Frontotemporal dementia with the C9ORF72 hexanucleotide repeat expansion: clinical, neuroanatomical and neuropathological features.

An expanded hexanucleotide repeat in the C9ORF72 gene has recently been identified as a major cause of familial frontotemporal lobar degeneration and motor neuron disease, including cases previously identified as linked to chromosome 9. Here we present a detailed retrospective clinical, neuroimaging and histopathological analysis of a C9ORF72 mutation case series in relation to other forms of genetically determined frontotemporal lobar degeneration ascertained at a specialist centre. Eighteen probands (19 cases in total) were identified, representing 35% of frontotemporal lobar degeneration cases with identified mutations, 36% of cases with clinical evidence of motor neuron disease and 7% of the entire cohort. Thirty-three per cent of these C9ORF72 cases had no identified relevant family history. Families showed wide variation in clinical onset (43-68 years) and duration (1.7-22 years). The most common presenting syndrome (comprising a half of cases) was behavioural variant frontotemporal dementia, however, there was substantial clinical heterogeneity across the C9ORF72 mutation cohort. Sixty per cent of cases developed clinical features consistent with motor neuron disease during the period of follow-up. Anxiety and agitation and memory impairment were prominent features (between a half to two-thirds of cases), and dominant parietal dysfunction was also frequent. Affected individuals showed variable magnetic resonance imaging findings; however, relative to healthy controls, the group as a whole showed extensive thinning of frontal, temporal and parietal cortices, subcortical grey matter atrophy including thalamus and cerebellum and involvement of long intrahemispheric, commissural and corticospinal tracts. The neuroimaging profile of the C9ORF72 expansion was significantly more symmetrical than progranulin mutations with significantly less temporal lobe involvement than microtubule-associated protein tau mutations. Neuropathological examination in six cases with C9ORF72 mutation from the frontotemporal lobar degeneration series identified histomorphological features consistent with either type A or B TAR DNA-binding protein-43 deposition; however, p62-positive (in excess of TAR DNA-binding protein-43 positive) neuronal cytoplasmic inclusions in hippocampus and cerebellum were a consistent feature of these cases, in contrast to the similar frequency of p62 and TAR DNA-binding protein-43 deposition in 53 control cases with frontotemporal lobar degeneration-TAR DNA-binding protein. These findings corroborate the clinical importance of the C9ORF72 mutation in frontotemporal lobar degeneration, delineate phenotypic and neuropathological features that could help to guide genetic testing, and suggest hypotheses for elucidating the neurobiology of a culprit subcortical network.

Introduction

The frontotemporal lobar degenerations (FTLD) are a common cause of young onset neurodegeneration (Ratnavalli ). FTLD is clinically and pathologically heterogeneous (Lillo ; Seelaar ) and genetic factors play a significant role, with a genetic basis implicated in up to 40% of cases in series from specialist centres (Rohrer ; Rohrer and Warren, 2011). Discovery of disease-causing mutations in the progranulin (GRN) and microtubulin-associated protein tau (MAPT) genes left a substantial minority of cases of autosomal dominant FTLD without an identified genetic substrate; however, linkage to chromosome 9p21 has been described in a number of families with phenotypes of frontotemporal dementia (FTD), motor neuron disease (MND) or overlap syndrome (Morita ; Vance ; Valdmanis ; Luty ; Le Ber ; Gijselinck ; Boxer ; Pearson ). This elusive chromosome 9 gene has recently been identified as an expanded hexanucleotide (GGGGCC) repeat insertion in a non-coding promoter region of open reading frame 72 (C9ORF72) and is likely to account for a substantial proportion of the remaining genetic susceptibility to FTLD and MND) (Dejesus-Hernandez ; Gijselinck ; Renton ). Cases with the C9ORF72 repeat expansion with histopathological correlation have had TAR DNA-binding protein (TDP)-43 deposition, however, the pathogenic mechanism whereby the expansion leads to neurodegeneration has not been defined.

As has been demonstrated previously in the case of MAPT and GRN-associated FTLD (Rademakers ; Beck ; Le Ber ; Rohrer and Warren, 2011) detailed phenotypic characterization of disease-causing mutations has both clinical and neurobiological implications. Clinically, such a characterization will help guide genetic screening and thereby facilitate early, accurate diagnosis, and ultimately the possibility of identifying individuals both with established and incipient disease for clinical trials of disease-modifying therapies. Neurobiologically, phenotypic information may help elucidate the pathophysiology of the disease, for example, by delineating the interaction of the molecular lesion with affected neural networks (Fletcher and Warren, 2011; Rohrer ). Here we describe the clinical, neuropsychological, neuroimaging and neuropathological features of a large series of cases with the C9ORF72 hexanucleotide repeat expansion ascertained at a tertiary referral centre.

