Participants
A total of 25 CADASIL patients from 14 families evaluated at Shanghai Ninth People’s Hospital between May 2016 and January 2019 were recruited for this study. For all patients, the diagnosis was confirmed by identification of pathogenic mutations in the NOTCH3 gene [19]. All subjects underwent detailed standard neurological examinations. Subjects were excluded from the study if they had severe depression or anxiety according to evaluation by two trained neuropsychologists using the Hamilton Depression Scale (HAMD) and the Hamilton Anxiety Scale (HAMA) [20, 21]. Subjects were diagnosed with severe depression and anxiety based on HAMD and HAMA scores > 17 and > 14, respectively. Neurological deficits in all subjects were assessed using the National Institutes of Health Stroke Scale (NIHSS) and the modified Rankin scale (mRs). Cognitive scores in all subjects were recorded by the Montreal Cognitive Assessment (MoCA) and Mini-Mental State Examination (MMSE). Most of the patients underwent both MRI and PET/computed tomography (PET/CT), but four underwent only MRI and another four underwent only PET/CT.
Forty-two healthy subjects were recruited as a control group based on the following criteria: no history of stroke, headache, cognitive impairment or vascular disease risk factors; no family history of cerebrovascular diseases or vascular disease risk factors; not taking medications and no substance addiction, such as drugs, cigarettes, or alcohol. All of the healthy subjects had normal results on neurological and general examinations. Twenty-one of the 42 controls underwent only MRI and the remaining 21 controls underwent only PET/CT. The sample size and demographic information of each group were listed in Table 1.
MRI acquisition
Subjects in the first control group and 21 CADASIL patients underwent MRI, including resting-state fMRI, structural MRI (T1-weighted, T2-weighted, and fluid-attenuated inversion recovery [FLAIR] imaging), and DTI on a 3.0 Tesla system (Trio Tim; Siemens Healthcare, Malvern, PA, USA) with a 12-channel head coil at East China Normal University. Soft earplugs and custom-fit foam were applied to reduce noise and movement artifacts. Resting-state fMRI was performed using a T2*-weighted gradient-echo echo-planar imaging pulse sequence with the following parameters: repetition time/echo time (TR/TE) = 2000/30 ms; flip angle = 90°; field of view (FOV) = 220 × 220 mm2; number of slices = 33; resolution = 3.44 × 3.44 × 4.38 mm3; total volume = 210. During the fMRI scan, the subjects kept their eyes closed but did not fall asleep. The whole-brain anatomical volume was obtained using a high-resolution T1-weighted 3-dimensional magnetization-prepared rapid- acquisition gradient-echo pulse sequence with the following parameters: TR = 2530 ms; TE = 2.34 ms; flip angle = 7°; FOV = 256 × 256 mm2; number of slices = 192; resolution = 1 × 1 × 1 mm3. T2-weighted imaging was obtained using turbo spin echo dark fluid sequence with the following parameters: TR/TE = 5500/83 ms; FOV = 220 × 220 mm2; number of slices = 33; resolution = 3.44 × 3.44 × 4.38 mm3. The parameters of FLAIR imaging were: TR/TE = 9000/93 ms; FOV = 220 mm2; number of slices = 33; resolution = 3.44 × 3.44 × 4.38 mm3. DTI was performed using a single-shot, spin-echo planar imaging sequence acquired in contiguous axial planes with the following parameters: 64 noncollinear directions, diffusion weighting of b = 1000 s/m2, an acquisition without diffusion weighting of b = 0; TR/TE = 8900/86 ms; FOV = 256 × 256 mm2, covered the whole brain; 70 contiguous slices; resolution = 2 × 2 × 2 mm3.
PET acquisition
18F-FDG PET data were acquired using a Siemens Biograph Truepoint HD 64 PET/CT at the PET Center of Huashan Hospital, Fudan University. 18F-FDG was synthesized and radiolabeled at the PET Center according to the manufacturer’s protocol under inspection of the Chinese Food and Drug Administration. Before 18F-FDG injection, subjects were asked to avoid strenuous physical activity and fast for about 6 hours to maintain blood glucose level < 8.0 mmol/L. After receiving injection of 18F-FDG at a dose of 5.55 MBq/kg [0.15 mCi/kg], subjects rested in a dimly lit room for 50 minutes. Before PET acquisition, a low-dose CT scan was performed for attenuation correction, and then 10-min PET images were reconstructed using a filtered back-projection algorithm. The matrix size of the reconstructed images was 168 × 168 × 148 with resolution of 2.04 × 2.04 × 1.5 mm3.
Data processing
Prior to preprocessing, all the raw DICOM data were converted to the Neuroimaging Informatics Technology Initiative format (NII) using MRICRON software (http://people.cas.sc.edu/rorden/mricron/install.html) and the quality of the images was checked visually.
fMRI data processing
The resting-state fMRI data were preprocessed using Data Processing Assistant for Resting-State fMRI (DPARSF; http://www.restfmri.net) [22, 23]. Data were preprocessed starting with removal of the first 10 volumes to ameliorate possible effects of scanner instability and the adaptation of subjects to the environment. Then, slice time correction was applied to reduce the effects of within-scan acquisition time differences between slices. To correct the effects of head motion, the fMRI images of each subject were realigned and registered. All subjects had head motions < 1.5° of rotation or 0.5 mm of mean frame wise displacement [24]. The fMRI images were then normalized into the Montreal Neurological Institute (MNI) space using the EPI template and smoothed by a full-width at half-maximum (FWHM) 8 mm Gaussian kernel. Following spatial smoothing, linear detrend was performed to remove noise due to long-term physiological shifts, movement-related noise remaining after realignment, and instrumental instability. To reduce further the effects of noise, the fMRI images were filtered with a temporal band-pass filter (0.01–0.08 Hz). Finally, the six head motion parameters, global mean signal, white matter signal, and cerebrospinal fluid signal were regressed out as nuisance covariates to remove these unwanted signals.
