Do Tau Deposition and Glucose Metabolism Dissociate in Alzheimer’s Disease Trajectory?

Abstract

Background:

Tau aggregation demonstrates close associations with hypometabolism in Alzheimer’s disease (AD), although differing pathophysiological processes may underlie their development.

Objective:

To establish whether tau deposition and glucose metabolism have different trajectories in AD progression and evaluate the utility of global measures of these pathological hallmarks in predicting cognitive deficits.

Methods:

279 participants with amyloid-β (Aβ) status, and T1-weighted MRI scans, were selected from the Alzheimer’s Disease Neuroimaging Initiative (http://adni.loni.usc.edu). We created the standard uptake value ratio images using Statistical Parametric Mapping 12 for [¹⁸F]AV1451-PET (tau) and [¹⁸F]FDG-PET (glucose metabolism) scans. Voxel-wise group and single-subject level SPM analysis evaluated the relationship between global [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET depending on the Aβ status. Linear models assessed whether tau deposition or glucose metabolism better predicted clinical progression.

Results:

There was a dissociation between global cerebral glucose hypometabolism and global tau load in amyloid-positive AD and amyloid-negative mild cognitive impairment (MCI) (p > 0.05). Global hypometabolism was only associated with global cortical tau in amyloid-positive MCI. Voxel-level single subject tau load better predicted neuropsychological performance, Alzheimer’s disease assessment scale-cognitive (ADAS-Cog) 13 score, and one-year change compared with regional and global hypometabolism.

Conclusions:

A dissociation between tau pathology and glucose metabolism at a global level in AD could imply that other pathological processes influence glucose metabolism. Furthermore, as tau is a better predictor of clinical progression, these processes may have independent trajectories and require independent consideration in the context of therapeutic interventions.

Keywords

Alzheimer’s disease fluorodeoxyglucose F18 positron-emission tomography tau proteins

INTRODUCTION

Hyperphosphorylation of tau leading to neurofibrillary tangles plays an important role in impaired synaptic function and cognitive dysfunction. 1 Tau propagates from the transentorhinal cortex to limbic regions and then spreads across the neocortex, initiating in prodromal phases earlier than previously thought.2 –5 Glucose metabolism measured by [¹⁸F]FDG positron emission tomography (PET) is a well-established hallmark of disease progression, with alterations in temporoparietal regions observed early in the disease trajectory. 6 The Alzheimer’s Association workgroup criteria for the intended use of tau PET in the staging, prognosis, and as an indicator of biological treatment effect. 7 Tau PET demonstrates clinical value in enhancing diagnostic accuracy, especially in those with an amyloid-positive status. 8 The appropriate use criteria of tau PET is to aid diagnosis in mild cognitive impairment (MCI) and Alzheimer’s disease (AD) and inform prognosis. 9 Tau pathology and hypometabolism have been shown to better predict cognitive deterioration than amyloid-β.10,11, 10,11 Hypometabolism has been associated with tau deposition in several brain regions in AD.12 –14 Researchers have since questioned whether [¹⁸F]FDG-PET is the inverse image of tau-PET. 15 However, no studies have evaluated the global changes in tau deposition and glucose metabolism at a voxel level using a single-subject design which should be able to answer that question at a voxel level and evaluate how these changes influence cognitive deterioration. Additionally, evaluating single-subject, voxel-level global tau load allows us to control for the off-target binding observed with tau-PET imaging effectively. 16

Changes in glucose metabolism and tau deposition may be affected independently by separate pathological processes: 1) Microglial activation may be contributing to [¹⁸F]FDG-PET signal; 17 2) The association between the tau and neurodegeneration biomarkers is likely to be influenced by the presence of amyloid-β.18,19, 18,19 As multiple factors influence changes in glucose metabolism and the effect of tau in the central nervous system, these pathological hallmarks are likely to have discordant trajectories in AD development.

To decipher whether global glucose hypometabolism dissociates from tau deposition in AD trajectory and to evaluate the utility of global measures of [¹⁸F]AV1451-PET and [¹⁸F]FDG-PET in predicting clinical status and the subsequent decline, we tested three main objectives. Firstly, we aimed to establish if hypometabolism dissociates with tau aggregation in AD by assessing the global relationship between [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET using single-subject statistical parametric mapping (SPM) analysis. Secondly, we compared the influence of tau deposition in subjects with amyloid-β deposition. Finally, we evaluated the utility of regional and global [¹⁸F]AV1451-PET and [¹⁸F]FDG-PET in predicting neuropsychological performance and longitudinal clinical progression.

MATERIALS AND METHODS

Participants

279 participants (44 AD, 95 amyloid-positive (MCI) with [¹⁸F]AV1451-PET and [¹⁸F]FDG-PET scans, and T1 weighted MRI; 70 cognitively normal (CN) healthy controls with an [¹⁸F]AV1451-PET scan and 70 CN healthy controls with an [¹⁸F]FDG-PET scan were evaluated from the Alzheimer’s Disease Neuroimaging Initiative (ADNI, http://adni.loni.usc.edu/). All participants with an [¹⁸F]AV1451-PET and an [¹⁸F]FDG-PET scan were selected. There was an average of 36 days (SD = 95.3) interval between [¹⁸F]AV1451-PET and [¹⁸F]FDG-PET scans.

Amyloid-β status (amyloid-positive/amyloid-negative) was evaluated for all subjects. The subject was considered amyloid-positive if the cortical SUVR (standard uptake value ratio over cerebellum was above 1.11 and 1.08 for [¹⁸F]florbetapir and [¹⁸F]florbetaben respectively (http://adni.loni.usc.edu). 20 All CN participants included were amyloid-negative with an [¹⁸F]AV1451-PET scan and an [¹⁸F]FDG-PET scan.

Standard protocol approvals, registrations, and patient consents

The ADNI study was approved by the institutional review boards of all participating institutions. Written informed consent was obtained from all participants at each site as a part of ADNI protocol.

