Association of a Common Polymorphism in the Methylenetetrahydrofolate Reductase (MTHFR) Gene With Bone Phenotypes Depends on Plasma Folate Status

Authors

  • Robert R McLean,

    Corresponding author
    1. Research and Training Institute, Hebrew Rehabilitation Center for Aged, Boston, Massachusetts, USA
    2. Division on Aging, Harvard Medical School, Boston, Massachusetts, USA
    • Address reprint requests to: Robert R McLean, MPH, Framingham Study, 73 Mt. Wayte Avenue, Framingham, MA 01702, USA
    Search for more papers by this author
  • David Karasik,

    1. Research and Training Institute, Hebrew Rehabilitation Center for Aged, Boston, Massachusetts, USA
    2. Division on Aging, Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author
  • Jacob Selhub,

    1. Jean Mayer US Department of Agriculture Human Nutrition Research Center on Aging, Tufts University, Boston, Massachusetts, USA
    Search for more papers by this author
  • Katherine L Tucker,

    1. Jean Mayer US Department of Agriculture Human Nutrition Research Center on Aging, Tufts University, Boston, Massachusetts, USA
    Search for more papers by this author
  • Jose M Ordovas,

    1. Jean Mayer US Department of Agriculture Human Nutrition Research Center on Aging, Tufts University, Boston, Massachusetts, USA
    Search for more papers by this author
  • Giuseppina T Russo,

    1. Jean Mayer US Department of Agriculture Human Nutrition Research Center on Aging, Tufts University, Boston, Massachusetts, USA
    2. Department of Internal Medicine, University of Messina, Messina, Italy
    Search for more papers by this author
  • L Adrienne Cupples,

    1. Department of Biostatistics, Boston University School of Public Health, Boston, Massachusetts, USA
    Search for more papers by this author
  • Paul F Jacques,

    1. Jean Mayer US Department of Agriculture Human Nutrition Research Center on Aging, Tufts University, Boston, Massachusetts, USA
    Search for more papers by this author
  • Douglas P Kiel

    1. Research and Training Institute, Hebrew Rehabilitation Center for Aged, Boston, Massachusetts, USA
    2. Division on Aging, Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author

  • The authors have no conflict of interest.

Abstract

A study of a polymorphism in the MTHFR gene, plasma folate, and bone phenotypes in 1632 individuals revealed that the genotype effect on BMD and quantitative ultrasound was dependent on the level of folate. Our findings support the hypothesis that the association between an MTHFR polymorphism and bone phenotypes depends on folate status.

Introduction: Genome-wide screens using quantitative ultrasound (QUS) and BMD phenotypes have shown suggestive linkage on chromosome 1pter-1p36.3, a region containing the methylenetetrahydrofolate reductase (MTHFR) gene. Individuals homozygous (TT) for the MTHFR C677T polymorphism who have low plasma folate concentrations exhibit elevated plasma homocysteine (tHcy) concentrations that may compromise bone quality. We hypothesized that folate status might modify an association between the C677T polymorphism and bone, possibly by influencing homocysteine concentrations.

Materials and Methods: QUS (broadband ultrasound attenuation [BUA], speed of sound, and quantitative ultrasound index) of the heel and BMD of the hip and spine were measured in 1632 male and female members of the Framingham Offspring Study (1996–2001). Participants were assessed for plasma folate concentration and genotyped for the MTHFR C677T polymorphism. TT participants were compared with individuals in the CC + CT group using analysis of covariance.

Results: Adjusted mean QUS and BMD measures did not differ between C677T groups. Although all participants with plasma folate concentrations ≥4 ng/ml had ∼2% higher QUS and BMD than those with folate <4 ng/ml, the association disappeared after controlling for tHcy. Suggestive interactions between folate status and the C677T group (CC + CT versus TT) were found for hip BMD (p ≤ 0.05) and BUA (p = 0.11). Compared with CC + CT participants, TT individuals had lower mean BUA (p = 0.06) and Ward's area BMD (p = 0.08) within the folate <4 ng/ml group and significantly higher hip BMD (p ≤ 0.05) within the folate ≥4 ng/ml group. For both folate groups, TT participants had higher age-adjusted mean plasma tHcy versus CC + CT participants. Controlling for tHcy in these models did not affect the statistical significance of the interaction effects.

