Activity in Memory Brain Networks During Encoding Differentiates Mild Cognitive Impairment Converters from Non-Converters

Abstract

Alzheimer’s disease (AD) has been associated with memory impairment due to alterations in the medial temporal lobe (MTL) and the precuneus. Therefore, the goal of this study was to investigate the effects of AD on the brain networks associated with the hippocampus and precuneus during an encoding memory task. 68 mild cognitive impairment patients (MCI), 21 AD patients, and 20 healthy controls (HC) were included. Participants were instructed to memorize landscapes while undergoing fMRI scanning, followed by a recognition test. MCI were followed up clinically for 18 months to track conversion status. Independent component analysis (ICA) was performed to investigate AD effects on precuneus and MTL networks during memory encoding. Behavioral analyses indicate that HC had a better performance than MCI converters (MCIc) and AD. ICA showed that MCIc had significantly higher activation in the MTL-associated network than MCI non converters (MCIn) and AD, including bilateral hippocampus, parahippocampus, and fusiform gyrus. Furthermore, the precuneus-associated network fitted the default mode network, showing a negative correlation with behavioral performance. These findings indicate that the hyperactivation of the hippocampal network displayed by MCIc has potential discrimination capacity to distinguish them of MCIn, and could be interpreted as a compensatory mechanism.

Keywords

Alzheimer’s disease hippocampus independent component analysis memory precuneus

INTRODUCTION

Alzheimer’s disease (AD) is a complex disease characterized by cognitive dysfunction, with memory impairment as the representative symptom. The relationship between this disease and life expectancy can be seen in the number of deaths attributed to AD, which increased by 71% between 2000 and 2013 [1]. Furthermore, the number of people who suffer from AD is expected to double in 20 years, reaching a total of 115.4 million affected people in 2050 [2, 3].

AD exhibits a general neuronal impairment that affects memory and other cognitive domains. Previous studies attributed the memory impairment to damage of the perforant pathway of the medial temporal lobe (MTL), leading to a disconnection between the entorhinal cortex and hippocampal structures [4 –7]. However, recent models suggest that clinical symptoms of AD may reflect a dysfunction of distributed brain networks, rather than a region-specific neuronal impairment [6, 8]. Thus, as Grady et al. [9] indicated, it is expected that “memory and other cognitive abilities are the result of the integrated activity in networks of regions, rather than activity in any one region in isolation”. In recent years, neuroimaging studies have contributed to understanding the neurobiological alterations associated with memory impairment in AD. Several studies have consistently reported decreased functional magnetic resonance imaging (fMRI) activation in MTL regions in patients with AD, compared to older control subjects, during episodic encoding tasks with a wide variety of stimuli, such as name-face associations, geometric shapes, verbal stimuli, or scenes [10 –17]. However, these studies focused on investigating segregated brain activations, ignoring how AD affects the underlying brain networks involved in memory encoding.

More recently, AD research has paid special attention to the default mode network (DMN) because one of the main nodes of this network, the precuneus/posterior cingulate, has been shown to be relevant in memory processes [18, 19]. The DMN includes the posterior cingulate, precuneus, lateral parietal lobe, and medial prefrontal cortex, and it is usually deactivated during memory encoding and other cognitively demanding tasks focused on processing external stimuli [20 –22]. The DMN shows abnormal responses in the posterior cingulate and precuneus during a memory task in AD patients and high-risk (of AD) subjects. Namely, these subjects tend to manifest a paradoxical increase in fMRI activity above baseline [21 –23], whereas healthy young people exhibit beneficial deactivations in these regions during memory tasks [24, 25].

Nevertheless, despite the remarkable advances in our understanding of the neurological alterations found in AD in the past two decades, at the time when a clinical diagnosis can be made, significant neuronal losses and irreversible neuropathological changes have already taken place [26]. Thus, it is crucial to work on developing early biomarkers that enable us to detect alterations in brain functions before the neuronal damage becomes irreversible, with the aim of successfully intervening in the course of the disease. For this reason, research on AD has shifted its focus to prodromal stages of AD, when mild cognitive impairment (MCI) patients are susceptible to developing AD [27 –30]. The results obtained in studies on the brain activation associated with memory encoding in MCI patients vary considerably. Whereas some studies report a decreased fMRI activation in the MTL in MCI patients compared to older control subjects [31 –34], other studies did not find any differences between these groups during memory encoding, although they reported a differential brain activation during memory recovery [35]. Furthermore, several studies reported an MTL hyperactivation in MCI patients compared to older control subjects during stimuli encoding [36 –39]. In this regard, the increased activity in the MTL might be interpreted as a compensatory mechanism associated with the prodromal phase of AD. In this way, the previous findings may be reconciled if the activation throughout the course of AD has an “inverse U curve” shape, because MCI patients could be distributed across a continuum in terms of their degree of MTL activation [18].

In the present study, we assumed the involvement of the hippocampus and precuneus in the development of AD within the context which considers this pathology as a dysfunction in distributed brain networks. For this reason, we used independent component analysis (ICA) to investigate the effects of AD on the brain networks associated with these structures. Unlike previous studies using conventional general linear model (GLM) analysis, ICA may serve to reveal the hidden factors that underlie fMRI signals, allowing the study of the specific time course associated with a component (i.e., a functional network) separately from the signals associated to other components (i.e., other networks and/or artifacts). Thus, the analysis of task-related modulations in the time courses of the functional networks generated by ICA can provide new insights into the brain’s functional organization that are not observed in conventional GLM analysis [40]. In this regard, we expected to identify the structures that work in a coordinated network with the precuneus and hippocampal areas during a memory encoding task. For the picture-encoding task selected for this study, strong bilateral activations have previously been reported in the temporal lobe [41, 42], especially in the parahippocampal formation, hippocampus, and fusiform and lingual gyrus. Furthermore, we investigated how AD impairs the activity of these networks. Based on the reviewed literature, we hypothesized that MCI patients would show a hyperactivation in these networks compared to other groups. We also expected that AD patients would display hypoactivation in the hippocampal network compared to healthy controls (HC) or MCI. Conversely, we expected that AD patients would not show proper deactivation of the DMN, compared to HC or MCI.

