The authors state that they have no conflicts of interest.
In this study, the associations of novel LRP5 variants with BMD variation were detected and some replicated in the two ethnic groups of Chinese and white origins, respectively. These data support the concept that LRP5 variation can contribute to minor and major variation in bone structure.
Introduction: Mutations in the low-density lipoprotein receptor-related protein 5 (LRP5) gene have been shown to cause both high and low bone mass. However, it is still controversial whether LRP5 is associated with normal BMD variation. This study explored the association of LRP5 with BMD phenotypes at three clinically important skeletal sites—the spine, hip, and ultradistal radius (UD)—in two independent populations of Chinese and white ethnicities, respectively.
Materials and Methods: The Chinese sample consisted of 733 unrelated subjects. The white sample was made up of 1873 subjects from 405 nuclear families. High-density single nucleotide polymorphisms (SNPs) across the whole LRP5 gene were genotyped and analyzed in both samples.
Results: Linkage disequilibrium (LD) analyses showed that the haplotype structures of LRP5 between Chinese and whites were in good agreement. Association tests showed that polymorphisms in block 5 spanning intron 7 to intron 19 of LRP5 significantly associated with spine BMD variation in both samples. Particularly, the significant association of SNP rs491347 in intron 7 with spine BMD in the Chinese sample (p = 0.002) was replicated in whites, even after adjusting for multiple testing (p = 0.005). Its strongly associated SNP rs1784235 could cause the loss of an estrogen receptor α (ERα) binding site in LRP5, which could partially explain the above replicated association. However, we did not observe any significant replication with BMD variation at the hip and UD. After accounting for multiple testing, associations with BMD variation at these two sites were mainly found in Chinese. Sex-stratified analyses further revealed that the LRP5 associations with BMD in Chinese and whites were driven by male and female subjects, respectively.
Conclusions: Our work supported LRP5 genetic variants as possible susceptibility factors for osteoporosis and fractures in humans. Especially, the SNP rs491347 and its strongly associated SNPs (e.g., rs1784235) could be important to human osteoporosis phenotypes.
BMD is the predominant study phenotype in osteoporosis genetics because it has a high heritability of ∼70%,(1) and low bone mass is one major risk factor for osteoporosis and low/no-trauma fractures.(2) Many genes have been examined for their association with normal BMD variation, which yields an ever-expanding candidate gene list.
However, a prominent osteoporosis candidate gene, the low-density lipoprotein receptor–related protein 5 (LRP5) gene, was identified because its mutations lead to severe Mendelian bone phenotypes. The LRP5 gene is composed of 23 exons and located to 11q12–13 in the human genome. LRP5 acts as a co-receptor for Wnt proteins that are required for the osteoblast proliferation and functions.(3) Different LRP5 mutations can cause either extremely high or low bone mass traits,(4–6) suggesting the prominent role of LRP5 in regulating bone mass.
Recently, common polymorphisms in LRP5 were also studied and associated with normal BMD variation in humans.(7–12) However, the association results for a particular locus usually cannot be replicated, especially among populations of distinct ethnicities.(12,13) In this study, using two large samples of Chinese and white ethnicities, respectively, we explored the role of common LRP5 variants in determining BMD variations at three clinically important skeletal sites—spine, hip, and ultradistal radius (UD).
MATERIALS AND METHODS
The Chinese part of this study was approved by the research administration of Hunan Normal University. The Chinese cohort consisted of 733 unrelated subjects (369 men and 364 women) 20–45 years of age, who belong to the Chinese Han ethnic group in Changsha and the surrounding areas of central south of China. Three ways of advertisement (internet, poster, and interview) were adopted to recruit subjects. After the subject signed the informed consent document, a questionnaire was given to obtain the subject's information on age, sex, medical history, family history, female history, physical activity, alcohol use, dietary habits, and smoking history under the direction of clinician. We adopted the exclusion criteria detailed elsewhere(14) to screen and recruit “healthy” subjects. The basic characteristics of the Chinese subjects are listed in Table 1.
Table Table 1.. Basic Characteristics of the Study Subjects From the Chinese and White Samples
BMD values of lumbar spine (L1–L4), total hip (neck, trochanter, and intertrochanter), and UD were measured by a Hologic QDR 2000 + DXA scanner (Hologic, Waltham MA, USA). The machine was calibrated daily, and the CV value of the DXA measurements for BMD was 1.33% on the Hologic 2000+.
