Bone Miner Res 2006;21: 1433–42. [7] Willaert A, Van Pottelbergh I, Zmierczak H, et al. A genome-wide linkage scan for low spinal bone mineral density in a single extended family confirms linkage to 1p36.3. Eur J Hum Genet 2008;16:970–6. [8] Lee YH, Rho YH, Choi SJ, et al. Meta-analysis of genome-wide linkage studies for bone mineral density. J Hum Genet 2006;51:480–6. [9] Huang QY, Kung AWC. Association of common polymorphisms in the QPCT gene with bone mineral density in Chinese. J Hum Genet 2007;52:757–62. [10] Ritchie MD, Hahn LW, Roodi N, et al. Multifactor-dimensionality reduction reveals high-order interactions among estrogen-metabolism genes in sporadic breast cancer. Am J Hum Genet 2001;69:138–47. [11] Tasker PN, Macdonald H, Fraser WD, et al. Association of PLOD1 polymorphisms with bone mineral density in a population-based study of women from the UK. Osteoporos Int 2006;17:1078–85. [12] Yamada Y, Ando F, Shimokata H. Association of candidate gene polymorphisms with bone mineral density in community-dwelling Japanese women and men. Int J Mol Med 2007;19:791–801. [13] Spotila LD, Rodriguez H, Koch M, et al. Association analysis of bone mineral density and single nucleotide polymorphisms in two candidate genes on chromosome 1p36. Calcif Tissue Int 2003;73:140–6. [14] Riancho JA, Valero C, Zarrabeitia MT. MTHFR polymorphism and bone mineral density: meta-analysis of published studies. Calcif Tissue Int 2006;79:289–93. [15] Ofek O, Karsak M, Leclerc N, et al. Peripheral cannabinoid receptor, CB2, regulates bone mass. Proc Natl Acad Sci USA 2006;103:696–701. [16] Karsak M, Cohen-Solal M, Freudenberg J, et al. Cannabinoid receptor type 2 gene is associated with human osteoporosis. Hum Mol Genet 2005;14:3389–96. [17] Spotila LD, Rodriguez H, Koch M, et al. Association of a polymorphism in the TNFRSF1B gene with low bone mineral density. J Bone Miner Res 2000;15:1376–83. [18] Albagha OM, Tasker PN, McGuigan FEA, et al. Linkage disequilibrium between polymorphisms in the human TNFRSF1B gene and their association with bone mass in perimenopausal women. Hum Mol Genet 2002;11:2289–95. [19] Tasker PN, Albagha OME, Masson CB, et al. Association between TNFRSF1B polymorphisms and bone mineral density, bone loss and fracture. Osteoporos Int 2004;15:903–8. [20] Chang YP, Liu X, Kim JD, et al. Multiple genes for essential-hypertension susceptibility on chromosome 1q. Am J Hum Genet 2007;80:253–64. [21] Gudmundsson J, Sulem P, Manolescu A, et al. Genome-wide association study identifies a second prostate cancer susceptibility variant at 8q24. Nat Genet 2007;39:631–7. [22] Haiman CA, Patterson N, Freedman ML, et al. Multiple regions within 8q24 independently affect risk for prostate cancer. Nat Genet 2007;39:638–44. [23] Yeager M, Orr N, Hayes RB, et al. Genome-wide association study of prostate cancer identifies a second risk locus at 8q24. Nat Genet 2007;39:645–9. [24] Tomlinson I, Webb E, Carvajal-Carmona L, et al. A genome-wide association scan of tag SNPs identifies a susceptibility variant for colorectal cancer at 8q24.21. Nat Genet 2007;39:984–8.
