Abstract
Background:
Recent advances in resting-state functional MRI have revealed altered functional networks in Alzheimer’s disease (AD), especially those of the default mode network (DMN) and central executive network (CEN). However, few studies have evaluated whether small vessel disease (SVD) or combined amyloid and SVD burdens affect the DMN or CEN.
Objective:
The aim of this study was to evaluate whether SVD or combined amyloid and SVD burdens affect the DMN or CEN.
Methods:
In this cross-sectional study, we investigated the resting-state functional connectivity within DMN and CEN in 37 Pittsburgh compound-B (PiB)(+) AD, 37 PiB(–) subcortical vascular dementia (SVaD), 13 mixed dementia patients, and 65 normal controls.
Results:
When the resting-state DMN of PiB(+) AD and PiB(–) SVaD patients were compared, the PiB(+) AD patients displayed lower functional connectivity in the inferior parietal lobule while the PiB(–) SVaD patients displayed lower functional connectivity in the medial frontal and superior frontal gyri. Compared to the PiB(–) SVaD or PiB(+) AD, the mixed dementia patients displayed lower functional connectivity within the DMN in the posterior cingulate gyrus. When the resting-state CEN connectivity of PiB(+) AD and PiB(–) SVaD patients were compared, the PiB(–) SVaD patients displayed lower functional connectivity in the anterior insular region. Compared to the PiB(–) SVaD or PiB(+) AD, the mixed dementia patients displayed lower functional connectivity within the CEN in the inferior frontal gyrus.
Conclusions:
Our findings suggest that in PiB(+) AD and PiB(–) SVaD, there is divergent disruptions in resting-state DMN and CEN. Furthermore, patients with combined amyloid and SVD burdens exhibited more disrupted resting-state DMN and CEN than patients with only amyloid or SVD burden.
Keywords
INTRODUCTION
Alzheimer’s disease (AD) results from an imbalance in the production and clearance of amyloid, which leads to synaptic and neuronal losses [1]. Subcortical vascular dementia (SVaD) is characterized by extensive small vessel disease (SVD) such as white matter hyperintensities (WMH) and multiple lacunes [2]. Using Pittsburgh compound-B (PiB)-PET, a sensitive method that detects amyloid [3], about 30% of clinically diagnosed SVaD patients have a significant amyloid burden and show more severe cognitive dysfunction and brain atrophy compared to patients with only amyloid or SVD burden [4].
Resting-state functional MRI (rs-fMRI) reveals the correlated spontaneous activity within cortical and subcortical regions that are functionally related [5]. With recent advances in rs-fMRI, many researchers have become interested in intrinsic network connectivity such as default mode network (DMN) and central-executive network (CEN) of aging and degenerative dementia [6, 7]. The DMN, which consists of the ventromedial prefrontal cortex and posterior cingulate cortex, contributes to both episodic memory and visual imagery [8]. On the contrary, the CEN, which consists of dorsolateral prefrontal cortex and lateral posterior parietal cortex, plays an important role in attention, working memory, planning, problem solving, and decision making [9].
Previous studies have reported that both the DMN and CEN are disrupted in AD patients [6, 10] and that SVD is associated with disruption in resting-state networks [11–15]. However, to our knowledge, there were few studies comparing the patterns of DMN or CEN disruption between PiB(+) AD and PiB(–) SVaD patients. It is possible that resting-state DMN is more disrupted in PiB(+) AD, while resting-state CEN is more disrupted in PiB(–) SVaD because memory dysfunction is predominant in AD patients while attention and executive dysfunctions are predominant in SVaD patients [16].
The clinical implication of amyloid burden in patients with SVD burden is an important issue. Pathological studies suggested that AD pathologies and SVD pathologies are independently associated with cognitive impairment in specific domains [17]. Furthermore, more recent studies revealed that amyloid and SVD synergistically affected cognitive dysfunction or structural abnormalities [18, 19]. It might be, therefore, reasonable to determine whether mixed dementia with combined AD and SVD pathologies may have more disrupted DMN and CEN than PiB(+) AD or PiB(–) SVaD.
We investigated clinically diagnosed AD and SVaD patients who underwent PiB-PET and rs-fMRI. First, we aimed to determine whether PiB(–) SVaD patients had distinctive patterns of resting-state DMN or CEN disruptions compared to those of PiB(+) AD patients. We also investigated whether patients with combined amyloid and SVD burdens (mixed dementia) displayed more severe resting-state network disruptions than patients with only amyloid (PiB(+) AD) or SVD burdens (PiB(–) SVaD).
