Skip to main content
NCBI home page
Alzheimer's & Dementia logoLink to Alzheimer's & Dementia
. 2025 Jan 8;21(2):e14499. doi: 10.1002/alz.14499

Mapping amyloid beta predictors of entorhinal tau in preclinical Alzheimer's disease

Jafar Zamani 1,✉, Amirali Vahid 1, Bárbara Avelar‐Pereira 1,2, Elveda Gozdas 1, S M Hadi Hosseini 1; for the Alzheimer's Disease Neuroimaging Initiative
PMCID: PMC11848422  PMID: 39777850

Abstract

INTRODUCTION

Amyloid beta (Aβ) plaques and hyperphosphorylated tau in the entorhinal regions are key Alzheimer's disease (AD) markers, but the spatial Aβ pathways influencing tau pathology remain unclear.

METHODS

We applied predictive modeling to identify Aβ standardized uptake value ratio (SUVR) spatial patterns that predict entorhinal tau levels, future hippocampal volume, and Preclinical Alzheimer's Cognitive Composite (PACC) scores at 5‐year follow‐up. The model was trained on Alzheimer's Disease Neuroimaging Initiative (ADNI) (N = 237), incorporating amyloid‐PET (positron emission tomography), tau‐PET, magnetic resonance imaging (MRI), and cognitive data, and validated on Harvard Aging Brain Study (HABS) (N = 276).

RESULTS

The model accurately predicted entorhinal tau levels (r = 0.48, p < 0.0001), future hippocampal volume (r = 0.24, p = 0.002), and PACC scores (r = 0.35, p < 0.0001) based on regional Aβ.

DISCUSSION

Aβ in the rostral middle frontal, medial orbitofrontal, and striatal regions predict entorhinal tau levels, future hippocampal volume, and PACC scores, indicating their potential as early biomarkers in AD prediction models.

Highlights

  • Positron emission tomography (PET) imaging reveals amyloid beta (Aβ) patterns predicting entorhinal tau levels in preclinical Alzheimer's disease (AD).

  • Aβ in medial orbitofrontal, rostral middle frontal, and nucleus accumbens best predicts tau.

  • Aβ distribution in these regions predicts future hippocampal neurodegeneration and cognitive decline.

  • Model validated with Alzheimer's Disease Neuroimaging Initiative (ADNI) and Harvard Aging Brain Study (HABS) data sets, showing robustness and reproducibility.

  • Findings suggest early Aβ patterns can aid in diagnosing AD and guide anti‐Aβ therapies

Keywords: Alzheimer's disease (AD), amyloid beta (Aβ), biomarkers, positron emission tomography (PET), tau pathology

1. BACKGROUND

Alzheimer's disease (AD) is the most prevalent form of dementia, accounting for more than 50% of cases. 1 The extracellular deposition of amyloid beta (Aβ) and intracellular aggregation of hyperphosphorylated tau proteins stand as histopathological hallmarks of AD. Positron emission tomography (PET) imaging studies illustrate that Aβ and tau spread systematically across the brain but exhibit distinctive propagation patterns. 2 Aβ pathology spreads from the neocortex, particularly the medial frontal and parietal regions, to all cortex, limbic regions, and then brainstem and cerebellum. 3 , 4 In contrast, tau pathology originates in the transentorhinal region in the medial temporal lobe (MTL) and extends through limbic and association areas before reaching the primary regions. 5 , 6 , 7

The spatial–temporal relationship between the Aβ and tau plays a crucial role in comprehending the pathogenesis and progression of AD. 8 The most accepted model suggests that Aβ pathophysiology initiates downstream events that could act as catalysts for subsequent molecular processes. 9 , 10 , 11 , 12 These include tau misfolding, tau‐mediated toxicity, tangle formation, and the spread of tau, ultimately leading to cortical neurodegeneration. 13 Cortical Aβ may create a permissive environment for the propagation of tau tangles from the MTL, which is closely linked to cognitive decline in AD. 14 Evidence indicates that individuals with the highest levels of Aβ also exhibit elevated levels of tau. 15 This idea is further supported by many studies showing the concomitant presence of Aβ and tau, rather than the presence of Aβ or tau in isolation, and is significantly associated with progression of AD, neurodegeneration, and subsequent cognitive decline. 16 , 17 , 18 This suggests that the combination of Aβ and tau leads to a phenotype that is distinct from Aβ alone. 19