Subjects and methods

Case ascertainment

The University College London FTLD DNA cohort comprises 273 cases (256 probands) with a syndromic diagnosis within the FTLD spectrum (Litvan ; Boeve ; Gorno-Tempini ; Rascovsky ), including 122 patients with behavioural variant FTD, 11 patients with FTD and MND (FTD–MND), 53 patients with semantic dementia, 49 patients with progressive non-fluent aphasia, 18 patients with corticobasal syndrome, eight patients with a progressive non-fluent aphasia–corticobasal overlap syndrome, 11 patients with progressive supranuclear palsy syndrome and one patient with the inclusion body myopathy–Paget's disease–frontal dementia syndrome. DNA is also held on a further 11 patients with the logopenic aphasia variant of primary progressive aphasia (Gorno-Tempini ). The cohort includes 22 cases with MAPT mutations (15 probands), 24 cases with GRN mutations (17 probands) and one case with a valosin-containing protein (VCP) mutation (Table 1). Thus, 223 probands with FTLD syndromes and 11 individuals with logopenic aphasia lacking a previously identified pathogenic mutation were included in the present screen for the C9ORF72 repeat expansion. Prior to screening, all cases within the cohort were assigned a modified Goldman score as a measure of the extent of family history of the disease (Goldman ; Beck ; Rohrer ). A score of 1 represents an autosomal dominant family history of FTLD or MND; 2 is familial aggregation of three or more family members with dementia; 3 is one other first-degree relative with dementia of young onset (<65 years); 3.5 is one other first-degree relative with dementia of onset >65 years of age; and 4 is no or unknown family history. Of the probands with a diagnosis within the FTLD spectrum included in the screen, six had a Goldman score of 1; 26 had a score of 2; 12 had a score of 3; 22 had a score of 3.5; and 157 had a score of 4 (Table 1).

Table 1

Numerical data on mutation frequency, demographics and family history for cases in this series

					n (% total) per modified Goldman score
Group	n	Male (%)	AAO (SD)	Probands (% totals)	1 (%)	2 (%)	3 (%)	3.5 (%)	4 (%)
C9ORF72	19	58	55.1 (6.6)	18 (7.0)	4 (23.5)	6 (17.6)	1 (5.6)	1 (4.2)	6 (3.7)
MAPT	22	59	51.7 (6.9)	15 (5.9)	6 (35.3)	3 (8.8)	3 (16.7)	1 (4.2)	2 (1.2)
GRN	24	50	57.6 (6.5)	17 (6.6)	4 (23.5)	5 (14.7)	3 (16.7)	1 (4.2)	4 (2.5)
VCP	1	100	44.0	1 (0.4)	1 (5.9)	0	0	0	0
Others	207	57	59.0 (8.6)	205 (80.1)	2 (11.8)	20 (58.8)	11 (61.1)	21 (87.5)	151 (92.6)
Total	273			256	17	34	18	24	163

In summary, if the modified Goldman score is 1, 88% of patients have a known genetic abnormality, and this percentage decreases with a higher score: 41% if the score is 2, 39% if the score is 3, 13% if the score is 3.5 and 7% if the score is 4. AAO = age at onset.

Ethical approval for the study was obtained from the local institutional ethics committee and carried out in accordance with the Declaration of Helsinki.

Genetic analyses

All samples with genetically unexplained FTLD based on previous screens were assessed for C9ORF72 expansions using the repeat-primed polymerase chain reaction as previously published (Renton ). Mutations were called if repeats of this assay consistently showed more than 30 repeats in total (Fig. 1).

Figure 1

Examples of electrophoresed products of repeat-primed polymerase chain reaction used to detect expansions in C9ORF72. Each panel was scaled to 800 fluorescent units on the y-axis and between 300–650 bp on the x-axis. (A) Data from a FTLD patient sample without an expansion and (B) data from a patient with FTLD with 2 and >60 repeats.

Clinical and neuropsychological analyses

A retrospective review of clinical case notes and neuropsychological reports (where available) was undertaken in those cases identified as having the C9ORF72 repeat expansion. Demographic features, family history and disease duration were recorded. Earliest clinical features were extracted by an experienced neurologist, and neuropsychological features at first assessment were also recorded; 13 participants had undergone more than one clinical assessment, and clinical features at the final assessment were also recorded. Demographic data for C9ORF72 cases were compared statistically with MAPT and GRN mutation groups using two-tailed t-tests within STATA 10 (Statacorp).

In addition, we recorded results of laboratory investigations undertaken as part of the clinical evaluation of C9ORF72 cases: electrophysiological assessments and CSF examination [including Aβ1–42 level (Innotest β-amyloid1–42, Innogenetics), total tau (Innotest hTau Ag, Innogenetics) and S100 beta (Abnova)].

Neuroimaging analyses

A majority of patients included in the University College London FTLD cohort have had at least one T1-weighted magnetic resonance brain volume acquired on either a 1.5T GE Signa scanner (General Electric) (256 × 256 matrix; 1.5 cm slice thickness) or a 3.0T Siemens Trio scanner (Siemens) (256 × 256 matrix; 1.1 cm slice thickness). Where available, volumetric brain MRI data for the C9ORF72 mutation cases were analysed using previously described semi-automated techniques and compared with the GRN and MAPT mutation groups.

Whole-brain segmentation yielded a whole-brain region separated from CSF, dura and skull (Freeborough ). Scans were then transformed into standard space by registration to the Montreal Neurological Institute template and hemispheric volumes were calculated for each individual. In order to assess hemispheric asymmetry, a left/right asymmetry ratio was derived in each case by dividing the left hemispheric volume by the right hemispheric volume, as previously described (Rohrer ). Repeat scans were transformed into standard space and underwent an affine registration (12 degrees of freedom) onto the baseline scan. Volume change was calculated directly from registered scans using the brain boundary shift integral (Freeborough and Fox, 1997). Brain boundary shift integral-derived whole-brain volume changes were expressed as an annualized volume change in ml per year. Volumetric data from a previously published cohort (Rohrer ) of GRN [n = 4, mean interval 1.3 years, standard deviation (SD) 0.2] and MAPT (n = 6, 2.1 years, SD 1.0) mutation carriers (all values derived using the same methods) were compared against C9ORF72 mutation carriers using two-sample t-tests with unequal variance within STATA 10 (Statacorp).

