White Matter Microstructure in Relation to Education ...

571

Journal of Alzheimer’s Disease 17 (2009) 571–583 DOI 10.3233/JAD-2009-1077 IOS Press

White Matter Microstructure in Relation to Education in Aging and Alzheimer’s Disease

1

Stefan J. Teipela,b,∗ , Thomas Meindlc , Maximilian Wagnerb , Thomas Kohlb , Katharina Bu¨ rgerb , Maximilian F. Reiserc , Sabine Herpertza , Hans-Ju¨ rgen Möllerb and Harald Hampelb,d,e a

Department of Psychiatry, University Rostock, Rostock, Germany Department of Psychiatry, Ludwig-Maximilian University, Munich, Germany c Department of Clinical Radiology, University Hospitals – Grosshadern, Ludwig-Maximilian University, Munich, Germany d Discipline of Psychiatry, School of Medicine and Trinity College Institute of Neuroscience (TCIN), Trinity College, Dublin, Ireland e The Adelaide and Meath Hospital incorporating the National Children’s Hospital (AMiNCH), Dublin, Ireland b

Accepted 6 January 2009 Communicated by Diana Woodruff-Pak

Abstract. The reduced risk of dementia in high-educated individuals has been suggested to reflect brain reserve capacity. In the present study, we determined the association between integrity of white matter microstructure and education separately in twentyone patients with clinically probable Alzheimer’s disease (AD) and 18 healthy elderly subjects. We used fractional anisotropy derived from high-resolution diffusion-tensor weighted imaging at 3 Tesla as an in vivo marker of white matter microstructure. Based on multivariate network analysis, more years of education were associated with reduced white matter integrity of medial temporal lobe areas and association fiber tracts when age, gender, and dementia severity had been controlled for (p < 0.001). In controls, higher education was associated with greater white matter integrity in medial temporal lobe areas and association fiber tracts (p < 0.001). In multiple regression models, education was the main factor accounting for fiber tract integrity even when occupation was taken into account. Reduced fiber tract integrity with higher education at the same level of cognitive impairment in AD patients and higher fiber tract integrity with higher education in similar white matter areas in cognitively healthy controls agrees with the hypothesis that white matter microstructure may contribute to brain reserve capacity in humans. Keywords: Aging, Alzheimer’s disease, brain reserve capacity, cortical connectivity, diffusion tensor imaging (DTI), education

INTRODUCTION Many epidemiological studies suggest a protective effect of higher education for the onset of dementia in elderly subjects [1–4]. Education may provide brain re1 An abstract of this work was presented at the International Conference on Alzheimer’s Disease, 27–31 July 2008, Chicago (Hot Topic Session II, Oral presentation). ∗ Corresponding author: Stefan J. Teipel, M.D., Department of Psychiatry and Psychotherapy, University Rostock, Gehlsheimer Str. 20, 18147 Rostock, Germany. Tel.: +49 1149 381 494 9610; Fax: +49 1149 381 494 9682; E-mail: [email protected].

serve against neurodegeneration [5]. Consistently, patients with Alzheimer’s disease (AD) with more years of education exhibit more functional [6,7] or structural brain lesions [8] in imaging studies at a comparable level of dementia severity compared to patients with lower education. It is not clear whether this represents a direct effect of education or if this effect is mediated by the effect of education on general health status [9] or occupational achievement [10]. Education is a risk factor of AD that is potentially accessible to intervention. Important risk factors for AD that are not accessible to intervention include age, female gender, and apolipoprotein E4 (ApoE4) genotype [11,12].

ISSN 1387-2877/09/$17.00  2009 – IOS Press and the authors. All rights reserved

572

S.J. Teipel et al. / Education and white matter microstructure

Diffusion tensor imaging (DTI) allows the determination of the integrity of subcortical fiber tracts in the living human brain [13]. Fractional anisotropy (FA), derived from DTI, is sensitive towards subcortical fiber changes with brain maturation [14,15] as well as towards fiber degeneration in AD [16–27]. Therefore, FA provides an interesting in vivo marker to determine the association between education and brain structure in aging and AD. In the present study, we investigated whether fiber tract integrity was associated with years of education in patients with AD using observer independent multivariate analysis [27]. We hypothesized that AD patients with more years of education would present a higher degree of fiber degeneration in regional predilection sites of AD pathology compared to patients with less years of education when we controlled for dementia severity, age, and gender. This would indicate that patients with higher education could maintain similar cognitive function compared to patients with lower education despite more severe impairment of fiber tract integrity. If the effect of education on brain reserve capacity was mediated by white matter microstructure, one would expect that white matter microstructure was increased with higher education in healthy control subjects. We tested this hypothesis in a group of healthy elderly subjects controlling for established risk factors of AD including, age, gender, and ApoE4 genoytpe. We used a multiple regression model to determine whether the effects of education on fiber tract integrity were dependent on occupational achievement.

