A, Criteria for defining A and T groups. B, Horizontal bar graph shows number of participants with extreme cortical residual values or asymmetry indices. Bottom right circles show intersections of various cortical combinations; circles are color coded to show subtype. Vertical bar graph shows number of participants with each intersection combination. C, Mean flortaucipir distribution for each A and T group. A– indicates normal β-amyloid; A+, elevated β-amyloid; A–T MTL–, normal β-amyloid without elevated MTL tau; A–T MTL+, normal β-amyloid and elevated MTL tau; A+T MTL–, preclinical AD without elevated MTL tau; A+T MTL+, preclinical AD and elevated MTL tau; MTL, medial temporal lobe; and SUVR, standardized uptake value ratio.
The circles are individual data points. The boxes and vertical lines are boxplots. A+T cortical+ indicates preclinical Alzheimer disease and divergent cortical tau; A+T MTL+, preclinical Alzheimer disease and elevated MTL tau; A–T MTL+, normal β-amyloid and elevated MTL tau; A–T MTL–, normal β-amyloid without elevated MTL tau; A–T MTL+, normal β-amyloid and elevated MTL tau; MTL, medial temporal lobe; S, subtype.
aP < .001.
bP < .01.
The circles are individual data points. The boxes and vertical lines are boxplots. A+T cortical+ indicates preclinical Alzheimer disease and divergent cortical tau; A+T MTL+, preclinical Alzheimer disease and elevated MTL tau; A–T MTL+, normal β-amyloid and elevated MTL tau; A–T MTL–, normal β-amyloid without elevated MTL tau; FCSRT, Free and Cued Selective Reminding Test; LM, logical memory; MTL, medial temporal lobe; S, subtype; and SUVR, standardized uptake value ratio.
bP < .05.
cP < .01.
A, Association between age and continuous global amyloid, measured in standardized uptake value ratios (SUVRs). B, Association between age and regional tau using SUVRs. C, Association between age and regional tau using SUVR controlling for global amyloid burden. All plots depict individuals with preclinical Alzheimer disease and medial temporal lobe (MTL) tau (A+T MTL+) and individuals with preclinical Alzheimer disease and divergent cortical tau (A+T cortical+). The orange lines correspond to the A+T cortical+ data points, and the black lines correspond to the A+T MTL+ data points. S indicates subtype.
eMethods 1. A4 Amyloid and Tau PET Processing
eMethods 2. Reproducibility of Findings
eMethods 3. Linear Regressions Across A+ Subsets
eFigure 1. Association Between MTL SUVR and Cortical SUVR
eFigure 2. Off-Target Flortaucipir (FTP) Binding Outside of the Brain in One A+TCortical+ Subject That Was Excluded From Subsequent Analyses
eFigure 3. Prevalence of A+TCortical+ Subtypes by Regional Criterion (Cortical Versus MTL and Cortical Asymmetry) in A4, ADNI, HABS, and Wisconsin Datasets
eFigure 4. Mean Tau PET Distributions for Each A+TCortical+ Subtype From A4 (Average Maps) As Well As Individual Subject Images From ADNI, HABS, and Wisconsin Datasets
eFigure 5. Conjunction Voxelwise Maps Showing Number of Subjects Within Each A+TCortical+ Subtype With Elevated FTP Signal
eFigure 6. Further Clustering of the A+TCortical+ Asymmetrical Right Subtype Showing Those With and Without High Cortical Tau
eFigure 7. Association Between Regional FTP Tau and AV45 Amyloid SUVRs Among (A) A+TCortical+ and (B) A+TMTL+ and A+TCortical+
eFigure 8. (A) Association Between MTL FTP SUVR and Hippocampal Volume, Adjusted for Intracranial Volume. (B) Association Between FTP SUVR and Medial Temporal Lobe (MTL)/Cortical Thickness in the Same Region Among A+TMTL+ and A+TCortical+ Subtypes
eFigure 9. Sample Tau PET (Flortaucipir) Scans from Subjects With Qualitatively Noisier Data in Comparison to Unsmoothed Sample Scans
eTable 1. Demographic Information for A4, ADNI, HABS, and Wisconsin Datasets
eTable 2. Mean (SD) Tau (Flortaucipir) SUVR Values for Each Region of Interest for Each A/T Group
eTable 3. Relation Between Tau (Flortaucipir [FTP]) SUVR and Hippocampal Volume As Well As Medial Temporal Lobe (MTL)/Cortical Thickness Among A+TMTL+ and A+TCortical+
eTable 4. Cognitive Differences Between A/T Groups
Customize your JAMA Network experience by selecting one or more topics from the list below.
Young CB, Winer JR, Younes K, et al. Divergent Cortical Tau Positron Emission Tomography Patterns Among Patients With Preclinical Alzheimer Disease. JAMA Neurol. 2022;79(6):592–603. doi:10.1001/jamaneurol.2022.0676
Do cortical tau positron emission tomography (PET) patterns vary across clinically unimpaired individuals with abnormal β-amyloid (ie, preclinical Alzheimer disease [AD])?
