Increased fMRI responses during encoding in mild cognitive impairment

Neurobiology of Aging 28 (2007) 1889–1903

Increased fMRI responses during encoding in mild cognitive impairment Anne H¨am¨al¨ainen a , Maija Pihlajam¨aki a,b , Heikki Tanila c , Tuomo H¨anninen d , Eini Niskanen e , Susanna Tervo f , Pasi A. Karjalainen e , Ritva L. Vanninen g , Hilkka Soininen a,b,f,∗ b

a Department of Neuroscience and Neurology, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland Department of Neurology, Brigham & Women’s Hospital, Harvard Medical School, 221 Longwood Avenue, Boston, MA 02115, USA c Department of Neurobiology, A.I. Virtanen Institute for Molecular Sciences, P.O. Box 1627, FIN-70211 Kuopio, Finland d Department of Neurology, Kuopio University Hospital, P.O. Box 1777, FIN-70211 Kuopio, Finland e Department of Physics, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland f Brain Research Unit, Clinical Research Centre, Mediteknia, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland g Department of Clinical Radiology, Kuopio University Hospital, P.O. Box 1777, FIN-70211 Kuopio, Finland

Received 27 March 2006; received in revised form 16 August 2006; accepted 21 August 2006 Available online 25 September 2006

Abstract Structural and functional magnetic resonance imaging (fMRI) was performed on 21 healthy elderly controls, 14 subjects with mild cognitive impairment (MCI) and 15 patients with mild Alzheimer’s disease (AD) to investigate changes in fMRI activation in relation to underlying structural atrophy. The fMRI paradigm consisted of associative encoding of novel picture-word pairs. Structural analysis of the brain was performed using voxel-based morphometry (VBM) and hippocampal volumetry. Compared to controls, the MCI subjects exhibited increased fMRI responses in the posterior hippocampal, parahippocampal and fusiform regions, while VBM revealed more atrophy in MCI in the anterior parts of the left hippocampus. Furthermore, the hippocampal volume and parahippocampal activation were negatively correlated in MCI, but not in controls or in AD. We suggest that the increased fMRI activation in MCI in the posterior medial temporal and closely connected fusiform regions is compensatory due to the incipient atrophy in the anterior medial temporal lobe. © 2006 Elsevier Inc. All rights reserved. Keywords: Functional magnetic resonance imaging; Voxel-based morphometry; Mild cognitive impairment; Alzheimer’s disease; Episodic memory; Encoding

1. Introduction Mild cognitive impairment (MCI) is defined as a syndrome of cognitive deficit greater than expected in relation to age, while the individual’s activities of daily living are substantially intact [54]. Thus, it is considered as a possible transitional state between cognitively normal aging and dementia. The syndrome can be divided into subgroups of amnestic and non-amnestic MCI according to whether the subject has memory impairment or not, and both subtypes ∗ Corresponding author at: Department of Neurology, Kuopio University Hospital, P.O. Box 1777, FIN-70211 Kuopio, Finland. Tel.: +358 17 173012; fax: +358 17 173019. E-mail address: [email protected] (H. Soininen).

0197-4580/$ – see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.neurobiolaging.2006.08.008

can further be classified as single or multiple domain based on the number of cognitive domains affected [21]. The outcome of non-amnestic MCI often is frontotemporal dementia or dementia with Lewy bodies [21], whereas subjects with the amnestic subtype of MCI, are estimated to have an annual conversion rate of 6–25% to Alzheimer’s disease (AD) [55]. Thus, the conversion rate of MCI to AD greatly exceeds the risk of conversion in healthy age-matched controls, and early identification of subjects at risk for AD may help in focusing the currently available treatments to the appropriate subject group in an early phase of AD. In AD, the first pathological changes appear in the transentorhinal cortex, and spread further to the hippocampus and other limbic structures and neocortex with a typical pattern [4]. The initial, so called transentorhinal Braak stages I–II

1890

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

have been linked with MCI in one neuropathological study [22]. The AD-related neuropathological changes eventually lead to structural atrophy, which has been widely investigated in the early diagnostics of MCI and AD using structural magnetic resonance imaging (MRI). Some studies have shown that volumetric measurements of the entorhinal cortex and hippocampus can differentiate MCI and AD subjects from controls [34,39,52]. Techniques that allow the assessment of whole-brain atrophy, such as voxel-based morphometry (VBM), have revealed atrophy in MCI subjects not only in the medial temporal lobe (MTL), but also in the lateral temporal, superior parietal, anterior cingulate, and thalamic areas [53]. However, structural imaging seems to be insufficient on its own for an early diagnosis of AD and its prodromal state as atrophy already is evidence of major neuronal loss, thus leaving little scope for preventive measures. In addition, there seems to be a remarkable overlap in the MTL volumetric measures among AD patients and controls, which complicates diagnostics at the level of individual subjects [10,34]. Besides leading to structural shrinkage, the earliest ADrelated neuropathological changes also most probably cause dysfunction that precedes measurable macroscopic structural alterations. Therefore, functional imaging methods such as functional MRI (fMRI) are becoming more and more popular in the investigation of potential biomarkers for AD. Indeed, the superiority of functional to volumetric markers has been reported in a PET study, where regional glucose metabolism measures in the MTL were reported to be better than volumetric measurements in most between-group comparisons of healthy elderly subjects versus MCI and AD patients [16]. The earliest site of AD neuropathology, the MTL, has dense and reciprocal connections to the association cortices [43], and therefore, the AD-related pathology present in the MTL structures already at the level of MCI could contribute to functional alterations of the cortical connecting areas that do not present structural atrophy at the time. It is possible that these functional alterations could be detected with sensitive whole-brain functional imaging methods such as fMRI. Currently the diagnostic of AD is primarily based on clinical evaluation and neuropsychological testing. However, in the early stages of AD it may be difficult to differentiate between aging-related memory deficits and pathological cognitive decline. It has been suggested that differential mechanisms cause the memory deficits in normal aging and in AD. The aging-related alterations have been linked to changes in frontal–striatal system whereas AD-related memory decline is caused by disruption of the MTL memory system [7]. However, functional imaging studies have also shown agingrelated functional changes in the MTL such that decreased hippocampal activity in elderly compared to young subjects has been detected during various memory tasks [9,13,24,51]. Moreover, in addition to weaker hippocampal activity, elderly subjects have shown increased parahippocampal fMRI activity during a memory retrieval task in comparison to young subjects [9]. As to frontal areas, a common aging-related finding has been an increased and a more bilateral pattern

of frontal activation, which has been suggested to present non-selective recruitment of the frontal cortex, or be compensatory [see for example, 9,44,51]. A recent study also shows an interesting negative correlation between the frontal and MTL activations in elderly subjects such that those subjects who least recruited the parahippocampal gyrus had the largest inferior frontal activations [29]. These studies did not, however, report possible underlying structural differences between the study groups. Taken together, these results suggest that the MTL memory system does experience some changes even along normal aging, and not only in relation to the pathological process of AD, thus making it evident that the differentiation between changes related to normal aging and changes related to cognitive decline in elderly, early stage AD subjects is challenging. There is, however, some evidence that the pattern of fMRI activation may differ between healthy elderly and MCI subjects. Recent fMRI studies have suggested increased MTL activations in MCI in comparison to controls while performing memory tasks [17,18]. Nonetheless, fMRI findings in MCI are discrepant as MTL hypoactivation, similar to that seen in AD patients [50,64], has also been reported [47]. These somewhat contradictory findings may be explained by the possibility that the MCI subjects in these studies fall to different stages along the aging-MCI-AD continuum. Another potential contributor to the incongruent results may relate to the subjects’ task performance, as relatively poor postscan memory test results have been related to diminished MTL activations in AD [59,66]. Taken together, it is possible that increased MTL activation in MCI, indicating a need to compensate for early AD-related pathology, is observed during a restricted period along the aging-MCI-AD continuum until the neural function is substantially diminished due to progression of neuropathological changes. Most of the studies on MCI have so far concentrated on MTL activations, and therefore knowledge on the functional properties of other cortical areas in MCI is scarce. Compensatory alterations in the connecting cortical areas of the MTL might, however, be expected as a result of neural reorganization or differential cognitive strategies required to successfully perform the fMRI activation tasks. The aim of our study was to explore the relations of structure and function in elderly controls and subjects with MCI and AD considering the whole brain. The fMRI task was a modification of the widely used Free and Cued Selective Reminding Test (FCSR) [26], a test that assesses cognitive functions known to deteriorate early in the course of AD, such as certain aspects of naming including lexical retrieval [31], as well as semantic [42] and visual [63] processing. Importantly, the task also assesses associative memory encoding, which is known to be an essential function of the MTL [57,68,69]. The MCI subjects in our study were selected from a population-based follow-up study, and all of them fulfilled the Petersen criteria [54] for amnestic MCI at the time of diagnosis. We hypothesized that in comparison to controls, the MCI subjects would show atrophy limited to the MTL,

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

whereas in AD the atrophy is more widespread. Furthermore, in MCI we expected to see compensatorily increased fMRI activation in the MTL and in its closest cortical connecting areas that do not show measurable atrophy yet. In AD, due to the extended structural atrophy, we expected to see diminished fMRI activations when compared to controls and MCI.

