Vitamin B12, folate, and homocysteine are implicated in pivotal neurodegenerative mechanisms and partake in elders’ mental decline. Findings on the association between vitamin-related biochemistry and cognitive abilities suggest that the structural and functional properties of the brain may represent an intermediate biomarker linking vitamin concentrations to cognition. Despite this, no previous study directly investigated whether vitamin B12, folate, and homocysteine levels are sufficient to explain individual neuropsychological profiles or, alternatively, whether the activity of brain regions modulated by these compounds better predicts cognition in elders. Here, we measured the relationship between vitamin blood concentrations, scores at seventeen neuropsychological tests, and brain activity of sixty-five elders spanning from normal to Mild Cognitive Impairment. We then evaluated whether task-related brain responses represent an intermediate phenotype, providing a better prediction of subjects’ neuropsychological scores, as compared to the one obtained considering blood biochemistry only. We found that the hemodynamic activity of the right dorsal anterior cingulate cortex was positively associated (
Vitamin deficiencies due to dietary habits, drug interactions, or genetic factors are quite common in the elderly and may contribute to the worsening of cognitive decline [
Despite the large amount of studies pointing to the same direction, other observations cast doubt on the reliability of the relationship between vitamin-related compounds and cognitive status [
To clarify whether and how vitamin-related compounds affect the cognitive status of elders, it may be necessary to consider also information related to brain structure and function as measured by
However, in sharp contrast with the abundance of evidence linking biochemical markers to brain morphological alterations, less is known about the relationship with brain functioning [
The two alternative hypotheses for the relationship between blood biochemistry, brain activity, and cognitive profiles. (a) Scheme depicting the potential role of B12, folate, tHcy, and their interactions as direct predictors of distinct cognitive profiles. These profiles are represented by principal components (PC) derived from subjects’ scores at seventeen neuropsychological tests. Age, years of education, Mini Mental State Examination scores and fMRI task accuracy are included in the model as nuisance variables. (b) Alternatively, brain hemodynamic activity can act as an intermediate phenotype in linking blood biochemistry to subjects’ cognitive profiles. First, brain regions modulated by B12, folate, tHcy, and their interactions are identified and hemodynamic activity is subsequently used as a predictor of cognitive profiles. The same nuisance variables are included in the model at both steps of analysis. tHcy: homocysteine; Fol: folate; MMSE: Mini Mental State Examination.
From the initial cohort of a longitudinal project [
In accordance with the European Consortium on Alzheimer’s Disease Working Group on MCI criteria [
Subjects gave their written informed consent to take part in the study and had the right to withdraw at any time. Protocol and procedures were approved by the local Ethical Committee for Clinical Experimentation and the study was conducted in accordance with the Declaration of Helsinki.
Prior to the analyses, we discarded nine subjects due to the following reasons: (i) their B12, folate, and tHcy values were higher or lower than 3 standard deviations from the group average (recursive outlier detection procedure;
To characterize the cognitive state of our sample, expert neuropsychologists administered a comprehensive battery of 18 tests, investigating multiple cognitive domains. A detailed description is provided in Supplementary Materials. Sample characteristics and test results are reported in Supplementary Table
Subjects underwent a fasting blood sampling procedure within one week prior to MRI acquisition. Blood analyses were performed by the Clinical Biochemistry Laboratory of the Azienda Ospedaliera Universitaria Pisana (Pisa, Italy) which also included routine blood tests. B12 and folate serum levels were evaluated by an electrochemiluminescence immunoassay (ECLIA) (Roche Diagnostics, Basilea, Switzerland) on a cobas immunoanalyzer (Roche Diagnostics, Basilea, Switzerland), whereas tHcy was evaluated in plasma by the automated latex enhanced immunoassay HemosIL (Werfen, Barcelona, Spain) on an ACL TOP 500 instrument (Werfen, Barcelona, Spain).
