Interaction Between the Vitamin D Receptor Gene and Collagen Type Iα1 Gene in Susceptibility for Fracture†
Preliminary results from this study were published in abstract form as Uitterlinden et al. Polymorphisms in the vitamin D receptor- and collagen type Iα1 gene predict osteoporotic fracture in women. Am J Hum Genet 1998;63(4):A223 (abstract 1279).
Osteoporosis is a common disease with a strong genetic component. Polymorphisms in the vitamin D receptor (VDR) gene have been implicated in osteoporosis but explain only a small part of the genetic effect on bone mineral density (BMD) while their effect on fractures is still uncertain. Recently, a G to T polymorphism in an Sp1 site in the collagen type Iα1 (COLIA1) gene was found to be associated with reduced BMD and with increased fracture risk. To analyze the combined influence of polymorphisms in the VDR gene and the COLIA1 gene in determining the susceptibility to osteoporotic fracture, we studied 1004 postmenopausal women. The “baT ” VDR haplotype, constructed from three adjacent restriction fragment length polymorphisms, was found to be overrepresented among fracture cases (p = 0.009). This corresponded to an odds ratio (OR) of 1.8 (95% CI, 1.0–3.3) for heterozygous carriers and 2.6 (95%CI, 1.4–5.0) for homozygous carriers of the risk haplotype. The effect was similar for vertebral and nonvertebral fractures and, most importantly, independent of BMD. We observed significant interaction (p = 0.03) between VDR and COLIA1 genotype effects. Fracture risk was not VDR genotype-dependent in the COLIA1 “reference” group (genotype GG) while in the COLIA1 “risk” group (genotypes GT and TT) the risk of fracture was 2.1 (95%CI, 1.0–4.4) for heterozygous and 4.4 (95%CI, 2.0–9.4) for homozygous carriers of the VDR risk haplotype. We conclude that both the VDR and the COLIA1 polymorphisms are genetic markers for osteoporotic fracture in women, independent of BMD. Our data indicate that interlocus interaction is likely to be an important component of osteoporotic fracture risk.
Osteoporosis is a common disease characterized by reduced bone mineral density (BMD), deterioration of bone microarchitecture, and increased fracture risk.(1) Low BMD is an important risk factor for fractures, the clinically most relevant feature of osteoporosis. Different aspects of osteoporosis have been shown to have genetic components. Twin and family studies have suggested that BMD has a strong genetic component(2–6) and is under polygenic control.(7,8) In addition, biochemical markers of bone turnover(9–11) and measurements of bone geometry and structure such as hip axis length and ultrasound measurements of the calcaneus(12) have been shown to have strong genetic components. Finally, maternal history of fracture is a strong risk factor for hip fracture in older women, independent of BMD, suggesting heritability of hip fracture.(13)
Several candidate genes have been analyzed in relation to BMD but the first and most widely studied gene in this respect, the vitamin D receptor (VDR) gene,(14) explains only a small part of the genetic effect on BMD.(15–17) Indeed, in our study population of elderly subjects we previously showed a small effect of VDR genotype on BMD.(18)
Most genetic analyses have focused on BMD as a determinant of fracture risk and not so much on fractures themselves as an endpoint in the analysis. Although a recent study(19) and an early unpublished report(20) suggested VDR genotype to predict osteoporotic fracture, other studies have not found significant associations between VDR polymorphisms and fracture.(21,22) Recently, the T allele of a polymorphism in an Sp1 binding site in the first intron of the COLIA1 gene, encoding the most abundant bone matrix protein, was found to be associated with reduced BMD and, more importantly, also with increased risk of osteoporotic fracture.(23,24) Interestingly, the COLIA1 genotype-dependent fracture risk was largely independent of BMD.(24) This suggests a genetic effect on the pathogenesis of osteoporotic fracture by mechanisms that are, at least in part, independent of an effect on BMD.
Osteoporosis can be considered a complex genetic trait with variants of several genes underlying the genetic determination of the variability of the phenotype. So far, no studies have addressed the possible interaction between osteoporosis candidate genes in relation to the clinically most important endpoint, that is, fractures. The COLIA1 gene and the VDR gene are interesting genes in this respect because the VDR is a transcription factor also regulating the expression of the COLIA1 gene.(25,26) There is now increasing evidence that a VDR allele carrying a particular haplotype of the three 3′ restriction fragment length polymorphisms (RFLPs; baT; haplotype 1) is associated with aberrant expression levels of VDR messenger RNA (mRNA), possibly through changes in mRNA stability.(14,27–29) Genetic variation in the VDR gene can therefore be expected to influence the association of COLIA1 gene variants with osteoporosis. Thus, we first analyzed the relation between fractures and VDR genotype and, second, studied interaction between the VDR and the COLIA1 polymorphisms.
