Widespread distribution of tauopathy in preclinical Alzheimer’s disease

Abstract

The objective of this study was to examine the distribution and severity of tau-PET binding in cognitively normal adults with preclinical Alzheimer’s disease as determined by positive beta-amyloid PET. ¹⁸F-AV-1451 tau-PET data from 109 cognitively normal older adults were processed with 34 cortical and 9 subcortical FreeSurfer regions and averaged across both hemispheres. Individuals were classified as being beta-amyloid positive (N = 25, A+) or negative (N = 84, A−) based on a ¹⁸F-AV-45 beta-amyloid-PET standardized uptake value ratio of 1.22. We compared the tau-PET binding in the 2 groups using covariate-adjusted linear regressions. The A+ cohort had higher tau-PET binding within 8 regions: precuneus, amygdala, banks of the superior temporal sulcus, entorhinal cortex, fusiform gyrus, inferior parietal cortex, inferior temporal cortex, and middle temporal cortex. These findings, consistent with preclinical involvement of the medial temporal lobe and parietal lobe and association regions by tauopathy, emphasize that therapies targeting tauopathy in Alzheimer’s disease could be considered before the onset of symptoms to prevent or ameliorate cognitive decline.

1. Introduction

Alzheimer’s disease (AD), the most common cause of dementia, is characterized by extracellular beta-amyloid plaques and intraneuronal neurofibrillary tangles (NFTs); the latter contain abnormal filaments of pathologic tau protein (Braak and Braak, 1991; Braak et al., 2006). In vivo neuroimaging and cerebrospinal fluid (CSF) quantification of both beta-amyloid and tau protein can now detect pathological aggregates and abnormal fluid concentrations of these proteins as early as 2 decades before the onset of clinical symptoms (Bateman et al., 2012; Sperling et al., 2011).

Multiple clinicopathologic studies demonstrate that AD pathology is present in individuals before clinical symptoms develop; such individuals are at elevated risk for progression to AD dementia (Fagan et al., 2007; Morris et al., 2009; Vos et al., 2013). The accumulation of beta-amyloid plaques and NFTs in cognitively normal (CN) individuals has been suggested to represent preclinical AD. The 2011 National Institute of Aging (NIA) criteria, and recently updated criteria (Jack et al., 2018a), further propose that preclinical AD be divided into separate stages: stage 1 with beta-amyloidosis only (A+); stage 2 with beta-amyloidosis and indicators of neurodegenerative pathology (N+); and stage 3 with beta-amyloidosis, neurodegeneration, and subtle cognitive decline (Jack et al., 2012; Sperling et al., 2011). These 3 proposed stages of preclinical AD can be described by surrogate markers in CSF; low CSF beta-amyloid-42 is generally accepted as a marker of beta-amyloidosis, and high CSF tau is generally accepted as a marker of neurodegeneration. However, there are several limitations of CSF markers, including lack of direct information about distributions or densities of beta-amyloid or of the NFTs. In addition, increased CSF total tau levels, commonly observed in AD, are similarly observed with other dementing diseases such as frontotemporal lobar degeneration and traumatic brain injury, as well as normal aging (Franz et al., 2003; Riemenschneider et al., 2002), making it nonspecific for AD. By contrast, PET tracers thought to bind to NFTs (Lowe et al., 2016; Marquie et al., 2017) reflect both total tauopathy burden and topography, although off target binding has been noted (Lemoine et al., 2018). In addition, spatial information from tau-PET will be critical for monitoring potential tau aggregate accumulation over time and for linking the regional spread of pathologic tau to other key mechanisms involved in AD progression. Furthermore, by providing this spatial information, tau-PET may help to distinguish preclinical AD and symptomatic AD from other tauopathy-related diseases that exhibit different anatomical patterns of tau aggregates. Therefore, PET measures of tau may be more suitable than CSF measures for applying NIA-AA criteria for preclinical AD or ATN classification.

While the regional deposition of beta-amyloid plaques in vivo, as detected by beta-amyloid-PET tracers, has been well established (Klunk et al., 2004; Villemagne et al., 2011), experience with tau-PET to describe the distribution of tauopathy in vivo is more limited. Nevertheless, findings with tau-PET to date appear to be fairly consistent with the spatial distributions of tauopathy that have been as described in neuropathologic studies (Braak and Braak, 1991; Braak et al., 2006). Initial ex vivo autoradiographic studies suggest tau-PET correlates with postmortem tauopathy (Lowe et al., 2016; Marquie et al., 2015). Furthermore, in vivo studies examining individuals with symptomatic AD compared with controls have found tau-PET ligand binding in temporal as well as neocortical areas in a spatial pattern that is generally consistent with, but not identical to, advanced Braak stages V and VI (Chien et al., 2013; Cho et al., 2016; Gordon et al., 2016; Johnson et al., 2016; Schwarz et al., 2016). The spatial pattern of binding also colocalizes with changes in hypometabolism (Bischof et al., 2016; Ossenkoppele et al., 2016), and atrophy (Wang et al., 2016), observed in atypical forms of AD (Day et al., 2017; Ossenkoppele et al., 2016), and mirrors the degree of selective cognitive impairment in those cases.

