Topographical Overlapping of the Amyloid-β and Tau Pathologies in the Default Mode Network Predicts Alzheimer’s Disease with Higher Specificity

Abstract

Background:

While amyloid-β (Aβ) plaques and tau tangles are the well-recognized pathologies of Alzheimer’s disease (AD), they are more often observed in healthy individuals than in AD patients. This discrepancy makes it extremely challenging to utilize these two proteinopathies as reliable biomarkers for the early detection as well as later diagnosis of AD.

Objective:

We hypothesize and provide preliminary evidence that topographically overlapping Aβ and tau within the default mode network (DMN) play more critical roles in the underlying pathophysiology of AD than each of the tau and/or Aβ pathologies alone.

Methods:

We used our newly developed quantification methods and publicly available neuroimaging data from 303 individuals to provide preliminary evidence of our hypothesis.

Results:

We first showed that the probability of observing overlapping Aβ and tau is significantly higher within than outside the DMN. We then showed evidence that using Aβ and tau overlap can increase the reliability of the prediction of healthy individuals converting to mild cognitive impairment (MCI) and to a lesser degree converting from MCI to AD. Finally, we provided evidence that while the initial accumulations of Aβ and tau seems to be started independently in the healthy participants, the accumulations of the two pathologies interact in the MCI and AD groups.

Conclusion:

These findings shed some light on the complex pathophysiology of AD and suggest that overlapping Aβ and tau pathologies within the DMN might be a more reliable biomarker of AD for early detection and later diagnosis of the disease.

Keywords

Alzheimer’s disease amyloid-β plaques default mode network mild cognitive impairment tau tangles

INTRODUCTION

Alzheimer’s disease (AD) is the most common cause of dementia in older adults and is associated with high morbidity and mortality [1]. An estimated five million people over the age of 65 in the US live with AD. This number is projected to rise to 13.8 million in the US and more than 130 million worldwide by 2050 [2]. Despite such high prevalence, the exact pathogenesis of AD is still under investigation. Small aggregates of extracellular Aβ peptide plaques in the form of oligomers and the deposition of the hyperphosphorylated form of tau protein in neural cytoplasm that forms a major component of neurofibrillary tangles, appears to be the two major proteinopathies involved in the pathogenesis of AD [3]. However, spatiotemporal discrepancies between these two proteinopathies are a key unresolved challenge in understanding the pathogenesis of AD. Spatially, amyloid-β (Aβ) is shown to be ubiquitously accumulated throughout the medial parietal, and prefrontal cortex where negligible tau pathology is found. Inversely, aggregation of tau pathology seems to follow a more region-specific pathway starting from the medial temporal lobe (MTL) and spreading to the inferior temporal gyrus where little or no Aβ accumulation has been reported [4 –7]. The temporal discrepancies are even more puzzling. While the amyloid cascade hypothesis of AD places accumulation of Aβ at the start of the AD pathophysiology about 15 years earlier than the onset of clinical symptoms at age 65 and older, there is now compelling evidence for the aggregation of tau pathology in the locus coeruleus and transentorhinal cortex in about 20%of the healthy population in their early adult lifespan (20 to 50 years) [8 –10]. In addition, while about 25%of the older population will never show any trace of Aβ in their brain (even at age 90 and higher), almost everyone at that age will have tau aggregation [8].

The default mode network (DMN) is a set of functionally interconnected brain regions whose activity consistently decreases during goal-oriented external tasks and/or increases during wakeful rest [11, 12]. The neurophysiological processes localized to DMN regions have been shown to be negatively task-responsive [13] as well as functionally connected [14, 15] independent of the type of the task being performed or even at rest. These two distinct characteristics of the DMN are often overlooked and considered to be a result of the same neurophysiological process [16, 17]. By measuring the DMN’s functional connectivity (FC) and task-evoked negative BOLD response (NBR) simultaneously, we have previously reported that the DMN has two dissociable functional levels that are topographically overlapping but have distinct roles in the functional architecture of this large-scale brain network [18]. Despite such ambiguity, numerous structural and functional measurements of the DMN (FC [19 –25], NBR [13 , 26–32], baseline hyper/hypometabolism [33 –35], and brain atrophy [36, 37]) have shown the DMN to be implicated in normal aging [19 , 27–29], the pre-clinical stage of AD [13 , 25], mild cognitive impairment (MCI) [20 , 31], and AD [16 , 32]. There is evidence that Aβ pathology reduces the FC of the DMN [38], whereas tau has negligible effect on DMN FC. Separately, tau pathology seems to disrupt NBR in the DMN but has negligible effects on the FC of the same regions [38]. Naturally, when both Aβ and tau spatially coincide within the DMN, in a double-insult, they disrupt the normal operations of both functional levels of the DMN, causing a complete breakdown of the network, which potentially can initiate a series of events resulting in AD. In contrast, a solo and/or spatially non-overlapping accumulation of Aβ and tau has milder consequences on the DMN functionality. Our hypothesis, if true, will provide the underlying reason for the weak relationship between the Aβ deposition and cognition [39], whereas its deposition level is shown to be strongly related to the FC of the DMN [21 , 25]. This is because, according to our hypothesis, the overlapping of the Aβ and tau pathologies is needed to breakdown the network and consequently cause any detectable decline in the cognitive performance.

In this study, we used our recently developed and validated quantification method and partial volume correction technique [40] for reconstructing Aβ and tau PET scans found in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database to provide preliminary evidence for our introduced double-insult hypothesis. By using our quantifying techniques, we can accurately measure the extent of Aβ and tau accumulation on the surface of the cerebral cortex of cognitively normal, MCI, and AD patients with an unprecedented spatial resolution (almost on a millimeter-by-millimeter scale). Achieving such spatial resolution is essential for accurately measuring the spatial overlap of the regions with both Aβ and tau accumulation. We first show that in AD patients the probability of observing spatially coinciding Aβ and tau pathologies within the DMN is much greater than in the rest of the brain. Then, we show the superior reliability of overlapping pathologies in the DMN for predicting the conversion from healthy control (HC) to MCI, and to a lesser degree from MCI to AD. Finally, we utilized Aβ-tau inter-regional correlation in different groups of subjects to show the increase of the association between these two pathologies from the HC to MCI stage, and from the MCI to AD stage. This evidence suggests that the overlapping tau and Aβ pathologies in the DMN regions might be a more reliable biomarker for early detection of AD, as well as for accurate diagnoses of the disease.