Contents lists available at ScienceDirect
Bone j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / b o n e
Multiple osteoporosis susceptibility genes on chromosome 1p36 in Chinese Qing-Yang Huang a,b,⁎, Gloria H.Y. Li a, Annie W.C. Kung a,b,⁎ a b
Department of Medicine, The University of Hong Kong, Hong Kong The Center of Heart, Brain, Hormone and Health Ageing, The University of Hong Kong, Hong Kong
a r t i c l e
i n f o
Article history: Received 29 January 2008 Revised 14 January 2009 Accepted 15 January 2009 Available online 30 January 2009 Edited by: B. Olsen Keywords: Osteoporosis BMD Association Candidate gene Interaction
a b s t r a c t Introduction: Chromosome 1p36 is a region that has previously shown good evidence of linkage to bone mineral density (BMD) in multiple studies, but the genes that are responsible for the linkage signals are unknown. Materials and methods: We performed a gene-wide and tag SNP-based association study of four positional and functional candidate genes (TNFRSF1B, PLOD, CNR2, and MTHFR) at 1p36 in 1, 243 case–control Chinese subjects. Twenty-three tag SNPs were selected and genotyped using the high-throughput Sequenom genotyping platform. Binary logistic regression analyses were performed to test for genotype associations between each SNP and BMD. Allelic and haplotype association analyses were conducted by Haploview. Gene– gene interactions were investigated using multifactor dimensionality reduction method. Results: The PLOD rs7529452 (C385T; F98F) and MTHFR rs1801133 (C677T; A429E) showed significant genotypic/allelic associations with BMDs at all sites measured (P = 0.08–0.001), and a promising two-locus gene–gene interaction for femoral neck BMD. The CNR2 rs2501431 (A592G; G155G) showed nominally significant allelic associations with trochanter and hip BMD. The TNFRSF1B rs976881 showed genotypic associations with BMDs (P = 0.08–0.04). Conclusions: Our results suggest that multiple genes at 1p36, individually or in different combinations, contribute to osteoporosis susceptibility in Chinese. © 2009 Elsevier Inc. All rights reserved.
Introduction Osteoporosis is a major public health problem worldwide and is highly heritable: there is an increased rate of concordance in MZ versus DZ twins and a substantially increased incidence in individuals with a positive family history. More than 20 genome-wide linkage scans have been launched to search for quantitative trait loci that underlie BMD variation. Several genomic regions, such as 1p36, 1q21–25, 2p22–24, 3p14–25, 4q25–34, 6p21, 7p14–21, 11q14–25, 12q23–24, 13q14–34 and 20p12, have been well validated [1]. Evidence for replication of linkage across populations seems to be strongest for a quantitative trait locus on chromosome 1p36. Devoto et al. [2] reported a genome-wide scan in 149 members of seven large pedigrees with the strongest evidence of linkage for hip BMD on chromosome 1p36 (LOD = 3.51). This finding was confirmed and extended in an expanded sample of 42 families [3]. Wilson et al. [4] performed a genome-wide screen of 1097 unselected female UK twin pairs, and observed suggestive linkage for whole-body BMD on chromosome 1 at 17 cM (LOD= 2.38). Xiao et al. [5] recently found suggestive evidence of linkage to wrist BMD on 1p36 in men (LOD = 2.81). Streeten et al. [6] reported suggestive linkage of total hip BMD to a quantitative trait locus on chromosome 1p36 at D1S1597 in ⁎ Corresponding authors. Fax: +852 28162187. E-mail addresses: [email protected] (Q.-Y. Huang), [email protected] (A.W.C. Kung). 8756-3282/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.bone.2009.01.368