SUBJECTS AND METHODS Subjects We examined 21 patients with the clinical diagnosis of probable AD (58–87 years of age; mean age: 76.1 (SD 7.4) years, 12 women) according to NINCDSADRDA criteria [28]. For comparison, we investigated 18 cognitively healthy elderly subjects (56–83 years of age; mean age: 66.2 (SD 7.3) years, 8 women). The AD and control groups differed significantly in age (ttest: p = 0.001, T = −4.20, 37 degrees of freedom), but showed a similar gender distribution (χ 2 = 0.63, 1 degree of freedom, p = 0.43). The Mini-Mental Status Examination (MMSE) was used to assess the degree of overall cognitive impairment [29]. Groups differed significantly in MMSE scores, with 22.9 (17–29, SD 3.0) points in AD and 29.1 (28–30, SD 0.6) points

in control subjects (p < 0.001, Mann-Whitney U = 8.5). Years of education were significantly different between AD patients and controls (p < 0.03, T = 2.4, 37 degrees of freedom) with 11.0 (8–18, SD 1.9) years in AD patients and 13.2 (8–20, SD 3.7) years in controls. Years of education were assessed as years attending school plus years of apprenticeship, technical school, college, and university. Occupation was determined as the most recent occupation or the last occupation before retirement. We used a modification of the Hollingshead occupational rating scale [30,31]. This scale has nine categories of occupational status ranging from unlearned worker to high level academic professions or senior manager level. As the number of subjects in our study was small and the range of professions limited, we collapsed the nine dimensions into four, leaving the category of unlearned worker (coded as 1), combining categories 2 to 4 (coded as 2), 5 to 7 (coded as 3) and 8 to 9 (coded as 4). The clinical assessment included detailed medical history; clinical, psychiatric, neurological, and neuropsychological examinations (CERAD battery [32], Clock-drawing-test [33], trail-making test [34]); and laboratory tests (complete blood count, electrolytes, glucose, blood urea nitrogen, creatinine, liver-associated enzymes, cholesterol, HDL, triglycerides, serum B12, folate, thyroid function tests, coagulation, serum iron). Additionally, ApoE4 genotyping was available in 17 of the 21 AD patients and in all control subjects, with 7 AD patients and 7 controls carrying at least one ApoE4 allele. Selection of subjects included a semiquantitative rating of T2-weighted MRI scans [35]. Only subjects who had no subcortical white matter hyperintensities exceeding 10 mm in diameter or 3 in number were included. All patients and controls were only examined if they gave their written informed consent. The study was approved by the institutional review board of the Medical Faculty of the University of Munich. MRI acquisition MRI acquisitions of the brain were conducted with a 3.0 Tesla scanner with parallel imaging capabilities (Magnetom TRIO, Siemens, Erlangen, Germany), maximum gradient strength: 45 mT/m, maximum slew rate: 200 T/m/s, 12 element head coil.


Subjects were scanned in a single session without changing their position in the scanner. The following sequences were used: for anatomical reference, a sagittal high-resolution 3-dimensional gradient-echo sequence was performed (magnetization prepared rapid gradient echo MPRAGE, field-of-view 250mm, spatial resolution 0.8 × 0.8 × 0.8 mm 3 , repetition time 14 ms, echo time 7.61 ms, flip angle 20 ◦ , number of slices 160). To identify white matter lesions a 2-dimensional T2-weighted sequence was performed (fluid attenuation inversion recovery FLAIR, field-of-view 230 mm, repetition time 9000 ms, echo time 117 ms, voxel size 0.9 × 0.9 × 5.0 mm, TA 3:20 minutes, flip angle 180 ◦, number of slices 28, acceleration factor 2). Diffusion-weighted imaging was performed with an echo-planar-imaging sequence (field-of-view 256 mm, repetition time 9300 ms, echo time 102 ms, voxel size 2.0 × 2.0 × 2.0 mm 3 , 4 repeated acquisitions, b-value 1 = 0, b-value 2 = 1000, 12 directions, noise level 10, slice thickness 2.0 mm, 64 slices, no overlap). DTI data processing Pre-processing: DTI data were preprocessed using the DTI toolbox of the FSL software (http://www.fmrib. ox.ac.uk/fsl/, written mainly by members of the Analysis Group, FMRIB, Oxford, UK, Version 3.2). After correcting for susceptibility artifacts [36], from the 12 gradient directions we derived a 12x12-tensor to extract eigenvalues and eigenvectors to determine fractional anisotropy maps [27]. The processing was implemented within Matlab 7.6 (MathWorks, Natwick, Mass.) through Statistical Parametric Mapping [37,38] (SPM 2, Wellcome Department of Imaging Neuroscience, London; available at http://www.fil.ion.ucl.ac.uk/spm),as described in a previous study [27]. In brief, we used low-dimensional normalization with a set of nonlinear basis functions [39,40] and highdimensional normalization with symmetric priors [41] to normalize the anatomical MPRAGE scans into standard space. The normalization parameters were sequentially applied to FA maps that had been spatially coregistered using affine transformation to the anatomical MPRAGE scans in native space. This procedure resulted in FA maps projected into standard space. Statistical analysis Spatial effects of education: Voxels from outside the white matter were removed from the spatially nor-