METHODS

Participants

Data used in the preparation of this article were obtained from the ADNI database (http://adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of the ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. To be able to compute the overlapping accumulation of tau and Aβ, we need the subjects to have both Aβ and tau scans as well as a high-resolution structural T₁-weighted scan being acquired within one year. We have identified 159 healthy controls (12 of them are the subjects who progress to MCI, healthy control converters), 127 MCI (14 of them are the subjects who progress to AD, MCI converters), and 17 AD participants who fit our requirements. All participants gave their written consent to participate in the ADNI repository, and all imaging protocols were approved by the local institutional review board of the site where the data were collected. The MCI converter patients were converted to AD between six and 36 months. AD patients had a Mini-Mental State Examination (MMSE) score of 20–26, a Clinical Dementia Rating (CDR) of 0.5 or 1.0, and met the National Institute of Neurological and Communicative Disorders and Stroke and the AD and Related Disorders Association (NINCDS/ADRDA) criteria for probable AD. MCI patients had MMSE scores between 24 and 30, a memory complaint, objective memory loss measured by education adjusted scores on Wechsler Memory Scale (WMS) Logical Memory (LM) II, a CDR of 0.5, absence of significant levels of impairment in other cognitive domains, essentially preserved activities of daily living, and an absence of dementia. The normal subjects were non-depressed, non-MCI, and non-demented, and had an MMSE score of 24–30 and a CDR close to zero.

Structural image reconstructions

Details of the MRI acquisition parameters can be found in the Supplementary Material. We used a newest version of the Freesurfer (Version 7) to re-process all the T1-weighted structural scans. Freesurfer [41] is an automated segmentation and cortical parcellation software package to reconstruct T1-weighted structural scans [42, 43]. All participants’ cortical surfaces are visually inspected/corrected by a demographic-blind technician. In the case of discrepancy, manual editing of the white and gray matter borders was conducted per the FreeSurfer manual editing guidelines [41]. Freesurfer segments the cortex into 33 different gyri/sulci-based regions in each hemisphere according to the Desikan-Killiany atlas [44] and calculates the cortical thickness at each vertex, which is at the millimeter-by-millimeter resolution. The maps produced can detect submillimeter differences between groups. In addition, the sub-cortical regions are also segmented using a Bayesian classification technique to give 37 subcortical regional masks. The vertex-wise data are not constrained to the pre-defined region of interest (ROI) and can be transferred to standard space using surface-based non-linear registration. To detect effects that are smaller in size or span over two or more pre-defined ROIs, utilization of the vertex-wise data is essential. The Freesurfer cortical and subcortical regions, as well as vertex-wise surface reconstruction, will be used to perform native space analysis of the PET data. Finally, all the vertex-wise quantification of the Aβ, tau, and their overlap, as well as the surface-based probabilistic atlases will be done using Freesurfer reconstructed surfaces.

Quantification process for PET data

Details of the PET acquisition parameters can be found in the Supplementary Material. An in-house developed, fully automatic quantification method has already been implemented and evaluated using histopathological data [40] and used in numerous published studies [34 , 45–47] for the quantification of PET scans. Briefly, it starts by aligning dynamic PET frames (six frames in tau-PET and four frames in amyloid-PET) to the first frame using rigid-body registration and averaging them to generate a static PET image. The PET image is then registered with the CT and merged to obtain a composite image in the PET static space. Each individual’s structural T1 image in FreeSurfer space is also registered to the same participant’s CT/PET composite image using normalized mutual information and six degrees of freedom to obtain a rigid-body transformation matrix to transfer all FreeSurfer regional masks to static PET image space. These regional masks in static PET space are used to extract the regional PET data. The standardized uptake value (SUV), defined as the decay-corrected brain radioactivity concentration normalized for injected dose and body weight, is calculated at selected regions, and then normalized to cerebellum gray matter to derive the standardized uptake value ratio (SUVR). The SUVR will be determined at both the voxel and ROI levels.

The substantial and strong spill-in signal from white matter’s non-specific binding in Florbetapir is addressed by discarding the uptake in the gray matter voxels located adjacent to white matter volume both in the cerebellum for computing the reference region uptake, and in the cerebral cortex for obtaining cortical regions’ SUVR. For AV1451, non-specific bindings occur mainly in the meninges; thus, the same approach is used to discard all voxels adjacent to the surface of the gray matter. For the surface reconstruction of the PET data, we sampled the Florbetapir on the gray matter surface and AV1451 on the white matter surface to address the spill-in issue from non-specific bindings. While these remedies helped compute the regional SUVRs, the vertex-wise SUVRs, which are needed for obtaining the topographical overlapping of the two pathologies, were still significantly demonstrating a strong spill-in signal. To completely address the issue, we developed a partial volume correction (PVC) method that took advantage of the available scans from young and healthy participants and removed the spill-in signal completely from the PET scans.

Partial volume correction

We developed simple but effective anatomy-driven partial volume correction. The details of the PVC method can be found elsewhere, but briefly, each gray matter voxel’s uptake is a combination of actual binding in that location and the spill-in from white-matter/meninges non-specific binding. Using PET scans from 35 young (< 30 years) and healthy participants we were able to estimate the white-matter/meninges spill-in signal. Since extremely young and healthy participants are not supposed to have any Aβ/tau accumulation, any gray matter uptake in these subjects can be considered a pure result of the spill-in from adjacent regions with non-specific binding. In each young subject, we used the white-matter/meninges mask to extract the spatial distribution of the non-specific binding within the white-matter/meninges region. Convolving this mask with the scanner residual point-spread function simulates the effect of the spill-in signal perfectly, since within the gray matter regions the correlation between actual uptake and the synthesized spill-in was more than 80%. Then we fitted a linear regression model for each voxel to predict the gray matter spill using the synthesized spill-in. Finally, using the fitted model parameters at each voxel, we estimated the gray matter spill-in given the synthesized spill-in and subtracted it from the actual gray matter uptake to completely remove its artifacts.

Topographical overlapping of Aβ and tau

To obtain topographical overlapping of the Aβ and tau, and to generate a probabilistic atlas for tau, and Aβ uptake as well as their overlap, PET data must be transferred to standard space (MNI152). For vertex-wise analysis, we project the surface-based reconstructed Aβ and tau to the surface of the MNI152 using spherical surface registration in FreeSurfer. Then, setting a threshold (SUVR > 1.3), we delineated vertices that have significant uptake of Aβ, tau, or both pathologies. The subject-wise binary masks of the delineated vertices not only give the extent and expression (level) of each pathology as well as their overlap, but they can also be used to generate the probabilistic atlas for each pathology and their overlap. To quantify the extent and expression of each pathology within the DMN, we combined 10 bilateral Freesurfer anatomical masks/labels (middle temporal, inferior parietal, lateral orbito-frontal, superior frontal, medial orbito-frontal, rostral middle frontal, precuneus, posterior cingulate, isthmus cingulate, and hippocampus) to generate a single binary mask to distinguish the vertices that fall within the DMN.

Aβ-tau inter-regional correlation

Using our in-house developed quantification method, we obtained the regional Aβ and tau SUVR in subjects’ native space which have been shown to be much more accurate than conventional standard space methods [40]. We then computed the subject-wise Pearson correlation coefficient between each cortical region’s tau SUVR and Aβ SUVRs of every cortical region given by Freesurfer (74 regions in total). Computing all possible inter-regional correlations between Aβ and tau gives a cross-correlogram for each group of participants which will be visualized using color-coded heat-map. We repeat the analysis for three different groups of participants: 1) HC that have not been converted to MCI in the next 36 months, 2) MCI patients that have not been converted to AD in the next 36 months, and 3) AD patients. It is noteworthy that we did not use any threshold for the regional SUVRs before computing the inter-regional correlations. The significant differences between groups’ inter-regional correlations are determined with an analysis of variance (ANOVA) test.

