Amyloid‐dependent and amyloid‐independent effects of Tau in individuals without dementia

Abstract Objective To investigate the relationship between the topography of amyloid‐β plaques, tau neurofibrillary tangles, and the overlap between the two, with cognitive dysfunction in individuals without dementia. Methods We evaluated 154 individuals who were assessed with amyloid‐β PET with [18F]AZD4694, tau‐PET with [18F]MK6240, structural MRI, and neuropsychological testing. We also evaluated an independent cohort of 240 individuals who were assessed with amyloid‐β PET with [18F]Florbetapir, tau‐PET with [18F]Flortaucipir, structural MRI, and neuropsychological testing. Using the VoxelStats toolbox, we conducted voxel‐wise linear regressions between amyloid‐PET, tau‐PET, and their interaction with cognitive function, correcting for age, sex, and years of education. Results In both cohorts, we observed that tau‐PET standardized uptake value ratio in medial temporal lobes was associated with clinical dementia rating Sum of Boxes (CDR‐SoB) scores independently of local amyloid‐PET uptake (FWE corrected at p < 0.001). We also observed in both cohorts that in regions of the neocortex, associations between neocortical tau‐PET and clinical function were dependent on local amyloid‐PET (FWE corrected at p < 0.001). Interpretation In medial temporal brain regions, characterized by the accumulation of tau pathology in the absence of amyloid‐β, tau had direct associations with cognitive dysfunction. In brain regions characterized by the accumulation of both amyloid‐β and tau pathologies such as the posterior cingulate and medial frontal cortices, tau’s relationship with cognitive dysfunction was dependent on local amyloid‐β concentrations. Our results provide evidence that amyloid‐β in Alzheimer’s disease influences cognition by potentiating the deleterious effects of tau pathology.


Abstract
Objective: To investigate the relationship between the topography of amyloid-β plaques, tau neurofibrillary tangles, and the overlap between the two, with cognitive dysfunction in individuals without dementia. Methods: We evaluated 154 individuals who were assessed with amyloid-β PET with [ 18 F]AZD4694, tau-PET with [ 18 F]MK6240, structural MRI, and neuropsychological testing. We also evaluated an independent cohort of 240 individuals who were assessed with amyloid-β PET with [ 18 F]Florbetapir, tau-PET with [ 18 F]Flortaucipir, structural MRI, and neuropsychological testing. Using the VoxelStats toolbox, we conducted voxel-wise linear regressions between amyloid-PET, tau-PET, and their interaction with cognitive function, correcting for age, sex, and years of education. Results: In both cohorts, we observed that tau-PET standardized uptake value ratio in medial temporal lobes was associated with clinical dementia rating Sum of Boxes (CDR-SoB) scores independently of local amyloid-PET uptake (FWE corrected at p < 0.001). We also observed in both cohorts that in regions of the neocortex, associations between neocortical tau-PET and clinical function were dependent on local amyloid-PET (FWE corrected at p < 0.001). Interpretation: In medial temporal brain regions, characterized by the accumulation of tau pathology in the absence of amyloid-β, tau had direct associations with cognitive dysfunction. In brain regions characterized by the accumulation of both amyloid-β and tau pathologies such as the posterior cingulate and medial frontal cortices, tau's relationship with cognitive dysfunction was dependent on local amyloid-β concentrations. Our results provide evidence that amyloid-β in Alzheimer's disease influences cognition by potentiating the deleterious effects of tau pathology.

