Inflection Point in Course of Mild Cognitive Impairment: Increased Functional Connectivity of Default Mode Network

Abstract

The alteration of the default mode network (DMN) functional connectivity (FC) has been reported in patients with amnestic mild cognitive impairment (aMCI) as a predictor of Alzheimer’s disease (AD). However, no studies exist that examined stage-dependent DMN FC changes throughout the course of aMCI. The present study aims to characterize patterns of DMN FC over three aMCI stages as first defined. Utilizing the extreme groups approach on the performance of memory tasks, aMCI subjects were divided into mild, moderate, and severe stages. Independent component analysis was used to assess DMN for individual patients in each of the three cross-sectionally defined stages. Instead of finding that continued monotonic decline was the case for the hippocampus volume, which we also investigated in this study, we observed an increase in DMN functional connectivity from mild aMCI to moderate aMCI and a decrease to severe aMCI, mainly in the left precuneus and superior parietal lobe. Moreover, the FC was significantly associated with cognitive performance. Though a longitudinal study is needed to confirm these results, our cross-sectional finding is that non-linear FC changes in DMN could be a characteristic of prodromal early disease development.

Keywords

Alzheimer’s disease amnestic mild cognitive impairment default mode network functional connectivity non-monotomic trajectory

INTRODUCTION

Alzheimer’s disease (AD) is a age-related neurodegenerative disease [1, 2]. It has tremendous implications for an already overburdened healthcare system [3]. For development and evaluation of possible early effective interventions that are not yet available, it is critical to identify early biomarkers that are strongly predictive of future progression of AD.

Amnestic mild cognitive impairment (aMCI) usually represents a transitional phase between normal cognitive function and AD [4]. Researchers proposed a hypothetical model of various neuroimaging biomarkers during the progression of MCI to AD [1, 5]. As the early pathological amyloid-β (Aβ) accumulation starts prior to AD, the volume of the hippocampus [6, 7] and metabolism [8, 9] decline over the progression. Interestingly and in contrast to the monotonic changes, there is first increased brain functional activation during cognitive tasks, followed by reduced activation in the temporo-parietal and frontal regions over the course of MCI [10, 11]. Such a brain functional inflection point of “increase followed by decrease” could serve as a marker of the onset of abnormal cognitive behavior.

In addition to the task functional magnetic resonance imaging (fMRI) techniques, resting-state (rs)-fMRI has showcased its potential for the examination of brain network. Default mode network (DMN) has been gaining attention as a potential noninvasive biomarker for AD. The DMN regions are comprised of the typical predilection sites of AD, such as pathological Aβ accumulation and gray matter (GM) atrophy [12]. AD is considered as a disconnection syndrome [13]. The functional connectivity (FC) in DMN was shown to distinguish healthy aging from AD [14] and distinguish the converted MCI from stable MCI [15]. The abnormality in DMN connectivity might be an early and suitable neuroimaging marker of the initial neurodegeneration.

Though the DMN FC appeared to show a downward trend throughout the MCI stage in Ewer’s Model, mixed findings were reported in different studies. Some studies demonstrated a loss of FC in the DMN in MCI [16, 17]. Other studies, on the other hand, reported increased connectivity in the anterior DMN, which contributed to memory deficits in MCI [18]. Such connectivity increase was interpreted as a way by which the brain reacts to and potentially adapt, in a detrimental way [18, 19]. We speculate that both the revealed consistencies and discrepancies in DMN FC changes in MCI were possibly due to the severity of variation among MCI patients.

AMCI severity led to the difference in incidence of AD and neuroimaging characteristics. In the well-known ADNI study, MCI were divided into late MCI (LMCI) and early MCI (EMCI) based on memory abilities [20]. The risk of AD and cortical thinning was greater in LMCI than EMCI [21, 22]. Compared with controls, less-impaired MCI showed functional hyperactivation, whereas more-impaired MCI demonstrated hypoactivation [23]. However, most previous studies utilized two MCI severity stages only and thus could not investigate, within MCI, the inflection point of “increase then decrease”. To better characterize the inflection point of the FC, a finer degree of MCI severity is needed.

In contrast to the possible “first increase then decrease” changes of functional characteristics, previous studies constantly reported that the GM volume declined throughout the progression of MCI, predominantly in the hippocampus [1 , 24]. The atrophy trend of brain structure is similar, with increased brain levels of Aβ and decreased cerebrospinal fluid Aβ levels [6, 25].

In this study, our primary aim is to confirm the existence of the DMN FC inflection point over aMCI severity using a finer staging for aMCI based on episodic memory measures. This finer severity definition allows us to better characterize the FC in DMN measured by rs-fMRI technique and GM volume measured by MRI technique. We hypothesized that there is a non-monotomic trajectory of FC and the monotonic changes of brain structure over the multistage grading of aMCI.

MATERIALS AND METHODS

Participants

All participants in the present study were from the Beijing Aging Brain Rejuvenation Initiative (BABRI), which aimed to investigate the aging and cognitive conditions of urban elderly adults in Beijing [26]. In this study, MCI participants were selected according to the following criteria: 1) age between 50 to 80 years old; 2) six or more years education; 3) met the published criteria of amnestic MCI, scored more than 1.5 standard deviation (SD) below the age- and education-adjusted norms on at least two of the score [27 –32] of long-delay recall (trial 5) and total word recall (trial 1–5) of the Rey Auditory Verbal Learning Test (R-AVLT) [33] and delay recall of the Rey-Osterrieth Complex Figure test (ROCF) [34]; 4) the general cognitive function relatively preserved, and had no problems in daily life (scoring 0 on the Activites of Daily Living scale); 5) no history of disease, such as neurologic, psychiatric, or systemic illnesses, that could influence cerebral function; and 6) no history of taking psychoactive medications.

