Calcified Tissue International

, Volume 81, Issue 4, pp 254–262

Variation in the BMP2 Gene: Bone Mineral Density and Ultrasound in Young Adult and Elderly Women

Authors

  • Fiona E. McGuigan
    • Clinical and Molecular Osteoporosis Research Unit, Department of Clinical Sciences MalmöLund University
    • Department of OrthopedicsMalmö University Hospital
  • Emma Larzenius
    • Clinical and Molecular Osteoporosis Research Unit, Department of Clinical Sciences MalmöLund University
    • Department of OrthopedicsMalmö University Hospital
  • Mattias Callreus
    • Clinical and Molecular Osteoporosis Research Unit, Department of Clinical Sciences MalmöLund University
    • Department of OrthopedicsMalmö University Hospital
  • Paul Gerdhem
    • Clinical and Molecular Osteoporosis Research Unit, Department of Clinical Sciences MalmöLund University
    • Department of OrthopedicsKarolinska University Hospital, Karolinska Institute
  • Holger Luthman
    • Medical Genetics Unit, Department of Clinical Sciences MalmöLund University
    • Clinical Research CenterMalmö University Hospital
    • Clinical and Molecular Osteoporosis Research Unit, Department of Clinical Sciences MalmöLund University
    • Department of OrthopedicsMalmö University Hospital
Article

DOI: 10.1007/s00223-007-9054-9

Cite this article as:
McGuigan, F.E., Larzenius, E., Callreus, M. et al. Calcif Tissue Int (2007) 81: 254. doi:10.1007/s00223-007-9054-9

Abstract

Bone morphogenetic protein-2 (BMP2) plays a key role in bone formation and maintenance. Studies of polymorphisms within the gene in relation to bone mineral density (BMD) and fracture have been inconsistent. Our aim was to investigate associations between polymorphisms in the BMP2 gene and bone mass, fracture, and quantitative ultrasound (QUS) measures at different stages of skeletal development. Study subjects were participants of two population-based cohorts of Swedish women: the PEAK-25 cohort of young adult women aged 25 years (n = 993) and the OPRA cohort of elderly women aged 75 years (n = 1,001). We analyzed four single-nucleotide polymorphisms (SNPs) across the BMP2 gene including the Ser37Ala SNP previously identified in relation to BMD, QUS of the calcaneus, and, in the elderly women, fracture. BMP2 gene variations were associated with QUS of bone, independent of BMD, but only in the young women. Even after adjusting for confounding factors, SNP rs235754 in the 3′ region of the gene was significantly associated with the ultrasound parameters speed of sound (P = 0.003) and stiffness (P = 0.002). The 5′ SNP rs235710 showed trends for QUS parameters (P = 0.02–0.07). No association with BMP2 SNPs was observed in either cohort for either BMD or fracture. While further, more extensive genotyping across the gene is recommended, as we may not have captured all information, our preliminary data suggest that variation in BMP2 may play a previously unidentified role in aspects of bone quality, which may be age- and site-dependent.

Keywords

BMP2PolymorphismPeak bone massUltrasoundFracture

Osteoporosis is a common skeletal disease characterized by low bone mineral density (BMD) and microarchitectural deterioration of bone tissue, leading to increased risk of fragility fracture. Osteoporosis risk has a strong genetic component, with both common and distinct genes associated with variations in BMD and fracture risk [1, 2]. Bone strength is influenced by the combined properties of bone mass and bone quality, the latter not fully defined but encompassing structural and biomechanical aspects. Quantitative ultrasound (QUS) is believed to provide additional information about these properties independent of BMD measurements by dual-energy X-ray absorptiometry (DXA) [3]. Attainment of peak bone mass is a major determinant of bone mass in adulthood, whereas rate of bone loss and the resultant BMD, in addition to nonskeletal factors, play an important role with advancing age. Bone strength, which itself has a strong hereditary element, is also an important contributor to bone fragility in both young and elderly women [4].