Unraveling the spatial distribution of Aβ that would reliably predict accumulation of tau in its epicenter is crucial for understanding AD progression, but also plays a key role in informing therapeutic strategies targeting Aβ and stratifying patients based on certain Aβ patterns. However, this question has received little scrutiny in previous research. Recent studies have used predictive modeling to improve predicting cognitive decline in AD. Hojjati et al. applied neural networks to neuroimaging data to distinguish healthy individuals, mild cognitive impairment (MCI), and patients with AD. 20 Minhas et al. used autoregressive models and support vector machines (SVMs) to identify MCI patients at higher risk of developing AD. 21 Similarly, Jang et al. applied logistic regression with multiple biomarkers to predict decline in amyloid‐positive MCI patients. 22 Vieira et al. used multitarget random forests to track cognitive decline over time. 23 Our study primarily aims is to use PET imaging to establish an in vivo pathological pattern of Aβ deposition in subjects along the AD spectrum to predict levels of entorhinal tau in preclinical AD. A robust predictive modeling approach was trained and tested on two large independent data sets for generalizability. To further demonstrate the implication of the proposed approach in predicting future neurodegeneration and cognitive decline, we utilized the Aβ pattern identified in the previous step to predict future hippocampal volume and Preclinical Alzheimer's Cognitive Composite (PACC) cognitive scores at the 5‐year follow‐up. The Alzheimer's Disease Neuroimaging Initiative (ADNI) and Harvard Aging Brain Study (HABS) were used for training and evaluating the accuracy of the model, respectively. 24 , 25 This approach improves robustness and ensures reproducibility. Furthermore, to demonstrate the utility of the proposed approach in the early detection of AD, while the training ADNI data set included individuals across the AD spectrum, we exclusively tested the performance of the model on HABS data that included a representative sample of cognitively normal (CN) older adults and preclinical AD. We expect that Aβ accumulation in the medial frontal and parietal regions highly contributes to the prediction of accumulation of tau in the entorhinal cortex. In clinical practice, the proposed method could help identify patients who would benefit from anti‐amyloid therapies or preventive treatments, enabling personalized care by mapping Aβ patterns linked to tau pathology, improving outcomes, and supporting early intervention to delay symptom progression.

2. METHODS

2.1. Participants

Data from two independent cohorts were utilized: the ADNI and the HABS data sets. 24 , 25 A total of 509 participants were included, with N = 237 from ADNI (53 CN, 87 significant memory concern [SMC], 74 with MCI, and 23 AD), and N = 276 from HABS (all CN, 93 Aβ+). Aβ+ indicates the presence of significant levels of Aβ plaques in the brain, suggesting pathological accumulation associated with AD, whereas Aβ– signifies the absence of significant Aβ plaques, with individuals typically showing a lower likelihood of developing Alzheimer's pathology and associated cognitive decline. ADNI was launched in 2003, with the goal of testing whether magnetic resonance imaging (MRI), PET, other biological markers, and clinical and neuropsychological assessments could be combined to improve the understanding of AD development and progression. HABS is a longitudinal observational research initiative designed to enhance our comprehension of brain aging and preclinical AD. Participants in the HABS cohort were considered CN, as determined by a Clinical Dementia Rating (CDR) assessment and their scores on the Logical Memory delayed‐recall test, Mini‐Mental State Examination (MMSE), and Geriatric Depression Scale (GDS). The diagnosis criteria for CN individuals included MMSE score >24, a CDR of 0, and normal memory function with no significant cognitive or daily living impairments. The eligibility criteria for MCI participants included the presence of SMCs, MMSE score >24, a CDR of 0.5, and abnormal memory function with preserved general cognition and functional performance. AD participants were characterized by SMCs, an MMSE score between 20 and 26, a CDR of 0.5 or 1.0, and meeting the National Institute of Neurological Disorders and Stroke(NINCDS)/ Alzheimer's Disease and Related Disorders Association(ADRDA) criteria for probable AD. PACC, which is a composite score on tests assessing episodic memory, global cognition, and executive function, was used as a measure of cognitive decline. PACC has been used frequently as the cognitive outcome across various clinical trials. 26 Details of participants’ demographics are provided in Table 1.

TABLE 1.

Demographics of the participants.

Aβ– Aβ+ Overall
ADNI N = 129 N = 104 N = 233
Sex (female/male) 70/59 59/45 129/104
Age (mean ± SD) 69.9 ± 7 72.3 ± 7.4 71 ± 7.2
Education, y (mean ± SD) 16.5 ± 2.3 16.2 ± 2.4 16.4 ± 2.3
MMSE (mean ± SD) 28.7 ± 1.7 26.8 ± 3 27.9 ± 2.5
HABS N = 183 N = 93 N = 276
Sex (female/male) 109/74 54/39 163/113
Age (mean ± SD) 76.7 ± 6.3 78.81 ± 5.53 77.4 ± 6.2
Education, y (mean ± SD) 15.9 ± 3.2 16.4 ± 2.7 16 ± 3
MMSE (mean ± SD) 29.1 ± 1.3 28.6 ± 2.5 28.9 ± 1.3
Open in a new tab

Abbreviations: ADNI, Alzheimer's Disease Neuroimaging Initiative; Aβ, amyloid beta; HABS, Harvard Aging Brain Study; MMSE, Mini‐Mental State Examination; N, number; SD, standard deviation; y, years.

2.2. PET and MRI Image Data Processing

For ADNI Aβ and tau‐PET data, we utilized Florbetapir (AV45) and Flortaucipir (AV‐1451) measures, respectively. Florbetapir PET imaging involved injecting the radiotracer Florbetapir and acquiring a static three‐dimensional (3D) scan. The total acquisition time was 20 min, divided into four, 5‐min frames. Imaging occurred between 50 and 70 min after injection. AV‐1451 PET imaging uses the radiotracer AV‐1451 to visualize tau protein aggregates in the brain. The scan protocol involved a dynamic 3D scan, lasting a total of 30 min with six, 5‐min frames. Imaging was conducted between 75‐ and 105‐min postinjection to capture the distribution and accumulation of tau pathology. Processing involved a native‐space MRI scan for each subject, which was then processed using FreeSurfer v7.1.1 to delineate a cortical summary region, with the whole cerebellum serving as the reference region. Subsequently, each PET data set was co‐registered to the MRI scan closest in time. PET standardized uptake value ratios (SUVRs) were computed by dividing the cortical summary region by the reference region. We obtained PET data from ADNI in the fully preprocessed format. The summary SUVR, based on the whole cerebellum reference region with a threshold of 1.11, was employed to categorize subjects as Aβ+ and Aβ–. 27 Specifically, we extracted 126 Aβ SUVR features and entorhinal tau SUVR levels based on FreeSurfer‐defined cortical and subcortical regions. 28 , 29 ,  

RESEARCH IN CONTEXT

  1. Systematic review: We conducted a comprehensive review of the literature on positron emission tomography (PET) imaging of amyloid beta (Aβ) and tau pathologies in Alzheimer's disease (AD), utilizing databases such as PubMed and relevant conference proceedings. Previous research has extensively documented the progression of Aβ and tau in AD, but the interplay between Aβ distribution patterns and tau accumulation, particularly in the preclinical stages, has been less explored.