MATERIAL AND METHODS

Participants

The study included 109 participants organized in three groups: 68 MCI patients (mean age: 73.34±5.17, gender (M/F): 28/40) and 21 AD patients (mean age: 75±3.83, gender (M/F): 8/13), both recruited from dementia units of the Valencian Community Healthcare System, and 20 healthy older volunteers (mean age: 72.85±5.51, gender (M/F): 10/10). AD and MCI diagnoses were made by experienced neurologists and based on clinical and neuropsychological evidence. The AD group was composed of patients who met revised criteria for probable AD [43] and showed a Clinical Dementia Rating (CDR) score of 1 (mild AD). For the MCI group, the inclusion criteria included 1) memory complaints (self-reported or confirmed by an informant); 2) objective memory impairment assessed with the logical memory subtest II from the Wechsler memory scale-III (WMS III); 3) essentially intact activities in daily living; 4) no evidence of dementia; and 5) a CDR score of 0.5. Cognitively normal individuals were included in the HC group if they had no memory complaints, normal performance on the neuropsychological assessment (see below), and a CDR score of 0. Participants were excluded if they had any of the following clinical characteristics: 1) other nervous system diseases such as a brain tumor, cerebrovascular disease, encephalitis, or epilepsy, or that they met the criteria for other dementias different from AD; 2) a Geriatric Depression Scale score ≥6 [44, 45]; 3) visible abnormalities reported by an experienced radiologist in magnetic resonance images, such as leukoaraiosis and infarction; 4) current psychiatric disorder or use of psychoactive medication.

All participants underwent a structured clinical interview and a neuropsychological assessment, which included the Mini-Mental State Examination (MMSE) [46, 47], Functional Activities Questionnaire (FAQ) [48], short form of the Boston naming test [49], Digit subtest (forward and backward) from the WMS-III [50], Similarities subtests from the Wechsler adult intelligence scale-III (WAIS-III) [51], and logical memory subtests (I and II) from the WMS-III [50]. The MCI patients were followed up clinically with periodic neuropsychological assessments and clinical interviews every 6 months for one and a half years, although the MR data was acquired only once in the first clinical visit. The MCI patients who received an AD diagnosis during the follow-up period (within a period of 18 months after the fMRI recording) were classified as MCI converters (MCIc; N = 21), whereas patients who remained MCI after this period were classified as MCI non-converters (MICn; N = 28). Nineteen MCI participants did not complete the follow-up period and, consequently, were excluded from the analyses involving the longitudinal recording. Statistics for neuropsychological assessment and demographic data are presented in Table 1.

Table 1

Demographic, clinical and neuropsychological data of all participants

Neuropsychological and demographic data	HC	MCIn	MCIc	AD
N	20	28	21	21	Differences
Age	72.85±5.50	71.89±5.38	75.81±4.68	75.00±5.01	F_3,86 = 3.21; p < 0.05
Sex (men/women)	10/10	13/15	7/14	8/13	χ² = 1.53; p = 0.68
MMSE	29.50±0.83	28.46±1.67	26.00±2.49	22.90±3.48	F_3,84 = 32.93; p < 0.001
FAQ	0.50±0.61	4.21±1.47	6.76±5.88	13.76±7.08	F_3,85 = 30.22; p < 0.001
Boston	11.95±0.22	9.64±1.16	8.57±1.78	7.38±3.67	F_3,85 = 16.57; p < 0.001
WMS-III: Digit subtest
Forward	7.25±0.44	5.11±0.83	5.14±1.42	4.81±1.78	F_3,85 = 19.15; p < 0.001
Backward	6.20±0.95	3.68±0.55	3.24±0.89	2.57±1.40	F_3,85 = 54.49; p < 0.001
WAIS-III: Similarities subtests	17.75±2.31	13.28±2.12	9.43±3.40	6.95±4.19	F_3,85 = 51.89; p < 0.001
WMS-III: Logical memory subtests
Short-term	13.35±1.60	10.79±2.41	8.85±3.59	3.55±2.91	F_3,84 = 47.04; p < 0.001
Long-term	11.20±2.46	8.36±1.97	6.85±3.04	2.05±2.31	F_3,84 = 50.86; p < 0.001
Memory task
Hits percentage (%)	77.43±9.75	71.82±16.01	63.82±11.11	65.07±11.23	F_3,84 = 3.37, p < 0.05

HC, healthy controls; MCIn, mild cognitive impaired no-converter; MCIc, mild cognitive impaired converter; AD, Alzheimer’s disease.

Participants were informed of the nature of the research, and they provided their written informed consent prior to their participation in the study. This research study was approved by the Institutional Review Board of the Jaume I of Castellón University. All the study procedures conformed to the Code of Ethics of the World Medical Association.

Experimental design and stimuli

The task was based on previous studies investigating memory function [37]. Participants performed a memory encoding task (see Fig. 1) while undergoing fMRI scanning. They had to memorize a total of 48 different landscape pictures displayed with VisuaStim goggles (Resonance Technology, Northridge, CA, USA). The activation task consisted of two conditions that were alternated in 16 blocks: 1) Baseline (8 blocks): Subjects viewed a white fixation cross centered on a black screen for 18 s; 2) Encoding (8 blocks): Subjects viewed 6 different landscapes, each presented for 2.5 s and with an interstimulus interval of 500 ms; this block had a duration of 18 s. The total duration of the task was 288 s. The task was programmed and presented with the Presentation software (Neurobehavioral Systems, Albany, CA, USA).

Fig.1

Experimental procedure (top) and fMRI encoding task (bottom). The encoding task was composed from a total of 16 blocks (8 encoding blocks + 8 baseline blocks). While encoding blocks, participants had to memorize 6 different landscapes, each displayed for 2.5 s with an interstimulus interval of 500 ms.

Immediately after the fMRI scan, the participants performed a recognition test. It consisted of distinguishing the landscapes they had seen during the scanner presentation from other landscapes that were not presented during the fMRI scan. Target and non-target images were randomly selected from the same pool of landscapes prior to the study. Correct responses (hits) were recorded for further analysis.