While there is an established literature describing tau-PET findings in the setting of AD dementia, the investigation of regional distribution of tauopathy in CN individuals with or without preclinical AD (A+ or A−) remains relatively unexplored. Initial studies investigating tau-PET ligand binding in small cohorts of CN adults have identified several brain regions that show evidence of tauopathy in older individuals (Hanseeuw et al., 2017; Jack et al., 2018b; Jacobs et al., 2018; LaPoint et al., 2017; Lowe et al., 2018a,b; Scholl et al., 2016; Vemuri et al., 2017) and report that increased tau-PET binding inversely correlates with beta-amyloid-42 levels in the cerebrospinal fluid (Chhatwal et al., 2016; Gordon et al., 2016). This early work suggests tau-PET binding is elevated in preclinical AD, but further work is needed to define the topography of tau-PET binding early in the disease course. The characterization of regional tau accumulation in CN adults in vivo using tau-PET could be important for identifying CN individuals most at risk of cognitive decline and might encourage the evaluation of tau-related pharmacological interventions early in the disease course.

2. Materials and methods

2.1. Participants

Data from 109 participants from studies at the Knight Alzheimer’s Disease Research Center, Washington University in St Louis (including the Adult Children Study and the Healthy Aging and Senile Dementia Study) were used. Inclusion criteria included cognitive normality [Clinical Dementia Rating score equals 0 (Morris, 1997)] and completion of both beta-amyloid and tau-PET scans. The Washington University in St. Louis Institutional Review Board approved all procedures and each participant was provided signed informed consent for the study.

2.2. MRI

Data were acquired on a Siemens Biograph mMR (n = 83) or Trio 3T scanner (n = 26). T₁-weighted images were acquired using a magnetization-prepared rapid acquisition gradient echo (MPRAGE) sequence with the following: repetition time = 2300 ms, echo time = 2.95 ms, flip angle = 9°, 176 slices, in plane resolution 240 × 256, slice thickness = 1.2 mm acquired in sagittal orientation. Images underwent volumetric segmentation using FreeSurfer 5.3 [mailto:http://freesurfer.net (Fischl et al., 2004)] to identify regions of interest (ROIs) used in the PET analyses.

2.3. PET imaging

2.3.1. Beta-amyloid-PET

Participants underwent beta-amyloid-PET imaging with ¹⁸F-AV-45 (florbetapir). Participants received a single intravenous bolus of 370 MBq (10 mCi) of florbetapir infused over 60 seconds. Scans were acquired on a Siemens Biograph mMR PET/MR scanner and attenuation corrected with a corresponding CT. Data were processed using an ROI approach using FreeSurfer software. As described previously (Gordon et al., 2016), data between the 50- to 70-minute postinjection window were examined. In each ROI, data were converted to standardized uptake value ratios (SUVRs) using the cerebellar gray as the reference region and partial volume corrected using a regional spread function approach (Rousset et al., 1998; Su et al., 2015, 2016).

2.3.2. Tau-PET

Tau-PET imaging was performed within 13 months (mean 44.0 days, range: 1–373 days) of the beta-amyloid-PET imaging session using ¹⁸F-AV-1451 (flortaucipir). Participants received a single 6.5–10 mCi intravenous bolus of flortaucipir infused over 20 seconds. Data were processed using an ROI approach using FreeSurfer and, as done in prior work using ¹⁸F-AV-1451 (Brier et al., 2016; Chien et al., 2013; Day et al., 2017; Gordon et al., 2016; Johnson et al., 2016; Wang et al., 2016), and data from the 80- to 100-minute postinjection window were examined. Scans were acquired on a Siemens Biograph 40 PET/CT scanner. As described previously, in each ROI, data were converted to SUVRs using the cerebellar gray as the reference region and partial volume corrected using a regional spread function approach (Rousset et al., 1998; Su et al., 2015, 2016). This partial volume correction method, including nonbrain region sampling, may additionally help minimize off-target binding. Tau-PET SUVRs for each cortical and subcortical ROI were extracted and averaged together from the left and right hemispheres to form 1 bilateral measure. The average amount of time between clinical assessment and tau-PET imaging session was 103 ± 57 days, and the average time between beta-amyloid-PET and tau-PET imaging sessions was 82 ± 84 days.

