Deficient levels of vitamin B12 or folate lead to increased levels of homocysteine, which is a risk factor for thrombosis, microbleeds, strokes, cognitive impairment and neuronal atrophy(Reference Smith and Refsum1,Reference Wang, Ou and Jiang2) . Vitamin B12, or cobalamin, normally contributes to the production of myelin in the central nervous system and fatty acid metabolism. Vitamin B12 is naturally occurring in meat, fish, milk and eggs(Reference Hunt, Harrington and Robinson3). When vitamin B12 is consumed, it first binds to haptocorrin (or transcobalamin I) to protect the vitamin B12 from stomach acid. Intrinsic factor is produced in the stomach and binds to vitamin B12 in the intestines to allow absorption into the ileum enterocytes via the membrane protein cubilin (CUBN)(Reference Kozyraki and Cases4,Reference Alpers5) . Importantly, vitamin B12 is transported to the liver via transcobalamin II, where it takes place in folate/vitamin B12-dependent remethylation to facilitate the conversion of homocysteine to methionine and the subsequent methylation of DNA, proteins and lipids(Reference Williams and Schalinske6).
The literature is currently mixed about the role of vitamin B12 in brain health and late-life adults with or without Alzheimerʼs disease (AD)-related cognitive impairment. On the one hand, vitamin B12 is significantly lower in both plasma(Reference Lopes da Silva, Vellas and Elemans7) and cerebrospinal fluid (CSF)(Reference de Wilde, Vellas and Girault8) of patients with AD v. normally ageing controls. Further, vitamin B12 deficiency in aged, cognitively normal (CN) adults with diabetes was associated with less grey matter (GM) volume in the left middle temporal pole and the left insula(Reference Deng, Wang and Wang9) suggesting that supplementation may be useful. Indeed, clinical trials have found that vitamin B12 supplementation may slow brain atrophy in mild cognitive impairment (MCI), which is often a precursor state to AD(Reference Petersen, Doody and Kurz10), when n-3 fatty acid levels are sufficiently high, perhaps by changing homocysteine levels(Reference Smith, Smith and de Jager11,Reference Douaud, Refsum and de Jager12) . On the other hand, aged adults with mildly elevated plasma homocysteine levels showed less total brain volume after 2 years of daily supplementation with 500 μg of vitamin B12 and 400 μg of folic acid v. placebo tablet(Reference van der Zwaluw, Brouwer-Brolsma and van de Rest13). This complication may in part be due to the APOE4 carrier status, the strongest genetic risk factor for developing AD, which may modify associations between vitamin B12 and regional GM(Reference Lee, Ha and Park14).
For cognitive function, the literature is also mixed regarding vitamin B12 supplementation efficacy or its use as a biomarker to track AD-related cognitive decline. Despite controversy(Reference Garrard and Jacoby15) surrounding the meta-analysis by Clarke et al. (Reference Clarke, Bennett and Parish16) in selecting rigorous clinical trials and sensitive global cognitive measures in normal ageing, meta-analyses indicate that vitamin B12 supplementation may not influence cognitive decline among CN aged adults with type 2 diabetes(Reference Clarke, Bennett and Parish16), perhaps due to the mild nature of cognitive decline in normal ageing and difficulty in controlling for nutritional status(Reference Morris and Tangney17). Vitamin B12 combined with folate does appear to have modest clinical efficacy in CN or MCI participants, however(Reference Butler, Nelson and Davila18). Qin et al. found that individuals in the top quintile for vitamin B12 intake showed increased performance in working memory, but no differences in memory or executive function tests(Reference Qin, Xun and Jacobs19). Thus, clinical status and perhaps underlying features of AD, such as amyloid-β (Aβ) and total tau, may modulate vitamin B12 supplementation. Although vitamin B12 in rodent models is protective against Aβ (Reference Alam, Siddiqi and Chaturvedi20) and total tau fibrillar accumulation(Reference Rafiee, Asadollahi and Riazi21), it is not clear what vitamin B12 is tracking in terms of neural or cognitive outcomes when clinically significant levels of these AD hallmarks are already present in the brain.
