Synaptic Loss in Primary Tauopathies Revealed by [11C]UCB‐J Positron Emission Tomography

Synaptic loss is a prominent and early feature of many neurodegenerative diseases.

The primary degenerative tauopathies of progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD) cause a severe combination of movement and cognitive impairment. [1][2][3][4] Pathologically, both are associated with a four-repeat (4R)-tauopathy. 5 We proposed that the neurophysiological and functional impairments in PSP and CBD are, at least in part, a consequence of synaptic loss. For example, at postmortem there is 50% loss of cortical synapses in PSP and CBD, 6,7 and in vivo there is limited evidence of a 20% loss of postsynaptic GABA A receptors as shown with [ 11 C]flumazenil PET. 8,9 Indeed, abnormal physiology in pathways involved in presynaptic function have been identified from transcriptomic studies in patients with mutations in the microtubule-associated protein tau gene. 10 Transgenic models of tauopathies (eg, rTg4510) confirm a synaptotoxic effect of oligomeric tau, before cell death. 11,12 Moreover, in other neurodegenerative dementias, such as Alzheimer's disease (AD), synaptic loss correlates better with cognitive dysfunction than atrophy. 13 We therefore tested the hypothesis that PSP and CBD reduce synaptic density, in proportion to disease severity. We included patients with the classic phenotype of PSP, PSP-Richardson's syndrome, which has a high clinicopathological correlation, 14 and presents with postural instability, supranuclear gaze palsy, axial rigidity, and cognitive impairment. 15 Other phenotypes of PSP are increasingly recognized, 3,16 but excluded here. We include patients with corticobasal syndrome (CBS), with combinations of asymmetric rigidity, apraxia, dystonia, alien limb, and cognitive impairment. 1,17 In order to identify those with probable underlying CBD, it is necessary to exclude the substantial minority of CBS caused by AD pathology. 18 We therefore used amyloid imaging to distinguish those with CBS attributed to CBD, versus AD; we refer to this group as the CBD cohort. Both PSP and CBD are associated with cortical and subcortical atrophy on MRI 19 ; and changes in neurophysiology and connectivity measured by magnetoencephalography and functional MRI. [20][21][22][23] However, functional changes are also observed in areas of the brain that are minimally atrophic.

Participants and Study Design
Fourteen patients with PSP-Richardson's syndrome and 15 patients with CBS were recruited from a tertiary specialist clinic for PSP/CBS at the Cambridge University Centre for Parkinson-Plus (Cambridge, UK). Fifteen healthy volunteers were recruited from the UK National Institute for Health Research Join Dementia Research register. Patients had either probable PSP-Richardson's syndrome, 3 or both probable CBS and probable CBD. 1 Healthy controls and patient volunteers were initially screened by telephone; our exclusion criteria were: current or recent history (within the last 5 years) of cancer, concurrent use of the medication levetiracetam, history of ischaemic or haemorrhagic stroke evident on MRI available from the clinic, any severe physical illness or co-morbidity that limited ability to fully participate in the study, and any contraindications to performing MRI. Eligible participants were invited for a research visit where they underwent clinical and cognitive assessment including measures of disease severity (Table 1); these included a neurological examination by a clinician including the PSP and CBD rating scales, the UPDRS (motor subsection III), the Schwab and England Activities of Daily Living (SEADL) and Clinical Dementia Rating Scale (CDR); cognitive testing included the revised Addenbrooke's Cognitive Examination (ACE-R), the Mini-mental State Examination (MMSE), the Montreal Cognitive Assessment (MoCA), and the INECO frontal assessment test. Patients' carers completed the revised Cambridge Behavioural Inventory (CBI).
All participants underwent simultaneous 3 Tesla MRI and [ 11 C]UCB-J PET. Patients with CBS also underwent amyloid PET imaging using Pittsburgh compound B ([ 11 C] PiB), and cortical standardized uptake value ratio (SUVR; 50-70 minutes postinjection; whole cerebellum reference tissue) was determined using the Centiloid Project methodology. 30 Only those with a negative amyloid status, as characterized by a cortical [ 11 C]PiB SUVR <1.21 (obtained by converting the Centiloid cutoff of 19 to SUVR using the Centiloid-to-SUVR transformation) 31 are included in the subsequent analysis, with the aim of excluding patients with CBS associated with Alzheimer's disease. We interpret this amyloid-negative group as having CBD, although acknowledge that other pathologies are possible.
The research protocol was approved by the local 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.