2.4. Amyloid positivity classification

As previously described, a composite beta-amyloid deposition measure was created using the average across the left and right lateral orbitofrontal, medial orbitofrontal, rostral middle frontal, superior frontal, superior temporal, middle temporal, and precuneus regions (Su et al., 2013, 2016). To identify individuals who were beta-amyloid positive, we split our sample based on a partial volume corrected florbetapir SUVR cutoff of 1.22 (Mishra et al., 2017). To generate a cutoff value for ¹⁸F-AV-45, a previously established cutoff using ¹¹C-Pittsburgh compound B [(Gordon et al., 2015; Vlassenko et al., 2011)] was translated using a linear regression from a separate cohort of 100 individuals who had both ¹⁸F-AV-45 and ¹¹C-Pittsburgh compound B imaging as part of a crossover study. Based on this cutoff, our current sample included 25 beta-amyloid-positive (A+) individuals and 84 beta-amyloid-negative (A−) participants.

2.5. Neuropsychological assessment

Each cohort at the Knight ADRC receives slightly different cognitive batteries, and in the interest of maximizing the available sample size, only tests that were common across all cohorts were considered for the present analyses. This resulted in a sample size of 108 individuals. Tests included a measure of episodic memory: the free recall score from the Free and Cued Selective Reminding Test (Grober et al., 1988); a measure of working memory: Letter Number Sequencing (Wechsler, 1997); a measure of semantic memory retrieval: category fluency for Animal Naming (Goodglass and Kaplan, 1983); a test of processing speed: Trail Making part A and a test of executive function: Trail Making part B (Armitage, 1946). Tests were standardized using the sample mean and standard deviation of the cognitive assessment that was nearest to the tau-PET scan and then averaged to form a cognitive composite score (Aschenbrenner et al., 2018).

2.6. Statistics

To test for group differences between A+ and A− cohorts on demographics, t-tests were performed for age, years of education, and Mini–Mental State Examination score; chi-squared tests were performed for sex and apolipoprotein E ε4 (APOE4) genotype (Table 1).

Table 1.

Participant characteristics

Characteristic	A−(N= 84)	A+ (N= 25)	p-value
Age, y	66.8 (8.6) [46–91]	71.9 (7.4) [58–90]	0.006
Female, % (n)	50.0 (42)	60.0 (15)	0.495
APOE4 positive, % (n)	23.2 (19)	56.0 (14)	0.003
Education, y	16.1 (2.2) [12–20]	16.6 (1.6) [12–18]	0.425
GDS	1.16 (1.8) [0–10]	0.68 (1.1) [0–3]	0.104
MMSE	29.3 (1.1) [25–30]	29.5 (1.0) [27–30]	0.228
Clinical assessment and tau-PET time interval, d	105.0 (54.8) [0–240]	94.9 (65.7) [0–328]	0.548
Beta-amyloid-PET and tau-PET time interval, d	76.4 (80.9) [1–353]	102.8 (94.4) [1–373]	0.212

Values are mean (SD) [range] unless otherwise indicated.

Key: GDS, Geriatric Depression Scale, APOE4, apolipoprotein E ε4; MMSE, Mini–Mental State Examination, A+, beta-amyloid-PET positive; A−, beta-amyloid-PET negative.

To visualize the collective anatomic distribution of tau-PET SUVR in each of the groups, we created representative group mean images. For each participant, their nonpartial volume corrected SUVR images were aligned to their individual MPRAGE using a rigid body transformation, subsequently transformed to MNI atlas space using a nonlinear warp, and resampled into a 2-mm isotropic resolution. Voxels across all participants in either the A+ or A− groups were then averaged together (Fig. 1A and B). Higher SUVRs indicate higher binding of the tracer relative to the cerebellar reference region, which reflects more tau pathology or greater nonspecific binding, in one group compared to the other.

Fig. 1.

Mean tau distribution in A+ and A− cohorts. Mean tau deposition represented as SUVRs for (A) A+ and (B) A− participants. Higher SUVRs indicate higher mean binding and more tau pathology. Abbreviations: SUVR, standardized uptake value ratio; A+, beta-amyloid-PET positive; A−, beta-amyloid-PET negative.

Next, to statistically compare regional tau-PET SUVRs in the A+ and A− groups, we ran linear regressions for each of the 34 cortical and 9 subcortical regions examined (see Table 2 for a list of regions), including age and sex as covariates in the model. Covariates were selected based on their established association with AD and brain measures. To correct for multiple comparisons, we implemented a Benjamini-Hochberg procedure with a false discovery rate of 5%.

Table 2.