Thus, wide variability in vitamin B12 associations with GM atrophy or fluorodeoxyglucose (FDG) metabolism may be due to vascular factors, clinical status, CSF levels of Aβ and total tau, and/or genetic methylation patterns specific to vitamin B12(Reference Rafiee, Asadollahi and Riazi21–Reference Mosconi, Murray and Davies23). Thus, beyond examining the main effects of vitamin B12 on neural outcomes of interest, we examined potential modulators such as: (1) the vascular risk marker homocysteine; (2) established cut-offs of Aβ and total tau accumulation relevant to AD(Reference Shaw, Vanderstichele and Knapik-Czajka24); (3) baseline clinical status and (4) SNP among four a priori selected genes involved in vitamin B12 transport, uptake, and metabolism, CUBN, methylenetetrahydrofolate reductase (MTHFR), 5-methyltetrahydrofolate-homocysteine methyltransferase reductase (MTRR) and transcobalamin II (TCN2)(Reference Surendran, Adaikalakoteswari and Saravanan25). To further elucidate conflicting findings, analyses were separately conducted in two large cohorts spanning North America and Australia that had similar MRI and cognitive data but different proportions of adults without impairment v. MCI or AD.
Materials and methods
Participants
Data from aged adults were obtained from the Alzheimerʼs Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu) and Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing (AIBL) study group(Reference Ellis, Bush and Darby26). The ADNI was launched in 2003 as a public–private partnership, led by principal investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial MRI, positron emission tomography, other biological markers and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. For up-to-date information, see http://www.adni-info.org. Written informed consent was obtained from all ADNI participants at their respective ADNI sites. The ADNI protocol was approved by site-specific institutional review boards. Data were collected in accord with the Helsinki Declaration of 1975. To eliminate the influence of vitamin B12 supplementation, especially in participants who were receiving increased care from physicians, analyses only included participants who did not report taking vitamin B12 supplements or multivitamins, excluding 299 ADNI participants and seventy-one AIBL participants.
Serum biomarkers
Vitamin B12 data were downloaded at baseline for ADNI participants and obtained through a data request from AIBL. Vitamin B12 was assayed via a Siemens ADVIA Centaur XP autoanalyzer immunoassay by Quest Diagnostics as of 16 April 2008 for ADNI participants. For the AIBL participants, vitamin B12 was assayed by the Royal Melbourne Pathology in Melbourne and PathWest Laboratory Medicine WA in Perth via ADVIA Centaur Assay – competitive immunoassay. Blood processing took place between 2007 and 2014. Homocysteine in ADNI was obtained through the ADNI Biomarker Core, using a validated enzyme immunoassay methodology(Reference Trojanowski, Vandeerstichele and Korecka27). Homocysteine in AIBL was obtained through a data request, where all assays were conducted in two laboratories via MMULITE 2000 – competitive immunoassay. ADNI vitamin B12 levels are reported in pg/ml, and AIBL levels were reported from our data request in pmol/l. All AIBL vitamin B12 values were converted to pg/ml for consistency.
APOE genotype
The ADNI Biomarker Core at the University of Pennsylvania conducted APOE ε4 genotyping. APOE genotypic data were downloaded for AIBL participants. We characterised participants as having zero APOE4 alleles, one APOE4 allele or two APOE ε4 alleles.
Amyloid and tau cerebrospinal fluid biomarkers
CSF sample collection, processing and quality control of p-tau, total tau and Aβ 1-42 are described in the ADNI1 protocol manual (http://adni.loni.usc.edu/) and Shaw et al. (Reference Shaw, Vanderstichele and Knapik-Czajka24), where CSF Aβ 1-42 and total tau cut-offs were <192 and >93 pg/ml, respectively. Amyloid and tau markers were only available in a very small subset of AIBL participants, and so these were not assessed.
Neuropsychological assessment
Cognitive testing was available for both ADNI and AIBL. ADNI utilises an extensive battery of assessments to examine cognitive functioning with particular emphasis on domains relevant to AD. A full description is available at http://www.adni-info.org/Scientists/CognitiveTesting.aspx. All subjects underwent clinical and neuropsychological assessment at the time of scan acquisition. Neuropsychological assessments included: The Clinical Dementia Rating sum of boxes (CDR-sob), Mini-Mental Status Exam, Auditory Verbal Learning Test and AD Assessment Schedule – Cognition. A composite memory score encompassing the Auditory Verbal Learning Test, AD Assessment Schedule – Cognition, Mini-Mental Status Exam and Logical Memory assessments was also utilised(Reference Crane, Carle and Gibbons28). Additionally, a composite executive function score comprising Category Fluency – animals, Category Fluency – vegetables, Trails A and B, Digit span backwards, Wechsler Adult Intelligence Scale (WAIS-R) Digit Symbol Substitution, Number Cancellation and five Clock Drawing items was used(Reference Gibbons, Carle and Mackin29). These composite scores were used in formal analyses to represent memory and executive function among subjects. Out of the cognitive tests that were available for ADNI, only Logical Memory – Immediate Recall, Logical Memory – Delayed Recall, Mini-Mental Status Exam and Global CDR scores were available for AIBL, although the same protocols were used. An executive function composite factor was available for AIBL, although it was comprised of CDR-sob, the Stroop test, the FAS test and Category Switch Total(Reference Burnham, Raghavan and Wilson30).