Data acquisition

Images were acquired on a 3T scanner (Siemens Trio). Participants were placed in a supine position in the MRI scanner, and their heads were immobilized with cushions to reduce motion artifacts. For the fMRI, a total of 120 volumes were recorded over 4’8 min using a gradient-echo T2*-weighted echo-planar imaging sequence (TR = 2400 ms; TE = 30 ms; matrix, 64×64; voxel size, 3.8×3.8 mm; flip angle, 90°; slice thickness, 3.5 mm; slice gap, 0.5 mm). We acquired 33 interleaved axial slices parallel to the anterior–posterior commissure plane covering the entire brain. Before the functional fMRI sequences, a high-resolution structural T1-weighted MPRAGE sequence was acquired (TR = 2300 ms, TE = 2.98 ms; flip angle 9°, matrix = 256×256; voxel size = 1 mm³).

Data processing

Functional MRI data were preprocessed using SPM12 (Wellcome Trust Center for Neuroimaging, London, UK). It included noise filtering (automatic detection and repairing bad slices) with the ArtRepair toolbox for SPM (http://cibsr.stanford.edu/tools/humanbrain-project/artrepair-software.html), realignment to correct for motion-related artifacts, spatial normalization into the standard Montreal Neurological Institute (MNI) space, and smoothing using a Gaussian kernel with a full-width at half-maximum (FWHM) of 8 mm. Inspection of realignment parameters showed that none of the participants moved their heads more than 2 mm/degrees in any of the six directions during functional data recording.

Group independent component analysis

Group independent components analysis was performed using the GIFT toolbox (v3.0b, http://icatb.sourceforge.net) and the Infomax algorithm [52] to obtain functional networks that underlie fMRI data [53]. The group spatial ICA is performed on all the subjects at once and provides an independent component spatial map and a single associated ICA time course for every component, subject, and session. Significant between-group differences are determined by a second-level analysis of the ICA results. The objective of our study was to investigate between-group differences in the activity of the precuneus and hippocampus networks during memory encoding. Thus, we performed a GLM analysis in the components’ time courses, estimated by ICA, to determine how the activity of the different brain networks was modulated by the experimental conditions. To study between-group comparisons, we used the beta-weights obtained from the GLM in second-level analyses. Fifty iterations of the ICA analysis were performed with the ICASSO software to ensure the stability of the estimated components. ICA dimensionality was set at 24 independent components based on minimum description length criteria [54]. Prior to the ICA, the intensity of images was normalized, and data dimensionality was reduced through a principal-components analysis (PCA). After ICA decomposition, individual independent component maps and time courses were computed using the GICA-3 back-reconstruction approach.

Component selection

In order to determine the functional networks comprising the hippocampus and precuneus, spatial correlations were run between the averaged spatial maps of each component and region of interest (ROI) of these structures. Probabilistic masks of the bilateral hippocampus and precuneus were extracted from subcortical and cortical Harvard-Oxford atlases, respectively. The components that showed the highest correlation with each ROI were selected for the subsequent analyses. After the identification of the components of interest, subject-specific spatial maps for these components were used to determine the regions belonging to the hippocampal and precuneus networks through whole brain voxel-wise one-sample t-test analyses. Following previous studies [55, 56], the statistical threshold for these analyses was set at p < 1×10^–13 FDR-corrected).

Statistical analyses

In order to study how the hippocampal and precuneus networks were modulated during the task, a GLM, as implemented in the GIFT toolbox, was performed for each subject and component. The subject-specific time courses for each component of interest were set as dependent variable. The GLM design matrix included as independent variables a regressor defining the occurrence memory blocks and the six parameters that modeled residual motion. The task regressor was convolved with the canonical hemodynamic response function. Once the analyses had been performed, the beta-weights associated with the memory task were used to perform the second-level analyses.

Second-level analyses were conducted to study between-group differences in the activity of the hippocampal and precuneus networks. In these analyses, the averaged beta-weights for each group derived from the longitudinal classification (HC, MCIn, MCIc, and AD) were compared using a one-way ANCOVA as implemented in SPSS 23 (IBM Corp.), with age as covariate.

Voxel based morphometry

To complement the functional measures obtained using ICA analysis, we studied regional atrophy in our sample. Specifically, we performed a voxel-based morphometry (VBM) using CAT12 (Computational Anatomy Toolbox; C. Gaser, Jena University Hospital, Jena, Germany; http://dbm.neuro.uni-jena.de/cat/). We used standard VBM preprocessing which included bias correction of intensity non-uniformities, spatial normalization to MNI template using the Diffeomorphic Anatomic Registration Through Exponentiated Lie algebra (DARTEL) algorithm [57], tissue segmentation into grey matter (GM), white matter, and cerebrospinal fluid, modulation using the Jacobian determinants and smooth with an 8 mm full-width-at-half-maximum Gaussian kernel.

We investigated focal GM volume differences in a priori ROIs, specifically using the same hippocampus and precuneus from the Harvard-Oxford that we used to select the components of interest in the previous ICA analysis, but, in this case, with a minimum 50% probability threshold (see Fig. 4a). The GM volumes were obtained for each ROI and each subject from the modulated and smoothed images. Next, a one-way ANCOVA was performed using IBM SPSS Statistics 23 (IBM Corp) in order to compare GM volumes in target regions between the four experimental groups (HC, MCIn, MCIc, and AD), controlling by age and total intracranial volume (TIV).

Recognition test analysis

SPSS 23 (IBM Corp) was used to process the behavioral data. The percentage of correct responses during the post-scanner recognition test was used as the accuracy index. A one-way ANCOVA [Group (HC, MCIn, MCIc, and AD); covariate: age] was conducted in order to assess potential differences among groups in the accuracy ratio. This analysis was followed by Bonferroni post-hoc tests.

Additionally, we studied the relationship between the participant’s performance and the activity of the components of interest. Specifically, we investigated whether the activity of the hippocampal and precuneus networks were positively and negatively correlated with the percentage of hits, respectively. To test these hypotheses, we performed partial correlations between beta-weights for each network of interest and the accuracy index, including age as covariate. Similar partial correlation analyses were performed to study the relationship between hippocampus and precuneus volumes obtained in VMB analysis and participants performance using age and TIV as covariates.