women (LOD= 2.02). A recent genome-wide linkage scan in a single extended family also revealed a maximum LOD score of 3.07 on 1p36.3 (D1S468) for spine BMD [7]. Bin 1.1 (1pter-1p36.22) lies above the 95% confidence level (P =0.05) in a meta-analysis of published data from whole genome scans [8]. Obvious functional candidate genes situated within this region include tumor necrosis factor receptor superfamily member 1B (TNFRSF1B, 1p36.3-p36.2), lysyl hydroxylase (PLOD, 1p36.3p36.2), cannabinoid receptor 2 (CNR2, 1p36.11), and methylenetetrahydrofolate reductase (MTHFR, 1p36.3). To identify the quantitative trait locus genes responsible for the linkage signals found at chromosome 1p36, we performed a gene-wide and tag SNP-based association study of four positional candidate genes (TNFRSF1B, PLOD, CNR, and MTHFR) in a sample of 1243 cases and matched controls. Associations with SNPs within multiple genes and gene–gene interaction were observed. Materials and methods Subjects Samples consisted of 1243 case–control Chinese subjects, with 909 spine case–control subjects, 792 femoral neck case–control subjects, 706 trochanter case–control subjects and 760 total hip case–control subjects (Table 1). Cases were arbitrarily defined as subjects with low BMD (Z score ≤ −1.28, equivalent to the lowest 10th percentile of the population). Controls were age and sex-matched subjects with high
Q.-Y. Huang et al. / Bone 44 (2009) 984–988
985
Table 1 Characteristics of 1243 case–control subjects Spine Subject number Postmenopausal women Premenopausal women Men Age (years) Height (m) Weight (kg) BMD (g/cm2) Spine Femoral neck Trochanter Total hip Z-score Spine Femoral neck Trochanter Total hip
Femoral neck
Trochanter
Total hip
Case
Control
Case
Control
Case
Control
Case
Control
476 270 137 69 53.3 ± 14.4 1.54 ± 0.08 49.7 ± 8.2
433 214 153 66 53.7 ± 14.6 1.58 ± 0.07 63.2 ± 10.7
427 235 160 32 51.4 ± 14.9 1.53 ± 0.08 48.6 ± 7.7
365 166 162 37 51.0 ± 14.0 1.58 ± 0.07 64.1 ± 10.9
369 216 117 36 52.8 ± 15.5 1.53 ± 0.08 48.5 ± 7.8
337 142 154 41 50.8 ± 14.3 1.58 ± 0.07 64.8 ± 10.7
420 241 144 35 52.5 ± 15.1 1.53 ± 0.08 48.3 ± 7.3
340 148 152 40 51.0 ± 14.2 1.58 ± 0.07 64.5 ± 10.7
0.67 ± 0.10 0.56 ± 0.09 0.47 ± 0.08 0.64 ± 0.10
1.13 ± 0.11 0.80 ± 0.13 0.71 ± 0.11 0.92 ± 0.13
0.72 ± 0.14 0.52 ± 0.08 0.46 ± 0.08 0.61 ± 0.10
1.07 ± 0.13 0.89 ± 0.10 0.75 ± 0.10 0.97 ± 0.11
0.69 ± 0.13 0.52 ± 0.09 0.43 ± 0.07 0.59 ± 0.09
1.11 ± 0.12 0.87 ± 0.11 0.77 ± 0.09 0.99 ± 0.10
0.70 ± 0.13 0.53 ± 0.08 0.45 ± 0.08 0.59 ± 0.09
1.10 ± 0.12 0.88 ± 0.10 0.76 ± 0.09 1.00 ± 0.10
− 1.99 ± 0.48 − 1.42 ± 0.70 − 1.44 ± 0.72 − 1.61 ± 0.78
1.73 ± 0.63 0.87 ± 0.91 1.03 ± 0.90 1.01 ± 0.92
− 1.62 ± 0.79 − 1.80 ± 0.41 − 1.63 ± 0.64 − 1.89 ± 0.63
1.15 ± 0.94 1.60 ± 0.52 1.38 ± 0.83 1.49 ± 0.78
− 1.80 ± 0.78 − 1.72 ± 0.60 − 1.87 ± 0.45 − 2.01 ± 0.57
1.40 ± 0.93 1.40 ± 0.79 1.66 ± 0.61 1.59 ± 0.72
− 1.76 ± 0.74 − 1.72 ± 0.53 − 1.73 ± 0.57 − 2.00 ± 0.52
1.35 ± 0.92 1.49 ± 0.67 1.56 ± 0.69 1.70 ± 0.58
Data are expressed as mean ± SD.
BMD (Z score N +1) at the corresponding sites. For a common SNP (allele frequency = 0.2), this association study had N90% power to detect a 40% difference in allele frequencies between the case and control groups at the 5% level of significance [9]. All participants gave informed consent and the study was approved by the Ethics Committee of the University of Hong Kong and conducted according to the Declaration of Helsinki. Tag SNP selection and genotyping We selected SNPs on the basis of the following principal criteria: (1) tag SNPs (tSNP) identified using genotype data from the CHB panel (Han Chinese in Beijing) of the phase II HapMap Project. The criterion for tagging was set at r2 N 0.8 and minor allele frequency N0.2. (2) functional relevance and importance. (3) SNPs significantly associated with BMD in previous studies. A total of 23 SNPs of the TNFRSF1B, PLOD, CNR2, and MTHFR genes (9 in the TNFRSF1B, 5 in the PLOD, 3 in the MTHFR, and 6 in the CNR2) were selected (Table 2). The 23 tag SNPs captured 83% of common SNPs (minor allele frequency N0.2) in HapMap Chinese database at r2 N 0.8 (78% in the TNFRSF1B, 70% in the PLOD, 100% in the MTHFR, and 100% in the CNR2). SNPs were genotyped using the high-throughput Sequenom genotyping platform. DNA from case and control subjects was randomly assigned to the 96 well plates, and genotyping was performed blind to the status of the samples. Genotyping was repeated in 5% of samples for verification and quality control. Quality control testing revealed that genotype data had an error rate b 0.1%.