The white part of this study was approved by the Creighton University Institutional Review Board. Signed informed consent documents were obtained from all study participants before they entered the study. People with chronic diseases and conditions that might potentially affect bone mass, structure, or metabolism were excluded as detailed previously.(14) All of the 1873 participants from 405 nuclear families were U.S. whites of European origin and recruited for BMD study by advertising. The general related characteristics are also listed in Table 1. The sample composition was the same as our previous study,(14) yielding 1512 sib-pairs and 2266 parent-offspring pairs in total.
BMD values of whites at spine (L1–L4), total hip (neck, trochanter, and intertrochanter), and UD were measured using a Hologic QDR 2000 + or 4500 DXA (Hologic). Both machines were calibrated daily. The CV values of the DXA measurements for BMD were 1.87% on the Hologic 2000+ and 1.98% on the Hologic 4500.
Single nucleotide polymorphism selection
We selected single nucleotide polymorphisms (SNPs) according to public information available in the dbSNP (http://www.ncbi.nlm.nih.gov/SNP) and HapMap (www.hapmap.org) databases. In the Chinese and white samples, the respective 29 and 41 SNPs in and around the LRP5 gene were determined on the basis of the following criteria: (1) validation status in Chinese and whites, (2) an average density of one SNP per 3–5 kb, (3) degree of heterozygosity, (i.e., minor allele frequencies [MAFs] > 0.05), (4) functional relevance and importance, and (5) reported to dbSNP by various sources of confirmation. In the random sample of Chinese ethnicity, we also selected another three SNP markers located in different genomic regions (GCR, HDC, CCR3) to control for the problem of population stratification.
In the Chinese, genomic DNA was isolated from whole blood using the phenol-chloroform extraction method. The concentration of DNA was quantified using a DU530 UV/VIS Spectrophotometer (Beckman Instruments, Fullerton, CA, USA). Twenty-five LRP5 SNPs (of the initial 29) and three SNP markers of other unlinked genes were successfully genotyped using the TaqMan assay in a 384-well microplate format. Briefly, 20 ng of DNA was amplified in a final volume of 5 μl containing 0.025 μl of 40× MGB probes and primers, 1 μl of 2× TaqMan Universal PCR Master Mix, and 0.004 μl of 100× BSA (New England Biolabs, Beverly, MA, USA). Amplification conditions were 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 s and 60°C for 1 minute. Allelic discrimination was performed using the ABI PRISM 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). SNPs genotyped by the Taqman technique were assayed in duplicate, and discordant genotypes were discarded. The genotype discordant rate was <0.3% in the Chinese sample.
In whites, genomic DNA was extracted from whole blood using a commercial isolation kit (Gentra Systems, Minneapolis, MN, USA). DNA concentration was assessed by a DU530 UV/VIS Spectrophotometer (Beckman Coulter, Fullerton, CA, USA). Of 41 LRP5 SNPs intended in whites, 27 SNPs were successfully genotyped using the high-throughput BeadArray SNP genotyping technology of Illumina (San Diego, CA, USA). The overall sample success rate, locus success rate, and genotype call rates were 98.24%, 85.16%, and 99.95%, respectively. The reproducibility rate as revealed through blind duplicating was 100%. The GC_Scores measuring the reliability of the genotypes called for LRP5 in whites ranged from 0.75 to 0.96, with a mean of 0.87 ± 0.08 (SD).
PedCheck(15) was used to check Mendelian consistency of SNP genotype data in the white sample, and any inconsistent genotypes were removed. The error checking option embedded in Merlin(16) was run to identify and disregard the genotypes flanking excessive recombinants, thus further reducing genotyping errors. Less than 0.02% of the overall genotypes analyzed (<10) were omitted because of the violation of any of the above two rules.
Hardy-Weinberg equilibrium, linkage disequilibrium, and population stratification analysis. In the overall Chinese sample of 733 unrelated subjects, the program PowerMarker(17) was used to calculate allele frequencies and χ2 test adopted to verify that the genotype data were in Hardy-Weinberg equilibrium (HWE). The program Haploview(18) was used to calculate pairwise linkage disequilibrium (LD) statistics (namely D' = D/Dmax, proportion of observed LD of maximum possible LD) among all the SNPs, select haplotype tagging SNPs (htSNPs), and plot haplotype block structure of LRP5 (through the criteria of Gabriel et al.(19)). Using genotype data of four markers located in different chromosomes (LRP5, GCR, HDC, CCR3), we ran the program STRUCTURE(20) to determine whether there was hidden population structure in our Chinese sample that could cause false-positive associations.