RESULTS

Neuropsychological results

As expected, neuropsychological results revealed statistically significant between-group differences. The MMSE scores distinguished: a) HC compared to the MCIc (p < 0.001) and AD groups (p < 0.001); b) MCIn compared to MCIc (p < 0.01) and AD (p < 0.001); and c) MCI compared to AD (p < 0.001). All the other neuropsychological tests revealed statistically significant differences (p < 0.001) between HC and each patient group (MCIn, MCIc, and AD). In addition, on these neuropsychological tests, a progressive performance impairment was observed across groups (HC>MCIn>MCIc>AD). Additionally, the MCIn group showed statistically significant better performance than the AD group (p < 0.01) on the FAQ, Boston, Backward digits, Similarities, and Logic Memory (short and long-term) tests. Finally, AD had statistically significant worse performance on the Logic Memory tests, compared to MCIc (p < 0.01).

Independent component analysis results

Spatial correlations revealed that C6 (r = 0.147) was the component that correlated most with the hippocampus, and C19 (r = 0.40) was the component that correlated most with the precuneus. As Fig. 2a and Table 2a show, the hippocampal network included the bilateral hippocampus, parahippocampus, lingual gyrus, fusiform gyrus, cerebellum (lobules IV and V), and cerebellar vermis (lobules III, IV and V). The one-way ANCOVA revealed significant differences between groups in the activity of this network during the task (F_3,85 = 3.38, p = 0.022). As Fig. 2b shows, Bonferroni post-hoc analysis revealed that MCIc patients have a significantly higher activation (p < 0.05) in this network than MCIn and AD patients.

Fig.2

a) Hippocampal network spatial map. Results FDR-corrected at p < 1×10–13; k = 30. Colored bars express t-scores. b) Between-group differences in hippocampal network activity during fMRI task (memory encoding > baseline). MCIc showed a significant hyperactivation compared to MCIn and AD patients. c) Non-significant correlation between hippocampal network activity and hits on the recognition memory task covariate by age.

Table 2

Brain regions belonging to hippocampus and precuneus networks (p < 1×10^–13 FDR-corrected, k = 30)

Regions	R/L	BA	MNI coordinates			Voxels number	T	K size
			x	y	z
a) Hippocampus Network (Component 6)
Cerebellum, lobules IV, V	L	–	–15	–31	–19	276	29.10	2638
Parahippocampus	L	30	–15	–31	–16	120	29.08
Parahippocampus	R	30	18	–25	–19	151	28.08
Fusiform gyrus	L	30	–15	–34	–16	90	27.58
Lingual gyrus	L	27/30	–12	–34	–10	117	26.00
Cerebellum, lobules IV, V	R	–	18	–34	–16	217	25.55
Fusiform gyrus	R	30	21	–37	–16	90	24.66
Lingual gyrus	R	30	18	–37	–13	161	23.39
Vermis, lobule III	–	–	6	–37	–13	62	22.67
Hippocampus	R	30	24	–22	–16	72	21.90
Hippocampus	L	30	–15	–28	–10	94	20.66
Vermis, lobule IV, V	–	–	–3	–49	–22	106	18.78
b) Precuneus Network (Component 19)
Precuneus	R	23	6	–55	35	595	39.47	9929
Angular gyrus	R	39	51	–55	38	381	37.67
Angular gyrus	L	39	–51	–64	32	349	36.61
Midcingulate gyrus	R	23	0	–22	35	398	33.98
Midcingulate gyrus	L	23	–3	40	38	384	32.14
Middle temporal gyrus	R	21	60	–52	11	586	22.08
Middle temporal gyrus	L	21	–63	–37	–7	569	18.80
Middle frontal lobe	R	9	42	20	44	147	13.88	315
Superior frontal lobe	R	8	21	26	59	67	12.47
Middle frontal lobe	L	9	–42	17	47	225	18.18	258
Superior frontal lobe	L	8	–15	26	62	30	10.41
Cerebellum Crus1	L	–	–33	–67	–34	60	14.66	161
Cerebellum, lobule VI	R	–	15	–67	–25	25	10.88
Vermis, lobule VI	–	–	0	–73	–16	26	9.97
Cerebellum, lobule VI	L	–	–9	–64	–19	27	9.69
Vermis, lobule VII	–	–	3	–70	–25	11	9.31
Orbitofrontal medial lobe	L	10	–9	47	–4	46	12.29	153
Superior medial frontal lobe	R	10	3	50	4	18	11.69
Superior temporal pole	L	38	–36	14	–22	81	15.70	116
Superior temporal pole	R	38	39	17	–25	32	12.01	38

R/L, Right/Left; BA, Brodmann Areas; MNI, Montreal Neurological Institute; Voxel number, Number of voxels of each structure into the cluster; K size, cluster size measured by voxels.

Table 3

Regional atrophy, GM-volume between groups differences (p < 0.001, FWEc corrected)

Between-group differences	R/L	K size	MNI coordinates			Z value	T value
			x	y	z
HC >MCIn
Fusiform	L	1041	–32	–18	–26	4.02	4.24
Hippocampus	L		–33	–23	–17	3.53	3.69
Parahippocampus	L		–24	6	–24	3.48	3.62
Amygdala	L		–29	0	–27	3.17	3.28
HC >MCIc
Inferior temporal lobe	L	1267	–62	–33	–20	4.81	5.19
Middle temporal lobe	L		–59	2	–29	3.99	4.21
Fusiform	L	2002	–30	3	–39	4.33	4.61
Amygdala	L		–26	3	–23	3.92	4.13
HC >AD
Superior temporal pole	R	5046	27	8	–23	5.17	5.64
Hippocampus	R		24	–11	–14	4.34	4.61
Fusiform	R		27	–35	–15	4.26	4.56
Inferior orbitofrontal	R		26	27	–15	4.21	4.46
Parahippocampus	R		35	–26	–17	3.88	4.07
Inferior temporal	R		35	9	–42	3.80	3.98
Middle temporal pole	R		30	14	–33	3.71	3.88
Superior temporal pole	L	5304	–26	5	–23	5.00	5.43
Fusiform	L		–27	–33	–18	4.92	5.33
Inferior temporal	L		–32	5	–39	4.42	4.72
Hippocampus	L		–18	–35	0	4.56	4.89
Parahippocampus	L		–30	–30	–12	4.55	4.87

R/L, Right/Left; MNI, Montreal Neurological Institute; K size, cluster size measured by voxels.