our association as a replication of previous studies. In this situation, a P value b 0.05 was considered statistically significant. The multifactor dimensionality reduction (MDR; http://www. epistasis.org) method was used to test for potential gene–gene interactions [10]. MDR first pooled multi-locus genotypes with high dimension into only one dimension. Cross-validation consistency and testing accuracy were calculated for each combination of a pool of genetic polymorphisms. The final best model was selected as the one with maximal cross-validation consistency and minimal prediction error. Statistical significance was assessed by comparing the average testing accuracy from the observed data with the distribution of average testing accuracy under the null hypothesis of no associations derived empirically from 1000 permutations. Two-locus interactions among the SNPs were tested using MDR. Conditional logistic regression was used to validate the reality of the interaction. If an interaction term is not significant (P N 0.05) by logistic regression, the corresponding significant MDR interaction model could be simply caused by the additive main effects from the component loci. Table 2 Tag SNP information Gene
TNFRSF1B
Statistical analyses The genotyping quality of each SNP was first checked for the call rate, minor allele frequency and Hardy–Weinberg equilibrium. SNPs that failed quality control checks were excluded from further consideration. The program Haploview (http://www.broad.mit.edu/ mpg/haploview/) was used to calculate allele frequencies, to verify that the genotype data were in Hardy–Weinberg equilibrium, to calculate pair-wise linkage disequilibrium statistics, to plot haplotype block structure, and to test allelic and haplotype associations. Haplotypes were estimated using an accelerated EM algorithm in the program Haploview. Binary logistic regression analyses were performed to test for associations between SNP genotype and BMD with adjustment for sex, height, weight and the menopause status. Three possible genetic models (additive, dominant and recessive) were tested for each SNP. All statistical analyses were performed using SPSS version 13.0 for Windows (SPSS Inc., Chicago, IL, USA). As our study is not the first report, we mainly evaluated the significance of
PLOD
MTHFR
CNR2
SNP ID
rs496888 rs976881 rs1201157 rs683240 rs653667 rs5746057 rs5746059 rs1061624 rs1061628 rs1208984 rs7529452 rs7551175 rs7514577 rs2273291 rs9651118 rs1801133 rs1801131 rs4649119 rs2229581 rs2229579 rs4649124 rs2501431 rs7530595
Genomic position (bp)
Genic position
Alleles (major/ minor)
MAFa
12167072 12168020 12183297 12184095 12186074 12196956 12197058 12201531 12202265 11929442 11944221 11944222 11958501 11959619 11796480 11790644 11788742 23916849 23946400 23946468 23946663 23946949 24026412
Intron 1 Intron 1 Intron 2 Intron 2 Intron 3 Intron 9 Intron 9 Exon 10 Exon 10 Intron 1 Exon 3 Exon 3 Exon 12 Intron13 Intron 2 Exon 5 Exon 8 3′ Exon 2 Exon 2 Exon 2 Exon 2 5′
A/G G/A C/T T/C A/C A/C A/G A/G C/T A/G C/T A/G C/T C/T T/C C/T A/C G/C C/G C/T A/G G/A C/T
0.18 0.19 0.27 0.15 0.34 0.01 0.15 0.15 0.31 0.22 0.39 0.48 0.40 0.28 0.46 0.24 0.23 0.45 0.45 0.22 0.46 0.46 0.33
P - values b 0.01 are highlighted in bold type. a MAF: minor allele frequency. b HWE: Hardy–Weinberg equilibrium.
