Participants
This research was carried out as part of the Sunnybrook Dementia Study, a prospective longitudinal cohort study of cognitively normal ageing, mild cognitive impairment, and neurodegenerative dementias conducted at Sunnybrook Health Sciences Centre, University of Toronto (ClinicalTrials.gov: NCT01800214). The design and methods have been previously published [11]. The study was approved by the Research Ethics Board at Sunnybrook Health Sciences Centre and all participants, or their caregivers when appropriate, provided written informed consent, in accordance with the Declaration of Helsinki.
Consecutive deceased participants with autopsy-confirmation of pure FTLD (i.e., without other pathological comorbidity) were retrospectively identified, including cases of FTLD due to Pick’s disease (i.e., FTLD-Tau [Pick’s]), TDP-43 proteinopathy (i.e., FTLD-TDP), Progressive Supranuclear Palsy (i.e., FTLD-Tau [PSP]), and Corticobasal Degeneration (i.e., FTLD-Tau [CBD]). Cognitively impaired participants with autopsy-confirmation of pure Alzheimer’s pathology without other proteinopathies (e.g., synucleinopathy or FTLD-TDP) or large vessel cerebrovascular disease were consecutively selected. Healthy matched controls were randomly selected from the Sunnybrook Dementia Study cohort and included in the present study for comparison purposes for the antemortem MRI study. For enrolment in the study, these participants had to be between the age of 40 and 90, be fluent in English, completed 8 years of education or higher, and have no significant memory complaints. They were excluded if they were being treated or had a history of being treated for a psychiatric or neurological illness (i.e., other than primary neurodegenerative diagnoses), history of alcohol or substance abuse or dependence, were currently using psychoactive medications not indicated for their primary neurodegenerative diagnoses, or had medical contraindications to MRI.
Genetic studies
Genomic DNA was extracted from whole blood using Qiagen kits. DNA from participants with a clinical diagnosis of a frontotemporal dementia spectrum disorder was screened for pathogenic mutations known to cause FTLD: C9orf72 [12], GRN [13], and MAPT [14]. A pathogenic expansion of C9orf72 was considered as having more than 30 repeats. All selected cases of AD were sporadic in nature, free of mutations in the APP, PSEN1, and PSEN2 genes [14].
Neuropathology
Autopsies limited to the brain and spinal cord were performed by an experienced neuropathologist (author JK). Neuropathological diagnoses and staging were made for the primary disease process and any co-existing neurodegenerative phenomena at the time of original autopsy based on a standardized blocking and staining protocol for dementia, applying consensus criteria for AD [15,16,17,18] and FTLD [19, 20]. Cases were included in the study and assigned into the following neuropathological diagnostic categories based on the original autopsy reports: (i) AD (n = 15), (ii) FTLD-Tau (Pick’s) (n = 10), (iii) FTLD-Tau (PSP) (n = 15), (iv) FTLD-Tau (CBD) (n = 11), and (v) FTLD-TDP (n = 22).
The pathologic classification of FTLD-TDP has evolved in recent years; thus, these cases were subjected to a central pathology review. Slides from the original autopsy were retrieved from the Sunnybrook pathology archive and reviewed by authors JK and AG. Additional slides were cut from selected original tissue blocks and stained with antibodies for TDP-43, alpha-synuclein, tau (AT8), and/or p62 (antibody information provided in the Additional file 1). Based on this central pathology review, cases within the FTLD-TDP category were re-classified using the harmonized consensus criteria for FTLD-TDP pathology as types A-D [21, 22].
Cases were excluded from the study if (i) neuropathological diagnosis could not be accurately assessed (n = 1), (ii) immunohistochemical staining examination was incomplete and could not be retrospectively completed (n = 2), and (iii) multiple co-morbid neuropathologies were present, none of which could be assigned unequivocally as the predominant cause of dementia (n = 25; 10 cases of FTLD had concomitant alpha-synucleinopathy, amyloid plaques, or infarcts, and 15 cases of AD had concomitant alpha-synucleinopathy, infarcts, or TDP43).
MRI acquisition and analysis
All participants underwent MRI on a 1.5 Tesla GE Signa (Milwaukee, WI, USA) system in compliance with consensus panel imaging recommendations for studies examining vascular cognitive impairment [23]. The following sequences were used for volumetric analysis: T1-weighted-axial three-dimensional (3D) Spoiled Gradient Recalled Echo (SPGR): 5 ms echo time (TE), 35 ms repetition time (TR), 1 number of excitations (NEX), 35o flip angle, 22 × 16.5cm (FOV), 0.859 × 0.859 mm in-plane resolution, 1.2 to 1.4 mm slice thickness depending on head-size, and a whole head interleaved proton density and T2 (interleaved axial dual-echo spin echo: TEs of 30 and 80 ms, 3 s TR, 0.5 NEX, 20 × 20 cm FOV, 0.781 × 0.781 mm in-plane resolution, 3-mm slice thickness with no gaps between slices).