In order to determine patterns of brain atrophy at group level, volumetric MRI data from healthy controls and the C9ORF72, MAPT and GRN mutation groups were compared using voxel-based morphometry with the DARTEL toolbox of SPM8 (http://www.fil.ion.ucl.ac.uk/spm) running under Matlab 7.0 (Mathworks). Segmentation, modulation and smoothing of grey and white matter images were performed using default parameter settings. Final images were normalized to standard space. In order to reduce the effects of atrophy, analysis was performed over voxels within a consensus mask, which includes all voxels with intensity of >0.1 in 70% of subjects (Ridgway ). A multiple regression analysis was performed with voxel intensity modelled as a function of group membership with age, total intracranial volume, gender and scanner field strength included as nuisance covariates. Results were then overlaid on the MNI152 template brain for anatomical analysis. Total intracranial volume was calculated by summing together grey matter, white matter and CSF segmentations.

Cortical thickness measurements were made using FreeSurfer v5.1 (http://surfer.nmr.mgh.harvard.edu/) following previously described methods (Dale ; Fischl and Dale, 2000). A surface-based Gaussian smoothing kernel of 20 mm full width at half-maximum was applied to reduce local variations in cortical thickness. Skull stripping used a locally generated brain mask, and FreeSurfer ventricular segmentations were added to the white matter mask to improve cortical segmentation. Between-group variations in cortical thickness were assessed using a vertex-by-vertex general linear model performed with SurfStat (http://www.math.mcgill.ca/keith/surfstat/). Regional differences in grey matter between control and disease groups were assessed using the same covariates as the voxel-based morphometry analysis. A significance threshold of P < 0.05 with false discovery rate correction was applied. Where this threshold was not met, percentage difference maps were used to illustrate trends in atrophy patterns.

Alterations of white matter tract integrity were assessed with diffusion tensor imaging using an unbiased group-wise analysis technique. Three individuals underwent diffusion tensor imaging scanning on the same 3T scanner (64 direction single shot, spin-echo-echo planar imaging sequence, 55 contiguous slices). Tensor eigenvalues (λ1 = axial diffusivity, λ2 and λ3), fractional anisotropy and radial diffusivity (RD = λ2 + λ3) were extracted at each voxel using CAMINO (Cook ). Tensor fitted images were then imported for voxel-wise analysis within tract-based spatial statistic (TBSS v1.1) (Smith ). Voxel intensity within tracts was modelled on disease group membership with the age, gender and total intracranial volume included as nuisance covariates. Statistical analysis was implemented using the permutation-based (non-parametric) randomize tool within FSL.

For all group-wise analyses, a significance threshold (P < 0.05) was applied following correction for multiple comparisons. Threshold-free cluster enhancement modulation was applied to diffusion tensor imaging results (Smith and Nichols, 2009).

Neuropathological protocol

Cases in the University College London FTLD cohort coming to post-mortem have undergone detailed histopathological assessment as previously described (Lashley ; Rohrer ). At the time of brain donation, brains were fixed in 10% buffered formaldehyde and sampled using standard protocols. Histological sections of hippocampus with parahippocampal and fusiform gyri and cerebellum, stained with antibodies to p62 (1:200, BD Transduction Laboratories) and TDP-43 (amino acids 1-261, Abnova, 1:800), were analysed. Selection of these anatomical regions was based on preliminary observations indicating that the number of p62-positive inclusions may exceed TDP-43-positive neuronal cytoplasmic inclusions in C9ORF72 mutation cases; using a semi-quantitative approach, the frequency of p62-positive neuronal cytoplasmic inclusions was compared with TDP-43-positive inclusions in the granule cell layer of the dentate fascia, hippocampal subregions and cerebellar granule cell layer.

Results

Genetics

Of the 223 cases with FTLD screened, 18 probands (8% of probands without previously identified mutations) were found to have the C9ORF72 repeat expansion. One further case in the cohort, from the same family as one of the probands, also had the C9ORF72 expansion. No mutations were identified in the 11 screened cases with logopenic aphasia. Alternatively expressed, 18/51 (35%) of all FTLD probands with identified mutations or 18/256 (7%) of all probands in our FTLD series had a C9ORF72 mutation. Of the 18 C9ORF72 probands, four patients had a modified Goldman score of 1, six had a score of 2, one had a score of 3, one had a score of 3.5 and six had a score of 4 (Fig. 2 and Table 1), indicating that in this series 6/163 (4%) of apparently sporadic cases of FTLD had a C9ORF72 mutation. If an individual had a Goldman score of 1 assigned, a genetic cause (C9ORF72, MAPT, GRN or VCP mutation) was identified in 88% of all FTLD probands, of which 35% were MAPT, 24% were C9ORF72, 24% were GRN and 6% were VCP mutations. If a score of 2 was assigned, a genetic cause was identified in 41% of cases; if a score of 3 was assigned, a genetic cause was identified in 39%; if a score of 3.5 was identified, a genetic cause was identified in 13% and if a score of 4 was assigned, a genetic cause was identified in 7% of cases. Representative pedigrees of C9ORF72 families are shown in Fig. 3. The pedigree data indicate that age at symptom onset and clinical syndrome are highly variable within a family.

