Elevated Tau PET Signal Depends on Abnormal Amyloid Levels and Correlates with Cognitive Impairment in Elderly Persons without Dementia

INTRODUCTION

Alzheimer’s disease (AD) is characterized by the pathological deposition of both amyloid-β (Aβ) and tau. In the research framework proposed by the National Institute on Aging-Alzheimer’s Association (NIA-AA) workgroup, Alzheimer’s pathologic change refers to the early stage of the Alzheimer’s continuum, defined in vivo by biomarkers as abnormal Aβ deposition but normal tau status. The term “Alzheimer’s disease” would be applied when biomarker evidence of both Aβ and pathologic tau was present [1]. The Aβ cascade hypothesis proposes that amyloidosis induces the spread of pathologic tau and that pathological tau subsequently leads to neurodegeneration which eventually causes cognitive decline [1–4].

Recently, the developed positron emission tomography (PET) ligands for Aβ [5, 6] and tau [7] permit the investigation of these two neuropathological hallmarks of AD in vivo. Previous PET imaging studies showed that in clinically normal subjects with a negative amyloid-PET scan, tau-PET retention was limited largely to medial temporal lobe [8], whereas in subjects with positive amyloid PET scans, cross-sectional comparisons suggested increasing and progressive tau-PET retention in widespread cortices [8].

Despite the growing cross-sectional tau-PET imaging literature, only a few studies examined longitudinal change in tau-PET signal. Previous longitudinal studies indicated that the rate of tau retention associated with baseline tau retention and Aβ status measured by amyloid-PET [8–10]. Further evaluation of the way in which tau changes over time could be critical to understand the interaction between amyloid and tau in promoting neurodegeneration and cognitive impairment in AD [8].

Cortical composite and regional tau-PET standard uptake value ratio (SUVr) has been reported to correlate with cognitive impairment [11–16], but the relationship between tau-PET SUVr and specific cognitive domains varies across studies [15–19]. Besides, inconsistent results were found on how Aβ and tau correlated with cognition. While some studies concluded that only tau tangles correlated with cognition and clinical symptoms [13, 20], others showed that Aβ and tau were synergistically associated with cognitive decline [20].

Here, we hypothesize that increased amyloid level can predict increased tau-PET deposit in certain regions of interest (ROI), and increased tau is associated with memory and executive function decline. We stratified a large cohort of non-demented elderly participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) into two groups by Aβ status using florbetapir PET imaging. The differences in baseline retention and prospective change of tau pathology as measured by [F-18]-AV-1451 PET between groups was investigated. We then evaluated the associations of baseline tau burden and baseline Aβ burden with the rate of prospective tau accumulation. Finally, we analyzed the relative contributions of Aβ, tau, and the potential synergy of these pathologies in the prediction of memory and executive function performance and longitudinal decline.

MATERIALS AND METHODS

RESULTS

Table 1 showed the demographics and baseline characteristics for the 389 subjects with available tau-PET and amyloid-PET scans, including 160 Aβ+ subjects and 229 Aβ– subjects. Aβ+ subjects were older (p < 0.001); were more likely to be APOE ɛ4 carriers (p < 0.001); and had lower MMSE score (p < 0.001), higher ADAS11 score (p = 0.001), lower composite memory score (p < 0.001), and lower composite executive function score (p < 0.001). There were no significant differences in educational years and gender composition between Aβ+ and Aβ– subjects. Demographics were also presented by clinical diagnostic groups (Supplementary Table 1).

Table 2 showed the different patterns of cross-sectional tau-PET signal in Aβ+ and Aβ– subjects. In Aβ+ subjects, the temporal composite and other regional tau-PET SUVr was elevated for MCI compared with the CN (Temporal composite: p < 0.001; Frontal: p = 0.002; Occipital: p = 0.02; Parietal: p = 0.002; Temporal: p < 0.001). Within the Aβ- group, the differences in tau-PET SUVr between CN and MCI were detected in temporal composite ROI (p = 0.031), frontal lobe (p = 0.014), and temporal lobe (p = 0.041). Within CN and MCI group, the tau-PET SUVr of Aβ+ subjects significantly elevated in the temporal composite (CN: p < 0.001; MCI: p < 0.001), frontal lobe (CN: p = 0.003; MCI: p = 0.003), parietal lobe (CN: p = 0.002; MCI: p = 0.001), occipital lobe (CN: p = 0.223; MCI: p = 0.007), and temporal lobe (CN: p < 0.001; MCI: p < 0.001). As shown in Fig. 1, temporal composite tau-PET SUVr was higher for MCI group compared to the CN group and for the Aβ+ compared to the Aβ–.