Neuropsychology

Mini-Mental State Examination and Alzheimer’s Disease Assessment Scale (ADAS-Cog 13) at baseline and one-year follow-up, and summary scores for neuropsychometric tests were obtained for all subjects. Neuropsychometric composite scores evaluated memory, executive function, language, and visuospatial performance.

Imaging

All participants underwent [¹⁸F]AV1451-PET, [¹⁸F]FDG-PET, and 3-Tesla T1-weighted scans as per ADNI protocol.

[¹⁸F]AV1451-PET and [¹⁸F]FDG-PET scans were processed using SPM12 (Wellcome Centre for Human Neuroimaging, UCL, London, UK) in MATLAB (v2018a) at Imperial College London to evaluate tau deposition and glucose metabolism respectively. PET data were coregistered to individual participants’ T1 volumetric MRI. The MRI and PET scans were then normalized to the T1 Montreal Neurological Institute (MNI) template. The normalized T1 volumetric MRI scan was segmented into grey matter, white matter, and cerebrospinal fluid. Individual object maps were created for each participant by applying the inverse deformation fields acquired from segmentation to a probabilistic brain atlas. 21 For [¹⁸F]AV145-PET, standard uptake value ratio (SUVR) images were created by dividing the target region by the cerebellar uptake, while for [¹⁸F]FDG-PET, pons was used as a reference region.

Image analysis and statistics

A flow diagram depicting the image processing and analysis pipelines is shown in Fig. 1. Once SUVR images were created, single subject analysis to assess the cluster level differences to a sample of healthy controls, voxel-wise regression between [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET SUVR images, and region of interest analysis were performed as described in detail below.

Fig. 1

Methods for [¹⁸F]FDG and [¹⁸F]AV1451 image processing pipeline and statistical analysis. As shown in the left and right panels, [¹⁸F]FDG and [¹⁸F]AV1451 PET scans were reregistered to MRI and normalized to MNI space and divided by pons for [¹⁸F]FDG and cerebellum for [¹⁸F]AV1451 to generate SUVR images. The central panel shows the methods of statistical analysis performed on the SUVR images; single subject analysis to assess the cluster level differences to a sample of healthy controls, voxel-wise regression between [¹⁸F]FDG and [¹⁸F]AV1451 SUVR images, and region of interest analysis to evaluate the relationship in specific areas of interest.

Single-subject voxel-level analysis

Single-subject analysis was performed to account for the individual variation between subjects enabling to distinguish increases/decreases at individual voxels. Additionally, the single-subject analysis accounts for the off-target binding of PET radiotracers, which is an important limitation of [¹⁸F]AV1451-PET. Single-subject SPM analysis was conducted on both [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET in FSL (the FMRIB Software Library). To perform this analysis, the [¹⁸F]FDG-PET SUVR image and [¹⁸F]AV1451-PET SUVR image of each cognitively impaired participant were compared against the respective CN group using a non-parametric two-sample t-test and a threshold-free cluster enhancement approach, implemented in the FSL randomize tool as detailed below. 22

Global hypometabolism

Each MCI and AD participants’ [¹⁸F]FDG-PET SUVR image was tested for the level of hypometabolism by performing this two-sample t-test against 70 cognitive normal healthy controls on a voxel-to-voxel basis. The statistical threshold of the resulting hypometabolic maps created in FSL was set at p < 0.05. Significant clusters of [¹⁸F]FDG-PET hypometabolic t-value maps were extracted as voxel of interest maps. The volume of significant hypermetabolic clusters (mm³) and the mean t-value of hypometabolic maps were calculated with fslstats. The volume of hypometabolic clusters was then multiplied by the mean t-value, to produce a measure to reflect the total volume x intensity of global hypometabolism. For each patient, this produced a single value which reflects the total number and intensity of significant voxels [¹⁸F]FDG-PET (decreases) [value used for all subsequent analysis, referred to as ‘global hypometabolism’].

Global cortical tau load

To calculate global tau burden, MCI and AD participants’ [¹⁸F]FDG-PET SUVR image was tested for the level of tau deposition by performing a two-sample t-test against 70 cognitive normal healthy controls on a voxel-to-voxel basis. Statistical threshold for resulting tau deposition maps created in FSL was set at p < 0.05. Significant clusters of [¹⁸F]AV1451-PET t-value maps were extracted as voxel of interest maps. The volume of significant clusters of tau deposition (mm³) and the mean t-value of tau deposition maps were calculated with fslstats. The volume of clusters of tau deposition was then multiplied by the mean t-value, to produce a measure to reflect the total volume x intensity of tau deposition. For each patient, this produced a single value which reflects the total number and intensity of significant voxels [¹⁸F]AV1451 (increased) [value used for all subsequent analysis, referred to as ‘global cortical tau load’].

Statistical analysis

Spearman’s correlations analysis was performed to identify the relationship between ‘global cortical tau load’ and ‘global hypometabolism’. To evaluate the global relationship between hypometabolism and tau load, linear regression corrected for age was also performed as follows: Global hypometabolism ∼ b₀ + b₁Global cortical tau load + b₂Age.

Group voxelwise linear regression

Voxelwise linear regression enables the evaluation the relationship between PET imaging biomarkers across cortical regions without the requirement for a priori assumptions where regional effects may occur. To explore the relationship between glucose metabolism and tau deposition at group-level across the cortex, voxelwise correlation between [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET SUVR images was performed using VoxelStats implemented on MATLAB 2018a. 23 Age was included as a confounding variable with a random field theory-based correction threshold of p < 0.001. We used a linear model using the following formulae: [¹⁸F]FDG ∼ b₀ + b₁[¹⁸F]AV1451 + b₂Age. The t-statistical outputs were saved in nifty format, and regions with significant voxels were localized.