Neuroimaging [ 11 C]UCB-J was synthesized at the Radiopharmacy Unit, Wolfson Brain Imaging Centre, Cambridge University (Cambridge, UK), using the methodology previously described. 32 Dynamic PET data acquisition was performed on a GE SIGNA PET/MR (GE Healthcare, Waukesha, WI) for 90 minutes, starting immediately after [ 11 C]UCB-J injection (median injected activity: 351 ± 107 MBq; injected mass: ≤10 μg), with attenuation correction including the use of a multisubject atlas method 33,34 and also improvements to the MRI brain coil component. 35 Each emission image series was aligned using SPM12 (www.fil.ion.ucl.ac.uk/spm/software/spm12/), then rigidly registered to a T 1 -weighted MRI acquired during PET data acquisition (repetition time = 3.6 msec, echo time = 9.2 msec, 192 sagittal slices, in-plane resolution 0.55 × 0.55 mm [subsequently interpolated to 1.0 × 1.0 mm]; slice thickness 1.0 mm). Using a version of the Hammersmith atlas (http://brain-development.org) with modified posterior fossa regions, combined regions of interest (ROIs; including aggregated regions for frontal, parietal, occipital, and temporal lobes; cingulate; and cerebellum) were spatially normalized to the T 1 -weighted MRI of each participant using Advanced Normalization Tools (ANTs) software. 36 Regional time-activity curves were extracted following the application of geometric transfer matrix (GTM) partial volume correction (PVC 37 ) to each of the dynamic PET images. ROIs were multiplied by a binary gray matter mask (>50% on the SPM12 gray matter probability map smoothed to PET spatial resolution), with the exception of the pallidum, substantia nigra, pons, and medulla because masking eliminated the ROI for some or all of the subjects. Multiple background gray matter, white matter, and cerebrospinal fluid regions were also defined to provide whole-brain coverage for GTM PVC. The mean gray matter/(gray matter + white matter) fraction in the masked ROIs was 0.97 ± 0.03, 0.96 ± 0.03, and 0.96 ± 0.03 for the control, CBD, and PSP groups, respectively, illustrating the predominance of gray matter in the masked ROIs. To assess the impact of PVC, time-activity curves were also extracted from the same ROIs without the application of GTM PVC.
To quantify SV2A density, [ 11 C]UCB-J nondisplaceable binding potential (BP ND ) was determined, both regionally and at the voxel level, using a basis function implementation of the simplified reference tissue model, 38 with the reference tissue defined in the centrum semiovale. 39,40 The volume-weighted average of the GTM PVC BP ND values in the masked ROIs was used as a global BP ND metric. Group average BP ND images (illustrated in Fig. 1A) were obtained by spatially normalizing each individual T 1 -weighted MRI (and thereby the coregistered BP ND map) to Montreal Neurological Institute (MNI) space, and then to the group template using ANTs.

Statistical Analysis
Statistical analyses used R software (version 3.6.2; R Foundation for Statistical Computing, Vienna, Austria), with analysis of covariance to compare regional [ 11 C] UCB-J BP ND between the three groups (control, CBD, and PSP), with age as a covariate of no interest. ROIs were: frontal, temporal, parietal, and occipital lobes; cingulate cortex, hippocampus, insula, amygdala, caudate nucleus, nucleus accumbens, putamen, pallidum, thalamus, cerebellum, substantia nigra, midbrain, pons, and medulla. The relationships between [ 11 C]UCB-J BP ND , disease severity (PSP and CBD rating scales), and cognition (ACE-R) were tested through linear models of the patient data, with age as a covariate of no interest.
The primary analyses used BP ND determined following GTM PVC, but all analyses were repeated using BP ND without PVC.

Results
Of the 15 patients with CBS, 6 had a cortical [ 11 C] PiB SUVR >1.21 and were therefore excluded from further analysis in this article. The remaining groups (9 CBD, 14 PSP, and 15 controls) were matched in age, sex, and education (Table 1). We observed typical cognitive profiles, as summarized in Table 1: Patients were impaired on memory, verbal fluency, language, and visuospatial domains of the ACE-R, MMSE, and MoCA. There were high endorsements on the CBI, and the CDR scale, with impairment of activities of daily living on the Schwab and England scale. Concurrent medications used by our participants at the time of the [ 11 C]UCB-J PET scan are outlined in Supporting Information Table S1. Four of our patients (1 PSP, 3 CBD) were on dopaminergic medication and 9 on amantadine (3 PSP, 6 CBD).
