Measurements of the homocysteine phenotype, the Alu element methylation phenotype, and the LINE-1 methylation phenotype were available for 760, 628 and 621 participants, respectively. All had genotype data, 533 men had data on all three phenotypes; each analysis included the maximum number possible. The phenotype groups had similar frequencies for the MTHFR
genotype, but differed by age and hence differed slightly on age-related variables (Table ). The MTHFR
genotype prevalence in the largest group, the plasma homocysteine group, was 12.2%, similar to the frequency reported in a large North American sample [7
Characteristics of Normative Aging Study participants, 1961-2001, with measurements on three phenotypes.
Age and current smoking status were associated with homocysteine (P ≤ 0.001), age was associated with Alu (P ≤ 0.005), and current smoking was associated with LINE-1 (P = 0.055). Folate, vitamin B-6, and vitamin B-12 were associated with homocysteine (P ≤ 0.005), vitamin B-6 was associated with Alu (P ≤ 0.05), and these biomarkers had little or no association with LINE-1. Models exploring the SNP--phenotype association were adjusted for age, smoking, and nutrient residuals. Adjusting for age and smoking made little difference to the coefficients for each SNP. The set of SNPs comprising the most significant associations was nearly identical with or without adjusting for nutrient residuals. Further adjustment for the MTHFR rs1801133 variant made little or no difference to the SNP regression coefficients. The most statistically significant SNPs for each phenotype were relatively common (MAF ≥13%), and the set of most significant SNPs was unique to each phenotype (Tables , , and and Figure ).
The most statistically significant associations (P ≤ 0.005) between single nucleotide polymorphisms and the plasma homocysteine phenotype a, e
The most statistically significant associations (P ≤ 0.005) between single nucleotide polymorphisms and the Alu methylation phenotype a, b, d, f
The most statistically significant association (P ≤ 0.005) between single nucleotide polymorphisms and the LINE-1 methylation phenotype a, b, d, e, f
Figure 1 Manhattan plot. Folate-related SNPs as predictors of plasma total homocysteine and global genomic DNA methylation phenotypes. Models adjusted for age, smoking status, and folate, vitamin B-6, and vitamin B-12 residuals. Horizontal lines represent nominal (more ...)
Total plasma homocysteine phenotype
Of the 20 SNPs with a nominal P ≤ 0.005, five were also significant at the FDR threshold (P ≤ 0.05) (Table ). These 5 SNPs comprise 3 genes: formiminotransferase cyclodeaminase (FTCD; 1 SNP, intronic), solute carrier family 19 (folate transporter), member 1 (SLC19A1, 3 SNPs, representing coding nonsynonymous, 5' region, and intronic variants), and solute carrier family 19, member 3 (SLC19A3, 1 SNP, intronic). Genetic variation in all 5 SNPs was positively associated with plasma homocysteine levels, and effects were similar in direction and magnitude (variant genotypes associated with a 4.9-7.2% higher plasma total homocysteine vs. the referent genotype). In each case, the association of the genotype with homocysteine was partially mediated by nutrients; when plasma folate and vitamin B-6 or B-12 biomarkers were added to the models, the regression coefficients were reduced by 29% for FTCD rs2277820, by 43% for SLC19A1 rs1051266, rs1131596, and rs4819130, and by 34% for SLC19A3 rs13007334 (data not shown). A model containing a nonredundant set of 3 of the top 5 FDR-significant SNPs (FTCD rs2277820, SLC19A3 rs13007334, SLC19A1 rs1051266) explained 3.6% of the variation in plasma homocysteine beyond that explained by age, smoking, and folate, B-6, and B-12 residuals (data not shown); the set of 3 SNPs was statistically significant (LRT = 17.6, P = 0.0005, 3 degrees of freedom, df), and the coefficients for each SNP were similar to coefficients from single SNP models. Considering the MTHFR genotype in more detail, the TT genotype group (vs. CC) had elevated homocysteine (nominal P = 0.0052), but the CT genotype had no association with homocysteine (nominal P = 0.8107); thus, the MTHFR genotype did not pass preset FDR thresholds.
In models investigating interactions between each SNP and MTHFR
rs1801133, 4 interaction terms were below the FDR threshold (FDR-adjusted P value ≤ 0.2) for the homocysteine phenotype (Additional file 4
). No SNP--nutrient (folate, B-6, or B-12) interaction coefficients reached FDR-significance (FDR-adjusted P value ≤ 0.2; Additional file 5
). The MTHFR
--folate interaction did not reach preset statistical thresholds (pnominal
= 0.0578), but the pattern of interaction supported a greater association of MTHFR TT
genotype with homocysteine conditional on lower folate status.
Global genomic DNA methylation phenotype: Alu elements
In analyses of the Alu element methylation phenotype, 8 SNPs were statistically significant with a nominal P ≤ 0.005; however, none were statistically significant at the FDR threshold (FDR-adjusted P value ≤ 0.05) (Table ). There was little or no mediation of the association by nutrients or plasma homocysteine levels (data not shown). There were no SNP--nutrient interactions with folate or B-12 that reached FDR thresholds for statistical significance (FDR-adjusted P ≤ 0.2) (Additional file 6
). Three SNPs had an FDR-significant interaction with plasma vitamin B-6 (Additional file 6
); these interactions involved 3 intronic SNPs in 2 genes, aminomethyltransferase (AMT
, rs1464567 and rs1464566) and DNA (cytosine-5-)-methyltransferase 3 beta (DNMT3B
, rs1883729). Comparing men with the AMT
genotype to the GG
genotype, the mean Alu element methylation was 0.4 SD higher at low B-6, 0.1 SD higher at median B-6, and 0.4 SD lower at high B-6. Comparing men with the AMT
genotype to the AA
genotype, the mean Alu element methylation was 0.4 SD higher at low B-6, 0.1 SD higher at median B-6, and 0.3 SD lower at high B-6. Comparing men with the DNMT3B
genotype to the AG/GG
genotype, the mean Alu element methylation was 0.1 SD lower at low B-6, 0.3 SD higher at median B-6, and 0.8 SD higher at high B-6.
Global genomic DNA methylation phenotype: LINE-1 elements
No SNP main effect associations reached the FDR-significance threshold for LINE-1 methylation (FDR-adjusted P ≤ 0.05; Table ). There were no SNP--nutrient interactions for folate or B-12 that reached FDR-significance levels (FDR-adjusted P ≤ 0.2) (Additional file 7
). An interaction of plasma B-6 with 1 SNP was significant at the FDR threshold of P ≤ 0.2 (rs17080689, an intronic SNP in methylenetetrahydrofolate dehydrogenase (NADP+ dependent) 1-like, MTHFD1L
) (Additional file 7
), suggesting that the relation of the SNP to LINE-1 methylation varied according to plasma levels of vitamin B-6. Comparing participants with the MTHFD1L
genotype to the CC/AA
genotype, mean LINE-1 element methylation was 0.6 SD higher at low B-6, 0.2 SD higher at median B-6, and 0.4 SD lower at high B-6.