Synaptic density in carriers of C9orf72 mutations: a [11C]UCB‐J PET study

Abstract Synaptic loss is an early and clinically relevant feature of many neurodegenerative diseases. Here we assess three adults at risk of frontotemporal dementia from C9orf72 mutation, using [11C]UCB‐J PET to quantify synaptic density in comparison with 19 healthy controls and one symptomatic patient with behavioural variant frontotemporal dementia. The three pre‐symptomatic C9orf72 carriers showed reduced synaptic density in the thalamus compared to controls, and there was an additional extensive synaptic loss in frontotemporal regions of the symptomatic patient. [11C]UCB‐J PET may facilitate early, pre‐symptomatic assessment, monitoring of disease progression and evaluation of new preventive treatment strategies for frontotemporal dementia.


Introduction
Synaptic dysfunction and loss are early events in the pathogenesis of many neurodegenerative diseases. Frontotemporal dementia is associated with changes in behaviour and language in which post mortem human studies 1,2 and animal models 3 suggest early and severe synaptic loss. In vivo methods for assessing synapse loss have recently emerged, including fluidic and radioligand biomarkers. 4,5 A fifth of frontotemporal dementia cases have autosomal dominant inheritance, 6 most commonly due to hexanucleotide expansions in the chromosome 9 open reading frame 72 (C9orf72), mutations of granulin (GRN), or microtubule-associated protein tau (MAPT) genes. Familial frontotemporal dementias provide the opportunity to study early pathogenic processes, and evaluate biomarkers in the pre-symptomatic phase of the disease. For example, pentraxin-2, which is implicated in synaptic plasticity, is decreased in symptomatic C9orf72 mutation carriers, but not pre-symptomatic carriers. 7 However, such fluidic markers for synaptic loss do not elucidate in vivo spatial distributions of pathology, and the restricted spatial extent of pre-symptomatic disease might reduce their sensitivity.
Previous volumetric analysis of C9orf72 mutation carriers indicated that the earliest changes are detected in the thalamus. [8][9][10][11] Here, we test the hypothesis that the earliest loss of synaptic density occurs in the thalamus. To do this, we used positron emission tomography (PET) with the radioligand [ 11 C]UCB-J, which binds to synaptic vesicle protein 2A, in three asymptomatic carriers of C9orf72 mutations, a patient with symptomatic frontotemporal dementia, and 19 healthy controls. Synaptic density was estimated through the determination of the binding potential of [ 11 C]UCB-J. This ligand offers a direct measure of synaptic density, 5 reflects disease-specific regional synaptic loss, and is related to other biomarkers and clinical severity. [12][13][14][15][16] Methods Participants Three pre-symptomatic carriers of C9orf72 mutations and one patient with a clinical diagnosis of probable behavioural variant of frontotemporal dementia (bvFTD) 17 were recruited from the Cambridge Centre for Frontotemporal Dementia at the University of Cambridge. Nineteen healthy volunteers (8 females and 11 males; age range: 68.0 AE 7.0) were recruited from the UK National Institute for Health Research Join Dementia Research (JDR) register.
Demographic and clinical features are given in Table 1; individual age (mean age C9orf72 carriers = 50.3 AE 4.2) and sex are not reported to protect anonymity. The C9orf72 carriers were asymptomatic, with positive family histories, and were clinically unimpaired. Carrier 1 and Carrier 3 had a family history of genetic motor neuron disease, and at the time of PET they were approximately  (Table 1). However, all carriers underperformed in one or more frontal assessment tests, as compared to controls (Table 1). In particular, Carrier 3 underperformed on the revised version of Addenbrooke's cognitive examination (ACE-R, a cognitive screening test across five domains), and on working memory-related tests. The patient with bvFTD presented with symptoms at the age of 58 manifesting as a progressive change in personality, apathy and executive dysfunction. There were initial cognitive deficits in executive tasks but with preservation of memory and visuospatial skills, praxis, semantics and normal pyramidal and extra-pyramidal motor examinations; they were diagnosed at the age of 62. At the time of the PET scan (aged 63), they presented with a lack of insight and empathy, severe apathy, agrammatism and semantic language deficits, performing poorly in all cognitive domains, especially in executive tasks. No mutations were identified in C9orf72, GRN, MAPT or TBK1. The research protocol was approved by the Cambridge Research Ethics Committee (REC: 18/EE/0059) and the Administration of Radioactive Substances Advisory Committee. All participants provided written informed consent in accordance with the Declaration of Helsinki.
To facilitate voxel-wise statistical testing, for each subject a [ 11 C]UCB-J non-displaceable binding potential (BP ND ) map was determined from dynamic images corrected for partial volume error at the voxel level using the iterative Yang method. 19 BP ND was calculated using a basis function implementation of the simplified reference tissue model, with centrum semiovale as the reference tissue. 20 Each BP ND map was warped to the ICBM 152 2009a asymmetric MR template using parameters from the spatial normalisation of the co-registered T1 MR image with Advanced Normalization Tools (ANTs; http:// www.picsl.upenn.edu/ANTS/). BP ND was quantified in sub-divisions of the thalamus by warping the scale IV ROIs from the Melbourne subcortex atlas 21 to the ICBM 152 2009a asymmetric MR template and extracting values from the partial volume corrected BP ND maps.
For whole-brain regional analysis, a version of the n30r83 Hammers atlas (http://brain-development.org) modified to include segmentation of posterior fossa regions was spatially normalised to the T1-weighted MRI of each participant. Regions were multiplied by a binary grey matter mask (>50% on the SPM12 grey matter probability map smoothed to PET spatial resolution) and geometric transfer matrix partial volume correction 19 was applied to each image of the dynamic series. Regional BP ND was determined using the same reference tissue approach as for the BP ND maps.