Conclusions: Our findings support the hypothesis that the association between the C677T MTHFR polymorphism and bone phenotypes depends on folate status. The mechanism mediating the association, however, remains unclear, but may be partially caused by homocysteine effects on bone.

INTRODUCTION

ACTIVITY OF THE enzyme methylenetetrahydrofolate reductase (MTHFR), which is determined by MTHFR gene polymorphisms, can affect the methylation of homocysteine to methionine. Interference with the methylation process could result in abnormal plasma homocysteine (tHcy) and methionine concentrations. Kang et al.(1, 2) first reported the thermolabile form of MTHFR, which was later discovered to result from a common missense substitution of C to T (alanine-to-valine) at position 677. Frosst et al.(3) suggested that the homozygous (TT) form of this mutation was associated with elevated tHcy concentration. Additional studies found high plasma homocysteine concentrations in individuals with thermolabile MTHFR, but only among those with reduced plasma folate concentrations.(4, 5) These findings indicate that MTHFR genotype and plasma folate concentration combine to influence the development of hyperhomocystinemia.

The observation of an increased prevalence of osteoporosis in individuals with homocystinuria(6, 7) suggests the possibility that high serum homocysteine concentrations have adverse effects on bone. One proposed mechanism is that homocysteine interferes with the cross-linking of collagen, decreasing the structural integrity of bone.(8, 9) Thus, tHcy concentrations may play a role in the development of osteoporosis in individuals possessing both the homozygous form of the MTHFR genotype and low plasma folate concentrations.

Although BMD is considered a major determinant of osteoporotic fracture risk,(10) quantitative ultrasound (QUS) may provide additional information about fracture risk.(11) Similar to BMD, QUS measures are associated with risk factors for osteoporosis,(12) and QUS predicts fracture as well as, and independent of, BMD.(13, 14)

Reported heritability (h2) of QUS (range, 0.45–0.82) and BMD (range, 0.34-0.84) measures suggest a strong genetic component for both of these phenotypes.(15-17) Recently, linkage studies have implicated a region on chromosome 1p as being related to bone phenotypes. A whole-genome scan of 42 families found evidence for linkage of femoral neck BMD on chromosome 1p36.2-p36.3.(18) An additional whole-genome scan of participants in the Framingham Osteoporosis Study, using calcaneal QUS phenotypes, found suggestive linkage to broadband ultrasound attenuation (BUA) on chromosome 1pter-1p36.3.(19) This region contains potential candidate genes for bone status, including the MTHFR gene.

Based on these observations, we studied the association of the C677T MTHFR genotype with ultrasound of the calcaneus and BMD of the hip and spine in men and women in the Framingham Osteoporosis Study. Based on the previously identified combined influences of folate status and MTHFR genotype, we hypothesized that low folate status might influence the association between the common C677T polymorphism and QUS and BMD measures. Because tHcy concentration was available for study participants who had plasma folate information, we examined whether tHcy could be the mechanism driving the joint effect of the MTHFR gene and plasma folate concentration.

MATERIALS AND METHODS

Study population

The Framingham Offspring Study began in 1971 with the primary goal of evaluating risk factors for heart disease among the adult children, and their spouses, of the original participants of the Framingham Heart Study Cohort. The 5124 who enrolled in the Offspring Study at baseline returned for examinations every 4 years. Valid QUS measurements of the calcaneus were obtained from 2795 Offspring Cohort members who participated in the Framingham Osteoporosis Study from 1996 to 2001. During this same time period, BMD measurements of the hip and spine were collected from 3035 participants. The assessments yielded 3066 participants with at least one bone phenotype measurement. We excluded 64 participants who reported current osteoporosis medication use at the time of phenotype measurement. After exclusions, 3002 participants had at least one bone phenotype measurement, of which 1896 had MTHFR C677T genotype information. A total of 1632 of these remaining participants also had plasma folate and homocysteine information. This study was approved by the appropriate Hebrew Rehabilitation Center for Aged, Boston University, and Tufts-New England Medical Center institutional review boards, and written informed consent was obtained for all study subjects.