573

malized FA maps by means of a mask derived from the white matter maps of the anatomical MPRAGE scans [42]. The masked FA maps were smoothed with a 12-mm FWHM Gaussian kernel. Images were scaled to the same mean value and standard deviation using a voxel-wise z-transformation. The multivariate approach followed three subsequent steps that will only briefly be described in the following section. For further details we refer to [27,43]. First, the high-dimensionally normalized FA maps were subjected to principal component analysis within the AD and control groups. Second, we determined the significance of the hypothesized effect of education on FA values using multivariate analysis of covariance (MANCOVA). We employed a linear model with years of education as independent factor, controlling for MMSE score, age, and gender in the AD patients and for age, gender, and ApoE4 genotype in the control group, and the principal components from the FA maps as multivariate dependent variable. Third, we characterized the spatial distribution of these effects using canonical variate analysis in terms of the canonical vector that best captured the effect of education. To this end, we defined canonical images in the observation space such that the variance ratio between the effect of interest and the total error sum of squares was maximized. Each canonical image has an associated canonical value that serves to estimate whether a particular canonical image is important. The canonical value can be compared to an F distribution with denominator degrees of freedom equal to the rank of the matrix of the effect of interest and nominator degrees of freedom equal to the number of scans minus the rank of the design matrix (= degrees of freedom of the error term). We considered a canonical image important if its canonical value exceeded the critical F threshold for p < 0.05. Calculations were carried out using an algorithm written in MATLAB v. 7.6 (Mathworks, Newton, MA). Effects of education controlled for occupation: To find the principal component that was significantly associated with years of education, we correlated the principal component scores of each scan with education. The scores from the principal component that were significantly correlated with education represented scalar markers for the effect of education on FA maps for each subject. Multiple regression models were applied with the principal component scores as dependent variables and age, gender, MMSE score (in the AD patients), education, and occupation as independent variables. In

574


the first step all independent variables were forced into the model; in subsequent steps those variables were removed that did not significantly contribute to the fit of the model according to the overall F-test. Robustness measures: To assess the stability of our multivariate solution, we followed an approach suggested by Zuendorf et al. [44]. First, we tested the data set for potential outliers, using Hotellings T 2 test [45], a multivariate generalization of Students t-test. T2 gives a measure of the distance of a vector from a vector whose elements represent the means of a multivariate normal distribution. In our case, it was given by T 2 = (n − 1)y T · y with n = number of subjects, y = a column vector of PC scores belonging to one subject whose elements are divided by the square root of the corresponding eigenvalues. The T2 statistics is related to the F-distribution by: p = Fn−p

(n − p) 2 T p(n − 1)

with n = number of subjects and p = number of principal component scores used for the test. Additionally, we carried out repeated MANCOVAs based on seven subsets of 18 scans each after randomly removing three different scans each time from the full data set in the AD patients and based on 6 subsets of 15 scans each after randomly removing three different scans each time from the full data set in the controls. The resulting overall effects of the MANCOVAs were compared to the χ 2 distribution for p < 0.001 at 7 df in AD patients and at 6 df in controls. Identification of voxel with peak loading The 98th percentile was used as threshold for voxel with highest positive or negative loading on the canonical image. The Talairach-Tournoux coordinates [46] of these voxel were determined using in-house software written in C. The program produces a listing of all local maxima and minima of the canonical image with each local peak at least 1 cm away from neighboring peaks.

RESULTS Effects of education in AD patients We assessed the significance of the effect of education with a design matrix having 21 rows (one for each

Table 1 Voxel with positive peak loadings on the second canonical image in AD patients, representing lower fractional anisotropy with higher education Region Inferior frontal gyrus

Side L

Superior temporal gyrus Lingual gyrus Cerebellum

L L L

Insula Inferior temporal gyrus Inferior parietal lobule Lingual gyrus Fusiform gyrus

R R R R R

Cerebellum

R

Coordinates (mm) x y z −38 26 −1 −37 4 18 −61 −50 21 −16 −83 5 −4 −86 −24 −11 −74 −12 −31 −78 −15 35 3 15 30 −14 −35 45 −35 24 23 −71 −6 36 −33 −17 47 −60 −18 37 −80 −16 55 −64 −20

Voxel with positive loading above the 98th percentile. Brain regions are indicated by Talairach and Tournoux coordinates, x, y, and z [46]: x = the medial to lateral distance relative to midline (positive = right hemisphere); y = the anterior to posterior distance relative to the anterior commissure (positive = anteror); z = superior to inferior distance relative to the anterior commissure-posterior commissure line (positive = superior). R/L = right/left.

scan) and 5 columns, one for the effect of education, and one each for the effects of the intercept, MMSE scores, age, and gender. The data were reduced to the first 7 eigenvectors having eigenvalues greater than unity. The resulting matrix was subject to MANCOVA. Wilk’s lambda after transformation corresponded to χ 2 = 36.2 with 7 degrees of freedom, p < 0.001. The effect of education was almost completely accounted for by the second canonical image. The canonical val1 ue was 272.2 and was larger than the threshold F 16 = 246 for p < 0.05. We divided the canonical image in one image containing only voxels with high positive loadings (positive components) and one image containing only voxels with high negative loadings (negative components) on the canonical image. Subsequently, we considered only the negative components representing a decline of FA associated with more years of education. Figure 1 shows the negative components of the second canonical image projected onto the T1-weighted MRI template in standard space (in red). The 80th percentile threshold was used to show only the most relevant features. Table 1 shows the Talairach-Tournoux coordinates of the voxel with the highest negative loadings on the canonical image lying above the 98th percentile. The canonical image compromised medial temporal lobes (includ-