Regional tau-PET SUVR in A+ and A− cohorts

Region	A− mean SUVR (SD)	A+ mean SUVR (SD)	Unstandardized B (SE)	95% CI	Partial eta squared	p- value
Accumbens	1.35 (0.30)	1.40 (0.26)	−0.004 (0.07)	−0.138 to 0.130	0.000	0.958
Amygdala	1.08 (0.17)	1.24 (0.28)	0.136 (0.05)	0.044 to 0.227	0.076	0.004
Caudate	1.33 (0.25)	1.44 (0.26)	0.044 (0.06)	−0.069 to 0.157	0.006	0.441
Choroid plexus	1.82 (0.65)	1.94 (0.93)	0.063 (0.17)	−0.274 to 0.400	0.001	0.712
Banks of superior temporal sulcus	1.23 (0.13)	1.36 (0.19)	0.094 (0.03)	0.027 to 0.160	0.069	0.006
Caudal anterior cingulate	1.11 (0.17)	1.10 (0.20)	−0.006 (0.04)	−0.089 to 0.076	0.000	0.879
Caudal middle frontal	1.00 (0.15)	1.08 (0.23)	0.082 (0.04)	0.003 to 0.162	0.039	0.042
Cuneus	1.22 (0.15)	1.21 (0.16)	−0.009 (0.04)	−0.080 to 0.061	0.001	0.792
Entorhinal cortex	1.03 (0.22)	1.21 (0.31)	0.168 (0.06)	0.055 to 0.281	0.077	0.004
Frontal pole	0.91 (0.34)	0.85 (0.26)	0.006 (0.06)	−0.120 to 0.133	0.000	0.921
Fusiform	1.22 (0.12)	1.33 (0.12)	0.095 (0.03)	0.04 to 0.150	0.100	0.001
Inferior parietal	1.23 (0.15)	1.35 (0.20)	0.100 (0.04)	0.030 to 0.170	0.070	0.006
Inferior temporal	1.26 (0.15)	1.39 (0.18)	0.111 (0.04)	0.040 to 0.181	0.085	0.002
Insula	0.98 (0.13)	0.98 (0.14)	0.005 (0.03)	−0.056 to 0.066	0.000	0.878
Isthmus cingulate cortex	1.20 (0.16)	1.32 (0.23)	0.090 (0.04)	0.010 to 0.170	0.045	0.028
Lateral occipital cortex	1.24 (0.23)	1.30 (0.22)	0.038 (0.05)	−0.062 to 0.139	0.005	0.450
Lateral orbital frontal cortex	1.18 (0.15)	1.25 (0.15)	0.053 (0.04)	−0.016 to 0.122	0.021	0.134
Lingual cortex	1.16 (0.14)	1.18 (0.14)	0.008 (0.03)	−0.056 to 0.072	0.001	0.808
Medial orbital frontal cortex	1.02 (0.16)	1.06 (0.17)	0.034 (0.04)	−0.043 to 0.111	0.007	0.382
Middle temporal cortex	1.21 (0.14)	1.31 (0.14)	0.098 (0.03)	0.039 to 0.158	0.092	0.001
Paracentral cortex	1.06 (0.18)	1.06 (0.19)	−0.005 (0.04)	−0.090 to 0.079	0.000	0.900
Parahippocampal gyrus	1.01 (0.15)	1.09 (0.17)	0.086 (0.04)	0.014 to 0.159	0.051	0.019
Pars opercularis	1.03 (0.15)	1.05 (0.19)	0.019 (0.04)	−0.055 to 0.092	0.002	0.617
Pars orbitalis	1.21 (0.23)	1.19 (0.19)	−0.004 (0.05)	−0.096 to 0.089	0.000	0.928
Pars triangularis	1.14 (0.18)	1.14 (0.22)	0.014 (0.04)	−0.071 to 0.099	0.001	0.747
Peri calcarine cortex	1.19 (0.14)	1.21 (0.16)	0.001 (0.03)	−0.066 to 0.067	0.000	0.985
Post central	0.98 (0.13)	0.98 (0.15)	0.003 (0.03)	−0.058 to 0.064	0.000	0.927
Posterior cingulate cortex	1.23 (0.14)	1.31 (0.19)	0.075 (0.04)	0.004 to 0.146	0.040	0.038
Precentral cortex	0.97 (0.12)	0.98 (0.17)	0.019 (0.03)	−0.044 to 0.082	0.003	0.556
Precuneus	1.20 (0.15)	1.32 (0.21)	0.110 (0.04)	0.034 to 0.187	0.072	0.005*
Rostral anterior cingulate cortex	1.06 (0.17)	1.05 (0.17)	−0.011 (0.04)	−0.090 to 0.068	0.001	0.784
Rostral middle frontal cortex	0.94 (0.18)	0.95 (0.17)	−0.029 (0.04)	−0.042 to 0.100	0.006	0.419
Superior frontal cortex	0.88 (0.15)	0.91 (0.18)	0.024 (0.04)	−0.049 to 0.098	0.004	0.428
Superior parietal cortex	1.05 (0.17)	1.14 (0.24)	0.072 (0.04)	−0.015 to 0.160	0.025	0.105
Superior temporal cortex	1.01 (0.12)	1.03 (0.13)	0.033 (0.03)	−0.021 to 0.087	0.014	0.225
Supramarginal gyrus	1.16 (0.13)	1.22 (0.15)	0.053 (0.03)	0.006 to 0.112	0.029	0.079
Temporal pole	1.00 (0.17)	1.06 (0.22)	0.077 (0.04)	0.002 to 0.157	0.034	0.057
Transverse temporal cortex	0.98 (0.16)	1.01 (0.23)	0.036 (0.04)	−0.048 to 0.120	0.007	0.395
Hippocampus	1.21 (0.19)	1.24 (0.15)	0.005 (0.04)	−0.078 to 0.088	0.000	0.905
Pallidum	1.91 (0.34)	1.98 (0.32)	−0.039 (0.07)	−0.177 to 0.100	0.003	0.643
Putamen	1.52 (0.24)	1.62 (0.23)	0.015 (0.05)	−0.079 to 0.109	0.001	0.751
Thalamus	1.25 (0.15)	1.28 (0.13)	0.020 (0.03)	−0.047 to 0.087	0.003	0.557
Ventral diencephalon	1.49 (0.21)	1.46 (0.18)	−0.041 (0.05)	−0.137 to 0.055	0.007	0.401