All study procedures and ethical aspects of this research were approved by the Institutional Review Board of the Beijing Normal University Imaging Center for Brain Research. All participants agreed and signed informed consent before participating in the study.

Defining aMCI severity

The episodic memory tests used for the diagnosis of aMCI included long-delay recall (trial 5) and total word recall (trial 1–5) of the R-AVLT and delay recall of the ROCF. In order to define aMCI severity, we transformed these three tests scores into a 0∼10 range using X’ = ([X-X_min]/[X_max-X_min])*10; the minimum and maximum were theoretical values here. We then averaged the three tests as the score of episodic memory in this study. Thus, the score of episodic memory of aMCI ranges from 0∼10.

Utilizing the extreme groups approach [35], we divided the aMCI patients into different severity stages based on the score of episodic memory defined above. Basically, we ranked the episodic memory scores of 124 aMCI patients in the BABRI Database in descending order and defined the highest 27% people as mild aMCI, the lowest 27% people as severe aMCI, and those in-between as moderate aMCI [35]. Efforts were made to match the demographic characteristics, such as gender, age, and year of education, among the newly defined mild, moderate, and severe aMCI sub-groups after excluding subjects who were refused to join or unsuitable for MRI, had significant head motion or were not able to finish the scan. A total of 45 patients who finished both structure MRI and functional MRI were included in this study, including 10 mild aMCI, 19 moderate aMCI, and 16 severe aMCI. None of the subjects reported a history of head trauma, psychiatric or neurological disorders or alcohol or drug abuse.

Neuropsychological assessment

All subjects underwent an extensive neuropsychological assessment, including the following tests. The general cognitive function measured by the Mini-Mental State Examination (MMSE). The episodic memory was assessed with the R-AVLT and the ROCF, as well digit span test assessing working memory. Language ability was assessed with the Boston Naming Test and Category Verbal Fluency Test (CVFT). Executive function was assessed with part B of the Trail Making Test (TMT) and part C of the Stroop Test; Spatial processing was assessed with the ROCF-copy and the Clock-Drawing Test. Attention tests included the Symbol Digit Modalities Test, part A of the TMT, and the Similarities subtest of the Wechsler Adult Intelligence Scale-Chinese revision.

MRI data acquisition

MRI data were acquired using a SIEMENS TRIO 3T scanner in the Imaging Center for Brain Research, Beijing Normal University and included high-resolution T1-weighted structure images and rsfMRI scan. Participants laid supine with their head snugly fixed by straps and foam pads to minimize head movement. Participants were asked to keep still with their eyes closed in the scanner, not to think about anything particular, and not to fall asleep. T1-weighted, sagittal 3D magnetization prepared rapid gradient echo (MP-RAGE) sequences were acquired and covered the entire brain [176 sagittal slices, repetition time (TR) = 1900 ms, echo time (TE) = 3.44 ms, slice thickness = 1 mm, flip angle = 9°, inversion time = 900 ms, field of view (FOV) = 256×256 mm², acquisition matrix = 256×256]. Resting state data were collected using a gradient echo EPI sequence [33 axial slices, TR = 2000 ms, TE = 30 ms, slice thickness = 3.5 mm, flip angle = 90°, FOV = 200×200 mm², matrix = 64×64].

Structural image analysis

The Matlab2012b and VBM8-toolbox (http://dbm.neuro.uni-jena.de/vbm) were used to preprocess the structural images. These images were segmented and spatially normalized into GM, white matter (WM), and cerebrospinal fluid in standard East Asian brains Montreal Neurological Institute (MNI) space using a unified segmentation algorithm. GM and WM segments were modulated only by nonlinear components to preserve regional GM and WM volume (modulated GM and WM volumes). All of the resultant images were resampled with 3-mm isotropic voxels and smoothed with a 6-mm full width and half maximum (FWHM) Gaussian kernel.

We used SPM8 software (http://www.fil.ion.ucl.ac.uk) to inspect the brains’ structural differences and segmented GM maps after preprocessing was entered into group analysis using the general linear model, as full-factorial analyses with age, gender, and years of education as covariates of no interest. A whole brain GM template was used as mask. We used p < 0.001 (uncorrected for multiple comparisons) with an extent threshold of 30 voxels as the threshold of significance for these analyses.

To more precisely describe the hippocampus variation among the three aMCI severity sub-groups, hippocampus segmentation was manually performed slice by slice using the MRIcron package and its subprogram dcm2niigui software (http://www.mccauslandcenter.sc.edu/mricro/mricron/dcm2nii.html) by two experienced raters who were well trained and knew the anatomy well, especially in terms of hippocampus volume for hippocampus manual segmentation (The alpha consistency coefficient of left hippocampus, right hippocampus, and total hippocampus volume are 0.936, 0.895, and 0.930, respectively, which means the manual segmentation of these two person were highly consistent). This manual outlining is done in sequential coronal T1-weighted MR images. For each subject, both the right and left hippocampi were segmented. The segmentation was performed with aMCI stage information blinded.

Functional image analysis

Matlab2012b, SPM8 (http://www.fil.ion.ucl.ac.uk/spm) and DPARSF (http://www.restfmri.net/forum/taxonomy/term/36) were used to preprocess the RS-fMRI data. To allow participants to adapt to the magnetic field, we discarded the first 10 volumes. Then, the remaining data were preprocessed by slice timing, within-subject interscan realignment to correct possible movement, spatial normalization to a standard brain template in the MNI coordinate space, resampling to 3×3×3 mm³ and smoothing with a 6-mm full-width half-maximum Gaussian kernel. In addition, they were processed with linear detrending and filtered by a phase-insensitive band-pass filter (0.01-0.08 Hz) to reduce the effect of low frequency drift and high frequency physiological noise. As nuisance covariates, we also included the six motion parameters, the global mean signal, the WM signal and the cerebrospinal fluid signal.

