Cortical tau is associated with microstructural imaging biomarkers of neurite density and dendritic complexity in Alzheimer's disease

Abstract Introduction In Alzheimer's disease (AD), hyperphosphorylated tau is closely associated with focal neurodegeneration, but the mechanism remains uncertain. Methods We quantified cortical microstructure using neurite orientation dispersion and density imaging in 14 individuals with young onset AD. Diffusion tensor imaging measured mean diffusivity (MD). Amyloid beta and tau positron emission tomography were acquired and associations with microstructural measures were assessed. Results When regional volume was adjusted for, in the medial temporal lobe there was a significant negative association between neurite density and tau (partial R2 = 0.56, p = 0.008) and between orientation dispersion and tau (partial R2 = 0.66, p = 0.002), but not between MD and tau. In a wider cortical composite, there was an association between orientation dispersion and tau (partial R2 = 0.43, p = 0.030), but not between other measures and tau. Discussion Our findings are consistent with tau causing first dendritic pruning (reducing dispersion/complexity) followed by neuronal loss. Advanced magnetic resonance imaging (MRI) microstructural measures have the potential to provide information relating to underlying tau deposition.

findings suggest an association between tau and dendritic morphology; however, in vivo confirmation is lacking. 3 The development of tau positron emission topography (PET) allows in vivo quantification of cortical tau. 2 Magnetic resonance imaging (MRI) methods for automated measurement of cortical thickness and/or volume allow the detection of macrostructural loss. 4 More recently, diffusion-weighted MRI (dMRI), primarily using diffusion tensor imaging (DTI), has enabled the assessment of cortical changes at the microstructural level, usually through the metric of mean diffusivity (MD). 5,6 However, DTI provides only a composite view of each voxel, so cannot distinguish the characteristics of dendritic trees from the surrounding cellular structures or cerebrospinal fluid (CSF). 7 It is, therefore, unclear to what extent the cortical MD signal relates to CSF partial volume due to macrostructural atrophy versus changes in neuronal microstructure. The development of multi-shell dMRI, using multiple diffusion weightings, allows for multi-compartmental modeling, with the most widely used and validated being neurite orientation dispersion and density imaging (NODDI). 8,9 NODDI models neural tissue as separate tissue-specific compartments within the same voxel. It becomes possible to explore specific disease-related processes, including estimating both the orientation dispersion index (ODI)-providing information relating to dendritic complexity-and the neurite density index (NDI), with both measures having undergone histological validation. 9, 10 We would hypothesize that as neurodegeneration progresses, dendrites, which are the predominant neurite in cortical gray matter, would reduce in density and structural complexity, causing a decrease in both NDI and ODI measures.
Here we used Aβ PET, tau PET, and multimodal MRI, including NODDI, in individuals with young onset AD (age at onset <65), capitalizing on the more variable clinical phenotype in younger patients.
We explore associations between molecular pathology and neuronal/dendritic microstructure, and compare with more conventional volume and DTI metrics.  NiftyReg software was used to estimate the rigid-body registrations between each individual's T1-weighted and PET/dMRI images using a symmetric block-matching approach, 16 with the resulting transform used to resample the ROIs using nearest-neighbor interpolation.

MATERIALS AND METHODS
GIFv3 was also used for estimation of total intracranial volume (TIV).
dMRI pre-processing included motion, eddy current, and susceptibility correction. 13 The

RESULTS
Participant demographics and clinical details are outlined in Table 1.
All 14 participants were Aβ PET positive, confirming underlying AD pathology. There was no evidence of a significant association between Aβ and tau SUVRs, either in the MTL or the cortical composite. In addition, there was no association between Aβ SUVR and any of the MRI metrics.
There was no significant association between tau SUVR and cortical volume in either ROI (Figure 1). In the MTL, there was a significant neg- Linear regression, adjusting for regional TIV-corrected cortical volume, confirmed a significant negative association between NDI and tau PET SUVR in the MTL (β = −0.14, partial R 2 = 0.56, p = 0.008). An association between MTL ODI and tau (β = −0.13, partial R 2 = 0.66, p = 0.002) was also seen, but not between MTL MD and tau. In the cortical composite, after adjusting for volume, there was a significant association between ODI and tau (β = −0.17, R 2 = 0.43, p = 0.030) but not between NDI and tau or MD and tau.

