409

Journal of Alzheimer’s Disease 39 (2014) 409–422 DOI 10.3233/JAD-131232 IOS Press

White Matter Hyperintensities are Positively Associated with Cortical Thickness in Alzheimer’s Disease Heidi I.L. Jacobsa,b,c,∗ , Lies Clerxa,b , Ed H.B.M. Gronenschilda,b , Pauline Aaltena,b and Frans R.J. Verheya,b a School

for Mental Health and Neuroscience, Alzheimer Centre Limburg, Maastricht University, Maastricht, The Netherlands b European Graduate School of Neuroscience EURON, Maastricht University, Maastricht, The Netherlands c Cognitive Neuroscience, Institute of Neuroscience and Medicine-3, Research Centre J¨ ulich, J¨ulich, Germany Handling Associate Editor: Sebastiaan Engelborghs

Accepted 17 September 2013

Abstract. White matter hyperintensities are associated with an increased risk of Alzheimer’s disease (AD). White matter hyperintensities are believed to disconnect brain areas. We examined the topographical association between white matter hyperintensities and cortical thickness in controls, mild cognitive impairment (MCI), and AD patients. We examined associations between white matter hyperintensities and cortical thickness among 18 older cognitively healthy participants, 18 amnestic MCI, and 17 mild AD patients. These associations were cluster-size corrected for multiple comparisons. In controls, a positive association between white matter hyperintensities and cortical thickness was found in lateral temporal gyri. In MCI patients, white matter hyperintensities were positively related to cortical thickness in frontal, temporal, and parietal areas. Positive associations between white matter hyperintensities and cortical thickness in AD patients were confined to parietal areas. The results of the interaction group by white matter hyperintensities on cortical thickness were consistent with the findings of positive associations in the parietal lobe for MCI and AD patients separately. In the frontal areas, controls and AD patients showed inverse associations between white matter hyperintensities and cortical thickness, while MCI patients still showed a positive association. These results suggest that a paradoxical relationship between white matter hyperintensities and cortical thickness could be a consequence of neuroinflammatory processes induced by AD-pathology and white matter hyperintensities. Alternatively, it might reflect a region-specific and disease-stage dependent compensatory hypertrophy in response to a compromised network. Keywords: Compensation, cortical thickness, dementia, hypertrophy, inflammation, white matter hyperintensities

INTRODUCTION White matter hyperintensities (WMHs), also referred to as leukoaraiosis or leukoencephalopathy, are areas of increased lucency visualized on T2weighted or FLAIR images. WMHs are a common ∗ Correspondence

to: Dr. Heidi I.L. Jacobs, School for Mental Health and Neuroscience, Alzheimer Centre Limburg, Maastricht University, PO BOX 616, 6200 Maastricht, The Netherlands. Tel.: +31 43 388 41 26; Fax: +31 43 388 40 92; E-mail: h.jacobs@ maastrichtuniversity.nl.

magnetic resonance imaging (MRI) finding in brains of older people [1]. The prevalence and severity of WMHs increases with age and their presence have been associated an increased risk of Alzheimer’s disease (AD) [1]. WMHs are also common in individuals with mild cognitive impairment (MCI), a transition stage between cognitively healthy aging and AD [2, 3]. Although not all MCI individuals progress to AD, MCI patients share cognitive and pathological features with AD [4]. The effect that WMHs exert on cognition is still not fully understood, as studies report inconsistent