Region of interest analysis

To compare global cortical load and global hypometabolism with regional uptake using the SUVR map, we evaluated [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET regional uptake for the whole cortex, frontal, temporal, parietal, and occipital lobes by sampling the corresponding SUVR maps. As the anterior and posterior cingulate, thalamus, striatum, medial temporal lobe (MTL), and hippocampus are of significant interest in AD, these regions were sampled separately. Pearson’s correlations analysis was performed to identify the relationship between [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET tracer uptake with a false discovery rate (FDR) correction applied (q < 0.05).

Relationships with neuropsychological performance and clinical progression

To evaluate the predictive value of [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET tracer uptake on performance in cognitive domains (memory, executive functioning, language, and visuospatial), a multivariate analysis was performed using the following formula: $\begin{matrix} Neuropsychometric {domain \sim b}_{0} {+ b}_{1} \\ Global cortical tau {load + b}_{2} \\ Global {hypometabolism + b}_{3} Age \end{matrix}$

To explore the predictive value of global cortical tau uptake/global hypometabolism as well as regional temporal lobe SUVR uptake on baseline ADAS-cog 13 performance and clinical deterioration, we performed a regression analysis using a linear model based on the change in 1-year cognitive scores as follows: $\begin{matrix} ADAS - Cog {13 \sim b}_{0} {+ b}_{1} Temporal lobe [18 F] AV 1451 \\ {SUVR + b}_{2} Temporal lobe [18 F] FDG {SUVR + b}_{3} \\ Age \\ ADAS - Cog 13 1 year {change \sim b}_{0} {+ b}_{1} Temporal \\ lobe [18 F] AV 1451 {SUVR + b}_{2} Temporal lobe \\ [18 F] FDG {SUVR + b}_{3} Age \\ ADAS - Cog {13 \sim b}_{0} {+ b}_{1} Global cortical tau \\ {load + b}_{2} Global {hypometabolism + b}_{3} Age \\ ADAS - Cog 13 1 year {change \sim b}_{0} {+ b}_{1} Global \\ cortical tau {load + b}_{2} Global {hypometabolism + b}_{3} \\ Age \\ {(Where b}_{0} = intercept and b_{n} = slope coefficients) \end{matrix}$

Correlation analysis and linear models were performed on r v4.0.4 (The R Foundation for Statistical Computing), with a p < 0.05 significance threshold and confidence intervals (CI) reported throughout.

RESULTS

Baseline demographics

AD and MCI participants did not differ in age. AD participants had lower Mini-Mental State Examination and higher ADAS-Cog 13 scores (Table 1).

Table 1

Descriptive statistics (mean±SD) of single subject analysis and MRI volumes

Diagnostic group	AD Aβ positive	MCI Aβ positive
N	44	95
Mean global cortical tau load^*	2395200±2398490	1116184±1199439
Percentage of subjects with increased global cortical tau load	100%	100%
Mean global hypometabolism^*	797017±376543	473776.7±391887.4
Percentage of subjects with global hypometabolism	100%	98%
Hippocampal tau^†	1.43±0.21	1.36±0.22
MTL tau^†	1.46±0.21	1.35±0.23
Temporal lobe tau^†	1.51±0.36	1.31±0.25
Hippocampal glucose metabolism^†	0.99±0.14	1.07±0.08
MTL glucose metabolism^†	0.97±0.14	1.04±0.08
Temporal lobe glucose metabolism^†	1.17±0.12	1.25±0.10
Hippocampus volume (mm³)	6174±1101	6982±948
MTL volume (mm³)	15581±2859	17479±2279
Temporal lobe volume (mm³)	93043±12501	99044±12640
Total grey matter volume (mm³)	552981±57678	576690±51132

^*Calculated from the single subject analysis, significant cluster increases/decreases in comparison to healthy controls. ^†denotes the standard uptake ratio values generated from the region of interest analysis of [¹⁸F]AV1451 and [¹⁸F]FDG as markers of tau deposition and glucose metabolism respectively.

Association between [¹⁸F]FDG and [¹⁸F]AV1451

Single-subject voxel-level analysis

Details on the number and percentage of participants with global cortical tau load increases and global hypometabolism are outlined in Table 1. All AD and amyloid-positive MCI subjects (100%) showed an increase in global cortical tau load. All amyloid-positive AD subjects (100%) and 98% of amyloid-positive MCI subjects demonstrated global hypometabolism.

Spearman’s correlations showed no correlation in the single-subject SPM analysis between global hypometabolism and global cortical tau load in amyloid-positive subjects (AD, r = 0.170, p = 0.271, 95% CI: –0.143 to 0.451; MCI, r = 0.087, p = 0.397, 95% CI: –0.120 to 0.287), implying that these processes are not closely associated.

On linear models controlling for age, global cortical tau load failed to predict global hypometabolism in amyloid-positive AD subjects (b = 0.165, p = 0.426, 95% CI: –0.25 to 0.58). In contrast, global cortical tau load was a significant predictor of global hypometabolism, in amyloid-positive MCI subjects (b = 0.257, p = 0.011, 95% CI: 0.06 to 0.45) (Fig. 2). Perhaps the influence of other pathological processes on glucose metabolism in later disease explains this dissociation in AD patients compared to MCI stage.

Fig. 2

Relationship between global cortical tau load and global hypometabolism. Figure 2 shows individual participants’ global cortical tau and corresponding level of global hypometabolism. Panel A shows a dissociation between global cortical tau and global hypometabolism in AD. Panel B shows a weak association between global tau load and global hypometabolism in amyloid-positive MCI.

Group-level voxel-wise associations

Voxel-wise regression using Voxelstats showed a negative correlation between glucose metabolism and tau deposition in small clusters localized in the frontal, temporal, and occipital lobe in AD. In amyloid-positive MCI subjects a stronger negative correlation (larger cluster) was shown with maximum correlations shown within the parietal and temporal lobe (Fig. 3).