Population-based association analysis. Normality test–Kolmogorov-Smirnov test implemented in Minitab (Minitab, State College, PA, USA) was conducted before association analysis. It was assumed that BMD residuals obtained by adjusting for significant covariates (here including age, sex, and weight) should obey normal distribution. After appropriate outlier deletion and/or data transformation, all kinds of the BMD residuals (spine, hip, and UD) met the normality criteria. The possible association of LRP5 SNPs and haplotypes with BMD variation (normalized BMD residuals) was analyzed using QTPHASE program within UNPHASED.(21) UNPHASED is a suite of programs able to conduct robust association analysis of multilocus haplotypes from unphased genotype data. QTPHASE within UNPHASED was used to conduct SNP or haplotype association with quantitative traits.(21) To circumvent the problem of false-positive results incurred by multiple testing, the program SNPSpD developed by Nyholt(22) was used to establish the experiment-wide significance threshold needed to keep type I error rate <5%, considering the redundancy of information provided by SNPs that were in LD.
HWE and LD analysis. The allele frequencies for each SNP were estimated in all the subjects of the white nuclear families through a maximum likelihood method implemented in the program SOLAR(23) (http://www.sfbr.org/sfbr/public/software/solar). HWE was tested using the PEDSTATS procedure embedded in Merlin(16) (http://www.sph.umich.edu/csg/abecasis/Merlin/). Similar to the analyses in Chinese, the program Haploview was adopted to compute pairwise LD statistics, select htSNPs, and generate graphical representation of LD structure using the group of 703 unrelated parents in the 405 white nuclear families.
Family-based association analysis. We conducted a family-based association test (FBAT)(24,25) for the quantitative trait BMD residuals adjusted by significant covariates including age, sex, height, and weight. The haplotype version of FBAT (HBAT)(26) was performed to obtain empirical global p values for single-SNP and haplotype markers (with adjustment for all possible alleles or haplotypes) by means of the Monte-Carlo permutation procedures implemented in HBAT (10,000 permutations were conducted). The program SNPSpD(22) was also used in whites to establish the experiment-wide significance threshold needed to keep type I error rate <5%.
All of the association analyses were also performed in the male and female subgroups of the Chinese and white samples separately to detect the potential sex-linked bone effects of LRP5. The procedures were the same as those done in the entire samples of Chinese and white ethnicities, respectively.
HWE tests, haplotype blocks, and population stratification
The MAFs of all the 25 SNPs successfully genotyped in Chinese were >5%. However, four SNPs of them were excluded from subsequent analysis because their genotype frequency distributions were not in HWE (p < 0.001). Of the 27 successfully genotyped SNPs in whites, only 1 was discarded because that its minor allele frequency was <1%. The genotype frequency distributions of the remaining 26 SNPs were in HWE.
Of the 35 SNPs finally analyzed (Table 2), 12 were studied in both Chinese and white samples. Nine SNPs were studied only in Chinese, whereas 14 were studied only in whites. The basic information of these SNPs was tabulated in Table 2. LD analyses using the program Haploview showed that the haplotype structures of the LRP5 gene were highly consistent between Chinese and whites (Fig. 1). Briefly, the entire LRP5 gene region was divided into six blocks (in Chinese, block 6 was not shown because no 3′-UTR SNPs were genotyped), with several “orphan” SNPs (i.e., those that do not belong to any LD block) lying between blocks 4 and 5.
Table Table 2.. General Information of the Analyzed LRP5 SNPs in This Study
To obviate possible population stratification in our Chinese sample, we ran the program STRUCTURE,(20) using genotype data from SNPs on four distinct independent loci (LRP5: rs607887; GCR: rs852977; HDC: rs854150; CCR3: rs1491962). The highest likelihood was obtained under the assumption of one population (K = 1), whereas none of the individuals was assigned to a particular population for higher values of K. These indicate that an unknown population substructure was unlikely to exist in our Chinese cohort. For whites, we did not attempt to detect population stratification because the FBAT approaches robust to such a problem were adopted.
Association of LRP5 polymorphisms with spine BMD variation—the replication found across the Chinese and white populations
In the following, all the associations referred to the residual BMD values unless otherwise specified. Of the 12 SNPs genotyped in both samples, 3 showed replicated associations with spine BMD variation (p < 0.05; Table 3). These SNPs are rs491347, rs607887, and rs901823, and all are present in the block 5 region of LRP5. The experiment-wide significance threshold p values were set at 0.006 and 0.005 by the program SNPSpD for the Chinese and white samples, respectively. Thus, the association of rs491347 with spine BMD variation remained significant in both Chinese (p = 0.002) and whites (p = 0.005) after correction for multiple testing. In sex-stratified analyses, experiment-wide significant association of rs491347 with spine BMD were still present in white women (p = 0.005) and Chinese men (p = 0.002). Moreover, the association of rs607887, rs901823, and the whole block 5 with spine BMD were all significant in Chinese men after multiple-testing correction (p = 0.001, 0.005, and 0.001, respectively).
