COL1A1 Sp1 Polymorphism Predicts Perimenopausal and Early Postmenopausal Spinal Bone Loss


  • Helen M. Macdonald,

    Corresponding author
    1. Department of Medicine and Therapeutics, University of Aberdeen, Foresterhill, Aberdeen, United Kingdom
    • Address reprint requests to: Helen Macdonald, BSc., MSc, Ph.D., Osteoporosis Research Unit, Victoria Pavilion, Woolmanhill Hospital, Aberdeen AB25 1LD, UK
    Search for more papers by this author
  • Fiona A. McGuigan,

    1. Department of Medicine and Therapeutics, University of Aberdeen, Foresterhill, Aberdeen, United Kingdom
    Search for more papers by this author
  • Susan A. New,

    1. Center for Nutrition and Food Safety, School of Biomedical and Life Sciences, University of Surrey, Guildford, Surrey, United Kingdom
    Search for more papers by this author
  • Marion K. Campbell,

    1. Health Services Research Unit, University of Aberdeen, Foresterhill, Aberdeen, United Kingdom
    Search for more papers by this author
  • Michael H. N. Golden,

    1. Department of Medicine and Therapeutics, University of Aberdeen, Foresterhill, Aberdeen, United Kingdom
    Search for more papers by this author
  • Stuart H. Ralston,

    1. Department of Medicine and Therapeutics, University of Aberdeen, Foresterhill, Aberdeen, United Kingdom
    Search for more papers by this author
  • David M. Reid

    1. Department of Medicine and Therapeutics, University of Aberdeen, Foresterhill, Aberdeen, United Kingdom
    Search for more papers by this author


Genetic factors play an important role in the pathogenesis of osteoporosis but the genes that determine susceptibility to poor bone health are defined incompletely. Previous work has shown that a polymorphism that affects an Sp1 binding site in the COL1A1 gene is associated with reduced bone mineral density (BMD) and an increased risk of osteoporotic fracture in several populations. Data from cross-sectional studies have indicated that COL1A1 Sp1 alleles also may be associated with increased rates of bone loss with age, but longitudinal studies, which have examined bone loss in relation to COL1A1> genotype, have yielded conflicting results. In this study, we examined the relationship between COL1A1 Sp1 alleles and early postmenopausal bone loss measured by dual-energy X-ray absorptiometry (DXA) in a population-based cohort of 734 Scottish women who were followed up over a 5- to 7-year period. The distribution of genotypes was as expected in a white population with 484 “SS” homozygotes (65.9%); 225 “Ss” heterozygotes (30.7%), and 25 “ss” homozygotes (3.4%). Women taking hormone-replacement therapy (HRT; n = 239) had considerably reduced rates of bone loss at the spine (−0.40 ± 0.06%/year) and hip (−0.56 ± 0.06%/year) when compared with non-HRT users (n = 352; spine, −1.36 ± 0.06%/year; hip, −1.21 ± 0.05%/year; p < 0.001 for both sites). There was no significant difference in baseline BMD values at the lumbar spine (LS) or femoral neck (FN) between genotypes or in the rates of bone loss between genotypes in HRT users. However, in non-HRT users (n = 352), we found that ss homozygotes (n = 12) lost significantly more bone at the lumbar site than the other genotype groups in which ss = −2.26 ± 0.31%/year compared with SS = −1.38 ± 0.07%/year and Ss = −1.22 ± 0.10%/year (p = 0.004; analysis of variance [ANOVA]) and a similar trend was observed at the FN in which ss = −1.78 ± 0.19%/year compared with SS = −1.21 ± 0.06%/year and Ss = −1.16 ± 0.08%/year (p = 0.06; ANOVA). The differences in spine BMD loss remained significant after correcting for confounding factors. Stepwise multiple regression analysis showed that COL1A1 genotype independently accounted for a further 3.0% of the variation in spine BMD change after age (4.0%), weight (5.0%), and baseline BMD (2.8%). We conclude that women homozygous for the Sp1 polymorphism are at significantly increased risk of excess rates of bone loss at the spine, but this effect may be nullified by the use of HRT.


