The heritability of plasma homocysteine, and the influence of genetic variation in the homocysteine methylation pathway

Abstract

Background: The extent of genetic influence on plasma homocysteine, a risk factor for ischaemic heart disease, is uncertain. Many association studies have investigated common polymorphisms and their role in hyperhomocysteinaemia, but only the thermolabile variant of methylene tetrahydrofolate reductase (MTHFR) has shown an association (small but robust).

Aim: To estimate the heritability of plasma homocysteine and the contributions of well-studied common SNPs in the three main candidate genes in the homocysteine methylation pathway.

Design: Twin study.

Methods: We studied 216 monozygotic and 790 dizygotic pairs of twins; all were women. Blood was collected after overnight fasting for measurement of homocysteine, folate, vitamin B12, and extraction of DNA. Heritability was estimated by structural modelling, including correction for known environmental influences, particularly serum folate. The frequency of a common coding SNP in MTHFR and methionine synthase (MTR), and two coding SNPs in methionine synthase reductase (MTRR) were measured in dizygotic twins by ABI 7700 Sequence Detection, and the contribution of each to homocysteine variance was determined.

Results: The heritability of homocysteine was 57% (95%CI 51–63%). The highest contribution to homocysteine was serum folate, accounting for 10.13% of variance. This was twice the total genetic contribution of 4.56%, and only the C1763T SNP of MTRR showed significant association with homocysteine.

Discussion: Homocysteine has one of the highest heritabilities of common risk factors for ischaemic heart disease. This is not accounted for by the commonly studied SNPs in MTHFR, MTR and MTRR.

Introduction

Homocysteine (Hcy) is an independent risk factor for ischaemic heart disease,^1–4 and a number of variants in the genes governing Hcy metabolism have been investigated as potential markers of risk, although the impact of these polymorphisms on Hcy levels remains unclear. The C677T polymorphism of the MTHFR gene results in (at most) a moderate (up to 15–19%) increase in mean Hcy levels between the CC and TT genotypes,⁵ and this effect is greatly attenuated in those individuals with mid-range to high folate levels.⁶ The effect of other common polymorphisms on Hcy levels remains unproven. It may be that genetic factors are not a major influence on Hcy levels; alternatively, the lack of consensus amongst the many association studies published may be due to confounding environmental factors, small study numbers that lack the power to detect relatively small gene effects, or to the presence of other unknown genetic variants that are exerting more powerful effects on the phenotype than those studied so far. Our main objective was to obtain a robust estimate of the inherited contribution to variance in plasma Hcy. Well-tried structural modelling in a large twin set allowed us to estimate heritability and the confidence limits around this. Our second aim was to estimate the contribution of commonly studied SNPs to the inherited component of Hcy variance.

Methods

Study population and design

We studied 1006 twin pairs (216 monozygotic, 790 dizygotic) selected at random from the St Thomas’ UK Adult Twin Registry. All participants in The St Thomas’ UK Adult Twin Registry are healthy, female, Caucasian adults aged 18–79 years, recruited from the UK general population using national media campaigns. A questionnaire was used to gather information on medical history, demographics and medication. Where there was uncertainty, zygosity was confirmed by DNA fingerprinting. This patient group has been described previously.⁷,⁸

At recruitment, we measured weight, height and blood pressure, and took venous blood samples from both twins at the same visit, after at least 9 h fasting. Aliquots of serum, plasma and cells were stored at –80°C until measurements were made. The use of small aliquots avoided the need for repeated freezing and thawing of samples.

Measurement of plasma total homocysteine concentration used high performance liquid chromatography and fluorescence detection.⁹ A competitive binding radioimmunoassay (BioRad) was used to measure serum B₁₂ and folate levels. Genotyping was performed on genomic DNA from 500 dizygotic pairs selected randomly from the above population. Four single nucleotide polymorphisms (SNPs), all encoding amino acid changes, were investigated: C677T in MTHFR, A66G and C1763T at opposite ends of methionine synthase reductase (MTRR), and A2756G in methionine synthase (MTR). These were selected for being common and (with the exception of C1763 MTRR) frequently studied; in addition, available haplotype information for MTR and MTRR suggested that these SNPs would be markers for the rest of the gene. Table 1 shows primers and probes used for TaqMan (ABI 7700 Sequence Detection) genotyping analysis for each of the polymorphisms studied.