This cross-sectional study found that divergent patterns of cortical tau uptake are present in approximately 10% of individuals with preclinical AD. Three main subtypes were identified: asymmetrical left, precuneus predominant, and asymmetrical right; individuals demonstrating these cortical tau PET patterns were younger and showed worse executive function than individuals with preclinical AD and typical medial temporal lobe uptake.
This study suggests that heterogeneous cortical tau PET patterns exist among patients with preclinical AD and are more common than currently appreciated.
Characterization of early tau deposition in individuals with preclinical Alzheimer disease (AD) is critical for prevention trials that aim to select individuals at risk for AD and halt the progression of disease.
To evaluate the prevalence of cortical tau positron emission tomography (PET) heterogeneity in a large cohort of clinically unimpaired older adults with elevated β-amyloid (A+).
Design, Setting, and Participants
This cross-sectional study examined prerandomized tau PET, amyloid PET, structural magnetic resonance imaging, demographic, and cognitive data from the Anti-Amyloid Treatment in Asymptomatic AD (A4) Study from April 2014 to December 2017. Follow-up analyses used observational tau PET data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), the Harvard Aging Brain Study (HABS), and the Wisconsin Registry for Alzheimer’s Prevention and the Wisconsin Alzheimer’s Disease Research Center (together hereinafter referred to as Wisconsin) to evaluate consistency. Participants were clinically unimpaired at the study visit closest to the tau PET scan and had available amyloid and tau PET data (A4 Study, n = 447; ADNI, n = 433; HABS, n = 190; and Wisconsin, n = 328). No participants who met eligibility criteria were excluded. Data were analyzed from May 11, 2021, to January 25, 2022.
Main Outcomes and Measures
Individuals with preclinical AD with heterogeneous cortical tau PET patterns (A+T cortical+) were identified by examining asymmetrical cortical tau signal and disproportionate cortical tau signal relative to medial temporal lobe (MTL) tau. Voxelwise tau patterns, amyloid, neurodegeneration, cognition, and demographic characteristics were examined.
The 447 A4 participants (A+ group, 392; and normal β-amyloid group, 55), with a mean (SD) age of 71.8 (4.8) years, included 239 women (54%). A total of 36 individuals in the A+ group (9% of the A+ group) exhibited heterogeneous cortical tau patterns and were further categorized into 3 subtypes: asymmetrical left, precuneus dominant, and asymmetrical right. A total of 116 individuals in the A+ group (30% of the A+ group) showed elevated MTL tau (A+T MTL+). Individuals in the A+T cortical+ group were younger than those in the A+T MTL+ group (t61.867 = –2.597; P = .03). Across the A+T cortical+ and A+T MTL+ groups, increased regional tau was associated with reduced hippocampal volume and MTL thickness but not with cortical thickness. Memory scores were comparable between the A+T cortical+ and A+T MTL+ groups, whereas executive functioning scores were lower for the A+T cortical+ group than for the A+T MTL+ group. The prevalence of the A+T cortical+ group and tau patterns within the A+T cortical+ group were consistent in ADNI, HABS, and Wisconsin.
Conclusions and Relevance
This study suggests that early tau deposition may follow multiple trajectories during preclinical AD and may involve several cortical regions. Staging procedures, especially those based on neuropathology, that assume a uniform trajectory across individuals are insufficient for disease monitoring with tau imaging.
Aggregation of tau into neocortical neurofibrillary tangles in association with amyloid, a key pathologic change of Alzheimer disease (AD), begins in older, clinically unimpaired individuals.1-6 Postmortem examination provides an incomplete picture of tau aggregates in the asymptomatic stages of AD because it is limited by sparse sampling.7 Most tau positron emission tomography (PET) studies of clinically unimpaired cohorts have focused on the medial temporal lobes (MTLs) and adjacent inferior temporal lobes, which are early regions consistent with Braak staging. However, both postmortem and imaging analyses demonstrate heterogeneity regarding MTL involvement8,9 as well as varying patterns of cortical involvement beyond the MTL that correspond to atypical nonamnestic clinical presentations of AD dementia, such as posterior cortical atrophy and logopenic variant primary progressive aphasia.10-15
It is currently unknown whether atypical tau patterns described in the context of clinically symptomatic AD dementia, often in association with atypical AD phenotypes, can be detected in preclinical AD cohorts. Estimates of tau positivity in clinically unimpaired individuals with elevated β-amyloid (A+) range from 15%, when positivity is based on a large temporal region consisting of the MTL and inferior and middle temporal regions,16 to approximately 40%, when positivity is based on the entorhinal cortex alone.17 However, to our knowledge, existing studies have not evaluated variability in the cortex despite suggestions of extratemporal involvement.17 The presence of divergent patterns of tau aggregation in individuals with preclinical AD has important implications for participant selection and evaluation of disease modification in prevention trials. Reliance on the same region of interest (ROI) across all individuals with preclinical AD in a secondary prevention trial may underestimate the effect of a drug that reduces the spread of tau if the pattern of spread varies across individuals. Thus, an understanding of the earliest patterns of tau deposition in individuals with preclinical AD has important implications for clinical trials and for understanding the factors associated with divergent tau patterns.