OSTEOPOROSIS IS a common condition characterized by reduced bone mass and an increased risk of fragility fractures.(1) Low bone mineral density (BMD) values in older individuals are one of the most important predictors of osteoporotic fracture risk2-4) and these may arise because of reduced peak bone mass attainment during childhood and adolescence or because of increased bone loss with aging. There is overwhelming evidence to show that genetic factors play an important role in determining peak bone mass(5, 6) but it is less clear whether genetic factors are important in predicting rates of bone loss. For example, although significant genetic effects on perimenopausal bone loss have been reported in female twins,(7) a similar study in aging male twins showed no such effect.(8) At present, there are only limited data showing which genetic factors might predispose to rapid bone loss around the time of the menopause. However, if genetic determinants of increased bone loss were identified, this could be of considerable clinical value in targeting individuals at risk for initiation of therapy aimed at preventing or reducing loss of bone.

Now, a large number of studies have been performed looking at the association between polymorphisms of candidate genes and BMD. Allelic variants of the vitamin D receptor have been associated with bone loss in several studies, and this association has been found by some investigators to be dependent on dietary calcium intake.9-11) There is evidence to suggest that polymorphisms of the estrogen receptor α also may be associated with bone loss12-14) and response to estrogen-replacement therapy(15) in some populations. Another widely studied candidate gene is the COL1A1 gene, which encodes the α1-chain of type 1 collagen. There is an increasing body of evidence to suggest that a polymorphism affecting an Sp1 binding site in COL1A1 is related to reduced BMD and an increased risk of osteoporotic fracture.16-25) Data from a large cross-sectional study also indicate that COL1A1 genotype may predict age-related bone loss, because the genotype-specific difference in BMD is higher in individuals with increasing age.(22) However, longitudinal studies of bone loss in relation to COL1A1 alleles have yielded conflicting results.(26, 27) In one study from the United States, Harris et al. reported that COL1A1 alleles predicted bone loss in older men and women,(26) whereas Heegaard et al. found no relationship between COL1A1 alleles in Danish women followed over an 18-year period.(27) To clarify the relationship between COL1A1 alleles and bone loss and to examine whether any relationship was altered by the use of hormone-replacement therapy (HRT), we have studied the association between the COL1A1 Sp1 polymorphism and early postmenopausal axial bone loss over a 5- to 7-year period in a population-based cohort of 735 white women living in a city northeast of Scotland.



Between 1990 and 1993, a subset of 1064 healthy, mainly premenopausal women who were not suffering from any condition or taking any medication likely to affect their bone metabolism were selected to participate in a study of dietary determinants of osteoporosis.(28, 29) This study was part of a larger population-based screening program for osteoporotic fracture risk involving over 5000 women drawn at random using Community Health Index (CHI) records from a 25-mile radius of Aberdeen, a city northeast of Scotland, with a population of 250,000.(30, 31) All participants underwent bone densitometry and completed a food frequency (FFQ) and risk factor questionnaire. The average age of participants at the start of the study was 48 ± 1.5 years and most were premenopausal. All women were invited to undergo further assessment between 1997 and 1999. Of the 907 women who reattended for evaluation (85% response rate), 738 (81%) consented to give a blood sample for DNA extraction. Four of these women were excluded from the study; 2 because they were taking bisphosphonates, 1 because she was wheelchair bound, and another because she had an unusually high calcium intake (2101 mg dietary calcium a day plus 800 mg calcium supplement), leaving a final study population of 734 women. Written informed consent was obtained for all the women and the study was approved by the Grampian Joint Ethical Committee.

