Longitudinal neuroimaging and neuropsychological profiles of frontotemporal dementia with C9ORF72 expansions

Twenty cases from a previously published DNA cohort comprising 227 cases within the frontotemporal lobar degeneration spectrum [6] were found to harbour a C9ORF72 expansion using the repeat-primed PCR as previously published [4]. Mutations were called where more than 30 repeats were shown consistently. For the purpose of reporting longitudinal change only cases with a minimum of two neuropsychological assessments or volumetric MRI scans were included. In total 12 individuals (mean age 59.4 years (± 6.8 years), seven male) had longitudinal neuropsychological data and six of these cases (mean age 62.7 years (± 7 years), five male) also had longitudinal volumetric MRI scans. All cases identified had been assessed in the Specialist Cognitive Disorders Clinic at the National Hospital for Neurology and Neurosurgery by an experienced cognitive neurologist and all met current consensus criteria for a diagnosis of behavioural variant FTD [1]. Two cases had additional clinical features of motor neuron disease at presentation.

The study was approved by the local research ethics committee under Declaration of Helsinki guidelines and all subjects gave informed consent for participation.

The mean duration between serial neuropsychological assessments was 1.4 years (± 0.7 years). General intellectual function was assessed using the Wechsler Adult Intelligence Scale - Revised or the Wechsler Abbreviated Scale of Intelligence [11, 12]. Executive function was assessed using the Weigl test, the Stroop colour-word test or the Hayling test [13‐15]. Verbal memory and visual memory were assessed with the Recognition Memory Test for words and faces respectively [16]. Naming was assessed using the Graded Naming Test or the Oldfield Naming Test [17, 18]. Visuospatial and visual perception skills were assessed using subsets of the Visual Object and Spatial Perception battery [19]. Calculation and spelling were assessed with the Graded Difficulty Arithmetic test [20] and the Baxter Spelling tests [21] respectively. Raw scores were converted into percentiles for reporting.

Serial T₁-weighted magnetic resonance volumetric brain MRI was performed using a Magnetization Prepared Rapid Gradient Echo sequence: three studies were acquired on a 1.5T GE Signa scanner (General Electric Milwaukee, WI, USA) (256 × 256 matrix; 1.5 mm slice thickness) and three studies acquired on a 3.0T Siemens Trio scanner (Siemens, Germany) (256 × 256 matrix; 1.1 mm slice thickness). Patient data were compared with data from 15 age-matched (mean age 57.7 years; 10 male, five female) healthy controls with two volumetric MRI scans (12 controls on the 1.5T scanner, three controls on the 3.0T scanner). The mean duration between scans was 1.0 years (± 0.2 years) for patients and 1.6 years (± 0.8 years) for controls. All images were visually inspected for alternative pathologies and motion artefact. Whole brain, ventricular and cerebellar segmentation was performed by an experienced segmentor using a semi-automated technique using the MIDAS software package [22]. Scans underwent affine registration to spatially align the repeat scan to baseline. Rates of whole brain and cerebellar atrophy and ventricular expansion were calculated using the boundary shift integral (BSI), utilising the more robust KN-BSI methodology to provide automatic quantification of volume change[23]. Rates of change are expressed as the percentage loss from baseline volume and adjusted to an annualised rate according to the interval. Scans were registered into standard space for ventricular and hemispheric segmentation. Ventricular regions included the lateral ventricles and temporal horn of the lateral ventricles but excluded the third and fourth ventricles. Visualised cerebellar regions were dissected from the brainstem at mid-pontine level and images underwent further manual editing in coronal and sagittal planes to remove any remaining areas of brainstem. Right and left cerebral hemispheric volumes were calculated by dividing the brain along the mid-sagittal section. Finally, total intracranial volumes were calculated by summing grey matter, white matter and cerebrospinal fluid volumes acquired using the New Segment toolbox within Statistical Parametric Mapping 8 [24, 25].

Cortical and subcortical regional volumes were obtained from each subject's baseline and repeat volumetric MRI images using FreeSurfer (v5.1) running the automated longitudinal processing stream [26]. Default parameters were used with the exception of applying custom brain masks defined from the whole brain segmentation step to improve anatomical accuracy. Segmentations were visually inspected and edited where necessary. Volumes from 34 cortical regions following the Desikan atlas [27] and six subcortical regions (thalamus, caudate, putamen, globus pallidus, amygdala, hippocampus) were extracted for each subject and time point.