Association studies of single-nucleotide polymorphisms (SNPs) in many osteoporosis candidate genes, including COL1A1 [5, 6], estrogen receptor-alpha (ERα) [7, 8], vitamin D receptor (VDR) [911], and transforming growth factor-beta (TGF-β1) [12, 13], indicate that variations in these genes account for only a small portion of the population variance in bone mass. Bone morphogenetic protein-2 (BMP2), a signaling molecule belonging to the TGF-β superfamily, is a strong candidate gene for osteoporosis. It plays a key role during embryogenesis but also contributes to skeletal maintenance later in life, stimulating osteoblast differentiation from osteoprogenitor cells [14]. BMP2 also affects the transcription of target genes via Smad pathways [15]. In a large genomewide linkage study performed in Iceland by Styrkarsdottir et al. [16], linkage was on chromosome 20p12 with both osteoporotic fracture and BMD. Associations between a missense polymorphism (Ser37Ala) in exon 2 of the BMP2 gene, low bone mass, and fracture risk were also reported. Despite this promising start, subsequent studies of this and other SNPs in the BMP2 gene have been unable to provide conclusive evidence for the importance of variations in this gene to the development of osteoporosis [1720].

Recently, it has been speculated that a function of BMPs is to accelerate bone turnover, thereby improving the mechanical strength and quality of bone [21]. In our study, we aimed to determine if variation in the BMP2 gene plays a more important role in bone strength, reflected in ultrasound measurements, than in BMD. We also examined the relative importance of BMP2 polymorphisms in women at different skeletal stages: young adult women at peak bone mass and women in late postmenopause with an elevated risk of fracture. To this end, we evaluated almost 2,000 women from the PEAK-25 and OPRA cohorts.

Materials and Methods

Subjects

The study subjects were participants in one of two population-based studies. The PEAK-25 cohort consists of young Caucasian women living in Malmö, Sweden, between 1999 and 2003. The women, all of whom were aged 25 years, were randomly selected from the city files. Initially, 2,394 women were invited by letter 1 week after their 25th birthday to attend the Osteoporosis Research Unit, Malmö University Hospital, within the following 12-month period. Of those invited, 1,059 agreed to participate (44%). Of the 1,335 nonparticipants, 809 (60.6%) could not be contacted or did not reply, 412 (30.9%) expressed that they did not want to participate, 12 (0.9%) no longer lived in Sweden, and 102 (7.6%) were excluded due to pregnancy or the recent birth of a child. The Malmö Osteoporosis Prospective Risk Assessment (OPRA) study consists of Caucasian women, all of whom were aged 75 years, who were randomly selected from the Malmö city files between 1995 and 1999. Full details of the cohort have been previously reported by Gerdhem et al. [6]. For both cohorts, the primary assessment included bone density measurements using DXA and QUS and completion of a comprehensive questionnaire concerning lifestyle factors, health and fitness, socioeconomic status, and nutritional factors. Peripheral whole-blood samples were collected for DNA analysis. Participants of both cohorts gave informed consent, and the studies were approved by the Lund University ethics committee. The participants who were included in the reported analysis comprised n = 1,001 from OPRA and n = 993 from PEAK-25, for whom a BMD measurement was available at either the hip or spine and who also had a BMP2 genotype result.

Bone Mineral Density

Measurements were obtained for lumbar spine (LS-BMD) and femoral neck (FN-BMD) using a Lunar (Madison, WI) Prodigy in the PEAK-25 cohort and a Lunar DPX-L scanner in the OPRA cohort. Calibrations were performed using a phantom supplied by the manufacturer. In this analysis, BMD measurements were available at the lumbar spine for 991 and at the femoral neck for 989 women from the PEAK-25 study. From the OPRA study, BMD measurements were available for 940 and 919 women at the hip and spine, respectively.