Key: SUVR, standardized uptake value ratio; SE, standard error; CI, confidence interval, A+, beta-amyloid-PET positive; A−, beta-amyloid-PET negative.

Next, because APOE4 status was significantly different between the A− and A+ groups, we repeated these multivariable linear regressions, comparing regional tau-PET SUVRs in the A+ and A− groups, for each cortical and subcortical region found to be significant in the primary analyses, but added APOE4 status as a covariate, along with age and sex. To correct for multiple comparisons, we implemented a Benjamini-Hochberg procedure with a false discovery rate of 5%.

Since the A+ classification was derived from a composite beta-amyloid-PET SUVR, we wanted to explore further the relationship between regional tau-PET and this composite beta-amyloid-PET SUVR value. We ran a multivariate model for each of the 8 regions identified as having significantly higher tau-PET SUVRs in the A+ cohort compared to controls, controlled for age and sex. We additionally report the within-group Pearson’s correlation of regional tau-PET and beta-amyloid-PET for these 8 regions. To correct for multiple comparisons, we implemented a Benjamini-Hochberg procedure with a false discovery rate of 5%.

Finally, to determine whether the regional tau deposition is related to cognition, we ran linear regression models between ¹⁸F-AV-1451 and the global cognition score, adjusting for age and sex. To correct for multiple comparisons, we implemented a Benjamini-Hochberg procedure with a false discovery rate of 5%.

3. Results

3.1. Participant characteristics

As shown in Table 1, participants in the A+ cohort were older (mean age = 71.9 years) compared to A− individuals (mean age = 66.8 years, F_df = 1.259_1,107, p = 0.008) and had a higher percentage of APOE4carriers (56.0%) than the A− group (23.2%, X² = 9.681, p = 0.002).

3.2. Tau distribution in A+ cohort

Linear regressions evaluating group differences between the A+ and A− groups show that the A+ cohort had significantly higher tau-PET SUVR in 9 regions (Table 2), including within the amygdala (B = 0.136, p = 0.004), banks of the superior temporal sulcus (B = 0.094, p = 0.006), entorhinal cortex (B = 0.168, p = 0.004), fusiform gyrus (B = 0.095, p = 0.001), inferior parietal cortex (B = 0.100, p = 0.006), inferior temporal cortex (B = 0.111, p = 0.002), parahippocampal gyrus (B = 0.086, p = 0.019), middle temporal cortex (B = 0.098, p = 0.001), precuneus (B = 0.110, p = 0.005). Effect size maps of regression coefficients (B values) from significant linear regressions are presented in Figure 2. Violin plots show the distribution of AV1451 in 9 significant regions of interest (Supplementary Material (SM1)). Furthermore, ¹⁸F-AV-1451 signal in these regions is highly correlated across all regions (Supplementary Material (SM2)).

Fig. 2.

Difference in tauopathy in A+ cohort compared to A−. Effect size maps, depicting regression coefficients (B values), from significant linear regressions, adjusting for age and sex, comparing A− and A+ participants. Abbreviations: AD, Alzheimer’s disease, A+, beta-amyloid-PET positive; A−, beta-amyloid-PET negative.