2. Methods 2.1. Study subjects A total of 50 subjects were examined in this study (for details, see Table 1). These included 21 right-handed controls (17 females, 4 males, age range 64–79, mean age 71.2), who were identified as cognitively normal in the neuropsychological tests, and scanned as a control group. The MCI group consisted of 14 subjects (10 females, 4 males, age range 57–81, mean age 72.4), and the AD group contained 15 subjects (10 females, 5 males, age range 62–83, mean age 71.3). None of the subjects had a history of neurological or psychiatric disease other than AD. At the time of image acquisition, 9 out of the 15 AD patients were not on cholinesterase inhibitor treatment; 2 patients were on donepezil, 1 patient on rivastigmine and 3 patients on galantamine. One AD patient was using 10 mg of citalopram daily. The other patients were not taking any medications known to affect cognition at the time of imaging. All the subjects were right-handed and did not need correction of visual acuity during functional imaging. Informed written consent was acquired from all the

1891

subjects, and in the case of AD patients, consent was obtained in the presence of a caregiver. The study was approved by the Ethics Committee of Kuopio University Hospital. 2.2. Recruitment and selection criteria of study subjects Control and MCI subjects were recruited from the third follow-up visit of a population-based longitudinal study running at the Brain Research Unit, University of Kuopio [30,70]. Follow-up visits consisted of a structured interview including CDR scale, demographic information, medical history, current medication, history of smoking and alcohol consumption, and a subjective assessment of memory disturbances and depression. Unlike during the first and second follow-up visits of the longitudinal study, during the third follow-up it was also possible to include an informant interview due to the smaller amount of subjects to be assessed. The informants were interviewed to corroborate the subject’s memory complaints, and they also completed the CDR interview. The evaluations were performed by a clinician (AH) and a neuropsychologist (ST), and they also made the diagnoses in consensus. The neuropsychological testing included the following tests—memory: visual reproduction test (VR, immediate and delayed recall) from the Wechsler memory scale [61], logical memory test (LM, immediate and delayed recall) from the Wechsler memory scale-revised [73], word list recall (immediate and delayed recall) from the CERAD neuropsychological assessment battery [49], delayed recall of the constructional praxis from CERAD [49], NYU paragraph recall (immediate

Table 1 Demographic and cognitive characteristics Controls

Age Female/male Education (years) MMSE CDR total score CDR sum of boxes Trail making test A Trail making test C Boston naming test Verbal fluency (animal) Verbal fluency (PAS) Clock drawing test Heaton, immediate Heaton, delayed Heaton, savings (%) Logical memory, immediate Logical memory, savings (%) Word list, delayed recall Word list, savings (%)

MCI

AD

n

Mean ± S.D.

n

Mean ± S.D.

n

Mean ± S.D.

21 21 21 21 21 21 20 20 20 20 20 20 20 20 20 20 20 20 20

71.2 ± 4.9 17/4 7.9 ± 2.9 27.7 ± 2.0 0.0 ± 0.0 0.1 ± 0.2 47.6 ± 13.6 116.3 ± 60.2 13.1 ± 1.7 23.20 ± 4.9 44.3 ± 11.9 5.6 ± 0.8 11.5 ± 2.4 9.7 ± 3.1 81.8 ± 15.4 12.6 ± 3.3 90.5 ± 11.1 6.7 ± 1.6 85.1 ± 14.2

14 14 14 14 14 14 14 13 13 14 12 11 9 9 9 13 12 13 12

72.4 ± 7.3 10/4 8.1 ± 2.6 25.6 ± 3.1* 0.5 ± 0.0# 1.6 ± 0.6# 59.5 ± 19.3 186.0 ± 78.6# 10.7 ± 2.4# 16.6 ± 5.5# 36.3 ± 14.7 4.7 ± 1.1 9.2 ± 2.5 5.2 ± 2.4# 57.4 ± 25.5* 6.7 ± 3.9# 82.3 ± 16.4 4.5 ± 1.5# 67.1 ± 19.0#

15 15 15 15 8 4 11 9 9 11 4 3 4 4 4 9 8 11 8

73.1 ± 6.7 10/5 8.2 ± 2.7 21.7 ± 3.7#,† 0.8 ± 0.3# 2.9 ± 1.4# 75 ± 23.6# 287.1 ± 21.8#,† 10.9 ± 2.4* 14.6 ± 5.6# 34.0 ± 14.9 3.7 ± 1.2 6.0 ± 3.4* 2.5 ± 2.9# 36.5 ± 47.6 5.6 ± 2.6# 25.6 ± 30.0#,‡ 1.6 ± 1.5#,‡ 38.8 ± 32.2#,†

MCI: mild cognitive impairment; AD: Alzheimer’s disease; MMSE: mini-mental state examination. * P < 0.05 vs. controls. † P < 0.05 vs. MCI. ‡ P < 0.005 vs. MCI. # P < 0.005 vs. controls.

1892

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

and delayed recall) [40]; language: abbreviated (15 items) Boston naming test [35], vocabulary subtest of the Wechsler adult intelligence scale-revised (WAIS-R) [72]; attention and executive function: verbal fluency test [3,8], trail making test parts A, B and C [58]; visuospatial skills: constructional praxis from CERAD [49], block design from the WAIS-R [72]; global functioning: clock drawing test [49], minimental state examination (MMSE) [20]. According to the national guidelines in Finland, the subjects have not been opted to perform the backwards spelling task in the MMSE assessment and the scoring for the corresponding section is based solely on the seven subtraction task. In the longitudinal study, diagnosis of MCI required a score of 0.5 in the Clinical Dementia Rating (CDR) scale [32], a score of 1.5 S.D. below the average of a normative age-matched sample group in at least one memory test, normal global cognitive function (MMSE ≥ 20) and no dementia according to the NINCDSADRDA criteria [48]. For the fMRI study, we selected MCI subjects with a total CDR rating of 0.5, and at least 0.5 in the memory subcategory, and thus the subjects belonged to the category of amnestic multidomain MCI. Patients with AD were recruited from the neurological outpatient clinic in the Kuopio University Hospital, and they underwent extensive diagnostic measures including neuropsychological testing, laboratory sampling, CT or MR imaging of the head as well as a clinical and neurological examination. Diagnoses were made by experienced neurologists according to the NINCDS-ADRDA criteria for probable AD [48]. For detailed cognitive characteristics of the study subjects see Table 1. 2.3. FMRI stimulus Prior to scanning, all subjects underwent a thorough training on the task until they could perform it satisfactorily. The activation paradigm (Fig. 1) was a modification of the FCSR [26], and consisted of three different conditions that alternated as blocks: (1) encoding, (2) cued retrieval and (3) baseline. This article focuses on the whole-brain results from the encoding-baseline comparisons. In the encoding condition, pictures of black-and-white line drawings were presented with a categorical cue word, which was contextually associated to the picture. In the two separate instruction slides preceding each encoding block, the subjects were instructed to name and encode the picture, and to press a button with their index finger as a response when they had accomplished the task. In the recall condition, the categorical cue words were shown to the subjects, and the instruction slides shown prior to each retrieval block encouraged the subjects to recall the pictures with the help of the cue, and to press the button each time they succeeded in retrieval. During the baseline condition, subjects were encouraged to focus their gaze at a simple fixation cross-hair and not to rehearse previous word-picture pairs. Subjects were instructed to perform the tasks silently to avoid motion artefacts related to overt speech [38]. The motivation for collecting behavioural

Fig. 1. Schematic figure of the activation task. The task consisted of three different conditions: encoding, retrieval and baseline with instruction slides shown in between. The cycle of these three conditions was repeated six times during a single run. This article presents the results of the encoding-baseline contrast. During the encoding task, the subjects were instructed to encode and name the picture with the help of the contextually related categorical cue. The text above the picture translated from Finnish means “which bird?”. During the baseline condition the subjects were instructed to focus their gaze at the crosshair and not to rehearse the picture-word pairs.