OPEN IN VIEWER

Fig.1

Temporal composite tau-PET SUVr by clinical diagnostic cohort and amyloid status. Box plot of temporal composite tau-PET SUVr by diagnostic cohort and amyloid status of all 389 participants. Participants with amyloid-PET cortical summary SUVr > 1.11 were classified as amyloid-β positive; those with amyloid-PET cortical summary SUVr≤1.11 were classified as the amyloid-β negative. p values are from the analysis of covariance with planned contrast. The box plot whiskers extend to the lowest and highest data points within 1.5 times the inter quartile range from the lower and upper quartiles. The dots represent individual points. Aβ(+), amyloid-β positive; Aβ(–), amyloid-β negative; CN, cognitively normal, MCI, mild cognitive impairment; SUVr, standard uptake value ratio.

Figure 2 showed the significant longitudinal change of temporal composite tau-PET SUVr over 24 months in Aβ+ but not Aβ– subjects. The baseline temporal composite tau retention was significantly higher in the Aβ+ subjects compared to Aβ– subjects [1.25(0.120) versus 1.14(0.083), between-group difference p < 0.001]. In the Aβ+ group, the tau-PET SUVr increased significantly [Estimated annualized change (95% CI)=0.01(0.004∼0.018), p = 0.003]. In the Aβ– group, the tau-PET SUVr remained unchanged over time [Estimated annualized change (95% CI)=0.00(–0.009∼0.006), p = 0.722].

OPEN IN VIEWER

Fig.2

Estimated temporal composite tau-PET SUVr annualized change. Participants with amyloid-PET cortical summary SUVr > 1.11 were classified as amyloid-β positive; those with amyloid-PET cortical summary SUVr≤1.11 were classified as the amyloid-β negative. Aβ(+), amyloid-β positive; Aβ(–), amyloid-β negative; SUVr, standard uptake value ratio.

As shown in Table 3, baseline cortical Aβ burden and baseline temporal composite tau burden were associated with longitudinal regional tau accumulation. Higher levels of Aβ were associated with greater longitudinal tau change in temporal composite ROI [β= 0.05±0.013, p < 0.001], occipital lobe [β= 0.04±0.012, p = 0.005], parietal lobe [β= 0.04±0.012, p = 0.005], and temporal lobe [β= 0.04±0.012, p = 0.001]. Greater baseline tau in temporal composite ROI also related to greater tau accumulation in temporal composite ROI [β= 0.05±0.015, p < 0.001], occipital lobe [β= 0.03±0.014, p = 0.020], parietal lobe [β= 0.04±0.014, p = 0.008], and temporal lobe [β= 0.05±0.014, p = 0.001]. The effect of baseline temporal composite tau on annualized Aβ change was additionally investigated and we did not find evidence of a significant relationship between prospective Aβ change and baseline temporal composite tau [β= 0.02±0.022, p = 0.265].

The associations of Aβ and tau with cognition were lastly examined (Table 4; Supplementary Table 2). Greater temporal cortical tau at baseline was related to poorer memory [β= –1.32±0.464, p = 0.004] and executive function [β= –1.47±0.604, p = 0.015]. In contrast, the relationships between baseline cortical Aβ and cognitive performances did not reach statistical significance and were considerably smaller in magnitude than those observed for baseline temporal composite tau [memory composite: β= –0.24±0.62, p = 0.693; executive functional composite: β= –0.41±0.807, p = 0.611]. In the longitudinal analyses, baseline temporal composite tau predicted greater longitudinal decline in the memory composite [β= –0.41±0.189, p = 0.033], but baseline cortical Aβ was not associated with the memory change. Neither Aβ nor tau was observed to be associated with prospective change of executive function. The interaction between Aβ and tau did not show significant effect on cross-sectional and longitudinal cognitive performance for composites of memory and executive function, which indicated that the effect of tau on cognition was not modified by Aβ.

DISCUSSION

This study investigated the different patterns of cross-sectional tau-PET signal in Aβ+ and Aβ– groups. In the longitudinal analysis, we observed a significant increase in temporal composite ROI over 24 months. This increase was amyloid-dependent, correlated with baseline tau burden. Finally, tau-PET retention significantly explained the variance in memory and executive function composite scores, and this was independent of the amyloid-PET retention. Elevated tau-PET signal could predict prospective memory decline.