Ultrasound measurements

We measured QUS of the right calcaneus using the Sahara bone sonometer (Hologic, Bedford, MA, USA), a water-free ultrasound device that contains a motorized caliper mechanism with a pair of sound transducers. The Sahara measured three parameters: speed of sound (SOS; m/s), the distance between the two opposing transducers divided by the time it takes for the signal to pass from one transducer, through the heel, to the other transducer; BUA (dB/MHz), the slope of the sound wave attenuation versus frequency curve;(20) and quantitative ultrasound index (QUI), a linear combination of the SOS and BUA measurements. Higher values for all three ultrasound measures indicate greater bone mass.

A participant's right heel was scanned by using standard positioning unless there was a history of amputation, deformity, or pain in the right foot. Measurements from 180 participants were excluded because of violation of the expected linear frequency-attenuation relation, indicating an invalid scan. Based on duplicate, same-day measurements on 29 subjects, the intramachine CVs for BUA, SOS, and QUI were 5.3%, 0.4%, and 4.3%, respectively.(21)

BMD measurements

BMD of the proximal right femur (femoral neck, trochanter, and Ward's area) as well as the lumbar spine (average BMD of L2-L4) was measured in grams per square centimeter, using a Lunar dual X-ray absorptiometer (DPX-L; Lunar Radiation Corp., Madison WI, USA). The right side was scanned unless there was a history of previous fracture or hip joint replacement. For these individuals, the left side was scanned. We used standard positioning as recommended by the manufacturer. For quality assurance, we scanned phantoms provided by the manufacturer under varying thicknesses of oil and water. No drift in BMD measurements was found over time. The CVs for the DPX-L measurements were 1.7% (femoral neck), 2.5% (trochanter), 4.1% (Ward's area), and 0.9% (lumbar spine).(22)

MTHFR genotyping

DNA was isolated from peripheral blood leukocytes from 5–10 ml whole blood samples collected between 1987 and 1991. MTHFR C677T genotyping was performed using the Perkin-Elmer/Applied Biosystems 7700 Sequence Detection Systems and TaqMan reagents. The technique involves the use of two probes labeled at the 5′ end with different reporter fluorescent dyes, one complementary to the wildtype DNA strand and the other complementary to the strand with the C>T mutation, and at the 3′ end with a fluorescent quencher. The sample was genotyped as CC if there was only fluorescence from the wildtype reporter, TT if there was only fluorescence from the reporter for the mutant allele, and CT if there was an intermediate fluorescence from both reporters.(23-25)

Laboratory measurements

From 1991 to 1995 and from 1995 to 1998, blood samples from fasting (>10 h) subjects were obtained and stored at −70°C. Plasma folate was determined by a 96-well plate microbial (Lactobacillus casei) assay.(26–28) Plasma total homocysteine (tHcy) was determined by HPLC with fluorometric detection.(29) The CVs for the assays were 13% for folate and 8% for tHcy.(28)