Fig. 3

Global voxel-level relationship between [¹⁸F]AV1451 and [¹⁸F]FDG. Figure 3 shows the significant negative association between tau deposition and glucose metabolism in grey matter voxels. Panel A shows the relationship between tau deposition and glucose hypometabolism in Alzheimer’s disease, limited to several small clusters in frontal, temporal, and occipital regions. Panel B shows the relationship in amyloid-positive MCI subjects. Age was included as a confounding factor with results shown after a random field theory correction was applied (p < 0.001).

ROI analysis

Consistent with the findings in the voxelwise analysis, at the ROI level, amyloid-positive AD subjects demonstrated that in the MTL, lower glucose metabolism correlated with higher tau (r = –0.471, adjusted p = 0.014, 95% CI: –0.673 to –0.202) (Fig. 4). However, there were no significant correlations in the other regions following FDR correction.

Fig. 4

Heatmaps depicting the Pearson’s correlation coefficients between glucose metabolism and tau deposition in all regions of interest. Panel A shows a negative correlation throughout most ROIs, most notably in temporal lobe structures in AD subjects. Panel B shows both positive and negative correlations in amyloid-positive MCI subjects. ACC, anterior cingulate cortex; FL, frontal lobe; Hipp, hippocampus; MTL, medial temporal lobe; OL, occipital lobe; PL, parietal lobe; PC, posterior cingulate; St, striatum; TL, temporal lobe; Ts, thalamus; WB, whole brain.

Similarly, in amyloid-positive MCI subjects, our regional analysis demonstrated that lower glucose metabolism uptake correlated with higher tau in the temporal lobe and the medial temporal lobe (r = –0.315, adjusted p = 0.010, 95% CI: –0.486 to –0.121 and r = –0.292, adjusted p = 0.015, 95% CI: –0.466 to –0.096, respectively).

As several regions of interest were evaluated, FDR multiple comparison correction may limit the statistical power to detect a region-of-interest correlation that were detected in small clusters such as frontal lobe regions in AD and MCI subjects.

Relationship between regional and global tau, glucose metabolism, and neuropsychological performance and clinical progression

Global hypometabolism and global cortical tau

The multivariate analysis showed that global cortical tau load was a strong predictor of memory, executive function, language, and visuospatial performance as well as baseline ADAS-Cog 13 score and clinical progression (1-year ADAS-Cog 13 change) in amyloid-positive subjects. Global hypometabolism was only a significant predictor of worse memory and executive function scores in amyloid-positive MCI subjects.

Regarding clinical status and progression, global cortical tau load effectively predicted ADAS-Cog 13 performance at baseline and clinical progression in amyloid-positive MCI and AD subjects. Global hypometabolism significantly predicted clinical progression in amyloid-positive MCI and AD subjects. Interestingly, global hypometabolism failed to predict baseline ADAS-Cog 13 scores (Table 2). Global cortical tau was the better indicator of neuropsychological performance and clinical progression.

Table 2

Multivariate analysis of glucose metabolism and tau deposition as predictors of neuropsychological performance and clinical progression in Aβ+ subjects