The group ICA toolbox (GIFT version 2.0e; http://mialab.mrn.org/software/gift/) was used for Group Independent Component Analysis (ICA). We estimated each group with 25 independent components separately. There are three main stages of GIFT procedure: 1) data reduction, 2) application of the FastICA algorithm, and 3) back-reconstruction for each individual subject. Then, the best-fit components for the DMN were identified by visual inspection.

We also extracted the FC though defining region of interest (ROI) upon the results of functional images’ group compare. Two important peaks of the DMN group difference were selected to define the center of a spherical ROI with a 6-mm radius (Left Superior Parietal: x = –15, y = –60, z = 54; Left Precuneus: x = –6, y = –60, z = 45). We used FC in these ROIs for our correlation analysis later.

Goodness-of-fit analysis

A goodness-of-fit (GOF) index reflecting the degree of DMN maps of each subject [36] was calculated by comparing to the DMN template created by all of the participants in this study. To accomplish this, the DMN template was first created in SPM8 using a one-sample t test of all of the aMCI subjects’ default mode component maps [p (False Discovery Rate (FDR)-corrected for multiple comparisons) <0.05 and cluster size >100]. The GOF was calculated in MATLAB as GOF = mean(component[mask]) –mean(component[∼mask]), or the mean z score of all voxels within the DMN mask minus the mean z score of all voxels outside the mask (among all in-brain voxels), as described previously [14, 37].

Statistical analyses

ANCOVA with group as the factor and age, gender, and formal education years as covariates was conducted to identify the GOF, the FC [p (FDR) <0.05] and the regional GM [p < 0.001 (uncorrected for multiple comparisons), cluster size >30] differences among the mild, moderate, and severe aMCI groups. To determine the pattern of hippocampus volume difference among these three groups, we also performed ANCOVA with age, gender, formal education years, and total brain volume as covariates. The Least Significant Difference (LSD) was used for all the post hoc analysis. Partial correlation analysis was performed to examine the relationship between the FC in two ROIs and the neuropsychological performances with gender, age, and years of education as covariates. At last, we did multiple regression analysis with the neuropsychological performances as dependent variables and the age, gender, years of education, value of GOF, FC of left superior parietal, FC of left precuneus, total brain volume, volume of left hippocampus, volume of right hippocampus as independent variables. SPSS 20.0 and SPM8 were used for complete these tests.

RESULT

Demographic information and neuropsychological characterizations

There are no differences in gender, age, or years of education among the three groups. They, however, showed significant linear trends in terms of decreased MMSE scores (F = 6.248, p = 0.004) in the order of mild aMCI, moderate aMCI, and severe aMCI. After controlling for the effects of gender, age, and years of education, only AVLT delay recall (F = 14.412, p < 0.001), AVLT total (F = 9.112, p = 0.001), and ROCFT recall (F = 11.363, p < 0.001) showed significant aMCI severity dependent difference in the same group order. Not surprisingly, the episodic memory used to define the aMCI severity (the score in Table 1 is the average Z score of AVLT delay recall, AVLT total and ROCFT recall plus 1) was significant among the three groups (F = 109.05, p < 0.001). All of the other neuropsychological measures showed no statistically significant group differences (Table 1).

Table 1

Demographics and Neuropsychological Characterizations of aMCI

Characteristic (mean±SD)	Total	Group			×²/F	p
		Mild aMCI	Moderate aMCI	Severe aMCI
Number	45	10	19	16
Gender (m/f)	17/28	4/6	5/14	8/8	2.10	0.350
Age (y)	64.28±6.65	61.60±7.53	64.94±6.30	66.06±8.49	1.15	0.326
Education (y)	11.49±2.74	10.30±2.98	11.47±2.27	12.06±3.28	1.21	0.309
Zmemory	0.53±0.41	1.066±0.134	0.610±0.176	0.089±0.193	109.05	<0.001^a,b,c
MMSE	25.80±1.83	27.60±1.07	25.47±1.65	25.06±1.73	7.57	0.002^a,c
Memory
AVLT delay recall	1.64±1.30	2.50±1.35	2.16±0.96	0.50±0.73	14.41	<0.001^b,c
AVLT Total	16.56±5.61	19.90±6.82	18.37±4.22	12.31±3.50	9.11	0.001^b,c
ROCFT recall	8.22±6.47	14.60±8.25	7.42±4.96	5.19±3.82	11.36	<0.001^a,c
Digit Span total	9.73±2.14	9.60±1.51	9.53±2.22	10.06±2.43	0.22	0.805
Digit Span forward	6.27±1.70	5.80±1.23	6.32±1.86	6.50±1.79	0.78	0.465
Digit Span backward	3.47±1.31	3.80±1.03	3.21±1.36	3.56±1.41	0.50	0.609
Language
CVFT	38.00±6.60	41.00±6.09	36.47±6.88	37.94±6.31	1.92	0.160
BNT	21.78±3.96	22.70±4.52	21.21±3.38	21.88±4.36	1.33	0.277
Executive Function
TMT b-a	150.69±62.66	123.50±50.54	171.16±66.49	143.38±59.98	1.63	0.209
SCWT-SIE	45.07±27.09	41.20±19.85	46.26±31.70	46.13±26.37	0.10	0.907
Spatial Processing
ROCFT copy	31.71±4.25	31.60±5.60	31.89±3.70	31.56±4.19	0.40	0.674
CDT	23.58±3.99	22.50±4.50	24.26±3.40	23.44±4.38	0.60	0.552
Attention
SDMT	28.93±10.75	33.20±11.82	29.63±10.98	25.20±9.03	2.53	0.093
TMT^a	68.49±29.29	64.50±30.05	60.00±19.34	81.06±35.42	2.80	0.074
Similarity	13.20±4.81	13.70±4.64	12.95±4.55	13.19±5.46	0.17	0.847

