This is a novel study aimed to discover characteristic voxel-wise patterns of brain atrophy associated with carrying the commonly prevalent MTHFR C677T gene variant, in 3D.
We found that a very commonly carried variant in the MTHFR gene, which is associated with high homocysteine levels in the blood, is significantly associated with brain structure variation, in particular with lower regional brain volumes, in subjects with MCI in both the ADNI and CHS cohorts. As these cohorts were independently collected and assessed, the results strongly corroborate our findings.
Consistent with our hypotheses, our main findings were:
(a) carrying MTHFR risk alleles was associated with having smaller brain volumes in MCI subjects from the ADNI cohort (); (b) carrying MTHFR risk alleles was associated with an accelerated rate of regional brain atrophy () in the same MCI subjects from the ADNI cohort followed up after 12 months; (c) the MTHFR associations with lower regional brain volumes were replicated in an independent set of subjects with mild cognitive impairment from the CHS cohort (); (d) the regions of associations were not influenced by carrying ApoE4 alleles in either cohort; (e) folate levels were controlled for in the CHS study and did not affect the results, so the MTHFR SNP effect on brain volumes was not dependent on the folate levels; (f) plasma homocysteine levels possibly mediate the effect: adjusting for homocysteine, as expected, weakens the associations of the MTHFR risk alleles with brain volumes in both cohorts.
The affected periventricular fronto-parietal white matter () in ADNI and parieto-occipital white matter brain regions in CHS () were identified as associated with homocysteine levels in our prior work on ADNI MCI subjects (Rajagopalan et al., 2011
). These are regions in which white matter hyperintensities (associated with vascular infarcts) are most commonly detected in elderly individuals (Enzinger et al., 2006
). However, the associations with white matter hyperintensities were not significant in our subjects. This is quite interesting in itself. It could either be that the WMH measure used in ADNI is fairly noisy — as it depends to some extent on thresholding the intensities of the scans to pick up what can sometimes be a very diffuse effect. Or it may be that the very subtle interval changes in brain volume are more precisely quantified by methods such as TBM, and the precision in the measures makes it easier to pick up a correlation with other pertinent biomarkers. There is some evidence in the latter direction, as drug trials using WMH volumes as an outcome measure sometimes need very large sample sizes, at least relative to measures of volumetric atrophy, which are easier to measure precisely.
It is somewhat disappointing that the brain regions showing associations with the MTHFR
allele did not overlap in the ADNI and the CHS datasets after statistical thresholding, although the allele effects are likely more pervasive than the regions that survive thresholding. The CHS MCI cohort tended to show more posterior periventricular deficits when compared to the ADNI MCI cohort that showed in the fronto-parietal region, after adjusting for age and sex in the regression models. Also, the MTHFR
risk allele associated brain volume deficits were somewhat stronger in the CHS (up to 12%; ) than in ADNI (up to 6%; ). These findings may reflect differences in the demographics of the two cohorts; the subtypes of MCI subjects in the ADNI (includes only amnestic MCI) is somewhat different from the CHS (includes probable and possible amnestic MCI). Effects of MCI subtypes may influence the strength and location of correlations between brain structure and the MTHFR
risk allele. Also, as we used FDR to assess effects in the brain, we did not expect a strong congruence in the localization of effects, bearing in mind that the peaks of effect size in any one sample depend strongly on the noise and any unmodeled biological variation in the data. A similar situation was found in Kohannim et al. (2012)
, where we found significant, diffuse effects of the autism risk gene, MACROD2
, on brain volumes in two different cohorts, but not in exactly the same locations in the brain. As such, we did not make a strong a priori
hypothesis about the exact location of the effects, as any clusters in the statistical fields should not be considered the only places where biological effects are found, but simply evidence that there is a distributed effect on the brain. The overall significant associations between the MTHFR
allele and brain volumes, in ADNI and CHS subjects alike, probably provide evidence that there is a distributed effect of the MTHFR
allele on the brain; thereby providing support for considering replication of the MTHFR
associations with brain volumes.