Covariate description

Gender, age, and other potential confounding variables were measured at the time of the phenotype assessment. Height without shoes was measured to the nearest 0.25 in using a stadiometer. Weight in pounds was measured using a standard balance beam scale. Body mass index (BMI) was calculated in kilograms per meters squared. A physical activity score was assessed using a validated questionnaire of self-reported activity in the past 7 days.(30) Smoking was assessed as current cigarette smoker (smoked regularly in the year before phenotype measurement), former smoker, or never smoked cigarettes. Total alcohol consumption was calculated based on an equation using self-report of the intake of beer, wine, and mixed drinks per week, providing an alcohol equivalent in ounces per week. Usual intakes of calcium and vitamin D from foods and dietary supplements for the previous 12 months were assessed using a 126-item semiquantitative food frequency questionnaire (FFQ), as described elsewhere.(31) A participant's FFQ was excluded from analyses if their reported energy intake was <600 or >4000 kcal or if more than 12 food items were left blank. This FFQ has been validated for numerous nutrients.(31, 32) For women, estrogen replacement therapy was evaluated as current use, former use, and never used oral conjugated estrogen, patch, or cream. Menopause status was defined as postmenopausal (periods stopped for at least 1 year or currently using estrogen replacement therapy) or premenopausal.

Statistical analysis

We used analysis of covariance (ANCOVA) to compare crude and least-squares adjusted mean QUS and BMD among MTHFR C677T genotype groups. In the adjusted model, we controlled for the potential confounders gender, age, height, BMI, physical activity score, smoking status, alcohol consumption, dietary intakes of calcium and vitamin D, and estrogen use and menopause status in women.

To define folate status, we dichotomized participants into two groups: plasma folate <4 ng/ml and plasma folate ≥4 ng/ml. We chose 4 ng/ml as a cut-off because this was still considered below the normal level of 5 ng/ml and was close to the median value for the population. To examine the potential effect of plasma folate status on bone phenotypes, we used ANCOVA to compare least-squares adjusted mean QUS and BMD between folate groups (<4 and ≥4 ng/ml). These analyses were repeated after adjustment for tHcy to determine if tHcy was mediating the folate association with BMD and QUS.

To determine whether plasma folate modified the association between genotype group and QUS or BMD, we tested for statistical interaction between MTHFR C677T group and folate (≥4 versus <4 ng/ml) for each bone phenotype, controlling for the previously listed covariates. In these analyses, we considered p values for the interaction term that were close to or less than 0.10 as suggestive of statistical interaction. Additional ANCOVAs were performed to compare the R2 values of models that included covariates with or without the addition of MTHFR genotype, folate status, and the interaction term. The differences between R2s were calculated to determine the variations in bone phenotypes caused by MTHFR genotype, folate status, and their interaction.

We subsequently used ANCOVA to compare least-squares adjusted mean QUS and BMD between genotypes within each folate group, adjusting for potential confounders. To assess potential confounding caused by inclusion of related individuals among study participants, these analyses were repeated using only unrelated individuals.

Implementation of nationwide folic acid fortification of grain products (1997) occurred during the period when bone measures were being performed (1996–2001). The plasma folate concentration measurements used in this study, however, were obtained before folic acid enrichment (1991-1995). To examine any effect on bone caused by potential changes in folate status during the time between folate measurement and bone phenotype assessment, we repeated our analyses using a subset of participants who had available post-fortification plasma folate concentrations contemporaneous with their bone phenotype measures.

Because we used the cut-off of 4 ng/ml to determine folate status, some study participants with plasma folate concentrations below “normal” (5 ng/ml) would be included in the “high” folate group (≥4 ng/ml). Therefore, we repeated our analyses defining folate status using the cut-off of 5 ng/ml.

To determine whether tHcy might be influencing the association among genotype, nutrient, and bone phenotypes, we calculated Pearson correlation coefficients (r) between tHcy and folate for each folate-genotype stratum and compared age-adjusted mean tHcy concentrations between genotype groups within each folate stratum. We also repeated the test for interaction between genotype and folate, adding tHcy to the model.

For all statistical analyses, we used release 8.1 of the SAS/STAT component of the SAS system (SAS Institute, Cary, NC, USA).

RESULTS

The 1106 study participants with measured phenotypes but no MTHFR C677T information were similar to the 1896 that were genotyped with respect to all phenotypes and covariates except mean age (genotyped, 59.4 years; not genotyped, 58.0 years; p < 0.001) and the proportion of women (genotyped, 0.54; not genotyped, 0.58; p = 0.03).