aMCI, amnestic mild cognitive impairment; MMSE, Mini-Mental State Examination; AVLT, Auditory Verbal Learning Test; ROCF, Rey-Osterrieth Complex Figure; CVFT, Category Verbal Fluency Test; BNT, Boston Naming Test; TMT, Trail Making Test; SCWT-SIE, Stroop interference effect of Stroop Color and Word Test; CDT, Clock-Drawing Test; SDMT, Symbol Digit Modalities Test. The difference of Zmemeory, MMSE and all the neuropsychological were adjusted with gender, age and years of education. Post hoc pair-wise comparisons were t performed using t tests. p < 0.05 was considered significant. ^a Post hoc paired comparisons showed significant group differences between mild aMCI and moderate aMCI, after least-significant difference (LSD) correction. ^b Post hoc paired comparisons showed significant group differences between moderate aMCI and severe aMCI, after LSD correction. ^c Post hoc paired comparisons showed significant group differences between mild aMCI and severe aMCI, after LSD correction.

Default mode network alteration in aMCI

The best-fit component for DMN of the three groups was obtained by group ICA (Fig. 1). Maps of DMN connectivity were compared with a template DMN map constructed from all participants to obtain GOF indices of DMN expression. The moderate aMCI group has the highest GOF indices, significantly higher than mild aMCI (p = 0.007), and there is a marginally significant difference between moderate aMCI and severe aMCI (p = 0.088) (Fig. 1d). Additionally, a trend analysis shows that there was a significant quadratic trend (p = 0.008) of GOF index along with the disease progression.

Fig.1

The functional connectivity (FC) in DMN of each group and difference between groups. a, b, c represent the default mode network of mild aMCI, moderate aMCI, and severe aMCI, respectively, q < 0.05 FDR corrected, cluster size >100; d, the group difference of GOF, the group means were significantly different between mild aMCI and moderate aMCI (p < 0.01). The right side (black) shows FC difference between mild aMCI and moderate aMCI, and between moderate aMCI and severe aMCI (cluster size >10, q < 0.05, FDR corrected). Mi., Mild aMCI; Mo., Moderate aMCI; Se., Severe aMCI; SPG, Superior parietal gyrus; PCUN, precuneus.

There were significant differences in FC of DMN among the mild aMCI patients, the moderate aMCI patients and the severe aMCI patients. These regions included the left superior parietal and left precuneus. Post-hoc pair-wise comparisons of FC in DMN are shown in Fig. 1B, C and Table 2. Compared with the mild aMCI patients, the moderate aMCI patients displayed significantly higher FC in the bilateral precuneus, middle occipital, left superior parietal, right post central, right angular, left medial superior frontal, and left orbital inferior frontal. Compared with the moderate aMCI patients, the severe aMCI patients displayed significantly lower FC in the left precuneus, left superior parietal and right orbital middle frontal [p (FDR-corrected for multiple comparisons) <0.05]. The left precuneus and left superior parietal showed significant differences among the three groups. This suggests these two regions may be very important in aMCI’s pathology progressing. The FC extracted by the method of region of interest of these regions (Left Superior Parietal: x = –15, y = –60, z = 54; Left Precuneus: x = –6, y = –60, z = 45) in moderate aMCI patients were significantly higher than both mild aMCI and severe aMCI. FC of left precuneus in moderate aMCI are significantly higher than both mild aMCI (p < 0.001) and severe aMCI (p < 0.01), and severe aMCI is higher than mild aMCI (p < 0.05). FC of left superior parietal in moderate aMCI is significantly higher than mild aMCI (p < 0.001) and severe aMCI (p < 0.001) (Fig. 1d).

Table 2

The difference of functional connectivity in DMN

Region (aal)	Cluster	x	y	z	T value
	sizes
Mild aMCI <Moderate aMCI
Right Precuneus	124	9	–63	48	–6.33
Left Superior Parietal		–15	–60	54	–4.82
Left Precuneus		–6	–60	45	–5.12
Right Postcentral	59	51	–18	42	–6.46
Left Middle Occipital	92	–39	–81	18	–6.41
Right Middle Occipital	83	45	–75	21	–5.72
Right Angular		45	–75	39	–4.77
Left Medial Superior Frontal	11	0	54	18	–5.10
Left Orbital Inferior Frontal	25	–42	27	–18	–4.50
Moderate aMCI >Severe aMCI
Left Superior Parietal	21	–18	–57	54	6.38
Left Precuneus	36	–12	–63	66	5.71
Right Orbital Middle Frontal	14	24	54	–18	5.69

The x, y, z coordinates is the primary peak locations in the MNI space. Cluster size >10, q < 0.05, FDR corrected.

Correlation between FC of DMN and cognitive performance

We tested whether variability in DMN functional connectivity predicted cognitive performance. We examined the relationship between cognitive performance and FC in regions where the group differences were observed. Two important peaks of the DMN group difference were selected to define the center of a spherical region of interest with a 6-mm radius (Left Superior Parietal: x = –15, y = –60, z = 54; Left Precuneus: x = –6, y = –60, z = 45). The results showed that only the FC of the left precuneus is significantly correlated with backward (r = –0.338, p = 0.029) and total digit span (r = –0.328, p = 0.034) (Fig. 2), but, nevertheless, this p value cannot survival with the FDR correction. No significant correlation was found between FC of the left superior parietal in DMN and cognitive performance.

Fig.2

The correlation between FC and cognition behavior. Significant negative correlations were observed between FC of the left precuneus and digit span backward (A), and between FC of the left precuneus and digit span total (B).

Gray matter volume in aMCI patients

Compared to mild aMCI patients, the moderate aMCI patients showed significantly lower GM volume in the right precentral, right postcentral, right insula, left anterior cingulum, right calcarine, and right orbital medial frontal. Compared to moderate aMCI patients, the severe aMCI patients showed significantly decreased GM volume in the left superior occipital, left lingual, part of vermis and bilateral hippocampus (Table 3 and Fig. 3).