Figure 2

FTLD DNA cohort by modified Goldman score. Each bar displays % of cohort by assigned Goldman score and mutation status.

Figure 3

Representative pedigrees from the Dementia Research Centre (DRC) cohort indicating inheritance, predominant clinical syndrome and, where known, age at onset (RIP indicates only age at death known). bvFTD = behavioural variant FTD.

Clinical and neuropsychological features

The average age at onset across the series was 55.0 years (SD 6.6, range 43–68), which did not differ significantly from MAPT or GRN mutation groups. Mean disease duration from symptom onset until death (n = 12) or time of last review (still alive n = 5) was 8.7 years (SD 4.9, range 1.7–22), which was not significantly different to either the MAPT mutation group (11.4 SD 5.3 years) or the GRN mutation group (6.8 SD 2.3 years). As a proportion of the entire FTLD cohort, 11% of cases with behavioural variant FTD, 36% with FTD–MND and 2% with progressive non-fluent aphasia were found to have the C9ORF72 repeat expansion.

Detailed clinical information was available on 16 of the 19 cases with the C9ORF72 expansion who had been assessed in the specialist cognitive disorders clinic at the National Hospital for Neurology and Neurosurgery. Fifteen patients had neuropsychological assessments. Group data for these cases are summarized in Fig. 4 and individual data in Table 2. The mean duration of follow-up was 2.3 (SD 1.4) years. At initial presentation, eight of these 16 patients (50% of cases) had a clinical syndromic diagnosis of behavioural variant FTD; three (19%) had a diagnosis of FTD–MND; two (13%) had a diagnosis of probable Alzheimer's disease; one had a diagnosis of progressive non-fluent aphasia, one an encephalitic illness and one atypical Parkinsonism. The final syndromic diagnoses were behavioural variant FTD in 12 cases, FTD–MND in three cases and progressive non-fluent aphasia in one case. For the remaining three cases without detailed clinical information, the final diagnoses were behavioural variant FTD in two cases and FTD–MND in one case.

Figure 4

Prevalence of symptoms and neurological signs (expressed as proportion of disease cohort) observed in C9ORF72 cases. Data for first and final assessments are coded separately.

Table 2

Individual clinical and neuropsychological data available for C9ORF72 cases

Case	Final diagnosis	AAO (years)	Duration (years)	Hand	Early symptoms	Neuropsychology profile
						VIQ	PIQ	RMTWa	RMTFa	GNTa	Calca	Percepa	Exec	Praxis
1	bvFTD	51	6.0	R	Anxiety, disinhibition, episodic memory impairment, speech adynamic	80	89	<5	<1	50–75	<1	>5	++	Norm
2	bvFTD	60	7.0b	R	Anxiety, delusions, disinhibition, episodic memory impairment, hyperorality	100	113	5–10	<5	95	<50	>5	++	Norm
3	bvFTD	56	8.0b		Anxiety, empathy loss, obsessionality, prosopagnosia	82	77	<5	10–25	<5	5–10	>5	++	Imp
4	FTD–MND	55	7.7		Disinhibition, obsessionality, speech effortful, weakness upper limbs									Norm
5	bvFTD	45	8.4	L	Anxiety, apathy, depression, involuntary movements	71	75	<5		<5	<5	>5	++
6	bvFTD	55	7.1		Disorganization, episodic memory impairment, obsessionality	87	78	<5	<1	<5	5–10	>5	++	Imp
7	bvFTD	55	16.6		Disorganization, empathy loss, episodic memory impairment, prosopagnosia	103	95	50–75	<1	75	25–50	>5	++
8	bvFTD	43	1.7	R	Anxiety, apathy, disinhibition, prosopagnosia, topographical memory impairment	UT	UT	UT	UT	5	<5	>5	+	Imp
9	FTD–MND	64	5.4	R	Apathy, delusions, prosopagnosia, speech effortful, topographical memory impairment	84	82	5–10	<1	<5	<5	>5	++	Norm
10	FTD–MND	54	5.0	R	Depression, disinhibition, speech effortful, weakness limbs	69	80	5–10	1–5	10–25	<5	>5	+	Imp
11	bvFTD	58	8.0b	R	Apathy, disinhibition, hyperorality, topographical memory impairment	85	69	10–25	50–75	10–25	<50	>5	+++	Imp
12	bvFTD	61	5.7	R	Episodic memory impairment, hyperorality, falls, somatic symptoms, weakness limbs	78	76	<1	<1	<5	10–25	>5	+++	Imp
13	bvFTD	68	7.1	R	Apathy, episodic and topographical memory impairment, speech adynamic	73	78	<5	<5	10–25	10–25	>5	++	Imp
14	bvFTD	53	14.8		Apathy, disinhibition, episodic memory impairment, prosopagnosia	UT	UT	<5	<5	<1	UT	>5	+	Imp
15	PNFA	55	5.0b	R	Comprehension impairment, disinhibition, speech effortful	55	70	25–50	<1	<5	<1	>5	+++	Norm
16	bvFTD	46	22.0b	R	Anxiety, disorganization, episodic memory impairment, obsessionality, prosopagnosia	116	113	75–95	50–75	75–90	>90	>5	++	Norm

Blank cells indicate testing not performed.

a Percentile scores are shown.

b Current duration for living cases.