  2. Interpretation: Our study identifies specific Aβ patterns in the medial orbitofrontal, rostral middle frontal, and nucleus accumbens regions that predict entorhinal tau levels, hippocampal neurodegeneration, and cognitive decline. This finding advances the understanding of how early Aβ deposition can serve as a marker for tau pathology and potential future cognitive impairment, offering new insights into AD progression and diagnosis.

  3. Future directions: Future research should focus on validating these Aβ patterns in diverse populations and exploring their utility in predicting AD onset in individuals with minimal Aβ pathology. In addition, investigating the underlying mechanisms linking Aβ distribution to tau spread and assessing the efficacy of targeted anti‐Aβ therapies based on these findings could further refine AD diagnostic and therapeutic strategies.

For HABS, PET data were obtained using a Siemens/CTI ECAT HR Scanner in 3D mode, with 63 image planes, a 15.2 cm axial field of view, 5.6 mm transaxial resolution, and a 2.4 mm slice interval. Prior to injection, 10‐min transmission scans were collected for attenuation correction. PET data underwent reconstruction, attenuation correction, and manual evaluation for count statistics and motion correction. In terms of data processing and region of interest (ROI)–based measures, a mean image was created after the acquisition, with specific processing for Pittsburgh compound B (PiB; first 8 min postinjection). The image was co‐registered to the corresponding FreeSurfer‐processed T1 image using the Statistical Parametric Mapping(SPM)12 package. The atlas was generated in FreeSurfer. Bilateral cerebellum gray matter was then used as the reference region for SUVR measurements. We employed the same parcellation and features utilized in the ADNI.

Hippocampal volume data were extracted from MRI T1‐weighted data in HABS. MRI was conducted on a Siemens TIM Trio 3T System with a 12‐channel head coil, using structural T1‐weighted volumetric magnetization‐prepared, rapid acquisition gradient echo (MPRAGE) scans. ROI labeling was performed using FreeSurfer v6.0, with cortical measurements based on the Desikan‐Killiany atlas and subcortical measurements based on the FreeSurfer aseg atlas. Hippocampal volume at the 5‐year follow‐up was used as a measure of future neurodegeneration. 24 This study employed two versions of FreeSurfer to analyze neuroimaging data from different data sets: FreeSurfer v7.1.1 for the ADNI data set and FreeSurfer v6.0 for the HABS data set. The choice of versions was based on their respective capabilities and validation status. Notably, minor differences exist between these versions, particularly in the volumetric values.

2.3. Predictive modeling

The Aβ SUVR levels for 126 ROIs spanning the whole brain, including cortical and subcortical areas, were utilized as input features to the model. Entorhinal tau SUVR level was considered as the response variable. First, we regressed out the effect of confounding factors including age, sex, and education from the input features and response variable. To regress out age, sex, and education, we created linear regression models for each feature and the response variable, using these factors as predictors. We then subtracted the predicted values from the original values to get residuals, which show the effects of the main variables without the influence of these confounders. This method helps ensure our analysis focuses on the true relationships in the data. Next, we implemented a correlation‐based ranking with forward feature selection to identify the optimal Aβ SUVR features that were highly correlated with entorhinal tau levels. 30 We applied this approach to ADNI data as our training set to identify the Aβ SUVR features that result in highest accuracy in predicting entorhinal tau levels within the ADNI data set. We then implemented a support vector regression (SVR) model for predictive modeling. SVR was chosen due to its effectiveness in handling high‐dimensional data and its capability to model complex relationships. 31 The identified Aβ SUVR features were utilized as input to train the SVR model on ADNI data. We then evaluated the performance and generalizability of the model using independent data from HABS. The same approach was applied for training and testing the models for predicting future hippocampal volume and PACC cognitive scores 5‐years later using the already identified Aβ SUVR features. The accuracy and statistical significance of the models were evaluated by testing the Pearson correlation coefficient between the predicted and observed values for entorhinal tau levels, future hippocampal volume, and PACC cognitive scores.

3. RESULTS

3.1. Association between regional Aβ SUVR and entorhinal tau levels

Regions that showed the highest association between Aβ SUVR and entorhinal tau levels within the ADNI data set were located primarily in the prefrontal cortex followed by regions in the parietal and temporal cortices (Table 2 and Figure 1A). Pearson correlation coefficients and p‐values between regional Aβ SUVR and entorhinal tau levels for ADNI and HABS are listed in Table S1. Table S2 presents the non‐parametric Spearman's rank correlation coefficients (r) and p‐values (p) between regional Aβ SUVR and entorhinal tau levels within the ADNI and HABS data sets.

TABLE 2.