Ultrasound

QUS measurements of the calcaneus were performed using a Lunar Achilles system. Results were obtained for speed of sound (SOS), broadband ultrasound attenuation (BUA), and stiffness index (SI), a derivative of BUA and SOS. QUS measurements were available for 832 (78.6%) women from the PEAK-25 study and 854 (81.8%) women from the OPRA study.

Fracture Ascertainment

Data on self-reported fracture history in the OPRA cohort were gathered at the baseline visit. For the purpose of analysis, women were categorized as (1) those having sustained at least one fracture during their lifetime and, to exclude fractures sustained at younger ages and less likely therefore to be osteoporotic, (2) those with at least one fracture sustained after the age of 50 [6].

BMP2 SNPs and Genotyping

BMP2, on chromosome 20p12.3, consists of three exons spanning approximately 10.6 kb. In addition to the SNP identified by Styrksdottir et al. [16], a T/G change in codon 37 resulting in a serine to alanine amino acid change (rs2273073), we selected three SNPs from Ensembl (http://www.ensembl.org/) to encompass the gene and its surroundings (Fig. 1). At that time, data on tagging SNPs were not yet available. On that basis, SNPs were selected on the following criteria: relative position to the gene, minor allele frequency in Caucasian populations, and distance between the polymorphisms. The first, rs235710, is a C/T base change located 10 kb upstream of the transcriptional start site; the second, rs235767, is a T/G base change located in intron 2; and the third and final SNP, rs235754, is a C/G base change located 8 kb downstream of the gene. The polymorphisms were all genotyped by the 5′ nuclease Taqman® assay (Applied Biosystems, Foster City, CA) in a 5-μL reaction volume on 384-well plates. Assay identification numbers from Applied Biosystems were rs2273073 (C_16177762_10) and rs235754 (C_2513518_10). The probe and primer sequences for SNP rs235710 were as follows: forward primer ACTTGCCCATTCATGTTATTTTCTATTAGC, reverse primer CTCTTCCAATCCAGGACACCAG, probe 1 VIC-TCGTATATGTGTAGGACTTT, probe 2 FAMTTATCGTATATGTATAGGACTTT. Those for SNP rs235767 were as follows: forward primer GATTGATATAGTATAGATTATTAAATTAGGACTTCACTTTTGGA, reverse primer TGGCTGCAAAAATGTGGGAAAG, probe 1 VIC-AATGATATCTGACCTTTTC, probe 2 FAM-TGATATCTGAACTTTTC. The cycling parameters were 50°C for 2 minutes, 10 minutes at 95°C, followed by 40 cycles of 15 seconds at 95°C and 1 minute at 60°C. Genotypes were detected by an end point read using an ABI Prism 7900HT sequence detector (Applied Biosystems). All samples were genotyped blind, with 1% routinely repeated to ensure agreement of results. Additional checks were performed for the Ser37Ala polymorphism since it appeared to violate Hardy-Weinberg equilibrium (HWE) in the PEAK-25 cohort. An additional 3% of all PEAK-25 samples and 4.5% of all OPRA samples were repeated by Taqman, with 100% agreement between results in those samples for which two results were available. Furthermore, individuals from the PEAK-25 cohort who were homozygous for the rare Ala/Ala genotype and a random selection of samples from both populations were sequenced, with complete concordance of the results.
https://static-content.springer.com/image/art%3A10.1007%2Fs00223-007-9054-9/MediaObjects/223_2007_9054_Fig1_HTML.gif
Fig. 1

Genomic organization of the BMP2 gene and location of SNPs. Exons are shown in black in relation to the location of the polymorphisms studied. The haplotypes, formed from polymorphisms rs2273073 and rs235767 are shown below pair-wise linkage disequilibrium values between all four polymorphisms represented by D′ values in the table

Statistics

Statistical analyses were performed using Statistica version 7.0 (Statsoft, Tulsa, OK) and SPSS version 14 (SPSS Inc., Chicago, IL) packages. HWE was calculated by the chi-squared test. Linkage disequilibrium (LD) between SNPs (Lewontin’s D′) was calculated using the EH and 2BY2 programs (http://linkage.rockefeller.edu/soft/) and verified by the Haploview program (http://www.broad.mit.edu/mpg/haploview/), which also provided the square of the correlation coefficient (r2) values for each SNP pair. The haplotypes defined by the two SNPs in LD (rs235767 and rs2273073) were determined by indirect haplotyping using the program PHASE version 2.02 (http://stephenslab.uchicago.edu/software.html) [22].