Statistical analysis

All statistical analyses and their visualization in this study are performed using Python and its main numeric modules numpy and matplotlib. The Student’s t-tests, ANOVA, and Chi-square tests are performed using SciPy statistical package (v6.1.1).

RESULTS

Table 1 lists the number and demographic of the subjects used in this study for all three groups (HC, MCI, AD). Our subject selection was only based on the availability of the imaging data in the ADNI dataset within less than one year between data acquisitions in each modality and successful processing and analyzing the imaging data.

Table 1

Demographics of the subjects used in the study

	HC-NC^a	HC-C^b	MCI-NC^c	MCI-C^d	AD^e	p
Number	147	12	113	14	17
Male/Female	60/87	5/7	68/45	6/8	12/5	0.01^f
Age (Mean±STD)	74.56±7.58	83.33±4.52	74.64±7.16	75.32±8.71	74.92±9.36	0.004^g

^aHealthy Controls –Non-Converters/ ^bHealthy Controls –Converters/ ^cMild Cognitive Impairment –Non-Converter/ ^dMild Cognitive Impairment –Converter/ ^eAlzheimer’s Disease/ ^fChi-square test/ ^gAVOVA test.

ANOVA test revealed that there is a significant difference in the age of the five groups listed in Table 1, the post hoc t-test determined that only the age of the converters from HC to MCI are significantly higher than the rest of the group. There were significantly more male participants in the MCI and AD groups whereas the number of females was higher in both converters as well as HC groups.

Using 17 AD patients’ data in our dataset, we first show that the probability of observing overlapping tau and Aβ pathologies within the DMN is significantly greater than observing it outside the DMN, hinting at their more influential role in the pathophysiology of the AD. If the overlapping Aβ and tau pathologies within the DMN are the true underlying pathophysiology of the AD, then all AD patients should have it. To show this, we first generated a vertex-wise probabilistic atlas for the AD group where the value at each vertex gives the probability of observing overlapping Aβ and tau in that group. Then, simply comparing the probabilities of the DMN vertices with the rest of the brain will determine whether observing the overlapping pathologies within the DMN is more likely than outside the DMN. Figure 1 shows the results of this examination. Figure 1a shows the probabilistic atlas of observing the overlapping pathologies obtained from the AD group and overlaid over a semi-inflated cerebral cortex surface reconstructed from the MNI152 template. The probabilities are color-coded with hot colors where darker-red and red colors indicate lower probabilities and bright-red and yellow indicate higher probabilities. As it is seen, the probability of observing overlapping Aβ and tau pathology is much higher in the DMN regions in comparison to the rest of the brain. Figure 1b quantifies this probabilistic atlas using a boxplot and a Student’s t-test for determining the significant differences. Each boxplot illustrates the distribution of the probabilities for vertices that fall within the DMN and compare it with the one that falls outside the DMN. Two hemispheres are plotted separately. Both right and left hemispheres show a significantly (right; t = 132.73, p < 0.0; left: t = 112.20, p < 0.0) higher probabilities of observing overlapping pathologies within the DMN than outside the DMN, providing preliminary evidence that overlapping pathology within the DMN plays a more significant role in the underlying pathophysiology of the AD.

Fig. 1

(a) The vertex-wise probabilistic atlas of observing overlapping pathologies throughout the entire cerebral cortex obtained from 17 AD patients. The probabilities at each vertex are color-coded with hot colors and overlaid over a semi-inflated cortical surface of the MNI152 template. (b) boxplots depict the distribution of the probabilities for vertices that fall within the DMN as well as the vertices that fall outside the DMN separately for each hemisphere. The significant of the differences are assessed using student t-test and reported for each hemisphere separately.

Next, we show that in comparison to Aβ, tau, and both pathologies, the topographical overlapping pathology within the DMN has more predictability power to identify individuals converting from HC to MCI as well as from MCI to AD. We first start by creating a probabilistic atlas for Aβ and tau pathology as well as their overlap for the HC and MCI groups. In addition, we generate a separate probabilistic atlas for individuals who are converting from HC to MCI and from MCI to AD. Figure 2 depicts these probabilistic atlases. The top row shows the probabilistic atlas for Aβ accumulation in the HC (left) and MCI (right) groups along with their converter which is depicted underneath. The middle row shows the same for tau accumulation and the bottom row illustrates the same for overlapping Aβ and tau. It is clear, even visually, that overlapping Aβ, and tau has more specificity for predicting the converters to MCI. While both Aβ and tau have very high sensitivity (true positive) for detecting the converters, they have extremely poor specificity (true negative) to identify non-convertors. Note almost no overlapping pathology for HC non-converters and very minimal overlapping pathology in the MCI converters, suggesting the high specificity of the overlapping pathology to predict conversion from HC to MCI as well as from MCI to AD.

Fig. 2

Vertex-wise probabilistic atlas for Aβ (top row), tau (middle row), and their overlap (bottom row) for 159 subjects in HC group (left) and 127 participants in MCI (right) group. The converters to MCI and AD are illustrated underneath each probabilistic atlas. The probabilities at each vertex are color-coded with heatmap and overlaid over a semi-inflated cortical surface of the MNI152 template.

To quantify the findings in Fig. 2, we use ROC curves. Briefly, we set different threshold for the extent of Aβ, tau, and their overlap to detect the converters, and compute the true-positive and false-positive rate of each classification and plot the true-positive rates in terms of false-positive rates. Figure 3 shows the resultant ROC curves for (a) predicting the HC to MCI conversion using the whole brain and (c) considering only the DMN regions; (b) predicting MCI to AD conversion using the whole brain, and (d) using only DMN accumulation. As it seen in Fig. 3a, the predictability of the Aβ, tau, and their overlap changes significantly across different ranges of the false-positive rate. For higher true-positive/false-positive rate the overlapping Aβ and tau outperforms the predictability of both Aβ and tau, whereas for the lower true-positive/false-positive rates its superiority reduces to level lower than tau predictability. We speculate that this loss of predictability in the lower range of true-positive/false-positive rate is due to massive accumulation of Aβ and tau in most of our participants. The same behavior can also be seen in the MCI to AD conversion as well as when we concentrate only to the DMN regions. The only difference is that, for MCI participants the outperformance of the overlapping Aβ and tau reduces substantially to the same level of tau and Aβ. Again, we argue this is mainly because these subjects already passed the disease initiation stage where we can accurately detect AD pathophysiology and subsequently predict the conversion.

Fig. 3

ROC curves obtained for predicting conversion (a and c) from HC to MCI and (b and d) from MCI to AD using the extent of the regions with accumulation of Aβ (SUVR > 1.3), Tau (SUVR > 1.3), and their overlap (a and b) throughout the entire cerebral cortex and (c and d) when they are bounded to the DMN regions. As shown, the prediction using the extent of the overlapped pathologies outperforms the prediction using the extent of either of the pathology alone or overlapping outside DMN regions. The prediction accuracy reduces in the lower range of the true-positive/false-positive rates, but it is always higher than predicting using only Aβ extent.