With reference to previous studies [18, 25, 26], 11 separate regions comprising the left DMN and 11 mirrored regions comprising the right DMN were defined as regions of interest (ROIs). The 11 ROIs were spheres of radius 8 mm in the dorsal MPFC (dMPFC), anterior MPFC (aMPFC), ventral MPFC (vMPFC), posterior IPL (pIPL), temporal parietal junction (TPJ), lateral temporal cortex (LTC), temporal pole (TempP), PCC, retrosplenial cortex (RSC), PHC, and hippocampal formation (HF) (See Supplemental Table 1 and Supplemental Figure 1 for coordinates and spatial positions). Average fMRI time-series were calculated across every voxel in each ROI. The absolute value of Fisher’s z-transformed Pearson’s correlation coefficient between each pair of time-series was defined as the functional connectivity (FC) strength.
Graph analysis of the pairwise (11 × 11) correlation matrixes was performed using GRETNA (v2.0.0; https://www.nitrc.org/projects/gretna/) [27]. Global and nodal network properties, including nodal degree centrality, nodal shortest path length, nodal clustering coefficient, nodal efficiency, nodal local efficiency, betweenness centrality, global efficiency, assortativity coefficient, and small-worldness, were calculated to delineate the integrative and local topological architecture of the DMN, respectively. Their definitions and calculations of the nodal and global network properties were summarized in Supplemental Table 2.
T1-weighted data processing
The T1-weighted MRI data were preprocessed using the Computational Anatomy Toolbox (CAT12; http://dbm.neuro.uni-jena.de/cat12) implemented in statistical parametric mapping software (SPM12; htttp://www.fil.ion.ucl.ac.uk/spm/). First, all T1-weighted MRI data were normalized into the MNI space using the Diffeomorphic Anatomic Registration Through Exponentiated Lie algebra algorithm (DARTEL). The bias field inhomogeneities were corrected to remove non-uniform intensities. Normalized images were then segmented into gray matter, white matter, and cerebrospinal fluid components. The total intracranial volume (TIV) of each participant was evaluated to correct for the effects of differences in brain size. The internal gray matter threshold was set to 0.2 to exclude artifacts on the gray–white matter border. Thereafter, all preprocessed scans were smoothed with the FWHM 6 mm Gaussian kernel. Finally, average GMV was calculated across every voxel in each ROI.
DTI data processing
The raw DTI data were preprocessed using FMRIB Software Library (FSL; http://www.fmrib.ox.ac.uk/fsl/index.html). First, eddy current correction was performed to correct for head motion artifacts and eddy current distortions. Then, the brain of each subject was extracted using the FSL Brain Extraction Tool (BET). Tensor reconstruction and fiber tracking were applied by Diffusion Toolkit TrackVis (https://www.nitrc.org/projects/trackvis). The Fiber Association Continuous Tracking (FACT) algorithm in Diffusion Toolkit was applied to obtain the whole-brain fiber tracts. The main parameters in fiber tractography were as follows: maximum turning angle threshold at 35°; minimum fractional anisotropy (FA) threshold of 0.2. Then, SPM12 was applied to bring all the individual tracts into the MNI space by nonlinear transformation methods. In the normalization step, tracts were spatially normalized by: coregistering T1-weighted MRI to the corresponding FA image; calculating the deformation field of the individual coregistered T1-weighted image space to the MNI space; and applying the deformation field to tracts and bringing them into the MNI space. Thereafter, TrackVis was used to record the number of tracts (NT) passing through each ROI.
PET data processing
First, PET images of each subject were processed using SPM12 software with spatial normalization and smoothing. The PET template in SPM12 was used in the spatial normalization step. The FWHM 8 mm Gaussian kernel was applied in the smoothing step. The average glucose metabolism was then calculated across every voxel in each ROI.
Statistical analysis
Statistical analysis was performed using IBM SPSS Statistics for Windows (SPSS, Chicago, IL, USA). The Chi-square tests and permutation tests (permutation times = 10,000) were used to compare demographic, clinical, and imaging characteristics between the CADASIL and control groups, as appropriate. Further, the two-sample t test was applied for voxel-wise metabolism comparisons between the CADASIL and corresponding control groups using SPM12 software. Subsequently, partial correlations were established to estimate the relations between the cognitive deficits and the imaging characteristics showing significant between-group differences. Age, sex, and education levels were entered as covariates in partial correlation analysis. Benjamini-Hochberg false discovery rate (FDR) correction was further used to avoid type-I errors in the multiple comparisons and correlations. The results of two-sample comparisons and partial correlations were regarded significant at p < 0.05 (two-tailed) with FDR correction.