data with the button press was to verify that the subjects were attending properly to the task, and to obtain the subjects’ own estimate of their performance, while also ensuring that the AD patients were able to execute the task adequately. The duration of the stimulus slides during activation blocks was 4.5 s and the stimuli were separated by a 0.5 s interstimulus fixation. The entire activation block including five picture slides, as well as the baseline block, lasted 25 s. Each instruction slide was shown for 9.5 s and was followed by a 0.5 s interstimulus fixation. The combination of the encoding/retrieval/baseline-block, including instruction slides, was repeated six times during one functional run, thus leading to a total functional imaging time of 13 min 30 s. The visual stimuli were presented using a laptop computer outside the scanning room and the task was projected to the subjects via a video projector (Lite Pro 620, In Focus Systems Inc, Wisconville, OR, USA) onto a translucent screen. The subjects viewed the stimuli through a mirror attached to the head coil. 2.4. MRI acquisition MR imaging was performed with a 1.5 T scanner (Magnetom Vision, Siemens Medical Systems, Erlangen, Germany) capable of echo-planar imaging (EPI). A circular-polarized head coil was used, and in order to minimize head motion, the subject’s head was carefully fixed with foam rubber pads. Functional imaging was conducted using a gradient echo EPI sequence sensitive to blood-oxygen-level-dependent

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

1893

(BOLD) contrast. The functional images were acquired in an oblique axial orientation aligned according to the anterior–posterior commissural (AC-PC) line. Imaging parameters were as follows: TR = 2500 ms, TE = 70 ms, flip angle = 90◦ , slice thickness = 5 mm plus 1 mm interslice gap, FOV = 256 mm, matrix size = 64 × 64, pixel size = 4.0 mm × 4.0 mm. According to visual inspection of the EPI images, the overall image quality was reasonable and in-plane distorsions insignificant. However, due to our oblique axial slice acquisition orientation, signal drop-out related to susceptibility artefacts in the very anteromedial parts of the MTL, such as the entorhinal cortex, are likely in some of our study subjects. Anatomic high resolution reference images were acquired using a T1-weighted 3D-MPRAGE (three-dimensional magnetization-prepared rapid acquisition gradient echo) sequence with parameters as follows: TR = 9.7 m, TE = 4.0 m, flip angle = 10◦ , slice thickness = 1.0 mm, FOV = 256 mm, matrix size = 256 × 256, pixel size = 1.0 mm × 1.0 mm. T2-weighted FLAIR images were acquired to exclude subjects with significant vascular pathology and the images were evaluated by an experienced neuroradiologist (R.L.V.).

ulated GM images were smoothed with a 12 mm Gaussian kernel. In the statistical analysis, a “single subject: conditions and covariates” model including age, gender and intracranial area as nuisance covariates was used to compare the GM volumes of the three study groups. Differences between groups were assessed using a t-test with a height threshold of P < 0.01 (uncorrected) and an extent threshold of 200 voxels. Statistical threshold for reporting significant differences in brain atrophy between study groups was a cluster-corrected P < 0.05. Individual volumes of the hippocampi in MCI subjects were further obtained by manually outlining the structure with a trackball cursor and an in-house developed software from T1-weighted 3D-MPRAGE images resampled to a slice thickness of 2.0 mm. The measurement of the hippocampus was performed in an anterior to posterior-direction and according to previously published methodology [41,65]. The coronal intracranial area at the level of the anterior commissure was measured and used for normalisation of the hippocampal volumes according to the formula: (volume/intracranial area) × 100.

2.5. Structural data analysis

2.6. Functional data analysis

Structural MRI data analysis was performed using optimized (VBM) [1,23]. At first, a customized template was created. The origin of the spatial coordinates in the individual images was manually set to the anterior commissure and images were reoriented coronally perpendicular to the intercommissural line. The images were then normalized to the Montreal Neurological Institute (MNI) T1-weighted template of SPM2 using 12 parameters affine transformation, and resampled to a voxel size of 2 mm × 2 mm × 2 mm with a bilinear interpolation algorithm. The customized template was obtained by smoothing the normalized images with an 8 mm isotropic Gaussian kernel and averaging the smoothed images. Customized prior probability maps were created by partitioning the normalized unsmoothed images into gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) segments, smoothing with an 8 mm Gaussian filter and averaging the segmented images, thus yielding specific customized prior probability maps for GM, WM and CSF. Original images were then segmented, and parameters for normalization were determined from the obtained GM segments, customized prior probability maps and customized template. Using these parameters, original images were normalized to the customized template through affine and nonlinear transformations, medium regularisation, resampling to 2 mm × 2 mm × 2 mm and using no masking. The normalized images were further segmented into GM, WM and CSF using the customized prior probability maps. A modulation was performed on the segmented GM images by multiplying the GM voxels by the Jacobian determinants derived from the non-linear step of spatial normalization. The mod-

Image preprocessing and data analysis were carried out with SPM2 (Wellcome Department of Imaging Neuroscience, London, UK; http://www.fil.ion.ucl.ac.uk/spm). First, all functional volumes were spatially realigned. Head motion was investigated using the data output of the motion correction algorithm and average values of the head motion in the three cardinal translational and rotational planes were calculated to assure that no group presented excessive movement in comparison to the other study groups (data not shown). Between-slice timing differences induced by differences in acquisition order were corrected. Functional volumes were coregistered to T1-weighted structural volumes oriented along the intercommissural line, and the coregistration success was visually controlled for each subject individually. Normalization parameters determined from the structural volumes were used to spatially normalize each functional volume to a standard template volume based on the Montreal Neurological Institute (MNI) reference brain. Spatial smoothing was performed with a 8 mm Gaussian filter. Functional image analysis was conducted by using general linear model on a voxel-by-voxel basis employing a random effects model implemented with a two level procedure. Functional volumes were sorted by condition and thus divided into encoding (ENC), immediate cued recall and baseline (BL) blocks. This article focuses solely on the ENC-BL comparisons. The haemodynamical BOLD responses were modelled with a canonical haemodynamic response function [15]. The ENC-BL contrast was first defined for each individual. In the within-group analysis, one-sample t-tests were carried out upon the resulting contrast images for each group, whereas in the between-group comparisons we applied two-

1894

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

sample t-tests implementing SPM random effects analysis. For both the within-group and between-group data analysis we used a height threshold of P < 0.01 (uncorrected) and an extent threshold of 200 voxels. Statistical threshold for reporting final significance of fMRI results in the within-group and between-group analyses was a cluster-corrected P < 0.05. The same statistical threshold was used for assessing significant differences in brain atrophy between study groups. The anatomical location of activation areas was performed visually from the statistical activation maps superimposed on the spatially normalized T1 images, using the atlas by Duvernoy [19] as a reference. To inspect the relation of hippocampal volumes and fMRI activation, a correlation analysis with a “single subject: conditions and covariates” model was performed on the MCI subjects’ fMRI data. Individual normalized left and right hippocampal volumes were averaged together and used as the covariate of interest, and encoding performance as a nuisance variable. The search space for possible correlations was restricted by masking the correlation contrast with the between-group encoding-baseline comparison of MCI and control subjects. Both positive and negative correlations were tested. Local peak coordinates were inspected using an uncorrected threshold of P < 0.01 and further performing a small volume correction (SVC) by placing a 10 mm sphere at the detected peak coordinates. SV-corrected voxel-level P values are reported for this analysis. 2.7. Statistical analysis Statistical analysis of demographic, behavioural and volumetric data was conducted with SPSS software (SPSS Inc., Chicago, IL, USA) using nonparametric Kruskal–Wallis and Mann–Whitney U-tests due to the non-normally distributed variables.

(P < 0.90), but there was a significant difference in the MMSE scores (controls versus MCI, P < 0.040; controls versus AD, P < 0.001; MCI versus AD, P < 0.007). From the neuropsychological test battery, only those tests that were also administered to most of the AD patients in the diagnostic testing are reported in order to depict the differences in cognitive profiles between the study groups. Table 1 lists the details of the demographics and neuropsychological test results. 3.2. Behavioural data The behavioural data (Table 2) during functional imaging was collected completely in all the MCI, in 19 control and 11 AD subjects. The missing data in 2 control and 4 AD subjects are due to technical problems in the recording of the reaction times. Despite the missing reaction times, documentation of adequate button presses was completely recorded in all the subjects. The reaction times did not differ significantly between the three study groups (P = 0.494). Statistical analyses of the motor responses, indicating the attention to the task and subjective evaluation of performance, showed no significant differences between the controls and MCI subjects (P = 0.133), but the controls were significantly better than the AD patients (P = 0.006). The amount of motor responses among the MCI and AD groups did not differ significantly (P = 0.186). The subsequent memory test, where the study subjects made a subjective evaluation of their memory performance, indicated that the controls felt they remembered 92.2% of the picture-word pairs, while the MCI subjects remembered 82.6% and AD patients only recalled 58.2%. There were statistically significant differences in the memory performance between the controls and MCI (P < 0.005), AD and controls (P < 0.002) and MCI and AD (P < 0.005). 3.3. Structural results

3. Results 3.1. Subject characteristics The study groups did not differ significantly from each other in terms of age (P < 0.40) or years of education

Detailed information on the localization and statistical significance of VBM structural results is presented in Table 3 and Fig. 2. Compared with controls, the MCI subjects revealed a significant GM volume reduction in the left superior frontal gyrus, in the left anterior hippocampus, and bilaterally in the thalamus.