S.J. Teipel et al. / Education and white matter microstructure (A)

575 (B)

Fig. 1. Effect of education on FA values. The component of the second canonical image of the AD patients (red colors) representing lower FA values with more years of education after controlling for age, gender, and MMSE score, and the component of the second canonical image of the controls (blue colors) representing higher FA values with more years of education after controlling for age, gender, and ApoE4 genotype, are shown. The components of the FA maps in voxel space are projected on the rendered axial sections (A), and on a coronal slice through the medial temporal lobe at Talairach-Tournoux coordinate y = −12 (B) of the T1-weighted template brain. Axial sections go from ventral at Talairach-Tournoux coordinate z = −17 to dorsal at z = 38, sections are 5 mm apart. Right of image is right of brain (view from posterior in the coronal section, from superior in the axial section). The white arrows point to the right and left parahippocampal gyrus on the coronal slice.

ing fusiform gyrus, parahippocampal gyrus, and hippocampus), subcortical white matter of insula cortex, as well as lateral temporal, occipital, frontal, and parietal lobes white matter. Additionally, the canonical image compromised white matter of the posterior lobe of the cerebellum. When we used the sum score from the CERAD battery based on z-score transformation of each subtest according to [47] instead of the MMSE score to control for cognitive impairment, the significance of the results and the spatial distribution of the effects were

unchanged (Fig. 2). Inclusion of ApoE4 genotype as covariate in the analysis of the effect of education on FA in the subset of 17 AD patients where ApoE4 was available did not alter the significance or spatial distribution of effects (data not shown). Effects of education in the controls We assessed the significance of the effect of education with a design matrix having 18 rows (one for each scan) and 5 columns, one for the effect of education,

576


Fig. 2. Effect of education on FA values controlling for MMSE score or CERAD sum score. The negative components of the second canonical image controlling for MSME score (red) or controlling for CERAD sum score (blue) are shown, representing lower FA values with more years of education after controlling for age, gender, and cognitive impairment. Overlapping areas are in pink. The effects are projected on the rendered axial sections of the T1-weighted template brain. Axial sections go from ventral at Talairach-Tournoux coordinate z = −17 to dorsal at z = 38, sections are 5 mm apart. Right of image is right of brain (view from superior in the axial sections). Please note that controlling for MMSE score or for CERAD score leads to nearly identical spatial pattern; compare also Fig. 1. (Colours are visible in the electronic version of the article at www.iospress.nl.)

and one each for the effects of the intercept, ApoE4 genotype, age, and gender. The data were reduced to the first 7 eigenvectors having eigenvalues greater than unity. The resulting matrix was subject to MANCOVA. Wilk’s lambda after transformation corresponded to χ 2 = 39.5 with 7 degrees of freedom, p < 0.001. The

effect of education was almost completely accounted for by the second canonical image. The canonical value was 820.0 and was larger than the threshold F 113 = 243 for p < 0.05. We divided the canonical image in one image containing only voxels with high positive loadings (positive components) and one image contain-

S.J. Teipel et al. / Education and white matter microstructure Table 2 Voxel with positive peak loadings on the second canonical image in controls, representing higher fractional anisotropy with higher education Region Middle frontal gyrus

Side L

Insula Superior temporal gyrus Postcentral gyrus Fusiform gyrus Precuneus Cuneus Corpus callosum Thalamus Projecting on fasc. occipito-frontalis Pons Middle frontal gyrus Medial frontal gyrus

L L L L L L L L L L R R

Anterior cingulate Cingulate gyrus Uncus Parahippocampal gyrus

R R R R

Superior temporal gyrus Fusiform gyrus Inferior parietal lobule Middle occipital gyrus Lingual gyrus

R R R R R

Fornix Thalamus

R R

Coordinates (mm) x y z −24 5 37 −24 23 30 −26 −10 44 −49 10 34 −42 −1 1 −30 6 −30 −33 −34 54 −50 −28 −25 −8 −55 50 −20 −89 25 −24 17 19 −18 −26 15 −27 −32 21 −6 32 2 16 20 10 30 21 36 10 47 40 36 10 7 9 1 15

−19 38 45 39 20 13 −7 41 −20 −9 1 −12 −30 −98 −84 −102 −16 −23

−36 34 −12 17 29 42 −32 −5 −18 −29 0 −38 41 16 −9 −9 17 10

Voxel with negative loading above the 98th percentile. Brain regions are indicated by Talairach and Tournoux coordinates, x, y, and z [46]: x = the medial to lateral distance relative to midline (positive = right hemisphere); y = the anterior to posterior distance relative to the anterior commissure (positive = anterior); z = superior to inferior distance relative to the anterior commissure-posterior commissure line (positive = superior). R/L = right/left.

ing only voxels with high negative loadings (negative components) on the canonical image. Subsequently, we considered only the negative components representing a higher FA associated with more years of education (Fig. 1 (blue) and Table 2). The canonical image compromised medial temporal lobes (including fusiform gyrus, parahippocampal gyrus, and uncus), subcortical white matter of insula, middle and medial frontal cortex, as well as lateral temporal, parietal, and occipital lobes white matter. Additionally, the canonical image compromised white matter of the pons and thalamus.