Introduction
The role of amyloid-β in the cognitive dysfunction which characterizes Alzheimer's disease (AD) has been a matter of extensive debate. 1 Current versions of the amyloid cascade hypothesis stipulate that amyloid-β is a disease trigger for numerous other pathophysiological processes leading to tau hyperphosphorylation, neuroinflammation, and neurodegeneration, 2,3 eventually resulting in cognitive dysfunction. A large body of literature has identified multiple neurotoxic roles for amyloid-β, including synaptic dysfunction 4 and synapse loss. 5,6 However, the frequent appearance of elevated amyloid-β in individuals without detectable cognitive impairment 7 challenges the purported relationship between amyloid-β and neural dysfunction suggested by experimental studies.
Recent human imaging studies suggest that aggregation of tau into neurofibrillary tangles, rather than amyloid-β, is closely linked with clinical status, with tau-PET patterns recapitulating regional glucose hypometabolism 8 and domain-specific cognitive dysfunction. 8,9 These studies, coupled with the observation of amyloid-β reaching a plateau early in the disease 10,11 have led to an emerging framework in which AD is characterized by amyloiddependent and amyloid-independent phases. 12,13 Postmortem 14 and in vivo 15,16 studies have documented a characteristic sequential pattern of tau aggregation beginning in the medial temporal lobes, eventually spreading to multisensory association areas and subsequently primary sensory areas of the neocortex. Amyloidβ aggregation, on the other hand, is characterized by early neocortical aggregation in regions such as the posterior cingulate, precuneus, and medial prefrontal cortices. 17,18 Building on reports of heightened toxicity in the presence of both amyloid-β and tau pathologies, the colocalization of amyloid-β and tau in neocortical regions highlights the possibility that the neurotoxic effects of tau may be potentiated by local amyloid-β in a region-dependent manner.
Here, we test the hypothesis that amyloid-β potentiates the effects of tau pathology on clinical function in AD. Based on the reported topographical patterns of amyloidβ and tau pathologies, we hypothesize that tau in the medial temporal lobes will be directly associated with cognitive dysfunction, while neocortical tau's effects will be potentiated by local amyloid-β. We measured amyloid-β and tau pathology with PET in two independent cohorts of cognitively unimpaired (CU) elderly and individuals with mild cognitive impairment (MCI). Using a novel analytical framework, we tested whether associations between tau pathology and clinical function are dependent on local amyloid-β concentrations.

TRIAD
The Translational Biomarkers in Aging and Dementia (TRIAD) cohort aims at describing biomarker trajectories and interactions as drivers of dementia. 19 20 Cognitively normal controls had a CDR of 0 and individuals with MCI had a CDR of 0.5. Inclusion criteria for all subjects are the ability to speak English or French, good general health (no diseases expected to interfere with study participation over time), absence of claustrophobia, and adequate visual and auditory capacities to follow neuropsychological evaluation. This study's protocol was approved by McGill University's Institutional Review Board and informed written consent was obtained from each subject. There was no attempt to match cases between cohorts.

ADNI
In this study, we also assessed cognitively normal individuals (n = 157) as well as individual with amnestic MCI (n = 83) individuals from Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort who underwent amyloid-β PET with [ 18 F]Florbetapir, tau-PET with [ 18 F] Flortaucipir, structural MRI, and genotyping for APOEε4. Cognitively normal controls had a CDR of 0, MCI subjects had a CDR of 0.5. The ADNI study was approved by the Institutional Review boards of all of the participating institutions. Informed written consent was obtained from all participants at each site. Full information regarding the ADNI inclusion and exclusion criteria can be accessed at http://adni.loni.usc.edu/.

TRIAD
Determination of APOE genotypes for subjects recruited at McGill was performed using the polymerase chain reaction amplification technique, followed by restriction enzyme digestion, standard gel resolution, and visualization processes. Full details of this procedure can be found elsewhere. 21