Table 2 Behavioural measures during functional imaging Elderly controls

Reaction time (ms) Encoding performance (%) Retrieval performance (%)

MCI

AD

n

Mean ± S.D.

n

Mean ± S.D.

n

Mean ± S.D.

19 21 21

1971.7 ± 680.0 97.5 ± 5.1 92.2 ± 6.4

14 14 14

2085.0 ± 469.0 94.8 ± 7.1 82.6 ± 11.9#

11 15 15

1956.3 ± 472.8 91.8 ± 7.0* 58.2 ± 25.8#,‡

Reaction times were collected during the encoding condition in which the subjects were instructed to press a button after they had named and memorized the word-picture pair. Performance indicates the number of responses the subjects made during the task. Retrieval performance indicates the results of the subsequent memory test (based on subjective evaluation). Only significant P-values are reported. ‡ P < 0.005 vs. MCI. # P < 0.005 vs. controls.

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

1895

Table 3 Regions of significant gray matter volume reductions in MCI and AD subjects relative to controls (SPM random effects analysis; P < 0.01) Brain region

R/L

MNI coordinate (x, y, z)

BA

T-value

Controls vs. MCI Superior frontal gyrus Middle temporal gyrus Hippocampus Thalamus

L* R R/L* R* /L*

−15, 39, 49 57, −20, −12 33, −17, −17/−30, −15, −16 5, −19, 7/−7, −21, 5

9 21

3.55 3.99 3.80/4.41 4.04/5.45

Controls vs. AD Superior frontal gyrus Middle frontal gyrus Gyrus rectus Cingulate gyrus Superior parietal gyrus Angular gyrus Intraparietal sulcus Superior temporal gyrus Middle temporal gyrus Inferior temporal gyrus Hippocampus (anterior) Hippocampus (posterior) Amygdala Thalamus Fusiform gyrus

L L R* L L R R R/L R R R* /L* R* /L* L* R* /L* R

−25, 19, 60 −56, 15, 29 6, 25, −27 −10, −54, 23 −38, −51, 70 52, −51, 27 37, −75, 16 46, −52, 11/−43, 18, −30 56, −43, −7 64, −46, −24 26, −10, −20/−23, −14, −17 35, −36, −6/−32, −36, −6 −24, −3, −21 1, −16, 2/−6, −13, 17 36, −16, −41

8 46 12 23 7 40 39 22/38 21 37

20

3.66 4.24 3.57 3.52 5.23 4.26 3.87 3.94/3.73 3.96 3.67 3.60/4.45 5.28/5.32 4.04 3.39/3.83 3.82

MCI vs. AD Superior parietal gyrus Inferior temporal gyrus

L R

−36, −54, 71 54, 17, −39

7 38

4.16 4.03

BA: Brodmann area; (*) cluster-level significance, P < 0.05 corrected.

Cortical GM volume reductions in AD patients relative to controls were detected in the right gyrus rectus, left amygdala, and in the thalamus bilaterally. The hippocampal formation was bilaterally affected along its entire long axis. There was a trend towards more atrophy in frontal as well as parietotemporal cortices in AD but these differences did not meet the

criteria of cluster-corrected P < 0.05 for statistical significance. When compared to MCI subjects, the AD patients showed a trend for GM volume loss in the left superior parietal cortex and in the right inferior temporal cortex. The mean normalized volume of the left and right hippocampi, obtained by manual outlining, were 13.8 ± 1.9

Fig. 2. Areas of significant atrophy in MCI subjects ((A) crosshair position −30, −12, −18) and AD patients ((B) crosshair position −32, −12, −20) compared to controls (P < 0.05, corrected) are superimposed on the group average anatomical image of the elderly control subjects. Left in the image is right in the brain. Gray matter loss in the MCI subjects is evident in the left anterior hippocampus, and in the thalamus while AD patients show, in addition to thalamic atrophy, bilateral hippocampal atrophy along the entire long axis of the hippocampal formation. Color bar presents T values of the atrophy clusters.

1896

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

(mean ± S.D.) and 14.5 ± 2.0 in control subjects, 13.4 ± 2.6 and 14.5 ± 2.2 in MCI subjects, and 11.5 ± 1.4 and 12.2 ± 2.6 in AD patients. There was a statistically significant difference in the hippocampal volumes between the controls and AD subjects (P < 0.005), and MCI and AD subjects (P < 0.05), but the difference between the hippocampal volumes of the control and MCI subjects did not meet statistical significance.

3.4. Functional imaging—within-group level results Detailed information on the within-group activation findings is provided in Table 4. In the control subjects, brain activations in the ENC-BL contrast were detected bilaterally in the superior frontal and precentral gyri, in the left middle and inferior frontal gyri, and in the left anterior cingulate gyrus. There was a trend for bilateral activation of

Table 4 Significant whole-brain activation clusters at the within-group level in the encoding-baseline comparison (SPM random effects analysis; P < 0.01) Brain region

R/L

Coordinate (x, y, z)

BA

T-value

Controls Superior frontal gyrus Middle frontal gyrus Inferior frontal gyrus Inferior frontal gyrus (orbitalis) Cingulate gyrus Precentral gyrus Postcentral gyrus Intraparietal sulcus Supramarginal gyrus Superior temporal gyrus Hippocampus Parahippocampal gyrus Thalamus Fusiform gyrus Cuneus Lingual gyrus Middle occipital gyrus

R* /L* R/L* R/L* L* L* R* /L* L* R* /L* L* L* R* /L* R* /L* L* R* /L* R* /L* R* /L* R* /L*

10, −2, 54/−10, 0, 54 38, 42, 30/ −42, 24, 36 42, 22, 0/−46, 8, 15 −46, 28, −9 −10, 10, 36 42, −6, 48/−48, −6, 48 −50, −26, 45 28, −68, 48/−26, −66, 45 −42, −42, 45 −58, −40, 15 34, −18, −21/−38, −16, −21 34, −26, −21/−32, −26, −21 −16, −20, 6 32, −54, −18/−32, −56, −18 4, −74, 6/−2, −76, 6 12, −78, −15/−10, −78, −15 24, −88, −15/−32, −86, −9

8 46/8 + 9 + 46 44 + 45 47 24 4 1+2 7 + 40 40 22 37 35 + 36

7.45/13.30 3.93/3.79 4.39/8.86 5.33 3.70 6.45/8.13 8.02 8.46/7.56 6.87 5.03 3.61/4.10 4.38/3.91 4.09 12.23/11.73 4.10/3.76 6.72/6.98 13.59/12.09

MCI Middle frontal gyrus Inferior frontal gyrus Inferior frontal gyrus (orbitalis) Precentral gyrus Postcentral gyrus Insula Angular gyrus Intraparietal sulcus Superior temporal gyrus Superior temporal sulcus Hippocampus Parahippocampal gyrus Thalamus Fusiform gyrus Cuneus Lingual gyrus Superior occipital gyrus Middle occipital gyrus Inferior occipital gyrus

R* /L* R* /L* L* R* /L* L* L* L* R/L* L R/L L* R* /L * R* /L* R* /L* R* /L* R* /L* R* /L* R* /L* R* /L*

42, −4, 51/−42, 0, 51 40, 14, 18/−46, 10, 12 −42, 26, −15 46, −14, 44/−44, −8, 48 −48, −28, 51 −42, 8, −9 −34, −74, 42 30, −72, 48/−22, −70, 45 −60, −48, 12 50, −30, −3/−58, −4, −15 −34, −32, −12 28, −32, −21/−28, −30, −24 8, −26, 0/−10, −20, 0 32, −64, −18/−40, −64, −12 10, −96, 0/−12, −96, 6 10, −76, −18/−10, −88, −12 24, −96, 15/−26, −92 18 14, −92, −6/−18, −96, −9 30, −70, −21/−40, −76, −18,

8/8 + 9 + 46 44 + 45/44 + 45 + 47 47 4 1+2

AD Inferior frontal gyrus Precentral gyrus Insula Superior temporal sulcus Superior temporal gyrus Hippocampus Parahippocampal gyrus Thalamus Fusiform gyrus Middle occipital gyrus Inferior occipital gyrus

R* /L* R* /L L* R L* R* R* L* R* /L* R* /L* R* /L*

42, 8, 27/−54, 16, −3 42, −10, 33/−48, −16, 48 −40, 22, −6 52, −40, 0 −64, −10, 6 34, −24, −18 30, −26, −21 −12, −18, 0 32, −62, −18/−34, −68, −18 42, −74, −12/−38, −80, −9 36, −82, −12/−38, −80, −15

BA: Brodmann area; (*) cluster-level significance, P < 0.05 corrected.