Fig.3

Gray matter difference between groups. Moderate aMCI had significant gray matter atrophy compared with mild aMCI at the right precentral, right postcentral, right insula, left anterior cingulum, right calcarine and right orbital medial frontal; Severe aMCI had significant gray matter atrophy compared with moderate aMCI at the left superior occipital, left lingual, part of vermis and bilateral hippocampus. Cluster size >30, p < 0.001.

Table 3

Brain areas with significant GM atrophy

Region (aal)	Cluster	x	y	z	T value
	level
Mild aMCI >Moderate aMCI
Right Precentral	440	53	–12	43	4.38
Right Postcentral		59	–18	45	4.20
Right Insula	87	39	26	–6	3.71
Left Anterior Cingulum	85	0	18	30	3.66
Right Calcarine	34	11	–72	15	3.56
Right Orbital Medial Frontal	30	8	45	–11	3.41
Moderate aMCI >Severe aMCI
Left Superior Occipital	134	–15	–88	31	3.89
Left Hippocampus	78	–29	–9	–14	3.79
Left Lingual	46	–21	–75	–3	3.75
Right Hippocampus	56	36	–12	–18	3.74
Vermis_4_5	33	5	–58	3	3.49

The x, y, z coordinates is the primary peak locations in the MNI space. Cluster size >30, p < 0.001.

Independent of the VBM analysis, we especially examined the hippocampus volumes that were carefully delimitated using the procedure described in the methods section. After adjusting for gender, age, years of education, and global volume, there were significant main effects among the three groups in left hippocampus volumes (F = 3.63, p = 0.037) and total hippocampus volumes (F = 3.14, p = 0.050). Post hoc tests found that the left hippocampus volumes were significantly lower in the severe aMCI compared to moderate aMCI (severe aMCI: 2051.75±492.24, moderate aMCI: 2572.12±458.43, p = 0.012), and the total hippocampus volumes were also significantly lower in the severe aMCI compared to moderate aMCI (severe aMCI: 4200.50±829.19, moderate aMCI: 5090.71±884.92, p = 0.023). Furthermore, using trend analysis in the general linear model, controlling for gender, age, and year of education, we found that left (p = 0.030), right (p = 0.029), and total (p = 0.024) hippocampus volumes had significant linear trend over the aMCI severity as mild >moderate>severe.

Influences of GM volume, FC in DMN, and GOF on cognitive performance

To evaluate the relationship between function connectivity, GM volume, and cognitive performance in the aMCI patients, we performed multiple regression analyses. The results show that FC in the left superior parietal lobe is positively correlated with AVLT delay recall (B = 0.42, p = 0.014), AVLT total (B = 0.39, p = 0.015). FC in the left precuneus is negatively correlated with MMSE (B = –0.56, p = 0.014). The global brain volume is positively correlated with MMSE (B = 0.56, P = 0.013). Right hippocampus volume is correlated with CVFT (B = 0.91, p = 0.025) positively, and there is a negative correlation between GOF and AVLT total (B = –0.49, p = 0.025) (Table 4).

Table 4

Regression with functional connectivity, gray matter volume, and neuropsychological characteristics

Dependent variable	MMSE		AVLT delay recall		AVLT total		CVFT
	Estimate	p-value	Estimate	p-value	Estimate	p-value	Estimate	p-value
FC of left superior Parietal	0.22	0.236	0.42	0.014	0.39	0.015	–0.08	0.682
FC of left Precuneus	–0.56	0.014	–0.30	0.129	–0.12	0.502	0.14	0.566
GOF	0.32	0.209	–0.16	0.475	–0.49	0.025	–0.08	0.765
Volume of left hippocampus	–0.07	0.872	0.63	0.090	0.60	0.087	–0.78	0.085
Volume of right hippocampus	0.16	0.667	–0.21	0.518	–0.17	0.568	0.91	0.025
Total brain volume	0.56	0.013	–0.08	0.690	–0.18	0.330	0.11	0.630
Education	0.01	0.955	0.06	0.734	0.13	0.432	–0.23	0.276
Age	0.04	0.837	–0.34	0.052	–0.58	0.001	0.29	0.177
Gender	0.38	0.094	0.20	0.334	0.07	0.710	0.12	0.605
Constant	5.77	0.487	4.73	0.246	54.35	0.002	11.68	0.623

MMSE, Mini-Mental State Examination; AVLT, Auditory Verbal Learning Test; CVFT, Category Verbal Fluency Test.

DISCUSSION

In this study, we used ICA to report DMN FC differences among three aMCI severity levels we defined to examine the possible existence of a nonlinear/non-monotonic trend. Moderate aMCI demonstrated evidence of the highest functional connectivity compared with mild aMCI and severe aMCI in default mode networks. Moreover, the FC was significantly associated with cognitive performance. However, the volume of the hippocampus continues to decline throughout the continuum of aMCI. These results strongly suggest that distributed DMN are altered in a non-linear fashion as aMCI becomes more severe in contrast to the trending hippocampus volume variation. The inflection point of FC in DMN could be very important in predicting MCI to AD.

Researchers have recently begun to focus on the stage of aMCI course. A dichotomy of early and late MCI was put forward by ADNI investigators and others [21]. They found that the risk of AD dementia was greater in late-MCI than early-MCI [21]. Celone et al. classified the MCI patients as less and more impaired MCI. They found a remarkable curve in the pattern of deactivation in medial and lateral parietal regions with greater deactivation in less-impaired MCI and loss of deactivation in more impaired MCI [23]. Through the three stages of aMCI, our current study also demonstrated a nonlinear trajectory of functional connectivity changes and a linear trajectory of structural atrophy.