MRI data were acquired on a GE HDxt 1.5 T Signa (General Electric Healthcare) system, equipped with an 8-channel phased-array head coil. For each subject, an exhaustive MR session was performed, including a clinical protocol (i.e., T2w FSE, FLAIR, and T2
The experimental apparatus (i.e., goggles and response pads) was connected to a workstation running MATLAB Release 2010b 64 bit (The MathWorks Inc., Natick, MA, USA). The visuospatial attention task was implemented in the Psychtoolbox v3.0.947 [
Scores at neuropsychological tests were adjusted for age and education and transformed into
In our final sample (
Moreover, we evaluated whether vitamin-related compounds would significantly predict subjects’ cognition, as measured by PCs (Figure
In the whole sample, ranging from healthy to mild cognitive impaired subjects, we tested the relationship between fMRI task performance and age, MMSE scores, clinical diagnosis, and blood biochemistry (see Supplementary Materials). Furthermore, we used 17 independent GLMs to measure the relationship between fMRI task performance and PC scores, while adjusting the results for age, years of education, and MMSE (Bonferroni corrected
MRI data were extensively preprocessed to rule out potential confounds related to physiological, as well as acquisition-related artifacts, and analyzed using ANTs [
Lastly, we assessed whether cerebrovascular activity acts as an intermediate phenotype, indirectly linking blood biochemistry to cognition in nondemented elders. To this purpose, the activity of brain regions showing significant group-level correlation with vitamin-related compounds was used as a predictor of cognitive profiles. Thus, for each of the identified regions, we performed 17 independent GLMs using subjects’ hemodynamic activity as the explanatory variable. Age, years of education, and MMSE scores as well as fMRI task accuracy were included as nuisance regressors and each PC score was included as the response variable. To robustly estimate the statistical significance of each test, we adopted a permutation approach in which both dependent and independent matrices of the GLMs were rearranged (100,000 permutations) by shuffling values within each column (i.e., permuting subject ID), separately. This procedure generated a null distribution of regression coefficients, against which the actual association was tested. The resulting
As depicted in Figure
Neuropsychological assessment results. (a) Correlations among all the seventeen neuropsychological tests as estimated by Pearson’s coefficient. Marked cells refer to significant correlations corrected for multiple comparisons through the False Discovery Rate procedure. (b) Matrix describing the results of the principal component analysis and showing seventeen uncorrelated cognitive profiles, tests loadings, and the proportion of variance explained by each component. As a matter of fact, the first principal component (explained variance: 29.1%) highlighted subjects’ global cognitive performance, since all the neuropsychological tests jointly contributed to it and given its high correlation with MMSE scores (
Results for the relationship between blood biochemistry and subjects’ cognitive performance as measured by the PCs are shown in Table
Predictive power of biochemical markers on cognitive profiles.
PC1 | PC7 | PC12 | PC15 | |||||
---|---|---|---|---|---|---|---|---|
tHcy | .011 | .9322 | -.068 | .6180 | -.138 | .3098 | -.076 | .5790 |
Folate | -.080 | .5563 | .021 | .8841 | .154 | .2616 | ||
B12 | .222 | .1034 | .171 | .2071 | -.147 | .2746 | .007 | .9567 |
B12 |
.189 | .1629 | -.117 | .3926 | .012 | .9254 | ||
B12 |
.037 | .7844 | .066 | .6272 | -.139 | .3076 | -.081 | .5522 |
tHcy |
.149 | .2689 | .007 | .9604 | -.253 | .0618 | ||
Age | -.154 | .2596 | -.175 | .2004 | .038 | .7802 | ||
Education | .101 | .4625 | .233 | .0840 | .031 | .8185 | ||
MMSE scores | -.042 | .7674 | -.042 | .7621 | ||||
fMRI task accuracy | .037 | .7912 | -.029 | .8342 | .126 | .3566 | -.060 | .6629 |
Model: PCj=
Average
Overall, during the execution of the visuospatial attention task, participants recruited a widespread bilateral network (Figure
Brain activity results. (a) fMRI task-evoked activity across subjects entails medial and lateral occipital regions (i.e., primary and motion-sensitive visual cortex), the dorsal parietal attention network (intraparietal sulcus and superior parietal lobule), frontal areas involved in motor control and focusing (ventral and dorsal premotor, as well as the supplementary motor), and nodes of the salience network (e.g., anterior insula). Regions enclosed in black survived the Bonferroni correction for multiple comparisons (
Interestingly, results for the relationship between vitamin-related compounds and brain activity highlighted a significant ROI within the task-positive network: hemodynamic activity of the right dorsal anterior cingulate cortex (dACC; center of gravity:
Resulting model for the relationship between blood biochemistry, right dACC activity, and cognitive profiles. (a) Scheme depicting the role of B12, folate, tHcy, and their interactions, as direct predictors of the cognitive profile expressed by the 11th component (PC). No significant associations were found between biochemical markers and cognitive status. Loadings of each neuropsychological test for the 11th component are color coded: the more intense the red is, the more positive is the loading of a test; vice versa, the more intense the blue is, the more negative the loading. Age, years of education, Mini Mental State Examination scores, and fMRI task accuracy are included in the model as nuisance variables. (b) Activity of the right dACC significantly acts as an intermediate phenotype in linking B12 serum levels to subjects’ spatial attention and search abilities. Indeed, the right dACC hemodynamic activity correlated with B12 serum levels (
Most importantly, dACC activity significantly predicted (
Predictive power of brain activity on cognitive profile. (a) Scatter plot depicting the relationship between hemodynamic activity of the right dorsal anterior cingulate cortex (dACC) and the cognitive profile expressed by the 11th component (PC), related to subjects’ spatial attention and search abilities (i.e., scores at the Trail Making Test A versus delayed recall of the Babcock story). (b) Null distribution of regression coefficients against which the relationship between right dACC activity and cognitive profile expressed by the 11th component was tested. The blue dot indicates the actual relationship between the right dACC and the 11th PC; dotted lines represent
Predictive power of right dACC activity on cognitive profiles.