37 18 18 18 + 19

39 7 + 40 22 22 35 + 36 20 + 37 18 18 + 17 19 18 19 44/44 + 45 4+6 22 22 35 + 36 37 19 19

4.40/4.54 6.88/9.13 6.93 5.65/8.36 7.21 4.08 3.96 7.08/9.47 4.38 5.08/5.12 4.06 4.09/4.03 4.77/5.35 14.41/15.87 4.01/5.36 8.47/8.04 5.78/5.88 7.13/12.13 9.41/9.72 8.35/5.67 4.91/3.89 4.03 4.55 5.41 4.83 3.88 5.15 4.61/5.61 10.09/8.39 9.32/6.52

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

the middle and inferior frontal gyri, but the activations on the right hemisphere did not reach cluster-level significance. The left postcentral, supramarginal and superior temporal gyri were activated, as well as areas around the intraparietal sulci bilaterally. Bilateral MTL activations were detected in the hippocampus and parahippocampal gyrus, whereas the thalamic activation was only left-sided. Additionally, the fusiform and lingual gyri and also the middle occipital gyrus were bilaterally activated. The whole-brain within-group activation results of the MCI group in the ENC-BL contrast comprised bilateral activation in the middle and inferior frontal and precentral gyri. Postcentral gyrus and insula were activated on the left. Parietal activations were detected around the left intraparietal sulcus and in the angular gyrus. A trend for activation not meeting cluster-level significance was detected around the right intraparietal sulcus and bilaterally around the superior temporal sulci, and in the left superior temporal gyrus. MTL activations included regions in the parahippocampal gyrus bilaterally and in the left hippocampus. The thalamic activation was bilateral. Large occipital and inferotemporal activation areas were detected bilaterally in the superior, middle and inferior occipital, as well as fusiform and lingual gyri. The within-group analysis of the AD group in the ENC-BL contrast revealed activation bilaterally in the inferior frontal gyri and in the right precentral gyrus, left insula, and in the left superior temporal gyrus. There was a trend for activa-

1897

tion around the right superior temporal sulcus. The right hippocampus and parahippocampal gyrus were activated, as well as the thalamus on the left. We also detected bilateral activations in the fusiform and middle and inferior occipital gyri, but no parietal activation areas were found in AD in the within-group analysis. 3.5. Functional imaging—between-group level results Brain activations related to encoding of novel pictureword pairs in the between-group comparisons using SPM random effects analyses are presented in Table 5. The MCI subjects when compared to controls, revealed increased activation along the left ventral visual stream in the fusiform gyrus extending further to the posterior parts of the parahippocampal gyrus and the hippocampus in the ENC-BL contrast (Fig. 3). They also recruited the thalamus bilaterally to a larger extent than the controls. Trend for activation in the right inferior frontal and left middle frontal gyri was also detected, which is consistent with our within-group results where the controls showed significant unilateral and the MCI subjects bilateral frontal activations. When using the cluster-corrected P < 0.05 as threshold for statistical significance we found no activation areas greater in controls than in MCI but there was a trend for activation in the anterior cingulate gyrus. In the two-sample t-test comparing MCI subjects to AD patients no significant clusters were detected. However, there

Table 5 Significant brain activation clusters at the between-group level in the encoding-baseline comparisons (SPM random effects analysis; P < 0.01) Brain region

R/L

MNI coordinate (x, y, z)

BA

T-value

Controls > MCI Cingulate gyrus MCI > controls Middle frontal gyrus Inferior frontal gyrus Precentral gyrus Insula Hippocampus Parahippocampal gyrus Thalamus Fusiform gyrus

L

−14, 0, 42

24

3.25

L R L R/L L* R/L* R* /L* R/L*

−38, 50, 9 42, 36, 0/48, 18, 9 −44, 6, 15 38, 6, −9/−42, 8, −9 −34, −32, −9 20, −34, −15/−24, −36, −15 4, −22, 0/−4, −22, 0 22, −50, −15/−16, −44, −9

10 44 + 45 4

3.03 3.14/3.77 4.29 3.71/3.51 2.90 2.57/2.64 3.56/4.10 4.06/3.87

Controls > AD Precentral gyrus Hippocampus Cuneus Middle occipital gyrus

L L R/L L

−54, 2, 33 −38, −16, −21 18, −74, 9/−6, −92, 15 −32, −88, 3

AD > controls Retrosplenium Thalamus

L R

−4, −40, 21 2, −12, 3

30

2.79 3.23

MCI > AD Intraparietal sulcus Intraparietal and intraoccipital sulcus Cuneus

R/L L R/L

30, −86, 27/−40, −58, 51 −22, −92, 18 4, −86, 18/−6, −92, 12

7 + 40 39 + 19 18

2.98/3.01 3.13 2.80/3.08

AD > MCI No significant activation areas BA: Brodmann area; (*) cluster-level significance, P < 0.05 corrected.

35 + 36 20 + 37 4 18 18 + 19

2.93 3.87 2.97/3.05 3.15

1898

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

Fig. 3. Between-group statistical activation map showing the increased left parahippocampal and fusiform activation (P < 0.05, corrected; crosshair position −36, −40, −12) in MCI when compared to controls in the encoding-baseline comparison. fMRI results are superimposed on the group average anatomical image of MCI subjects. Left in the image is right in the brain. Color bar presents T values.

was a trend for stronger activation in the parietal areas, situated mostly around the intraparietal and intraoccipital sulcus. The AD patients did not reveal greater activations in any brain areas when compared to MCI. Controls compared to AD patients did not show any stronger activations meeting the criteria for cluster-level significance, but there was a trend for recruiting the left hippocampus more. Trend for stronger activation was also detected in the left precentral and middle occipital gyri, and in the cuneus bilaterally. The AD patients did not show any greater activations than the controls.

3.6. Correlation of hippocampal volumes to brain activations in MCI subjects Interestingly, we found a negative correlation between the hippocampal volume and activity in the left parahippocampal gyrus (peak coordinate −24, −36, −12, uncorrected P = 0.001; SV-corrected P = 0.044). The right fusiform gyrus also showed a trend for negative correlation (peak coordinate 32, −48, −18, uncorrected P = 0.002; SV-corrected P = 0.101). The parahippocampal correlation result is pre-

Fig. 4. Negative correlation between the hippocampal volume and brain activation in MCI subjects during encoding is present in the left posterior parahippocampal gyrus ((A) P = 0.044, small volume corrected; crosshair position −24, −36, −12). Area of significant correlation is superimposed on the group average anatomical image of MCI subjects. Left in the image is right in brain. Color bar presents the T values. The scatter plot (B) further illustrates the inverse relationship between hippocampal volume and parahippocampal activation in MCI. The hippocampal volumes (HC vol.; mm3 ) on the Y-axis were normalized with the intracranial area (ICA; mm2 ). The weighted parameter estimate on the X-axis is an estimate of the magnitude of the activation in the encoding-baseline comparison.

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

sented in Fig. 4. There were no positive correlations between the hippocampal volume and brain activations in the MCI subjects. Furthermore, we did not find correlations between the hippocampal volumes and parahippocampal activation in controls or in AD patients.