METHODS
Participants
In this cross-sectional study, 65 normal controls (NCs), 37 patients with PiB(+) AD, 37 patients with PiB(–) SVaD, and 13 patients with PiB(+) SVaD (mixed dementia) were recruited from the memory clinic at Samsung Medical Center.
We selected patients with PiB(+) AD, PiB(–) SVaD, or mixed dementia according to clinical diagnosis and PiB-PET status (see the below methods for PiB-PET). PiB(+) AD patients were those who (1) met probable AD dementia criteria proposed by the National Institute of Neurological and Communicative Disorders and Stroke and the AD and Related Disorders Association [20]; (2) had minimal WMH (periventricular WMH <5 mm and deep WMH <5 mm); (3) and had significant amyloid burden (PiB SUVR ≥1.5) as measured by PiB-PET. SVaD patients were those who (1) met the diagnostic criteria for vascular dementia as determined by the Diagnostic and Statistical Manual of Mental Disorders–Fourth Edition; (2) had focal neurological symptoms and signs; (3) had severe WMH, defined as periventricular WMH ≥10 mm and deep WMH ≥25 mm, as modified from the Fazekas ischemia criteria [21]. Then according to amyloid burden status, SVaD patients were divided into PiB(–) SVaD (PiB SUVR <1.5) and mixed dementia (PiB SUVR ≥1.5).
Patients were excluded if they had territorial infarctions, WMH due to radiation injury, multiple sclerosis, vasculitis, or leukodystrophy. Blood tests included complete blood count, blood chemistry test, vitamin B12/folate measurement, syphilis serology, thyroid function test, and APOE genotyping. The detailed diagnostic process and neuropsychological test are provided in Supplementary Material 1.
The NC subjects were recruited from the community. Volunteers were screened using brain MRI and detailed neuropsychological tests including Korean version of Mini-Mental State Examination (K-MMSE) and Clinical Dementia Rating-sum of boxes (CDR-SOB). Those who had minimal WMH and whose neuropsychological test results were within normal range (above 16th percentile of age and education matched norms) were enrolled as NC subjects.
We obtained written consent from each participant. This study protocol was approved by the Institutional Review Board of Samsung Medical Center.
[11C] PiB-PET
All patients completed the [11C] PiB-PET scan. The methods of PiB-PET acquisition and analysis are described in detail in Supplementary Material 2. PiB standardized uptake value ratio (SUVR) ≥1.5 was considered as PiB-positive [4].
MR imaging techniques
All imaging was carried out at Samsung Medical Center with a Philips Intera Achieva 3.0 Tesla scanner equipped with an 8-channel SENSE head coil (Philips Healthcare, the Netherlands). Whole-brain Echo planar imaging (EPI) time series scans (TR = 3 s; TE = 35 ms; flip angle = 90°; 1.7 × 1.7 × 4 mm3 voxel resolution) consisted of 100 volumes and lasted for approximately five minutes, which is known to be sufficient to evaluate resting-state functional connectivity [22]. During each scan, participants were instructed to rest with their eyes open and fixated on a centrally placed, white crosshair against a black background. A high-resolution T1-weighted anatomical image was also acquired using a magnetization-prepared gradient echo (MPRAGE) sequence (TR = 9.9 ms; TE = 4.6 ms; flip angle = 8°; 0.5 × 0.5 × 0.5 mm3 voxel resolution).
Preprocessing of fMRI data
Analysis of Functional NeuroImages (AFNI) software (http://afni.nimh.nih.gov/afni) was used to preprocess the fMRI data [23]. The first three volumes of each functional image were discarded to allow for stabilization of the magnetic field. The residual images were de-spiked, and were then corrected for slice-time acquisition differences and head motion [24]. To exclude the possibility that differences in motion can introduce artifactual group differences in functional connectivity [25, 26], we screened for head motion artifacts using estimated displacement through the motion correction parameters of x, y, z translation and three rotation axes at the motion correction stage. The estimated displacement was less than 0.3 mm for the Euclidean L2 norm of motion displacement between successive time-series and also less than 2 mm translations in any of the three directions or 2° rotations around any of the axes. The Local Pearson Correlation cost function was used to co-register the functional images of each subject to the structural brain [27]. Next, images were normalized to the Montreal Neurological Institute (MNI) 152 template space [28]. Subsequently, a 6 mm full-width-at-half-maximum (FWHM) Gaussian kernel was used for spatial smoothing.