Table 1 displays the descriptive statistics for the study participants by C677T genotype. The MTHFR C677T genotype frequencies in our sample were 39%, 47%, and 14% for CC, CT, and TT, respectively. This distribution follows the Hardy-Weinberg equilibrium. For the 1896 participants (882 men and 1014 women), the mean ± SD age was 59 ± 9.4 years (range, 32–87 years). Genotype groups were similar for all characteristics with the exception of plasma folate, which showed a significant decreasing linear trend from CC to TT (p = 0.03), and tHcy, which displayed a significant increasing linear trend from CC to TT (p < 0.001). Because of similar distributions of characteristics and our a priori hypothesis that only the homozygous TT group was known to display high tHcy in the presence of low folate, we combined the CC and CT genotypes and compared QUS and BMD measures to the main group of interest, TT.

Table Table 1.. Characteristics of Framingham Osteoporosis Study Participants by MTHFR Genotype (Mean ± SD, Unless Indicated Otherwise)*†
original image

Table 2 shows the results of the unadjusted and adjusted models comparing mean QUS and BMD measures between the two genotype groups. There were no differences in mean QUS measures between genotype groups for either the unadjusted (p value range, 0.11–0.18) or the adjusted model (p value range, 0.52-0.94). Comparisons of unadjusted and adjusted BMD measures between genotype groups were similar to those for QUS measures (unadjusted p values = 0.72-0.93; adjusted p values = 0.17-0.35).

Table Table 2.. Unadjusted and Least Squares Mean ± SE QUS and BMD, According to MTHFR Genotype, in Framingham Osteoporosis Study Participants
original image

We then examined the main effects of folate on QUS and BMD by comparing adjusted mean QUS and BMD according to plasma folate status. The folate ≥4 ng/ml group had approximately 2% higher QUS and BMD measures than the folate <4 ng/ml group. p values for differences in means ranged from 0.02 (trochanter) to 0.13 (BUA), with most values close to 0.05. When we added tHcy to this model, folate was no longer significantly associated with any bone phenotype.

To test our hypothesis of a statistical interaction between the MTHFR gene and plasma folate, we included in the regression model an interaction term between MTHFR C677T genotype (CC + CT versus TT) and plasma folate (<4 versus ≥4 ng/ml). For BUA, there was a borderline significant interaction (p = 0.11). The formal test for an interaction for SOS (p = 0.31) and QUI (p = 0.20) did not approach significance at the p = 0.10 level. For BMD phenotypes, we observed statistically significant interactions for femoral neck (p = 0.02), Ward's area (p = 0.01), and trochanter (p = 0.05), but not for lumbar spine (p = 0.41). MTHFR genotype, folate status, and their interaction explained about 0.5% (difference in R2 without and with MTHFR, folate status, and their interaction = 0.005) of the total variation for all bone phenotypes.

To illustrate the effects of interaction, the results for comparisons between genotype groups stratified by folate group are displayed in Fig. 1. In the folate <4 ng/ml group, bone phenotypes tended to be lower in TT individuals compared with CC + CT individuals. Differences in adjusted mean BUA and Ward's area BMD approached statistical significance (BUA, 71.6 versus 75.5, p = 0.06; Ward's area, 0.719 versus 0.750, p = 0.08), whereas differences for all other phenotypes were not statistically significant (p value range, 0.14–0.86). Among individuals with folate ≥4 ng/ml, TT individuals had significantly higher adjusted mean femoral neck, Ward's area, and trochanter BMD compared with CC + CT individuals (femoral neck, 0.951 versus 0.926, p = 0.03; Ward's area, 0.796 versus 0.762, p = 0.01; trochanter, 0.830 versus 0.809, p = 0.05). There were no differences in lumbar spine BMD or any QUS measures between genotype groups in the folate ≥4 ng/ml group. Similar trends for differences in bone phenotypes were observed in the analyses using only unrelated individuals. At the femoral neck, for example, TT individuals in the folate ≥4 ng/ml group had higher BMD (0.946 versus 0.924; p = 0.09), whereas TT participants in the folate <4 ng/ml group had significantly lower BMD (0.877 versus 0.917; p = 0.03).