DISCUSSION
In this study of individuals with young onset AD, using amyloid PET and tau PET alongside NODDI, we found higher tau deposition in the MTL to be associated with lower neurite density, a finding that persisted after correction for regional volume. This in vivo finding is consistent with there being a direct association at the microstructural level between focal tau pathology and focal loss of neurites, with implications for our understanding of the mechanisms underpinning focal neuronal loss in neurodegenerative disease.
There was no relationship between regional Aβ PET tracer uptake and either tau PET or any MRI measure. This was not unexpected given that the participants were at a disease stage when it would be expected that amyloid deposition had plateaued, and is consistent with multiple studies that demonstrated Aβ to not be tightly topographically coupled with either tau or neurodegeneration. 2,19 Higher tau load was also associated with lower ODI in the cortex; and in the MTL when volume was accounted for. Lower ODI is thought to reflect a decreased organizational complexity of neurites.
Dendrites are the predominant neurite within cortical gray matter, and, therefore, are thought to be the main contributor to cortical ODI and NDI measures. These findings would be consistent with focal tau deposition leading to dendritic degeneration or pruning, building on previous imaging work to clarify the specific link between tau and microstructure, while also replicating in vivo what has been observed in post-mortem histopathology. 3,20,21 The fact that ODI changes were generalized throughout the cortex while NDI associations were seen only in the MTL, a region typically affected early in AD, is consistent with tau deposition leading first to pruning, and then to frank neurite density changes.
Our findings are also consistent in part with a previous study of NODDI and CSF biomarkers of AD pathology (which do not provide anatomic information), which found that alterations in Aβ alone were not associated with lower NDI but alterations in both p-tau and Aβ in combination were. 22 Of interest, however, the study did not find an association between AD CSF biomarkers and ODI, contrary to the associations we observe between ODI and tau-PET.
Our results suggest that NODDI measures are more closely associated with AD tau pathology than either volume or DTI measures are.
A study that used both NODDI and DTI in a mouse model of AD found that cortical NDI correlated with histological measurements of hyperphosphorylated tau, whereas other metrics including MD did not. 23 Combined with our results, this raises the possibility that NODDI metrics, derived from standard 3T MRI multi-shell acquisitions, may be able to provide indirect measurement of underlying tau without the expense and radiation exposure of a PET scan. If validated, such a tool would have significant implications for clinical practice, research, and therapeutic trials.
This study has a number of limitations. The cohort is small, and the results should, therefore, be interpreted with a degree of caution.
Replication in larger AD patient groups will be valuable. The data presented are cross-sectional only, and future longitudinal studies will be needed to confirm the temporal relationship between cortical tau and changes in NODDI metrics. We included patients with young onset disease, a substantial proportion of whom had an atypical, PCA phenotype where medial temporal loss typically occurs later in the disease; however, despite this we still saw a relationship between MTL NDI and tau deposition.
In conclusion, our findings demonstrate, for the first time, a close association between focal tau and in vivo measures of both dendritic complexity and dendritic density in AD. The results support the hypothesis that microstructural cortical neuronal changes are mediated by focal tau deposition. Multi-compartmental modeling of multi-shell dMRI, using approaches such as NODDI, 8 has the potential to provide unique information on different aspects of microstructural degeneration, with additional sensitivity and specificity to degenerative changes compared with conventional imaging measures, and the potential to provide information pertaining to underlying tau deposition.

CONSENT STATEMENT
All participants provided written informed consent.