Bone mineral densitometry

The BMD measurements of the left proximal femur (the femoral neck [FN]) and lumbar spine (LS; L2-L4) were performed by dual-energy X-ray absorptiometry (DXA) using one of two Norland XR26 or XR36 densitometers (Norland Corp., Fort Atkinson, WI, USA). Of the 734 women whose genotype was determined, 679 (92.5%) were measured (at baseline and follow-up) on the XR26. The remaining 55 women (35 SS, 14 Ss, and 6 ss) were measured on the XR26 at baseline and the XR36 at follow-up. Calibration of the machines was performed daily, and quality assurance was checked by measuring the manufacturer's LS phantom at daily intervals and a Hologic spine phantom at weekly intervals. The in vivo precision in our hands for the XR36 is 1.2% for the LS and 2.3% for the FN. Corresponding values for the XR26 are 1.95% and 2.31% for the LS and FN, respectively. A comparison between the XR26 and XR36 was made using 50 phantom spine measurements from each machine. It appeared that the XR36 (mean ± SD, 0.7963 ± 0.0068 g/cm2) was giving slightly higher measurements than the XR26 (0.7771 ± 0.0054 g/cm2). Therefore, BMD measurements from the XR36 were corrected by dividing by 1.02478509, the mean difference between the two machines.

Menopausal status and HRT use

The majority (90%) of women were premenopausal at the baseline visit and none had taken HRT. At the follow-up visit only 6% were menstruating regularly, 10% were perimenopausal (irregular periods), and 39% were postmenopausal, that is, with no menstrual loss for 6 months. None of the women described previously had ever taken HRT. The remaining 45% had taken HRT since the baseline visit. Of these, one-third had stopped taking HRT and are defined as past HRT users and two-thirds are still taking HRT and are defined as present HRT users. Of those currently taking HRT, 229 of 239 women gave precise details on dates of usage; 9.6% women had taken HRT for <1 year and for the rest, the median time of usage was 4.4 years (mean, 4.5 ± 2.1 years). For the past HRT users, 86 of 98 women gave dates of usage and 58% women had used HRT for <1 year. The remaining women had taken HRT for a median of 3 years (mean, 3.3 ± 1.6 years)

COL1A1 genotyping

Genotyping for the COL1A1 Sp1 polymorphism was carried out on DNA extracted from peripheral blood using standard techniques as previously described.(25)

Anthropometric measurements

The women were weighed on both occasions wearing light clothing and no shoes on a set of balance scales (Seca, Hamburg, Germany) calibrated to 0.05 kg. Height and sitting height were measured with a stadiometer (Holtain, Ltd., Crymych, UK).

Dietary intake and physical activity

Usual dietary intake was assessed by the same FFQ (validated using 7-day weighed records and biochemical markers of antioxidant status) that had been used in the baseline studies.(28, 29) Because the FFQ is used to assess complete dietary intake, dietary calcium can be measured as crude intake or adjusted for energy intake (an important confounder in diet and bone health studies). Physical activity levels (PALs) were obtained using the same questions as used for the Scottish Heart Health Study.(32) The PAL is calculated from the numbers of hours in a 24-h period doing heavy, moderate, or light activities and how many hours are spent sleeping or resting in bed. These questions were asked separately for working and nonworking days. Normally, PAL is defined as the ratio of overall daily energy expenditure to basal metabolic rate. Seated work with discretion/requirement to move around with little or no strenuous activity is consistent with a PAL of 1.6-1.7 and standing work (e.g., housewife and shop assistant) is consistent with PAL between 1.8 and 1.9.(33)

Statistical methods

The statistical package SPSS version 9.0 (SPSS, Inc., Chicago, IL, USA) was used for all statistical analysis. Differences in BMD among the genotypes were tested using one-way analysis of variance (ANOVA) with the Scheffe test for multiple comparisons where appropriate and adjustment for confounding variables was made using analysis of covariance (ANCOVA). Multiple regression was undertaken to identify independent predictors of bone loss. For categorical variables such as COL1A1 and menopausal/HRT status, dummy variables were constructed for use in the multiple regression analysis. For example, for the three-category COL1A1 variable, two dummy variables had to be constructed for inclusion in the multiple regression. For the first dummy variable, women with Ss were coded as 1 and all remaining women were coded as 0 and for the second dummy variable, ss women were coded as 1, with the remaining women again coded as 0.

To examine whether the effect of COL1A1 status differed depending on HRT/menopausal status, we also analyzed the effect of COL1A1 within each HRT/menopausal status group separately.


Genotype distributions in the 734 women in relation to other relevant characteristics are shown in Table 1. The groups were similar in terms of weight, body mass index (BMI), calcium intake, and BMD values at the spine and hip. There were fewer smokers in the ss genotype group and more women had been educated to University standard in this group, but the differences were not statistically significant.