ADNI
Determination of APOE genotypes for ADNI subjects took place at the University of Pennsylvania Alzheimer's Disease Biomarker Laboratory. Complete details of genetic methods employed in ADNI can be accessed at http://adni.loni.usc.edu/data-samples/clinical-data/.
PET image acquisition and processing TRIAD All subjects had a T1-weighted MRI which was used for coregistration. PET imaging acquired in the TRIAD cohort has been described previously. 22 [ 18 F]MK6240 images were acquired 90-110 min postinjection and scans were reconstructed with the OSEM algorithm on a 4D volume with four frames (4 × 300 sec). 23 [ 18 F]AZD4694 images were acquired 40-70 min postinjection and scans were reconstructed with the OSEM algorithm on a 4D volume with three frames (3 × 600 sec). 24 The reconstruction algorithm is a 3D ordinary Poisson ordered subset expectation maximization (OP-OSEM) 25 with point spread function 26 modeling, using 16 subsets and 10 iterations. Immediately following each PET acquisition, a 6-min transmission scan was conducted with a rotating 137 Cs point source for attenuation correction. Additionally, the images underwent correction for dead time, decay, and random and scattered coincidences. T1weighted images were nonuniformity and field-distortion corrected and processed using an in-house pipeline. Then, PET images were automatically registered to the T1weighted image space, and the T1-weighted images were linearly and nonlinearly registered to the ADNI template space. Subsequently, a PET nonlinear registration was performed using the linear and nonlinear transformations from the T1-weighted image to the ADNI space and the PET to T1-weighted image registration, using ANTs. The PET images were spatially smoothed to achieve a final resolution of 8 mm full-width at half maximum. [ 18 F] MK6240 standardized uptake value ratio (SUVR) maps were generated using the inferior cerebellar grey matter as a reference region and [ 18 F]AZD4694 SUVR maps were generated using the cerebellar grey matter as a reference region. 27  In each cohort, we tested whether amyloid-β potentiates relationships between tau pathology and cognitive dysfunction.
Baseline demographic and clinical data were assessed using t tests and χ 2 tests. Neuroimaging analyses were carried out using the VoxelStats toolbox (https://github.c om/sulantha2006/VoxelStats), a MATLAB-based analytical framework that allows for the execution of multimodal voxel-wise neuroimaging analyses. 30 Other statistical analyses were performed using the R Statistical Software Package version 3.5.3 (http://www.r-project.org/). Amyloid-PET and tau-PET images were centered on the mean of each cohort in order to improve coefficient interpretability numerical stability for estimation associated with multicollinearity. 31 Given the large number of covariates in the statistical models, model diagnostics were carried out using the car package in R to determine the presence of multicollinearity.
In the TRIAD cohort, the voxel-based model outlined below was built to test whether main effects and interactive effects of [ 18 F]AZD4694 SUVR and [ 18 F]MK6240 SUVR on CDR-SoB. The model was also adjusted for sex, years of education, and age. Statistical parametric maps were corrected for multiple comparisons using Random Field Theory 32 with a voxel threshold of p < 0.001 and a cluster threshold of p < 0.05. In every brain voxel, the model was of the form: Next, we tested the same hypothesis in the ADNI database, examining main and interactive effects of [ 18 F]Florbetapir and [ 18 F]Flortaucipir on CDR-SoB. This model was also adjusted for sex, years of education, and age. Statistical parametric maps were corrected for multiple comparisons using Random Field Theory 32 with voxel threshold of p < 0.001 and a cluster threshold of p < 0.05. In every brain voxel, the model was of the form: In each cohort, we further tested the adequacy of the models the interaction terms using an analysis of variance, comparing the interaction model with each reduced model testing amyloid-PET plus tau-PET, as well as the inclusion of the interaction term. 33 Statistical analyses were repeated using MMSE score and rey auditory verbal learning test (RAVLT) delayed score as outcome measures. Statistical analyses were also repeated using global measures of amyloid-PET.