Compared to controls, in patients there was a significant global reduction in [ 11 C]UCB-J BP ND (Fig. 1A-C) across all major cortical and subcortical areas (P < 0.05 false discovery rate [FDR] corrected for all ROIs shown in Fig. 1C); regional BP ND values for the three groups are reported in Table 2. BP ND in PSP and CBD was 20% to 50% lower than controls (P < 0.01), with the most severe median reduction observed in the medulla, substantia nigra, pallidum, midbrain, pons, and caudate nucleus in patients with PSP and in the medulla, hippocampus,  amygdala, caudate nucleus, insula, and thalamus in patients with CBD. Post-hoc analysis revealed that the significant differences in BP ND between patients and controls in the pallidum and substantia nigra were mainly driven by the PSP cohort. Using data without GTM PVC, the pattern of statistically significant differences in BP ND for the reported regions in Table 2 remains, P < 0.001. The reduction in synaptic density was noted even in areas of the brain that did not show significant gray matter atrophy. Figure 2A shows the group differences in gray matter volume normalized against the mean of the control group; the significant areas of gray matter volume loss were in the caudate nucleus (P = 0.01) and thalamus (P = 0.04) in the CBD cohort and in the frontal (P < 0.01), temporal (P = 0.04), parietal (P < 0.01), and occipital lobes (P < 0.01), caudate nucleus (P < 0.001), and thalamus (P < 0.01) in the PSP cohort. The reduction in [ 11 C]UCB-J BP ND , however, was more extensive and consistently significantly different across all major cortical and subcortical areas, as shown in the normalized plot in Figure 2B (binding potentials were normalized against the mean binding potential of the control cohort for each ROI).
Correlations between [ 11 C]UCB-J BP ND and both global cognition and disease severity are given in Figure 3. A significant positive correlation was observed between [ 11 C]UCB-J BP ND and the ACE-R total score (R = 0.52; P = 0.01; Fig. 3A). There was a significant negative correlation between [ 11 C]UCB-J binding and the PSP (R = -0.61; P < 0.01) and CBD (R = -0.72; P < 0.001) rating scales (Fig. 3B,C).

Discussion
The principal result of this study is a widespread reduction in synaptic density in PSP-Richardson's syndrome and amyloid-negative CBS (which we define as CBD). This accords with postmortem estimates of synaptic loss in PSP and CBD, using synaptophysin immunohistochemistry, 6 imaging of neurite density in PSP, 41 and morphological studies of cortical dendrites in the closely related condition of frontotemporal lobe dementia. 42 Indirect evidence of synaptic loss, from consequential reduction in metabolism, comes from [ 18 F]FDG PET changes in frontal, temporal, and parietal lobes. [43][44][45][46] However, PET imaging with the ligand [ 11 C]UCB-J provides direct evidence in vivo of severe and extensive loss of cortical and subcortical synapses, including areas of the brain that are minimally atrophic. 47 PSP and CBD are progressive, with an average disease duration of 5 to 8 years from symptom onset. 48 In our clinically diagnosed CBD and PSP groups, mean symptom duration at the time of PET was 3.5 years, and our patients were likely to be approximately midway through their symptomatic disease course (not including a potentially long presymptomatic period). The median reduction of 20% (and maximal 50%) in [ 11 C]UCB-J binding observed in vivo, compared to controls, is therefore in keeping with the predictions from postmortem data.
The synaptic loss observed in our study was widespread, extending beyond the regions that are arguably most associated with the diseases. In PSP, from postmortem studies, these include basal ganglia, thalamus, substantia nigra, premotor cortex, as well as the dentate nucleus and cerebellar white matter. In CBD, areas associated with the disease include cortex, thalamus, basal ganglia, and brainstem, without cerebellar involvement. [48][49][50] However, in our study, the loss of synapses in PSP is global across the cortex, and not confined to the premotor and motor areas, and extends beyond the substantia nigra in the brainstem with pontine and medullary involvement. Loss of synapses in the cerebellum in PSP echoes pathological studies of tau distribution in this disease. 49 Interestingly, the cerebellum was also markedly abnormal in CBD; although cerebellar atrophy and tau accumulation are not typical associations of CBD. 49 Cerebellar synaptic loss in CBD may therefore represent cerebellar diaschisis in response to widespread cortical pathology and loss of corticocerebellar projections; a small minority of persons in an amyloid-negative CBS cohort may have PSP as the underlying cause for their CBS, although this is unlikely to be sufficient to drive the group-wise effect.