HWEb P
Call rate
0.26 0.76 0.11 0.18 0.00 0.00 0.92 0.00 1.00 0.10 0.51 0.00 0.14 0.05 0.80 0.06 0.69 0.16 0.00 0.38 0.04 0.01 0.51
0.97 0.98 0.94 0.99 0.42 0.96 0.94 0.52 0.94 0.88 0.90 0.81 0.98 0.92 0.96 0.97 0.96 0.96 0.95 0.99 0.96 0.94 0.98
986
Q.-Y. Huang et al. / Bone 44 (2009) 984–988
Results The clinical characteristics of the study subjects are summarized in Table 1. Among the 23 HapMap tSNPs selected, genotyping call rate N90%, minor allele frequency N5%, duplicate error rate b 2% and Hardy–Weinberg equilibrium P values N 1% were achieved in 18 tSNPs. Five tSNPs failed quality control checks, with significant discrepancy with Hardy–Weinberg equilibrium, and were excluded from further analyses. The final coverage of the genes with 18 out of 23 original tag SNPs was 65%. Details for all tSNPs (including genomic and genic position, minor allele frequency, Hardy–Weinberg equilibrium, call rate) are shown in Table 2. Of these, 10 tSNPs were mapped within introns, and 11 were in exons, and 2 were in regulatory regions. Single marker analysis Results of the single marker association test are shown in Table 3. For the PLOD gene, SNP rs7529452 (C385T; F98F) showed consistent genotypic/allelic association with BMD at all sites measured
(P = 0.08–0.001). The SNPs rs2273291 showed significant allele association with spine BMD (P = 0.025), and rs7514577 showed significant genotypic associations with femoral neck BMD and trochanter BMD (P = 0.02). For the MTHFR gene, the SNP rs1801133 (C677T; A429E) showed nominally significant allelic associations with BMD at all sites measured (P = 0.02–0.05) and a weak genotype association trend. For the CNR2 gene, the SNP rs2501431 (A592G; G155G) showed nominally significant allelic associations with trochanter BMD and hip BMD (P = 0.02 and 0.04, respectively). For the TNFRSF1B gene, rs976881 showed genotypic associations with BMD (P = 0.04–0.08). For spine BMD, the genotype frequencies for CC, TC and TT of PLOD rs7529452 were 36.9%, 46.5%, 16.7% in case samples, and 40.8%, 49%, 10.2% in control samples. Frequencies of the TT homozygotes were significantly higher among case than control subjects (P = 0.008; OR = 1.6; 95% CI 1.14–2.35). For femoral neck BMD, the genotype frequencies for CC, TC and TT of MTHFR rs1801133 were 61.3%, 34.4%, 4.4% in case samples, and 52.2%, 42.4%, 5.3% in control samples. Individuals with CC genotype showed a significantly higher risk of
Table 3 Results of single marker association tests
Three possible genetic models (additive, dominant and recessive) were tested and the lowest P values with adjustment for sex, height, weight and the menopause status were shown for each genotype association. P - values V 0.05 are highlighted in bold type.
Q.-Y. Huang et al. / Bone 44 (2009) 984–988 Table 4 Best two-locus gene–gene effect models identified by MDR method
Locus combination
Femoral neck BMD Spine BMD Hip BMD
Trochanter BMD
rs7529452 rs1801133 0.0013 20/20
P value Cross-validation consistency Testing accuracy (%) 59.18
rs2273291 rs1208984 0.4119 15/20
rs2273291 rs7529452 0.0207 15/20
rs1201157 rs1801133 0.8684 10/20
50.23
53.90
49.92
P - values b 0.05 are highlighted in bold type.
osteoporosis compared with those with TC and TT genotypes (P = 0.012; OR = 1.17; 95% CI 1.03–1.33). The T allele of PLOD rs7529452 was associated with low femoral neck BMD (P = 0.04). For trochanter BMD, genotype associations for the PLOD rs7529452 and rs7514577, TNFRSF1B rs976881, and CNR2 rs7530595 and allelic associations for the PLOD rs7529452, MTHFR rs1801133 and CNR2 rs2501431 were observed (Table 3). For hip BMD, the PLOD rs7529452 and rs7514577 showed genotypic associations, and PLOD rs7529452, MTHFR rs1801133 and CNR2 rs2501431 showed marginally significant allelic associations (Table 3). Haplotype analysis Pair-wise linkage disequilibrium between the 18 SNPs is shown in Table 3. For the MTHFR gene, three SNPs were in strong linkage disequilibrium. We identified four haplotypes, each with frequency greater than 5%. Combined, these four haplotypes accounted for 99.6% of the chromosomes in this population. The haplotype ATT showed consistent association with high BMD at all sites measured (P b 0.05). For the CNR2 gene, we identified three common haplotypes, and found no association between any of them and BMD. For the TNFRSF1B gene, two linkage disequilibrium blocks were observed: no association was found between any of these TNFRSF1B haplotypes and BMD. For the PLOD gene, all SNPs were in low linkage disequilibrium. Gene–gene interaction analyses The best two-locus gene–gene effect models identified by MDR method are shown in Table 4. The most promising gene–gene interaction was rs7529452 (PLOD) and rs1801133 (MTHFR) for femoral neck BMD (P = 0.0013). These two SNPs showed significant consistent genotypic/allelic associations with femoral neck BMD in single marker association analyses. The other significant gene–gene interactions were rs2273291 (PLOD) and rs7529452 for hip BMD (P = 0.021). These significant interactions were supported by conditional logistic regression analyses. The P values for the interaction terms of the femoral neck BMD and hip BMD were 0.006 and b0.001, respectively. Discussion