Multivariate analysis	Mean	Standardized	t	p
coefficient β (95% CI)
AD subjects
Neuropsych Memory (n = 44)	–0.764±0.58
Global cortical tau load^*	2395200±2398490	–0.458 (–0.82, –0.09)	–2.516	0.016^*
Global hypometabolism^*	797017±376543	–0.137 (–0.41, 0.14)	–0.997	0.325
Age	76.7±8.86	0.021 (–0.34, 0.39)	0.118	0.907
Neuropsych Executive function (n = 44)	–0.817±1.11
Global cortical tau load^*	2395200±2398490	–0.599 (–0.96, –0.23)	–2.319	0.002^*
Global hypometabolism^*	797017±376543	–0.163 (–0.44, 0.11)	–1.194	0.240
Age	76.7±8.86	–0.249 (–0.61, 0.11)	–1.391	0.172
Neuropsych language (n = 44)	–0.684±1.01
Global cortical tau load^*	2395200±2398490	–0.462 (–0.85, –0.07)	–2.404	0.021^*
Global hypometabolism^*	797017±376543	–0.176 (–0.47, 0.12)	–1.210	0.233
Age	76.7±8.86	–0.254 (–0.64, 0.13)	–1.328	0.192
Neuropsych Visuospatial (n = 44)	–0.720±1.11
Global cortical tau load^*	2395200±2398490	–0.284 (–0.67, 0.11)	–1.463	0.151
Global hypometabolism^*	797017±376543	0.040 (–0.26, 0.34)	0.273	0.786
Age	76.7±8.86	0.151 (–0.24, 0.54)	0.783	0.438
Baseline ADAS-cog 13 (n = 43)	33.2±7.22
Temporal AV1451 SUVR	1.481±0.31	0.461 (0.12, 0.80)	2.751	0.009^*
Temporal lobe FDG SUVR	1.179±0.11	–0.221 (–0.51, 0.07)	–1.534	0.133
Age	76.9±8.81	0.239 (–0.10, 0.58)	1.438	0.158
Baseline ADAS-cog 13 (n = 43)	33.2±7.22
Global cortical tau load^*	2273020±2284131	0.520 (0.14, 0.90)	2.754	0.009^*
Global hypometabolism^*	788412±376597	0.116 (–0.18, 0.41)	0.797	0.430
Age	76.9±8.81	0.327 (–0.05, 0.71)	1.734	0.098
ADAS-cog 13 score 1-year increase (n = 30)	4.34±5.70
Temporal lobe AV1451 SUVR	1.527±0.33	0.504 (0.06, 0.95)	2.329	0.028^*
Temporal lobe FDG SUVR	1.178±0.12	–0.071 (–0.40, 0.26)	–0.436	0.666
Age	76.9±8.98	–0.081 (–0.52, 0.36)	–0.379	0.708
ADAS-cog 13 score 1-year increase (n = 30)	4.34±5.70
Global cortical tau load^*	2539200±2478189	0.805 (0.33, 1.28)	3.510	0.002^*
Global hypometabolism^*	838937±387222	0.421(0.12, 0.72)	2.885	0.008^*
Age	76.9±8.98	0.292(–0.19, 0.77)	1.250	0.222
MCI subjects
Neuropsych Memory (n = 95)	0.148±0.60
Global cortical tau load^*	1116184±1199439	–0.432 (–0.63, –0.24)	–4.442	<0.001^*
Global hypometabolism^*	473776.7±391887.4	–0.210 (–0.41, –0.01)	–2.126	0.036^*
Age	74.3±6.88	–0.171 (–0.37, 0.03)	–1.671	0.098
Neuropsych Executive function (n = 95)	0.328±0.97
Global cortical tau load^*	1116184±1199439	–0.347 (–0.54, –0.15)	–3.555	<0.001^*
Global hypometabolism^*	473776.7±391887.4	–0.237 (–0.43, –0.04)	–2.389	0.019^*
Age	74.3±6.88	–0.271 (–0.47, –0.07)	–2.631	0.010^*
Neuropsych language (n = 95)	0.325±0.83
Global cortical tau load^*	1116184±1199439	–0.283 (–0.49, –0.07)	–2.649	0.010^*
Global hypometabolism^*	473776.7±391887.4	–0.113 (–0.32, 0.10)	–1.038	0.302
Age	74.3±6.88	–0.172 (–0.40, 0.05)	–1.525	0.131
Neuropsych Visuospatial (n = 95)	0.012±0.73
Global cortical tau load^*	1116184±1199439	–0.255 (–0.47, –0.04)	–2.353	0.021^*
Global hypometabolism^*	473776.7±391887.4	–0.021 (–0.24, 0.20)	–0.191	0.849
Age	74.3±6.88	–0.222 (–0.45, 0.01)	–1.937	0.056
Baseline ADAS-cog 13 (n = 94)	20.8±6.77
Temporal lobe AV1451 SUVR	1.314±0.25	0.474 (0.29, 0.66)	5.025	<0.001^*
Temporal lobe FDG SUVR	1.251±0.10	–0.229 (–0.42, –0.04)	–2.410	0.018^*
Age	74.3±6.92	0.204 (0.02, 0.39)	2.202	0.030^*
Baseline ADAS-cog 13 (n = 94)	20.8±6.77
Global cortical tau load^*	1118399±1205675	0.476 (0.29, 0.67)	4.988	<0.001^*
Global hypometabolism^*	473991±393983	0.114 (–0.01, 0.38)	1.091	0.064
Age	74.3±6.92	0.229 (0.02, 0.43)	2.706	0.025^*
ADAS-cog 13 score 1-year increase (n = 67)	1.990±5.62
Temporal lobe AV1451 SUVR	1.301±0.24	0.271 (0.02, 0.53)	2.128	0.037^*
Temporal lobe FDG SUVR	1.260±0.10	–0.179 (–0.44, 0.08)	–1.385	0.171
Age	74.3±6.85	0.063 (–0.20, 0.32)	0.480	0.633
ADAS-cog 13 score 1-year increase (n = 67)	1.990±5.62
Global cortical tau load^*	1074894±1207688	0.293 (0.05, 0.54)	2.378	0.021^*
Global hypometabolism^*	426768±379653	0.283 (0.03, 0.53)	3.823	0.024^*
Age	74.3±6.85	0.058 (–0.21, 0.31)	1.763	0.693

^*Calculated from the single subject analysis, significant cluster increases/decreases of [¹⁸F]AV1451/[¹⁸F]FDG signal respectively in comparison to a sample of healthy controls.

Regional analysis

At a regional level, temporal lobe tau was a significant predictor of baseline ADAS-Cog 13 score and clinical progression (1-year ADAS-Cog 13 change) in amyloid-positive subjects. However, in AD and MCI subjects, global cortical tau was a better predictor of progression and associated better with the clinical status compared to the temporal lobe tau, as shown by the standardized betas (Baseline ADAS-cog 13: AD, Global cortical tau b = 0.520, Temporal tau b = 0.461, MCI, Global cortical tau b = 0.476, Temporal tau b = 0.474. ADAS-cog 13 score 1-year increase: AD, Global cortical tau b = 0.805, Temporal tau b = 0.504, MCI, Global cortical tau b = 0.293 Temporal tau b = 0.271).

Regionally, temporal lobe [¹⁸F]FDG-PET only predicted baseline ADAS-Cog 13 score in amyloid-positive MCI subjects but did not predict clinical progression (Table 2). Global hypometabolism outperformed regional [¹⁸F]FDG-PET as a predictor of clinical progression in AD and MCI. (Baseline ADAS-cog 13: AD, Global hypometabolism b = 0.116, Temporal FDG b = –0.221, MCI, Global hypometabolism b = 0.114, Temporal FDG b = –0.229. ADAS-cog 13 score 1-year increase: AD, Global hypometabolism b = 0.421, Temporal FDG b = –0.071, MCI, Global hypometabolism b = 0.283 Temporal FDG b = –0.179).

DISCUSSION

In this study, we have demonstrated that at the global level, using single-subject SPM analysis, there was no association between tau deposition and glucose metabolism in AD, whereas, in amyloid-positive MCI subjects, a minimal correlation was found. Tau deposition better predicted neuropsychological performance and clinical progression compared to glucose metabolism. Global increases of tau showed higher accuracy than regional analysis in predicting clinical status and progression in AD and MCI.

Thus, we demonstrate that tau and glucose hypometabolism have distinct trajectories during the disease. The initial neurofibrillary tangle deposition in the neuron may lead to changes in glucose metabolism. However, as the disease progresses, glucose metabolism may be influenced by the interaction between tau and amyloid deposition. Additionally, glucose hypometabolism may be affected by the changes in microglial and astrocyte activation, oxidative stress, and neuroinflammation.