We also demonstrate that the probability of observing overlapping pathology within DMN is much higher than the rest of the brain in HC and MCI converters, suggesting the more influential effect of overlapping Aβ and tau when it occurs within DMN. Figure 4 shows the results of this analysis using boxplots and the Student’s t-test. As is seen, the probability of observing overlapping Aβ and tau in the DMN is significantly higher when it is compared with the rest of the brain in both participants converting from HC to MCI (t = 79.90, p < 0.0) and converters from MCI to AD (t = 227.14, p < 0.0). This result suggests that the overlapping Aβ and tau in the DMN is more influential for conversion from HC to MCI or from MCI to AD.

Fig. 4

Boxplots illustrate the distribution of the vertex-wise probabilities of observing both Aβ and tau for vertices within the DMN and in comparison, with distribution of the probabilities of vertices falling outside DMN for participants who converting from HC to MCI and from MCI to AD. The significant of the differences are assessed using student t-test and reported for each hemisphere separately.

Finally, inter-regional correlations between Aβ and tau uptake across different groups of subjects (HC, MCI, and AD) are depicted in Fig. 5. The x-axis and y-axis represented the 74 cortical regions. Each element of the cross-correlogram represents the subject-wise correlation between x-axis region’s Aβ uptake and y-axis region’s tau uptake which is color-coded with heatmap. Red color indicates highest correlation value and blue color indicates zero correlation. Based on Fig. 5a, it is apparent that in HC subject’s correlations are weak between the two pathologies hinting at their independently initiated depositions in different brain regions. On the other hand, we can see the increase in the inter-regional correlations in the MCI group (Fig. 5b) in comparison with the HC group. Moreover, in Fig. 5c, strong correlations between Aβ and tau pathologies in several brain regions can be seen in the AD group. In Fig. 5d, each boxplot illustrates the distribution of the regional correlation values between different groups of subjects (HC, MCI, and AD). ANOVA test reveals that there is a significant difference in correlation values of these three groups (f = 10528.14, p < 0.0).

Fig. 5

Inter-regional cross-correlogram between Aβ and tau accumulations for (a) HC, (b) MCI, and (c) AD group. Each color-codded cell presents the subject-wise correlation between crossing regions’ Aβ (vertical axis) and tau (horizontal axis) accumulations. (d) boxplots depict the distribution of the Aβ-tau subject-wise inter-regional correlation values for 74 cortical regions in different groups. The significant of the differences are assessed using ANOVA.

DISCUSSION

We have introduced the double-insult hypothesis in this work, where instead of Aβ and tau pathology alone, the interaction of these two pathologies within the DMN plays the critical role in the underlying pathophysiology of AD. Using a publicly available dataset (ADNI) and available in vivo imaging data of the two pathologies (303 participants; 17 AD, 127 MCI, and 159 HC) which were quantified using our newly developed and evaluated technique, we provided preliminary evidence in support of our hypothesis. We first showed that, among 17 AD patients, the probability of observing overlapping Aβ and tau pathologies is significantly higher within the DMN in comparison to the rest of the brain. Next, we showed that at least in the high true-positive/false-positive range, the true positive-rate of detecting conversion from HC to MCI as well as from MCI to AD has significantly increased. We argued that the reason such superiority is diminishing in the lower range of the true-positive/false positive rate is because of massive Aβ and tau uptake in some of the non-converter participants, suggesting that there might be more converters in the HC group that have yet to be clinically diagnosed with AD. Finally, we used Aβ-tau inter-regional correlation in different groups of subjects to show that the interaction between these two pathologies increases with severity of dementia. Future work is required to test our hypothesis on a younger population with a lower level of Aβ and tau accumulation.

AD is a progressive neurodegenerative disease that is characterized by the presence of Aβ plaques and tau tangles [48, 49]. PET studies have been shown that interaction between Aβ and tau is associated with brain dysfunction, changing of the brain atrophy, and cognitive disorders [50 –53]. Since these two pathologies can be initiated independently at different times and regions of the brain [3], how the accumulation of these two individual proteinopathies together causes enough neurological damage to lead to AD is an unresolved challenge. Recent report suggests that at some point the independently initiated tau and Aβ start interacting with each other, which subsequently accelerates not only the deterioration in their associated brain measurements but also the spread of the pathologies to other regions [3, 54]. Previous longitudinal tau PET work has also shown that tau increased faster in high-Aβ than in low-Aβ HC individuals [55]. Thus, we hypothesize that it is the effects of this coupling between the two misfolded proteins that could potentially initiate a series of events that eventually results in significant neurodegeneration and cognitive decline. To understand the coupling effect, in the current study for the first time, we examined the conversion probability of a relatively healthy participant to MCI, or even an MCI participant to AD, when one proteinopathy or both are individually present in the DMN and compared that with the probability of observing the same pathology outside the DMN. In this study, we aimed at exploring the spatial distribution and overlap of tau with Aβ pathologies, to determine and effectively predict the conversion of the examined cognitively healthy subjects to MCI, and MCI subjects to AD. We have shown that individuals with topographically overlapping Aβ and tau accumulation within the DMN regions are more likely to convert to MCI or AD than the subject that has equivalent levels of, but non-overlapping, pathologies.

Both HC and MCI participants who converted to MCI and AD, respectively, had higher probabilities of spatial overlap inside the DMN regions. Therefore, the presence of overlap between tau and Aβ depositions inside the DMN region of the MCI and HC subjects can predict conversion to AD, whereas the presence of only tau or Aβ in the DMN regions is not a powerful predictive factor of the conversion. This evidence suggests some modifications to the previous studies that hypothesized that Aβ serves as an initiator of a pathogenic cascade that affects the spread of tau aggregation [56], as well as studies suggesting that the presence of Aβ accumulation is required for the appearance of high-grade cortical tau pathology [57 –59]. Aβ plaques with associated tau pathology correlate more closely with neuronal loss and dementia in AD than either Aβ alone or tau alone pathologies [60]. Thus, Aβ deposition may cause a mild, but progressive, pathological state before neurodegeneration, and tauopathy may act as a trigger for severe neurodegeneration [61, 62]. It should be emphasized that while double-insult hypothesis highlights the importance of the overlapping Aβ and tau pathologies in the initiation of the disease and its progression afterward, the underlying process by which the overlapping Aβ and tau pathology starts interacting and whether or not Aβ is the accelerator for accumulation of tau and/or there are other intervening factors is still under investigation. For instance, some studies have shown preliminary evidence that mild neuroinflammation might be the missing but key factor in the Aβ and tau relationship [63 –69].