Following affine registration and bias correction of scan pairs, each scan set underwent cropping using subject-specific masks to exclude nonbrain regions. Fluid registration was performed to visualise intra-subject changes in brain morphology [28]. Briefly this involves nonlinear warping of each individual's repeat scan to match their baseline scan, generating a deformation field for each subject that allows visualisation of voxel-level expansion or contraction.

Group-level performance is displayed in Figure 1 and individual data in Table 1. At baseline the mean general intellectual function, as reflected in verbal IQ and performance IQ, was in the low average range (mean baseline verbal IQ = 83 (± 14); mean baseline performance IQ = 83 (± 15)); over the period of follow-up, both verbal IQ and performance IQ became impaired, declining by an average of 11 points (mean follow-up verbal IQ = 72 (± 19); mean follow-up performance IQ = 71 (± 23)).

Executive function was severely impaired in the majority of subjects at baseline (7/12 subjects scored < 5th percentile on at least one executive measure) and deficits became more frequent over the period of follow-up (10/12 < 5th percentile). Recognition memory was frequently weak at baseline, with deficits in verbal (7/12 < 5th percentile) and visual memory (8/12 < 5th percentile); over the period of follow-up, visual memory deficits became more frequent (10/12 < 5th percentile). Naming was impaired in one-half of the patients at baseline (6/12 < 5th percentile); at follow-up, naming deficits were evident in the majority (8/12 < 5th percentile).

Dominant parietal skills were assessed in only five patients longitudinally; however, three had evidence of dyscalculia and/or dysgraphia at baseline and four exhibited at least one of these deficits at follow-up. Visual perceptual functions remained largely stable over the period of follow-up, only one subject becoming impaired.

Individual and group volumetric data and rates of whole brain, hemispheric and ventricular change measured using BSI are displayed in Figure 2 and Table 2. Rates of whole brain atrophy varied widely between subjects; these data have been reported previously for five of the cases in the present series [6]. The most consistent finding (present in 5/6 cases) was an increased rate of ventricular enlargement in patients with a C9ORF72 mutation: patients had a mean annualised ventricular expansion rate of 3.2 (± 2.0) ml/year compared with controls at 0.7 ml/year (± 0.6) (P = 0.001), despite substantial individual variation. Longitudinal cerebellar atrophy was present in the majority of individual patients (4/6 cases); the mean annualised rate of cerebellar atrophy in the patients was also significantly higher (1.0%/year) than in controls (0.1%/year; P = 0.02). In addition, the mean annualised rate of whole brain atrophy in the patients as measured using KN-BSI (1.4%) was significantly higher than in controls (0.4%; P = 0.04) Mean atrophy rates for the cerebral hemispheres considered separately were similar for each hemisphere (left 2.4%/year; right 2.1%/year) and similar to the whole brain atrophy rate; atrophy in individual patients was symmetrical between the hemispheres across the C9ORF72 mutation group (inter-hemispheric volume ratio 0.99) and did not become more asymmetric over the follow-up interval.

Detailed data on subcortical volume change are displayed in Table 3. Compared with healthy controls, significant subcortical volume loss over time was detected in the C9ORF72 mutation group in the right thalamus (P = 0.006), left thalamus (P = 0.03) and left globus pallidus (P = 0.04). No significant change over time was detected in cortical regions when compared with controls.

Fluid-based nonrigid registrations in individual patients (Figure 3) revealed heterogeneous patterns of whole brain atrophy across subjects. Over the period of follow-up, most patients showed a diffuse but dorsally directed pattern of cerebral parenchymal loss, with more variable involvement of the temporal lobe regions; ventricular expansion and cerebellar volume loss were consistent features. A pattern of generalised progressive atrophy was apparent in Cases 4 to 6; Case 2 had prominent bifrontal volume loss, particularly implicating orbitofrontal cortices; and Cases 1 and 3 had more posterior atrophy, although expansion of the frontal horns of the lateral ventricles was also prominent in Case 1.