Results
Demographic and clinical information for both samples examined in this study are summarized in Table 1. Demographic comparisons between cohorts are reported in Table S1. CU individuals in TRIAD had higher baseline CDR-SoB scores than did with CU individuals in ADNI. Variance inflation factors (VIFs) for all variables were between 1 and 2, indicating that problematic levels of multicollinearity are not present in our analyses. 34 In the TRIAD cohort, no significant relationships between amyloid-PET SUVR and clinical function were observed (Fig. 1A). Higher tau-PET SUVR in medial temporal and inferior temporal cortices was associated with impaired clinical function (Fig. 1B). Voxel-level interactions between continuous measures of amyloid-PET and continuous measures of tau-PET in the medial prefrontal, dorsolateral prefrontal, anterior cingulate, posterior cingulate, and precuneus cortices were associated with impaired clinical function (Fig. 1C). In regions where the interaction term was significant, the main effects of amyloid-β and tau-PET SUVR on impaired clinical status were negligible. Results using MMSE and RAVLT delayed scores in the TRIAD cohort are displayed in Figures S1A-C In the ADNI cohort, no significant relationships between amyloid-PET SUVR and clinical function were observed ( Fig. 2A). Higher tau-PET SUVR in medial temporal, inferior temporal, and occipital cortices was associated with impaired clinical function (Fig. 2B). Voxel-level interactions between continuous measures of amyloid-PET and continuous measures of tau-PET in the medial prefrontal, orbitofrontal, superior frontal, anterior cingulate, posterior cingulate, and precuneus cortices were associated with impaired clinical function (Fig. 2C).  Similar to the results obtained in the TRIAD cohort, the main effects of amyloid-β and tau-PET SUVR on impaired clinical status were negligible in regions where the interaction term was significant. Results using MMSE and RAVLT delayed scores in the ADNI cohort are displayed in Figures S1D-F Figures S3, S4.
Analysis of variance supported that in both cohorts, the model including the interaction term best described the relationship between amyloid-PET, tau-PET, and clinical status ( Table 2). Summaries of statistical outcomes from both TRIAD and ADNI cohorts are reported in Table 3. A schematic of the topographical overlap of results obtained from TRIAD and ADNI cohorts is displayed in Figure 3.