When APOE4 status was included as a covariate, regional tau-PET SUVRs for the A+ cohort remained significantly higher compared to the A− group in 8 of the 9 regions including the amygdala (B = 0.141, p = 0.005), banks of the superior temporal sulcus (B = 0.091, p = 0.013), entorhinal cortex (B = 0.152, p = 0.014), fusiform gyrus (B = 0.084, p = 0.005), inferior parietal cortex (B = 0.101, p = 0.009), inferior temporal cortex (B = 0.103, p = 0.008), middle temporal cortex (B = 0.098, p = 0.003), and precuneus (B = 0.109, p = 0.010), with the exception being the parahippocampal gyrus (p = 0.110). Furthermore, in these models, APOE4 status is not significantly associated with tau-PET SUVRs in any regions examined. Full models are presented in Supplementary Material (SM3).

3.3. Relationship between beta-amyloid- and tau-PET

There were significant associations between regional tau-PET SUVR and composite beta-amyloid SUVR levels in all 8 regions examined (Fig. 3), including the amygdala (B [standard error {SE} = 0.153 [0.04], p = <0.001, η_p² = 0.121), banks of the superior temporal sulcus (B[SE] = 0.090[0.03], p = 0.003, η_p² = 0.080), entorhinal cortex (B[SE]= 0.168[0.05], p = 0.001, η_p² = 0.098), fusiform gyrus (B[SE] = 0.087[0.03], p = 0.001, η_p² = 0.106), inferior parietal (B[SE] = 0.102[0.03], p = 0.002, η_p² = 0.092), inferior temporal (B[SE] = 0.106[0.03], p = 0.001, η_p² = 0.098), middle temporal (B[SE] = 0.093[0.03], p = 0.001, η_p² = 0.104), and precuneus (B[SE] = 0.135[0.03], p = <0.001, η_p² = 0.136). In addition, there were within-group correlations between tau-PET and beta-amyloid-PET in A+ individuals in the inferior parietal (r = 0.498, p = 0.011), inferior temporal (r = 0.453, p = 0.023), and middle temporal (r = 0.531, p = 0.006). There was no correlation in A− individuals between tau-PET and beta-amyloid-PET in any region examined.

Fig. 3.

Association between regional tau-PET and beta-amyloid-PET. Relationship between composite beta-amyloid-PET measure and regional tau-PET SUVR from (A) amygdala, (B) banks of superior temporal sulcus, (C) entorhinal cortex, (D) fusiform, (E) inferior parietal, (F) inferior temporal, (G) middle temporal, and (H) precuneus. Abbreviations: Red, beta-amyloid-PET positive; A+ and blue, beta-amyloid-PET negative; A− PET, positron emission tomography; SUVR, standardized uptake value ratio.

3.4. Relationship between cognition and tau-PET

There was no relationship between tau-PET and the cognitive composite score in any of the 8 regions examined (p >0.101, Table 3).

Table 3.

Associations between regional tau-PET SUVR and cognition

Region	Unstandardized B (SE)	p- value	95% CI	Partial eta squared
Amygdala	0.049 (0.03)	0.101	−0.010 to 0.108	0.026
Banks of superior temporal sulcus	0.035 (0.02)	0.102	−0.007 to 0.078	0.026
Entorhinal	0.010 (0.04)	0.789	−0.064 to 0.083	0.001
Fusiform	−0.012 (0.02)	0.529	−0.048 to 0.025	0.004
Inferior parietal	−0.007 (0.02)	0.755	−0.053 to 0.039	0.001
Inferior temporal	−0.008 (0.02)	0.735	−0.054 to 0.038	0.001
Middle temporal	−0.001 (0.02)	0.943	−0.041 to 0.038	0.000
Precuneus	0.024 (0.03)	0.327	−0.025 to 0.073	0.009

Key: SUVR, standardized uptake value ratio; SE, standard error; CI, confidence interval.

4. Discussion

Clinicopathologic studies of AD propose a stereotypical spread of tauopathy based on postmortem pathological studies. The Braak and Braak staging scheme envisions spread of NFTs from the brainstem and transentorhinal cortex to entorhinal cortex, then into neocortical regions, including the fusiform gyrus, medial temporal gyrus, and insular cortex, and, later, into frontal, parietal, and occipital cortices (Braak and Braak, 1991; Braak et al., 2006). Recent studies (Lowe et al., 2016; Marquie et al., 2017) comparing ¹⁸F-AV-1451 autoradiography with tau immunohistochemistry have found high colocalization of ¹⁸F-AV-1451 with tauopathy in AD compared to other non-AD tauopathies, particularly in the brains with advanced Braak NFT stages, supporting the use of this tracer to map AD-associated tauopathy.