Haplotype analyses were performed according to whether individuals possessed zero, one, or two copies of the haplotype allele. Genotype- and haplotype-specific differences in BMD were investigated using general linear model analysis of variance, correcting for confounding factors including body mass index (BMI), estrogen use, years since menarche or menopause, and smoking but not age since all individuals within each cohort were of the same age (25.5 years, range 25.03–25.99, in PEAK-25 and 75.2 years, range 75.03–75.99, in OPRA). Comparison between groups was by Scheffe’s post hoc test. Predictors of fracture were identified by logistic regression analysis. The primary power analysis for the studies was based on BMD, with an estimated power of 80% to detect differences of 0.14 SD units in BMD, between genotypes occurring at frequencies of 4% or less. Although we performed multiple statistical tests (four polymorphisms, bone mass measurements at two sites, and ultrasound parameters), these factors are not independent of each other. Therefore, taking into consideration the extent of LD between the SNPs, the significant correlation between BMD measurements at the spine and femoral neck (r = 0.63), and the significant correlation between ultrasound parameters (r = 0.64), a total of 7.7 independent tests were performed. Applying a Bonferroni correction using this figure, the adjusted level of significance was therefore P = 0.007, equivalent to P = 0.05 taking into account multiple testing [23].

Results

The baseline clinical characteristics of the two study populations are shown in Table 1. BMI was 13.9% lower in the younger women, accounted for by their 4% greater height and 5% decreased weight. Among the PEAK-25 cohort, 26% were current smokers compared to 13.8% in the OPRA cohort. Age at menarche was also lower in the younger population.
Table 1

Baseline characteristics of the young PEAK–25 and elderly OPRA study populations

 

PEAK-25 (n = 993)

OPRA (n = 1,001)

Age (years)

25.5 ± 0.2

75.2 ± 0.1

Height (cm)

167.6 ± 6.1

160.6 ± 5.8

Weight (kg)

64.6 ± 11.0

67.8 ± 11.5

BMI (kg/m2)

22.9 ± 3.7

26.3 ± 4.2

Menarche (age, years)

12.7 ± 1.3

14.0 ± 1.4

Menopause (age, years)

n/a

49.1 ± 4.9

HRT/contraceptive usea

626 (63%)

17 (1.7%)

Current smokers

258 (26%)

138 (13.8%)

BMD

  

  Spine (g/cm2)

1.239 ± 0.13

0.992 ± 0.19

  Femoral neck (g/cm2)

1.054 ± 0.12

0.748 ± 0.13

QUS

  BUA (dB/mHz)

1,17.6 ± 10.7

99.1 ± 10.0

  SOS (m/s)

1,575 ± 32.5

1,502 ± 27.8

  SI

99.2 ± 14.7

66.9 ± 13.0

Fracture

  Any site/age

n/a

434 (43.4%)

  After age 50

n/a

361 (36.1%)

  At osteoporotic site

n/a

227 (22.7%)

aHRT, hormone replacement therapy, currently or previously

All genotype frequencies were in apparent HWE (Table 2) with the exception of rs2273073 in the PEAK-25 cohort (χ= 7.7, P = 0.005). This could be accounted for, however, by the low frequency of the rare allele, such that the Ala/Ala genotype was observed in only two individuals. Importantly, the minor allele frequency of 1.9% in both the young and elderly cohorts of women is not significantly different from that reported in the control population of the original Icelandic study (0.8%) or from other northern European Caucasian populations (1–2.5%). Only rs235767 differed significantly between the two cohorts, with the young population having an underrepresentation of the A allele (χ= 5.9, P = 0.05). Neither the OPRA nor the PEAK-25 cohorts displayed any genotype-related differences in the general clinical characteristics.
Table 2

