This study tested the hypothesis that changes in synaptic density are closely linked to MRI measures of cortical microstructure, in the neurodegenerative tauopathies of CBD and PSP-RS. The main insights are that (i) in the neurodegenerative tauopathies of Corticobasal Degeneration (CBD) and Progressive Supranuclear Palsy (PSP-RS), widespread changes in grey matter dendritic complexity are observed even in areas without significant atrophy; and (ii) Orientation Dispersion Imaging (ODI), an MRI-based estimate of post-synaptic dendritic structure and complexity, correlates with the loss of presynaptic density estimated by the PET radioligand [¹¹C]UCB-J; this effect is not a result of changes in grey matter atrophy. This extends previous investigations of ODI microstructural change in PSP-RS and CBS/CBD (Mitchell et al., 2019), demonstrating the in vivo coupling of dendritic complexity to presynaptic density, in line with preclinical models of tauopathies (Hoffmann et al., 2014; Rocher et al., 2010).

Grey matter atrophy within the motor cortex, as well as the basal ganglia and brainstem are common MRI findings in PSP-RS and CBD; the pattern of atrophy shown in figure 1 and 2 concords with previous studies (Jabbari et al., 2019). In this study however we show that, both within areas of the brain where there is atrophy, and in areas with absent atrophy, there is significant and more severe loss of dendritic complexity and, as previously shown (Holland et al., 2020) and replicated here, presynaptic density in the patient cohort. The widespread pattern of cortical ODI and presynaptic deficits in PSP-RS and CBD matches the expected pattern of tau pathology, both in the subcortical and cortical area (Dickson et al., 2011; Kovacs et al., 2020). Loss of dendritic complexity and presynaptic loss, in non-atrophied areas of the brain, for example the occipital lobes, potentially reflects early changes in synaptic function, in response to tau pathology (Kovacs et al., 2020), or toxic oligomers of tau, that later in the disease process progress to atrophy. Indeed, in preclinical models, pathological tau oligomers induce synaptic degeneration (Usenovic et al., 2015) and interfere with synaptic function and density (Yoshiyama et al., 2007), in the absence of neuronal loss.

The effect size for the comparison between ODI and [¹¹C]UCB-J BP_ND in patients relative to controls, was stronger than for atrophy. Together, the larger extent of ODI reductions relative to grey matter atrophy indicate that NODDI can reveal new aspects of the cellular pathology of tauopathies, in keeping with the recent demonstration that changes in ODI parameters closely reflect complex histological changes (Grussu et al., 2017). Notwithstanding the cross-sectional design of this study, this observation highlights the early structural changes in disease pathogenesis that may underlie the emergence of cognitive and motor symptoms not attributable to atrophy. Indeed studies of cortical physiology in PSP-RS have illustrated abnormal electrophysiology in the absence of atrophy (Hughes et al., 2013; Sami et al., 2018). Our findings of reduced ODI beyond atrophy are in agreement with those published from the Alzheimer’s disease literature (Parker et al., 2018) which is a related tauopathy to PSP-RS and CBD, where there is a stronger relationship between cognitive function and synaptic density than with atrophy (Terry et al., 1991).

Our observation of a tight coupling between dendritic complexity (figure 3 and 4), as measured with ODI, and presynaptic density, as measured with [¹¹C]UCB-J, in the patient cohort, echo preclinical findings in animal models of tauopathy (Harris and Kater, 1994), and post mortem studies (Bigio et al., 2001; Lipton et al., 2001). There are two potential explanation for this tight coupling: first, tau protein is enriched in axons and associated with axonal growth and transport; the synaptic toxicity associated with pathological tau is observed both in pre- and post-synaptic structures (see review by Mang & Mendelkow 2015 (Wang and Mandelkow, 2016)), leading to both a reduction in dendritic complexity and presynaptic density. Second, while a cross-sectional design precludes inferences of causality, alterations in synaptic function in CBD and PSP-RS (Holland et al., 2020) are expected to induce dendritic morphological alterations with early loss of dendritic spines and reduced dendritic branching. This has been illustrated in animal studies, where post-synaptic dendritic morphology correlates with the numbers of presynaptic vesicles and synaptic strength (Schikorski and Stevens, 1999). It is possible that primary dendritic degeneration in CBD and PSP-RS causes a reduction in the density of the presynaptic contacts on the distal dendrites. These alternative accounts are not mutually exclusive and could act in parallel to impair effective transneuronal connectivity, and subsequently motor and cognitive function.

