OBJECTIVE: Synaptic loss is an early feature of neurodegenerative disease models, and is severe in post mortem clinical studies, including frontotemporal dementia. Positron emission tomography (PET) with radiotracers that bind to synaptic vesicle glycoprotein 2A enables quantification of synaptic density in vivo. This study used [11 C]UCB-J PET in participants with behavioural variant frontotemporal dementia (bvFTD), testing the hypothesis that synaptic loss is severe and related to clinical severity. METHODS: Eleven participants with clinically probable bvFTD and 25 age- and sex-matched healthy controls were included. Participants underwent dynamic [11 C]UCB-J PET, structural MRI and a neuropsychological battery, including the Addenbrooke's Cognitive Examination (ACE-R), and INECO frontal screening (IFS). General linear models compared [11 C]UCB-J binding potential maps and grey-matter volume between groups, and assessed associations between synaptic density and clinical severity in patients. Analyses were also performed using partial-volume corrected [11 C]UCB-J binding potential from regions of interest (ROIs). RESULTS: Patients with bvFTD showed severe synaptic loss compared to controls. [11 C]UCB-J binding was reduced bilaterally in medial and dorsolateral frontal regions, inferior frontal gyri, anterior and posterior cingulate gyrus, insular cortex and medial temporal lobe. Synaptic loss in the frontal and cingulate regions correlated significantly with cognitive impairments. Synaptic loss was more severe than atrophy. Results from ROI-based analyses mirrored the voxel-wise results. INTERPRETATION: In accordance with preclinical models, and human post mortem evidence, there is widespread frontotemporal loss of synapses in symptomatic bvFTD, in proportion to severity. [11 C]UCB-J PET could support translational studies and experimental medicine strategies for new disease-modifying treatments for neurodegeneration. This article is protected by copyright. All rights reserved.