The DMN is particularly relevant for aging and dementia because the regions of DMN are vulnerable to atrophy and, deposition of the amyloid protein, and they generally show a reduced glucose metabolism [38]. The disruption of functional connectivity is also one of the important features of AD pathology. MCI converters showed more severely decreased functional connectivity compared to non-converters [37]. The brain’s functional activations in DMN were reported to be both increased and decreased in MCI patients [19, 39]. Previous studies also found a variation of FC in DMN due to the heterogeneity of MCI [40, 41]. Our results provided more detailed information for the non-linear pattern of FC in DMN over MCI severity levels.

The two major brain regions where the group differences existed were the left precuneus and superior parietal lobe. Past studies have shown that the precuneus is involved with episodic memory and visuospatial processing [42]. It has been suggested to be the ‘core node’ of the default mode network. In preclinical and early AD [43 –45], the precuneus exhibited high levels of amyloid deposition and may be selectively vulnerable to structural and functional alterations [46]. There is reduced functional activation [47] and connectivity [48] of the precuneus in aMCI, which is a syndrome that leads to high risk of developing AD. The superior parietal lobule is involved in the spatial orientation [49] and manipulation of information in working memory [50].

An important question naturally relates to what these non-linear changes represent. Historically, a higher DMN FC in older brains has been interpreted as representing compensatory processes [51, 52]; that is, DMN increases FC levels to compensate for failing activity in other region(s), allowing for the maintenance of cognitive function. Other explanations of increasing activity in AD and its prodromal stage have also been proposed [53, 54]. It is also possible that this increased FC is the result of additional conscious efforts or affective valence from mild to moderately impaired MCI who may be particularly concerned about their memory [23]. From moderate to severe aMCI patients, the decreased FC may reflects the exhaustion of the cognitive reserve and of cognitive strategies used to compensate the monotonic decline of memory and other cognitive functions in moderate aMCI patients [55, 56].

The Ewers’ model indicated that both hippocampus volumes and FC of DMN presented declined trends throughout the preclinical, MCI and AD stages [1]. However, these findings of FC decline were inferred from studies about elderly cognitively normal subjects with amyloid deposition [10, 17] and limited to MCI patients without MCI severity variation considered [48]. The heterogeneity of MCI could induce the variability in MCI studies. Previous studies have also reported the heterogeneity of brain function activation [57, 58] and functional connectivity [40, 41] in MCI patients. Our study first graded the aMCI into three groups according to memory performance and added additional details for Ewer’s hypothetical model (Fig. 4).

Fig.4

The modified Ewer’s Hypothetical model. A) Hypothetical model of various neuroimaging biomarkers and their predicted utility during disease progression [11]. B) Patterns of DMN FC and hippocampus volume among over 3 aMCI multistage: DMN FC increased from mild aMCI to the moderate aMCI and then decreased to severe aMCI (blue line). In contrast, these three aMCI groups showed progressive decline in hippocampus volumes, which consist with Ewers’ model (green line).

As our study sample is relatively small, the results in this study may need further verification in a large sample. Additionally, a longitudinal study should also be performed to overcome the limitation of a cross-sectional study.

Conclusions

Our study first characterized the functional connectivity variation in DMN along with the disease severity as defined based on the three stages among community-based aMCI patients. Though these three aMCI groups showed progressive decline in hippocampus volumes, changes in functional connectivity in DMN that first increased and then decreased were also apparent. The changed functional connectivity and hippocampus volumes were significantly correlated with cognitive performance in multistage grading of aMCI. In the face of accumulating neuropathology and hippocampus atrophy, compensatory brain functional connectivity supports the maintenance of cognitive function and impairment begin after the early compensatory mechanisms ultimately failed. The non-linear trajectory of FC could be an important neuroimaging marker for the prediction of aMCI progression.

Footnotes

ACKNOWLEDGMENTS

This work was supported by National Science Fund for Distinguished Young Scholars (grant number 81625025), State Key Program of National Natural Science of China (grant number 81430100), Beijing Municipal Science & Technology Commission (grant number Z161100000216135), the Natural Science Foundation of China (grant number 31500922 and 31571129) and the Fundamental Research Funds for the Central Universities (grant number 2017XTCX04).

Authors’ disclosures available online ().

References

Ewers

, Sperling

, Klunk

, Weiner

, Hampel

(2011) Neuroimaging markers for the prediction and early diagnosis of Alzheimer’s disease dementia. Trends Neurosci 34, 430–442.

Petersen

, Stevens

, Ganguli

, Tangalos

, Cummings

, DeKosky

(2001) Practice parameter: Early detection of dementia: Mild cognitive impairment (an evidence-based review) Report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology 56, 1133–1142.

Batsch

, Mittelman

(2012) World Alzheimer Report 2012: Overcoming the stigma of dementia. Alzheimer’s Disease International.

Petersen

, Stevens

, Ganguli

, Tangalos

, Cummings

, DeKosky

(2001) Practice parameter: Early detection of dementia: Mild cognitive impairment (an evidence-based review). Report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology 56, 1133–1142.

Jack

, Knopman

, Jagust

, Shaw

, Aisen

, Weiner

, Petersen

, Trojanowski

(2010) Hypothetical model of dynamic biomarkers of the Alzheimer’s pathological cascade. Lancet Neurol 9, 119–128.

Mormino

, Kluth

, Madison

, Rabinovici

, Baker

, Miller

, Koeppe

, Mathis

, Weiner

, Jagust

(2009) Episodic memory loss is related to hippocampal-mediated β-amyloid deposition in elderly subjects. Brain 132, 1310–1323.

Schuff

, Tosun

, Insel

, Chiang

, Truran

, Aisen

, Jack

, Weiner

, Alzheimer’s Disease Neuroimaging Initiative (2012) Nonlinear time course of brain volume loss in cognitively normal and impaired elders. Neurobiol Aging 33, 845–855.