Table Table 3.. Associations of Shared SNPs With Spine BMD in the Chinese and White Samples
For the SNP rs491347 (A/G), showing the strongest and replicated evidence of association, its minor allele G was associated with lower spine BMD values in both ethnic groups. Using the raw BMD data unadjusted by significant covariates, in the total Chinese sample, we found that subjects with the GG or GA genotypes had significant lower spine BMD than those with the AA genotype (raw BMD values were 0.92:0.95:0.97 g/cm2 for GG:GA:AA genotypes, p = 0.030), showing an allele dose effect of 0.025 g/cm2 lower spine BMD value per copy of the G allele. In Chinese men, this phenomenon was more obvious (raw BMD values were 0.93:0.97:1.00 g/cm2 for GG:GA:AA genotypes, p = 0.048), yielding an allele dose effect of 0.035 g/cm2 lower spine BMD value per copy of the G allele. In the overall white offspring sample and the female offspring subsample, similar patterns were observed (in the former, raw BMD values were 1.02:1.05:1.07 g/cm2 for GG:GA:AA genotypes, p = 0.035; in the latter, raw BMD values were 1.01:1.04:1.06 g/cm2 for GG:GA:AA genotypes, p = 0.025).
Association of LRP5 polymorphisms with hip and UD BMD variation
For BMD variation at the hip and UD, there was no significant replication detected across the Chinese and white samples. However, significant associations after multiple testing correction (threshold p = 0.006) were detected in the Chinese. As shown in Table 4, rs3781590 in block 4 was associated with hip BMD variation in both the total and male samples of the Chinese (p = 0.0001). Another two SNPs, rs643981 in block 3 and rs901823 in block 5, significantly associated with hip BMD in Chinese men (p = 0.001 and 0.002, respectively). Blocks 3 and 4 as a whole were also associated with hip BMD variation after accounting for multiple testing (p = 0.001/0.002 and 0.0001/0.0009 in the total/male samples, respectively). For UD BMD variation, corrected significant associations were found in the total Chinese sample at rs643981 in block 3 (p = 0.002) and rs671191 in block 4 (p = 0.002), as well as the whole block 4 region (p = 0.004). All the significant associations of LRP5 with BMD variation at the hip and UD in the Chinese, including the nominal ones (p < 0.05), are listed in Table 4.
Table Table 4.. Significant Associations of LRP5 With Hip and UD BMD in the Chinese Sample
In whites, two SNPs, rs682429 and rs312016 in block 1, were nominally associated with hip BMD variation (p = 0.030 and 0.036). In the sex-stratified analyses, these two also showed nominal association with hip BMD in women (p = 0.026 and 0.029). Another SNP, rs312786 in block 3, was nominally significant for hip and UD BMD variation in white women (p = 0.043 and 0.050). However, these associations were not significant after multiple testing correction and not replicated in the Chinese. None of the LRP5 SNPs studied were significant for hip and UD BMD variation in white men.
We tested whether LRP5 gene polymorphisms contributed to normal peak BMD variation in human populations. This study had several notable strengths. First, two large populations of different ethnic background were examined, with certain significant association replicated in both groups. Second, a large number of SNPs distributed densely across LRP5 were studied in each ethnic group (Chinese: 21; white: 26). Third, SNPs having high heterozygosity were used to maximize the power to find association with BMD. Fourth, LD analysis was conducted, and the haplotype information was used in the association tests. Fifth, the problems often afflicting the genetic association study, such as population stratification and multiple testing, were circumvented to minimize the number of false-positive results. The detected replications further added to the robustness of our findings.
We found that the LD patterns of LRP5 between Chinese and whites were quite similar. Our identified LRP5 haplotype block structures were in excellent agreement with those reported previously.(9,27) Both our and others' studies showed extensive substantial LD throughout the majority of the LRP5 gene and the presence of a region of very low LD in the middle of LRP5 region. However, the exact picture of the haplotype structure of LRP5 could be established only when all of the LRP5 polymorphisms were studied.