Distribution within and between cohorts and BMD values in relation to BMP2 genotype

 

Cohort distribution and HWE

PEAK-25

OPRA

PEAK-25

OPRA

Between cohorts

LS-BMD (g/cm2)

FN-BMD (g/cm2)

LS-BMD (g/cm2)

FN-BMD (g/cm2)

rs235710 (G/A)

  GG

279

276

 

1.244 ± 0.03

1.059 ± 0.13

0.999 ± 0.19

0.744 ± 0.13

  GA

494

480

P = 0.40

1.235 ± 0.14

1.052 ± 0.12

0.990 ± 0.20

0.751 ± 0.13

  AA

214

241

 

1.244 ± 0.13

1.058 ± 0.13

0.981 ± 0.19

0.746 ± 0.13

  Total

987*

997*

 

P = 0.69

P = 0.75

P = 0.25

P = 0.89

rs2373073 (T/G)

  TT

955

970

 

1.238 ± 0.13

1.055 ± 0.12

0.992 ± 0.19

0.748 ± 0.13

  TG

34

30

P = 0.27

1.247 ± 0.14

1.050 ± 0.13

0.928 ± 0.16

0.713 ± 0.11

  GG

2

0

 

1.293 ± 0.003

0.990 ± 0.06

0.990 ± 0.19

 

  Total

993**

1,000*

 

P = 0.88

P = 0.56

P = 0.11

P = 0.22

rs235767 (A/C)

  AA

355

407

 

1.239 ± 0.13

1.054 ± 0.12

0.976 ± 0.19

0.738 ± 0.13

  AC

498

450

P = 0.05

1.239 ± 0.13

1.057 ± 0.12

0.998 ± 0.19

0.750 ± 0.13

  CC

142

144

 

1.235 ± 0.13

1.043 ± 0.12

1.006 ± 0.21

0.764 ± 0.12

  Total

993*

1,001*

 

P = 0.99

P = 0.64

P = 0.40

P = 0.32

rs235754 (G/C)

  GG

264

288

 

1.241 ± 0.12

1.049 ± 0.12

0.980 ± 0.21

0.755 ± 0.13

  GC

495

498

P = 0.38

1.236 ± 0.12

1.054 ± 0.12

0.998 ± 0.18

0.749 ± 0.12

  CC

231

211

 

1.239 ± 0.14

1.057 ± 0.12

0.984 ± 0.19

0.732 ± 0.13

  Total

990*

997*

 

P = 0.72

P = 0.84

P = 0.72

P = 0.28

Values are least square means ± SD adjusted for BMI, years since menarche or menopause, smoking, contraceptive pill use (young women only)

*P ≥ 0.1, ** P < 0.05

LD was detected only between SNPs Ser37Ala (rs2273073) and rs235767 (T/G), although given the low r2 value, this could be considered weak LD (D′ = 0.91 and r= 0.02). PHASE analysis identified two common haplotypes derived from these SNPs, accounting for >98% of alleles at the BMP2 locus: haplotype 1, Ser/T (61.5%), and haplotype 2, Ser/G (36.5%). The rare alanine variant occurred at very low frequency, as haplotype 3, Ala/T, only in the PEAK-25 cohort (0.05%) and as haplotype 4, Ala/G (1.5% and 1.8%, respectively), in the PEAK-25 and OPRA cohorts.