Table Table 1.. Subject Characteristics at Follow-Up Visit (Unless Otherwise Indicated) by COL1A1 Genotype
original image

The relationship between COL1A1 genotype and bone loss is shown in Table 2. Past HRT users are included for completeness only but because there were only 2 women with ss in this group, no conclusions can be drawn from these results. When all women were considered together there was a weak but significant relationship between bone loss at the LS and COL1A1 genotype such that the ss group lost significantly more bone than the other genotype groups (Scheffe posttest for ANOVA, p = 0.036 for SS vs. ss and p = 0.026 for Ss vs. ss). Then, the relationship between COL1A1 genotype and bone loss was analyzed for current HRT users compared with women who had never used HRT. This showed a highly significant difference in rate of spine bone loss in non-HRT users according to COL1A1 genotype but no significant difference in HRT users. As expected, the rates of bone loss in HRT users was significantly less than for non-HRT users in all genotype groups (p < 0.001; Table 2).

Table Table 2.. Changes of BMD in Relation to COL1A1 Genotype (Mean ± SEM)
original image

The differences between genotypes remained significant after adjustment for confounding factors including age, height, weight, weight change, current smoking habits, baseline BMD, and PAL, and also after further adjustment for menopausal status and HRT use for the full group. Similar trends were seen at the FN but even in non-HRT users, these failed to reach statistical significance (p = 0.066).

Multiple regression was undertaken to identify independent predictors of bone loss (from age, weight, height, current smoking, annual weight change, baseline BMD at the appropriate site, PAL, PAL difference, menopausal status/HRT use, and COL1A1 genotypes) for all women (Table 3). Menopausal status and HRT use accounted for most of the variation. The regression for non-HRT users only showed that weight accounted for 5.0% of the variation in LS BMD change, age accounted for an additional 4.0%, baseline LS BMD accounted for a further 2.8%, and COL1A1 accounted for a further 3.0% of the variation in LS BMD change (Table 4). Again, similar but nonsignificant trends were seen at the FN where COL1A1 genotype was responsible for only 0.7% of the variability in change of BMD. At this site, only age (3.3%), smoking (1.0%), and annual weight change (2.3%) were the significant predictors. None of the foregoing variables were identified as independent predictors of LS BMD change and FN BMD change in current HRT users. Repeating the multiple regression analysis without the correction factor for the XR36 measurements confirmed the significance of COL1A1 s genotype (p = 0.021 for all women and p = 0.009 for the non-HRT subgroup). Also, completely excluding those women who were analyzed on the XR36 gave similar results (p = 0.008 for the whole group and p = 0.026 for the non-HRT subgroup).

Table Table 3.. Results of Multiple Regression Analyses to Identify Independent Predictors of Bone Loss* For All Women (n = 734)
original image
Table Table 4.. Results of Multiple Regression Analyses To Identify Independent Predictors of Bone Loss* For Non-HRT Users Only (n = 352)
original image


The s allele of the COL1A1 Sp1 polymorphism has been associated with reduced BMD and an increased risk of osteoporotic fractures in several studies.16-25) However, other investigators have not found an association between COL1A1 genotype and bone mass or osteoporotic fracture.(34, 35) One of these studies was very small(35) and in the other, COL1A1 genotype distributions in elderly patients with fracture were compared with premenopausal twins as controls.(34) In a large study reported by Uitterlinden et al., the genotype-specific differences in BMD increased markedly with age, raising the possibility that COL1A1 genotype may predict bone loss.(22) This raises the possibility that the strong association that has been observed between COL1A1 alleles and osteoporotic fractures(16, 19, 21, 22, 25) is not mediated by an effect on peak bone mass, but rather by an effect on bone loss. However, previous longitudinal studies on the relationship between the COL1A1 Sp1 binding site polymorphism and bone loss have yielded conflicting results.(26, 27) In a recent study of 243 men and women of mean age 65 years, Harris et al. reported that individuals with the ss genotype had significant accelerated total body bone loss over a 5-year period when compared with the other genotype groups. Although a similar trend was observed at the FN, no association was found between COL1A1 genotype and change in spine BMD.(26) However, this discrepancy might be explained by spinal artifacts in older subjects.(36) In another study of Danish women initially aged 45-54 years, Heegard and colleagues(27) found no significant relationship between change in bone mineral content (BMC) at the forearm either in the initial 2 years after the menopause or over an 18-year follow-up period. BMD was only assessed at the spine and hip over a 6-year period when the women were aged 63-69 years. Again, no significant association was found between the rates of bone loss and the COL1A1 genotype. It is important to note that the number of women investigated was small (n = 133, of whom only 2 women had the ss genotype) and the longitudinal study was performed in only 42% of the original cohort. Our study on 734 individuals is the largest longitudinal study of bone loss in relation to COL1A1 alleles or indeed any other candidate gene.