The role of the LRP5 gene in normal BMD variation had been intensively studied in recent years.(7–11,13,28–30) Previous studies listed several possible loci underlying BMD variation, such as c.2047G>A, c.4037C>T, c.3357A>G, A1330V, and C171346A. However, positive results usually could not be replicated, especially when across populations of different ethnicities. For example, the c.4037C>T missense substitution was not significantly associated with BMD variation in Asians(13) and Africans,(31) although it was significant in whites.(7,12,29) Here we reported the successfully replicated association with spine BMD in the LRP5 block 5 region, especially for the SNP rs491347 (Table 4). Bioinformatics analysis using the Vista program(32) to compare the human and mouse LRP5 genomic sequences showed that the block 5 region was highly conserved,(33) suggesting the existence of functional variants within this block that may be the “causal” ones influencing BMD variations. The spine BMD-decreasing effect of the G allele of rs491347 was obvious in both Chinese and whites, with GG genotypes having ∼5–7% lower BMD values than AA genotypes. Such replication for LRP5 across different ethnic groups was discovered for the first time. Although rs491347 is in intron 7, it is still possible for such intronic polymorphisms to influence gene expression, mRNA splicing, export, and decay.(8) An alternative explanation is that rs491347 is in LD with other spine BMD-associated variants in block 5. Actually, SNP rs1784235 that was in strong LD with rs491347 in our Chinese sample (pairwise r2 = 0.91) also showed an experiment-wide significant association with spine BMD. Although rs1784235 was not genotyped in our white sample, because of the perfect LD of it with rs491347 in whites as revealed by HapMap data (pairwise r2 = 1.0), we could safely conclude the replicated association of rs1784235 with spine BMD in our two ethnic groups. More interestingly, the MAPPER program showed that rs1784235 was in the middle of a putative estrogen receptor α (ERα) binding site in the LRP5 gene, and the change of the major allele T of rs1784235 (in LD with A of rs491347) to the minor allele C (in LD with G of rs491347) could cause the loss of this ERα binding site. Given the well-known function of ERα in regulating various gene transcription in a wide variety of issues and its prominent role in osteoporosis, this led us to hypothesize that rs1784235 may be a functional site through which ERα regulates the gene expression of LRP5, which finally influences spine BMD variation in human populations. However, such a hypothesis needs to be substantiated by molecular and cellular studies both in vivo and in vitro.
As for BMD at other skeletal sites, the most significant hip BMD-associated SNPs in the Chinese, rs3781590 and rs643981 (p = 1 × 10−5 and 1 × 10−3, respectively), could not be replicated in whites from both our own and other groups' studies.(9) The same phenomenon was observed for UD BMD-associated SNPs: rs643981 and rs671191. This could be explained by the significant allele frequency differences for these markers between the Chinese and white samples (MAFs of rs3781590, rs643981, and rs671191 in the two ethnic groups were 0.125/0.340, 0.164/0.410, and 0.142/0.350, p < 1 × 10−8; Table 2). Therefore, although these markers may be in strong LD with the causal variants in the Chinese, they cannot capture the same causal variants in whites because of their much inflated MAFs. Thus, the use of population-specific tagging SNP marker sets instead of a universal set when conducting genetic association studies in different ethnic groups should be the norm. Under such circumstances, HapMap information is particularly useful.
In sex-stratified analyses, we noticed that the majority of our significant findings in the Chinese were mainly present in men (Tables 3 and 4), whereas those in whites were mainly present in women (Table 3).(34) This might reflect that the sex and ethnicity could interact to influence the LRP5 associations with BMD. However, more data are needed to justify this point.
Overall, our findings indicated that LRP5 variants significantly associated with the determination of vertebral BMD in both Chinese and whites. We successfully replicated the association of certain LRP5 polymorphisms with spine BMD variation, on the basis of which future molecular genetic studies may be conducted to unravel the underlying mechanisms. We also detected the significant associations of LRP5 variants with BMD at the hip and UD, especially in the Chinese. Our data, combined together with the ever-increasing evidence of LRP5 function in bone metabolism, supported LRP5 variants as important genetic susceptibility factors for bone mass and osteoporosis.
Investigators of this work were partially supported by grants from the NIH (R01 AR050496, K01 AR02170-01, R01 AR45349-01, and R01 GM60402-01A1) and an LB595 grant from the State of Nebraska. The study also benefited from grants from National Science Foundation of China, Huo Ying Dong Education Foundation, HuNan Province, Xi'an Jiaotong University, and the Ministry of Education of China.