The Pearson rank correlation coefficients (r) and p‐values (p) between regional Aβ SUVR and entorhinal tau levels within the ADNI and HABS data sets.

Features name ADNI HABS
r p r p
Medial orbitofrontal 0.6724 <0.000001 0.4849 <0.000001
Right medial orbitofrontal 0.6698 <0.000001 0.5106 <0.000001
Right rostral middle frontal 0.6690 <0.000001 0.4989 <0.000001
Rostral middle frontal 0.6679 <0.000001 0.4960 <0.000001
Left medial orbitofrontal 0.6671 <0.000001 0.4456 <0.000001
Accumbens area 0.6605 <0.000001 0.4596 <0.000001
Left rostral middle frontal 0.6586 <0.000001 0.4792 <0.000001
Right accumbens area 0.6579 <0.000001 0.4753 <0.000001
Left lateral orbitofrontal 0.6572 <0.000001 0.4400 <0.000001
Lateral orbitofrontal 0.6572 <0.000001 0.4658 <0.000001
Right frontal pole 0.6543 <0.000001 0.5018 <0.000001
Right pars triangularis 0.6512 <0.000001 0.4510 <0.000001
Frontal pole 0.6510 <0.000001 0.4905 <0.000001
Right lateral orbitofrontal 0.6485 <0.000001 0.4759 <0.000001
Right pars orbitalis 0.6473 <0.000001 0.4277 <0.000001
Pars orbitalis 0.6466 <0.000001 0.4474 <0.000001
Pars triangularis 0.6461 <0.000001 0.4810 <0.000001
Left accumbens area 0.6419 < 0.000001 0.4221 < 0.000001
Right middle temporal 0.6369 < 0.000001 0.4616 < 0.000001
Middle temporal 0.6369 < 0.000001 0.4651 < 0.000001
Open in a new tab

Note: The FDR correction was applied to the p‐values. Only the top 20 brain regions with significant p‐values are included.

Abbreviations: ADNI, Alzheimer's Disease Neuroimaging Initiative; Aβ, amyloid beta; FDR, false discovery rate; HABS, Harvard Aging Brain Study; SUVR, standardized uptake value ratio.

FIGURE 1.

FIGURE 1

Open in a new tab

Brain maps showing correlation of Aβ SUVR with entorhinal tau levels, future hippocampal volume, and future PACC cognitive scores. Brain surface render maps (cortical and subcortical) illustrating the correlation coefficients between Aβ SUVR across all brain regions and (A) entorhinal tau levels, (B) future hippocampal volume at 5‐year follow‐up, and (C) future PACC cognitive scores at 5‐year follow‐up, within the ADNI data set. ADNI, Alzheimer's Disease Neuroimaging Initiative; Aβ, amyloid beta; PACC, Preclinical Alzheimer's Cognitive Composite; SUVR, standardized uptake value ratio.

3.2. Aβ SUVR pattern predicts entorhinal tau levels

The correlation‐based ranking feature selection algorithm revealed the regional Aβ SUVR features that resulted in the highest accuracy for predicting entorhinal tau levels in ADNI (Figure S1). These regions included the bilateral rostral middle frontal cortex, bilateral medial orbitofrontal cortex, and nucleus accumbens. The Aβ SUVR values for these regions were utilized for training a model that predicts entorhinal tau level in ADNI. The performance and generalizability of the trained model was then tested on HABS. The model showed significant predictive performance with correlation accuracy of r = 0.48 (p < 0.0001) for predicting entorhinal tau levels. Figure 2A shows the contribution of each region Aβ SUVR in predicting tau levels. The SVR model weights for Aβ SUVR patterns displayed distinct spatial distributions across cortical and subcortical regions, predicting key AD markers. For entorhinal tau levels (Figure 3A), significant model weights were identified in regions closely associated with early tau accumulation, underscoring the predictive role of specific Aβ SUVR patterns for tau pathology in the entorhinal cortex.

FIGURE 2.

FIGURE 2

Open in a new tab

Prediction of entorhinal tau levels, future hippocampal volume, and future PACC scores using Aβ patterns. Prediction of (A) entorhinal tau levels, (B) future hippocampal volume, and (C) future PACC cognitive scores based on Aβ‐identified patterns. The graphs show the association between measured and predicted values. Aβ, amyloid beta; PACC, Preclinical Alzheimer's Cognitive Composite.

FIGURE 3.

FIGURE 3

Open in a new tab

SVR model weights for Aβ SUVR patterns in predicting entorhinal tau levels, future hippocampal volume, and future PACC scores. Brain surface render maps (cortical and subcortical) illustrating the SVR model weights for Aβ SUVR‐identified patterns in predicting (A) entorhinal tau levels, (B) future hippocampal volume at 5‐year follow‐up, and (C) future PACC cognitive scores at 5‐year follow‐up. Aβ, amyloid beta; PACC, Preclinical Alzheimer's Cognitive Composite; SVR, support vector regression; SUVR, standardized uptake value ratio.