In agreement with the findings of Harris et al., we found a significant association between the COL1A1 ss genotype and accelerated bone loss, but in our study the effect mainly was observed at the LS (although a similar but nonsignificant trend was observed at the FN). Similarly, the association remained significant after correcting for confounding factors and, overall, the COL1A1 genotype was found to account for 3.0% of the variance in spine bone loss along with body weight (5.0%), age (4.0%), and baseline BMD (2.8%) for women who did not take HRT. The site-specific differences that were observed in relation to genotype and bone loss probably are explained by the fact that early postmenopausal bone loss mainly effects sites that are rich in trabecular bone such as the spine. The power that our study had to detect differences in the generally slower rates of bone loss, which are observed at the hip, therefore was relatively low.

The women who took part in the screening study were selected at random from the population and the subset was selected at baseline to include only healthy women (excluding those with disease or those taking medication that would affect their bone health). Baseline BMD measurements were identical for all three genotypes and we are confident that the findings were not a result of selection bias. Although two different Norland DXA scanners were used at follow-up and phantom measurements indicated a significant difference between measurements with these two scanners, only 7.5% of the population was measured on the second scanner and only they had their BMD results adjusted. Reanalysis either excluding the women who were measured on the second scanner or excluding the correction factor in the multiple linear regression model for LS bone loss rates confirmed the significance of COL1A1 genotype.

The mechanism by which women with the ss genotype have greater bone loss is not clear. Garnero et al.(23) found an association between ss genotype and reduced levels of serum C-terminal extension propeptide of type I collagen (PICP), a marker of bone formation. Furthermore, a finding that the s allele is associated with an abnormal ratio of collagen α1- to α2-chains has been published by Dean et al. in abstract form.(37) Whether the presence of the s allele leads to a reduction in bone formation and accelerated bone loss in the early postmenopausal years is uncertain and clearly further work is required to assess fully the pathogenesis of the greater spinal bone loss in these women.

An important observation to emerge from this study was the fact that women with the unfavorable ss genotype who took HRT were protected against accelerated bone loss. Although all women who were current users of HRT had significantly reduced rates of bone loss compared with non-HRT users, our data suggest that individuals with the ss genotype who are particularly at risk of increased bone loss after the menopause may derive particular benefit from HRT use. From a clinical viewpoint, COL1A1 genotyping may be of value in identifying such individuals and targeting the use of HRT, although such a suggestion needs to be validated in further cohorts of women examined prospectively.


We acknowledge the excellent technical assistance of Grace Taylor, Ewan McLeod, and Lorna Smith for DNA extraction and genotyping and the assistance of Dr. Rob Van't Hof with database management. We also are very grateful to David Grubb of the Rowett Research Unit for his assistance with analysis of the FFQ. These studies were supported in part by a project grant to D.M.R., S.A.N., and M.H.N.G. from the Medical Research Council/Department of Health Nutrition Program Phase II; a cooperative group grant to S.H.R. and D.M.R. from the Medical Research Council; and an Integrated Clinical and Academic Center (ICAC) grant from the Alcohol Research Center (ARC) to D.M.R. and S.H.R. The Health Services Research Unit is funded by the Chief Scientist Office of the Scottish Executive Health Department. Any views expressed are the authors' own.