ISSN 1387-2877/14/$27.50 © 2014 – IOS Press and the authors. All rights reserved

410

H.I.L. Jacobs et al. / White Matter Lesions and Cortical Thickness

findings, ranging from an inverse relationship [2, 5] to no relationship [6]. A functional dissociation between periventricular and subcortical WMH have been reported, in which periventricular WMH would be more related to cognitive dysfunction and subcortical WMH more to emotional disturbances [7–10]. Furthermore, the underlying etiology of WMHs is also under investigation. Pathogenic mechanisms underlying WMHs are heterogeneous and possible roles have been attributed to ischemia, blood-brain barrier dysfunction, microglial activation, changes in cerebral blood flow autoregulation, fiber tract demyelination, gliosis, or toxic effects from amyloid pathology [11]. Again, etiological and neuropathological distinctions have been made between periventricular and subcortical WMH. Periventricular WMH are believed to more related to myelin pallor, while the proposed mechanism underlying the functional consequences of subcortical WMH is believed to be related to microangiopathy and ischemia [12, 13]. Presumably, the effect of WMHs in general on cognition is related to disconnection of functionally related cortical and subcortical structures through fiber tract demyelination and gliosis [14], as indicated by diffusion tensor imaging [15]. This idea is reflected in the disconnection hypothesis, in which white matter breakdown and the associated loss of connectivity is thought to be a major factor in cognitive decline [16]. Brain atrophy is also a common finding in healthy and pathological aging [17–19], and it has been suggested that WMH are associated with brain atrophy [13, 20]. The sequential relationship between WMH and atrophy is not yet known, as it requires longitudinal and large sample size studies. But it has been hypothesized that WMHs induce axonal or white matter tract disruption. Such a tract disruption causes Wallerian and retrograde degeneration and in turn induces distant grey matter atrophy leading to cognitive decline. It may also be that white matter changes are secondary to grey matter changes. Other hypotheses consider white and grey matter changes as two independent processes, potentially arising from a common etiological process [21]. Several studies have investigated the relationship between WMH and grey matter changes and generally found an inverse relationship: a higher level of WMH was associated with lower levels of grey matter volume or density. These associations have been reported in healthy elderly [21, 22] or in dementia patients with a mixed, AD and vascular, etiology [20, 23, 24]. So far, only two studies examined the relationship between WMH and atrophy in AD patients

[20, 23]. Studies investigating distinct relationships between either periventricular or subcortical WMH with grey matter changes provide inconsistent results, ranging from a stronger association for subcortical WMH [25] to a stronger association with periventricular WMH [24]. Independent of the sample size, healthy or demented elderly, a strong relationship between WMH and frontal atrophy has been found [20, 22–24]. However, it is generally known that AD patients show more increased posterior atrophy than healthy elderly [26–30]. The fact that this spatial progression is not seen in association with WMH might be related to the instrument used in measuring WMH or grey matter, or might be related to the fact that patients in various stages of AD are concatenated or not represented. In most of these studies, WMH was measured with visual analog scales [22, 23] and grey matter measures either consisted of large regions of interest [20, 31] or voxelbased morphometric (VBM) analyses [22, 23, 25]. Measuring WMH continuously might provide more information than visual ordinal scales. On the other hand, VBM analyses are associated with limitations, in that regional artifacts can be induced from imperfect spatial normalization [32, 33]. Cortical thickness has the advantage that it measures a physical property, making it easy to interpret, and that it has the capability to detect subtle tissue changes at a sub-millimeter level [34]. Furthermore, to the best of our knowledge, no study so far examined these associations in patients of varying degree of cognitive decline related to AD pathology. Considering the hypothesis that WMHs interrupt neuronal pathways connecting cortical regions, the association between atrophy and WMHs can be expected to show regional differences as well as disease stage differences. Our aim was to investigate the association between total WMH, periventricular WMH, or subcortical WMH volumes and cortical thickness in healthy elderly, patients with amnestic MCI, as these patients are believed to have the highest risk for conversion to AD, and patients with AD. We expected that WMH causes cortical thinning and that this thinning would be increased with disease progression, especially in typical AD-affected areas. MATERIALS AND METHODS Participants Three groups of older male participants were included in this study: 18 healthy participants without memory impairments (mean age: 64.6 years ± 3.4