As Fig. 3a and Table 2b show, the precuneus network matched the key regions of the DMN, and included the right precuneus, bilateral angular gyrus, midcingulate gyrus, middle temporal gyrus, middle frontal lobe, superior frontal lobe, cerebellum (lobule VI), vermis (lobules VI and VII), superior temporal pole, and left orbitofrontal medial. The one-way ANCOVA revealed no significant differences between groups in the activity of this network during the task (F_3,85 = 0.64, p = 0.577; see Fig. 3b).

Fig.3

a) Precuneus network spatial map. Results FDR-corrected at p < 1×10–13; k = 30. Colored bars express t-scores. b) Non-significant between-group differences in precuneus network activity during fMRI task (memory encoding > baseline). c) Negative correlation between precuneus network activity and hits on the recognition memory task covariate by age.

Finally, given that our groups presented differences in age, we replicated the analyses investigating between-group differences in network activity, but using a subsample of participants matched on this variable. The matching was carried out with the MatchIt function implemented in R Project for Statistical Computing and using the “nearest” method. This procedure provided four matched groups of 20 participants each (HC = 72.85±5.509, MCIn = 72.25±5.004, MCIc = 75.45±4.501, AD = 74.80±3.820; F_(3,76) = 2.075, p = 0.111). The analyses using this subsample revealed similar results that the analyses with the whole sample: MCIc patients showed a greater activity compared to MCIn patients (t₃₈ = 2.146, p < 0.05) and also compared to AD patients (t₃₈ = 2.224, p < 0.05) in the hippocampus network. Furthermore, no significant differences were found in the precuneus network.

Voxel based morphometry results

One-way ANCOVAs revealed statistically significant differences across groups in GM hippocampus volume (F3,82 = 6.938, p < 0.001), but not in GM precuneus volume (F3,82 = 2.016, p = 0.118). Specifically, the main differences in GM hippocampus volume were found between HC and MCIn patients (p < 0.05), HC and MCIc patients (p < 0.05), and HC and AD patients (p < 0.001). Overall, these results showed hippocampus atrophy in patients compared to HC (see Fig. 4b).

Fig.4

a) Hippocampus (red) and precuneus (green) ROIs taken from the Harvard-Oxford atlas used for VBM analysis. b) Between-group differences in GM hippocampus volume (F3,82 = 6.938, p < 0.001) and GM precuneus volume (F3,82 = 2.016, p = 0.118).

Recognition test results

Task performance of each group was measured by the mean of the percentage of hits: HC = 77.43±9.75%; MCIn = 71.82±16.01%; MCIc = 63.82±11.11% and AD = 65.07±11.23%. The one-way ANCOVA revealed a statistically significant difference between groups (F_3,83 = 3.37, p < 0.05).

Additionally, as Fig. 2c shows, the activity of the precuneus network showed a negative correlation with participants’ performance (r = –0.223, p < 0.05). No significant correlations were observed between the hippocampal network and accuracy (r = 0.117, p = 0.23). Furthermore, a significant positive correlation was found between GM hippocampus volume and participants’ performance (r = 0.385, p < 0.001).

DISCUSSION

The present fMRI research focused on studying the underlying networks associated with the key regions related to memory impairments in AD: hippocampus and precuneus [21–23 , 58–60]. To accomplish this, we performed an ICA analysis and obtained the functional connectivity networks associated with these structures, along with their specific temporal courses during a memory encoding task (landscape encoding versus fixation). Additionally, we performed a VBM analysis to obtain complementary brain atrophy measures. On the one hand, our results showed that the hippocampal network revealed differential activation between groups; specifically, MCIc patients showed hyperactivation in this network compared to MCIn and AD patients. On the other hand, the precuneus network matched the key regions of the DMN, and its activity was negatively correlated with task performance. Additionally, we found GM volume differences between HC and all the patient groups (MCIn, MCIc, and AD) in the hippocampus. Overall, our findings support the hypothesis that DMN deactivation is relevant in memory encoding, and they suggest that hippocampal network activity may serve to detect MCI subjects at risk of conversion to AD.

Classically, the MTL has been the key region in the search for predictive biomarkers during the prodromal phase of AD. It has been studied from a volumetric perspective, where we can find lost hippocampal volume to be an early biomarker of AD in subjects with MCI [61 –64]. Moreover, functional research has associated AD patients with an MTL hypoactivation compared to healthy older controls [10 –17]. Regarding MCI, patients’ results have been heterogeneous, and several studies have associated MCI patients with increased MTL activation during a memory task when compared to healthy-older controls. To explain this, Sperling et al. [65, 66] proposed the inverse-U hypothesis, which argues that MTL activation could be course-dependent, such that MTL activation increases as degeneration of this region begins, acting as a compensatory mechanism. The increasing activity continues until reaching a breaking point when MTL degeneration is too severe, making a compensatory mechanism impossible. A recent meta-analysis [67] collected different studies that compared brain activity during episodic memory tasks in healthy older controls, MCI patients, and AD patients, using fMRI. This study concluded that HC had significantly higher activation in the hippocampus than AD patients, whereas MCI patients showed a significant hyperactivation in the cerebellum and lingual gyrus compared to HC. The regions reported in this meta-analysis corresponded to the hippocampal network obtained in the present study, which included the cerebellum (lobules IV and V), lingual gyrus, parahippocampus, vermis (lobules III, IV and V), hippocampus, and fusiform gyrus. The ICA procedure followed in our study enables the estimation of the specific temporal signal associated with this network. Thus, our study adds to the previous literature by demonstrating that: 1) these regions make up a functional network and 2) that this network shows a hyperactivation in MCIc patients during memory encoding. Our findings do not reveal significant differences between HC and MCIn or between HC and AD patients; nevertheless, the differential hyperactivation shown by the MCIc group is able to discriminate between “low-risk” (MCIn) and “high-risk” (MCIc) MCI patients. Following previous research, hyperactivation of the MTL during the task is predictive of greater brain atrophy [37 , 68]. Additionally, event-related fMRI studies [38 , 69] found that this hyperactivation occurred specifically on successful trials; therefore, Sperling suggested that “the increased activity may serve as a compensatory mechanism in the setting of early AD pathology” [18].