+/++/+++ = defective performance on at least one/two/three tests, respectively, of executive function (Weigl sorting test; Wisconsin Card Sorting Test; Stroop colour interference; Verbal Fluency).

AAO = age at onset; bvFTD = behavioural variant FTD; Calc = calculation; Dur = total/current disease duration; Exec = executive functions; GNT = Graded Naming Test; Hand = handedness data available; Imp = impaired; Norm = normal; Percep = perception; PIQ = performance IQ; PNFA = progressive non-fluent aphasia; RMTF/W = recognition memory test for faces/words; UT = untestable; VIQ = verbal IQ.

The most common symptoms at the time of first presentation (Fig. 4) were disinhibition (63%) and apathy (49%); however, forgetfulness (44%) and anxiety (38%) were also prominent. Based on the history obtained from patients’ carers, the most common initial symptom was anxiety (33%), followed by disinhibition (13%), effortful speech (13%) and impaired topographical memory (13%). At the time of final assessment, the most frequent symptoms were disinhibition (76%), apathy/adynamia (63%), and anxiety (52%). The most common neuropsychological features (Table 2) were executive dysfunction (100%), impaired verbal and/or visual episodic memory (87%), apraxia (57%), anomia (47%) and dyscalculia (38%). Object and space perception were unimpaired in all patients assessed. The most common physical neurological sign at presentation was brisk deep tendon reflexes (40%); fewer than 20% of patients had features of bulbar or lower motor neuron involvement. At the time of final assessment (or first if only one assessment), the most frequent neurological signs were brisk tendon reflexes (60%), spasticity (47%), fasciculations (33%) and primitive reflexes (33%).

Four patients (two with an initial clinical diagnosis of FTD–MND) underwent electrophysiological studies; one individual had EMG evidence of MND, one had borderline features and two had no features of MND. Five patients underwent CSF examination. Basic CSF constituents including protein, cells and glucose were within normal limits for all patients examined; further analyses comprised Aβ1–42 (mean = 465 pg/ml, SD 303, in-house normal range = 220–2000; all within the normal range), total tau (mean = 190 pg/ml, SD 103, in-house normal range = 0–32; 80% of cases within the normal range) and S100 beta (mean = 0.33 ng/ml, SD 0.1, in-house normal range = 0.05–0.61; all within the normal range). The mean tau-to-beta-amyloid ratio was 0.5 (SD 0.3) and also within the normal limits for the assaying laboratory for all subjects.

Neuroimaging features

Volumetric data

Eleven patients with the C9ORF72 repeat expansion (mean age 58.5, SD 8.3 years) underwent volumetric brain MRI: eight at 1.5T and three at 3.0T. Five patients had repeat volumetric MRI on the same scanner (mean interval 1.2 years, SD 0.6), four at 1.5T and one at 3T. Inspection of individual brain magnetic resonance scans revealed highly variable profiles of frontal, temporal and parietal lobe atrophy; representative sections are shown in Fig. 5.

Figure 5

Representative T1-weighted coronal magnetic resonance brain images for cases with C9ORF72 mutation. Numbering of cases is as per Table 2. Patterns of atrophy were highly variable across the group, and no clear consistent profile of atrophy was evident visually. The distribution of atrophy was not prominently asymmetric between the cerebral hemispheres in most cases. Several cases had predominantly frontal lobe atrophy (Cases 1, 2 and 13), while temporal lobe atrophy was prominent in other cases (Cases 6 and 9). Other cases showed more generalized involvement with more prominent central atrophy (Cases 8 and 15) or minimal cross-sectional atrophy to visual inspection (Case 16).

For the purposes of the group-wise analysis, neuroimaging data from the C9ORF72 mutation group were compared with a group of 15 age- and gender-matched healthy control subjects (mean age 58.6, SD 5.7 years); brain images were acquired for 12 controls on the same 1.5T magnetic resonance scanner and for three controls on the same 3T scanner as the patients, and all group-wise analyses were adjusted for variation in field strength. Mean brain volume corrected for total intracranial volume was significantly lower in the C9ORF72 mutation group (Fig. 6) (mean brain/total intracranial volume ratio = 0.68, SD 0.07; P = 0.001) than the control group (mean brain/total intracranial volume = 0.77, SD 0.03); mean brain volume did not differ significantly between the C9ORF72 mutation group and the MAPT or PGRN mutation groups. Mean annualized atrophy rate was greatest in the GRN mutation group (Fig. 6) (brain boundary shift integral change 3.4%/year, SD 0.6) followed by the C9ORF72 mutation group (brain boundary shift integral change 1.6%/year, SD 1.6) and the MAPT mutation group (brain boundary shift integral change 1.4%, SD 0.6), but mean atrophy rates did not differ significantly between groups. As indexed using the hemispheric asymmetry ratio, cerebral atrophy was relatively symmetric in the C9ORF72 mutation group (Fig. 6) (hemispheric asymmetry ratio range 0.96–1.04) and the MAPT mutation group (range 0.97–1.05), similar to the healthy control group (range 0.98–1.02); whereas cerebral atrophy was strongly asymmetric in the GRN mutation group (range 0.86–1.33).