Other studies have investigated the spatial pattern of ¹⁸F-AV-1451 binding tauopathy in vivo, using ¹⁸F-AV-1451 and PET, comparing CN individuals to older adults with symptomatic AD. Johnson and colleagues (2016) found that participants with mild cognitive impairment and AD had significantly higher tau-PET SUVRs in regions including the inferior temporal lobe, fusiform gyrus, posterior cingulate cortex, occipital cortex, parahippocampal gyrus, and entorhinal cortex compared with CN peers. Similarly, our group (Brier et al., 2016; Gordon et al., 2016) and others (Cho et al., 2016; Johnson et al., 2016) have shown patterns of increased tau-PET SUVRs in cognitively impaired individuals compared to controls in temporal and occipital neocortical areas. This homogeneity across multiple centers suggests a consistent spatial pattern of tau-PET binding in symptomatic AD cohorts relative to cognitively normal controls. This spatial pattern of tau-PET binding in symptomatic AD has been stereotyped into an estimated tau-PET Braak staging scheme (Maass et al., 2017; Scholl et al., 2016; Schwarz et al., 2016) in which the regional tau burden in symptomatic AD participants appears consistent with higher Braak stages (V and VI).

This study adds to converging literature examining tau-PET in A+ and A− CN individuals (Gordon et al., 2016; Hanseeuw et al., 2017; Jack et al., 2018b; Jacobs et al., 2018; Lowe et al., 2018a,b; Mishra et al., 2017; Scholl et al., 2016; Sepulcre et al., 2016; Vemuri et al., 2017; Villemagne et al., 2017). For example, Schöll and colleagues (2016) examined ¹⁸F-AV-1451 tau-PET, mapped onto the Braak NFT staging scheme, being distributed across all Braak stages, in a cohort of 5 CN young adults, 33 CN older adults, and 15 symptomatic AD patients. Among them, CN older adults ranged across stages 0, I/II, and III/IV. Other studies also described regional associations with other core AD biomarkers including beta-amyloid-PET (Brier et al., 2016; Lockhart et al., 2017; Sepulcre et al., 2016), CSF tau levels (Chhatwal et al., 2016; Gordon et al., 2016; Wang et al., 2016), gray matter volumes and cortical thinning (LaPoint et al., 2017; Sepulcre et al., 2016), and functional connectivity MRI in CN individuals (Schultz et al., 2017). Specifically, a study investigating difference in inherent tau-PET signal across the brain in younger adults proposed implementation of region-specific z-score values to assess severity of NFT burden (Vemuri et al., 2017). Results focus on the entorhinal cortex as an AD-specific tau-PET signature. Similarly, we find an increase in tau-PET signal in the entorhinal cortex in A+ individuals compared to A−. However, we importantly show that there are a number of other regions also significantly elevated in this A+ cohort, which extend outside the medial temporal lobe. Recent studies investigating the cross-sectional and longitudinal tau-PET signal in CN and cognitively impaired individuals additionally support our results. Using a meta-region of interest approach to assess longitudinal change in tau, results were in alignment with the current findings, suggesting increase in the rates of tau accumulation in regions other than the entorhinal cortex, including midtemporal, retrosplenial, and posterior cingulate (Jack et al., 2018b). However, the regions characterized as an ”early-AD” meta-region for longitudinal analyses included the fusiform and posterior cingulate gyrus, therefore not completely consistent with our results.

The present study extends the work of those studies in several critical ways. First, it focuses on the generation of a topographical map of ¹⁸F-AV-1451 signal in A+ CN individuals. Such work aids in our interpretation of tau accumulation in preclinical AD. Second, with its restriction specifically on a large A+ cohort, characterized clinically by CDR and for AD pathology by beta-amyloid-PET, this study provides strong evidence that regions of tauopathy detected by tau-PET in A+ are comparable to those identified in recent studies of AD dementia and are comparable, but not identical to, Braak NFT staging. It is possible that the discrepancy in our results compared to Braak NFT staging is due to differences in specificity of ¹⁸F-AV-1451 binding of certain tau isoforms compared to classical histochemical staining or signal contamination in ¹⁸F-AV-1451 from nonspecific binding in nearby regions. Alternatively, this discordance could be related to subtle difference in the cohorts studied and comparisons examined. While Braak NFT staging is primarily a classification of presence or absence of regional NFT, our analyses more specifically examined not only the distribution of NFT in A+ CN individuals but also the density of NFTs in these regions, as compared to an A−cohort. We therefore provide evidence for significant regional differences in ¹⁸F-AV-1451 signal in A+ CN individuals, compared to A− individuals.