Previous linkage studies in multiple populations have shown well-replicated evidence of linkage to chromosome 1p36 for BMD [2–7]. To identify the quantitative trait locus gene underlying BMD variation in Chinese at chromosome 1p36, we targeted four positional and functional candidate genes (TNFRSF1B, PLOD, CNR2, and MTHFR), and performed a gene-wide and tSNP-based association study in a sample of 1243 cases and matched controls. The SNP rs7529452 of PLOD and the SNP rs1801133 of MTHFR showed consistent genotypic/allelic associations with BMD at all sites measured (P = 0.08–0.001). The CNR2 rs2501431 showed nominally significant allelic associations with trochanter BMD and hip BMD (P = 0.02 and 0.04, respectively). The TNFRSF1B rs976881 showed genotypic associations with BMDs (P = 0.08–0.04). In addition, we identified a promising two-locus gene–gene interaction between
987
PLOD and MTHFR. These findings strengthen the potential importance of chromosome 1p36 in contributing to BMD variation. It is likely that the 1p36 locus harbors multiple osteoporosis susceptibility genes that affect BMD. Multiple testing is a major issue in association studies. Permutation testing, Bonferroni correction, and correction by false-discovery rate are effective strategies to address the problems of multiple comparisons. Replication may nonetheless be the best alternative to establish a valid genotype–phenotype association. Although smaller P-values generally provide greater support for a true association, it is the consistency and strength of the association across one or more replication studies rather than the strength of the P value in a single study that is critical to filter out a false-positive association. Instead of a purely positional-cloning approach, we opted to prioritize our association study by focusing on genes with known or putative functions involved in bone metabolism. The associations between these four genes and BMD have been replicated in other populations. Analysis of all genes in chromosome 1p36 can only confirm and will not diminish our findings. Polymorphisms in the PLOD gene and BMD The PLOD gene encodes the enzyme procollagen-lysine, 2oxoglutarate 5-dioxygenase, that catalyses the hydroxylation of lysine residues during the posttranslational modification of type 1 collagen, the major protein of bone matrix. The rs7551175 of PLOD was associated with spine BMD in Scottish women [11] and femoral neck and trochanter BMD in Japanese women [12]. Spotila et al. [13] detected an association between spine BMD and a T→G polymorphism in intron 6 of PLOD which was in complete linkage disequilibrium with rs7551175. In our case or control subjects, the distribution of rs7551175 genotypes was not in Hardy–Weinberg equilibrium (P b 0.001), and did not allow us to test its association with BMD although it is nominally significantly associated with spine BMD (P = 0.04). The SNP rs7529452, which is 1 bp from SNP rs7551175, nonetheless showed consistent genotypic/allelic association with BMD at all sites measured (P = 0.08–0.001). Tasker et al. [11] also observed a weak association between rs7529452 and spine BMD, which was primarily driven by low BMD values in heterozygotes. Thus our results, together with previous observations, support PLOD as a susceptibility gene for osteoporosis. Polymorphisms in the MTHFR gene and BMD MTHFR catalyzes the conversion of 5, 10-methylenetetrahydrofolate to 5-methylenetetrahydrofolate. The rs1801133 polymorphism of MTHFR has previously been associated with bone status in some studies, but the results have been mixed. A recent meta-analysis of studies about the association of the rs1801133 polymorphism and BMD indicated that individuals with TT genotype showed a small but significantly reduced BMD compared with those with TC and CC genotypes [14]. In the current study, the SNP rs1801133 of the MTHFR gene showed weak and consistent genotypic/allelic associations with BMD at all sites measured in single marker and haplotype analyses. Our present results and the previous association studies thus suggest that MTHFR is a susceptibility gene for osteoporosis in Chinese. Polymorphisms in the CNR2 gene and BMD CNR2 knockout mice have decreased bone mass that resembles human osteoporosis [15]. Karsak et al. [16] detected an association of multiple SNPs and haplotypes of CNR2 with osteoporosis, rs4649124 and rs2501431 showing most significant P-value for allele (0.0026 and 0.0014, respectively) and genotype (0.0023 and 0.00073, respectively). The SNP rs2501431 has recently been associated with BMD in Japanese, with the G allele related to reduced BMD [12]. In this study,
988
Q.-Y. Huang et al. / Bone 44 (2009) 984–988
we selected 6 tag SNPs of CNR2 from HaMap to test associations with osteoporosis. CNR2 is encoded by a single exon (exon 2). These SNPs are in tight linkage disequilibrium. The rs2501431 showed nominally significant allelic associations with trochanter and hip BMD. rs4649124 nonetheless showed no evidence of association with BMD in our samples. We also found no association between any of the CNR2 haplotypes and BMD. The role of the CNR2 gene in the etiology of Chinese osteoporosis thus requires further study in larger samples.