Figure 6

Individual brain volumes corrected for total intracranial volume (TIV) by group (top). Annualized percentage change in brain boundary shift integral (BBSI) by group (middle). Hemispheric asymmetry expressed as the ratio of left/right hemisphere volumes by subject and group (bottom).

Unbiased analyses

In the voxel-based morphometry analysis, relative to the healthy control group (n = 15), the C9ORF72 mutation group (n = 11) showed a regional profile of grey matter atrophy including thalamus bilaterally, left opercular cortex, left orbitofrontal cortex and bilateral posterior cerebellum (all false discovery rate corrected P < 0.05; Fig. 7). Direct comparisons between the C9ORF72 mutation group and the MAPT mutation group (n = 11) (Fig. 8) showed the MAPT group had decreased grey matter bilaterally in anterior temporal lobe, amygdala, hippocampus and parahippocampal gyrus (all false discovery rate corrected P < 0.05), while the C9ORF72 group had decreased grey matter in prefrontal cortex and cerebellar vermis (uncorrected P < 0.001). Direct comparisons between the C9ORF72 mutation group and the GRN mutation group (n = 8) (Fig. 8) showed that the GRN group had decreased grey matter asymmetrically in left inferior and middle temporal gyri, post-central gyrus and inferior frontal gyrus, whereas the C9ORF72 group had decreased grey matter in right prefrontal cortex (all uncorrected P < 0.001).

Figure 7

(A) Profiles of grey matter (GM) atrophy (n = 11) and increased radial diffusivity (RD) maps (n = 3) in C9ORF72 cases versus healthy controls. MNI co-ordinates are displayed. (B) Mean difference in cortical thickness expressed as a percentage of the mean thickness in C9ORF72 cases (n = 11) versus controls. A = anterior; FDR = false discovery rate; FWE = family-wise error; L = left; P = posterior.

Figure 8

Top: Voxel-based morphometry (VBM) profiles of grey matter (GM) atrophy by genetic subtype; P < 0.001. MNI co-ordinates are displayed. Middle: regional reduction of cortical thickness in MAPT mutation carriers compared with C9ORF72 mutation carriers. Colour scale for statistical difference represents FDR-corrected P < 0.05. Bottom: regional variation of cortical thickness in GRN mutation carriers compared with C9ORF72 mutation carriers. Colour bar for per cent difference represents magnitude of cortical thickness group difference expressed as a percentage of the mean group thickness. Red and yellow (positive values) represent lower cortical thickness in the GRN group, whereas dark to light blue (negative values) represent lower cortical thickness in the C9ORF72 group. A = anterior; FDR = false discovery rate; L = left; P = posterior; Ucorr = uncorrected.

In the cortical thickness analyses, relative to healthy controls, the C9ORF72 mutation group showed a symmetrical pattern of reduced thickness of frontal, anterior and superior temporal and parietal cortices (Fig. 7). Direct comparisons between the C9ORF72 and MAPT mutation groups showed the MAPT group had reduced thickness of anterior temporal cortex bilaterally (false discovery rate corrected P < 0.05) (Fig. 8). Direct comparisons between the C9ORF72 and GRN mutation groups showed the GRN group had asymmetrically reduced thickness of left inferior frontal, temporal and parietal cortices, whereas the C9ORF72 group had reduced thickness of right dorsal frontal, temporal and parietal cortices (Fig. 8).

In the diffusion tensor imaging analysis (Fig. 7), relative to healthy controls, the cases with C9ORF72 mutation showed increased radial diffusivity and decreased fractional anisotropy bilaterally in anterior thalamic radiations, uncinate fasciculus, anterior cingulum and anterior corpus callosum; and in right posterior corpus callosum, posterior inferior longitudinal fasciculus and superior longitudinal fasciculus (all family-wise error corrected P < 0.05). In addition, the C9ORF72 cases showed increased radial diffusivity in left frontal superior longitudinal fasciculus and posterior corpus callosum, and in the right corticospinal tract.

Histopathological features

Genetic studies of the neuropathological FTLD cohort identified six cases with the C9ORF72 repeat expansion. These cases were assessed in relation to a group of 53 ‘control’ cases with FTLD-TDP [four cases with type B TDP-43 pathology (Mackenzie ), 11 type A cases with a GRN mutation, 17 type A cases without a GRN mutation and 21 type C cases]. TDP-43 pathology was evident in the hippocampus in all cases with FTLD-TDP irrespective of the presence of the C9ORF72 repeat expansion. Four of the six cases with C9ORF72 mutation were previously classified as FTLD-TDP type A while two cases had type B pathology. All six cases contained sparse to moderate numbers of compact and/or ‘star-shaped’ neuronal cytoplasmic inclusions in the granule cell layer of the hippocampal dentate gyrus. Sparse TDP-43-positive neuronal cytoplasmic inclusions were also found in the CA4 to CA1 hippocampal subregions with two cases also containing an occasional neuronal intranuclear inclusion. Variable numbers of TDP-43-positive neurites were also seen. No TDP-43 pathology was observed in the cerebellum of either the cases with C9ORF72 mutation or FTLD-TDP controls.