MR images were analyzed with the semi-automatic brain region extraction (SABRE) and Lesion Explorer (LE) processing pipeline [24], which permits semi-automatized segmentation and parcellation procedures and to obtain regionalized and whole-brain volumetrics for normal appearing tissues and WMH. An automated 3D connectivity algorithm was applied to segment periventricular from deep WMH. Volumes for gray matter, normal appearing white matter, and WMH were obtained in 26 regions of interest, 13 per hemisphere (frontal: superior, middle, inferior, medial inferior, medial superior, medial middle; parietal: superior, inferior; occipital; temporal: anterior, posterior; basal ganglia/thalamus: anterior, posterior). Intracranial volumetric data (gray matter and WMH) were normalized for total intracranial volume (TIV). For analysis, all WMH values were log-transformed after normalizing for TIV due to their known skewed distribution [24].
Neuropsychological and neuropsychiatric assessments
Participants underwent a standardized clinical evaluation at baseline within 12 weeks of MRI acquisition. This comprised a medical history, physical examination, and a neuropsychological and neuropsychiatric battery [25]. The following vascular risk factors were collected: hypertension, hyperlipidemia, diabetes mellitus, and history of stroke and/or transient ischemic attack. For the purpose of this study, cognitive and neuropsychiatric testing results were retrieved for the following: (i) the Mini-Mental Status Examination (MMSE) [26], (ii) the Dementia Rating Scale (DRS) [27], and (iii) the Neuropsychiatric Inventory (NPI) [28]. For the latter, the total score (maximum of 144 points) and the 12 items, i.e., neuropsychiatric symptom subscores (maximum of 12 points for each item), were obtained, as well as the caregiver distress subscore (maximum of 60 points).
Image-guided neuropathology review of white matter regions with highest burden of WMH
For each neuropathologic group (FTLD-TDP, FTLD-Tau [PSP], FTLD-Tau [CBD], and FTLD-Tau [Pick’s], AD with CAA, and AD without CAA), antemortem T2-weighted MRI images were examined to determine the one case per group with the greatest volume of WMH. These cases subsequently underwent further pathologic evaluation of the affected white matter using annotated coronal MRI images as a guide (see Additional Figure 1, in Additional file 1). One periventricular white matter region that was heavily affected by WMH on MRI was identified within the previous coronally dissected formalin-fixed archived cadaveric brain tissue. For one case, no white matter region of interest could be found in the remaining brain tissue. For the other seven cases, this single region of interest was sampled. FFPE sections were cut at 6 microns and stained with H&E/LFB as well as immunohistochemistry for neurofilament, GFAP, CD68, Tau (AT8), and TDP43 (antibody information provided in Additional file 1). These slides were scanned at 40X on an Aperio ScanScope AT Digital Pathology Slide Scanner and examined digitally synchronously by two experienced neuropathologists (authors AG and JK) who were blinded to the neuropathologic diagnosis/group of each case. Correlating the H&E/LFB stained slide to the annotated coronal MRI image, the white matter region of interest was located within the sampled tissue. On H&E/LFB, pallor of myelin staining was assessed semi-quantitatively (0–3: none, mild, moderate, severe) using subcortical U-fibers as the internal control for none. Arteriolosclerosis, hemosiderin deposition, and collagenosis of the small and large caliber periventricular veins were determined to be present or absent. TDP43 and tau (AT8)-positive inclusions, axonal loss (neurofilament), gliosis (GFAP), and macrophage/microglial infiltration/activation (CD68) were assessed semi-quantitatively (0–3) on immunohistochemistry.
Statistical analyses
We compared baseline characteristics between each of the neuropathological groups and the healthy control group using ANOVA with post hoc Bonferroni tests for continuous, normally distributed variables, χ2/Fisher exact tests for categorical/dichotomous variables, respectively, and Kruskal-Wallis with post hoc Mann-Whitney U tests for non-normally distributed data. Differences in total and regional volumes of WMH on T2-weighted imaging (i.e., dependent variables) among neuropathological groups and the healthy control group were assessed by using ANCOVA, controlling for age at imaging and vascular risk factors.
We also assessed for the association between regional WMH volumes and corresponding regional gray matter volumes using multiple linear regression analyses, controlling for age, education, sex, and vascular risk factors, with regional WMH volume as the independent variable and the corresponding grey matter volume as the dependent variable.
We conducted linear regressions to assess for associations between global and regional WMH volumes and scores on the NPI (total scores and 12 subscale scores) across all neuropathological groups. For the linear regressions, a model was designed a priori and contained age, sex, vascular risk factors, and corresponding regional gray matter volumes as covariates, with WMH volume as the independent variable and NPI score as the dependent variable. Considering (1) the exploratory nature of our study, (2) that NPI subscores are highly correlated with each other, and (3) that regional WMH volumes are also highly correlated with each other, the Benjamini-Hochberg procedure was used to control for the false discovery rate across all regression analyses for each brain region, with a false discovery rate (FDR; Q) set at 0.10. A Bonferroni correction is too strict due to the non-independence of these variables as outlined above. Statistical analyses were performed with the Statistical Package for the Social Sciences, version 24.0.