As expected, the precuneus network matched the key regions of the DMN, including the precuneus, bilateral angular gyrus, and medial regions of the prefrontal cortex. The pattern of activity of this network is inconsistent with our hypothesis that MCI and AD patients would show less hypoactivation than HC in this network. Nevertheless, the activation pattern showed a significant negative correlation with the participants’ performance; that is, the worse the performance, the greater the activation. As we have seen in the recognition test results, MCIc and AD patients obtained significantly less accuracy than HC, but these groups do not show a differential activation in the DMN. This finding supports the idea that AD patients show a decline of the beneficial deactivation of DMN regions during task performance, which has been previously associated with better performance on memory tasks [21 , 70–72].

As a complementary analysis, we investigated volumetric measures of hippocampus and precuneus in our sample. Previous research indicates that hippocampus atrophy may serve as an early biomarker of AD in subjects with MCI [61 –64]. In this study we found significant higher GM hippocampal volume in HC than MCIn, MCIc, and AD. Furthermore, the volume of hippocampus correlated with the percentage of hits of the recognition task. However, we did not find significant differences between our MCIn and MCIc groups. Taken together, the results of this study may have clinical implications. In our sample, hippocampus volume was able to discriminate between patients and controls, while hippocampus network activity was able to discriminate between MCIn and MCIc. These results suggest that the hyperactivity of the hippocampus network may be a potential marker in detecting individuals at risk of conversion to AD, at least in the following 18 months after recording. By contrast, this measure would not be sensitive enough to discriminate between HC and early AD. Given the inverse-U shaped pattern, HC and early AD may have similar activation indices. In this regard, the study of hippocampus volume would be a better measure, given that it follows a linear descending pattern. Finally, our study is limited in the use of a fixation cross as a baseline condition. A scrambled image rather than a black screen with a white fixation cross would have been more appropriate to disentangle pure vision effects from landscape encoding.

In conclusion, our findings show differential hippocampal network activity between MCI patients at “high-risk” of developing AD and “low-risk” MCI patients. Specifically, we found that “high-risk” MCI patients present a hyperactivation in this network which is in line with the assumption of a compensatory mechanism within this stage of AD. Furthermore, although no significant differences in activation were found between groups, the precuneus network was inversely related to participants’ performance, which means that participants with greater precuneus-network deactivation obtained better results on the memory task.

Footnotes

ACKNOWLEDGMENTS

This research was supported by the project (201410-30-31) provided by Fundació Marató TV3 awarded to C.A. NA was supported by an FPU grant from the Spanish Ministry of Education (FPU16/01525), and V.C was supported by a Juan de la Cierva post-doctoral graduate program grant from the Spanish Ministry of Economy, Industry and Competitiveness (IJCI-2016-29247).

Authors’ disclosures available online ().

References

Alzheimer’s Association (2016) 2016 Alzheimer’s disease facts and figures. Alzheimers Dement 12, 459–509.

Tomaskova

, Kuhnova

, Cimler

, Dolezal

, Kuca

(2016) Prediction of population with Alzheimer’s disease in the European Union using a system dynamics model. Neuropsychiatr Dis Treat 12, 1589–98.

Martin Prince

, Albanese

, Guerchet

, Prina

, Richard Pender

, Ferri

, Mazzotti

, Piovezan

, Padilla

, Luchsinger

(2014) World Alzheimer Report 2014: Dementia and Risk Reduction. An Analysis of Protective and Modifiable Factors. Alzheimer’s Disease International, London, UK.

Hyman

, Van Hoesen

, Damasio

, Barnes

(1984) Alzheimer’s disease: Cell-specific pathology isolates the hippocampal formation. Science 225, 1168–1170.

Sperling

, Bates

, Chua

, Cocchiarella

, Rentz

, Rosen

, Schacter

(2003) fMRI studies of associative encoding in young and elderly controls and mild Alzheimer’s disease. J Neurol Neurosurg Psychiatry 74, 44–50.

Delbeuck

, Linden

M Van der

, Collette

(2003) Alzheimer’s disease as a disconnection syndrome? Neuropsychol Rev 13, 79–92.

Sperling

, Dickerson

, Pihlajamaki

, Vannini

, LaViolette

, Vitolo

, Hedden

, Becker

, Rentz

, Selkoe

, Johnson

(2010) Functional alterations in memory networks in early Alzheimer’s disease. Neuromolecular Med 12, 27–43.

Brier

, Thomas

, Ances

(2014) Network dysfunction in Alzheimer’s disease: Refining the disconnection hypothesis. Brain Connect 4, 299–311.

Grady

, Furey

, Pietrini

, Horwitz

, Rapoport

(2001) Altered brain functional connectivity and impaired short-term memory in Alzheimer’s disease. Brain 124, 739–756.

10.

Pariente

, Cole

, Henson

, Clare

, Kennedy

, Rossor

, Cipoloti

, Puel

, Demonet

, Chollet

, Frackowiak

RSJ

(2005) Alzheimer’s patients engage an alternative network during a memory task. Ann Neurol 58, 870–879.

11.

Rombouts

, Barkhof

, Goekoop

, Stam

, Scheltens

(2005) Altered resting state networks in mild cognitive impairment and mild Alzheimer’s disease: An fMRI study. Hum Brain Mapp 26, 231–239.

12.

Sperling

, Chua

, Cocchiarella

, Rand-Giovannetti

, Poldrack

, Schacter

, Albert

(2003) Putting names to faces: Successful encoding of associative memories activates the anterior hippocampal formation. Neuroimage 20, 1400–1410.

13.

Golby

, Silverberg

, Race

, Gabrieli

, O’Shea

, Knierim

, Stebbins

, Gabrieli

(2005) Memory encoding in Alzheimer’s disease: An fMRI study of explicit and implicit memory. Brain 128, 773–787.

14.

Rombouts

, Barkhof

, Veltman

, Machielsen

, Witter

, Bierlaagh

, Lazeron

, Valk

, Scheltens

(2000) Functional MR imaging in Alzheimer’s disease during memory encoding. AJNR Am J Neuroradiol 21, 1869–1875.

15.

Kato

, Knopman

, Liu

(2001) Dissociation of regional activation in mild AD during visual encoding: A functional MRI study. Neurology 57, 812–816.

16.

Solé-Padullés

, Bartrés-Faz

, Junqué

, Vendrell

, Rami

, Clemente

, Bosch

, Villar

, Bargalló

, Jurado

, Barrios

, Molinuevo

(2009) Brain structure and function related to cognitive reserve variables in normal aging, mild cognitive impairment and Alzheimer’s disease. Neurobiol Aging 30, 1114–1124.