Dissociation between global tau and glucose hypometabolism

To the best of our knowledge, this is the first paper that evaluates the relationship between global tau load and global cerebral glucose metabolism in a single-subject voxel-level design, enabling a novel evaluation of these pathological hallmarks. We demonstrate a general dissociation between global cortical tau load and global hypometabolism along the AD trajectory. Studies have previously demonstrated an inverse relationship between glucose metabolism and increased tau in AD at a regional level, primarily in limbic regions.24 –26 In contrast, global tau deposition fails to display a close relationship with glucose metabolic changes in advanced AD. While tau represents the main driver of neurodegeneration in AD leading to neuronal loss and subsequent hypometabolism, we indicate that this detrimental influence of tau aggregation on glucose metabolic deficits only displays a strong correlation within temporal lobe regions in region-of-interest analysis. 27 Voxelwise linear regression did reveal a relationship within occipital and frontal lobe regions in AD, but this was confined to relatively small clusters of voxels. Glucose metabolism is influenced by several additional factors in AD, including amyloid deposition, inflammation, and oxidative stress.28 –30 As neuroinflammation increases, brain energy metabolism diminishes as the disease progresses. 29 Oxidative damage affects processes that result in synaptic and neuronal loss, which ultimately reduces brain glucose utilization. 30 It is also suggested that glial activation plays a significant role in glucose metabolism. 17 Thus, tau deposition and glucose metabolism can provide independent information which requires separate consideration in the evaluation of AD. It should be considered that the dissociation between tau and glucose metabolism may also reflect non-AD pathology. [¹⁸F]FDG-PET signal is abnormal in synucleinopathies and following stroke which could account for the dissociation in the MCI stage.31,32, 31,32

Aggregation of tau predicts neuropsychological performance and clinical progression

We identified that global tau is a significant predictor of memory, language, visuospatial, and executive function impairment in AD/MCI stage, whereas, glucose hypometabolism was only associated with dysfunctional memory and executive function scores in MCI subjects in this population. Literature has previously demonstrated that temporoparietal tau is closely related to cognitive functioning in prodromal AD, whilst measures of cortical thickness and amyloid-β deposition have failed to predict a subject’s cognitive status. 10 Spatiotemporal differences in tau deposition influence the cognitive domains most greatly affected in AD patients. 33

To the best of our knowledge, we further provide the first examination of the predictive utility of single-subject global tau load and global [¹⁸F]FDG-PET on clinical progression. We demonstrate that global cortical increases of tau in our single-subject analysis were strongly predictive of cognitive decline as shown by increased ADAS-Cog13 scores in patients with AD and amyloid-positive MCI. Temporal lobe [¹⁸F]FDG-PET failed to significantly predict clinical status at baseline and participants’ longitudinal progression in MCI and AD. While global hypometabolism failed to predict baseline ADAS-Cog13 scores, it was a significant predictor of clinical progression, despite displaying a weaker association than global cortical tau load. In support, a recent study demonstrated that tau PET was the best predictor of clinical progression in MCI and dementia cases compared to amyloid PET, FDG PET, and structural MRI. 34 This demonstrates the clinical utility of tau PET in the diagnosis and prognosis of AD, providing an added value to established neuroimaging markers of the disease. Tau aggregation is, therefore, a better predictor of clinical function and clinical decline than glucose metabolism in the AD trajectory, further highlighting that [¹⁸F]FDG-PET is not simply the inverse mirror image of [¹⁸F]AV1451-PET. While [¹⁸F]FDG-PET imaging is an established tool to explore neurodegenerative processes and consistently correlates with cognitive function we demonstrate that [¹⁸F]FDG-PET cannot replace [¹⁸F]AV1451-PET as an indicator of tau deposition and that [¹⁸F]AV1451-PET is a superior predictor of cognitive function and clinical progression. 35 As such we indicate that these PET biomarkers require independent evaluation.

By combining global tau deposition as well as the intensity of tau uptake in individual voxels, we offer a novel evaluation of tau accumulation in AD. Single-subject voxel-level [¹⁸F]AV1451-PET analysis is an effective method to predict neuropsychological performance, clinical status, and progression, outperforming regional analysis. Original research and clinical intervention studies frequently employ a region-of-interest analysis when utilizing PET tracers, whereby, researchers select one or several a priori regions of interest.36 –38 This can induce bias by the choice of ROI and can cause an issue of multiple comparisons, thus, reducing statistical power.39,40, 39,40 Furthermore, a regional analysis fails to account for the off-target binding limited by PET radiotracers and assumes where differences will exist.16,41, 16,41 Previous literature employing a regional approach frequently fails to acknowledge when ROIs were selected from research conception to the data analysis, which could inflate the reported effect size. 40 A single-subject analysis can provide a single marker of the extent of global binding and intensity, that prevents the confounding issues created by making a priori assumptions. The development of a single-subject global measure could elevate statistical power, account for off-target binding, and reduce researcher bias. With several promising anti-tau agents in clinical development, future trials should consider adopting this methodology. 42

Limitations

One of the limitations of our study is that a follow-up period longer than a 1-year change of ADAS-Cog 13 to track clinical progression would have been ideal, however, it has been demonstrated that the ADAS-Cog 13 is sensitive to detect a mean 4-point decline in mild AD over 12 months. 43

Conclusion

In this study, we demonstrate a dissociation between global hypometabolism and tau deposition in a large sample of cognitively impaired subjects. Retention of [¹⁸F]AV1451-PET closely predicts cognitive impairment and future clinical progression. Single-subject analysis of [¹⁸F]FDG-PET and [¹⁸F]AV1451-PET is a superior technique to establish the expected rate of clinical decline compared to standard regional analysis. Clinical trials must consider that [¹⁸F]FDG-PET is not simply the inverse mirror image of [¹⁸F]AV1451-PET; thus, we argue that glucose metabolism and tau deposition require independent evaluation with single-subject analysis enhancing the sensitivity of clinical studies to demonstrate significant differences.