577

Effect of education relative to occupation In the AD patients, the third and the 17th principal components, in the controls the fourth principal component were significantly correlated with education (Fig. 3). Within the AD group, occupation was only significantly correlated with the scores for the 17th principal component, within the control group occupation was significantly correlated with the scores for the 4th principal component (Fig. 3). After stepwise backward selection only education remained in the model for both principal component 1 = 4.7, p < 0.05 for the scores in the AD group (F 19 1 third, F19 = 11.0, p < 0.005 for the 17th principal component) and for the 4th principal component in the 1 = 11.7, p < 0.003). healthy control group (F 16 Robustness measures Hotelling’s T2 test demonstrated the absence of outliers at the p = 0.25 level (suggesting a low type II error of erroneously assuming no outliers in the data) both in the AD patients and in the controls (Fig. 4). Repeated MANCOVAs based on seven subsets of 18 scans each in AD patients and six subsets of 15 scans each in the controls, respectively, showed that the effect of education controlling for dementia severity, age and gender was above the critical threshold for all subsets of the data (Fig. 5).

DISCUSSION We assessed the association between years of education and subcortical fiber tract integrity in patients with AD. The study is based on a large body of evidence that years of education impacted the age of onset and rate of progression of dementia in elderly subjects. Fiber tract integrity in characteristic predilection sites of AD was more impaired with higher education (controlled for dementia severity, age, and gender). In a multiple regression model, education, not occupation, was the main predictor of fiber tract integrity. In a group of healthy elderly subjects, after controlling for age, gender, and ApoE4 genotype as major risk factors of AD, we found that higher education was associated with higher fiber tract integrity in several brain areas, including areas that showed reduced fiber tract integrity with higher education in AD. Fractional anisotropy was more reduced in fiber tracts of parietal, temporal, and occipital lobes, includ-

578


(A)

(B) Fig. 3. Correlation between principal component scores and education. The third and 17th principal components of the FA maps of the AD patients (upper row) and the 4th principal component of the FA maps of the controls (lower row) are correlated with education at the level of p < 0.05. Vertical lines indicate the threshold for the correlation coefficient at p < 0.05. Red bars indicate the correlation coefficients between occupation and those principal component scores that were significantly associated with education. (Colours are visible in the electronic version of the article at www.iospress.nl.)

ing fusiform and parahippocampal gyrus, with more years of education. These fiber tracts have previously been found to be involved in AD in independent samples [16–27], suggesting that the FA reductions in our

sample reflect lesions to predilection sites of AD. The finding of increased lesion load in patients with higher education is in line with positron emission tomography (PET) studies showing reduced parieto-


579

(A)

(B) Fig. 4. Hotelling T2 -test for the first 8 principal components. Hotelling T2 -test for the first 8 principal component scores showing no outliers at p = 0.25 for the AD patients (A) and the controls (B).

temporal cortical metabolism [6] and increased frontal lobe amyloid binding [7] in AD patients with higher compared to patients with lower education at the same level of dementia severity. In an MRI-based study, AD patients with higher education showed more ventricular enlargement compared to patients with lower educa-

tion after controlling for dementia severity [8]. These observations suggest that patients with higher education maintain a comparable level of cognitive function despite more severe structural and functional brain lesions. Our study provides the first evidence for education

580


p < 0.001 threshold at 7 df

(A)

p < 0.001 threshold at 6 df

(B) Fig. 5. Wilk’s lambda for subsets of the data, randomly leaving out three different scans each. Distribution of Wilk’s lambda after transformation corresponding to χ2 at 7 (A) or 6 (B) df derived from subsets of the data after randomly removing three different scans each from the entire data set of the AD patients (A) or controls (B). (Colours are visible in the electronic version of the article at www.iospress.nl.)

related differences in the fiber tract integrity that may underlie differences in brain reserve capacity. DTI is a sensitive technique to assess mechanisms of brain maturation [14,15]. The neuron populations that mature

later during brain development are the earliest affected by characteristic AD neuropathology [48]. Environmental factors influence the formation and myelination of axons during early childhood and adolescence [49].