Preclinical models of tauopathy suggest early synaptotoxicity with reduced plasticity and density, 11 in response to soluble oligomeric tau aggregates 12 and inflammation. 51 The toxicity associated with tau pathology leading to synapse loss is complex and involves direct and indirect pathways (reviewed in Spires-Jones and colleagues 52 ). Naturally occurring tau plays a role in synaptic function through modulating microtubule and axonal stability; disruptions to this machinery lead to prevention of the trafficking of essential components to synapses, such as synaptic receptors 53 and mitochondria. Indeed, overexpression of tau interferes with mitochondria transport 54 and contributes to hyperexcitability of neurons and impaired calcium influx in transgenic mouse models (rTg4510). 55 The global nature of synaptic reduction suggests a more widespread pathology in the primary tauopathies of PSP and CBD beyond the areas that are histologically reported as harboring a high tau burden, such as the basal ganglia, thalamus, and brainstem. 56 This may, in part, be explained by the global damage caused by oligomers of tau, which are not easily visible on tau PET imaging or histology. In support of this are biochemical studies that report tau accumulation in both gray and white matter by western blot in PSP, but not necessarily by immunohistochemistry. 57 We observed a significant correlation between synaptic loss and disease severity in PSP and amyloid-negative CBS. Synaptic loss correlates with cognitive impairment in another clinical tauopathy, AD, 13,58 and preclinical models of this. 59,60 Our in vivo PET results support the potential use of synaptic PET as a marker of disease and progression, but longitudinal data are required. Synaptic PET may support early-stage clinical trials in PSP and CBS/CBD; it is encouraging, in this latter respect, that [ 11 C]UCB-J PET is sensitive to changes in synaptic density; for example, in response to treatment with the synaptic modulator, Saracatinib. 61 Our study has several limitations. Although the sample size is small, it is adequately powered in view of the large effect sizes predicted. However, subtler relationships with mild disease, progression, or individual clinical features, or phenotypic variants of PSP and CBS, require larger studies. We acknowledge the potential for off-target binding, but preclinical data indicate very high correlations between UCB-J and synaptophysin, a marker of presynaptic vesicular density. 25 Our diagnoses were clinical, without neuropathology, although the clinicopathological correlations of PSP-Richardson's syndrome are very high, and in the absence of AD, the clinicopathological correlation of CBS with a 4Rtauopathy (CBD or PSP) is also high. 18 Binding potentials for SV2A radioligands such as [ 11 C]UCB-J can be confounded by the use of concurrent medication that may bind to SV2A. We did not enroll any individuals taking levetiracetam or any member of this family of drugs that are SV2A-specific ligands. 62 Previously reported studies using [ 11 C]UCB-J in disease have usually not commented on medications used by participants; however, one study using this ligand in major depressive disorders reports exclusion of participants on psychotropic medications in the 2 months preceding PET scanning 63 ; whereas many of our PSP and CBD patients are on medications falling under the psychotropic umbrella, to our knowledge, none of these bind to SV2A.
Arterial blood sampling was not carried out in this study; we used reference tissue modeling to reduce the demand on our patient cohort. Reference tissue modeling of [ 11 C]UCB-J with the centrum semiovale as the reference tissue has been verified against arterial input function compartmental modeling in healthy controls 39,40 and in AD. 64 To assess the validity of the centrum semiovale in our cohort, we determined the mean total distribution volume (V T ) for each of our subject groups using standard arterial input function data from the literature 26,65 ; this approach assumed that the standard input function was equally valid for all groups. This analysis indicated a small positive bias in centrum semiovale V T for CBD (5%) and less so PSP (2%) relative to that in controls, which would lead to a commensurate reduction in BP ND under the assumption that the nondisplaceable distribution volume in the target ROIs remains invariant. These biases cannot, however, explain the much greater BP ND reductions observed for CBD and PSP, which is especially true for PSP. Indeed, scaling BP ND in the CBD and PSP cohorts to account for the above biases in centrum semiovale V T produced a similar pattern of significant global reduction in BP ND for patients compared to controls, except that the significant differences in the midbrain, pons, substantia nigra, pallidum, and occipital lobe were primarily driven by the PSP cohort in the post-hoc analysis.
The therapeutic challenge in tauopathies is partly attributable to the complex nature of the underlying pathology. Early-stage trials will require early accurate diagnosis, although diagnosis is typically made 3 years after symptom onset. 66,67 It is unlikely that synaptic PET could provide presymptomatic diagnosis in rare conditions, but it is a promising tool to characterize pathogenetic mechanisms, monitor progression, and assess response to experimental medicines. 68