Participants
We enrolled 33 bvFTD patients from July 1, 2017, to December 31, 2020, at the Department of Neurology of Xuanwu Hospital, who fulfilled the 2011 consensus probable bvFTD criteria [19]. We also recruited 33 age- and sex-matched healthy individuals, who were unrelated to the patients, to serve as the controls for bvFTD patients. All participants underwent clinical interviews, physical examinations, neuropsychological assessments, genetic testing, and a brain [18F]-FDG PET/MRI.
In addition, 18 asymptomatic participants were recruited from the Department of Neurology of Xuanwu Hospital in September 2017, who belonged to a family with an autosomal dominant P301L mutation of the MAPT gene. We defined participants as asymptomatic when both they and their spouse denied cognitive and behavioral disturbances and had normal scores on neuropsychiatric measures. All participants underwent genetic screening, and six participants were found to be carriers of the mutation. The remaining 12 were mutation-negative and were used as controls owing to their similar early environment, genetic background, and demographics. Each participant underwent clinical interviews, physical examinations, neuropsychological assessments, and a brain [18F]-FDG PET/MRI. All participants had been followed up prospectively with annual clinical examinations between September 2017 and October 2021 at Xuanwu Hospital. During the 4-year follow-up period, all subjects remained symptom-free and none developed any bvFTD symptoms or any other neurodegenerative disease.
Neuropsychological assessments
The neuropsychological test battery consisted of widely used neuropsychological assessments that measure the cognitive function in the domains of memory, language, and behavioral abnormalities. Global cognitive screening measures comprised the Mini-Mental State Examination (MMSE), the Montreal Cognitive Assessment (MoCA), and the Frontotemporal Lobar Degeneration-Clinical Dementia Rating scale (FTLD-CDR). Word-list memory was evaluated using Rey’s Auditory-Verbal Learning Test (AVLT). Language was evaluated using the Boston Naming Test (BNT). The severity of behavioral abnormalities was assessed using the Frontal Behavior Inventory (FBI), which is divided into the negative apathy symptom subscale (first 12 items) and the positive disinhibition symptom subscale (last 12 items).
PET/MRI acquisition parameters
All images were acquired on a hybrid 3.0 T TOF PET/MRI scanner (SIGNA PET/MR, GE Healthcare, WI, USA) [20]. PET and MRI data were acquired simultaneously using a vendor-supplied 19-channel head and neck union coil. Subjects were injected intravenously with [l8F]-FDG (3.7 MBq/kg), and underwent three-dimensional (3D) T1-weighted sagittal imaging and [l8F]-FDG-PET imaging 40 min later during the same session.
A 3D T1-weighted fast field echo sequence (repetition time [TR] = 6.9 ms, echo time [TE] = 2.98 ms, flip angle = 12°, inversion time = 450 ms, matrix size = 256 × 256, field of view = 256 × 256 mm2, slice thickness = 1 mm, 192 sagittal slices with no gap, voxel size = 1 × 1 × 1 mm3, and acquisition time = 4 min 48 s) was used for data acquisition. Static [l8F]-FDG-PET data were acquired using the following scanning parameters: matrix size = 192 × 192, field of view = 350 × 350 mm2, and pixel size = 1.82 × 1.82 × 2.78 mm3, and included corrections for random coincidences, dead time, scatter, and photon attenuation.
Structural image preprocessing
Data were preprocessed using the Computational Anatomy Toolbox (CAT12) toolbox segment data pipeline implemented within Statistical Parametric Mapping 12 (SPM12, www.fl.ion.ucl.ac.uk/spm). Structural MRI images were normalized to standard Montreal Neurological Institute (MNI) space using diffeomorphic anatomical registration through exponentiated lie algebra normalization as implemented in SPM12. The images were then smoothed using an 8-mm full-width half-maximum isotropic Gaussian kernel for all directions.
PET/MRI image preprocessing
The [l8F]-FDG PET image processing and analyses were performed using SPM12 implemented in the Matlab software (Mathwork, Inc., Natick, MA, USA). After normalizing the structural MRI images, the transformation parameters determined by the T1-weighted image spatial normalization were applied to the co-registered PET images for PET spatial normalization. The images were then smoothed using an isotropic Gaussian kernel with an 8-mm full-width half-maximum. The FDG-PET scan intensity was normalized using a whole cerebellum reference region to generate standardized uptake value ratio (SUVR) images.
Analysis at the striatal subregion level
We used the substriatal regions of interest (ROIs) from the Oxford-GSK-Imanova Striatal Connectivity Atlas [9], which is a probabilistic atlas of substriatal regions segmented according to their white-matter connectivity to cortical regions. Based on the differential cortical connectivity patterns, the atlas subdivides the striatum into seven subregions: limbic, executive, rostral-motor, caudal-motor, parietal, occipital, and temporal subregions (Fig. 1A). We focused on the limbic striatum because it is the principal subcortical relay of the limbic CSTC circuits. Mean [18F]-FDG PET SUVRs were determined separately in the unilateral limbic portion of the striatum using ROIs provided by the atlas.
Metabolic connectivity analysis of the limbic CSTC circuit
We used sparse inverse covariance estimation (SICE), which is a method previously validated by Huang et al. [21] A series of nodes (N = 172) that represent brain ROIs for the connectivity analysis were selected to cover the whole brain [22, 23]. Because our hypothesis was specifically focused on the limbic CSTC circuit, we selected six ROIs for the functional connectivity analysis, which included the vmPFC, OFC, rectus gyrus, ACC, limbic striatum, and thalamus (i.e., the main components of the limbic CTSC circuitry). We performed a seed-based analysis with the six ROIs to investigate the connectivity between the frontal cortex and the limbic striatum, the limbic striatum and the thalamus, and the thalamus and the frontal cortex.
Statistical analyses
The GraphPad Prism software (version 8.3.0, GraphPad Software Inc, La Jolla, CA, USA) was used for all statistical analyses. The normality of the distribution for all numerical variables was evaluated using Shapiro-Wilk tests, and homogeneity of variance was assessed by F tests. Numerical variables are presented as means ± standard deviations or medians (Q1–Q3). Group differences were assessed using an independent t-test for normally distributed and homogeneous variance data, otherwise, the Mann-Whitney test was used. Comparisons of categorical variables were analyzed using chi-square and Fisher’s exact tests.
The structural and [18F]-FDG PET data were subjected to voxel-wised whole-brain two-sample t-tests based on the framework of a general linear model (GLM) in SPM12, with age and sex as covariates. Brain regions with significant volume and FDG changes were determined using a voxel threshold of p < 0.05 (familywise error [FWE]-corrected). We then conducted the atlas-based ROI analysis of the PET images to extract the regional SUVRs of the relays in limbic CSTC circuit for further correlation analyses. To compare metabolic connectivity between groups, we used non-parametric permutation tests with 5000 permutations to determine significance. The p-values were calculated as the fraction of the difference in distribution values that exceeded the difference value between the actual groups.
For the bvFTD group, we performed a Pearson’s correlation between the [18F]-FDG SUVR of the relays in the limbic CSTC circuit and neuropsychiatric assessment scores using a threshold of p < 0.05 (false discovery rate [FDR]-corrected). Furthermore, multivariable linear regression analyses were performed to examine whether the severity of behavior disruption (FBI total score, FBI apathy, FBI disinhibition) was associated with involvement of limbic striatum covarying out the general cognition (MMSE) or disease severity (FTLD-CDR). For all analyses, a p-value < 0.05 indicated statistical significance.