AUTHOR CONTRIBUTIONS

Joseph Nowell (Conceptualization; Data curation; Formal analysis; Methodology; Validation; Writing – original draft); Sanara Raza (Formal analysis; Methodology); Nicholas R. Livingston (Formal analysis; Investigation; Methodology); Shayndhan Sivanathan (Formal analysis; Investigation; Methodology); Steve Gentleman (Conceptualization; Supervision; Writing – review & editing); Paul Edison (Conceptualization; Resources; Supervision; Visualization; Writing – review & editing).

Footnotes

ACKNOWLEDGMENTS

Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and 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 (). 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.

FUNDING

Joseph Nowell was funded by the Imperial College President’s PhD scholarship.

CONFLICT OF INTEREST

Paul Edison was funded by the Medical Research Council and now by Higher Education Funding Council for England (HEFCE). He has also received grants from Alzheimer’s Research, UK, Alzheimer’s Drug Discovery Foundation, Alzheimer’s Society, UK, Alzheimer’s association, US, Medical Research Council, UK, Novo Nordisk, Piramal Life Sciences and GE Healthcare. P.E. is a consultant to Roche, Pfizer and Novo Nordisk. He has received speaker fees from Novo Nordisk, Pfizer, Nordea, Piramal Life Science, Glaxo Smith Kelin (GSK). He has received educational and research grants from GE Healthcare, Novo Nordisk, Piramal Life Science/Life Molecular Imaging, Avid Radiopharmaceuticals and Eli Lilly. He is an external consultant to Novo Nordisk and Cytodyn and is a member of their Scientific Advisory Board.

The other authors report no disclosures.

DATA AVAILABILITY

All data necessary for the conclusion of this paper are presented in the manuscript. Data used for the current manuscript are available at .

References

Naseri

, Wang

, Guo

, et al. The complexity of tau in Alzheimer’s disease. Neurosci Lett 2019; 705: 183–194.

Braak

, Thal

, Ghebremedhin

, et al. Stages of the pathologic process in Alzheimer disease: age categories from 1 to 100 years. J Neuropathol Exp Neurol 2011; 70: 960–969.

Braak

, Alafuzoff

, Arzberger

, et al. Staging of Alzheimer disease-associated neurofibrillary pathology using paraffin sections and immunocytochemistry. Acta Neuropathol 2006; 112: 389–404.

Schultz

, Gordon

, Mishra

, et al. Widespread distribution of tauopathy in preclinical Alzheimer’s disease. Neurobiol Aging 2018; 72: 177–185.

Lowe

, Wiste

, Senjem

, et al. Widespread brain tau and its association with ageing, Braak stage and Alzheimer’s dementia. Brain 2018; 141: 271–287.

Chetelat

, Arbizu

, Barthel

, et al. Amyloid-PET and (18)F-FDG-PET in the diagnostic investigation of Alzheimer’s disease and other dementias. Lancet Neurol 2020; 19: 951–962.

Jack

Jr , Andrews

, Beach

, et al. Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup. Alzheimers Dement 2024; doi: 10.1002/alz.13859.

Smith

, Hägerström

, Pawlik

, et al. Clinical utility of tau positron emission tomography in the diagnostic workup of patients with cognitive symptoms. JAMA Neurology 2023; 80: 749–756.

Rabinovici

, Knopman

, Arbizu

, et al. Updated appropriate use criteria for amyloid and tau PET. Alzheimers Dement 2023; 19: e078912.

10.

Ossenkoppele

, Smith

, Ohlsson

, et al. Associations between tau, Abeta, and cortical thickness with cognition in Alzheimer disease. Neurology 2019; 92: e601–e612.

11.

Ottoy

, Niemantsverdriet

, Verhaeghe

, et al. Association of short-term cognitive decline and MCI-to-AD dementia conversion with CSF, MRI, amyloid- and (18)F-FDG-PET imaging. Neuroimage Clin 2019; 22: 101771.

12.

Chiotis

, Saint-Aubert

, Savitcheva

, et al. Imaging in-vivo tau pathology in Alzheimer’s disease with THK5317 PET in a multimodal paradigm. Eur J Nucl Med Mol Imaging 2016; 43: 1686–1699.

13.

Hanseeuw

, Betensky

, Schultz

, et al. Fluorodeoxyglucose metabolism associated with tau-amyloid interaction predicts memory decline. Ann Neurol 2017; 81: 583–596.

14.

Sintini

, Schwarz

, Martin

, et al. Regional multimodal relationships between tau, hypometabolism, atrophy, and fractional anisotropy in atypical Alzheimer’s disease. Hum Brain Mapp 2019; 40: 1618–1631.

15.

van Eimeren

, Bischof

and Drzezga

. Is tau imaging more than just upside-down (18)F-FDG imaging? J Nucl Med 2017; 58: 1357–1359.

16.

Lemoine

, Leuzy

, Chiotis

, et al. Tau positron emission tomography imaging in tauopathies: The added hurdle of off-target binding. Alzheimers Dement (Amst) 2018; 10: 232–236.

17.

Xiang

, Wind

, Wiedemann

, et al. Microglial activation states drive glucose uptake and FDG-PET alterations in neurodegenerative diseases. Sci Transl Med 2021; 13: eabe5640.

18.

Therriault

, Pascoal

, Sefranek

, et al. Amyloid-dependent and amyloid-independent effects of Tau in individuals without dementia. Ann Clin Transl Neurol 2021; 8: 2083–2092.

19.

Pooler

, Polydoro

, Maury

, et al. Amyloid accelerates tau propagation and toxicity in a model of early Alzheimer’s disease. Acta Neuropathol Commun 2015; 3: 14.

20.

Jagust

, Landau

, Koeppe

, et al. The Alzheimer’s disease neuroimaging initiative 2 PET core: 2015. Alzheimers Dement 2015; 11: 757–771.

21.

Hammers

, Allom

, Koepp

, et al. Three-dimensional maximum probability atlas of the human brain, with particular reference to the temporal lobe. Hum Brain Mapp 2003; 19: 224–247.

22.

Winkler

, Ridgway

, Webster

, et al. Permutation inference for the general linear model. NeuroImage 2014; 92: 381–397.