Figure FIG. 1..

Least-squares adjusted mean (A) BUA, (B) femoral neck BMD, (C) Ward's area BMD, and (D) trochanter BMD for genotype group stratified by folate status (folate <4 ng/ml: CC + CT, n = 409, TT, n = 90; folate ≥4 ng/ml: CC + CT, n = 989, TT, n = 144). Comparisons of least-squares means for CC + CT vs. TT were performed using ANCOVA.ap = 0.06;bp < 0.05;cp = 0.05.

Although the sample size for the analysis of the study participants with contemporaneous folate and bone phenotype measures was not sufficient to detect any meaningful, statistically significant differences, we did observe the same trends as in the results from the analysis using the pre-fortification folate concentrations. For instance, TT individuals in the folate ≥4 ng/ml group had 2% higher trochanter BMD than CC + CT persons, whereas in the folate <4 ng/ml group, TT participants had 20% lower trochanter BMD compared with CC + CT participants.

The results of the analysis using a plasma folate cut-off of 5 ng/ml showed that bone phenotypes still differed between genotype groups. These differences, however, were of lesser magnitude and no longer statistically significant. Femoral neck BMD, for example, was not significantly different between CC + CT and TT individuals within either the folate <5 ng/ml group (CC + CT = 0.917, TT = 0.931; p = 0.35) or the folate ≥5 ng/ml group (CC + CT = 0.925, TT = 0.916; p = 0.55).

To determine if tHcy might explain the observed interaction between MTHFR genotype and folate, we examined tHcy concentrations in the four folate-genotype strata. TT participants with folate <4 ng/ml exhibited the greatest correlation between plasma tHcy and folate (−0.40, p < 0.001), whereas correlations in the other three strata were similar (CC + CT, folate <4 ng/ml, −0.23, p < 0.001; TT, folate ≥4 ng/ml, −0.25, p = 0.002; CC + CT, folate ≥4 ng/ml, −0.29, p < 0.001). For both folate levels, TT participants had higher age-adjusted mean plasma tHcy compared with CC + CT participants (folate <4 ng/ml: 13.27 versus 11.42, p < 0.001; folate ≥4 ng/ml: 9.99 versus 9.21, p = 0.01). Including tHcy in the above interaction models resulted in negligible changes in the regression coefficients and p values for the MTHFR × folate interaction terms.

DISCUSSION

Study participants homozygous for the common C677T mutation in the MTHFR gene (TT) had similar QUS and BMD values to those possessing at least one wildtype allele (CC + CT). Among those with plasma folate concentrations <4 ng/ml, participants with the TT genotype had lower BUA and Ward's area BMD than CC + CT individuals. Among participants with folate concentrations ≥4 ng/ml, however, TT individuals had higher femoral neck, Ward's area, and trochanter BMD than CC + CT individuals. Although these results support the hypothesis that the association between the C677T mutation and bone phenotypes depends on folate status, the additional proportion of variance explained by the interaction was quite modest. Relying on the R2 as a measure of effect, however, may be misleading because of its dependence on the SDs of the interacting variables and the outcome of interest.(33)

Whereas TT individuals had higher mean tHcy concentrations compared with CC + CT individuals at both folate levels, the effect of the TT genotype on bone phenotypes was reversed depending on folate concentration. Furthermore, inclusion of tHcy in the models testing for genotype-folate interaction did not affect the interaction terms. Thus, tHcy did not seem to fully explain the joint effect of the TT genotype and folate status on bone phenotypes.