MRI acquisition and pre-processing
MRI scans were available for both ADNI and AIBL. T1-weighted MRI scans were acquired within 10–14 d of the screening visit following a back-to-back 3D magnetisation prepared rapid gradient echo scanning protocol described elsewhere(Reference Jagust, Bandy and Chen31). Images were pre-processed using techniques previously described(Reference Willette, Xu and Johnson32). Briefly, the SPM12 ‘New Segmentation’ tool was used to normalise images and extract modulated GM and white matter (WM) volume maps to Montreal Neurological Institute space. Maps were smoothed with an 8 mm Gaussian kernel and then used for voxel-wise analyses.
Fluorodeoxyglucose-positron emission tomography
FDG-positron emission tomography images were available only for ADNI. FDG-positron emission tomography acquisition and pre-processing details have been described previously(Reference Jagust, Bandy and Chen31). Briefly, 185 MBq of [18-153-F]-FDG was injected intravenously. After 30 min, six 5-min frames were acquired. Frames of each baseline image series were co-registered to the first frame and combined into dynamic image sets. Each set was averaged, reoriented to a standard 160 × 160 × 96 voxel spatial matrix of resliced 1·5 mm3 voxels, normalised for intensity and smoothed with an 8 mm full width at the half maximum (FWHM) kernel. In order to derive the standardised uptake value ratio, pixel intensity was normalised according to the pons since it demonstrates preserved glucose metabolism in AD(Reference Dowling, Hermann and La Rue33). Normalisation to the pons removed inter-individual tracer metabolism variability. The Montreal Neurological Institute template space was used to spatially normalise images using SPM12 (http://www.fil.ion.ucl.ac.uk/spm/software/spm12/).
Genomic data processing and quality control
Genomic data were only available from ADNI. Quality control of these data was conducted by analysing Hardy–Weinberg equilibrium accepted data for Mendelian inheritance errors. From the entire data set, SNP were selected for further analyses based on Hardy–Weinberg equilibrium P > 0·00001, MAF > 0·05 %, call rate 95 %. Samples with >5 % missingness were removed. Sample genotypes were imputed using 1000Genomes data with Shapeit/Impute2 software following the protocol described previously(Reference van Leeuwen, Kanterakis and Deelen34). SNP with call rates <95 % or R 2 ≤ 0·3 were withdrawn, leaving 2 976 223 imputed and genotyped SNP after quality control. Subsequently, we a priori examined SNP that comprised the CUBN, MTHFR, MTRR and TCN2 genes.
Statistical analysis
For voxel-wise analysis, second-level linear mixed models in SPM12 (http://www.fil.ion.ucl.ac.uk/spm/software/spm12/) tested main effects of vitamin B12 on regional GM and WM volume as well as FDG, controlling for age, sex, BMI, baseline diagnosis and APOE4 status. Significance thresholds were set at P < 0·005 (uncorrected) and P < 0·05 (corrected) for voxels and clusters, respectively. Results were considered significant at the cluster level. As described previously(Reference Willette, Bendlin and Starks35), in order to reduce type 1 error, we utilised a GM threshold of 0·2 to ensure that voxels with <20 % likelihood of being GM were not analysed. For GM and WM, Monte Carlo simulations in ClusterSim (http://afni.nimh.nih.gov/afni/doc/manual/3dClustSim) were used to estimate that 462 contiguous voxels were needed for such a cluster to occur at P < 0·05. For FDG voxel-wise analyses, Monte Carlo simulations in ClusterSim were used to estimate that 224 contiguous voxels were needed for such a cluster to occur at P < 0·05.