4. Discussion 4.1. Relation of structural and functional imaging findings The main finding of the present study is that the fMRI activation in the posterior MTL and fusiform regions is increased during memory encoding in MCI at the time when the structural atrophy is limited primarily to the anterior parts of the MTL. In clinical AD with more extensive MTL atrophy, this phenomenon of compensatorily increased activation was not present anymore. This study aimed to localize areas of significant cortical atrophy in the MCI and AD when compared to controls, and to assess the relation of structure and function in these study groups. The areas of GM loss detected in our MCI subjects using VBM are substantially in line with previous studies comparing MCI and control subjects. The most common findings, as in our study, have been atrophy of the thalamus [37,53] and hippocampus [11,37]. The results of manual hippocampal volumetry showed a similar trend for smaller hippocampal volumes in MCI subjects in this relatively small MCI sample. If one evaluates the structural changes with the functional changes detected in our study between the MCI and controls, only the hippocampus and the thalamus revealed both structural and functional alterations whereas the increased cortical activations detected in MCI subjects should not, based on our VBM results, be affected by significant degrees of GM loss. When compared to controls, our AD patients exhibited significant atrophy bilaterally along the entire hippocampal long axis, as well as in the left amygdala, gyrus rectus, and thalamus bilaterally. Some cortical areas, including the temporal neocortex, parietal association areas, and posterior cingulate cortices, also showed GM loss in AD, but did not reach the relatively conservative cluster-corrected significance. These reported areas of GM loss are in agreement with the results of previous studies [2,36], as well as with the knowledge on the areas affected by AD pathology [4]. Therefore, it is possible that differences in the functional imaging findings between the controls and AD patients are confounded by structural atrophy in AD. In the present study, this concerns especially the hippocampus and the thalamus, in which both structural and functional differences among controls and AD patients were detected. When comparing the MCI subjects to AD patients, we found subtle structural and functional differences in parietal areas: the MCI subjects tended to show increased activation in the parietal association cortices, where they also showed less atrophy than AD patients. However, neither of

1899

these findings reached the conservative cluster-level significance and therefore these results need to be interpreted with some caution. 4.2. Increased fMRI responses in MCI The MCI subjects exhibited significantly higher fMRI responses in the fusiform and posterior parahippocampal gyri and in the hippocampus when compared to controls. Previous results from fMRI studies in MCI are conflicting, as both decreased [47] and increased [17,18] fMRI responses have been reported. The seeming discrepancy in the previous studies may partly be related to differences in the degree of cognitive decline in the subjects of these studies as the diagnostic criteria used for MCI are variable. Additionally, differential fMRI data analysis methods have been utilized in previous studies, which may lead to difficulties in comparing the results with each other. Our findings of increased activation in the posterior MTL and fusiform gyrus in MCI compared to controls may reflect compensatory mechanisms, i.e. activation of differential neuronal networks in order to compensate for the evolving dysfunction of the MTL while trying to achieve the level of control subjects in the behavioural performance. Increased parahippocampal activation during a memory task in healthy elderly subjects has been reported earlier as well [9]. In the present study, we did not have young control subjects and thus it is not possible to comment on the level of the activation of the elderly controls compared to young subjects. The mean ages of the control and MCI groups were, however, similar and thus it is unlikely that the finding of increased MTL and fusiform activation in the present study would be age-related. We suggest that in MCI even more pronounced compensatory mechanisms are needed in order to try to keep a proper level of task performance than in healthy aging. A recent study has reported a loss of outer molecular layer neurons in the dentate gyrus in 75% of MCI subjects when compared to the control group mean [62], and thus the function of neural networks in the MTL may have altered to compensate for the evolving synaptic and neuronal loss. Altogether compensatory activations in MCI subjects may be a combination of aging-related shifts in the function of the MTL substructures [13] and dysfunction caused by neurofibrillary pathology [4]. Our hypothesis of compensatory activation was further supported by the finding of a negative correlation between hippocampal volume and fMRI activation in the posterior parahippocampal gyrus, indicating that the subjects with smaller hippocampal volumes elicited stronger parahippocampal activation. Importantly, there was no correlation between the hippocampal volumes and parahippocampal activation either in the elderly controls or AD patients, and therefore the inverse relationship between the structural and functional imaging results was restricted to the MCI subjects. The VBM-results further indicated that the hippocampal atrophy in MCI was located in the anterior part, and thus the increased posterior MTL activation may be an

1900

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

attempt to compensate for the atrophy in the anterior MTL structures. The MCI subjects also showed a trend towards increased frontal activations compared to controls. This pattern was evident also at the within-group level, where the MCI subjects showed greater bilateral effects in the middle and inferior frontal gyri than the controls, although this difference was not significant at the cluster-level in the SPM random-effects between group analyses. Increased recruitment of frontal cortical areas during memory tasks has been detected in elderly healthy subjects [9,44,51], but it remains unclear whether additional frontal compensatory mechanisms are present in MCI in comparison to healthy aging. Altogether, the MCI subjects showed compensatory activations in the MTL memory system and along areas of the ventral visual stream during a visually presented memory encoding task. This suggests that they may present both aging-related changes as well as alterations linked to early AD-type pathology in brain activations. Follow-up of our MCI subjects will reveal whether the observed increased activations could be related to future cognitive decline. To better be able to perceive the functional alterations occurring along the continuum from normal aging to AD, further longitudinal studies investigating subjects both at the very mild as well as severe end of MCI with complementary imaging methods enabling the analysis of both vascular and neuronal changes, as well as neuronal connectivity is warranted. 4.3. Activations in AD patients At the within-group level, the activation pattern observed in the AD patients was typical for picture naming and encoding tasks. Activations were detected in the inferior frontal and lateral temporal cortices, in the MTL, and in the fusiform and occipital cortices. In the between-group comparisons, the AD patients showed decreased activations compared to controls in the occipital and frontoparietal cortex and in the MTL. The AD patients also showed some evidence towards both structural atrophy and diminished fMRI activation in the parietal cortices in comparison to MCI subjects. These findings are consistent with previous results [6,16,34] and with the known pattern of progressive deterioration in AD [4,22], however, these activations did not meet cluster-level significance and therefore must be interpreted with some caution. There were no differences between the AD and MCI subjects in MTL activations in the strictly thresholded SPM random effects analysis. It is possible that in the present study the MCI and AD patient groups were relatively similar in terms of their way of activating the hippocampus. The neuropsychological test profile of the MCI subjects supports this assumption as in some tests the MCI subjects performed almost as poorly as the AD patients. Some of these tests, in which the MCI subjects showed impaired performance, are known to activate MTL structures as evidenced by a previous fMRI study regarding verbal fluency [56]. On the other hand, the AD patients did not present similar posterior MTL areas of increased activation

in comparison to controls than the MCI subjects did, and the MCI subjects had better capability of recruiting the parahippocampal regions in order to compensate for the evolving memory impairment. 4.4. Methodological issues The usage of congruent criteria for defining MCI is extremely important when evaluating the results of different studies. In our study, we used the commonly acknowledged Petersen criteria for the identification of MCI subjects. However, the MMSE score cutoff we used (i.e. 20) for this study is somewhat lower than what is used in some previous studies (i.e. 24) [27]. Four out of our 14 subjects had an MMSE score less than 24. Generally, cognitive measures clearly demonstrated that the MCI subjects in our study are placed between the controls and AD patients. Nonetheless, it may be possible that the MCI subjects of the present study are more severely impaired compared to MCI subjects in some previous fMRI studies as evaluated by MMSE [17,18,47] or CDR Sum of Boxes scores [18,47]. Including a few more severely impaired MCI subjects in the study would cause the differences between controls and MCI be more pronounced while the differences between MCI and AD would be diminished. The compensatory activations were, however, only detected in MCI compared to controls. Thus, the phenomenon of compensatorily increased fMRI activation is likely to represent changes related to the transitional stage of MCI between normal aging and AD. In addition, the structural and functional changes we detected between the MCI and AD subjects are logical when considering the Braak staging of the distribution of AD-related neuropathological changes [4]. The BOLD contrast is known to indirectly measure neuronal activity through neurovascular coupling [45], which may, however, undergo changes during aging and AD-related pathological processes. Previously it has been shown that altered neurovascular coupling led to decreased signal-tonoise ratio of the BOLD response in the motor cortex in healthy elderly controls [12], and in the visual cortex in AD patients [6]. In AD, the presence of beta-amyloid (A␤) in the brain impairs neuronal and glial function, and therefore the effect of A␤ on cerebral blood flow would be attenuation [33], which could lead to diminished BOLD responses. In addition, diminished cholinergic activity may have an attenuative effect on neurovascular coupling and further on BOLD responses. On the other hand, it has been suggested that the activity of acetylcholine may be up-regulated in MCI [14], and thus might enhance regional blood flow and evoke increased BOLD responses. Considering these issues, it is possible that the differences in brain activations between controls, MCI and AD may partially be due to alterations in the physiological properties of haemodynamic responses measured by BOLD imaging. The results in our study derive from a relative comparison of active encoding and a more passive baseline condition, as