Independent component analysis
The resting-state DMN and CEN were assessed by group spatial independent component analysis (sICA). The group sICA was carried out using Group ICA for the fMRI toolbox (GIFT) (http://mialab.mrn.org/software/gift/) software [29]. GIFT software includes group data reduction using principal component analysis, ICA calculation and back-reconstruction for each subject. Minimum description length criteria were used to determine the optimal number of components [30]. Forty-four estimated independent components were identified. An independent component estimation was carried out using the Infomax algorithm [31]. For each component, the z-value reflected the degree to which the time series of each voxel was associated with one of the specific components. The functional connectivity levels were determined using the spatial differences across the z-values of each component. A combination of manual and template-matching procedures was used to choose a component of the DMN and CEN [32]. The component that included the ventromedial prefrontal cortex and posterior cingulate cortex was selected as DMN component and the component that included the dorsolateral prefrontal cortex and posterior parietal cortex was selected as CEN component.
W-score maps
To obtain measurements of functional alteration in patient groups compared with NCs, W-score maps were computed for each patient using the NCs as reference [33]. To obtain the W-score maps, multiple regressions were performed using age, gender, and education as covariates in the NCs. Then, W-score maps were computed as follow: W-score = [(patient’s functional connectivity value) - (value predicted in the NCs for the patient’s age, gender, education)]/standard deviation of residuals in control groups. The W-scores were reversed so that positive W-scores represent fewer functional connections compared with NCs. Then, individual W-score maps were averaged in the each patient group.
Neuropsychological tests
All patients underwent neuropsychological tests [34, 35]. Memory was assessed using delayed recall scores of the Seoul Verbal Learning Test (SVLT) and Rey-Osterrieth Complex Figure Test (RCFT). Frontal-executive function was assessed using the phonemic Controlled Oral Word Association Test (COWAT) and Stroop color reading test. General cognition and disease severity were assessed using the K-MMSE and CDR-SOB, respectively. The presence of depression was assessed by geriatric depression scale (GDS) ≥18 [36].
Statistical analysis
One-way analysis of variance and Bonferroni’s post hoc analyses were performed to evaluate the differences in demographics and cognitive test results among the four groups.
To determine the differences in the DMN and CEN connectivity among the four groups, an analysis of covariance was performed using age, gender and education as covariates. Monte Carlo simulations were performed using the AFNI’s 3dClustSim program to control for Type I errors (parameters: individual voxel p-value = 0.02, simulated 10,000 times iteratively, 6 mm FWHM Gaussian filter width with a whole-brain mask). The 3dClustSim program estimates the overall significance level achieved with various combinations of individual voxel probability thresholds and cluster size thresholds [37]. We obtained a corrected significance level of Pα< 0.05 (uncorrected individual voxel height threshold of p < 0.02 (F > 3.38)). Subsequently, to examine the inter-group differences, post hoc two-sample t tests were performed between pairs of groups for voxel-wise statistics (corrected significant level of Pα< 0.05), after adjusting for age, gender, and education. The significant regions between the groups that have >120 voxel size were clinically interpreted.
To evaluate the correlation between resting state DMN/CEN connectivity and neuropsychological test results among patients, we performed multiple linear regression analyses after adjusting for age, gender, and education.
RESULTS
Demographics of subjects
As shown in Table 1, there were age differences between NCs, PiB(–) SVaD, PiB(+) AD, and mixed dementia patients. The education level was lower in PiB(–) SVaD patients compared to NCs.
Comparisons of functional connectivity within the resting-state default mode network
Compared to the NCs, the PiB(+) AD patients displayed lower functional connectivity within the DMN in the left supramarginal gyrus (Table 2, Fig. 1A) and the PiB(–) SVaD patients displayed lower functional connectivity within the DMN in the left superior frontal gyrus (Table 2, Fig. 1B). There were no differences in functional connectivity within the DMN between the NCs and mixed dementia patients. However, when we used W-scores, mixed dementia patients displayed higher W-scores in widespread areas, representing lower functional connectivity within the DMN compared to the NCs (Supplementary Figure 1). Further analyses without depressed patients (GDS <18) showed similar results (Supplementary Figure 2).