The downstream effect of the above is a loss of functional connectivity. Indeed in PSP-RS there is reduced resting state connectivity between cortical and subcortical areas - reviewed in (Filippi et al., 2019). In CBD, the evidence is mixed as, thus far, functional MRI studies in this disease have not differentiated between amyloid positive and amyloid negative participants (also reviewed in (Filippi et al., 2019); in the former where Alzheimer’s pathology is the most likely finding at post mortem, functional connections are strengthened; however in the latter where a 4R tauopathy is likely, functional connections may become inefficient as seen in PSP-RS (Cope et al., 2018).

Synaptic pathology is a common feature in many neurodegenerative diseases (Clare et al., 2010; Herms and Dorostkar, 2016; Kweon et al., 2017). In the related tauopathy of Alzheimer’s disease, widespread reduction in synapses have been shown using [¹¹C]UCB-J PET in patients (Mecca et al., 2020), as well as changes to dendritic complexity seen both in vivo as reduced ODI (Parker et al., 2018) and in animal models of Alzheimer’s disease (Dorostkar et al., 2015). Similar pre- and postsynaptic losses are seen in the substantia nigra of patients with Parkinson’s disease (Bellucci et al., 2016; Reeve et al., 2018), and within the hippocampi of patients with Lewy Body Dementia (Revuelta et al., 2008). The tight coupling of the presynaptic and post-synatpic compartments shown in our study, may therefore extend to pathologies other than primary 4R-tauopathies, where synaptic dysfunction is one of the earliest stages of disease, regardless of the culprit protein aggregate in question. If so, the opportunity to utilise both MRI and PET to understand disease pathogenesis would greatly facilitate cohort studies of individual phenotypic differences, or repeat testing in clinical trials monitoring. Furthermore, the in vivo observation of the tight coupling between the pre- and postsynaptic compartments, offer a new angle of interpretation for future studies utilising either of these methods in isolation; for example if cost, scalability and resources are limited for larger cohort studies, MRI measures of dendritic complexity can offer a close surrogate of presynaptic density, although we appreciate it is not a substitute given the significantly larger effect sizes seen with [¹¹C]UCB-J (Figure 3C and 4C). This latter observation, which replicates our previous study in a smaller cohort of patients, may indicate a more severe toxic effect at the presynapse compared to dendritic pathology, although longitudinal studies are required to confirm this. However, it is important to note that ODI should not be taken as a direct representation of the post-synaptic density seen in histological studies. Rather, it provides complementary information to the PET data, with support from histological studies that have shown ODI to be a close surrogate (Grussu et al., 2017).

There are several strengths of this study. First, our NODDI resolution of 1.75mm was specifically optimised for the investigation of cortical microstructure, and is higher than previous NODDI studies in early onset AD (2.5 mm) (Parker et al., 2018), Parkinson’s disease (5 mm) (Kamagata et al., 2016) and PSP-RS (2.0 mm) (Mitchell et al., 2019). Second, we use amyloid imaging to exclude patients with CBS likely due to Alzheimer’s disease. This helps in reducing the potential underlying pathologies at play in the CBS cohort to 4R tauopathy, although we acknowledge other pathologies are possible but less likely (Alexander et al., 2014) – in this regard it is reassuring that the spatial patterns of reductions in ODI and [¹¹C]UCB-J binding are not substantially different between the PSP-RS and CBD cohorts. Thirdly, the use of non-parametric permutation analyses also confer additional robustness to our statistical results, since this approach is not reliant on data distributions (Winkler et al., 2014).

We acknowledge the limitations of our study including the relatively small sample size in our CBD cohort. However, given the large effect sizes seen of ODI and [¹¹C]UCB-J binding reduction in both our patient groups, our study was sufficiently powered to examine the relationships between both imaging markers. Secondly, we acknowledge that the MRI measure of ODI is not a measure of post-synaptic density but rather dendritic complexity; PET radioligands targeting the post-synaptic density may therefore provide further useful insights into post-synaptic pathology. Furthermore, although we illustrate a tight correlation between our two measures of synaptic health, we do not show causality in this cross-sectional study design, but are working towards this aim through our longitudinal studies. Lastly, we highlight the limitations of clinicopathological correlations, but take reassurance from previous validation studies illustrating a high consistency in the clinicopathological correlations in PSP-RS and in amyloid negative CBS.

Understanding the pathological processes that precede atrophy in neurodegeneration is key not only in expanding our knowledge of the pathophysiology of disease but also in informing the design of clinical trials both in terms of the imaging options for measuring disease-related changes and the optimal timing of intervention in the disease process. Our data implicate correlated changes in dendritic microstructure and synaptic density in patients with primary degenerative tauopathies including PSP-RS and CBD. Further cross-validation of ODI with [¹¹C]UCB-J BP_ND may help further our understanding of the pathophysiology of neurodegeneration, applicable to future studies of early neurodegeneration with a safe and widely available MRI platform.