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

often is the case in fMRI studies. Therefore, diminished baseline activity might be misinterpreted as increased task-related activity. Task-induced deactivations have been detected in the so-called “default mode network” that comprises structures such as the posterior midline parietal and ventromedial frontal cortices, as well as the hippocampus [25,28]. Alterations in the default mode network have been detected in MCI and AD patients such that the patients show diminished deactivation responses when compared to controls [46,60]. A recent study has also detected a linear increase in the default mode activity with age during memory tasks, as well decreased activity in the fusiform gyrus in relation to aging [24]. When considering our main results, i.e. increased activation in the fusiform, parahippocampal, and hippocampal regions in MCI, it appears possible that this finding might be explained by increased deactivation in MCI in these areas as well. The analysis approach in the present study does not provide direct evidence of deactivation findings, but inspection of the results in the opposite baseline-encoding contrast did not reveal significantly different MTL or fusiform deactivation responses in these three study groups using the present statistical criteria. Thus, it is unlikely that our main findings would primarily be due to differences in the default mode activity in the MTL/fusiform areas. As none of the subject groups differed by age, aging-related alterations in the brain function are also a very unlikely explanation for our findings. Regarding the paradigm used in the present study there are some other possible limitations. The comparison of an active task to a low-level control task such as visual fixation yields less precision in evaluating specific cognitive processes. As a consequence of such a broad cognitive comparison, the observed activations may reflect many different underlying processes such as simple motor responses given during the task but not during the fixation. The AD patients are not always, however, very capable to co-operate under fMRI experimental conditions, and therefore the fMRI tasks have to be kept simple enough. Another possible limitation in our study is the usage of block designed paradigm which does not allow the separation of successful and unsuccessful encoding trials. Previous literature has shown that the MTL activation relates to subsequent memory performance [5,67,71], and in MCI, both hippocampal and parahippocampal activation correlates positively to postscan memory performance [17]. Correspondingly, it has been shown that in AD, decreased MTL activity is associated to relatively poor subsequent memory test results [59,66]. Therefore, it is suggested that the increased activations detected in our MCI subjects present a compensatory response that enabled to reach a fMRI task performance level relatively close to the control subjects. A recent eventrelated study in AD patients reported similar hippocampal fMRI responses for successful and unsuccessful encoding trials [50], but further research is needed to address the neural underpinnings of successful and unsuccessful memory encoding in MCI.

1901

5. Conclusion This modification of the FCSR [26], applied in the context of an fMRI experiment, has proven to be feasible for examining subjects with memory impairment. Compared to controls, stronger activations were detected in the MCI subjects in the posterior parts of the MTL and in the fusiform gyrus, whereas there was no significant atrophy in these cortical regions. These focal increased fMRI activations may represent compensatory changes in response to the evolving neuropathological processes. Our findings are intriguing in relation to the possibility of combining structural and functional imaging markers for early detection of AD. Acknowledgements This study was supported by Health Research Council of Academy of Finland, Kuopio University Hospital EVO grants 477311 and 5772720, National Graduate School of Clinical Investigation, Nordic Centre of Excellence in Neurodegeneration, FinnWell program of the National Technology Agency of Finland, EU Regional funding 1247/31/04, Lilly Foundation and Research Foundation of Orion Corporation. References [1] Ashburner J, Friston K. Voxel-based morphometry-the methods. Neuroimage 2000;11:805–21. [2] Baron JC, Chetelat G, Desgranges B, Perchey G, Landeau B, De La Sayette V, et al. In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer’s disease. Neuroimage 2001;14:298–309. [3] Borkowski JG, Benton AL, Spreen O. Word fluency and brain damage. Neuropsychologia 1965;5:135–40. [4] Braak H, Braak E. Neuropathological staging of Alzheimer-related changes. Acta Neuropathol 1991;82:239–59. [5] Brewer JB, Zhao Z, Desmond JE, Glover GH, Gabrieli JD. Making memories: brain activity that predicts how well visual experience will be remembered. Science 1998;281:1185–7. [6] Buckner RL, Snyder AZ, Sanders AL, Raichle ME, Morris JC. Functional brain imaging of young, nondemented and demented older adults. J Cogn Neurosci 2000;12(Suppl. 2):24–34. [7] Buckner RL. Memory and executive function in aging and AD: multiple factors that cause decline and reserve factors that compensate. Neuron 2004;44:195–208. [8] Butters N, Granholm E, Salmon DP, Grant I. Episodic and semantic memory: a comparison of amnesic and demented patients. J Clin Exp Neuropsychol 1987;9:479–97. [9] Cabeza R, Daselaar SM, Dolcos F, Prince SE, Budde M, Nyberg L, et al. Task-independent and task-specific age effects on brain activity during working memory, visual attention and episodic retrieval. Cereb Cortex 2004;14:364–75. [10] Chan D, Fox NC, Scahill RI, Crum WR, Whitwell JL, Leschziner G, et al. Patterns of temporal lobe atrophy in semantic dementia and Alzheimer’s disease. Ann Neurol 2001;49:433–42. [11] Chetelat G, Desgranges B, De La Sayette V, Viader F, Eustache F, Baron JC. Mapping gray matter loss with voxel-based morphometry in mild cognitive impairment. Neuroreport 2002;13:1939–43. [12] D’Esposito M, Deouell LY, Gazzaley A. Alterations in the BOLD fMRI signal with aging and disease: a challenge for neuroimaging. Nat Rev Neurosci 2003;4:863–72.

1902

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903

[13] Daselaar SM, Fleck MS, Dobbins IG, Madden DJ, Cabeza R. Effects of healthy aging on hippocampal and rhinal memory functions: an eventrelated fMRI study. Cereb Cortex 2006 (Epub ahead of print). [14] DeKosky ST, Ikonomovic MD, Styren SD, Beckett L, Wisniewski S, Bennett DA, et al. Upregulation of choline acetyltransferase activity in hippocampus and frontal cortex of elderly subjects with mild cognitive impairment. Ann Neurol 2002;51:145–55. [15] Della-Maggiore V, Chau W, Peres-Neto PR, McIntosh AR. An empirical comparison of SPM preprocessing parameters to the analysis of fMRI data. Neuroimage 2002;17:19–28. [16] De Santi S, de Leon MJ, Rusinek H, Convit A, Tarshish CY, Roche A, et al. Hippocampal formation glucose metabolism and volume losses in MCI and AD. Neurobiol Ag 2001;22:529–39. [17] Dickerson BC, Salat DH, Bates JF, Atiya M, Killiany RJ, Greve DN, et al. Medial temporal lobe function and structure in mild cognitive impairment. Ann Neurol 2004;56:27–35. [18] Dickerson BC, Salat DH, Greve DN, Chua EF, Rand-Giovannetti E, Rentz DM, et al. Increased hippocampal activation in mild cognitive impairment compared to normal aging and AD. Neurology 2005;65:404–11. [19] Duvernoy HM. The human brain. Surface, blood supply, and threedimensional sectional anatomy. Wien: Springer-Verlag; 1999. [20] Folstein MF, Folstein SE, McHugh PR. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J Psychiat Res 1975;12:189–98. [21] Gauthier S, Reisberg B, Zaudig M, Petersen RC, Ritchie K, Broich K, et al. On behalf of the participants of the international psychogeriatric association expert conference on mild cognitive impairment 367. Mild cognitive impairment. Lancet 2006:1262–70. [22] Gold G, Bouras C, K¨ovari E, Canuto A, Glar´ıa BG, Malky A, et al. Clinical validity of Braak neuropathological staging in the oldest-old. Acta Neuropathol 2000;99:579–82. [23] Good CD, Johnsrude I, Ashburner J, Henson RN, Friston KJ, Frackowiak RS. A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage 2001;14:21–36. [24] Grady CL, Springer MV, Hongwanishkul D, McIntosh AR, Winocur G. Age-related changes in brain activity across the adult lifespan. J Cogn Neurosci 2006;18:227–41. [25] Greicius MD, Srivastava G, Reiss AL, Menon V. Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: evidence from functional MRI. Proc Natl Acad Sci 2004;101:4637–42. [26] Grober E, Buschke H. Genuine memory deficits in dementia. Develop Neuropsychol 1987;3:13–36. [27] Grundman M, Petersen RC, Ferris SH, Thomas RG, Aisen PS, Bennett DA, et al. For the Alzheimer’s Disease Cooperative Study. Mild cognitive impairment can be distinguished from Alzheimer’s disease and normal aging for clinical trials. Arch Neurol 2004;61:59–66. [28] Gusnard DA, Raichle ME. Searching for a baseline: functional imaging and the resting human brain. Nat Rev Neurosci 2001;2:685–94. [29] Gutchess AH, Welsh RC, Hedden T, Bangert A, Minear M, Liu LL, et al. Aging and the neural correlates of successful picture encoding: frontal activations compensate for decreased medial-temporal activity. J Cogn Neurosci 2005;17:84–96. [30] Hanninen T, Hallikainen M, Tuomainen S, Vanhanen M, Soininen H. Prevalence of mild cognitive impairment: a population-based study in elderly subjects. Acta Neurol Scand 2002;106:148–54. [31] Hodges JR, Salmon DP, Butters N. The nature of naming deficit in Alzheimer’s and Huntington’s disease. Brain 1991;114:1547–58. [32] Hughes CP, Berg L, Danziger WL, Coben LA, Martin RL. A new clinical scale for the staging of dementia. Br J Psychiat 1982;140:566–72. [33] Iadecola C. Neurovascular regulation in the normal brain and in Alzheimer’s disease. Nat Rev Neurosci 2004;5:347–60. [34] Juottonen K, Laakso MP, Partanen K, Soininen H. Comparative MR analysis of the entorhinal cortex and hippocampus in diagnosing Alzheimer disease. Am J Neuroradiol 1999;20:139–44. [35] Kaplan EF, Goodglass H, Weintraub S. The Boston naming test. Boston: Kaplan & Goodglass; 1991.