All genetic association analyses were conducted using PLINK v1.9 (http://www.cog-genomics.org/plink2). The following genes were analysed through linear associations in White participants of European ancestry with a phenotype of the predicted GM and FDG uptake in maxima from voxel-wise analyses: CUBN, MTHFR, MTRR and TCN2. Covariates for PLINK analyses included sex, clinical diagnosis, intracranial volume and APOE4 status. A Holm–Bonferroni threshold for significance was set of 0·05/4, P < 0·0125(Reference Holm36).
Non-voxel linear mixed regression was conducted using SPSS 25 (IBM Corp.) to test vitamin B12 main effects and interactions with baseline diagnosis and APOE4 status, on cognitive scores and biomarkers. Covariates included age, sex and BMI. Years of education were also added as a covariate in models with cognitive scores. Binomial and multinomial logistic regressions were used to assess the OR of a given participant being diagnosed as MCI or AD v. CN, or of diagnosis between CN v. MCI v. AD.
Results
Data summary
ADNI and AIBL clinical, demographics and biomarker data are separately presented in Table 1. A sub-sample of ADNI participants had FDG data, where sub-sample clinical and demographic data are listed in online Supplementary Table S1. Three outliers of each cohort were removed for having vitamin B12 levels 3 standard deviations from the mean. Vitamin B12 values ranged from 99 to 1146 pg/ml for ADNI participants and for each diagnostic group as follows: CN (157–1146 pg/ml); MCI (157–1084 pg/ml) and AD (99–1121 pg/ml). Vitamin B12 values for AIBL participants ranged from 117 to 1149 pg/ml and for each diagnostic group as such: CN (121–1149 pg/ml); MCI (176–1139 pg/ml) and AD (117–895 pg/ml). The reference range is 200–950 pg/ml(Reference Hunt, Harrington and Robinson3,Reference Andrès, Serraj and Zhu37) . In the ADNI sub-sample, eleven participants had values below the reference range and four participants had values above the reference range. In the AIBL sub-sample, sixteen participants had values below the reference range and four participants had values above the reference range.
Table 1. Demographics for Alzheimerʼs Disease Neuroimaging Initiative (ADNI) and Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing (AIBL) participants with grey matter (GM) images†
(Mean values and standard deviations; numbers; percentages)
CN, cognitively normal; MCI, mild cognitive impairment; AD, Alzheimerʼs disease.
**P < 0·01, ***P < 0·001.
† χ 2 Analyses were conducted to examine differences between sex and APOE4 status. ANOVA was otherwise used.
Baseline diagnosis: differences in vitamin B12 levels
Binary and multinomial logistic regression indicated that serum vitamin B12 levels did not predict a higher likelihood of being diagnosed as MCI, AD or cognitively impaired (MCI + AD) v. the CN reference group.
Vitamin B12 and regional grey matter volume
Voxel-wise analyses regressed serum vitamin B12 against regional GM at baseline separately for 144 participants from ADNI and 401 participants from AIBL. For ADNI participants, higher plasma vitamin B12 was correlated with less GM in three clusters (Table 2). The most significant cluster consisted of 507 voxels primarily in the right precuneus and posterior cingulate gyrus (Fig. 1). Other clusters spanned the left middle and inferior frontal gyri. When examining each diagnostic group separately, only the MCI participants showed significant associations between vitamin B12 and GM. Specifically, higher vitamin B12 in MCI was associated with less GM in the right thalamus, precuneus and calcarine cortex (k = 1023).
Table 2. Regional associations of higher serum vitamin B12 and less grey matter volume in the Alzheimerʼs Disease Neuroimaging Initiative*
R, right hemisphere; L, left hemisphere.
* This table depicts regions where all subjects had less grey matter volume per unit increase in vitamin B12. The highest t value for a given cluster of significant, contiguous voxels is shown. For clusters that extended over more than 15 mm, the highest t value in those areas is indicated. Coordinates are in Montreal Neurological Institute atlas space. Brains are oriented in neurological space.
Fig. 1. Brain areas showing less grey matter (GM) corresponding to increased vitamin B12 in Alzheimerʼs Disease Neuroimaging Initiative participants. The graph depicts the relationship at the maximum voxel in the right precuneus. AU, arbitrary units.
For AIBL participants, higher serum vitamin B12 was correlated with more GM (k = 559) in the right amygdala and superior temporal pole. See Table 3 for a full listing of significant clusters. When examining each diagnostic group separately, only the CN participants showed significant associations, with more vitamin B12 associated with more GM in three significant clusters: one in the right superior frontal gyrus (k = 1186), one in the right precuneus (k = 2105) and one in the right supplementary motor area (k = 496).