Acknowledgments This project is supported by Hong Kong Research Grant Council and seed funding for basic research, The University of Hong Kong, the Bone Health Fund, Hong Kong University Foundation, Matching Grant and the Osteoporosis and Endocrine Research Fund, The University of Hong Kong. QY Huang is partially supported by The KC Wong Education Foundation. References
Polymorphisms in the TNFRSF1B gene and BMD TNFRSF1B, which encodes the 75Kd TNF receptor (TNFR2), regulates the effects of TNF on osteoclastogenesis. Several previous studies have shown evidence of an association between BMD and haplotypes defined by polymorphisms at positions 593, 598 and 620 in the TNFRSF1B 3′ untranslated region. In one study, an association was observed between spine BMD and the A593-G598-T620 (AGT) haplotype [16], whereas another showed an association between femoral neck BMD and the A593-T598-C620 (ATC) haplotype [13, 17–19]. In the present study, TNFRSF1B haplotypes were not associated with BMD, but rs976881 showed genotypic associations with BMDs (P = 0.08–0.04). This suggests that the TNFRSF1B gene may contribute to BMD variation. It has become widely recognized that individually, most susceptibility alleles of complex diseases such as osteoporosis have modest effects. Hence, the well-replicated linkage signals determined by genome wide linkage studies likely arise from multiple genes in the region. The presence of multiple susceptibility genes in the same region makes the region more likely to be detected by multiple linkage studies. In this study, we established that there are multiple osteoporosis susceptibility genes in the 1p36 linkage region. Most of our results were nonetheless not significant with use of a strict Bonferroni correction for the number of tests, although this is a very conservative evaluation of significance. Given the consistency of effects at several sites of BMD measurement, and compelling biological relationships with osteoporosis, the associations we observed are likely to be true. One possibility is that the observed linkage comes from the contribution of multiple alleles and, therefore, no single SNP can reach statistical significance. Another possibility is that there may be other, unidentified osteoporosis genes in this wellreplicated region that we had not evaluated in present study. Genome/regional wide association studies that exhaustively survey all variations in this important interval are needed to demonstrate this in the future. The combined contributions of three genes have recently been reported to explain a major portion of blood pressure variations attributed to the 1q linkage region [20]. Similarly, multiple regions within 8q24 contribute to prostate/colorectal cancer [21–24]. Clustering of multiple genes that affect the same phenotype might be evolutionarily advantageous, and not just an exception. In summary, we have tested the association of four positional and functional candidate genes at 1p36 using a tSNP approach. PLOD and MTHFR showed consistent genotypic/allelic associations with BMD at all sites measured. CNR2 showed nominally significant allelic associations with trochanter BMD and hip BMD and TNFRSF1B showed genotypic association with BMDs. Significant gene–gene interactions between PLOD and MTHFR were identified. Our results suggest that multiple genes at 1p36, individually or in combination, contribute to osteoporosis susceptibility.