[36] Karas GB, Burton EJ, Rombouts SARB, van Schijndel RA, O’Brien JT, Scheltens P, et al. A comprehensive study of gray matter loss in patients with Alzheimer’s disease using optimized voxel-based morphometry. Neuroimage 2003;18:895–907. [37] Karas GB, Scheltens P, Rombouts SARB, Visser PJ, van Schijndel RA, Fox NC, et al. Global and local gray matter loss in mild cognitive impairment and Alzheimer’s disease. Neuroimage 2004;23:708–16. [38] Kemeny S, Ye FQ, Birn R, Braun AR. Comparison of continuous overt speech fMRI using BOLD and arterial spin labeling. Human Brain Mapp 2005;24:173–83. [39] Killiany RJ, Hyman BT, Gomez-Isla T, Moss MB, Kikinis R, Jolesz F, et al. MRI measures of entorhinal cortex versus hippocampus in preclinical AD. Neurology 2002;58:1188–96. [40] Kluger A, Ferris SH, Golomb J, Mittelman MS, Reisberg B. Neuropsychological prediction of decline to dementia in nondemented elderly. J Geriatr Psychiat Neurol 1999;12:168–79. [41] Laakso MP, Soininen H, Partanen K, Helkala EL, Hartikainen P, Vainio P, et al. Volumes of the hippocampus, amygdala and frontal lobes in the MRI-based diagnosis of early Alzheimer’s disease: correlation with memory functions. J Neural Transm 1995;9:73–86. [42] Lambon Ralph MA, Patterson K, Hodges JR. The relationship between naming and semantic knowledge for different categories in dementia of Alzheimer’s type. Neuropsychologia 1997;35:1251–60. [43] Lavenex P, Amaral DG. Hippocampal-neocortical interaction: a hierarchy of associativity. Hippocampus 2000;10:420–30. [44] Logan JM, Sanders AL, Snyder AZ, Morris JC, Buckner RL. Underrecruitment and non-selective recruitment: dissociable neural mechanisms associated with aging. Neuron 2002;33:827–40. [45] Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A. Neurophysiological investigation of the basis of the fMRI signal. Nature 2001;412:150–7. [46] Lustig C, Snyder AZ, Bhakta M, O’Brien KC, McAvoy M, Raichle ME, et al. Functional deactivations: change with age and dementia of the Alzheimer type. Proc Natl Acad Sci USA 2003;100:14504–9. [47] Machulda MM, Ward HA, Borowski B, Gunter JL, Cha RH, O’Brien PC, et al. Comparison of memory fMRI response among normal, MCI, and Alzheimer’s patients. Neurology 2003;61:500–6. [48] McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: report of the NINCDSADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s disease. Neurology 1984;34:939–44. [49] Morris JC, Heyman A, Mohs RC, Hughes JP, van Belle G, Fillenbaum G, et al. the CERAD investigators: the consortium to establish a registry for Alzheimer’s disease (CERAD). Part 1. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurology 1989;39:1159–65. [50] Pariente J, Cole S, Henson R, Clare L, Kennedy A, Rossor M, et al. Alzheimer’s patients engage an alternative network during a memory task. Ann Neurol 2005;58:870–9. [51] Park DC, Welsh RC, Marshuetz C, Gutchess AH, Mikels J, Polk TA, et al. Working memory for complex scenes: age differences in frontal and hippocampal activations. J Cogn Neurosci 2003;15: 1122–34. [52] Pennanen C, Kivipelto M, Tuomainen S, Hartikainen P, H¨anninen T, Laakso MP, et al. Hippocampus and entorhinal cortex in mild cognitive impairment and early AD. Neurobiol Aging 2004;25:303–10. [53] Pennanen C, Testa C, Laakso MP, Hallikainen M, Helkala EL, Hanninen T, et al. A voxel based morphometry study on mild cognitive impairment. J Neurol Neurosurg Psychiat 2005;76:11–4. [54] Petersen RC, Smith GE, Ivnik RJ, Tangalos EG, Schaid DJ, Thibodeau SN, et al. E status as a predictor of the development of Alzheimer’s disease in memory-impaired individuals. JAMA 1995;273:1274–8. [55] Petersen RC, Stevens JC, Ganguli M, Tangalos EG, Cummings JL, DeKosky ST. Practice parameter: early detection of dementia: mild cognitive impairment (an evidence-based review). Neurology 2001;56:1133–42.

A. H¨am¨al¨ainen et al. / Neurobiology of Aging 28 (2007) 1889–1903 [56] Pihlajam¨aki M, Tanila H, H¨anninen T, K¨on¨onen M, Laakso M, Partanen K, et al. Verbal fluency activates the left medial temporal lobe: a functional magnetic resonance imaging study. Ann Neurol 2000;47:470–6. [57] Pihlajam¨aki M, Tanila H, Hanninen T, Kononen M, Mikkonen M, Jalkanen V, et al. Encoding of novel picture-pairs activates the perirhinal cortex: an fMRI study. Hippocampus 2003;13:67–80. [58] Reitan RM. Validity of the trail making test as an indicator of organic brain damage. Percept Mot Skills 1958;8:271–6. [59] Rombouts SA, Barkhof F, Veltman DJ, Machielsen WC, Witter MP, Bierlaagh MA, et al. Functional MR imaging in Alzheimer’s disease during memory encoding. AJNR Am J Neuroradiol 2000;21:1869–75. [60] Rombouts SARB, Barkhof F, Goekoop R, Stam CJ, Scheltens P. Altered resting state networks in mild cognitive impairment and mild Alzheimer’s disease: an fMRI study. Hum Brain Mapp 2005;26:231–9. [61] Russel EW. A multiple scoring method for the assessment of complex memory functions. J Cons Clin Psychol 1975;43:800–9. [62] Scheff SW, Price DA, Schmitt FA, Mufson EJ. Hippocampal synaptic loss in early Alzheimer’s disease and mild cognitive impairment. Neurobiol Aging 2006;10:1372–84. [63] Silveri MC, Leggio MG. Influence of disorders of visual perception in word-to-picture matching tasks in patients with Alzheimer’s disease. Brain Lang 1996;54:326–34. [64] Small SA, Perera GM, DeLaPaz R, Mayeux R, Stern Y. Differential regional dysfunction of the hippocampal formation among elderly with memory decline and Alzheimer’s disease. Ann Neurol 1999;45:466–72. [65] Soininen HS, Partanen K, Pitk¨anen A, Vainio P, H¨anninen T, Hallikainen M, et al. Volumetric MRI analysis of the amygdala and the

[66]

[67]

[68] [69] [70]

[71]

[72] [73]

1903

hippocampus in subjects with age-associated memory impairment: correlation to visual and verbal memory. Neurology 1994;44:1660–8. Sperling RA, Bates JF, Chua EF, Cocchiarella AJ, Rentz DM, Rosen BR, et al. fMRI studies of associative encoding in young and elderly controls and mild Alzheimer’s disease. J Neurol Neurosurg Psychiat 2003;74:44–50. Sperling R, Chua E, Cocchiarella A, Rand-Giovannetti E, Poldrack R, Schacter DL, et al. Putting names to faces: successful encoding of associative memories activates the anterior hippocampal formation. Neuroimage 2003;20:1400–10. Squire LR, Stark CE, Clark RE. The medial temporal lobe. Annu Rev Neurosci 2004;27:279–306. Squire LR, Zola-Morgan S. The medial temporal lobe memory system. Science 1991;253:1380–6. Tervo S, Kivipelto M, Hanninen T, Vanhanen M, Hallikainen M, Mannermaa A, et al. Incidence and risk factors for mild cognitive impairment: a population-based three-year follow-up study of cognitively healthy elderly subjects. Dement Geriatr Cogn Disord 2004;17:196–203. Wagner AD, Schacter DL, Rotte M, Koutstaal W, Maril A, Dale AM, et al. Building memories: remembering and forgetting of verbal experiences as predicted by brain activity. Science 1998;281: 1188–91. Wechsler D. Wechsler adult intelligence scale-revised. Cleveland: Psychological Corporation; 1981. Wechsler D. Wechsler memory scale—revised manual. San Antonio: Psychological Corporation; 1987.