MATERIALS AND METHODS
The Rotterdam Study is a population-based cohort study of 7983 subjects aged 55 years or more, residing in the Ommoord district of the city of Rotterdam in the Netherlands. The study was designed to document the occurrence of disease in the elderly in relation to several potential determinants.(30) A total of 10,275 persons, of whom 9161 (89%) were living independently, were invited to participate in the study in 1991. In the independently living subjects, the overall response rate was 77% for home interview and 71% for examination in a research center, including measurement of anthropometric characteristics, BMD, and blood sampling. The Rotterdam Study was approved by the Medical Ethics Committee of the Erasmus University Medical School and written informed consent was obtained from each subject. The analysis of the association between VDR genotype, COLIA1 genotype, and osteoporotic fracture was performed in a subgroup of women participating in the study. Baseline measurements of BMD were available for 5931 independently living subjects from the study, but 1453 of these were excluded based on age (>80 years); use of a walking aid; diabetes mellitus; or use of diuretic, estrogen, thyroid hormone, or cytostatic drug therapy. From the 4478 remaining subjects, we studied a random sample of 1500 women aged 55–80 years. Anthropometric data, DNA samples, or genotype data for both loci were not available for 481 women, and we excluded women with the rare VDR haplotypes 4 and 5 (n = 15), resulting in a final study group of 1004 women.
Height and weight were measured at the initial examination. BMD (in g/cm2) was determined by dual energy X-ray absorptiometry (Lunar DPX-L densitometer; Lunar Corp., Madison, WI, USA) at the femoral neck and lumbar spine (vertebrae L2-L4) as described elsewhere.(31) Dietary intakes of calcium (mg/day) during the preceding year were assessed by food frequency questionnaire and adjusted for energy intake. Age at menopause and current cigarette smoking were assessed by questionnaire. For 732 women (73%), lateral radiographs of the spine from the fourth thoracic to the fifth lumbar vertebrae were obtained at baseline examination and analyzed for the presence of prevalent vertebral fractures by morphometric analysis as previously described.(32) The occurrence of incident nonvertebral fractures, including hip, wrist, and other fractures, was recorded, confirmed, and classified by a physician over a mean follow-up period of 3.8 years, as described previously.(33) In brief, follow-up for fractures was achieved through a link with the computer systems of the general practitioners of the district and on hospital admission data, covering about 80% of the study population. For all participants not covered by this system, annual checks were performed on the complete medical records of their general practitioners. Reported fractures were verified by retrieval and review of the appropriate discharge reports from the patient record. In total, 49 prevalent vertebral fracture cases and 52 incident nonvertebral fracture cases were recorded (7 hip, 6 upper humerus, 22 wrist, 4 hand, 4 ankle, 3 foot, and 5 other fractures). Four subjects, in whom both a vertebral and a nonvertebral fracture were present, were each counted as 1 fracture case, resulting in 97 cases with one or more fractures.
Determination of COLIA1 and VDR genotypes
Genomic DNA was extracted from peripheral venous blood samples according to standard procedures and the polymorphism in the COLIA1 gene was detected by polymerase chain reaction (PCR) with a mismatched primer that introduces a diallelic restriction site, as previously described.(23) The test discriminates two alleles named S and s, corresponding to nucleotides G and T, respectively, at the first base of the Sp1 binding site in the first intron of the gene for COLIA1. Three anonymous polymorphic restriction enzyme recognition sites at the 3′ end of the VDR gene, that is, for BsmI, ApaI, and TaqI, were assessed in relation to each other by a direct molecular haplotyping PCR procedure that we developed.(18) This allowed us to determine the phase of the alleles at each of the RFLP loci and, as a result, three frequent haplotype alleles are discerned: encoded 1 (baT; frequency 48%), 2 (BAt; frequency 40%), and 3 (bAT; frequency 10%) combining to six genotypes encoded 11, 12, 13, 22, 23, and 33. We excluded the less frequent haplotypes 4 and 5 from the analysis. Women with genotypes containing these haplotypes (n = 15) represent 1.5% of this population. Detailed information on haplotype alleles and genotype frequencies in the Rotterdam Study can be found elsewhere.(18)
Clinical variables were compared between the genotype groups by analysis of covariance to adjust for confounding factors. For the comparisons, we made reference heterozygote and homozygote groups for each of the COLIA1 and VDR alleles. For COLIA1, the groups comprised the GG genotype group for the reference group, GT for the heterozygote risk group, and TT for the homozygote risk group. For VDR haplotype 1, the groups comprised genotypes 22, 23, and 33 for the reference group, genotypes 12, and 13 for the heterozygote risk group, and genotype 11 for the homozygote risk group. The likelihood ratio test statistic was used to test for genotype distribution in women with and without fractures. Odds ratios (ORs; with 95% CI) were calculated by multivariate logistic regression analysis to estimate the relative risk of osteoporotic fracture by genotype. For regression analysis using combinations of VDR and COLIA1 genotype, we defined the reference group to include women with the COLIA1 GG genotype in combination with the VDR 22, or 23 or 33 genotype. The regression analysis included an interaction term defined as VDR genotype multiplied with COLIA1 genotype. All p values for statistical tests were two-sided.