Table 1

Polymorphisms studied and primers and probes used for TaqMan analysis

Polymorphism	Amino acid change	Forward (F) and reverse (R) primers, and probes (P)
MTHFR C677T (rs1801133)	Alanine to valine at position 222	F: GCACTTGAAGGAGAAGGTGTCT*
		R: CCTCAAAGAAAAGCTGCGTGATG*
		P: ATGAAATCGGCTCCCGC*
		P: ATGAAATCGACTCCCGC*
MTR A2576G (rs1805087)	Aspartate to glycine at position 919	F: GGATGAATACTTTGAGGAAATCATGGAAGA
		R: TGTTTCTACCACTTACCTTGAGAGACT
		P: ATTAGACAGGACCATTATG
		P: ACAGGGCCATTATG
MTRR C1763T (rs10380)	Histidine to tyrosine at position 622	F: AGCTTGACAACCTTTTAGTGATCCATTA*
		R: GGAGCATCTCTTGAGAAGGAAACC*
		P: CATGCTTAAGGAAATGTCTGA*
		P: ATGCTTAAGGAAATATCTGA*
MTRR A66G (rs1801394)	Isoleucine to methionine at position 22	F: AGCAGGGACAGGCAAAGG
		R: GCAGAAAATCCATGTACCACAGCTT
		P: ATCGCAGAAGAAATATGTGA
		P: ATCGCAGAAGAAATGTGTGA

*Denotes primers and probes designed in the reverse orientation. Polymorphisms are bolded and underlined. MTHFR, methylene tetrahydrofolate reductase; MTR, 5-methyltetrahydrofolate homocysteine methyltransferase or methionine synthase; MTRR, 5-methyltetrahydrofolate homocysteine methyltransferase reductase or methionine synthase reductase.

Statistical analysis

Homocysteine and B₁₂ and folate levels were log-transformed before statistical analysis, to normalize distribution. The frequent use of folic acid as a vitamin supplement distorted the distribution of serum folate; therefore folate levels were used as quartiles. Structural equation modelling using Mx software (Virginia Commonwealth University) was used to calculate heritability, as in previous studies of the St Thomas’ twins.⁸ Structural equation modelling uses the information of the variance covariance matrix from monozygotic and dizygotic twins to estimate the extent of genetic and environmental components upon homocysteine levels, and their confidence intervals.

To determine the contribution of each SNP to homocysteine levels, the differences in homocysteine levels (corrected for folate) between dizygotic twins were regressed upon differences between indicator variables for the additive and dominance effects of each SNP. Such an analysis is not susceptible to bias due to population stratification.¹⁰ The percentage of variance accounted for by each SNP is measured relative to the total within-twinship variance, and is an estimate of that SNP's contribution to heritability.

Results

Demographic data on the twin population is shown in Table 2. Twins in the highest quartile of folate had 10% lower Hcy than twins in the lowest quartile (p < 0.001) and a small negative linear relationship between Hcy and serum B₁₂ was found (r = –0.15, p < 0.001). Heritability of Hcy was estimated at 57% (95%CI 51–63%). Age accounted for 4% of the variation, and correction for folate or B₁₂ had no effect. Multiple regression analysis showed 10.13% of the variance in Hcy to be explained by folate. The total contribution of the four SNPs was much lower, at 4.56% (p = 0.0014) than the contribution from serum folate. The contribution of individual SNPs is shown (Table 3). Only the MTRR C1763T polymorphism demonstrated significant association with Hcy; however, the variance attributable to this polymorphism was only 1.95%.

Table 2

Patient demographic data

Variable	Mean (SD)
Homocysteine (mmol/l)	8.53 (3.3)
B12 (pg/ml)	586 (294)
Folate (ng/ml)	12.37 (6.1)
Weight (kg)	67 (12.7)
Height (cm)	163 (6.8)
BMI (kg/m2)	25.3 (5.5)
Systolic blood pressure (mmHg)	121 (18)
Diastolic blood pressure (mmHg)	77 (11)
Serum total cholesterol (mmol)	5.4 (1.3)

Table 3

Non-synonymous single nucleotide polymorphisms (SNPs) in the homocysteine methylation pathway, and their contribution to variance in plasma homocysteine

Gene/SNP	Allele/number of Patients	SNP number and amino acid change	Contribution to variance (%)		p*
			Between twins	In group
MTHFR/C677T	CC/436	rs1801133	0.37	0.77	0.019
	CT/483	Alanine/valine
	TT/106
MTRR/A66G	AA/663	rs1801394	0.21	0.07	0.67
	AG/333	Methionine/isoleucine
	GG/30
MTRR/C1763T	CC/805	rs10380	1.95	0.41	0.0073
	CT/201	Histidine/tyrosine
	TT/12
MTR/A2756G	AA/166	rs1805087	1.56	0.53	0.019
	AG/420	Aspartate/glycine
	GG/242

*For the larger contribution to variance. MTHFR, methylene tetrahydrofolate reductase; MTR, 5-methyltetrahydrofolate homocysteine methyltransferase or methionine synthase; MTRR, 5-methyltetrahydrofolate homocysteine methyltransferase reductase or methionine synthase reductase.