The frequency of the T allele among our study participants was 0.37, with 14% of the participants homozygous for the T allele. These percentages were lower than those reported in a study of postmenopausal Japanese women (0.44, 18.6%),(34) but higher than those reported among Danish postmenopausal women (0.29–0.30, 6.5-8.7%).(35, 36)

The study by Miyao et al.(34) among postmenopausal Japanese women found evidence that women homozygous for the C677T mutation had lower lumbar spine BMD than CC women and lower total body BMD than both CC and CT women. In a study of postmenopausal Danish women, the TT genotype was also found to be associated with lower femoral neck, total hip, and spine BMD and a 2-fold increased risk of fracture.(36) The association among both the Japanese and Danish women was demonstrated without accounting for folate status, whereas we did not see any effect of the TT genotype in the absence of stratification. This inconsistency could be because of differential plasma folate levels between our study population and the previously published studies. Possibly, lower mean plasma folate levels among Japanese and Danish women, relative to our study participants, could account for the observed association.

Consistent with our findings, a study of Danish postmenopausal women found that forearm BMD and heel BUA and SOS did not differ significantly across MTHFR C677T genotype groups.(35) They did find, however, that the T allele was protective against risk of osteoporotic fracture. The authors presented a number of possible mechanisms for this seemingly contradictory result, including that the T allele influences factors other than BMD or QUS that could affect risk of fracture. Again, folate status was not assessed; therefore, the possibility cannot be ruled out that the fractures reported among TT women occurred only among those with inadequate plasma folate levels.

The most widely recognized proposed mechanism behind the observed association between the MTHFR C677T mutation and bone phenotypes is that elevated tHcy concentrations in TT individuals interfere with collagen cross-linking, resulting in structurally weaker bones. Because controlling for tHcy attenuated the main effect of folate on bone phenotypes, we initially suspected that tHcy could be mediating the association between folate and bone traits. Indeed, among individuals with plasma folate <4 ng/ml, TT individuals not only exhibited lower values for QUS and BMD, but also had mildly elevated mean tHcy. TT individuals also had increased mean tHcy among participants with folate ≥4 ng/ml. In this folate group, however, these participants had the highest QUS and BMD. Furthermore, the addition of tHcy to the model containing the interaction between MTHFR and folate did not attenuate the interaction, implying that tHcy is not the sole mediator of the relation between MTHFR genotype and plasma folate. Thus, the mechanism underlying our findings is still uncertain.

An alternate hypothesis explaining the observed association between the MTHFR C677T mutation and bone may lie in the process of DNA hypomethylation. DNA methylation is a critical epigenetic mechanism that affects genetic structure and stability(37–39) and influences expression, as well as silencing of genes.(40-42) Changes in methylation patterns may contribute to the development of cancer and autoimmune disease and may influence abnormalities associated with aging.(43) Hypomethylation, in particular, has been associated with diseases such as gastritis and colitis and with different types of tumors.(39) Studies have shown that, in the presence of low folate, TT individuals experience diminished DNA methylation compared with CC individuals.(44, 45) Additional research is needed to determine whether hypomethylation resulting from the MTHFR C677T mutation affects bone tissue or phenotype expression.

A second possible hypothesis behind the association is that a different polymorphism in linkage disequilibrium with the MTHFR gene could be driving the observed association. Known candidate genes that are in close proximity to MTHFR on chromosome 1 include TNF receptor superfamily members 2, 9, and 18. Further study is needed to examine these possibilities.

This study had several potential limitations. First, plasma folate status was assessed at least 4 years before measurement of bone phenotypes and other covariates and before government-mandated, nationwide fortification of enriched grain products with folic acid in 1996. Consequently, plasma folate levels at the time of bone phenotype assessment for participants measured post-fortification may be significantly different from the pre-fortification levels used in this study. It has been shown that fortification has indeed improved folate status among middle-aged and older adults, significantly decreasing the prevalence of individuals with low plasma folate.(46) It is therefore possible than an improvement in folate status among study participants with “low” plasma folate may have had a positive effect on bone mass in the time between folate and bone phenotype measurements.