23.

Mathotaarachchi

, Wang

, Shin

, et al. VoxelStats: A MATLAB package for multi-modal voxel-wise brain image analysis. Front Neuroinform 2016; 10: 20.

24.

Ossenkoppele

, Schonhaut

, Scholl

, et al. Tau PET patterns mirror clinical and neuroanatomical variability in Alzheimer’s disease. Brain 2016; 139: 1551–1567.

25.

Pascoal

, Mathotaarachchi

, Mohades

, et al. Amyloid-beta and hyperphosphorylated tau synergy drives metabolic decline in preclinical Alzheimer’s disease. Mol Psychiatry 2017; 22: 306–311.

26.

Bischof

, Jessen

, Fliessbach

, et al. Impact of tau and amyloid burden on glucose metabolism in Alzheimer’s disease. Ann Clin Transl Neurol 2016; 3: 934–939.

27.

, Rodriguez

, Herman

, et al. Tau pathology induces excitatory neuron loss, grid cell dysfunction, and spatial memory deficits reminiscent of early Alzheimer’s disease. Neuron 2017; 93: 533–541e535.

28.

Carbonell

, Zijdenbos

, Bedell

, et al. Spatially distributed amyloid-beta reduces glucose metabolism in mild cognitive impairment. J Alzheimers Dis 2020; 73: 543–557.

29.

Takkinen

, Lopez-Picon

, Al Majidi

, et al. Brain energy metabolism and neuroinflammation in ageing APP/PS1-21 mice using longitudinal (18)F-FDG and (18)F-DPA-714 PET imaging. J Cereb Blood Flow Metab 2017; 37: 2870–2882.

30.

Butterfield

and Halliwell

. Oxidative stress, dysfunctional glucose metabolism and Alzheimer disease. Nat Rev Neurosci 2019; 20: 148–160.

31.

Meyer

, Frings

, Rücker

, et al. 18F-FDG PET in parkinsonism: differential diagnosis and evaluation of cognitive impairment. J Nucl Med 2017; 58: 1888–1898.

32.

Bunevicius

, Yuan

and Lin

. The potential roles of 18F-FDG-PET in management of acute stroke patients. Biomed Res Int 2013; 2013: 634598–634598.

33.

Vogel

, Young

, Oxtoby

, et al. Four distinct trajectories of tau deposition identified in Alzheimer’s disease. Nat Med 2021; 27: 871–881.

34.

Boccalini

, Ribaldi

, Hristovska

, et al. The impact of tau deposition and hypometabolism on cognitive impairment and longitudinal cognitive decline. Alzheimers Dement 2024; 20: 221–233.

35.

, Xu

, Li

, et al. FDG-PET as an independent biomarker for Alzheimer’s biological diagnosis: a longitudinal study. Alzheimers Res Ther 2019; 11: 57.

36.

Gejl

, Gjedde

, Egefjord

, et al. In Alzheimer’s disease, 6-month treatment with GLP-1 analog prevents decline of brain glucose metabolism: randomized, placebo-controlled, double-blind clinical trial. Front Aging Neurosci 2016; 8: 108.

37.

Lee

, Brown

, Kim

, et al. Regional Aβ-tau interactions promote onset and acceleration of Alzheimer’s disease tau spreading. Neuron 2022; 110: 1932–1943.e1935.

38.

Rondina

, Ferreira

, de Souza Duran

, et al. Selecting the most relevant brain regions to discriminate Alzheimer’s disease patients from healthy controls using multiple kernel learning: A comparison across functional and structural imaging modalities and atlases. Neuroimage Clin 2018; 17: 628–641.

39.

Proesmans

, Raedt

, Germonpré

, et al. Voxel-based analysis of [18F]-FDG brain PET in rats using data-driven normalization. Front Med (Lausanne) 2021; 8: 744157.

40.

Gentili

, Cecchetti

, Handjaras

, et al. The case for preregistering all region of interest (ROI) analyses in neuroimaging research. Eur J Neurosci 2021; 53: 357–361.

41.

Maschio

and Ni

. Amyloid and tau positron emission tomography imaging in Alzheimer’s disease and other tauopathies. Front Aging Neurosci 2022; 14: 838034.

42.

Jadhav

, Avila

, Scholl

, et al. A walk through tau therapeutic strategies. Acta Neuropathol Commun 2019; 7: 22.

43.

Podhorna

, Krahnke

, Shear

, et al. Alzheimer’s Disease Assessment Scale-Cognitive subscale variants in mild cognitive impairment and mild Alzheimer’s disease: change over time and the effect of enrichment strategies. Alzheimers Res Ther 2016; 8: 8.

Do Tau Deposition and Glucose Metabolism Dissociate in Alzheimer’s Disease Trajectory?

Abstract

Background:

Objective:

Methods:

Results:

Conclusions:

Keywords

INTRODUCTION

MATERIALS AND METHODS

Participants

Standard protocol approvals, registrations, and patient consents

Neuropsychology

Imaging

Image analysis and statistics

Single-subject voxel-level analysis

Global hypometabolism

Global cortical tau load

Statistical analysis

Group voxelwise linear regression

Region of interest analysis

Relationships with neuropsychological performance and clinical progression

RESULTS

Baseline demographics

Association between [18F]FDG and [18F]AV1451

Single-subject voxel-level analysis

Group-level voxel-wise associations

ROI analysis

Relationship between regional and global tau, glucose metabolism, and neuropsychological performance and clinical progression

Global hypometabolism and global cortical tau

Regional analysis

DISCUSSION

Dissociation between global tau and glucose hypometabolism

Aggregation of tau predicts neuropsychological performance and clinical progression

Limitations

Conclusion

AUTHOR CONTRIBUTIONS

Footnotes

ACKNOWLEDGMENTS

FUNDING

CONFLICT OF INTEREST

DATA AVAILABILITY

References

Association between [¹⁸F]FDG and [¹⁸F]AV1451