Bozoki

, Korolev

, Davis

, Hoisington

, Berger

(2012) Disruption of limbic white matter pathways in mild cognitive impairment and Alzheimer’s disease: A DTI/FDG-PET Study. Hum Brain Mapp 33, 1792–1802.

Landau

, Harvey

, Madison

, Koeppe

, Reiman

, Foster

, Weiner

, Jagust

, Initiative AsDN (2011) Associations between cognitive, functional, and FDG-PET measures of decline in AD and MCI. Neurobiol Aging 32, 1207–1218.

10.

Sperling

, LaViolette

, O’Keefe

, O’Brien

, Rentz

, Pihlajamaki

, Marshall

, Hyman

, Selkoe

, Hedden

(2009) Amyloid deposition is associated with impaired default network function in older persons without dementia. Neuron 63, 178–188.

11.

Miller

, Fenstermacher

, Bates

, Blacker

, Sperling

, Dickerson

(2008) Hippocampal activation in adults with mild cognitive impairment predicts subsequent cognitive decline. J Neurol Neurosurg Psychiatry 79, 630–635.

12.

Mosconi

(2005) Brain glucose metabolism in the early and specific diagnosis of Alzheimer’s disease. Eur J Nucl Med Mol Imaging 32, 486–510.

13.

Delbeuck

, Van der Linden

, Collette

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

14.

Greicius

, Srivastava

, Reiss

, Menon

(2004) Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: Evidence from functional MRI. Proc Natl Acad Sci U S A 101, 4637–4642.

15.

Petrella

, Sheldon

, Prince

, Calhoun

, Doraiswamy

(2011) Default mode network connectivity in stable vs progressive mild cognitive impairment. Neurology 76, 511–517.

16.

Sheline

, Raichle

, Snyder

, Morris

, Head

, Wang

, Mintun

(2010) Amyloid plaques disrupt resting state default mode network connectivity in cognitively normal elderly. Biol Psychiatry 67, 584–587.

17.

Hedden

, Van Dijk

, Becker

, Mehta

, Sperling

, Johnson

, Buckner

(2009) Disruption of functional connectivity in clinically normal older adults harboring amyloid burden. J Neurosci 29, 12686–12694.

18.

Gardini

, Venneri

, Sambataro

, Cuetos

, Fasano

, Marchi

, Crisi

, Caffarra

(2015) Increased functional connectivity in the default mode network in mild cognitive impairment: A maladaptive compensatory mechanism associated with poor semantic memory performance. J Alzheimers Dis 45, 457–470.

19.

, Wu

, Wang

, Zhang

, Dong

, Yao

, Li

(2010) Impairment and compensation coexist in amnestic MCI default mode network. Neuroimage 50, 48–55.

20.

Aisen

, Petersen

, Donohue

, Gamst

, Raman

, Thomas

, Walter

, Trojanowski

, Shaw

, Beckett

(2010) Clinical core of the Alzheimer’s Disease Neuroimaging Initiative: Progress and plans. Alzheimers Dement 6, 239–246.

21.

Jessen

, Wolfsgruber

, Wiese

, Bickel

, Mösch

, Kaduszkiewicz

, Pentzek

, Riedel-Heller

, Luck

, Fuchs

, Weyerer

, Werle

, van den Bussche

, Scherer

, Maier

, Wagner

; German Study on Aging, Cognition and Dementia in Primary Care Patients (2014) AD dementia risk in late MCI, in early MCI, and in subjective memory impairment. Alzheimers Dement 10, 76–83.

22.

, Seo

, Yang

, Kim

, Yoon

, Cho

, Noh

, Kim

, Chin

(2014) Comparison of cortical thickness in patients with early-stage versus late-stage amnestic mild cognitive impairment. Eur J Neurol 21, 86–92.

23.

Celone

, Calhoun

, Dickerson

, Atri

, Chua

, Miller

, DePeau

, Rentz

, Selkoe

, Blacker

(2006) Alterations in memory networks in mild cognitive impairment and Alzheimer’s disease: An independent component analysis. J Neurosci 26, 10222–10231.

24.

Bookheimer

, Strojwas

, Cohen

, Saunders

, Pericak-Vance

, Mazziotta

, Small

(2000) Patterns of brain activation in people at risk for Alzheimer’s disease. N Engl J Med 343, 450–456.

25.

Schuff

, Woerner

, Boreta

, Kornfield

, Shaw

, Trojanowski

, Thompson

, Jack

, Weiner

, Alzheimer’s Disease Nueroimaging Initiative (2009) MRI of hippocampal volume loss in early Alzheimer’s disease in relation to ApoE genotype and biomarkers. Brain 132, 1067–1077.

26.

, Ma

, Zhang

, Liang

, Chen

, Wang

, Zhang

, Wang

(2013) Prevalence of and potential risk factors for mild cognitive impairment in community-dwelling residents of Beijing. J Am Geriatr Soc 61, 2111–2119.

27.

Guo

(2007) Norm of auditory verbal learning test in the normal aged in China community. Chin J Clin Psychol 15, 132.

28.

Guo

(2006) Boston Naming Test in Chinese elderly, patient with mild cognitive impairment and Alzheimer’s dementia. Chin Ment Health J 20, 81.

29.

Zhao

(2007) Supermarket verbal fluency test in identification and differential diagnosis of dementia. Chin J Clin Psychol 15, 233.

30.

Guo

, Hong

, Lv

, Zhou

, Lu

, Ding

(2005) Application of Stroop color-word test on Chinese elderly patients with mild cognitive impairment and mild Alzheimer’s dementia. Chin J Neuromed 4, 701–704.

31.

Guo

, Zhao

, Chen

, Ding

, Hong

(2009) A comparison study of mild cognitive impairment with 3 memory tests among Chinese individuals. Alzheimer Dis Assoc Disord 23, 253–259.

32.