H.I.L. Jacobs et al. / White Matter Lesions and Cortical Thickness

SD), 18 patients with amnestic MCI (aMCI) (mean age: 65.1 years ± 4.5 SD), and 17 patients with AD (mean age: 70.6 years ± 9.1 SD). MCI and AD patients were recruited from the Memory Clinic of the Maastricht University Medical Center (MUMC+). All participants were administered an extensive neuropsychological test battery and MRI scanning session. All diagnoses were made by a multidisciplinary team under the supervision of an experienced neuropsychiatrist from the Memory Clinic (FRJV) according to the Petersen criteria for aMCI (with at least an impairment in the memory domain (−1.5 SD)) [35], and a Clinical Dementia Rating (CDR) score of 0.5, and according to the DSM-IV and NINCDS-ADRDA criteria for AD [36]. The control group participants were recruited by means of advertisements in local newspapers. Controls were required to have a CDR of 0, no cognitive complaints, and no evidence of cognitive deficits on testing. Hypertensive status was recorded based on the medical history (coded as yes or no). Participants were excluded from the study if they had a history of psychoactive medication use, history of severe depression, abuse of alcohol or drugs, past or present psychiatric or neurological disorders (i.e., epilepsy, stroke, Parkinson’s disease, multiple sclerosis, brain surgery, brain trauma, electroshock therapy, kidney dialysis, heart disease, Meni`ere’s disease, or brain infections), structural abnormalities of the brain, presence of depressive symptoms as indicated by the Hamilton Depression Rating Scale (score ≥17, according to [37]), or contraindications for scanning. A neuroradiologist reviewed the MR images to confirm the absence of clinically relevant neuropathology, such as neoplasms or infarctions. Two control participants were excluded from this study due to structural abnormalities in the brain and were replaced. Education was assessed by using a standardized eight-point scale (1 = primary school, 8 = university). The local Medical Ethics Committee (Clinical Trial Center Maastricht) approved the study (in accordance to the 1964 Declaration of Helsinki and its later amendments) and written informed consent was obtained from all participants and from the primary caregiver of the AD patients. Biomarker status Medial temporal lobe atrophy was also assessed blinded to group adherence by using a qualitative visual rating scale (done by LC) [38]. Rating was performed on coronal T1-weighted images using a 5-point visual

411

Table 1 Demographical and cognitive characteristics of the three groups Controls (n = 18) Age Educational level MMSE score (score) 15 WLT learning (score) 15 WLT memory (score) Fluency animals (number) Hypertension (n (%)) Total WMH (volume, ml) MTA score, left MTA score,right MTA score, sum

64.56 (3.4) 4 (1.4) 28.89 (0.9) 37.50 (7.6) 8.56 (1.9) 23 (5.3) 2 (11.11%) 6.39 (4.1) 0.67 (0.7) 0.44 (0.5) 1.11 (0.9)

MCI (n = 18)

AD (n = 17)

65.11 (4.5) 70.59 (9.1) 4 (1.8) 4 (1.9) 27.61 (2.3) 21.18 (3.9) 26.06 (9.8) 23.47 (11.7) 3.67 (2.8) 1.73 (2.4) 21 (5.4) 13.93 (4.7) 4 (22.22%) 2 (11.77%) 7.93 (11.9) 15.97 (22.9) 2.11 (0.5) 2.88 (0.8) 1.56 (0.7) 3.06 (0.8) 3.67 (1.0) 5.94 (1.3)

Values are mean (sd). AD, Alzheimer’s disease; MCI, mild cognitive impairment; MMSE, Mini-Mental State Examination; MTA, medial temporal lobe atrophy; WLT, word learning task; WMH, white matter hyperintensities.

scale (medial temporal lobe atrophy (MTA) scores), ranging from 0 (no atrophy) to 4 (severe atrophy) based on the height of the hippocampal formation and the surrounding cerebrospinal fluid spaces. According to recent work from our group, patients with a MTA score of 3 or more (left and right scores summed) were considered as positive for AD-risk [39]. In our control group, one person had a score of 3; none of the subjects had a score above 3. In the MCI group, 16 out of 18 patients (89%) has a score equal to or higher than 3. Six of these 18 MCI patients converted to AD within 1.5 year (there was no follow-up data available for 3 patients). All AD patients had a summed MTA score higher than 3. This indicates that the MCI and AD patients included in this study are most likely to have AD pathology. This information is added to the methods section and the mean and standard deviations for left, right, and summed MTA scores are added to Table 1. According to recent criteria [40, 41], these patients can also be termed as prodromal AD patients, which increases the probability that we indeed recruited early AD patients. MRI acquisition For this study, we merged two different datasets, one consisting of the controls and MCI patients and the other containing the AD patients. Both datasets were acquired on the same MRI scanner: a 3.0T whole-body MR system release 2.0 (Philips Achieva, Philips Medical Systems, Best, The Netherlands) using a body coil for RF transmission and equipped with an eight-element head coil (SENSE, factor 2) for signal detection.