Statistical analysis
To test our primary hypothesis, we assessed thalamic synaptic density. We compared the unsmoothed partialvolume corrected BP ND map of each carrier versus controls, using the randomise function in FSL (version 5.0.10) within a thalamic regional mask to generate a voxel-wise non-parametric t-test (Conditional Monte Carlo permutation test, 5000 permutations). Cluster significance was defined at p < 0.05 family-wise error rate (FWE) and clusters identified using the threshold-free cluster estimation method. We also calculated z-scores for BP ND in the thalamic sub-regions in comparison to control data.
Next, we performed exploratory, illustrative analyses across the whole brain using smoothed BP ND maps (isotropic 6 mm full width at half maximum Gaussian). First, for the patient and each mutation carrier, we calculated voxel-wise z-score maps of BP ND , to reveal voxels with reduced BP ND in each carrier/patient. Next, we calculated BP ND z-scores for each region of the modified Hammers atlas. In both cases, we considered a statistical threshold of z = À1.645 (p < 0.05 uncorrected) which corresponds to the 95th percentile of the normal distribution for a one-tailed test, based on the hypothesis of reduced synaptic density in patients and carriers compared to controls.  Table S1.

Reduction of thalamic [ 11 C]UCB-J BP ND was seen in each
Single subject and control mean [ 11 C]UCB-J BP ND maps are displayed in Figure 2 (panel A), with individual z-score maps for the three mutation carriers and patient (panel B). The participant with bvFTD showed widespread severe reduction of synaptic density, especially in the left frontal and temporal cortex, subcortical regions and, to a lesser extent, parietal cortex. For all C9orf72 mutation carriers, the most prominent reduction in [ 11 C] UCB-J binding was evident in the thalamus, along with a scattered and lesser reduction in cortical regions.
Regional [ 11 C]UCB-J BP ND values for the Hammers atlas are reported in Figure 3 (panel A), together with rendering of individual regional z-scores (panel B). The bvFTD patient showed reduced [ 11 C]UCB-J binding in the left frontal, temporal and parietal regions, cingulum, insula and thalamus, and bilaterally in middle and superior frontal gyri, the superior parietal gyrus, pallidum and substantia nigra. The C9orf72 mutation carriers had reduced [ 11 C]UCB-J binding in thalamus, reaching statistical threshold in Carrier 2 (left z = À1.77; right z = À1.26) and Carrier 3 (left z = À3.46; right z = À2.99). A significant reduction of [ 11 C]UCB-J binding was also found in the cerebellar dentate nucleus of Carrier 1 (right z = À1.77) and Carrier 2 (right   25), and in the pre-subgenual frontal cortex of Carrier 3 (right z = À1.71). Regional BP ND values and zscores are tabulated in Table S2.