3.3. Predicting future neurodegeneration and cognitive decline

Neurofibrillary tau tangles are believed to be the main driver of neurodegeneration and future cognitive decline. Thus, we utilized the regional Aβ SUVR features that were significant predictors of entorhinal tau levels as inputs in the model predicting future hippocampal volume and PACC cognitive scores. These models were trained on ADNI and were tested on HABS. Using baseline Aβ SUVR pattern from HABS, the models predicted future volume and PACC cognitive scores at 5‐year follow‐up with significant correlation accuracy of r = 0.24 (p = 0.002), and r = 0.35 (p < 0.0001). The brain surface render maps showing the contribution (feature weights) of each region Aβ for predicting tau in the HABS data set are shown in  Figure 2B and C.  We also explored the associations between baseline regional Aβ SUVR and future hippocampal volume, and PACC cognitive scores (Figure 1B,C and Table S3). Of interest, the baseline Aβ SUVR in the rostral middle frontal, medial orbitofrontal, and nucleus accumbens were highly associated with future hippocampal volume and PACC cognitive scores at the 5‐year follow‐up. In future hippocampal volume and PACC cognitive scores at the 5‐year follow‐up (Figure 3B,C), the model highlighted regions contributing to hippocampal atrophy and cognitive decline, indicating Aβ SUVR's influence on neurodegeneration and cognitive processing.

3.4. Distribution of Aβ SUVR load across brain regions

Finally, we examined the regional distribution of Aβ SUVR load across cortical and subcortical regions. Regions in the medial temporal, parietal, and frontal cortices showed the highest amount of Aβ SUVR load (Figure S2).

Our analyses revealed the Aβ distribution patterns that exhibit significant contributions for predicting entorhinal tau levels, future hippocampal volume, and PACC cognitive scores 5 years later.

4. DISCUSSION

The main goal of this study was to utilize PET imaging to establish an in vivo pathological Aβ deposition pattern for prediction of entorhinal tau levels in individuals in the preclinical AD stage. This is essential for understanding disease progression. Our results revealed that Aβ in the bilateral medial orbitofrontal, rostral middle frontal, and nucleus accumbens contributed the most to the prediction of entorhinal tau levels in preclinical AD. The Aβ SUVR load in these regions also predicted future hippocampal volume neurodegeneration and PACC cognitive scores 5 years later. Most importantly, these results were validated using an independent representative sample of older adults.

Although several studies have delved into Aβ staging and its association with tau pathology, the exploration of an Aβ pattern intricately linked to tau has remained a largely neglected facet in existing research. 32 , 33 We trained our model to identify Aβ patterns that predict entorhinal tau levels, considering that the entorhinal cortex is the epicenter of tau pathology and that tau spreading follows initial Aβ accumulation in AD. Moreover, the distribution and burden of tau might align more closely with clinical symptoms and neurodegenerative processes. 34 Although not all individuals with Aβ pathology will progress to AD, the majority of those with entorhinal tau levels pathology are likely to experience neurodegeneration and memory decline. 32 , 33 , 35 , 36 Our findings indicate that Aβ patterns can serve as markers of tau load in the entorhinal cortex and predict future hippocampal volume degeneration and PACC cognitive scores. The proposed method holds significant potential for early diagnosis of AD, given that the pathological accumulation of Aβ and tau proteins begins 10–20 years prior to the clinical onset of dementia.

Previous research has focused mainly on the posterior parietal (precuneus) and posterior cingulate regions as the hubs of Aβ accumulation. 37 , 38 , 39 We identified regions in the bilateral medial orbitofrontal cortex, rostral middle frontal cortex, and nucleus accumbens where the amyloid load exhibits the highest predictive accuracy for entorhinal tau levels pathology. Early postmortem neuropathological studies showed that the basal portion of the medial frontal cortex is one of the earliest sites of Aβ accumulation in AD, preceding Aβ build‐up in the precuneus and posterior cingulate. 2 Recent multi‐cohort neuroimaging studies also show that the medial orbitofrontal cortex is one of the earliest sites of Aβ accumulation in individuals with no signs of elevated Aβ in the cerebrospinal fluid (CSF) and PET. However, these individuals eventually go on to develop a high Aβ load. 40 , 41 The identified regions have strong connections with MTL structures and are involved in learning and memory. The orbitofrontal cortex has strong connections with MTL structures including the entorhinal and perirhinal cortices. 42 The rostral middle frontal gyrus is part of the dorsolateral prefrontal cortex and is connected to parietal and temporal lobes through the arcuate fasciculus. 43 Previous studies have shown significant reduction in the rostral middle frontal cortical thickness as Aβ levels increase. 44 , 45 , 46 There was a negative linear relationship between amyloid load in the rostral middle frontal cortex and worsening executive performance. 47 Nucleus accumbens is a central structure of the mesolimbic dopaminergic pathway and plays a critical role in reinforcement learning in connection with MTL structures. 48 This region is also critical for processing motivation and rewards and has been shown to exhibit AD‐associated pathological changes in humans. 49 , 50

The fact that Aβ load in areas remote to the entorhinal cortex are the best predictors of entorhinal tau levels is supported by previous evidence. 40 , 51 Animal studies suggest that injection of Aβ into the cortex of the mouse brain results in creation of tau in distal but connected subcortical areas. 51 One theory suggests that tau pathology may already exist in the brainstem and that the accumulation of Aβ in distant regions creates an environment, perhaps driven by inflammatory processes, that promotes the spread of tau pathology to the neocortex. 9 The connections between the identified regions and MTL may also contribute to this process.

Entorhinal tau level pathology is considered the most proximal measure of neurodegeneration compared with amyloid pathology. Using the features that were best predictors of entorhinal tau levels, our model successfully predicted future hippocampal volume and PACC cognitive scores. This approach is promising, considering that Aβ pathology is the earliest marker of AD, and our model could potentially help with predicting AD development before tau pathology starts to spread to the neocortex.