17.

Rémy

, Mirrashed

, Campbell

, Richter

(2005) Verbal episodic memory impairment in Alzheimer’s disease: A combined structural and functional MRI study. Neuroimage 25, 253–266.

18.

Sperling

, Dickerson

, Pihlajamaki

, Vannini

, LaViolette

, Vitolo

, Hedden

, Becker

, Rentz

, Selkoe

, Johnson

(2010) Functional alterations in memory networks in early Alzheimer’s disease. Neuromolecular Med 12, 27–43.

19.

Cavanna

, Trimble

(2006) The precuneus: A review of its functional anatomy and behavioural correlates. Brain 129, 564–583.

20.

Raichle

, MacLeod

, Snyder

, Powers

, Gusnard

, Shulman

(2001) A default mode of brain function. Proc Natl Acad Sci U S A 98, 676–682.

21.

Buckner

, Andrews-Hanna

, Schacter

(2008) The brain’s default network: Anatomy, function, and relevance to disease. Ann N Y Acad Sci 1124, 1–38.

22.

Schwindt

, Black

(2009) Functional imaging studies of episodic memory in Alzheimer’s disease: A quantitative meta-analysis. Neuroimage 45, 181–190.

23.

Petrella

, Prince

, Wang

, Hellegers

, Doraiswamy

(2007) Prognostic value of posteromedial cortex deactivation in mild cognitive impairment. PLoS One 2, e1104.

24.

Daselaar

, Fleck

, Dobbins

, Madden

, Cabeza

(2006) Effects of healthy aging on hippocampal and rhinal memory functions: An event-related fMRI study. Cereb Cortex 16, 1771–82.

25.

Krajcovicova

, Mikl

, Marecek

, Rektorova

(2014) Disturbed default mode network connectivity patterns in Alzheimer’s disease associated with visual processing. J Alzheimers Dis 41, 1229–1238.

26.

DeKosky

, Marek

(2003) Looking backward to move forward: Early detection of neurodegenerative disorders. Science 302, 830–834.

27.

Petersen

, Smith

, Waring

, Ivnik

, Kokmen

, Tangelos

(1997) Aging, memory, and mild cognitive impairment. Int Psychogeriatrics 9(Suppl 1), 65–69.

28.

Petersen

, Smith

, Waring

, Ivnik

, Tangalos

, Kokmen

(1999) Mild cognitive impairment: Clinical characterization and outcome. Arch Neurol 56, 303–308.

29.

Petersen

(2000) Mild cognitive impairment: Transition between aging and Alzheimer’s disease. Neurologia 15, 93–101.

30.

Kantarci

, Jack

, Xu

, Campeau

, O’Brien

, Smith

, Ivnik

, Boeve

, Kokmen

, Tangalos

, Petersen

(2000) Regional metabolic patterns in mild cognitive impairment and Alzheimer’s disease: A 1H MRS study. Neurology 55, 210–217.

31.

Machulda

, Ward

, Borowski

, Gunter

, Cha

, O’Brien

, Petersen

, Boeve

, Knopman

, Tang-Wai

, Ivnik

, Smith

, Tangalos

, Jack

(2003) Comparison of memory fMRI response among normal, MCI, and Alzheimer’s patients. Neurology 61, 500–506.

32.

Johnson

, Schmitz

, Trivedi

, Ries

, Torgerson

, Carlsson

, Asthana

, Hermann

, Sager

(2006) The influence of Alzheimer disease family history and apolipoprotein E4 on mesial temporal lobe activation. J Neurosci 26, 6069–6076.

33.

Parra

, Pattan

, Wong

, Beaglehole

, Lonie

, Wan

, Honey

, Hall

, Whalley

, Lawrie

(2013) Medial temporal lobe function during emotional memory in early Alzheimer’s disease, mild cognitive impairment and healthy ageing: An fMRI study. BMC Psychiatry 13, 76.

34.

van der Meulen

, Lederrey

, Rieger

, van Assche

, Schwartz

, Vuilleumier

, Assal

(2012) Associative and semantic memory deficits in amnestic mild cognitive impairment as revealed by functional magnetic resonance imaging. Cogn Behav Neurol 25, 195–215.

35.

Petrella

, Krishnan

, Slavin

, Tran

T-TT

, Murty

, Doraiswamy

(2006) Mild cognitive impairment: Evaluation with 4-T functional MR imaging. Radiology 240, 177–186.

36.

Dickerson

, Salat

, Greve

, Chua

, Rand-Giovannetti

, Rentz

, Bertram

, Mullin

, Tanzi

, Blacker

, Albert

, Sperling

(2005) Increased hippocampal activation in mild cognitive impairment compared to normal aging and AD. Neurology 65, 404–411.

37.

Hämäläinen

, Pihlajamäki

, Tanila

, Hänninen

, Niskanen

, Tervo

, Karjalainen

, Vanninen

, Soininen

(2007) Increased fMRI responses during encoding in mild cognitive impairment. Neurobiol Aging 28, 1889–1903.

38.

Heun

, Freymann

, Erb

, Leube

, Jessen

, Kircher

, Grodd

(2007) Mild cognitive impairment (MCI) and actual retrieval performance affect cerebral activation in the elderly. Neurobiol Aging 28, 404–413.

39.

Kircher

, Weis

, Freymann

, Erb

, Jessen

, Grodd

, Heun

, Leube

(2007) Hippocampal activation in patients with mild cognitive impairment is necessary for successful memory encoding. J Neurol Neurosurg Psychiatry 78, 812–818.

40.

, Potenza

, Calhoun

(2013) Spatial ICA reveals functional activity hidden from traditional fMRI GLM-based analyses. Front Neurosci 7, 154.

41.

Barrós-Loscertales

, Forn

, Mallo

, Parcet

, Belloch

, Campos

, Feliu-Tatay

, González-Darder

(2006) Memory lateralization with 2 functional MR imaging tasks in patients with lesions in the temporal lobe. AJNR Am J Neuroradiol 27, 498–503.

42.

Detre

, Maccotta

, King

, Alsop

, Glosser

, D’Esposito

, Zarahn

, Aguirre

, French

(1998) Functional MRI lateralization of memory in temporal lobe epilepsy. Neurology 50, 926–932.

43.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

, Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

44.