Table 3. Regional associations of higher serum vitamin B12 and more grey matter volume in the Australian Imaging, Biomarker & Lifestyle Flagship Study of Ageing*
R, right hemisphere.
* This table depicts regions where all subjects had more predicted grey matter volume per unit increase in vitamin B12. The highest t value for a given cluster of significant, contiguous voxels is shown. For clusters that extended over more than 15 mm, the highest t value in those areas is indicated. Coordinates are in Montreal Neurological Institute atlas space. Brains are oriented in neurological space.
In ADNI where CSF was largely available, we then tested interactions with vitamin B12 and binary cut-offs for CSF AD biomarkers(Reference Shaw, Vanderstichele and Knapik-Czajka38), For CSF Aβ 1-42, there was a positive interaction (P < 0·05, family-wise error (FWE) corrected), such that higher vitamin B12 was related to more GM volume in adults with v. without high amyloid accumulation. One cluster was present in the right middle frontal gyrus (k = 522), one in the right supramarginal gyrus (k = 739) and left superior frontal gyrus (k = 464). This interaction was not significantly associated with less GM. For CSF total tau, there was a negative interaction (P < 0·05, FWE corrected), such that vitamin B12 was related to less GM in aged adults with high tau accumulation. One cluster spanned the left superior frontal gyrus (k = 604). This interaction was not significantly associated with more GM.
Vitamin B12 and regional white matter volume
Voxel-wise analysis was used to regress plasma vitamin B12 against regional WM at baseline for ADNI and AIBL participants, separately. In ADNI, higher vitamin B12 was associated with more WM in two small clusters spanning the cerebellum. In AIBL participants, increased vitamin B12 was associated with more WM in one small cluster spanning the left insula. For CSF Aβ 1-42, there was a negative interaction (P < 0·05, FWE corrected), such that vitamin B12 was related to less WM in individuals with high amyloid accumulation. The significant cluster spanned the cerebellum (k = 735). The interaction was not significantly associated with more WM. For CSF total tau, there was a negative interaction (P < 0·05, FWE corrected), such that vitamin B12 was related to less WM in individuals with high tau accumulation. The significant cluster spanned the left cingulum (k = 2024). The interaction was not significantly associated with more WM.
Vitamin B12 and regional fluorodeoxyglucose metabolism
Voxel-wise analyses regressed plasma vitamin B12 concentrations against FDG uptake in 151 ADNI participants. Higher plasma vitamin B12 was correlated with less glucose metabolism in the right calcarine and precuneus, where Fig. 2 illustrates the relationship at the maximum voxel in the right calcarine (24, −48, 6). See Table 4 for a full listing of significant clusters. When conducting the analyses stratified by baseline diagnosis group, no clusters survived correction for CN and AD participants. For MCI participants, higher vitamin B12 was associated with less FDG uptake in one large cluster (k = 2166) spanning the right lingual gyrus, precuneus and posterior thalamus mostly in the pulvinar nucleus.
Fig. 2. Brain areas showing less fluorodeoxyglucose (FDG) metabolism corresponding to increased vitamin B12. The graph depicts the relationship at the maximum voxel in right calcarine cortex. SUVR, standardised uptake value ratio.
Table 4. Regional associations of higher serum vitamin B12 and less fluorodeoxyglucose (FDG) glucose uptake in the Alzheimerʼs Disease Neuroimaging Initiative*
R, right hemisphere.
* This table depicts regions where all subjects had less predicted FDG glucose uptake per unit increase in vitamin B12. The highest t value for a given cluster of significant, contiguous voxels is shown. For clusters that extended over more than 15 mm, the highest t value in those areas is indicated. Coordinates are in Montreal Neurological Institute atlas space. Brains are oriented in neurological space.
Similar to GM, interactions were then tested with vitamin B12 and cut-offs for AD biomarkers. For CSF Aβ 1-42, a positive interaction ((P < 0·05, FWE corrected) indicated that higher vitamin B12 was related to more GM in adults with high amyloid accumulation in one cluster spanning mid-cingulate gyrus (k = 499). This interaction was not significantly associated with less FDG metabolism. For total tau, by contrast, higher vitamin B12 in adults with high tau accumulation was related to less FDG metabolism in the same mid-cingulate cluster (k = 244). The interaction was not significantly associated with more FDG metabolism.