Brain maturation may even proceed into the fifth decade of life [50]. It is tempting to assume that the changes in FA reflect the more severe changes to originally stronger myelinated and more richly connected fiber tracts in patients with higher education that are required to induce the same level of dementia in these patients. This interpretation would agree with the finding that some of the brain areas regions that were negatively associated with education in the AD group were positively associated with education in the healthy control group, including medial temporal lobe areas (entorhinal cortex and hippocampus), fusiform gyrus, insula, superior temporal gyrus, and lingual gyrus. These areas belong to typical predilection sites of AD pathology and seem to have higher fiber tract integrity with higher education in cognitively healthy elderly subjects after controlling for major risk factors of AD. The spatial pattern, however, additionally involved pons and postcentral gyrus white matter, suggesting that the effect of education was not specific to predilection sites of AD in healthy elderly controls. Given the age range of the healthy control subjects, we cannot exclude the possibility that some of these subject harbor preclinical AD. Therefore, to further address the question whether education contributes to fiber tract integrity, one will have to study a group of young healthy subjects (adolescence to early adulthood) comparing fiber tract integrity in higher and lower educated subjects. However, the association between education and onset of dementia and severity of underlying structural and functional brain lesions is likely more complex. A young person’s educational achievement is influenced by their innate intelligence, as well as environmental factors, such as their parents’ educational and socioeconomic background, and available schooling. In turn, education influences the subsequent occupation, environment, and the likelihood for continued mentally and physically healthy and stimulating activities (such as continuous learning, exercise, and nutrition). All these factors in themselves may contribute to neurogenesis, and the organization, myelination, and maintenance of subcortical fiber tracts over the lifetime. Complexity of occupational work lowered the risk of dementia in monozygotic twin pairs discordant for AD [10], where complexity of work was related to educational achievement. We used linear multiple regression models to determine whether the effect of education was mediated by occupation. Interestingly, education was the main predictor of fiber tract integrity. We would, however, need larger samples to allow for cross validation of these effects. Therefore, education still serves

581

as an adjunct for the overall concept of brain reserve capacity. Within this conceptual framework, our findings suggest that the integrity of subcortical fiber tracts underlies brain reserve capacity. One could argue that the association between education and dementia is accounted for by an ascertainment bias: higher education could raise skills of subjects to score higher on diagnostic tests [51]. However, in an earlier study subjects with higher education showed higher rates of decline in cognitive performance compared to lower educated subjects [52]. This observation would be unlikely if performance in cognitive tests was mainly determined by educational background once patients got demented. Years of education in our sample were similar to those in a previous study on the association between education levels and cortical glucose consumption in a memory clinic sample of 96 AD patients [6], suggesting that our subjects represent a typical memory clinic sample. Although the number of scans was small in comparison to the dimensionality of the imaging data, Hotellings T2 statistic suggested that the DTI data represented a homogeneous data set and that the significance of the multivariate solution remained stable across different subsets of the data. We had excluded AD patients with evidence for significant cerebrovasuclar disease based on the rating of white matter hyperintensities, to be able to study the association of education and white matter microstructure in predominant neurodegenerative disease. Inclusion of patients with extended white matter changes would be required to further study the association between education and cerebrovascular disease as a measure of vascular reserve capacity. In summary, our findings agree with the hypothesis that education contributes directly to brain reserve capacity. They complement earlier findings on the structural and functional basis of reserve capacity of the human brain by providing insight into integrity of subcortical fiber tracts. The maturation of cerebral fiber tracts is known to proceed at least into early adulthood if not even into higher age. Our findings support a close association between mechanisms of brain maturation and plasticity. Therefore, our data should stimulate further research on the dynamic basis of brain reserve capacity to help in the design of interventions to strengthen individual brain reserve capacity in order to prevent the initiation and progression of pathological mechanisms in AD.

582


ACKNOWLEDGMENTS Part of this work was supported by grants of the Medical Faculty of the Ludwig-Maximilian University (Munich, Germany) to S.J.T., of the Hirnliga e. V. (Nürmbrecht, Germany) to S.J.T., an investigator initiated unrestricted research grant from Janssen-CILAG (Neuss, Germany) to H.H. and S.J.T., and a grant from the Bundesministerium f u¨ r Bildung und Forschung (BMBF 01 GI 0102) awarded to the dementia network “Kompetenznetz Demenzen”. There are no conflicts of interest associated with the work presented in this article. The corresponding author had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

[10]

[11]

[12]

[13]

[14]

[15] [16]

REFERENCES [1]

[2]

[3]

[4] [5]

[6]

[7]

[8]

[9]

Ampuero I, Ros R, Royuela A, Abraira V, del Ser T, GarciaRibas G, Garcia de Yebenes J (2008) Risk factors for dementia of Alzheimer type and aging-associated cognitive decline in a Spanish population based sample, and in brains with pathology confirmed Alzheimer’s disease. J Alzheimers Dis 14, 179-191. McDowell I, Xi G, Lindsay J, Tierney M (2007) Mapping the connections between education and dementia. J Clin Exp Neuropsychol 29, 127-141. Chin AV, O’Connell H, Kirby M, Denihan A, Bruce I, Walsh JB, Coakley D, Lawlor BA, Cunningham C (2006) Co-morbid and socio-demographic factors associated with cognitive performance in an elderly community dwelling Irish population. Int J Geriatr Psychiatry 21, 1150-1155. Katzman R (1993) Education and the prevalence of dementia and Alzheimer’s disease. Neurology 43, 13-20. Stern Y, Alexander GE, Prohovnik I, Stricks L, Link B, Lennon MC, Mayeux R (1995) Relationship between lifetime occupation and parietal flow: implications for a reserve against Alzheimer’s disease pathology. Neurology 45, 55-60. Perneczky R, Drzezga A, Diehl-Schmid J, Schmid G, Wohlschlager A, Kars S, Grimmer T, Wagenpfeil S, Monsch A, Kurz A (2006) Schooling mediates brain reserve in Alzheimer’s disease: findings of fluoro-deoxy-glucosepositron emission tomography. J Neurol Neurosurg Psychiatry 77, 1060-1063. Kemppainen NM, Aalto S, Karrasch M, Nagren K, Savisto N, Oikonen V, Viitanen M, Parkkola R, Rinne JO (2008) Cognitive reserve hypothesis: Pittsburgh Compound B and fluorodeoxyglucose positron emission tomography in relation to education in mild Alzheimer’s disease. Ann Neurol 63, 112118. Kidron D, Black SE, Stanchev P, Buck B, Szalai JP, Parker J, Szekely C, Bronskill MJ (1997) Quantitative MR volumetry in Alzheimer’s disease. Topographic markers and the effects of sex and education. Neurology 49, 1504-1512. Del Ser T, Hachinski V, Merskey H, Munoz DG (1999) An autopsy-verified study of the effect of education on degenerative dementia. Brain 122 ( Pt 12), 2309-2319.