Zhao

, Ding

, Guo

, Mortimer

, Copenhaver

, Salmon

, Galasko

, DeCarli

, Dai

, Petersen

(2009) Shanghai community brain health initiative pilot study: Neuropsychological results. Alzheimers Dement 5, P14–P15.

33.

Rosenberg

, Ryan

, Prifitera

(1984) Rey Auditory-Verbal Learning Test performance of patients with and without memory impairment. J Clin Psychol 40, 785–787.

34.

Tupler

, Welsh

, Asare-Aboagye

, Dawson

(1995) Reliability of the Rey-Osterrieth Complex Figure in use with memory-impaired patients. J Clin Exp Neuropsychol 17, 566–579.

35.

Preacher

, Rucker

, MacCallum

, Nicewander

(2005) Use of the extreme groups approach: A critical reexamination and new recommendations. Psychol Methods 10, 178.

36.

Greicius

, Srivastava

, Reiss

, Menon

(2004) Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: Evidence from functional MRI. Proc Natl Acad Sci U S A 101, 4637–4642.

37.

Petrella

, Sheldon

, Prince

, Calhoun

, Doraiswamy

(2011) Default mode network connectivity in stable vs progressive mild cognitive impairment. Neurology 76, 511–517.

38.

Hafkemeijer

, van der Grond

, Rombouts

(2012) Imaging the default mode network in aging and dementia. Biochim Biophys Acta 1822, 431–441.

39.

Bai

, Zhang

, Yu

, Shi

, Yuan

, Zhu

, Zhang

, Qian

(2008) Default-mode network activity distinguishes amnestic type mild cognitive impairment from healthy aging: A combined structural and resting-state functional MRI study. Neurosci Lett 438, 111–115.

40.

Bai

, Liao

, Watson

, Shi

, Wang

, Yue

, Teng

, Wu

, Yuan

, Jia

(2011) Abnormal whole-brain functional connection in amnestic mild cognitive impairment patients. Behav Brain Res 216, 666–672.

41.

Zhang

H-Y

, Wang

S-J

, Liu

, Ma

Z-L

, Yang

, Zhang

Z-J

, Teng

G-J

(2010) Resting brain connectivity: Changes during the progress of Alzheimer disease. Radiology 256, 598–606.

42.

Cavanna

, Trimble

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

43.

Buckner

, Snyder

, Shannon

, LaRossa

, Sachs

, Fotenos

, Sheline

, Klunk

, Mathis

, Morris

(2005) Molecular, structural, and functional characterization of Alzheimer’s disease: Evidence for a relationship between default activity, amyloid, and memory. J Neurosci 25, 7709–7717.

44.

Karas

, Scheltens

, Rombouts

, van Schijndel

, Klein

, Jones

, van der Flier

, Vrenken

, Barkhof

(2007) Precuneus atrophy in early-onset Alzheimer’s disease: A morphometric structural MRI study. Neuroradiology 49, 967–976.

45.

Wolk

, Price

, Saxton

, Snitz

, James

, Lopez

, Aizenstein

, Cohen

, Weissfeld

, Mathis

(2009) Amyloid imaging in mild cognitive impairment subtypes. Ann Neurol 65, 557–568.

46.

Ikonomovic

, Klunk

, Abrahamson

, Wuu

, Mathis

, Scheff

, Mufson

, DeKosky

(2011) Precuneus amyloid burden is associated with reduced cholinergic activity in Alzheimer disease. Neurology 77, 39–47.

47.

, Wang

, Zang

, Tian

, Zhang

, Li

, Jiang

(2007) Regional coherence changes in the early stages of Alzheimer’s disease: A combined structural and resting-state functional MRI study. Neuroimage 35, 488–500.

48.

Sorg

, Riedl

, Mühlau

, Calhoun

, Eichele

, Läer

, Drzezga

, Förstl

, Kurz

, Zimmer

(2007) Selective changes of resting-state networks in individuals at risk for Alzheimer’s disease. Proc Natl Acad Sci U S A 104, 18760–18765.

49.

Corbetta

, Shulman

, Miezin

, Petersen

(1995) Superior parietal cortex activation during spatial attention shifts and visual feature conjunction. Science 270, 802–805.

50.

Koenigs

, Barbey

, Postle

, Grafman

(2009) Superior parietal cortex is critical for the manipulation of information in working memory. J Neurosci 29, 14980–14986.

51.

Grady

, Mitchell

(1999) Complete primary repair of exstrophy. J Urol 162, 1415–1420.

52.

Madden

, Turkington

, Provenzale

, Denny

, Hawk

, Gottlob

, Coleman

(1999) Adult age differences in the functional neuroanatomy of verbal recognition memory. Hum Brain Mapp 7, 115–135.

53.

Palop

, Mucke

(2010) Amyloid-beta-induced neuronal dysfunction in Alzheimer’s disease: From synapses toward neural networks. Nat Neurosci 13, 812–818.

54.

Reiman

, Jagust

(2012) Brain imaging in the study of Alzheimer’s disease. Neuroimage 61, 505–516.

55.

Sperling

, Dickerson

, Pihlajamaki

, Vannini

, LaViolette

, Vitolo

, Hedden

, Becker

, Rentz

, Selkoe

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

56.

Johnson

, Fox

, Sperling

, Klunk

(2012) Brain imaging in Alzheimer disease. Cold Spring Harb Perspect Med 2, a006213.

57.

, Cao

, Zhang

, Chen

, Ma

, Fleisher

, He

, Zhang

(2014) Structural and functional brain changes in the default mode network in subtypes of amnestic mild cognitive impairment. J Geriatr Psychiatry Neurol 27, 188–198.

58.

, Zheng

, Zhang

, Zhou

, Ma

, Chen

, Shu

, Zhang

(2013) Differences in functional brain activation and hippocampal volume among amnestic mild cognitive impairment subtypes. Curr Alzheimer Res 10, 1080–1089.