Discussion
This study supports the hypothesis that amyloid-β contributes to clinical symptoms by potentiating taudependent cognitive dysfunction. In two cohorts, we observed that while amyloid-PET SUVR at the voxel level was not associated with cognitive dysfunction, tau-PET SUVR in the medial temporal lobes had a direct relationship with cognitive dysfunction. Moreover, in regions such as the medial prefrontal cortex, anterior cingulate, posterior cingulate, and precuneus cortices, amyloid-PET levels potentiated tau's relationship with clinical function. Taken together, these findings build on recent tau-PET studies in humans 35 to suggest that amyloid-β is associated with cognition by potentiating tau-dependent cognitive dysfunction.
In our study, the amyloid-independent effects of tau pathology on cognitive dysfunction were largely confined to the medial temporal lobes. Evidence from autopsy studies 36 as well as in vivo studies 15 suggests that the medial temporal lobes are a site of early tau aggregation. Crucially, the medial temporal regions are also regions in which amyloid-β plaques aggregate later in the course of AD, and in lower concentrations. 14,37 Correspondingly, it is plausible that the amyloid-independent effects of tau pathology on cognitive dysfunction are related to the lower concentrations of amyloid-β in these regions. Further supporting this idea is the finding that the amyloiddependent effects of tau pathology were observed in regions of the brain's default mode network, characterized by significant and early amyloid-β accumulation. 17,38,39 This study extends previous research conducted using CSF concentrations of phosphorylated tau which reported interactions between amyloid-β and tau concentrations  associated with cerebral metabolic dysfunction as well as longitudinal cognitive dysfunction. 33,40 Building on these studies, our study leverages the topographical information garnered by PET imaging to provide evidence of specific regional patterns of amyloid-dependent and amyloidindependent associations of tau with cognitive dysfunction, in which both the quantity and localization of amyloid-β modulate the effects of tau. Contemporary versions of the amyloid cascade hypothesis of AD posit that amyloid-β leads to AD through initiating a series of events including tau hyperphosphorylation, neuroinflammation, and neurodegeneration, among other events, eventually leading to cognitive dysfunction. 3 Our study contributes to this model by providing evidence that in addition to acting as a disease trigger, amyloid-β contributes to cognitive impairment through local interactions with tau pathology. The finding of deleterious interaction between amyloid-β and tau pathologies in humans is in line with studies from experimental animals in which molecular interactions between amyloid-β and tau peptides lead to synaptic 41 and neural circuit dysfunction. 42 Furthermore, cell culture studies have provided evidence that amyloid-tau interactions are associated with deficits in axonal transport 43 and exacerbate neuronal death. 44 Moreover, our results are in line with accepted AD biomarker models which suggest that amyloidosis alone is not sufficient for cognitive dysfunction. 45,46 However, our results contribute to this framework by providing evidence that amyloid-β may be more than a disease trigger: amyloid-β contributes to cognitive dysfunction through potentiating tau's effects on cognitive dysfunction.
Tau accumulation in the medial temporal cortex was associated with cognitive dysfunction. These results are in line with studies of individuals with primary age-related tauopathy (PART), characterized by medial temporal tau accumulation accompanied by mild cognitive dysfunction, in the absence of amyloid-β. 47 Individuals with PART, display slower rates of cognitive dysfunction 48 and rarely progress to dementia. 47 Taken together, these findings further support the role of amyloid-β in the cognitive dysfunction that characterizes AD.
From a therapeutic perspective, our findings highlight the possibility that anti-amyloid therapies may be beneficial to slow cognitive dysfunction in the symptomatic phase of AD by reducing amyloid-β potentiating of tau's pathological effects. However, based on the results presented in this study, anti-amyloid therapies may be less effective at slowing the progression of memory dysfunction mediated by medial temporal cortical regions. More studies directly assessing longitudinal cognitive decline, as well as memory decline specifically, are needed to further support this notion. As disease-modifying trials continue to shift toward earlier disease phases, 49 targeting amyloidβ before the appearance of tau pathology remains a promising strategy. 50 Finally, it is also important to consider the role of neurodegeneration in cognitive decline. Because neurodegeneration is considered to at least partially mediate associations between AD pathology and cognition, 51 future longitudinal studies including neurodegeneration biomarkers are needed.
While the majority of associations between PET measures and cognitive dysfunction were observed in both TRIAD and ADNI cohorts, some results were only observed in one cohort. For example, tau accumulation in occipital cortices was associated with cognitive dysfunction in ADNI, but not in TRIAD. Furthermore, in some areas such as the precuneus and lateral temporal cortices, little physical overlap was observed, but significant clusters were observed in the same region in both cohorts. It is conceivable that these differences in results are attributable to differences in the individuals enrolled in each study. More studies probing the specific nature of cognitive dysfunction may find that tau accumulation in these regions are associated with cognitive dysfunction in specific domains. 8 Methodological limitations should be considered when interpreting this study. The first is that the study is not designed to uncover biological mechanisms for the potentiation of tau's effects by amyloid-β. Moreover, TRIAD and ADNI are both research cohorts consisting of highly motivated individuals to participate in AD research and may not reflect the general population. Third, the spatial resolution of PET imaging places limitations on the capacity to describe molecular interactions; the colocalization of elevated amyloid-PET and tau-PET uptake within a voxel is not the same as identifying amyloid-tau interactions at the molecular level. However, our study builds on previous preclinical studies 41 to identify amyloid-β's potentiation of tau in living humans. Fourth, it is also plausible that cytoarchitectural differences of the medial temporal lobes, 52 rather than concentrations of amyloid-β, are related to differential vulnerability to amyloid-dependent versus amyloidindependent effects of tau reported in this study. Future experiments using preclinical models may shed light on this question. Methodological advantages of the study include large samples, replication in an independent multicenter study despite in baseline CDR-SoB scores, the use of continuous variables and replication of tau-PET results with both first-generation and secondgeneration radiotracers.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Figure S1. Regional associations between amyloid-β, tau, and MMSE in the TRIAD and ADNI cohorts. Figure S2. Regional associations between amyloid-β, tau, and RAVLT delayed recall in the TRIAD and ADNI cohorts. Figure S3. 3D scatter plot of the distribution of amyloid-PET and tau-PET on CDR Sum of Boxes. Figure S4. 3D scatter plot of main effects of medial temporal tau-PET on CDR Sum of Boxes. Table S1. Between-sample demographic comparisons.