Discussion
Each of three pre-symptomatic carriers of pathogenic mutations in C9orf72 showed in vivo evidence of synaptic loss in the thalamus. The restricted thalamic reduction in pre-symptomatic carriers contrasts with the extensive cortical and subcortical synaptic loss in a patient with symptomatic frontotemporal dementia.
In the C9orf72 carriers, synaptic density was most reduced in the ventral-posterior thalamic sub-region and the pulvinar nuclei (dorsal-posterior thalamic sub-region). Despite age differences between carriers and controls, previous evidence suggests that [ 11 C]UCB-J binding reflecting synaptic density does not decline with age. 22 In familial FTD, thalamic atrophy has been commonly reported in C9orf72 mutation carriers, 9,10,[23][24][25][26] early in the presymptomatic stage before the age of 40 11 and decades prior to symptom onset. 8 Thalamic neurodegeneration may not be specific to C9orf72 mutations. Thalamic atrophy can occur across the whole spectrum of frontotemporal dementia, [27][28][29][30] especially in those caused by TDP-43 pathology. 31 However, among familial frontotemporal dementia, those with C9orf72 mutation show the earliest and the most severe thalamic volume loss. Our finding, with partial volume correction applied to counter the known effect of thalamic atrophy, is consistent with a selective involvement of the pulvinar nucleus in C9orf72 expansion carriers. 10 The [ 11 C]UCB-J PET findings also accord with pathological evidence of RNA foci, dipeptide repeat protein inclusions and TDP-43 pathology in the thalamus. 32 The functional significance is suggested by the correlation between pulvinar atrophy and salience network connectivity in patients with C9orf72 mutations. 24 Carriers presented with a heterogenous family and clinical profile. Two carriers had a family history of motor neuron disease, and one of bvFTD. Carrier 3 underperformed in frontal tests, and presented the most severe pattern of synaptic loss in thalamic regions, despite being the carrier with the highest number of years of education and estimated years from onset; albeit it is well established that estimated years of symptom onset, deduced from affected family members, is a weak predictor of age of onset in C9orf72 carriers. 33 Carrier 3 also obtained low memory ACE-R sub-score, which is not a typical first domain of impairment in the FTD/ALS spectrum, although is common later in disease. Despite this heterogeneity, synaptic loss in the thalamus was seen across all subjects. Comparing our findings with those from [ 11 C] UCB-J PET in symptomatic C9orf72 carriers will help to further clarify the longitudinal impact of early thalamic impairment.
In summary, this study indicates that thalamic synaptic loss occurs early in C9orf72, before symptom onset, with a thalamic focus that has the potential to disrupt connectivity in multiple neural circuits and spread to widespread cortical regions. We suggest that [ 11 C]UCB-J is a useful biomarker for in vivo quantification of synaptic loss in frontotemporal dementia. Further investigations with other genetic aetiologies would be required to examine the specificity of these effects to C9orf72 mutations.
Darwin College and the PSP Association (UK). He provides consultancy to Asceneuron, Biogen, UCB and has research grants from AZ-Medimmune, Janssen and Lilly as industry partners in the Dementias Platform UK. TR has received honoraria from Biogen, Oxford Biomedica and the National Institute for Health and Clinical Excellence (NICE). JTO has received honoraria for work as DSMB chair or member for TauRx, Axon, Eisai and Novo Nordisk and, has acted as a consultant for Biogen, Roche, and has received research support from Alliance Medical and Merck.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Thalamic sub-division regional [ 11 C]UCB-J binding potential values (BP ND ) and z-scores (Z) for each carrier and the bvFTD patient versus controls. Regions with z-scores <À1.645, corresponding to the 95th percentile of a normal distribution for a one-tailed test, are shown in red font. Table S2. Whole-brain regional [ 11 C]UCB-J binding potential values (BP) and z-scores (Z) for each carrier and the bvFTD patient versus controls. Regions with zscores <À1.645, corresponding to the 95th percentile of a normal distribution for a one-tailed test, are shown in red font.