When we analyzed the distribution of fractures in women grouped according to VDR genotype, we observed an overrepresentation of fractures in women carrying the haplotype 1 (Table 1). Subsequently, women were grouped according to carrier status for this VDR haplotype as heterozygous carriers (including the genotypes 12 and 13) and homozygous carriers (consisting of genotype 11) of the risk haplotype and compared with women not carrying the haplotype (including genotypes 22, 23, and 33). No significant differences in known risk factors for osteoporosis could be observed between women grouped according to VDR haplotype 1 (Table 2). Similar results were obtained when the women were grouped according to VDR haplotypes 2 or 3 (data not shown).
Table Table 1.. Number of Postmenopausal Women with Fractures According to VDR Genotype
Table Table 2.. Characteristics of 1004 Postmenopausal Women According to Their VDR Haplotype 1 Genotype
We went on to determine the distribution of fractures in women according to their carrier status for VDR haplotype 1 (Table 3A). Significantly more women who were heterozygous for VDR haplotype 1 had fractures than the women in the reference group, and for women homozygous for the VDR haplotype 1 this difference further increased. When women were grouped according to VDR haplotype 2, we observed an underrepresentation in fracture cases (p = 0.002) while for VDR haplotype 3 no differences were observed (p = 0.65; data not shown). Logistic regression analysis showed that women heterozygous for the VDR haplotype 1 had 1.8 times the risk for fractures compared with women in the reference group. This was further increased to 2.6 times the risk for fracture for women homozygous for the VDR haplotype 1 compared with women in the reference group (Table 3A).
Table Table 3.. Number of Postmenopausal Women with Fractures and Odds Ratios for Fracture According to VDR Haplotype 1 Genotype and According to COLIA1 Genotype
When we analyzed by type of fracture, we observed the VDR genotype effect to be similar for prevalent vertebral fracture cases and incident nonvertebral fracture cases. The age-adjusted OR for vertebral fracture for women who are heterozygous or homozygous for VDR haplotype 1 is 1.7 (95% CI, 0.7–3.9) and 2.7 (95% CI, 1.1–6.6), respectively. The age-adjusted relative OR for nonvertebral fracture for women who are heterozygous or homozygous for VDR haplotype 1 is 2.1 (95% CI, 0.9–4.9) and 3.0 (95% CI, 1.2–7.3), respectively. When we limited the analysis of fracture risk to the group of women for whom we had radiographic data (n = 732 with 88 fracture cases) the data remained essentially unchanged. In this group the age-adjusted OR for any fracture for women who are heterozygous or homozygous for VDR haplotype 1 is 1.7 (95% CI, 0.9–3.2) and 2.6 (95% CI, 1.3–5.0), respectively. Also, when we analyzed the risk for nonvertebral fractures in this group (43 fracture cases), we found that the age-adjusted OR for women who are heterozygous or homozygous for VDR haplotype 1 was 1.7 (95% CI, 0.7–4.2) and 2.7 (95% CI, 1.1–6.6), respectively. Thus, for reasons of power we combined the prevalent vertebral and incident nonvertebral fracture cases in one group of “any fracture.” The risk of fracture essentially did not change after adjustment for potential confounding factors such as age, weight, and bone density in the multivariate regression analysis (Table 3).
In this group of women we also determined the distribution of fractures according to COLIA1 genotype (Table 3B). In correspondence with what we previously found(24) we observed the COLIA1 T allele to be associated with increased fracture risk, independent of BMD. To assess whether there was interaction between the VDR haplotype effect and the COLIA1 genotype effect on fracture, we determined the distribution of fractures according to VDR haplotype 1 in the different COLIA1 genotype groups (Table 4). The distribution of fracture cases according to VDR genotype did not differ in the group of women with the COLIA1 GG genotype. However, in the COLIA1 risk groups of women with the GT and TT genotypes the distribution of fracture cases was strongly VDR genotype dependent (Table 4A). Logistic regression analysis showed that the effect of VDR genotype on fracture risk is absent in women with the COLIA1 GG genotype while the VDR genotype effect is large in the COLIA1 heterozygous GT and homozygous TT risk group (Table 4B). When age, VDR genotype, COLIA1 genotype, and fracture were considered together in a multivariate regression model, we found that VDR genotype significantly modified the COLIA1 genotype effect (p = 0.03 for the interaction term). The effect was found to be similar for incident nonvertebral fracture cases and prevalent vertebral fracture cases and when bone density was entered into the model the results did not change, indicating the interaction effect to be independent of bone density.