We used ROC curves to find the predictive factor of conversion by comparing the converters and non-converters. What this experiment demonstrated was that the extents of the tau and Aβ pathologies in the brain does not reliably predict that the person will convert to MCI and AD. It was shown in Fig. 3 that regional coinciding of Aβ and tau pathologies results in a better performance in such prediction within higher true-positive/false-positive range. But in lower true-positive/false-positive range almost all brain regions were contaminated and distinguishing the converters subjects becomes challenging. It is also noteworthy that we used subjects who converted between six to 36 months after the acquisition date of the Aβ and tau scans. Thus, patients who may have converted to MCI and AD in a later time (more than 36 months) will not have a similar brain pathology at baseline compared to non-converter HC and AD patients, respectively. On the other hand, brain pathology of those converter patients who converted to MCI and AD in a shorter time (e.g., at six months) may be similar to that of the MCI and AD subjects, respectively. Since AD pathologies shown to be initiated decades before onset of the disease, sampling a population at the later adult life span essentially precluded ADNI to capture the full dynamic of the AD pathologies and its relationship with other brain biomarkers as well as cognitive and behavioral measurements [70]. In addition, ADNI neuroimaging datasets were acquired with relatively older MRI technologies and the first generation of PET tracers for Aβ and tau imaging. Therefore, its performance in predictions tasks is likely to be poor. Still, the results in this manuscript provide preliminary evidence for our hypothesis that the key component of the conversion is not the two pathologies, but rather is the overlap of the two pathologies and the location of their overlap. Finally, in our analysis the HC converters were significantly older than the other three groups, which could be considered a contributing factor to our findings. Thus, further experiment and analysis are warranted to provide more evidence for the validity of the double-insult hypothesis in the future. Such comprehensive validation should take advantage of the state-of-the-art imaging methods and technologies to target specifically the double-insult hypothesis using other brain and behavioral outcome as the early biomarker of the disease.

Both FC [16 , 71] and NBR [13 , 32] in DMN regions have been reported to be disrupted with normal aging [19 , 27–29], the pre-clinical stage of AD [13 , 25], MCI [24 , 71], and AD [13 , 32]. Existing reports show that early deposition of Aβ plaques tends to be concentrated in some DMN regions [38, 72]. Furthermore, while there is clear evidence that accumulation of Aβ causes disruption in the DMN spontaneous activity measured by FC [19 , 25], the effect of tau on DMN functioning is not clear [52, 73]. However, there are limited studies reporting that hypermetabolism of the DMN regions seems to be altered by the early accumulation of tau in the medial temporal lobe [74]. Our recent findings show that the DMN has two dissociable functional levels that are topographically overlapping but have distinct roles in the functional architecture of this large-scale brain network [75]. These findings, coupled with the experiment outlined in this study, indicate that overlapping accumulation of Aβ and tau within the DMN can interfere with functionalities of both levels causing more severe cognitive decline. Therefore, overlapping pathologies has more predictability power than the same level of non-overlapping Aβ and tau pathologies.

We used the inter-regional correlation analysis to provide further evidence for the double-insight hypothesis and the relation between the interacting Aβ and tau pathologies and severity of dementia. What this analysis presented was the fact that in the AD, the Aβ and tau pathology distribution has stronger inter-regional correlation. In the current study, our results showed that there was a steady increase in the inter-regional correlations of Aβ and tau uptakes from HC to AD groups. This is an important consideration because it suggests that Aβ and tau pathologies commenced independently in different regions of the brain [38]; and then start interacting strongly during the early stage of the disease [76, 77]. While a subtle cognitive using student t-test and reported for each hemisphere separately may be associated with Aβ [78], it is the overlapping tau accumulation that does correlate strongly with cognitive impairment [35 , 79]. Despite of the fact that existence Aβ and tau pathologies have influence on the progression of different disease stages [60], interaction between these two pathologies may have a resonance effect on progression of AD.

The pathophysiology of AD is still unclear, and there is an essential need to establish methods for predicting the progression of normal and MCI subjects to AD. This method will be clinically useful in detecting high-risk populations when they have neuropathological changes in their brain before their clinical symptoms of AD appear. Early detection of AD brain pathologies is of particular concern when new treatment methods have higher efficacy on early detected cases of AD.

CONCLUSION

The present study demonstrates that topographically overlapping pathologies, and not the existence of each AD proteinopathy alone, is a more accurate biomarker of AD. Thus, it is a more reliable predictor of a cognitively normal or MCI subject converting to an AD patient when the overlap is within the DMN regions of the brain. We considered the various evidence of brain changes that have been observed in different stages of AD to provide a perspective for supporting our hypothesis on effect of overlapping accumulation of Aβ and tau in initiation of AD pathophysiology. Based on our previous findings, we also suggested an underlying mechanism that may explain how the topographically overlapping pathologies in the DMN generate such a severe effect on brain and consequently on cognition. We conclude that the overlapping between these two pathologies has potential to be a hallmark AD prognosis. These findings shed light on the underlying pathophysiology of AD and pave the way to introduce a more accurate tool for the early detection of AD, and a diagnosis path that can be investigated in future studies.

Footnotes

ACKNOWLEDGMENTS

We would like to thank Cynthia Fox for carefully editing this manuscript. This work is made possible with the funding support from NIH/NIA under R01 AG057962 grant award.

Data collection and sharing for this project were funded by the Alzheimer’s Disease Neuroimaging Initiative (National Institutes of Health Grant U01AG024904). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; Bio Clinical, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson &Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neuro track Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

Authors’ disclosures available online ().

INFORMED CONSENT,DATA ACCESS,AND CONFIDENTIALITY OF THE DATA

According to ADNI protocol, “Informed consent was obtained in accordance with US 21 CFR 50.25, the TriCouncil Policy Statement: Ethical Conduct of Research Involving Humans and the Health Canada and ICH Good Clinical Practice.

Patient confidentiality was strictly held in trust by the participating investigators and research staff. This confidentiality was extended to cover testing of biological samples and genetic tests in addition to the clinical information relating to participants. All data will be transmitted securely via the Internet to ATRI at USC. Access is granted to study team members based on role. Data transmission will occur through a secure internet connection-https (hypertext transfer protocol secured) at 128-bit SSL. The study protocol, documentation, data and all other information generated will be held in strict confidence. No information concerning the study or the data will be released to any unauthorized third party, without prior written approval of the sponsoring institution. All ADNI data are shared without embargo through the LONI Image and Data Archive (IDA), a secure research data repository. Interested scientists may obtain access to ADNI imaging, clinical, genomic, and biomarker data for the purposes of scientific investigation, teaching, or planning clinical research studies. Access is contingent on adherence to the ADNI Data Use Agreement and the publications’ policies outlined in the documents listed below. Note: documents are subject to updates by ADNI. The application process includes acceptance of the Data Use Agreement and submission of an online application form. The application must include the investigator’s institutional affiliation and the proposed uses of the ADNI data. ADNI data may not be used for commercial products or redistributed in any way.

The supplementary material is available in the electronic version of this article: .

References

Ballard

, Gauthier

, Corbett

, Brayne

, Aarsland

, Jones

(2011) Alzheimer’s disease. Lancet 377, 1019–1031.

Alzheimer’s Association (2020) 2020 Alzheimer’s disease facts and figures. Alzheimers Dement 16, 391–460.