412

H.I.L. Jacobs et al. / White Matter Lesions and Cortical Thickness

Anatomical images were acquired with a T1 sequence: TR = 8 ms, TE = 3.7 ms, FA = 8◦ , FOV = 240 × 240 mm, voxel size = 1 mm isotropic, matrix size = 240 × 240 and number of slices = 180. The T1-sequence parameters were the same for each group. For the assessment of WMH, we used a T2 sequence with slice thickness of 3 mm and 0.5 × 0.5 mm voxel size, no slice gap, TE = 100 ms, TR = 2500 ms and FA = 90◦ , and a FLAIR sequence with a slice thickness of 3 mm and 0.5 × 0.5 mm voxel size, no slice gap, TE = 120 ms, TR = 11000 ms, inverse time = 2800 ms and FA = 90◦ . For both sequences the FOV was 256 × 256 mm, the matrix size 512 × 512, the slice thickness 3 mm with no slice gap and the voxel size 0.5 × 0.5 × 3.0 mm in controls and MCI patients. In AD patients, the T2 and FLAIR sequences had a slice thickness of 5 mm, with a voxel size of 0.2246 × 0.2246 and 0.4492 × 0.4992 mm respectively. Cortical thickness analyses Cortical reconstruction was performed with Freesurfer version 4.5.0 on Mac OSX 10.6 [42]. Cortical thickness analyses procedures have been validated against histological analysis and manual measurements and have good test-retest reliability across scanner manufacturers and field strengths [43]. The technical details of these procedures have been described in prior publications (for a recent overview, see [34] or see http://surfer.nmr.mgh.harvard.edu/). Briefly, this processing included motion correction, intensity normalization [44], removal of non-brain tissue, segmentation of the subcortical white matter and subcortical grey matter volumetric structures, tessellation of the grey matter-white matter boundary, automated topology correction, and surface deformation. Once the cortical models were complete, a number of deformable procedures were performed for further processing and analysis, including surface inflation, registration to a spherical atlas, and parcellation of the cortex into units based on gyral and sulcal structure. This method used both intensity and continuity information from the entire volume to produce representations of cortical thickness, calculated as the closest distance from the grey/white to grey/cerebrospinal fluid boundary at each vertex on the surface. The maps produced were not restricted to the voxel resolution and were thus capable of detecting submillimeter group differences. Cortical thickness measures were mapped on the inflated surface of each

participant’s reconstructed brain to allow visualization across the surface without interference from cortical folding. Maps were smoothed using a circularly symmetric Gaussian kernel with a full width half maximum (FWHM) of 20 mm and averaged across participants using a non-rigid high-dimensional spherical averaging method to align cortical folding patterns. This procedure provided accurate matching of morphologically homologous cortical locations among participants on the basis of each individual’s anatomy while minimizing metric distortion, resulting in a mean measure of cortical thickness at each point on the reconstructed surface. Clusterwise correction for multiple comparisons Multiple comparison correction was performed using a cluster-wise procedure described previously [45] and adapted for cortical surface analysis. This procedure, available as part of the FreeSurfer processing stream, is a method that utilizes a simulation to get a measure of the distribution of the maximum cluster size under the null hypothesis. Briefly, a z-map is synthesized and smoothed using a residual FWHM, and then thresholded at a given level. Areas of maximum clusters are then recorded, under these specifications, and the procedure is repeated for a given number of iterations. Only clustered vertices are retained and the assumption is that false positive vertices (i.e., vertices in which a significant relationship between the factor score and thickness is due only to chance) would not appear next to each other. Once the distributions of the maximum cluster size are obtained, correction for multiple comparisons is performed by finding clusters in the statistical maps using the same threshold as was given in the simulation procedure. For each cluster, the p value is the probability of seeing a maximum cluster of that size or larger during the simulation. Clusters remaining in similar areas of significance to the original cortical thickness maps would indicate that the result is not likely due to chance. For these analyses, a total of 5000 iterations of simulation were performed for each comparison, using a threshold of p = 0.05. The simulation cluster analysis was applied to cortical thickness maps for every statistical comparison. Intracranial volume (ICV) Calculation of the ICV with FreeSurfer could cause a bias in patients with severe atrophy [46]. Furthermore, ICV calculations are also influenced by the magnetic field strength of the MRI scanner [47]. Therefore, we