Genotype analyses for vitamin B12 and predicted differences in grey matter and fluorodeoxyglucose
Next, in ADNI, SNP for genes associated with vitamin B12 uptake, transport and metabolism were used as predictors of interest, to see if genotypes might explain the wide variance seen in vitamin B12 associations. Linear regression in PLINK tested the additive genetic model of each SNP for main effect associations with GM and FDG predicted values, from the voxel-wise analyses reported above. Nine SNP in the CUBN gene were significantly associated with GM and passed Holm–Bonferroni correction, seven of which were detrimental for individuals who had the minor allele and two of which were beneficial (Table 5). One SNP, rs7918972 in the CUBN gene, was significantly associated with FDG and passed Holm–Bonferroni correction, which was associated with less FDG uptake (β = −0·094, P = 0·0065). MTHFR, MTRR and TCN2 SNP were not significantly associated with GM or FDG vitamin B12 predicted values.
Table 5. Association between cubilin SNP and predicted grey matter at the maximal voxel in the right precuneus
A1, minor allele.
* β Values represent the difference in the predicted value of grey matter with an increase from no risk alleles to one risk allele or two risk alleles.
For vitamin B12 levels and genotype, minor allele counts of the ten significant CUBN SNP identified in our GM and FDG analyses were not significantly associated with vitamin B12 levels.
Vitamin B12 and associations with cognition and biofluid markers
Vitamin B12 was not significantly correlated with CDR Global scores, CDR-Sob 11, Mini-Mental Status Exam, composite executive factors, or the composite memory factor for ADNI or AIBL participants. There was a significant baseline diagnosis × vitamin B12 interaction for predicting the Auditory Verbal Learning Test learning score for ADNI participants (P = 0·048). Higher vitamin B12 was associated with better scores in AD (β 0·0027 (se 0·001), P = 0·042) but trending worse scores in MCI (β −0·0022 (se 0·001), P = 0·061).
For biomarkers in plasma, homocysteine was first regressed against vitamin B12. The range of homocysteine values for ADNI participants was 3·9–25·9 μmol/l, by diagnosis: CN (6·1–18·1 μmol/l); MCI (3·9–22·0 μmol/l) and AD (6·1–25·9 μmol/l). The range of homocysteine for AIBL participants was 2·8–35·0 μmol/l, by diagnosis: CN (3·0–35·0 μmol/l); MCI (5·5–26·4 μmol/l) and AD (2·8–18·8 μmol/l). Higher vitamin B12 was associated with lower plasma homocysteine levels in ADNI participants (β −0·0034 (se 0·001), P = 0·001) and AIBL participants (β −0·0021 (se 0·001), P < 0·001). For biomarkers in CSF, there were no significant associations between vitamin B12 and CSF total tau, P-tau-181 or Aβ 1-42 for ADNI participants. Only a very small subset of AIBL participants had available CSF data, so these associations were not assessed.
Discussion
We hypothesised that serum vitamin B12 levels may be a useful biomarker for AD-related brain atrophy, hypometabolism and cognitive decline. We originally predicted that there would be an inverse correlation between higher serum vitamin B12 levels and lower levels of AD markers, due to previous work from other groups(Reference Cho, Huang and Lee39,Reference Hooshmand, Mangialasche and Kalpouzos40) . This hypothesis was true for AIBL participants, where higher vitamin B12 was related to more GM in the right amygdala and superior temporal pole, which has been shown to progressively decrease in volume across the AD spectrum(Reference Hanggi, Streffer and Jancke41). This leads us to the hypothesis that vitamin B12 is protective against GM deterioration in the cognitively unimpaired population, as the majority of AIBL participants were cognitively unimpaired. Conversely, in ADNI participants, higher serum vitamin B12 was instead related to less GM volume and FDG metabolism in precuneus, as well as middle and inferior frontal gyri, where these regions show atrophy, hypometabolism and less default mode neural network strength in aged adults with MCI or AD(Reference Jones, Machulda and Vemuri42–Reference Ye, Seo and Kim44). These results contrast with our AIBL findings and lead us to our second hypothesis that vitamin B12 correlates with worse neurological outcomes in the cognitively impaired population, as the majority of the ADNI participants in this cohort were cognitively impaired.