Ehrenberg

, Khatun

, Coomans

, Betts

, Capraro

, Thijssen

, Senkevich

, Bharucha

, Jafarpour

, Young

PNE

, Jagust

, Carter

, Lashley

, Grinberg

, Pereira

, Mattsson-Carlgren

, Ashton

, Hanrieder

, Zetterberg

, Schöll

, Paterson

(2020) Relevance of biomarkers across different neurodegenerative diseases. Alzheimers Res Ther 12, 56.

Hansson

, Grothe

, Strandberg

, Ohlsson

, Hägerström

, Jögi

, Smith

, Schöll

(2017) Tau pathology distribution in Alzheimer’s disease corresponds differentially to cognition-relevant functional brain networks. Front Neurosci 11, 167.

Vogel

, Young

, Oxtoby

, Smith

, Ossenkoppele

, Aksman

, Strandberg

, La Joie

, Grothe

, Medina

, Rabinovici

, Alexander

, Evans

, Hansson

(2020) Spatiotemporal imaging phenotypes of tau pathology in Alzheimer’s disease. Alzheimers Dement 16, e045612.

Dronse

, Fliessbach

, Bischof

, von Reutern

, Faber

, Hammes

, Kuhnert

, Neumaier

, Onur

, Kukolja

, van Eimeren

, Jessen

, Fink

, Klockgether

, Drzezga

(2017) In vivo patterns of tau pathology, amyloid-β burden, and neuronal dysfunction in clinical variants of Alzheimer’s disease. J Alzheimers Dis 55, 465–471.

Farrell

, Jiang

, Schultz

, Properzi

, Price

, Becker

, Jacobs

HIL

, Hanseeuw

, Rentz

, Villemagne

, Papp

, Mormino

, Betensky

, Johnson

, Sperling

, Buckley

(2021) Defining the lowest threshold for amyloid-PET to predict future cognitive decline and amyloid accumulation. Neurology 96, e619–e631.

Arnsten

, Datta

, Tredici

, Braak

(2021) Hypothesis: Tau pathology is an initiating factor in sporadic Alzheimer’s disease. Alzheimers Dement 17, 115–124.

Duyckaerts

(2011) Tau pathology in children and young adults: Can you still be unconditionally baptist? Acta Neuropathol 121, 145–147.

10.

Braak

, Del Tredici

(2011) The pathological process underlying Alzheimer’s disease in individuals under thirty. Acta Neuropathol 121, 171–181.

11.

Raichle

, MacLeod

, Snyder

, Powers

, Gusnard

, Shulman

(2001) A default mode of brain function. Proc Natl Acad Sci U S A 98, 676–682.

12.

Shulman

, Fiez

, Corbetta

, Buckner

, Miezin

, Raichle

, Petersen

(1997) Common blood flow changes across visual tasks: II. Decreases in cerebral cortex. J Cogn Neurosci 9, 648–663.

13.

Pihlajamaki

, K

, Bertram

, Tanzi

, Dickerson

, Blacker

, Albert

, Sperling

(2010) Evidence of altered posteromedial cortical FMRI activity in subjects at risk for Alzheimer disease. Alzheimer Dis Assoc Disord 24, 28–36.

14.

Andrews-Hanna

, Reidler

, Sepulcre

, Poulin

, Buckner

(2010) Functional-anatomic fractionation of the brain’s default network. Neuron 65, 550–562.

15.

Greicius

, Krasnow

, Reiss

, Menon

(2003) Functional connectivity in the resting brain: A network analysis of the default mode hypothesis. Proc Natl Acad Sci U S A 100, 253–258.

16.

Greicius

, Srivastava

, Reiss

, Menon

(2004) Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: Evidence from functional MRI. Proc Natl Acad Sci U S A 101, 4637–4642.

17.

Sperling

, Laviolette

, O’Keefe

, O’Brien

, Rentz

, Pihlajamaki

, Marshall

, Hyman

, Selkoe

, Hedden

, Buckner

, Becker

, Johnson

(2009) Amyloid deposition is associated with impaired default network function in older persons without dementia. Neuron 63, 178–188.

18.

Parker

, Razlighi

(2019) Task-evoked negative BOLD response and functional connectivity in the default mode network are representative of two overlapping but separate neurophysiological processes. Sci Rep 9, 1–17.

19.

Andrews-Hanna

, Snyder

, Vincent

, Lustig

, Head

, Raichle

, Buckner

(2007) Disruption of large-scale brain systems in advanced aging. Neuron 56, 924–935.

20.

Bai

, Zhang

, Yu

, Shi

, Yuan

, Zhu

, Zhang

, Qian

(2008) Default-mode network activity distinguishes amnestic type mild cognitive impairment from healthy aging: A combined structural and resting-state functional MRI study. Neurosci Lett 438, 111–115.

21.

Hedden

, Van Dijk

, Becker

, Mehta

, Sperling

, Johnson

, Buckner

(2009) Disruption of functional connectivity in clinically normal older adults harboring amyloid burden. J Neurosci 29, 12686–12694.

22.

Mormino

, Smiljic

, Hayenga

, Onami

H.S

, Greicius

, Rabinovici

, Janabi

, Baker

, Yen

V.I

, Madison

(2011) Relationships between beta-amyloid and functional connectivity in different components of the default mode network in aging. Cereb Cortex 21, 2399–2407.

23.

Razlighi

, Habeck

, Steffener

, Gazes

, Zahodne

, Mackay-Brandt

, Stern

(2014) Unilateral disruptions in the default network with aging in native space. Brain Behav 4, 143–157.

24.

Rombouts

, Barkhof

, Goekoop

, Stam

, Scheltens

(2005) Altered resting state networks in mild cognitive impairment and mild Alzheimer’s disease: An fMRI study. Hum Brain Mapp 26, 231–239.

25.

Sheline

, Raichle

(2013) Resting state functional connectivity in preclinical Alzheimer’s disease. Biol Psychiatry 74, 340–347.

26.

Celone

, Calhoun

, Dickerson

, Atri

, Chua

, Miller

, DePeau

, Rentz

, Selkoe

, Blacker

, Albert

, Sperling

(2006) Alterations in memory networks in mild cognitive impairment and Alzheimer’s disease: An independent component analysis. J Neurosci 26, 10222–10231.

27.

Daselaar

, Prince

, Cabeza

(2004) When less means more: Deactivations during encoding that predict subsequent memory. Neuroimage 23, 921–927.

28.

Lustig

, Snyder

, Bhakta

, O’Brien

, McAvoy

, Raichle

, Morris

, Buckner

(2003) Functional deactivations: Change with age and dementia of the Alzheimer type. Proc Natl Acad Sci U S A 100, 14504–14509.

29.

Miller

, Celone

, DePeau

, Diamond

, Dickerson

, Rentz

, Pihlajamaki

, Sperling

(2008) Age-related memory impairment associated with loss of parietal deactivation but preserved hippocampal activation. Proc Natl Acad Sci U S A 105, 2181–2186.

30.