In the direct comparison between the PiB(+) AD and PiB(–) SVaD patients, the PiB(+) AD patients displayed lower functional connectivity within the DMN in the left inferior parietal lobule while the PiB(–) SVaD patients displayed lower functional connectivity in the right medial frontal and left superior frontal gyri (Table 2, Fig. 1C). Compared to the PiB(–) SVaD patients, the mixed dementia patients displayed lower functional connectivity within the DMN in the right posterior cingulate gyrus while there was no region where the PiB(–) SVaD patients displayed lower functional connectivity compared to the mixed dementia patients (Table 2, Fig. 1D). Compared to the PiB(+) AD patients, the mixed dementia patients displayed lower functional connectivity within the DMN in the left posterior cingulate gyrus and left superior frontal gyrus while there was no region where the PiB(+) AD patients displayed lower functional connectivity compared to the mixed dementia patients (Table 2, Fig. 1E).
There was no correlation between resting-state DMN connectivity and neuropsychological test results (Supplementary Table 1).
Comparisons of functional connectivity within the resting-state central executive network
Compared to the NCs, the PiB(–) SVaD patients displayed lower functional connectivity within the CEN in the left insular area (Table 3, Fig. 2A). There were no differences in functional connectivity within the CEN between the NCs and the PiB(+) AD or the mixed dementia patients. However, when we used W-score, the PiB(+) AD patients displayed higher W-score in several areas and the mixed dementia patients displayed higher W-score in widespread areas,representing lower functional connectivity within the CEN compared to the NCs (Supplementary Figure 3). Further analyses without depressed patients (GDS <18) showed similar results (Supplementary Figure 4).
In the direct comparison between the PiB(+) AD and PiB(–) SVaD patients, the PiB(–) SVaD patients displayed lower functional connectivity within the CEN in the left insular area while there was no region where the PiB(+) AD patients displayed lower functional connectivity (Table 3, Fig. 2B). Compared to the PiB(–) SVaD patients, the mixed dementia patients displayed lower functional connectivity within the CEN in the left inferior frontal gyrus while there was no region where the PiB(–) SVaD patients displayed lower functional connectivity compared to the mixed dementia patients (Table 3, Fig. 2C). Compared to the PiB(+) AD patients, the mixed dementia patients displayed lower functional connectivity within the CEN in the left inferior frontal gyrus while there was no region where the PiB(+) AD patients displayed lower functional connectivity (Table 3,Fig. 2D).
There was no correlation between resting-state CEN connectivity and neuropsychological test results (Supplementary Table 1).
DISCUSSION
In this study, we reported novel findings concerning altered resting-state DMN and CEN connectivities in the PiB(+) AD, PiB(–) SVaD and mixed dementia patients, who were analyzed using rs-fMRI and amyloid imaging. The major findings of this study were as follows: 1) When the resting-state DMN of PiB(+) AD and PiB(–) SVaD patients were compared, the PiB(+) AD patients displayed lower functional connectivity especially in the inferior parietal lobule while the PiB(–) SVaD patients displayed lower functional connectivity especially in the medial frontal and superior frontal gyri. Compared to the PiB(–) SVaD or PiB(+) AD patients, the mixed dementia patients displayed lower functional connectivity within the DMN especially in the posterior cingulate gyrus. 2) When the resting-state CEN connectivity of PiB(+) AD and PiB(–) SVaD patients were compared, the PiB(–) SVaD patients displayed lower functional connectivity especially in the anterior insular region. Compared to the PiB(–) SVaD or PiB(+) AD patients, the mixed dementia patients displayed lower functional connectivity within the CEN especially in the inferior frontal gyrus. Taken together, our findings suggested that the DMN is disrupted in both PiB(+) AD and PiB(–) SVaD patients with disease specific vulnerability while the CEN disruption is more unique to PiB(–) SVaD. Furthermore, patients with combined AD and SVD burdens exhibited more disrupted DMN and CEN than patients with only AD or only SVD burdens.
Our first major finding was that within DMN, the PiB(+) AD patients had lower functional connectivity in the inferior parietal lobule while the PiB(–) SVaD patients displayed lower functional connectivity in the medial frontal and superior frontal regions. To our knowledge, this is the first study suggesting that both the PiB(+) AD and PiB(–) SVaD patients had disease specific vulnerability in the DMN disruption. The DMN is known to comprise several distinct interacting subsystems [8]. In particular, the DMN includes the medial prefrontal regions, which are responsible for self-referential judgments and self-relevant mental simulations [38]. On the contrary, the inferior parietal lobule is responsible for monitoring the external environment [8]. Further studies are needed to examine whether PiB(+) AD and PiB(–) SVaD patients demonstrate the distinct deficits in these specific tasks.