Discussion

Where possible, the confirmation of heritability is an important prelude to any study of associations between DNA polymorphisms and disease, as it reduces the risk of false-positive and false-negative associations. Twin studies have an advantage over other methods of estimating heritability, in that they enable calculation of both additive and dominance variance. Early twin studies yielded variable estimates for the heritability of homocysteine: Reed et al.¹¹ and Berg et al.¹² estimated heritability values of 0.75 and 0.5, respectively, whereas Cesari et al.¹³ found no evidence for genetic influence of homocysteine levels. One criticism of such studies has been their small size; power calculations suggest that in order to obtain a reasonably accurate estimate of the magnitude of genetic influence on a quantitative trait, at least 200 pairs need to be studied for a trait of high heritability, whereas at least ten time this number need to be studied for traits of intermediate or low heritability.¹⁴ Our large twin study suggests that, after correction for known influences, genetic factors account for about 57% (95%CI 51–63%) of the variability in Hcy levels seen in this population.

Several previous studies have attempted to assess the influence of common polymorphisms upon the variation in homocysteine levels. Tsai et al. reported that the A2756G mutation of the methionine synthase gene, the C677T polymorphism of the MTHFR gene and three common mutations of the cystathionine beta synthase genes collectively contributed to only 1.5% of the variance seen in homocysteine levels in a randomly selected population with and without premature heart disease.¹⁵ Kluijtmans et al. found that five common SNPs accounted for 9% of the variation in homocysteine levels.¹⁶ These studies are smaller than ours and, without an estimate of overall heritability of homocysteine in their chosen populations, did not aim to compare the relative genetic contribution of each SNP to overall inherited variance. In addition, some authors did not take into account the influence of serum folate upon homocysteine in the study population.¹⁵

Our results suggest that a sibling's plasma Hcy is a much stronger predictor of a person's plasma Hcy than any known environmental factor. However, all our tested SNPs were weaker predictors of plasma Hcy than the strongest environmental factor (serum folate) which implies that studies to date have not yet studied the correct SNPs, or that other genes are involved. The study predates robust information about tag SNPs for the candidate genes. However, assuming that heritability is mostly dictated by common SNPs, and given the available information about common coding SNPs and linkage disequilibrium (LD) in our three genes, we have excluded most coding SNPs in the three main genes of the methylation pathway as the main site of genetic variance in Hcy. We have not however excluded rare variants in the coding region, and the latest Hapmap release indicates that we have not excluded SNPs contributing to Hcy variance in the upstream regions of MTHFR and MTRR. In the latter, a small LD block with a common ser/leu SNP (rs1532268) is also now predicted between the two SNPs we studied at either end of the gene. The lack of significant influence of the MTHFR polymorphism is consistent with a recent meta-analysis of >100 association studies that identified only a weak association between the MTHFR C677T polymorphism and elevated Hcy.¹⁷ The degree of heritability shown by Hcy in our study suggests that functional variants of genes controlling homocysteine metabolism will be found either in regulatory regions of the same genes (but in a different haplotype block from our SNPs) or in different genes altogether. A recent linkage study has found a potential area of interest on chromosome 16q (LOD score 1.76),¹⁸ but this area does not contain any obvious candidate genes.

One potential corollary of a substantial genetic contribution to plasma Hcy is the enhanced likelihood that Hcy is a causal (rather than a marker) risk factor for IHD. This was also suggested by a recent Mendelian randomization analysis, which reported consistency between genotype and phenotype odds ratios of disease.¹⁷ The modest size, however, of these odds ratios (∼1.1 for each 1 μmol/l increase in plasma Hcy) suggests that larger outcome studies than so far completed may be required to detect benefit from the small reductions in Hcy achieved with vitamin treatments.

Acknowledgements

We thank the Wellcome Trust, EU Biomed programs (Genomeutwin and EuroClot), the Bristol Myers Squibb Cardiovascular Fellowship and the British Heart Foundation for support, and Miss Isobel Ramsay and Mr Andrew Wood for their help with Taqman assays.