Persson

, Lind

, Larsson

, Ingvar

, Sleegers

, Van Broeckhoven

, Adolfsson

, Nilsson

, Nyberg

(2008) Altered deactivation in individuals with genetic risk for Alzheimer’s disease. Neuropsychologia 46, 1679–1687.

31.

Petrella

, Prince

, Wang

, Hellegers

, Doraiswamy

(2007) Prognostic value of posteromedial cortex deactivation in mild cognitive impairment. PLoS One 2, e1104.

32.

Pihlajamaki

, Sperling

(2009) Functional MRI assessment of task-induced deactivation of the default mode network in Alzheimer’s disease and at-risk older individuals. Behav Neurol 21, 77–91.

33.

, Xu

, Rodrigue

, Kennedy

, Cheng

, Flicker

, Hebrank

, Uh

, Park

(2011) Alterations in cerebral metabolic rate and blood supply across the adult lifespan. Cereb Cortex 21, 1426–1434.

34.

, Madison

, Baker

, Rabinovici

, Jagust

(2016) Dynamic relationships between age, amyloid-beta deposition, and glucose metabolism link to the regional vulnerability to Alzheimer’s disease. Brain 139, 2275–2289.

35.

Vaishnavi

, Vlassenko

, Rundle

, Snyder

, Mintun

, Raichle

(2010) Regional aerobic glycolysis in the human brain. Proc Natl Acad Sci U S A 107, 17757–17762.

36.

Dickerson

, Bakkour

, Salat

, Feczko

, Pacheco

, Greve

, Grodstein

, Wright

, Blacker

, Rosas

, Sperling

, Atri

, Growdon

, Hyman

, Morris

, Fischl

, Buckner

(2009) The cortical signature of Alzheimer’s disease: Regionally specific cortical thinning relates to symptom severity in very mild to mild AD dementia and is detectable in asymptomatic amyloid-positive individuals. Cereb Cortex 19, 497–510.

37.

Raz

, Rodrigue

(2006) Differential aging of the brain: Patterns, cognitive correlates and modifiers. Neurosci Biobehav Rev 30, 730–748.

38.

Jagust

(2018) Imaging the evolution and pathophysiology of Alzheimer disease. Nat Rev Neurosci 19, 687–700.

39.

Brier

, Gordon

, Friedrichsen

, McCarthy

, Stern

, Christensen

, Owen

, Aldea

, Su

, Hassenstab

, Cairns

, Holtzman

, Fagan

, Morris

, Benzinger

, Ances

(2016) Tau and Abeta imaging, CSF measures, and cognition in Alzheimer’s disease. Sci Transl Med 8, 338ra366.

40.

Tahmi

, Bou-Zeid

, Razlighi

(2019) A fully automatic technique for precise localization and quantification of amyloid-beta PET scans. J Nucl Med 60, 1771–1779.

41.

Fischl

(2012) FreeSurfer. Neuroimage 62, 774–781.

42.

Fischl

, Salat

, Busa

, Albert

, Dieterich

, Haselgrove

, van der Kouwe

, Killiany

, Kennedy

, Klaveness

, Montillo

, Makris

, Rosen

, Dale

(2002) Whole brain segmentation: Automated labeling of neuroanatomical structures in the human brain. Neuron 33, 341–355.

43.

Fischl

, van der Kouwe

, Destrieux

, Halgren

, Ségonne

, Salat

, Busa

, Seidman

, Goldstein

, Kennedy

, Caviness

, Makris

, Rosen

, Dale

(2004) Automatically parcellating the human cerebral cortex. Cereb Cortex 14, 11–22.

44.

Desikan

, Ségonne

, Fischl

, Quinn

, Dickerson

, Blacker

, Buckner

, Dale

, Maguire

, Hyman

(2006) An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage 31, 968–980.

45.

Brickman

, Guzman

, Gonzalez-Castellon

, Razlighi

, Gu

, Narkhede

, Janicki

, Ichise

, Stern

, Manly

, Schupf

, Marshall

(2015) Cerebral autoregulation, beta amyloid, and white matter hyperintensities are interrelated. Neurosci Lett 592, 54–58.

46.

, Razlighi

, Zahodne

, Janicki

, Ichise

, Manly

, Devanand

, Brickman

, Schupf

, Mayeux

, Stern

(2015) Brain amyloid deposition and longitudinal cognitive decline in nondemented older subjects: Results from a multi-ethnic population. PLoS One 10, e0123743.

47.

, Steffener

, Razlighi

, Habeck

, Liu

, Gazes

, Janicki

, Stern

(2015) Abeta-related hyperactivation in frontoparietal control regions in cognitively normal elderly. Neurobiol Aging 36, 3247–3254.

48.

LaFerla

, Oddo

(2005) Alzheimer’s disease: Aβ, tau and synaptic dysfunction. Trends Mol Med 11, 170–176.

49.

Oddo

, Caccamo

, Shepherd

, Murphy

, Golde

, Kayed

, Metherate

, Mattson

, Akbari

, LaFerla

(2003) Triple-transgenic model of Alzheimer’s disease with plaques and tangles: Intracellular Aβ and synaptic dysfunction. Neuron 39, 409–421.

50.

Hanseeuw

, Betensky

, Schultz

, Papp

, Mormino

, Sepulcre

, Bark

, Cosio

, LaPoint

, Chhatwal

(2017) Fluorodeoxyglucose metabolism associated with tau-amyloid interaction predicts memory decline. Ann Neurol 81, 583–596.

51.

Wang

, Benzinger

, Su

, Christensen

, Friedrichsen

, Aldea

, McConathy

, Cairns

, Fagan

, Morris

(2016) Evaluation of tau imaging in staging Alzheimer disease and revealing interactions between β-amyloid and tauopathy. JAMA Neurol 73, 1070–1077.

52.

Schultz

, Chhatwal

, Hedden

, Mormino

, Hanseeuw

, Sepulcre

, Huijbers

, LaPoint

, Buckley

, Johnson

(2017) Phases of hyperconnectivity and hypoconnectivity in the default mode and salience networks track with amyloid and tau in clinically normal individuals. J Neurosci 37, 4323–4331.

53.

Albert

, Zhu

, Moghekar

, Mori

, Miller

, Soldan

, Pettigrew

, Selnes

, Li

, Wang

M-C

(2018) Predicting progression from normal cognition to mild cognitive impairment for individuals at 5 years. Brain 141, 877–887.

54.

Hanseeuw

, Betensky

, Jacobs

, Schultz

, Sepulcre

, Becker

, Cosio

DMO

, Farrell

, Quiroz

, Mormino

(2019) Association of amyloid and tau with cognition in preclinical Alzheimer disease: A longitudinal study. JAMA Neurol 76, 915–924.

55.

Jack

Jr , Wiste

, Schwarz

, Lowe

, Senjem

, Vemuri

, Weigand

, Therneau

, Knopman

, Gunter

(2018) Longitudinal tau PET in ageing and Alzheimer’s disease. Brain 141, 1517–1528.

56.

Musiek

, Holtzman

(2015) Three dimensions of the amyloid hypothesis: Time, space and ‘wingmen’. Nat Neurosci 18, 800–806.