There was no association between BMP2 genotype or derived haplotypes and bone density at the femoral neck or lumbar spine in either cohort (Table 2). The analyses remained essentially the same when correction for confounding factors was considered. No relationship between ultrasound properties of the calcaneus and BMP2 genotype were detected among the elderly women; however, an association with BMP2 genotype was observed in the PEAK-25 population. As shown in Table 3, SNP rs235710 was nominally associated with SI (P = 0.02, P = 0.05 after correction for confounding factors, including BMD) and with BUA (P = 0.02, P = 0.04 after correction). A trend with SOS was also observed (P = 0.07 and P = 0.1). In all instances, the highest values were observed in individuals homozygous for the minor A allele. Stronger genotype-related differences in ultrasound parameters were observed with rs235754, located 3′ of BMP2. The strongest associations were with SOS (P = 0.01) and SI (P = 0.01). Even after correction for confounders (BMI, contraceptive pill use, years since menarche, and smoking), the relationships persisted (P = 0.003 and P = 0.002), falling within the adjusted threshold of significance (P = 0.007). There was no evidence of a gene dose effect, and heterozygotes had the lowest values. Combining the heterozygotes and variant homozygotes in the analysis identified carriers of the variant G allele to have the lowest SOS values (1,572.8 compared to 1,579.8, P = 0.001). No differences were observed with the Ser37Ala genotype or alleles, although haplotype 2 (formed together with rs235767 and carrying the common Ser allele) was nominally related to BUA (P = 0.04) but only under a recessive model. Individuals with two copies of this haplotype had the lowest values.
Table 3

Association of BMP2 SNPs with QUS parameters of bone in the PEAK-25 cohort

 

BUA (dB/mHz)

SOS (m/second)

SI

rs235710 (G/A)

  GG

116.7 ± 12.1

1,573.8 ± 33.3

98.3 ± 15.9

  GA

117.4 ± 9.8

1,573.4 ± 31.3

98.6 ± 13.9

  AA

119.4 ± 10.5

1,579.7 ± 34.4

101.7 ± 15.2

 

P = 0.037

P = 0.13

P = 0.049

rs2373073 (T/G)

  TT

117.4 ± 10.6

1,574.8 ± 32.7

99.1 ± 14.8

  TG

119.3 ± 11.7

1,571.6 ± 26.5

99.4 ± 14.3

  GG

124.3 ± 0.4

1,586.7 ± 7.9

106.9 ± 2.4

 

P = 0.35

P = 0.62

P = 0.48

rs235767 (A/C)

  AA

118.6 ± 10.6

1,576.8 ± 33.3

100.4 ± 15.1

  AC

116.6 ± 10.7

1,572.8 ± 32.3

97.9 ± 14.5

  CC

117 ± 10.9

1,575.8 ± 31.3

99.5 ± 14.7

 

P = 0.13

P = 0.76

P = 0.34

rs235754 (G/C)

  GG

118.6 ± 10.6

1,579.9 ± 33.4

101.3 ± 14.9

  GC

116.6 ± 10.5

1,571.3 ± 30.9

97.5 ± 14.2

  CC

118.3 ± 10.9

1,575.9 ± 34.3

99.9 ± 15.5

 

P = 0.016

P = 0.003

P = 0.002

Values are least square means ± SD adjusted for BMI, contraceptive pill use, years since menarche, smoking, and FN-BMD, assuming a codominant model

Fractures were studied only in the OPRA cohort. Genotypes were available for 434 women who had sustained a fracture, and of these, 361 (83%) fractured after the age of 50 years. Logistic regression analysis including the variables genotype, BMI, years since menopause, current smoker status, and BMD showed that LS-BMD was the single most important predictor of fracture ever sustained (P = 0.002) and of fracture after the age of 50 years (P = 0.001). Variation at the BMP2 locus did not independently predict fracture in this population.