In support of our first hypothesis and the positive correlation between vitamin B12 and neurological outcomes among AIBL participants, factor analysis of dietary patterns among a group of cognitively unimpaired adults showed that diets that were especially rich in vitamin D, vitamin B12 and Zn were associated with increased GM volume in the temporal and frontal cortices(Reference Berti, Murray and Davies22). Additionally, Erickson et al. conducted a study with 3-d food recalls and cross-sectional MRI scans among thirty-two cognitively unimpaired older adults(Reference Erickson, Suever and Prakash45). The group found that the individuals with higher vitamin B12 intake had increased GM in both the left and right superior parietal sulcus. Lastly, in patients with their first lacunar stroke, lower serum vitamin B12 was correlated with more severe Fazekas scale graded periventricular WM lesions(Reference Pieters, Staals and Knottnerus46).
Supplementation with vitamin B12 in individuals with dementia has been inconclusive. For example, there was no significant improvement in memory or cognition in individuals with dementia who had low levels of vitamin B12 and were subsequently supplemented; however, the supplemented individuals showed better verbal fluency scores compared with non-supplemented controls(Reference Eastley, Wilcock and Bucks47). However, the Homocysteine and B Vitamins in Cognitive Impairment (VITACOG) trial showed that B-vitamin supplementation was related to positive cognitive outcomes in individuals in the top tertile of baseline n-3 fatty acids, compared with individuals with lower baseline n-3 fatty acid levels, perhaps due to a synergistic effect on phospholipid production for the brain(Reference Oulhaj, Jernerén and Refsum48).
For further explanation of our second hypothesis, vitamin B12 levels in the context of other disease states and AD may support it being an indicator of pathological load later in a given disease process. For example, plasma vitamin B12 levels were correlated with increased all-cause mortality in women aged 85 years or greater(Reference Mendonça, Jagger and Granic49). In addition, Arendt et al. found that in a population of Danish patients who had serum vitamin B12 tested and recorded in their electronic medical record, patients with higher vitamin B12 had a higher overall cancer risk within 1 year of follow-up(Reference Arendt, Pedersen and Nexo50). The mechanism behind this relationship was not known; however, the authors postulated that vitamin B12 may rise along with malignant processes(Reference Arendt, Pedersen and Nexo50). Also, individuals with hyperlipidaemia and non-insulin-dependent diabetes had significantly higher levels of vitamin B12 compared with healthy controls(Reference Wasilewska, Narkiewicz and Rutkowski51). Additionally, higher serum vitamin B12 may be a sign of its decreased cellular uptake(Reference Pennypacker, Allen and Kelly52). Vascular damage is a common feature of AD(Reference Kirschen, Kéry and Ge53), and the vascular endothelium via the CD320 receptor may mediate the homoeostasis between the serum and tissue homoeostasis of vitamin B12(Reference Hannibal, Bolisetty and Axhemi54). Finally, we found that higher serum vitamin B12 was related to better memory performance and more GM in AIBL and seemed to be driven by CN participants, which make up a majority of the AIBL cohort. By contrast, higher vitamin B12 was associated with worse memory performance, less GM and less FDG in MCI participants, which constitute the majority of the ADNI1 cohort. These differing patterns of association may be due to MCI v. CN participants usually having more tau accumulation(Reference Jack, Knopman and Jagust55).
As an alternative or concurrent explanation for our ADNI findings, it has also been shown that vegetarians, though unlikely to be classified as clinically deficient, are more likely to have low-normal levels of vitamin B12(Reference Gilsing, Crowe and Lloyd-Wright56). This could likely be extended to individuals who consume plants as a larger portion of their meals, compared with meat. Perhaps, vitamin B12 status may act as an indicator of the ratio of meat intake:plant intake in the diet, and this may manifest in predicting AD outcomes, which have also been linked to plant-based dietary habits(Reference Pistollato, Iglesias and Ruiz57). It is still puzzling that we found a negative relationship between vitamin B12 and neural outcomes in light of the inverse correlation between serum vitamin B12 and homocysteine levels. Hooshmand et al. found that among 2570 individuals 60 years and older, higher serum methionine:serum total homocysteine ratios were associated with lower risk for dementia and AD and higher serum vitamin B12 was positively correlated with methionine:homocysteine ratios. The vast majority of patients in this study were free of dementia or AD(Reference Hooshmand, Refsum and Smith58). Potentially in our ADNI cohort, vitamin B12 still is taken up by the liver in individuals with higher serum vitamin B12 to reduce homocysteine production from methionine, but their high vitamin B12 levels may be indicative of other tissues, such as the brain, being unable to transport vitamin B12 or utilise it properly.