According to the research framework proposed by NIA-AA, cognitively unimpaired individuals with normal amyloid PET were not in the Alzheimer’s continuum and cognitively unimpaired individuals with abnormal amyloid PET were early in the Alzheimer’s continuum [1]. We found that elevated tau retention in CN Aβ+ subjects relative to CN Aβ– subjects. This replicated previous reports that Aβ and tau PET measures were correlated, even among CN [16]. Furthermore, this change was not confined to the temporal lobe, suggesting that early tau accumulation may not be as spatially restricted as implied in Braak staging [10, 28]. Taken together, subjects with abnormal amyloid-PET scans may have increasing tau-PET retention in widespread cortices, even at very early stage of AD.

In subjects with positive amyloid-PET scans, cross-sectional comparisons suggested as the disease progresses, pathological tau accumulated within and beyond the temporal lobe. In subjects with negative amyloid-PET scans, the increasing tau retention was limited to temporal lobe and frontal lobe. These findings suggested that Aβ plaques may potentiate the spread of tau beyond the temporal lobe [8]. Previous studies identified a propagation pathway for tau accumulation. Particularly, media/inferior temporal lobe areas projected pathways of tau-propagation toward the anterior pole, lateral and postero-medial temporal cortex, and midline frontal regions such as the orbitofrontal cortex [29]. In the present study, we found the different spatial pattern of tau progression in Aβ+ and Aβ– subjects. This suggested that Aβ might be associated with the tau propagation patterns. Further studies are wanted to investigate of association of Aβ and tau propagation pathway.

Early in the disease, tau-PET SUVr was elevated in the temporal composite ROI (CN Aβ– versus CN Aβ+). This indicated that this ROI could capture the change in tau PET seen early in the Alzheimer’s continuum. While later in the disease, the temporal composite tau-PET SUVr increased as the disease progressed (CN Aβ+ versus MCI Aβ+). These results suggested that rather than measure progression as change from lower to higher Braak stages, progression could also be measured as increasing tau-PET SUVr within this temporal composite ROI. Thus, quantification of change in tau-PET can be accomplished [10].

The absence of measurable change in temporal composite tau in Aβ- subjects, in contrast to the significant change in Aβ+ subjects (Fig. 2), suggested that amyloid deposition may be a necessary antecedent for tau accumulation [4, 9]. Mixed effect model also pointed to baseline cortical Aβ as the predictor of the rate of tau accumulation (Table 3). The amyloid-cascade hypothesis suggests that Aβ accelerates the spread of tau, disrupting function and initiating neurodegeneration in distributed brain networks, resulting in cognitive decline. This hypothetical model remains to be tested with clinical trails with anti-Aβ therapeutic agents [30]. Additionally, we did not find evidence of significant association between Aβ change and baseline tau burden. This may also support the temporal ordering in the above model.

The rate of different regional tau accumulation varied as a function of the baseline composite temporal tau burden (Table 3). This result was consistent with the hypothesis of trans-synaptic spread of tau [31, 32], in which it would be expected that the higher the baseline tau burden, the greater the number of neurons affected. This, in turn, would cause a greater number of at-risk synapses and thus, a greater degree of elevation in tau over the observation period [8]. This finding may also suggest that temporal cortex areas are the main “giver” regions for tau signal and the temporal lobe pathology progresses distantly [29].

We found baseline temporal composite tau was significantly associated with memory and executive function impairment. This was in agreement with the pathology literature, which suggested that tau tangles but not Aβ plaques correlated with cognition [33]. Tau tangles are also associated with neurodegeneration [34–36] and are likely to be part of the chain of events leading to cortical dysfunction and cognitive impairment. We also found higher levels of tau may demonstrate prospective memory decline. Our results raised the possibility that halting tau accumulation would prevent cognitive decline, and trials would benefit from tau-PET to identify participants at risk of cognitive decline [20]. Furthermore, baseline cortical Aβ was not closely associated with memory and executive function decline, and the association between tau and cognitive decline was independent of the amyloid-PET retention. This was consistent with previous work suggesting that tau-PET signal was more proximal to cognitive decline than amyloid-PET signal [20]. However, there were other studies showing both Aβ and tau were required for memory decline during preclinical stages of AD [11, 16].

Our study had some limitations. In this study, we only focused on the tau accumulation in composite regions of interest and did not capture the detailed topographic characteristics of tau change. Thus, further longitudinal tau-PET studies may test the sequential pattern of tau accumulation and its associations with amyloid and cognition. Additionally, our analyses considered data without partial volume correction. Partial volume correction can reduce contamination from regions in which off-target binding is common into neighboring regions of interest. Our inferences of these PET imaging need to be scrutinized further in data with partial volume correction.