H.I.L. Jacobs et al. / White Matter Lesions and Cortical Thickness

have chosen to calculate ICV from the inner skull contours produced by FSL Brain Extraction Tool (BET) [48] and visually checked. This has been shown to be almost as equally accurate as the manual estimations [47]. White matter hyperintensities The volumes of WMHs were quantified on the FLAIR images using a semi-automatic tool (GIANT) developed at the Maastricht School for Mental Health and Neuroscience (developed by EHBMG; see [49] for more details). In a preparatory step, the algorithm was trained to classify WMHs correctly. This proceeded as follows. First, the voxel intensities of the T2-weighted and FLAIR images were standardized [50] and corrected for intensity non-uniformity [44]. Next, ten T2-weighted and corresponding FLAIR scans with a substantial amount of WMHs were selected. The axial T2-weighted slices and corresponding FLAIR slices were displayed and aligned side by side on the computer screen, allowing visual inspection and easy identification of WMHs, after which the WMHs were traced manually. Finally, these manual tracings were used to derive parameters for the automatic classification of the WMHs. The actual quantification of WMHs was performed semi-automatically. Again, the axial FLAIR and T2-weighted slices were displayed and aligned side by side on the computer monitor. In each slice, a WMHs was indicated manually by clicking within its boundaries, thus generating a seed point and providing starting parameters for a subsequent automatic region growing. Manual corrections were applied when necessary. Finally, the volume of the WMHs (periventricular, subcortical, and total) was calculated. On ten randomly selected brains, the WMHs quantification was performed twice by the same rater (SP) blinded to group membership, which yielded high test-retest reliability (intraclass correlation coefficient = 0.99). Removal of the outliers produced no statistically different results as with inclusion of the outliers. We will therefore present the results with all patients. Statistical analyses Behavioral data analysis was performed with the Statistical Package for the Social Sciences (SPSS Inc., Chicago) version 19.0 for Windows. We first investigated group differences in terms of demographic characteristic, cognition, hypertension, and amount of WMHs using ANOVA for continuous variables and χ2

413

for dichotomous variables. The statistical significance threshold was set at p < 0.05. To approach a normal distribution, WMHs volumes were log-transformed. Since WMHs volumes can be zero and since zero-values cannot be log-transformed, 1 cc was added to all WMHs volumes (WMHsvolume + 1) and this result was log-transformed. Surface maps of cortical thickness effects were generated by applying a general linear model with WMHs volume (total, periventricular, or subcortical) as the independent variable and age and ICV (both centered) as covariates for each group separately. These analyses were repeated with Mini-Mental State Examination (MMSE) scores as extra covariate, to investigate a potential mediating effect of global cognition. We also generated surface maps for the interaction group by WMHs volume on cortical thickness (F-test) to investigate regional differences between the three groups. After the cluster-wise correction for multiple comparisons, the remaining significant clusters for each hemisphere were defined as regions of interest and statistics (mean p-values expressed in -log(p), SD, and cluster size) were extracted. To confirm the patterns of these analyses, we extracted the average cortical thickness values per lobe per subject. The outlines of the four lobes were determined by the Desikan atlas [51]. The WMH maps were divided according to this atlas, in order to be able to extract the amount of WMH volumes per lobe. The association between lobar WMH volume and average cortical thickness was further tested in a multivariate general linear model in SPSS and examined with scatter plots. To examine cortical thickness differences between the three groups, we additionally applied a general linear model to the pair wise group differences at each vertex (with age and ICV centered as a covariate). RESULTS Demographical and cognitive group differences Demographical and cognitive group characteristics are shown in Table 1. The three groups differed significantly with respect to age (MCI < AD, p = 0.03; controls < AD, p = 0.02), MMSE score (controls > MCI, p = 0.03; MCI > AD, p < 0.001; controls > AD, p < 0.001), and score on the delayed recall task of the 15 word learning task (controls > MCI, p < 0.001; MCI > AD, p < 0.05; controls > AD, p < 0.001), but not with respect to educational level, presence of hypertension, or WMHs volumes (ps > 0.05). MTA scores left, right, and