[17]

[18]

[19]

[20]

[21]

[22]

[23]

[24]

Andel R, Crowe M, Pedersen NL, Mortimer J, Crimmins E, Johansson B, Gatz M (2005) Complexity of work and risk of Alzheimer’s disease: a population-based study of Swedish twins. J Gerontol B Psychol Sci Soc Sci 60, P251-258. Gao S, Hendrie HC, Hall KS, Hui S (1998) The relationships between age, sex, and the incidence of dementia and Alzheimer disease: a meta-analysis. Arch Gen Psychiatry 55, 809-815. Slooter AJ, Cruts M, Kalmijn S, Hofman A, Breteler MM, Van Broeckhoven C, van Duijn CM (1998) Risk estimates of dementia by apolipoprotein E genotypes from a populationbased incidence study: the Rotterdam Study. Arch Neurol 55, 964-968. Beaulieu C (2002) The basis of anisotropic water diffusion in the nervous system - a technical review. NMR Biomed 15, 435-455. Lebel C, Walker L, Leemans A, Phillips L, Beaulieu C (2008) Microstructural maturation of the human brain from childhood to adulthood. Neuroimage 40, 1044-1055. Huppi PS, Dubois J (2006) Diffusion tensor imaging of brain development. Semin Fetal Neonatal Med 11, 489-497. Stahl R, Dietrich O, Teipel SJ, Hampel H, Reiser MF, Schoenberg SO (2007) White matter damage in Alzheimer’s disease and in mild cognitive impairment: assessment with diffusion tensor MRI using parallel imaging techniques. Radiology 243, 483-492. Bozzali M, Falini A, Franceschi M, Cercignani M, Zuffi M, Scotti G, Comi G, Filippi M (2002) White matter damage in Alzheimer’s disease assessed in vivo using diffusion tensor magnetic resonance imaging. J Neurol Neurosurg Psychiatry 72, 742-746. Fellgiebel A, Wille P, Muller MJ, Winterer G, Scheurich A, Vucurevic G, Schmidt LG, Stoeter P (2004) Ultrastructural hippocampal and white matter alterations in mild cognitive impairment: a diffusion tensor imaging study. Dement Geriatr Cogn Disord 18, 101-108. Head D, Buckner RL, Shimony JS, Williams LE, Akbudak E, Conturo TE, McAvoy M, Morris JC, Snyder AZ (2004) Differential vulnerability of anterior white matter in nondemented aging with minimal acceleration in dementia of the Alzheimer type: evidence from diffusion tensor imaging. Cereb Cortex 14, 410-423. Rose SE, Chen F, Chalk JB, Zelaya FO, Strugnell WE, Benson M, Semple J, Doddrell DM (2000) Loss of connectivity in Alzheimer’s disease: an evaluation of white matter tract integrity with colour coded MR diffusion tensor imaging. J Neurol Neurosurg Psychiatry 69, 528-530. Takahashi S, Yonezawa H, Takahashi J, Kudo M, Inoue T, Tohgi H (2002) Selective reduction of diffusion anisotropy in white matter of Alzheimer disease brains measured by 3.0 Tesla magnetic resonance imaging. Neurosci Lett 332, 45-48. Yoshiura T, Mihara F, Ogomori K, Tanaka A, Kaneko K, Masuda K (2002) Diffusion tensor in posterior cingulate gyrus: correlation with cognitive decline in Alzheimer’s disease. Neuroreport 13, 2299-2302. Fellgiebel A, Muller MJ, Wille P, Dellani PR, Scheurich A, Schmidt LG, Stoeter P (2005) Color-coded diffusion-tensorimaging of posterior cingulate fiber tracts in mild cognitive impairment. Neurobiol Aging 26, 1193-1198. Müller MJ, Greverus D, Dellani PR, Weibrich C, Wille PR, Scheurich A, Stoeter P, Fellgiebel A (2005) Functional implications of hippocampal volume and diffusivity in mild cognitive impairment. Neuroimage 28, 1033-1042.