Numerous prevention trials are currently underway, and a few have already been approved, to examine the hypothesis that early implementation of anti‐Aβ approaches may delay the future onset of clinical AD symptoms. 35 The premise of these trials is that reducing Aβ in the brain to levels below the threshold that triggers the spread of tau pathology may confer therapeutic benefits. 52 However, solely reducing Aβ may not consistently correlate with tangible clinical improvements. 53 Our results suggest that Aβ load in certain frontal and striatal regions is a strong predictor of entorhinal tau levels, future hippocampal volume neurodegeneration, and PACC cognitive scores. These findings can help with identification of individuals who are at higher risk of developing entorhinal tau level pathology and can potentially inform which individuals would benefit the most from anti‐Aβ therapies.

Although our study provides valuable insights into the spatial patterns of Aβ load that are best predictors of entorhinal tau level pathology, future research could utilize other biomarkers to unravel the potential intermediate processes that link amyloid and tau pathologies. Furthermore, longitudinal studies and larger data sets with even longer follow‐ups may further elucidate the temporal dynamics of AD progression.

We identified spatial patterns of Aβ load that were able to successfully predict entorhinal tau levels, future hippocampal volume, and PACC cognitive scores. Particularly, Aβ levels in certain parts of the frontal and striatal regions were the most important predictors of entorhinal tau levels. These findings not only help with our understanding of AD development in the early stages of the disease, they also have implications in identifying individuals who may develop neurodegeneration as well as for stratification of individuals who would benefit the most from anti‐amyloid therapies.

AUTHOR CONTRIBUTIONS

Jafar Zamani and S. M. Hadi Hosseini designed the study and conducted formal analyses and investigations. Amirali Vahid contributed analytical tools. S. M. Hadi Hosseini provided resources and supervision. Jafar Zamani, Bárbara Avelar‐Pereira, Elveda Gozdas, and S. M. Hadi Hosseini wrote and reviewed the paper.

CONFLICT OF INTEREST STATEMENT

No conflicts of interest, financial or otherwise, are declared by the authors. The authors report no competing interests. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All human subjects in this study provided written informed consent prior to their participation.

Supporting information

Supporting Information

ALZ-21-e14499-s001.docx (76.2KB, docx)

Supporting Information

ALZ-21-e14499-s002.pdf (762.7KB, pdf)

ACKNOWLEDGMENTS

We acknowledge the individuals who participated in the Alzheimers's Disease Neuroimaging Initiative(ADNI) and Harvard aging Brain Study(HABS) and the investigatory teams who made this work possible. ADNI data collection and sharing was funded by the ADNI (National Institutes of Health Grant U01 AG024904) and the Department of Defense (DOD) ADNI (Department of Defense award number W81XWH‐12‐2‐0012). 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; BioClinica, Inc.; Biogen; Bristol‐Myers Squibb Company; CereSpir, Inc.; Cogstate; 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; Neurotrack 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 (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer's Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. This study was partly funded by the National Institute on Aging (R01AG073362). S.M.Hadi.Hosseini.’s effort was supported in part by the National Institute on Aging (NIA; R21AG073973, R21AG064263, R01AG072470) and National Institute of Mental Health (NIMH; R61MH119289, R21MH123873).

Zamani J, Vahid A, Avelar‐Pereira B, Gozdas E, Hosseini SMH. Mapping amyloid beta predictors of entorhinal tau in preclinical Alzheimer's disease. Alzheimer's Dement. 2025;21:e14499. 10.1002/alz.14499

Data used in preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wpcontent/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf

DATA AVAILABILITY STATEMENT

Data used in preparation of this article were obtained from ADNI (adni.loni.usc.edu). As such, the investigators within ADNI contributed to the design and implementation and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wpcontent/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf. In addition, data from the HABS is available upon request at: https://habs.mgh.harvard.edu/researchers/request‐data.