Alternatively, genetic polymorphisms in B vitamin uptake, transport and metabolism may modify how vitamin B12 is utilised and affect vitamin B12 levels themselves. We found that minor allele polymorphisms in the CUBN gene tracked the association between high vitamin B12 and less regional GM. Besides its role in receptor-mediated uptake of vitamin B12 in the ileum, CUBN is also an apo receptor and is involved in the absorption of high-density lipoproteins in the kidney(Reference Hammad, Stefansson and Twal59). Perhaps, CUBN polymorphisms, some of which we have shown are associated with higher serum vitamin B12 levels, may lead to decreased apo reuptake in the kidneys, which may increase dementia risk.
There are several limitations of this secondary data analysis study. First, only vitamin B12 was measured in the participants. Ideally, holotranscobalamin and methylmalonic acid would also have been measured, which would indicate the level of vitamin B12 available for cellular uptake, and the absence of vitamin B12 from necessary methylation reactions, respectively(Reference Harrington60). Current practice in the healthcare field is to test serum vitamin B12 levels or holotranscobalamin first (if a patient has risk factors for vitamin B12 deficiency); a serum vitamin B12 level of <200 pg/ml is considered deficient(Reference Hunt, Harrington and Robinson3). Subsequently, current practice is for clinicians to test an additional metabolic indicator, either methylmalonic acid or homocysteine(Reference Yetley, Pfeiffer and Phinney61). Second, it is difficult to determine the mechanisms involved that affect an individualʼs vitamin B12 levels. This can be impacted by proton pump inhibitors(Reference Maes, Fixen and Linnebur62), level of animal product intake(Reference Gille and Schmid63) and genetic variability(Reference Surendran, Adaikalakoteswari and Saravanan25) as we have illustrated. Perhaps, the interaction between vitamin B12 and one or more of these variables may play a role in cognitive decline. We were also unable to assess associations with physical activity, circulating vitamin B12, and neurological outcomes, as these data were not collected in ADNI or available to us through AIBL. Additionally, FDG scans, genetic and CSF biomarker data were not available in AIBL, which limited our ability to compare ADNI and AIBL. Lastly, because the overwhelming majority of ADNI participants are of European ancestry, we were only able to reliably test the interactions between genetic data and volumetric/function brain outcomes in those individuals. It would be worthwhile to determine if similar genes are implicated in an African American cohort, as the incidence of AD is much higher among African Americans compared with Whites of European descent(Reference Weuve, Barnes and Mendes de Leon64).
Conclusion
This study showed contrasting associations between vitamin B12 and neurological outcomes among an Australian cohort with a strong cognitively unimpaired makeup and a North American cohort made up mostly of participants with some degree of cognitive impairment. While we found that vitamin B12 was associated with positive structural associations in the brain in the Australian cohort which was mostly cognitively unimpaired, we found that vitamin B12 was correlated with detrimental outcomes among the North American cohort, which consisted mainly of participants with some degree of cognitive impairment. This lends to our hypothesis that vitamin B12 may be protective of neurological decline in the healthy population, but may increase for an undetermined reason in those that are experiencing cognitive decline. Additionally, we have shown that these results may be influenced by genetic mutations related to vitamin B12 uptake, as well as CSF markers of amyloid and total tau accumulation that may reflect or work in concert with baseline clinical diagnosis. Future research should focus on the rate of uptake of vitamin B12 into healthy and diseased neuronal cells, the role of other B vitamin markers in AD onset and progression, and determine if high amyloid or tau accumulation affects the function or efficacy of vitamin B12.
Acknowledgements
This study was funded by Iowa State University, NIH R00 AG047282 and AARGD-17-529552. Neither funding source had any involvement in the report. Data collection and sharing for this project were funded by the Alzheimerʼs Disease Neuroimaging Initiative (ADNI) (National Institutes of Health grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimerʼs Association; Alzheimerʼs Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organisation is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimerʼs Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.
K. E. M. and A. A. W. formulated the research question and design and carried out the study. K. E. M., A. A. W. and A. D. C. M. analysed the data. K. E. M. and A. A. W. wrote the manuscript in consultation with J. P. M. and K. A. All authors provided critical feedback.
The authors have no conflicts of interest to report.
Supplementary material
For supplementary material referred to in this article, please visit https://doi.org/10.1017/S0007114520000951