S.J. Teipel et al. / Education and white matter microstructure [25]

[26]

[27]

[28]

[29]

[30] [31]

[32]

[33]

[34]

[35]

[36]

[37]

Naggara O, Oppenheim C, Rieu D, Raoux N, Rodrigo S, Dalla Barba G, Meder JF (2006) Diffusion tensor imaging in early Alzheimer’s disease. Psychiatry Res 146, 243-249. Medina D, Detoledo-Morrell L, Urresta F, Gabrieli JD, Moseley M, Fleischman D, Bennett DA, Leurgans S, Turner DA, Stebbins GT (2006) White matter changes in mild cognitive impairment and AD: A diffusion tensor imaging study. Neurobiol Aging 27, 663-672. Teipel SJ, Stahl R, Dietrich O, Schoenberg SO, Perneczky R, Bokde AL, Reiser MF, Moller HJ, Hampel H (2007) Multivariate network analysis of fiber tract integrity in Alzheimer’s disease. Neuroimage 34, 985-995. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM (1984) Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of the Department of Health and Human Services Task Force on Alzheimer’s disease. Neurology 34, 939-944. Folstein MF, Folstein SE, McHugh PR (1975) Mini-mentalstate: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189-198. Hollingshead AB (1975) Four factor index of social status., Department of Sociology, Yale University New Haven, CT. Davis J, Smith T, Hodge R, Nakao K, Treas J (1991) Occupational prestige ratings from the 1989 general social survey. Ann Arbor MI: Inter-university Consortium for Political and Social Research. Berres M, Monsch AU, Bernasconi F, Thalmann B, Stahelin HB (2000) Normal ranges of neuropsychological tests for the diagnosis of Alzheimer’s disease. Stud Health Technol Inform 77, 195-199. Shulman KI, Shedletsky R, Silver IL (1986) The challenge of time: Clock drawing and cognitive function in the elderly. Int J Geriatr Psychiatry 1, 135-140. Chen P, Ratcliff G, Belle SH, Cauley JA, DeKosky ST, Ganguli M (2000) Cognitive tests that best discriminate between presymptomatic AD and those who remain nondemented. Neurology 55, 1847-1853. Scheltens P, Barkhof F, Leys D, Pruvo JP, Nauta JJP, Vermersch P, Steinling M, Valk J (1993) A semiquantitative rating scale for the assessment of signal hyperintensities on magnetic resonance imaging. J Neurol Sci 114, 7-12. Mangin JF, Poupon C, Clark C, Le Bihan D, Bloch I (2002) Distortion correction and robust tensor estimation for MR diffusion imaging. Med Image Anal 6, 191-198. Friston K, Holmes AP, Worsley K, Poline J-B, C.D. F, Frackowiak RSJ (1995) Statistical parametric maps in functional imaging: a general linear approach. Hum Brain Mapp 2, 189210.

[38]

[39] [40]

[41]

[42]

[43]

[44]

[45] [46] [47]

[48]

[49] [50]

[51]

[52]

583

Friston K, Ashburner J, Frith CD, Poline J-B, Heather JD, Frackowiak RSJ (1995) Spatial registration and normalization of image. Hum Brain Mapp 2, 165-189. Ashburner J, Friston KJ (2000) Voxel-based morphometry–the methods. Neuroimage 11, 805-821. Ashburner J, Neelin P, Collins DL, Evans A, Friston K (1997) Incorporating prior knowledge into image registration. Neuroimage 6, 344-352. Ashburner J, Andersson JL, Friston KJ (1999) Highdimensional image registration using symmetric priors. Neuroimage 9, 619-628. Ashburner J, Friston K (1997) Multimodal image coregistration and partitioning–a unified framework. Neuroimage 6, 209-217. Friston KJ, Poline JP, Holmes AP, Frith C, Frackowiak RSJ (1996) A multivariate analysis of PET activation studies. Hum Brain Map 4, 140-151. Zuendorf G, Kerrouche N, Herholz K, Baron JC (2003) Efficient principal component analysis for multivariate 3D voxelbased mapping of brain functional imaging data sets as applied to FDG-PET and normal aging. Hum Brain Map 18, 13-21. Hotelling H (1931) A generalization of Student’s ratio. Ann Math Stat 2, 360-378. Talairach J, Tournoux P (1988) Co-Planar Stereotaxic Atlas of the Human Brain., Thieme, New York. Chandler MJ, Lacritz LH, Hynan LS, Barnard HD, Allen G, Deschner M, Weiner MF, Cullum CM (2005) A total score for the CERAD neuropsychological battery. Neurology 65, 102-106. Arendt T (2000) Alzheimer’s disease as a loss of differentiation control in a subset of neurons that retain immature features in the adult brain. Neurobiol Aging 21, 783-796. Levitt P (2003) Structural and functional maturation of the developing primate brain. J Pediatr 143, S35-45. Bartzokis G, Beckson M, Lu PH, Nuechterlein KH, Edwards N, Mintz J (2001) Age-related changes in frontal and temporal lobe volumes in men: a magnetic resonance imaging study. Arch Gen Psychiatry 58, 461-465. O’Connor DW, Pollitt PA, Treasure FP (1991) The influence of education and social class on the diagnosis of dementia in a community population. Psychol Med 21, 219-224. Bruandet A, Richard F, Bombois S, Maurage CA, Masse I, Amouyel P, Pasquier F (2008) Cognitive decline and survival in Alzheimer’s disease according to education level. Dement Geriatr Cogn Disord 25, 74-80.