We additionally report that APOE e4 may not be associated with tau-PET signal above and beyond that of A− and A+ classification. This is in contrast to recent studies in animal models suggesting a role of APOE e4 exacerbating NFT accumulation (Shi et al., 2017). It is possible, however, that the present study may be underpowered to detect a modulating effect of APOE e4 status within A− and A+ groups. Larger studies with a sufficient sample of APOE e4 A− individuals are needed to determine whether there are interactions between APOE e4 status and beta-amyloid load on regional tau-PET signal.

Furthermore, we report no associations between regional ¹⁸F-AV-1451 and global cognition in our cohort. These results support findings by Schöll and colleagues (2016) who similarly reported a lack of association between cross-sectional global cognition and ¹⁸F-AV-1451.

Overall, results from prior studies, in addition to the present study, provide evidence that there are associations with tau-PET and established biomarkers of AD in preclinical stage, further strengthening the utility of tau-PET in AD research and clinical trials.

Overall, by emphasizing the presence of widespread tauopathy early in the disease course, these findings should inform treatment strategies for preclinical AD. Having learned from previous clinical trials that unsuccessfully targeted mild-to-moderate AD dementia (Wang et al., 2017), ongoing clinical trials are now administering anti-beta-amyloid therapies in asymptomatic individuals who are determined to be positive by beta-amyloid-PET. The results of those trials will provide insight into the potential benefit of targeting beta-amyloid pathology early in the disease course. However, the results of this present study demonstrate that many clinical trial participants who are beta-amyloid-PET positive are also likely to have a significant tauopathy burden. Given that both tauopathy and beta-amyloid deposits are present in presymptomatic stages, it may be worthwhile to consider anti-tau therapies at this early stage—either alone, or in combination with anti-beta-amyloid therapy. Indeed, combined therapy might be more effective than either single therapy alone.

A particular strength of the present study is its large, well-characterized sample compared with prior studies (Johnson et al., 2016; Ossenkoppele et al., 2016). However, its sample size remains relatively modest for evaluating effects at a whole-brain level, underscoring the future need for a increasingly larger sample.

There remain several limitations of the present study. First, the current interpretation of NFT pathophysiology through utilization of ¹⁸F-AV-1451-PET in CN individuals is a potential limitation. It has been reported that ¹⁸F-AV-1451 may bind nonspecifically to neuromelanin, MAO, and iron deposits, in regions including putamen and thalamus. Furthermore, most investigations on off-target binding of ¹⁸F-AV-1451 have been conducted primarily in individuals with impairment or dementia, and little is known about the contribution of off-target binding in studies of CN individuals. However, the largest factor so far tied to this nonspecific binding has been age; thus, we included age as a covariate. Even so, there is evidence, as seen in Figure 1, of off-target binding, including parts of the basal ganglia and brainstem. Strong correlations of ¹⁸F-AV-1451 between brain regions suggest that associations of primary interest are not due to off-target binding of the basal ganglia and brainstem, yet we acknowledge there may be minimal bleeding effect of off-target binding on some regions examined, but largely controlled by partial volume correction methods including nonbrain region sampling.

Another potential limitation of our study is that the tau-PET imaging sessions were more likely to occur after the collection of beta-amyloid-PET data and clinical examination; however, we limited this interval to 12 months, so any overestimation of tauopathy in relation to beta-amyloid and clinical data should be very modest. In addition, similar to many other studies in preclinical AD and AD dementia, our A+ group was older than A− group. We accounted for this difference by including age as a covariate in the model; however, future studies would be improved by age-matching A− and A+ groups.

Finally, because we classified our participants categorically as A+ or A− based on beta-amyloid status, it is possible that some of our beta-amyloid-negative participants are actually subtly beta-amyloid-positive, but below threshold (Palmqvist et al., 2016; Vlassenko et al., 2016). Furthermore, as depicted in Figure 3, there are a few participants who may have elevated ¹⁸F-AV-1451 levels in contrast to their low beta-amyloid-PET levels, representing a potential primary age-related tauopathy subset in our A− group or generally noisy binding properties of the tracer in the absence of AD pathology. In addition, it is possible that accumulation of tau pathology might be happening in parallel, and A− individuals who present with high AV1451 might convert to A+ in the future. Longitudinal beta-amyloid- and tau-PET studies would support more confident identification of participants with preclinical AD.

In summary, our findings contribute to the understanding of tau pathology and illustrate in vivo that tauopathy is widespread in preclinical AD, encompassing both the temporal and parietal lobes. This and future studies of pathologic tau-PET in preclinical AD will be useful in designing clinical trials for AD dementia, especially when tauopathy-related therapies are administered. These results suggest that it may be worthwhile to consider antitauopathy therapies early in the disease course to prevent cognitive decline due to Alzheimer’s disease.