Table Table 4.. Number of Postmenopausal Women with Fractures and ORs for Fractures According to Combined VDR Haplotype 1 and COLIA1 Genotype
Although VDR gene polymorphisms have been implicated in the genetic regulation of BMD,(14,15) meta analyses showed that the effects on BMD are small(16,17) as we also indicated earlier in our study population.(18) Although two reports suggested VDR genotype to predict osteoporotic fracture,(19,20) no associations with osteoporotic fracture have been reported in two other studies.(21,22) However, the studies were small and used only the BsmI RFLP in their analysis. As we argued previously,(18) analyzing only the BsmI RFLP can compromise the outcome of such studies because heterogeneous groups are compared. For example, the “b ” allele (58.9% in our population)(18) is in fact a group of three alleles when defined as haplotypes, that is, baT (48.2%), bAT (10.5%), and bAt (0.2%). Thus, there is extensive linkage disequilibrium at the 3′ end of the VDR gene,(14,18,34,35) which can be measured accurately by the molecular haplotypes constructed from the three 3′ RFLPs for BsmI, ApaI, and TaqI. Thus, these haplotypes, which by themselves are not functional polymorphisms, can be used as good markers for truly functional polymorphisms elsewhere in the 3′ end of the VDR gene. This notion is underlined by the findings of some studies that a particular haplotype baT (or haplotype 1) is associated with aberrant mRNA expression and/or stability levels.(14,27–29)
Our findings suggest the VDR to be involved in bone metabolic pathways other than those reflected in BMD but still leading to increased fracture risk. The VDR genotype-dependent increased fracture risk is especially pronounced in interaction with COLIA1 genotype of which we already reported the COLIA1 Sp1 T allele to increase fracture risk in our study population.(24) The interaction between VDR and COLIA1 genotype we describe here raises the possibility of biological interaction of the gene products. The VDR is a member of the steroid transcription factors, known to be important regulators of gene expression. Vitamin D-dependent regulation of expression of bone-specific genes, such as osteocalcin, has been well documented(36) and also includes regulation of the expression of the collagen type Iα1 at the level of transcription.(25,26) Furthermore, in reverse-transcription (RT)-PCR experiments the Sp1 polymorphism has been shown to lead to differential binding affinity of the Sp1 transcription factor,(23) to genotype-dependent differences in COLIA1 mRNA and protein expression levels, and to differences in bone strength.(37) Therefore, the VDR-regulated expression of the collagen type Iα1 gene may differ across COLIA1 alleles and VDR alleles and could be an important factor in the interaction we observe. However, although the exact molecular mechanism underlying the associations we describe here remains to be elucidated, our observations on the interaction should be considered as preliminary.
It is likely that interactions between genetic loci involved in a complex trait are a common phenomenon and several examples have already been shown but mostly in model organisms. Our data represent the first example of interlocus interaction in relation to fractures between two well-known candidate genes in osteoporosis, a complex trait in humans. We show that the interaction leads to increases in the risk of fracture, the clinically most relevant feature of osteoporosis, and that this increase in risk is independent of BMD, the most widely used diagnostic criterion for osteoporosis. This has important consequences not only for the analysis of the genetic basis of osteoporosis but also for the identification of individuals at risk of the disease. However, larger studies will be required to assess the attributable risk of these genetic variations, including their interaction, also in relation to other known risk factors for fracture. Currently, we are collecting fracture data for the complete cohort in the Rotterdam Study to tighten the fracture risk estimates we observed here. Finally, our observation also raises the possibility of developing new therapeutic intervention strategies based on the known involvement of the VDR and the COLIA1 gene in bone metabolism and the determination of fracture risk.
The authors dedicate this publication to the memory of Professor Jan C. Birkenhäger, who died in September 1999. They gratefully acknowledge his inspirational contribution to their research throughout the years. This study was supported by ZorgOnderzoek Nederland (002824890), the Nestor stimulation program for geriatric research in the Netherlands (Ministry of Health and Ministry of Education), the municipality of Rotterdam, the Netherlands Organization for Scientific Research, and the European Commission.