Discussion

A recent genomewide screen indicated linkage with fracture and BMD at the 20p12 region, which contains BMP2 as the only candidate gene involved in bone regulation. In vivo and in vitro experiments implicate BMP2 as a potent osteotropic factor promoting bone [24] and cartilage [25] formation and, in animal and human studies, fracture healing [2632]. The original association study suggested a relationship with reduced BMD and fracture in Icelandic and Danish populations, but like many association studies of promising osteoporosis candidate genes, follow-up studies have been unable to confirm the observed associations with bone mass [15], possibly as a result of overestimation of the effect size in the initial study [33]. Like the studies of Medici et al. [19] and Ickikawa et al. [18], we were unable to detect any relationship between polymorphisms of BMP2 and BMD at the lumbar spine or the femoral neck in either the young or elderly women, although both of our cohorts had sufficient power to detect such relationships. Neither did we find any evidence of a relationship with fracture in the elderly women. Instead, we present indications of a relationship between BMP2 and bone ultrasound properties and raise the possibility of age- and site-specific effects for this gene.

Data modeled in twins indicate that both specific and shared genetic factors act on different bone phenotypes [2, 34]. Although, to our knowledge, this study is the only one of ultrasound parameters of bone and SNPs in the BMP2 region, many of the candidate genes studied to date contribute to both bone mass and bone quality. This may explain the partially BMD-independent associations commonly observed with the clinical phenotype fracture [3540]. Our observation is suggestive that variation in BMP2 contributes to aspects of bone other than density. We are confident that the BMP2 polymorphisms studied are not simply markers of bone mass measured by ultrasound since the association was still apparent after correction for BMD, indicating an association that exists independent of bone density. While we acknowledge the issues related to multiple testing, we believe we have identified a relationship with ultrasound parameters among young women that is apparent even after correction for the number of tests performed. Although the magnitude of the effect of the BMP2 polymorphism rs235754 is relatively modest (SOS 4 m/second and stiffness 1.2 units between homozygotes), this is not unexpected in a multigenic disease. Utilizing other methods of measuring bone microarchitecture such as high-resolution peripheral quantitative computed tomography would help to clarify the nature of the relationship, while additionally studying the rate of bone loss sustained at the menopause might provide a clearer picture of the role which BMP2 plays in relation to osteoporosis phenotypes.

There are numerous studies reporting that the genes regulating BMD act in a gender-specific, age-specific, and site-specific manner both in animals and in humans and furthermore that different loci may act at different skeletal sites [4144]. Even within a given skeletal site differences between cortical and trabecular bone exist [42]. This has clear implications in terms of identifying the genes controlling bone morphology at a given site and, indeed, the selection of measurement sites to accurately diagnose BMD status. We hypothesize that BMP2 variants may be more important at peripheral compared to axial sites, especially during the period following accrual of maximum bone density. This may reflect differences in growth patterns at different skeletal sites and between trabecular and cortical bone. Although a direct comparison is difficult to make since different phenotypes are measured, our observation of an association between BMP2 and the calcaneus, as measured by QUS, is to some extent supported by data from Choi et al. [17], who identified a relationship between BMP2 genotype and DXA-measured calcaneal BMD in young Korean men. Similar to our observations, Ichikawa et al. [18], in their study of young American men and women, found no association with BMD measured by DXA at the hip or spine; but again, the study design is not strictly comparable to ours and should be considered cautiously. To the best of our knowledge, there are no cohorts comparable in terms of sample size, genetic background, and phenotypes measured in which to replicate the ultrasound findings in young women; and there have as yet been insufficient studies using meta-analysis to elucidate the effect of BMP2 polymorphisms on variation in bone mass in the elderly.

The inconsistent results observed in studies to date could be attributed to differences in the polymorphisms studied, bone parameters measured, and population heterogeneity. While some differences in the genetic background of Swedish and Icelandic women would be expected, heterogeneity between Swedish and Danish women should be minimal and, therefore, is unlikely to explain the fact that we could not replicate the original findings in this southern Swedish cohort of elderly women. It is interesting to note that the Ser37Ala variant was not associated with QUS except as part of a haplotype containing the common allele along with the intron 2 SNP (rs235767) and, even then, only weakly while the strongest association with QUS parameters observed was with an SNP located 8 kb downstream of the BMP2 transcribed region. Elements within the 3′-untranslated region can modulate transcriptional regulation [45], and recently published data [46] indicate that variation within this region may be important for control of BMP2 gene expression. Further studies are necessary to determine if, individually or in combination, polymorphisms in this region influence susceptibility to osteoporosis.