All six cases with C9ORF72 mutation contained p62-positive small, ‘star-shaped’ neuronal cytoplasmic inclusions in the granule cells of the hippocampal dentate gyrus. The numbers of such inclusions ranged from moderate to frequent where virtually all neurons contained neuronal cytoplasmic inclusions (Fig. 9) and similar p62-positive neuronal cytoplasmic inclusions were also found in variable numbers throughout the hippocampus from the CA4 to the CA1 subregions. P62 immunohistochemistry also highlighted neuritic structures. In all cases with C9ORF72 mutation, p62-positive inclusions were strikingly more frequent than TDP-43-positive inclusions (Fig. 9A and B). In the control FTLD-TDP cases, the number of the p62-positive neuronal cytoplasmic inclusions in the hippocampal neurons including the granule cells of the dentate gyrus mirrored the frequency of TDP-43 immunoreactive neuronal cytoplasmic inclusions.

Figure 9

P62 and TDP-43 immunohistochemistry in FTLD-TDP with the C9ORF72 hexanucleotide repeat expansion. Frequent p62-positive (A) and sparse TDP-43-positive (B, arrow) neuronal cytoplasmic inclusions were present in the granule cells of the dentate gyrus in a case with the C9ORF72 repeat expansion. Numerous p62-positive neuronal cytoplasmic inclusions were also present in the cerebellar granule cell (C), which were negative for TDP-43. Scale bars: A = 50 µm; B = 30 µm; C = 80 µm.

Small, p62-positive, but TDP-43-negative neuronal cytoplasmic inclusions were found in the cytoplasm of the cerebellar granule cells in all cases with C9ORF72 mutation, but not in the control cases. The severity of the cerebellar pathology ranged from rare p62-positive neuronal cytoplasmic inclusions to severe involvement when nearly all granule cells appeared to contain an inclusion. It is of note that cases with high numbers of the p62-positive hippocampal inclusions also tended to have frequent p62-positive inclusions in the cerebellar granule cells.

Discussion

Here we present a detailed clinical, neuroanatomical and neuropathological analysis of a series of 19 cases with FTLD and the C9ORF72 hexanucleotide repeat expansion representing 18 families in the UK. We confirm initial reports (Dejesus-Hernandez ; Gijselinck ; Renton ) that this mutation is an important cause of FTLD and FTLD–MND overlap syndromes, representing ∼7% of our entire FTLD series, 8% of FTLD not otherwise accounted for by previously identified mutations, 35% of mutation-associated FTLD, 36% of FTLD with clinical evidence of MND and 4% of apparently sporadic FTLD. Identification of the C9ORF72 mutation has allowed attribution of a specific genetic cause in 88% of all cases of autosomal dominant FTLD in this series. However, some 33% of C9ORF72 mutation cases in this series had no known family history of FTLD or MND. In line with previous evidence in chromosome 9-linked FTLD and FTD–MND, age at onset (43–68 years), clinical duration (1.6 to >22 years) and clinical features were highly variable between and within families. The most common presenting syndrome was behavioural variant FTD, though around half of cases had an alternative syndrome at presentation, and over the course of follow-up ∼60% of cases manifested at least some clinical features of MND. This substantial clinical heterogeneity is in line with previous reports describing chromosome 9-linked FTLD (Momeni ; Morita ; Vance ; Valdmanis ; Luty ; Le Ber ; Gijselinck ; Boxer ; Pearson ), and was reflected in a range of profiles of atrophy distribution and severity on neuroimaging of individual patients here. However, group-wise analyses comparing the C9ORF72 mutation cases with healthy controls and with other mutation groups suggested certain common features that may constitute a core phenotype of this mutation.

Clinically, anxiety and agitated behaviour were early and prominent in a substantial proportion of our C9ORF72 mutation cohort. These features were of sufficient significance here to lead to early contact with psychiatric services in several cases and a presumptive diagnosis of encephalitis in one case; it is possible that similar but less severe symptoms were exhibited by other cases but were de-emphasized in the context of other behavioural and cognitive deficits (a limitation of retrospective analyses). Neuropsychologically, alongside executive dysfunction, deficits of episodic memory were salient. Indeed, memory impairment dominated the clinical presentation in several cases, leading to an initial clinical diagnosis of Alzheimer's disease in two patients. Several previous detailed phenomenological and neuropsychological studies of FTLD have included patients presenting with prominent neuropsychiatric features (Omar ; Lillo ; Loy ) or episodic memory impairment (Hornberger ), raising the possibility that apparently sporadic cases of behavioural variant FTD sharing such phenotypic features may carry the C9ORF72 repeat expansion. Dominant parietal lobe deficits (acalculia, apraxia) were frequent in the present C9ORF72 mutation cohort; it is of interest that clinical parietal lobe involvement was previously also found to be a common feature of GRN mutations (Rohrer ). Previous clinical series of chromosome 9-associated FTLD and FTD–MND have reported similar findings, with initial neuropsychiatric contact in a number of cases and clinical diagnoses including Alzheimer's disease and corticobasal degeneration (Morita ; Luty ; Lillo ; Boxer ; Murray ; Pearson ).