When controlling for the effects of the ApoE4
allele, in addition to age and sex, there was no effect on the associations of the MTHFR
risk allele and brain structure. ApoE4
risk alleles are associated with temporal lobe volume deficits and ventricular expansion (Hua et al., 2008a; Schuff et al., 2009
), but the profile of these deficits does not appear to interact with, or underlie, the associations of MTHFR
risk allele. Therefore, MTHFR
risk alleles are likely to exert their influence on brain structure independent of ApoE4
Folate is essential for the stability and synthesis of myelin basic protein, which, in turn, is essential for white matter structure in the central nervous system. It acts as a coenzyme for a large number of metabolic reactions in the body and affects brain structure independent of plasma homocysteine levels. Also, the effects MTHFR
variant on plasma homocysteine levels are pronounced when serum folate levels are low. Therefore, we adjusted for folate in the MTHFR
SNP associations with brain structure in the CHS study but found that it did not affect the associations. In the ADNI study – unlike the CHS – folate levels were not measured in the subjects; so we could not test for MTHFR
associations with brain structure, independent of folate levels in ADNI subjects. Folate supplementation is known to improve cognitive performance in the elderly with dementia and elevated plasma homocysteine (Nilsson et al., 2001
). We were unable to look into folate supplements as we did not have the necessary data in this imaged cohort.
When we adjusted for homocysteine levels in ADNI and CHS, the correlations for the MTHFR
C677T risk allele with brain volumes weakened, suggesting that the MTHFR
effect on brain structure may in part be due to homocysteine elevation induced by the risk allele. The MTHFR
risk allele may also directly affect brain structure independent of the homocysteine pathway. The MTHFR
risk allele that is known to increase homocysteine levels has been shown, albeit inconsistently, to be associated with AD (Anello et al., 2004; Brunelli et al., 2001; Mansoori et al., 2011; Prince et al., 2001; Wang et al., 2005
), vascular dementia (Chapman et al., 1998; Pollak et al., 2000
), silent brain infarcts (Kohara et al., 2003
) and white matter hyperintensities (Hong et al., 2009; Kohara et al., 2003
). This, by itself, may increase the MTHFR
variant associated risk for brain atrophy through various mechanisms including plaques and tangles (Vermeer et al., 2007
), in addition to increased homocysteine levels.
Our finding of the MTHFR
gene variant associations with brain atrophy may have implications for randomized controlled trials of medications aiming to lower homocysteine levels to resist brain degeneration. Some clinical trials target the folate pathway specifically, and advocate taking high doses of B vitamins as a preventative measure against further brain atrophy (Smith et al., 2010
) and cognitive decline (Jager et al., 2011
) in subjects with mild cognitive impairment. In such trials, and in epidemiological studies, it may help to genotype at this MTHFR
SNP, as these genotypes may influence homocysteine levels and the observed pattern of atrophy. Further cross-sectional and prospective studies will be helpful in replicating these findings.
Our study was carried out in a large well-characterized cohort (ADNI n = 359; CHS n = 51) with high-quality standardized brain MRI, and well-validated computational methods to map patterns and rates of brain atrophy at a local level. We found significant associations of the MTHFR SNP with brain volumes in the subjects diagnosed with MCI in both ADNI and CHS studies, even though the replication sample was relatively small (n = 51).
The minor allele frequency for the MTHFR
SNP varies moderately across different populations worldwide, and is reported to be 31% in Western and Central Europeans, 33% in US Chinese, 16% in US Gujarati Indians, 41% in US Mexicans, and 12% in African Americans (Altshuler et al., 2010
). The subjects analyzed in the study were of Caucasian origin in both the cohorts to avoid spurious results due to population stratification (Lander and Schork, 1994
). As a result, care must be exercised in generalizing our findings to ethnic groups with different allele frequencies and possibly different environmental influences.
Homocysteine levels in the CSF and plasma are highly correlated (r
= 0.85) (Selley et al., 2002
) but it is difficult to ascertain whether plasma homocysteine and folate levels correlate between blood and cerebrospinal fluid, especially when their values are not very high, as in the subjects studied here. Therefore in addition to elevated homocysteine, several other mechanisms such as dietary methionine intake, toxic habits, and pyridoxal-phosphate levels may contribute towards the effect of the MTHFR
SNP on brain structure. Unfortunately, these measures were not available to us in the current study.
These brain maps reveal that a commonly carried susceptibility allele for higher homocysteine is also associated with structural brain volume deficits and with accelerated rates of brain atrophy over time. As homocysteine is involved in this pathway and in the level of atrophy, the deficits may be resisted even in MTHFR
gene risk allele carriers, at least in principle, via efforts to lower homocysteine levels. Healthy lifestyle changes such as increasing dietary folate and supplementation with B vitamins (Blaise et al., 2007; Smith et al., 2010
) in carriers of the MTHFR
risk variant may help slow the rate of atrophy, especially in elderly subjects with mild cognitive impairment.