57.

Knopman

, Parisi

, Salviati

, Floriach-Robert

, Boeve

, Ivnik

, Smith

, Dickson

, Johnson

, Petersen

(2003) Neuropathology of cognitively normal elderly. J Neuropathol Exp Neurol 62, 1087–1095.

58.

Petersen

, Parisi

, Dickson

, Johnson

, Knopman

, Boeve

, Jicha

, Ivnik

, Smith

, Tangalos

(2006) Neuropathologic features of amnestic mild cognitive impairment. Arch Neurol 63, 665–672.

59.

Price

, Morris

(1999) Tangles and plaques in nondemented aging and preclinical Alzheimer’s disease. Ann Neurol 45, 358–368.

60.

Tiraboschi

, Hansen

, Thal

, Corey-Bloom

(2004) The importance of neuritic plaques and tangles to the development and evolution of AD. Neurology 62, 1984–1989.

61.

Donohue

, Sperling

, Salmon

, Rentz

, Raman

, Thomas

, Weiner

, Aisen

(2014) The preclinical Alzheimer cognitive composite: Measuring amyloid-related decline. JAMA Neurol 71, 961–970.

62.

Chetelat

, Villemagne

, Villain

, Jones

, Ellis

, Ames

, Martins

, Masters

, Rowe

(2012) Accelerated cortical atrophy in cognitively normal elderly with high β-amyloid deposition. Neurology 78, 477–484.

63.

Rhein

, Song

, Wiesner

, Ittner

, Baysang

, Meier

, Ozmen

, Bluethmann

, Drose

, Brandt

, Savaskan

, Czech

, Gotz

, Eckert

(2009) Amyloid-beta and tau synergistically impair the oxidative phosphorylation system in triple transgenic Alzheimer’s disease mice. Proc Natl Acad Sci U S A 106, 20057–20062.

64.

Garwood

, Pooler

, Atherton

, Hanger

, Noble

(2011) Astrocytes are important mediators of Abeta-induced neurotoxicity and tau phosphorylation in primary culture. Cell Death Dis 2, e167.

65.

, Yang

, Rosario

, Ubeda

, Beech

, Gant

, Chen

, Hudspeth

, Chen

, Zhao

, Vinters

, Frautschy

, Cole

(2009) Beta-amyloid oligomers induce phosphorylation of tau and inactivation of insulin receptor substrate via c-Jun N-terminal kinase signaling: Suppression by omega-3 fatty acids and curcumin. J Neurosci 29, 9078–9089.

66.

Hur

, Frost

, Wu

, Crump

, Pan

, Wong

, Barros

, Li

, Nie

, Zhai

, Wang

, Tcw

, Guo

, McKenzie

, Ming

, Zhou

, Wang

, Sagi

, Renton

, Esposito

, Kim

, Sadleir

, Trinh

, Rissman

, Vassar

, Zhang

, Johnson

, Masliah

, Greengard

, Goate

, Li

(2020) The innate immunity protein IFITM3 modulates gamma-secretase in Alzheimer’s disease. Nature 586, 735–740.

67.

Kitazawa

, Oddo

, Yamasaki

, Green

, LaFerla

(2005) Lipopolysaccharide-induced inflammation exacerbates tau pathology by a cyclin-dependent kinase 5-mediated pathway in a transgenic model of Alzheimer’s disease. J Neurosci 25, 8843–8853.

68.

Kitazawa

, Yamasaki

, LaFerla

(2004) Microglia as a potential bridge between the amyloid beta-peptide and tau. Ann N Y Acad Sci 1035, 85–103.

69.

Nisbet

, Polanco

, Ittner

, Gotz

(2015) Tau aggregation and its interplay with amyloid-beta. Acta Neuropathol 129, 207–220.

70.

Koychev

, Gunn

, Firouzian

, Lawson

, Zamboni

, Ridha

, Sahakian

, Rowe

, Thomas

, Rochester

, Ffytche

, Howard

, Zetterberg

, MacKay

, Lovestone

; Deep and Frequent Phenotyping study team, (2017) PET tau and amyloid-β burden in mild Alzheimer’s disease: Divergent relationship with age, cognition, and cerebrospinal fluid biomarkers. J Alzheimers Dis 60, 283–293.

71.

Binnewijzend

, Schoonheim

, Sanz-Arigita

, Wink

, van der Flier

, Tolboom

, Adriaanse

, Damoiseaux

, Scheltens

, van Berckel

, Barkhof

(2012) Resting-state fMRI changes in Alzheimer’s disease and mild cognitive impairment. Neurobiol Aging 33, 2018–2028.

72.

Jagust

, Landau

, Koeppe

, Reiman

, Chen

, Mathis

, Price

, Foster

, Wang

AYJAs

, Dementia (2015) The Alzheimer’s Disease Neuroimaging Initiative 2 PET core: 2015. Alzheimers Dement 11, 757–771.

73.

Schöll

, Maass

, Mattsson

, Ashton

, Blennow

, Zetterberg

, Jagust

(2019) Biomarkers for tau pathology. Mol Cell Neurosci 97, 18–33.

74.

Leuzy

, Rodriguez-Vieitez

, Saint-Aubert

, Chiotis

, Almkvist

, Savitcheva

, Jonasson

, Lubberink

, Wall

, Antoni

(2018) Longitudinal uncoupling of cerebral perfusion, glucose metabolism, and tau deposition in Alzheimer’s disease. Alzheimers Dement 14, 652–663.

75.

Seeley

, Crawford

, Zhou

, Miller

, Greicius

(2009) Neurodegenerative diseases target large-scale human brain networks. Neuron 62, 42–52.

76.

Lee

, Lee

, Seo

, Ko

, Kang

, Kim

, Seo

, Lee

, Park

, Yeon

, Okamura

, Na

, Seong

, Noh

(2020) Identification of heterogeneous subtypes of mild cognitive impairment using cluster analyses based on PET imaging of tau and astrogliosis. Front Aging Neurosci 12, 615467.

77.

Takahashi

, Capetillo-Zarate

, Lin

, Milner

, Gouras

(2010) Co-occurrence of Alzheimer’s disease β-amyloid and tau pathologies at synapses. Neurobiol Aging 31, 1145–1152.

78.

Price

, McKeel

, Buckles

, Roe

, Xiong

, Grundman

, Hansen

, Petersen

, Parisi

, Dickson

, Smith

, Davis

, Schmitt

, Markesbery

, Kaye

, Kurlan

, Hulette

, Kurland

, Higdon

, Kukull

, Morris

(2009) Neuropathology of nondemented aging: Presumptive evidence for preclinical Alzheimer disease. Neurobiol Aging 30, 1026–1036.

79.

Nelson

, Abner

, Schmitt

, Kryscio

, Jicha

, Smith

, Davis

, Poduska

, Patel

, Mendiondo

, Markesbery

(2010) Modeling the association between 43 different clinical and pathological variables and the severity of cognitive impairment in a large autopsy cohort of elderly persons. Brain Pathol 20, 66–79.