To examine this possible bias caused by fortification, we repeated our analysis including only those participants with contemporaneous folate and bone phenotype measurements after fortification. Contemporaneous measures were available for a subset of study participants who had bone phenotypes measured from 1996 to 1998. Although folic acid fortification was intended to be phased in over a 2-year period, with the goal of enrichment of all products by March 1998, it seemed that most products in New England contained folic acid by the summer of 1997.(46) Therefore, for the analysis using contemporaneous measures, we included only the 621 participants with both folate and bone phenotype measurements after June 1997, 14 (2%) of whom had plasma folate concentrations <4 ng/ml. Although the extremely low prevalence of low folate status did not allow us to detect any meaningful joint effects in this smaller subset of individuals, we did observe the same trends that were apparent in the analysis using the pre-fortification folate measures. Therefore, although individual folate concentrations did increase from the time of the first folate assessment to bone measurement, it did not seem that these increases significantly affected bone phenotypes in the time between fortification and bone measurement. Thus, the pre-fortification measurements better represented long-term folate status compared with the contemporaneous measurements. Furthermore, results generated from pre-fortification folate concentrations in the United States are relevant for other populations around the world in which there is currently no folic acid supplementation.

A second limitation of our study was the potential for a biased genotype-phenotype association because 32% of study participants were related. To assess the possible bias, we repeated all ANCOVAs using only the unrelated individuals among the study participants. The results from the analyses conducted among unrelated participants were similar to those of the models using all study participants, making it unlikely that the inclusion of related individuals was influencing the results.

Third, the cut-point used for our two folate groups was based on a compromise between a valid representation of low folate status and the need to have adequate numbers of individuals within groups. Serum folate levels of 5–16 ng/ml are considered normal, whereas levels <3 ng/ml are characterized as negative nutrient balances.(47) Only 17% of study participants had folate concentrations <3 ng/ml. Thus, we chose to use the cut-off of 4 ng/ml because it was still considered below “normal,” and it increased the number of participants in the “low” plasma folate group (31%), enhancing our power to detect differences in bone phenotypes between folate groups and between genotypes when stratified by folate status. Furthermore, the results of our analysis using the 5 ng/ml cut-off yielded a lesser effect than that observed from the analysis using the 4 ng/ml cut-off. This would suggest that the effects of low folate are not detectable with the “normal” cut-off because a number of individuals classified as “low” are truly replete with folate.

Fourth, 204 participants were not included in the fully adjusted models because of missing covariates, reducing our sample size and our power to detect differences in bone phenotypes. Finally, all study participants were white; therefore, the results of this study may not be generalizable to all populations. There is, however, large variability in the frequency of individuals homozygous for the mutation among different ethnic populations throughout the world, ranging from less than 1% in Africans to more than 20% in U.S. Hispanics.(48, 49) The ethnic homogeneity among our study participants may have aided in protecting against potential confounding by population stratification.

Our findings support the hypothesis that the association between a common mutation in the MTHFR gene and bone phenotypes is modified by plasma folate status. While the TT genotype was associated with lower values for bone phenotypes in the presence of low plasma folate concentration, it was also associated with higher values at elevated folate levels. The mechanism whereby folate status modified the association between genotype and bone density remains unclear. Although previously published studies have demonstrated an association between MTHFR genotype and fracture, we are unable to predict whether the genotype-folate interaction contributes meaningfully to risk of fracture, especially if underlying mechanisms involve collagen alterations and mineralization effects of the MTHFR genotype-folate interaction. Future studies should explore the relation between tHcy concentrations, bone phenotypes, and fracture risk and should investigate alternate mechanisms that may drive the association between the MTHFR C677T polymorphism and bone status.

Acknowledgements

The authors thank the clinic technician, Mary Hogan, for invaluable assistance and support and Hologic for supplying the Sahara bone sonometer. This study was funded by National Institutes of Health Grants AR/AG 41398 and HL54776; the Framingham Heart Study Contract N01-HC-25195; U.S. Department of Agriculture Research Service Agreement 58–1950-9-001 and Contract 53-3K06-01.

Ancillary