414

H.I.L. Jacobs et al. / White Matter Lesions and Cortical Thickness

summed were significantly different between groups (AD < MCI 0.05, respectively). In the frontal lobe, there is a positive association between WMH load and cortical thickness in MCI patients, no association for controls and a negative association for AD patients (F = 0.034, p > 0.05). Examining the separate contribution of either periventricular or subcortical WMHs showed that the positive effect of periventricular WMHs were more widespread over the cortical surface than subcortical WMHs in both MCI and AD patients (Fig. 3). Nonetheless, there is a substantial amount of overlap between periventricular and subcortical WMHs as indicated by the green overlays. Periventricular WMHs had a very regional specific effect in controls, as it was associated with cortical thinning in the

medial frontal regions. In contrast, subcortical WMHs in controls were associated with higher cortical thickness values in a lateral temporal and lateral parietal cluster. Comparing the effect of WMHs on cortical thickness patterns between the three groups In order to find out with regions of the brain show different patterns among the three groups, we performed an F-test, investigating the interaction between WMHs and group on cortical thickness (see Fig. 4 and Table 2). Significant interactions, where controls showed an inverse association between WMH and cortical thickness, MCI patients showed positive associations, and AD patients showed a trend toward a positive association, were found in the left and right parietal areas and left temporal regions. In the frontal areas, controls and AD patients showed an inverse association between WMH and cortical thickness, but MCI patients showed a positive association. DISCUSSION The aim of this study was to investigate the association between WMHs and cortical thickness on a whole brain level in cognitively healthy older persons,

416

H.I.L. Jacobs et al. / White Matter Lesions and Cortical Thickness

Table 2 Effect of WMHs on cortical thickness comparing controls, MCI and AD patients following a cluster-wise corrected whole brain analysis Regions (% of the cluster)

Cluster size

Mean p-value (10-x) of the clusters

SD

CLUSTER 1 Parietal lobe (100%) Superior parietal lobule (89.23%) Precuneus (7.85%) Inferior parietal lobule (2.91%)

4152

3.40

0.0001

CLUSTER 2 Parietal lobe (45.88%) Inferior parietal lobule (36.54%) Superior parietal lobule (9.34%) Occipital lobe (41.45%) Lateral occipital gyrus (41.25%) Cuneus (0.20%) Temporal lobe (12.67%) Middle temporal gyrus (7.41%) Bank superior temporal sulcus (2.67%) Inferior temporal gyrus (2.59%)

7559

3.70

0.0003

CLUSTER 3 Frontal lobe (100%) Superior frontal gyrus (81.29%) Rostral middle frontal gyrus (18.71%)

2523

2.70

0.0003

CLUSTER 1 Frontal lobe (100%) Superior frontal gyrus (98.32%) Precentral gyrus (1.68%)

2019

1.37

White matter hyperintensities are positively associated with cortical thickness in Alzheimer's disease.

White matter hyperintensities are associated with an increased risk of Alzheimer's disease (AD). White matter hyperintensities are believed to disconn...
499KB Sizes 0 Downloads 0 Views