PC1 | PC7 | PC11 | ||||
---|---|---|---|---|---|---|
Right dACC | .188 | .1467 | .210 | .1049 | ||
Age | -.233 | .0713 | .142 | .2737 | ||
Education | .104 | .4276 | -.023 | .8642 | ||
MMSE scores | -.055 | .6728 | .113 | .3870 | ||
fMRI task accuracy | -.001 | .9918 | -.045 | .7291 | .007 | .9592 |
Model: PCj=
In the present study, we used magnetic resonance imaging, blood biochemistry, and neuropsychological assessment to investigate whether functional characteristics of the brain, mediated by B12, folate, and tHcy concentrations, represent an intermediate phenotype able to significantly predict single-subject cognitive profiles. Our results demonstrated that, in sixty-five nondemented elders performing a visuospatial attention task, B12 levels were positively associated to hemodynamic responses in the right dACC. Crucially, the activity of this brain region, but not B12 levels
Some authors suggested that in old age B12 serum levels together with folate concentrations have a fundamental role in genomic and nongenomic methylation, and their deficiencies trigger homocysteine-mediated neurotoxicity [
This relationship has been further corroborated by reports linking B12 concentrations to brain structural properties, mainly focusing on global parenchymal atrophy and white-matter lesions. Vogiatzoglou et al. showed that healthy elders with lower B12 levels had higher global atrophy scores and increased rate of brain volume loss [
Despite the abundance of structural MRI findings, less is known about how B12 affects brain functioning. To our knowledge, only two studies investigated this issue: one showed that regional homogeneity of resting state hemodynamic activity is reduced in patients suffering B12 deficiencies, especially within the default mode and the cinguloopercular and frontoparietal networks [
In the current report, we found that B12 serum concentrations are positively associated to the right dACC hemodynamic activity during a visuospatial attention task that engages attention-, action-planning-, and motor-control-related brain regions [
In our study, we aimed to overcome the dichotomy imposed by the clinical evidence and to ascertain whether the task-related brain activity mediates the effects of blood biochemistry on individual neuropsychological profiles, regardless of the diagnosis. It is worth mentioning that it was not possible to predict subjects’ performance at neuropsychological tests by simply considering B12 serum concentrations (Figure
In addition, the inclusion of B12 as a confound in the correlation analysis between the right dACC and the cognitive profile did not alter the significance of their association (
From a more general perspective, our results suggest that only task-positive brain activity significantly contributes to solve the indirect relationship between blood biochemistry and cognitive profiles, while the hemodynamic activity of task-negative regions may not (see also Supplementary Materials). Indeed, further studies including different fMRI paradigms (e.g., verbal naming and lexical decision) and blood tests are required to better characterize whether only the task-positive activity represents a proper brain marker, linking biochemistry to cognitive profiles. We trust that the adoption of other tasks (e.g., memory encoding) may unveil relationships between clinical biochemistry, the activity of specific brain regions (e.g., the hippocampus), and subjects’ neuropsychological profiles (e.g., scores at verbal memory tests versus performance at executive functions tests).
Furthermore, it is relevant to note that subjects’ accuracy at the visuospatial attention task recorded during fMRI acquisition did not correlate with any of the cognitive profiles, or with any of the vitamin-related compounds except for the B12
In summary, our study is the first to prove that modelling brain activity as an intermediate factor between biochemical markers and cognition unveils their indirect relationship. This evidence suggests that the endophenotypic approach, successfully adopted to explore associations between genes, brain, and complex behaviors [
The biochemical, neuropsychological, and MRI data used to support the findings of this study are available from the corresponding author upon request.
The authors declare that they have no conflicts of interest.
Luca Cecchetti and Giada Lettieri equally contributed as first authors.
This work was supported by the “Fondazione Pisa” through the project “Train the Brain” (Bando Ricerca Scientifica in Neuroscienze 2007 of Fondazione Cassa di Risparmio di Pisa).
Supplementary Methods: a paragraph detailing the neuropsychological assessment, clinical biochemistry evaluation, and MRI data analysis. Supplementary Results: all the results that are not included in the main manuscript. Appendix: list of the members and affiliations of the Train the Brain Consortium. Supplementary Figure 1: the flowchart for the selection of participants. Supplementary Figure 2: the stimuli employed in the fMRI experiment. Supplementary Figure 3: additional results for the correlation between brain activity and blood biochemistry markers. Supplementary Figure 4: resulting model for the relationship between blood biochemistry, right vmPFC activity, and cognitive profiles. Supplementary Figure 5: resulting model for the relationship between blood biochemistry, bilateral vmPFC activity, and cognitive profiles. Supplementary References. Supplementary Table 1: demographic and neuropsychological data. Supplementary Table 2: predictive power of biochemical markers on cognitive profiles. Supplementary Table 3: predictive power of right dACC activity on cognitive profiles. Supplementary Table 4: predictive power of vmPFC activity on cognitive profiles. Supplementary Table 5: predictive power of right dACC activity on cognitive profiles, controlling for vitamin B12. Supplementary Table 6: predictive power of vmPFC activity on cognitive profiles, controlling for blood biochemistry. Supplementary Table 7: predictive power of fMRI task accuracy on cognitive profiles.
Complete list of the members and affiliations of the Train the Brain Consortium.