It is, of course, difficult to fully elucidate the role of a single gene and its effect on bone phenotypes in isolation, when its regulation is controlled by the interplay of complex signaling pathways. BMP2 has been implicated in steroid receptor signaling pathways, especially the ERα-dependent response pathway; and there is evidence of both inhibitory and stimulatory effects on BMP2 gene and protein expression [47, 48]. Furthermore, the expression profile varies throughout life, from fetal development to adulthood to the menopause and beyond, and may be influenced by nutritional factors and physical activity, not to mention gene-gene and gene-environment interactions, of which, at this time, we know little. An association with QUS was not evident in the elderly women in our study, possibly because the gene effect was overshadowed by environmental factors.

Our study has a number of important strengths: the homogeneous ethnicity of both cohorts; the identical age of the participants within each cohort (exactly 25 or 75 years), providing large age-specific samples at peak bone mass and at late postmenopause; the large amount of clinical information on the participants; and lastly, the inclusion of additional SNPs across the BMP2 locus. We considered it important to cover genetic variation in both the 5′ and 3′ regions of the gene, hence our selection of SNPs, which cover 10 kb upstream of the gene and 8 kb downstream. At the time when genotyping was completed, no functional SNPs apart from Ser37Ala had been identified. Five common haplotypes estimated from the selected SNPs represent 76% of the genetic variation in the locus. Had we used tag SNPs from the conserved haplotype block covering the gene, we would have captured an additional 15%. Even the most extensive coverage of BMP2 to date utilizing tag SNPs incorporating this haplotype block found no association with individual SNPs or haplotypes [18], although a large-scale family-based association analysis identified an SNP (rs1980499) in this proximal promoter region suggestive for BMD [49]. Our 3′-most SNP (rs235754) is located less than 2 kb from an SNP (rs235753) that has been positively associated with ultradistal osteoporosis [49], lending weight to the likelihood of a true association within this region. Despite this, caution should be exercised and the possibility recognized that in studies such as these the chosen SNPs may not fully capture all genetic information in the region and as yet unidentified linked polymorphisms associated with the disease phenotype.

The participation rate is a potential limitation to our study, particularly for the young women. In population-based studies, this may induce selection bias; however, a participation rate of 44% is still high compared to most population-based studies, and 65%, as in the OPRA cohort, can be regarded as very high. In the analysis of the initial response of the OPRA cohort, a slightly higher number of responders compared to nonresponders had had a fracture, but 1-year mortality was higher in nonresponders, suggesting that we included women who are somewhat healthier and able to attend the research unit (data not shown).

We conclude on the basis of this preliminary research in 25-year-old and 75-year-old women that, in this Scandinavian population, variations in the BMP2 gene may play a role in aspects of bone quality, especially in young adult women, but not of bone density in women of either age or fracture risk in the elderly. Given the importance of BMP2 to bone regulation, further studies of the effects of variation in this gene are merited.

Acknowledgements

Support for the study was received from the Swedish Research Council, Greta and Johan Kock Foundation, A. Påhlsson Foundation, A. Osterlund Foundation, Swedish Center for Sports Medicine Research, Malmö University Hospital Research Foundation, and Swedish Medical Society. We thank Dr. Alison Stewart (University of Aberdeen) for advice on QUS, Lisa Jansson for genotyping support, the research nurses at the Clinical and Molecular Osteoporosis Research Unit, and all the women who kindly participated in the study.

Copyright information

© Springer Science+Business Media, LLC 2007