Yesavage

, Brink

, Rose

, Lum

, Huang

, Adey

, Leirer

(1982) Development and validation of a geriatric depression screening scale: A preliminary report. J Psychiatr Res 17, 37–49.

45.

Aguado

, Martínez

, Onís

(2000) Adaptation and validation in Spanish of the abbreviated version of the “Geriatric Depression Scale” (GDS) of Yesavage (in Spanish). Aten Primaria 26, 328.

46.

Folstein

, Folstein

, McHugh

(1975) Mini-mental state. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–98.

47.

Lobo

, Saz

, Marcos

(2002) MMSE: Mini-Mental State Exam (in Spanish). TEA Ediciones.

48.

Pfeffer

, Kurosaki

, Harrah

, Chance

, Filos

(1982) Measurement of functional activities in older adults in the community. J Gerontol 37, 323–329.

49.

Serrano

, Allegri

, Drake

, Butman

, Harris

, Nagle

, Ranalli

(2001) A shortened form of the Spanish Boston naming test: A useful tool for the diagnosis of Alzheimer’s disease. Rev Neurol 33, 624–627.

50.

Wechsler

(1997) Wechsler Memory Scale, 3rd ed. Psychological Corporation.

51.

Wechsler

(1997) Wechsler Adult Intelligence Scale, 3rd ed. Psychological Corporation.

52.

Bell

, Sejnowski

(1995) An information-maximization approach to blind separation and blind deconvolution. Neural Comput 7, 1129–1159.

53.

Calhoun

, Adali

, Pearlson

, Pekar

(2001) A method for making group inferences from functional MRI data using independent component analysis. Hum Brain Mapp 14, 140–151.

54.

Y-O

, Adali

, Calhoun

(2007) Estimating the number of independent components for functional magnetic resonance imaging data. Hum Brain Mapp 28, 1251–1266.

55.

Kim

, Manoach

, Mathalon

, Turner

, Mannell

, Brown

, Ford

, Gollub

, White

, Wible

, Belger

, Bockholt

, Clark

, Lauriello

, O’Leary

, Mueller

, Lim

, Andreasen

, Potkin

, Calhoun

(2009) Dysregulation of working memory and default-mode networks in schizophrenia using independent component analysis, an fBIRN and MCIC study. Hum Brain Mapp 30, 3795–3811.

56.

Costumero

, Rodríguez-Pujadas

, Fuentes-Claramonte

, Ávila

(2015) How bilingualism shapes the functional architecture of the brain: A study on executive control in early bilinguals and monolinguals. Hum Brain Mapp 36, 5101–5112.

57.

Ashburner

(2007) A fast diffeomorphic image registration algorithm. Neuroimage 38, 95–113.

58.

Baxter

(2009) Involvement of medial temporal lobe structures in memory and perception. Neuron 61, 667–677.

59.

Squire

, Stark

CEL

, Clark

(2004) The medial temporal lobe. Annu Rev Neurosci 27, 279–306.

60.

Lustig

, Buckner

(2004) Preserved neural correlates of priming in old age and dementia. Neuron 42, 865–875.

61.

Visser

, Verhey

FRJ

, Hofman

PAM

, Scheltens

, Jolles

(2002) Medial temporal lobe atrophy predicts Alzheimer’s disease in patients with minor cognitive impairment. J Neurol Neurosurg Psychiatry 72, 491–497.

62.

DeCarli

, Frisoni

, Clark

, Harvey

, Grundman

, Petersen

, Thal

, Jin

, Jack

, Scheltens

, Alzheimer’s Disease Cooperative Study Group (2007) Qualitative estimates of medial temporal atrophy as a predictor of progression from mild cognitive impairment to dementia. Arch Neurol 64, 108–115.

63.

Rusinek

, Endo

, De Santi

, Frid

, Tsui

W-H

, Segal

, Convit

, de Leon

(2004) Atrophy rate in medial temporal lobe during progression of Alzheimer disease. Neurology 63, 2354–2359.

64.

Clerx

, van Rossum

, Burns

, Knol

, Scheltens

, Verhey

, Aalten

, Lapuerta

, van de Pol

, van Schijndel

, de Jong

, Barkhof

, Wolz

, Rueckert

, Bocchetta

, Tsolaki

, Nobili

, Wahlund

L-O

, Minthon

, Frölich

, Hampel

, Soininen

, Visser

(2013) Measurements of medial temporal lobe atrophy for prediction of Alzheimer’s disease in subjects with mild cognitive impairment. Neurobiol Aging 34, 2003–2013.

65.

Sperling

, Dickerson

, Pihlajamaki

, Vannini

, LaViolette

, Vitolo

, Hedden

(2011) Functional alterations in Memory Networks in Early AD. Neuromolecular Med 12, 27–43.

66.

Dickerson

, Sperling

(2009) Large-scale functional brain network abnormalities in Alzheimer’s disease: Insights from functional neuroimaging. Behav Neurol 21, 63–75.

67.

Terry

, Sabatinelli

, Puente

, Lazar

, Miller

(2015) A meta-analysis of fMRI activation differences during episodic memory in Alzheimer’s disease and mild cognitive impairment. J Neuroimaging 25, 849–860.

68.

Dickerson

, Salat

, Bates

, Atiya

, Killiany

, Greve

, Dale

, Stern

, Blacker

, Albert

, Sperling

(2004) Medial temporal lobe function and structure in mild cognitive impairment. Ann Neurol 56, 27–35.

69.

Dickerson

, Sperling

(2008) Functional abnormalities of the medial temporal lobe memory system in mild cognitive impairment and Alzheimer’s disease: Insights from functional MRI studies. Neuropsychologia 46, 1624–1635.

70.

Anticevic

, Repovs

, Shulman

, Barch

(2010) When less is more: TPJ and default network deactivation during encoding predicts working memory performance. Neuroimage 49, 2638–2648.

71.

Daselaar

, Prince

, Cabeza

(2004) When less means more: Deactivations during encoding that predict subsequent memory. Neuroimage 23, 921–927.

72.

White

, Jansen

, Doege

, Mullinger

, Park

, Liddle

, Gowland

, Francis

, Bowtell

, Liddle

(2013) Theta power during encoding predicts subsequent-memory performance and default mode network deactivation. Hum Brain Mapp 34, 2929–2943.