REFERENCES

  • 1. Shin JH. Dementia epidemiology fact sheet 2022. Annals of Rehabilitation Medicine. 2022;46(2):53‐59. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Braak H, Braak E. Neuropathological stageing of Alzheimer‐related changes. Acta Neuropathol. 1991;82(4):239‐259. [DOI] [PubMed] [Google Scholar]
  • 3. Thal DR, Rüb U, Orantes M, Braak H. Phases of Aβ‐deposition in the human brain and its relevance for the development of AD. Neurology. 2002;58(12):1791‐1800. [DOI] [PubMed] [Google Scholar]
  • 4. Hampel H, Hardy J, Blennow K, et al. The amyloid‐β pathway in Alzheimer's disease. Mol Psychiatry. 2021;26(10):5481‐5503. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Cai Y, Du J, Li A, et al. Initial levels of β‐amyloid and tau deposition have distinct effects on longitudinal tau accumulation in Alzheimer's disease. Alzheimer's Research & Therapy. 2023;15(1):30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Kaufman SK, Del Tredici K, Thomas TL, Braak H, Diamond MI. Tau seeding activity begins in the transentorhinal/entorhinal regions and anticipates phospho‐tau pathology in Alzheimer's disease and PART. Acta Neuropatho. 2018;136:57‐67. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Insel PS, Young CB, Aisen PS, et al. Tau positron emission tomography in preclinical Alzheimer's disease. Brain. 2023;146(2):700‐711. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Pontecorvo MJ, Devous MD Sr, Navitsky M, et al. Relationships between flortaucipir PET tau binding and amyloid burden, clinical diagnosis, age and cognition. Brain. 2017;140(3):748‐763. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Busche MA, Hyman BT. Synergy between amyloid‐β and tau in Alzheimer's disease. Nat Neurosci. 2020;23(10):1183‐1193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Choi SH, Kim YH, Hebisch M, et al. A three‐dimensional human neural cell culture model of Alzheimer's disease. Nature. 2014;515(7526):274‐278. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Bennett RE, DeVos SL, Dujardin S, et al. Enhanced tau aggregation in the presence of amyloid β. Am J Pathol. 2017;187(7):1601‐1612. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Vergara C, Houben S, Suain V, et al. Amyloid‐β pathology enhances pathological fibrillary tau seeding induced by Alzheimer PHF in vivo. Acta Neuropathol. 2019;137:397‐412. [DOI] [PubMed] [Google Scholar]
  • 13. Brettschneider J, Tredici KD, Lee VMY, Trojanowski JQ. Spreading of pathology in neurodegenerative diseases: a focus on human studies. Nat Rev Neurosci. 2015;16(2):109‐120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Jacobs HI, Hedden T, Schultz AP, et al. Structural tract alterations predict downstream tau accumulation in amyloid‐positive older individuals. Nat Neurosci. 2018;21(3):424‐431. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Quiroz YT, Sperling RA, Norton DJ, et al. Association between amyloid and tau accumulation in young adults with autosomal dominant Alzheimer disease. JAMA Neurol. 2018;75(5):548‐556. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Hanseeuw BJ, Betensky RA, Schultz AP, et al. Fluorodeoxyglucose metabolism associated with tau‐amyloid interaction predicts memory decline. Ann Neurol. 2017;81(4):583‐596. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Pascoal TA, Mathotaarachchi S, Mohades S, et al. Amyloid‐β and hyperphosphorylated tau synergy drives metabolic decline in preclinical Alzheimer's disease. Mol Psychiatry. 2017;22(2):306‐311. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Wang L, Benzinger TL, Su Y, et al. Evaluation of tau imaging in staging Alzheimer disease and revealing interactions between β‐amyloid and tauopathy. JAMA Neurol. 2016;73(9):1070‐1077. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Busche MA, Wegmann S, Dujardin S, et al. Tau impairs neural circuits, dominating amyloid‐β effects, in Alzheimer models in vivo. Nat Neurosci. 2019;22(1):57‐64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Hojjati SH, Babajani‐Feremi A; Alzheimer's Disease Neuroimaging Initiative . Prediction and modeling of neuropsychological scores in Alzheimer's disease using multimodal neuroimaging data and artificial neural networks. Front Comput Neurosci. 2022;15:769982. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Minhas S, Khanum A, Riaz F, Khan SA, Alvi A. Predicting progression from mild cognitive impairment to Alzheimer's disease using autoregressive modelling of longitudinal and multimodal biomarkers. IEEE J Biomed Health Inf. 2017;22(3):818‐825. [DOI] [PubMed] [Google Scholar]
  • 22. Jang H, Park J, Woo S, Kim S, Kim HJ, Na DL, Alzheimer's Disease Neuroimaging Initiative . Prediction of fast decline in amyloid positive mild cognitive impairment patients using multimodal biomarkers. NeuroImage: Clinical. 2019;24:101941. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Vieira BH, Liem F, Dadi K, et al. Predicting future cognitive decline from non‐brain and multimodal brain imaging data in healthy and pathological aging. Neurobiol Aging. 2022;118:55‐65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Dagley A, LaPoint M, Huijbers W, et al. Harvard aging brain study: dataset and accessibility. Neuroimage. 2017;144:255‐258. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Mueller SG, Weiner MW, Thal LJ, et al. Ways toward an early diagnosis in Alzheimer's disease: the Alzheimer's Disease Neuroimaging Initiative (ADNI). Alzheimer's & Dementia. 2005;1(1):55‐66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Donohue MC, Sperling RA, Salmon DP, et al. The preclinical Alzheimer cognitive composite: measuring amyloid‐related decline. JAMA Neurol. 2014;71(8):961‐970. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Landau SM, Lu M, Joshi AD, Pontecorvo M, Mintun MA, Trojanowski JQ; Alzheimer's Disease Neuroimaging Initiative . Comparing positron emission tomography imaging and cerebrospinal fluid measurements of β‐amyloid. Ann Neurol. 2013;74(6):826‐836. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Landau S, Murphy A, Lee JQ, Ward J, Jagust W. Florbetapir (AV45) processing methods. USC Loni; 2021. https://downloads.loni.usc.edu/download/files/study/6d969531‐cf84‐4ed2‐aedc‐f8f41567566a/file/adni/UCBERKELEY_AV45_Methods_11,15 [Google Scholar]