We also found that the PiB(–) SVaD patients had lower functional connectivity in the anterior insular region compared to PiB(+) AD patients, suggesting that CEN disruption is more unique to PiB(–) SVaD. Accumulating evidence suggests that the anterior insular region plays a critical role in attention and executive function [39]. We have previously reported that executive dysfunction is a core feature of PiB(–) SVaD that distinguishes it from PiB(+) AD [16]. Thus, the lower functional connectivity within the CEN in the anterior insular region might be one of the pathophysiologic mechanisms that would explain executive dysfunction in PiB(–) SVaD patients.
There are several possible mechanisms that explain why PiB(–) SVaD patients had a unique pattern of altered resting-state DMN and CEN connectivity. It is possible that SVD might have disconnected white matter tracts in the frontal region, which leads to functional connectivity problem especially in the anterior part of DMN or CEN. Alternatively, it is possible that the lower connectivity within the networks shown in PiB(–) SVaD are more reflective of alterations in vascular physiology rather than underlying differences in neural activation [12]. Indeed, previous studies suggested that patients with microangiopathy have reduced intrinsic microvascular oscillation [40].
Our second major finding was that patients with combined AD and SVD burdens exhibited more disrupted DMN and CEN than patients with only AD or only SVD burdens. This is in line with our previous study which showed that mixed dementia causes more memory dysfunction than PiB(–) SVaD [4]. The lower functional connectivity within the DMN and CEN in the frontal region of the mixed dementia patients compared to the PiB(+) AD patients might be explained by the extra SVD effects on the frontal region in the mixed dementia patients.
The reason that the mixed dementia patients revealed more disrupted DMN in the posterior cingulate region compared to the PiB(+) AD and more disrupted CEN in the inferior frontal gyrus compared to the PiB(+) AD or PiB(–) SVaD remains unclear. However, considering that these regions are characteristically apt to be affected in AD or SVaD patients, this finding might be related to the synergistic effects of AD and SVD pathologies on resting-state network of DMN and CEN in these regions. These findings are consistent with previous studies showing that amyloid and SVD burdens synergistically affect visuospatial and cortical thinning in the posterior region [18, 19]. Further investigation is needed to determine whether amyloid and SVD synergistically disrupt the DMN or CEN connectivity in these regions.
There are some limitations in this study. First, PiB-PET scans was not performed in the NCs. Previous studies showed that approximately 10 to 40% of cognitively normal elderly people are PiB-positive [41]. Thus, the degree of decreased functional connectivity in the patients’ groups compared to NCs may have been underestimated. Second, the lack of significant difference between mixed dementia and controls raises concerns about the internal validity of the results. However, it might be related to the small number of mixed dementia patients. When W-scores were calculated in each patient using NCs as reference and entering age, gender, and education as covariates, mixed dementia patients exhibited lower functional connectivities within DMN and CEN compared to NCs. Third, the demographics showed some differences across the four groups. However, we believe this reflects the characteristics of participants, and we controlled for these factors in the analyses. Finally, PiB-PET may not be sufficiently sensitive to detect soluble amyloid oligomers or very low levels of compact plaques.
In summary, our findings suggest that PiB(+) AD and PiB(–) SVaD have distinct patterns of altered connectivity in the resting-state DMN and CEN. Furthermore, the DMN and CEN in patients with combined AD and SVD burdens (mixed dementia) are more disrupted than in patients with only AD (PiB(+) AD) or only SVD burdens (PiB(–) SVaD). Our findings highlight that there may be distinct pathobiological mechanisms of intrinsic network connectivity in AD and SVaD patients.
Footnotes
ACKNOWLEDGMENTS
This study was supported by Basic Research Program through the National Research Foundation of South Korea (NRF) funded by the Ministry of Education (NRF-2013R1A1A2065365), the Korean Healthcare Technology R&D Project, Ministry for Health and Welfare Affairs (HIC120713), a Korean Science and Engineering Foundation (KOSEF) NRL program grant funded by the Korean government (MEST; 20110028333), a Samsung Medical Center Clinical Research Development Program grant (CRL108011 & CRS 110-14-1), the Korea Ministry of Environment (MOE) as the Environmental Health Action Program (2014001360002), the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health & Welfare, Republic of Korea (HI14C2768), and the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (NRF-2015R1C1A2A01053281).