[1] Huang QY, Kung AWC. Genetics of osteoporosis. Mol Genet Metab 2006;88: 295–306. [2] Devoto M, Shimoya K, Caminis J, et al. First-stage autosomal genome screen in extended pedigrees suggests genes predisposing to low bone mineral density on chromosomes 1p, 2p and 4q. Eur J Hum Genet 1998;6:151–7. [3] Devoto M, Specchia C, Li HH, et al. Variance component linkage analysis indicates a QTL for femoral neck bone mineral density on chromosomes 1p36. Hum Mol Genet 2001;10:2447–52. [4] Wilson SG, Reed PW, Bansal A. Comparison of genome screens for two independent cohorts provides replication of suggestive linkage of bone mineral density to 3p21 and 1p36. Am J Hum Genet 2003;72:144–55. [5] Xiao P, Shen H, Guo YF, et al. Genomic regions identified for bone mineral density in a large sample including epistatic interactions and gender-specific effects. J Bone Miner Res 2006;21:1536–44. [6] Streeten EA, McBride DJ, Pollin TI, et al. Quantitative trait loci for bone mineral density identified by autosome-wide linkage scan to chromosomes 7q and 21q in men from the Amish Family Osteoporosis Study. J Bone Miner Res 2006;21: 1433–42. [7] Willaert A, Van Pottelbergh I, Zmierczak H, et al. A genome-wide linkage scan for low spinal bone mineral density in a single extended family confirms linkage to 1p36.3. Eur J Hum Genet 2008;16:970–6. [8] Lee YH, Rho YH, Choi SJ, et al. Meta-analysis of genome-wide linkage studies for bone mineral density. J Hum Genet 2006;51:480–6. [9] Huang QY, Kung AWC. Association of common polymorphisms in the QPCT gene with bone mineral density in Chinese. J Hum Genet 2007;52:757–62. [10] Ritchie MD, Hahn LW, Roodi N, et al. Multifactor-dimensionality reduction reveals high-order interactions among estrogen-metabolism genes in sporadic breast cancer. Am J Hum Genet 2001;69:138–47. [11] Tasker PN, Macdonald H, Fraser WD, et al. Association of PLOD1 polymorphisms with bone mineral density in a population-based study of women from the UK. Osteoporos Int 2006;17:1078–85. [12] Yamada Y, Ando F, Shimokata H. Association of candidate gene polymorphisms with bone mineral density in community-dwelling Japanese women and men. Int J Mol Med 2007;19:791–801. [13] Spotila LD, Rodriguez H, Koch M, et al. Association analysis of bone mineral density and single nucleotide polymorphisms in two candidate genes on chromosome 1p36. Calcif Tissue Int 2003;73:140–6. [14] Riancho JA, Valero C, Zarrabeitia MT. MTHFR polymorphism and bone mineral density: meta-analysis of published studies. Calcif Tissue Int 2006;79:289–93. [15] Ofek O, Karsak M, Leclerc N, et al. Peripheral cannabinoid receptor, CB2, regulates bone mass. Proc Natl Acad Sci USA 2006;103:696–701. [16] Karsak M, Cohen-Solal M, Freudenberg J, et al. Cannabinoid receptor type 2 gene is associated with human osteoporosis. Hum Mol Genet 2005;14:3389–96. [17] Spotila LD, Rodriguez H, Koch M, et al. Association of a polymorphism in the TNFRSF1B gene with low bone mineral density. J Bone Miner Res 2000;15:1376–83. [18] Albagha OM, Tasker PN, McGuigan FEA, et al. Linkage disequilibrium between polymorphisms in the human TNFRSF1B gene and their association with bone mass in perimenopausal women. Hum Mol Genet 2002;11:2289–95. [19] Tasker PN, Albagha OME, Masson CB, et al. Association between TNFRSF1B polymorphisms and bone mineral density, bone loss and fracture. Osteoporos Int 2004;15:903–8. [20] Chang YP, Liu X, Kim JD, et al. Multiple genes for essential-hypertension susceptibility on chromosome 1q. Am J Hum Genet 2007;80:253–64. [21] Gudmundsson J, Sulem P, Manolescu A, et al. Genome-wide association study identifies a second prostate cancer susceptibility variant at 8q24. Nat Genet 2007;39:631–7. [22] Haiman CA, Patterson N, Freedman ML, et al. Multiple regions within 8q24 independently affect risk for prostate cancer. Nat Genet 2007;39:638–44. [23] Yeager M, Orr N, Hayes RB, et al. Genome-wide association study of prostate cancer identifies a second risk locus at 8q24. Nat Genet 2007;39:645–9. [24] Tomlinson I, Webb E, Carvajal-Carmona L, et al. A genome-wide association scan of tag SNPs identifies a susceptibility variant for colorectal cancer at 8q24.21. Nat Genet 2007;39:984–8.