  • 29. Landau S, Jagust W. Flortaucipir (AV‐1451) processing methods. Alzheimer's Disease Neuroimaging Initiative; 2016. [Google Scholar]
  • 30. Yang W, Wang K, Zuo W. Neighborhood component feature selection for high‐dimensional data. J Comput. 2012;7(1):161‐168. [Google Scholar]
  • 31. Smola AJ, Schölkopf B. A tutorial on support vector regression. Statistics and Computing. 2004;14:199‐222. [Google Scholar]
  • 32. Pichet Binette A, Franzmeier N, Spotorno N, et al. Amyloid‐associated increases in soluble tau relate to tau aggregation rates and cognitive decline in early Alzheimer's disease. Nat Commun. 2022;13(1):6635. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Hansson O. Biomarkers for neurodegenerative diseases. Nat Med. 2021;27(6):954‐963. [DOI] [PubMed] [Google Scholar]
  • 34. Nelson PT, Alafuzoff I, Bigio EH, et al. Correlation of Alzheimer disease neuropathologic changes with cognitive status: a review of the literature. Journal of Neuropathology & Experimental Neurology. 2012;71(5):362‐381. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Sperling RA, Mormino EC, Schultz AP, et al. The impact of amyloid‐beta and tau on prospective cognitive decline in older individuals. Ann Neurol. 2019;85(2):181‐193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Budd Haeberlein S, Aisen PS, Barkhof F, et al. Two randomized phase 3 studies of aducanumab in early Alzheimer's disease. The journal of Prevention of Alzheimer's Disease. 2022;9(2):197‐210. [DOI] [PubMed] [Google Scholar]
  • 37. Pascoal TA, Therriault J, Mathotaarachchi S, Kang MS, Shin M, Benedet AL, Alzheimer's Disease Neuroimaging Initiative . Topographical distribution of Aβ predicts progression to dementia in Aβ positive mild cognitive impairment. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring. 2020;12(1):e12037. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Sepulcre J, Grothe MJ, d'Oleire Uquillas F, et al. Neurogenetic contributions to amyloid beta and tau spreading in the human cortex. Nat Med. 2018;24(12):1910‐1918. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Jagust W. Imaging the evolution and pathophysiology of Alzheimer disease. Nat Rev Neurosci. 2018;19(11):687‐700. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Palmqvist S, Schöll M, Strandberg O, et al. Earliest accumulation of β‐amyloid occurs within the default‐mode network and concurrently affects brain connectivity. Nat Commun. 2017;8(1):1214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Collij LE, Salvadó G, Wottschel V, Mastenbroek SE, Schoenmakers P, Heeman F, Alzheimer's Disease Neuroimaging Initiative; for the ALFA study . Spatial‐temporal patterns of β‐amyloid accumulation: a subtype and stage inference model analysis. Neurology. 2022;98(17):e1692‐e1703. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Petrides M. The orbitofrontal cortex: novelty, deviation from expectation, and memory. Ann NY Acad Sci. 2007;1121(1):33‐53. [DOI] [PubMed] [Google Scholar]
  • 43. Rilling JK, Glasser MF, Preuss TM, et al. The evolution of the arcuate fasciculus revealed with comparative DTI. Nat Neurosci. 2008;11(4):426‐428. [DOI] [PubMed] [Google Scholar]
  • 44. Gomez G, Beason‐Held LL, Bilgel M, et al. Metabolic syndrome and amyloid accumulation in the aging brain. Journal of Alzheimer's Disease. 2018;65(2):629‐639. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Kirova AM, Bays RB, Lagalwar S. Working memory and executive function decline across normal aging, mild cognitive impairment, and Alzheimer's disease. Biomed Res Int. 2015;2015(1):748212. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Becker JA, Hedden T, Carmasin J, et al. Amyloid‐β associated cortical thinning in clinically normal elderly. Ann Neurol. 2011;69(6):1032‐1042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47. Garon M, Weis L, Fiorenzato E, et al. Quantification of brain β‐amyloid load in Parkinson's disease with mild cognitive impairment: a PET/MRI study. Frontiers in Neurology. 2022;12:760518. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48. Ibrahim KM, Massaly N, Yoon HJ, et al. Dorsal hippocampus to nucleus accumbens projections drive reinforcement via activation of accumbal dynorphin neurons. Nat Commun. 2024;15(1):750. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49. Guo C, Wen D, Zhang Y, et al. Amyloid‐β oligomers in the nucleus accumbens decrease motivation via insertion of calcium‐permeable AMPA receptors. Mol Psychiatry. 2022;27(4):2146‐2157. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50. Pavuluri M, Volpe K, Yuen A. Nucleus accumbens and its role in reward and emotional circuitry: a potential hot mess in substance use and emotional disorders. AIMS Neuroscience. 2017;4(1):52‐70. [Google Scholar]
  • 51. Gotz J, Chen FV, Van Dorpe J, Nitsch R. Formation of neurofibrillary tangles in P301L tau transgenic mice induced by Aβ42 fibrils. Science. 2001;293(5534):1491‐1495. [DOI] [PubMed] [Google Scholar]
  • 52. Karran E, De Strooper B. The amyloid hypothesis in Alzheimer disease: new insights from new therapeutics. Nat Rev Drug Discovery. 2022;21(4):306‐318. [DOI] [PubMed] [Google Scholar]
  • 53. Betthauser TJ, Koscik RL, Jonaitis EM, et al. Amyloid and tau imaging biomarkers explain cognitive decline from late middle‐age. Brain. 2020;143(1):320‐335. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information

ALZ-21-e14499-s001.docx (76.2KB, docx)

Supporting Information

ALZ-21-e14499-s002.pdf (762.7KB, pdf)

Data Availability Statement

Data used in preparation of this article were obtained from ADNI (adni.loni.usc.edu). As such, the investigators within ADNI contributed to the design and implementation and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wpcontent/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf. In addition, data from the HABS is available upon request at: https://habs.mgh.harvard.edu/researchers/request‐data.

Articles from Alzheimer's & Dementia are provided here courtesy of Wiley

RESOURCES