The group-wise neuroanatomical analyses here suggest candidate brain substrates for these clinical features. The C9ORF72 mutation group showed extensive cortical loss involving frontal, temporal and parietal lobes that was relatively symmetric between cerebral hemispheres; while the voxel-based morphometry analysis showed grey matter loss in thalamus and cerebellum, and tractography demonstrated associated damage involving long subcortical projection pathways in the superior and inferior longitudinal fasciculi, cingulum, thalamic radiations and descending corticospinal tract, as well as the corpus callosum. The atrophy profile of C9ORF72 mutation cases contrasts both with the relatively localized temporal lobe involvement associated with MAPT mutation cases and the strongly asymmetric profile in GRN cases. We speculate that the relatively sparse cortical correlates shown in the voxel-based morphometry analysis of the C9ORF72 mutation cohort here may reflect wide inter-subject variation in cortical loss, with the more consistent involvement of subcortical grey matter structures (including thalamus and cerebellum) being a more robust signature of this mutation. We do not argue that the neuroanatomical signature of the C9ORF72 repeat expansion is restricted to these regions; subcortical structures typically act as hubs within structural and functional networks (Van der Werf ; Schmahmann and Pandya, 2008; Tedesco ), with projection targets that include other sites of pathological involvement (for example, the thalamus participates in a diencephalic network that also includes the hippocampi). The diffuse projections of these subcortical structures may account for the wide range of cognitive findings and cortical damage exhibited across the C9ORF72 mutation cohort here. Involvement of thalamus and cerebellar connections could underpin the prominent neuropsychiatric features shown by these cases (Andreasen ; Schmahmann, 2005; Goossens ; Straube ; Schmahmann and Pandya, 2008; Tedesco ), as well as episodic memory deficits (Van der Werf ; de Jong ) somatic complaints, hallucinations and delusions (Kessler ; Tedesco ). Although none of our cases had frank cerebellar ataxia, this has been described in previous cases of chromosome 9 linked FTLD (Pearson ).

Neuropathological assessment of six cases with the C9ORF72 repeat expansion demonstrated a heterogeneous profile, in that both FTLD-TDP type A and type B morphological patterns were represented. Similar histomorphological heterogeneity has been emphasized in previous neuropathological analyses of C9ORF72 mutation cases (Murray ). In addition, all our C9ORF72 mutation cases had p62-positive neuronal cytoplasmic inclusions prominently in excess of TDP-43-positive neuronal cytoplasmic inclusions in the granule cells of the hippocampal dentate gyrus and other hippocampal subregions as well as p62-positive, TDP-43-negative inclusions in cerebellar granule cells. Our findings are supported by a recent publication that proposes this pattern of proteinopathy as pathognomonic for the C9ORF72 hexanucleatide repeat insertion (Al-Sarraj ). The ubiquitin-binding protein 62 (p62) or sequestosome 1 (like ubiquitin itself) is a common component of neuronal and glial proteinaceous inclusions in a range of neurodegenerative diseases including FTLD-TDP. It has been previously shown in a proportion of cases with FTLD-TDP that p62 (and, to a lesser extent, ubiquitin) immunohistochemistry may highlight a larger number of neuronal cytoplasmic inclusions in the dentate granule cells than are demonstrated on TDP-43 immunohistochemistry and that small dot-like or star-shaped p62-positve, TDP-43-negative inclusions may be present in the granule cells of the cerebellar cortex (Pikkarainen ; King ). Furthermore, ubiquitin-positive neuronal cytoplasmic inclusions in cerebellar granule cells have also been documented in FTD–MND linked to chromosome 9 due to the C9ORF72 repeat expansion (Dejesus-Hernandez ). The present neuropathological data support the notion that ‘excess’ p62 pathology may be a characteristic feature of the C9ORF72 repeat expansion. It is noteworthy that the sites of heavy p62 involvement identified here, including hippocampi and cerebellum, are likely to be pathophysiologically relevant based on clinical and neuroimaging findings in the present cohort. Moreover, these findings corroborate previous neuropathological evidence in other cohorts. However, the origin and true significance of the excess p62 labelling identified here require further investigation, including quantitative analysis of p62 deposition within and between brain regions. In addition, the present series comprised exclusively cases presenting with FTLD; there may be a stratification of histopathological findings across the broader phenotypic spectrum of the C9ORF72 mutations, which also includes cases presenting with MND (Al-Sarraj ; Murray ).

In summary, our findings suggest that clinicians should consider the C9ORF72 repeat expansion in patients with FTLD, particularly with clinical evidence (or a family history) of MND or with prominent anxiety or memory impairment. From a neurobiological perspective, our data suggest that involvement of a subcortical network may underpin the wide-ranging phenotypic features of the C9ORF72 repeat expansion. Recent clinico-anatomico-pathological evidence in FTLD suggests that phenotypes arise from an interaction of molecular and network-specific factors (Schmahmann and Pandya, 2008; Rohrer ). This hypothesis could be tested specifically in C9ORF72 mutation cases by targeting potentially relevant cognitive operations and network anatomy (Tedesco ). Neuropathologically, excess p62 deposition within the culprit network may be a signature of the C9ORF72 mutation. Future work should analyse these cases prospectively and in relation to other mutation groups, with joint longitudinal behavioural and neuroimaging (including connectivity-based) methods, and detailed genotypic stratification and histopathological correlation.

Funding

This work was undertaken at UCLH/UCL who received a proportion of funding from the Department of Health's NIHR Biomedical Research Centres funding scheme. The Dementia Research Centre is an Alzheimer's Research UK Co-ordinating Centre. This work was also funded by the Medical Research Council UK and by the Wellcome Trust. C.J.M. is supported by an MRC programme grant. N.C.F. is a NIHR senior investigator and MRC Senior Clinical Fellow. J.D.W. is supported by a Wellcome Trust Senior Clinical Fellowship (091673/Z/10/Z).