Given the 6% prevalence of atypical late-onset AD,18 we hypothesized that a similar percentage of clinically unimpaired A+ participants would show varying levels of cortical tau uptake that mimicked spatial patterns found in individuals with atypical clinical presentations of AD. We also expected that those with elevated cortical distributions would be younger,19 would be less likely to be APOE4 (OMIM 107741) carriers,12 and would have subtle deficits in nonmemory cognitive domains.
Primary analyses for this study used prerandomized data collected from April 2014 to December 2017 from the Anti-Amyloid Treatment in Asymptomatic AD (A4) Study,20 a secondary prevention trial that focused on participants with preclinical AD.21 All participants included in this study (n = 447; Table) were 65 to 85 years of age, clinically unimpaired (Clinical Dementia Rating score = 0, Mini-Mental State Examination score = 25-30, and Logical Memory Delayed Recall score = 6-18), completed amyloid PET scans ([18F]florbetapir) and tau PET scans ([18F]flortaucipir), and provided written informed consent before participation. Amyloid ([18F]florbetapir) status was determined using a hybrid quantitative and qualitative method established by the A4 Study,21 and regional standardized uptake value ratios (SUVRs) for both amyloid and tau PET were extracted using previously published processing pipelines (eMethods 1 and eFigure 9 in Supplement 1). The A4 Study, Alzheimer’s Disease Neuroimaging Initiative (ADNI), Harvard Aging Brain Study (HABS),22 and the Wisconsin Registry for Alzheimer’s Prevention23 and the Wisconsin Alzheimer’s Disease Research Center (together hereinafter referred to as Wisconsin) obtained institutional review board approvals for their respective studies.
Given that individuals with posterior cortical atrophy24 and logopenic variant primary progressive aphasia25 often show asymmetrical tau patterns11 and given that abnormal cortical involvement relative to MTL burden is characteristic of postmortem-defined hippocampal-sparing AD,9 we focused on these features to identify those with divergent cortical tau patterns. Mimicking procedures by Ossenkoppele et al,11 we used FreeSurfer-defined regions to create the following ROIs: (1) medial temporal–entorhinal, parahippocampal; (2) lateral temporal–superior, middle, and inferior temporal; (3) medial parietal–posterior cingulate, precuneus; (4) lateral parietal–inferior and superior parietal, supramarginal; (5) frontal–superior frontal, rostral and caudal middle frontal, lateral and medial orbitofrontal; and (6) occipital–pericalcarine, cuneus, lateral occipital. A hemispheric asymmetry index11 was calculated for each ROI: asymmetry index (%) = 200 × (right − left)/(right + left). Analogous to procedures used to define hippocampal sparing in postmortem data,9 a residual approach in which bilateral cortical ROIs were regressed against bilateral MTL was used to quantify disproportionate cortical involvement relative to MTL8,9 (eFigure 1 in Supplement 1).
Participants who were at least 3 SDs away from the A+ group mean in any of the cortical asymmetry indices or cortical vs MTL residual metrics were considered to have divergent cortical tau patterns (A+T cortical+ [Figure 1A]); a cutoff of 3 SDs was used given the relative rarity of atypical AD phenotypes. All A+T cortical+ tau PET data were visually inspected, and 1 individual with off-target binding was excluded (eFigure 2 in Supplement 1). Asymmetry indices and MTL residual metrics in the 5 cortical ROIs were z score normalized against the entire A+ group and used as features for hierarchical clustering based on Euclidean distance using the Ward minimum variance method26 to identify subgroups within the A+T cortical+ group. In addition, we identified a previously described pattern27-29 of a “typical” elevated tau PET signal that focused on the MTL (A+T MTL+). Given that this subtype should be more prevalent than cortical subtypes, we defined this group as including any individual in the A+ group who did not show divergent cortical tau patterns and who had MTL SUVRs that were at least 2 SDs greater than the mean among the group without elevated β-amyloid (A−).
Statistical analyses were performed from May 11, 2021, to January 25, 2022. A sequential statistical approach using R, version 4.0.4 (R Group for Statistical Computing) was used to ensure focus on primary outcomes of interest while simultaneously limiting the number of comparisons. First, the A+T cortical+, A+T MTL+, A+T MTL–, and A–T MTL– groups were compared. If group differences were significant across these 4 groups, the following post hoc tests were used to isolate specific outcomes of interest: A+T MTL+ group vs A–T MTL– group (outcomes of amyloid and MTL tau), A+T MTL+ group vs A+T MTL– group (outcome of MTL tau in preclinical AD), and A+T MTL+ group vs A+T cortical+ group (outcome of MTL vs cortical tau signal in preclinical AD). If the A+T MTL+ vs A+T cortical+ group comparison was significant, we then examined whether there were significant differences between the 3 A+T cortical+ subtypes. Nonparametric bootstrapping with pooled resampling based on 1-way analysis of variance and t tests that are recommended for small samples30 were used to determine significance for continuous variables. The Fisher exact test with additional false discovery rate correction for post hoc comparisons was used for categorical variables. P < .05 was considered significant, and all statistical tests were 2-sided.
Cognitive scores reflecting memory (ie, Logical Memory Delayed Recall, Free and Cued Selective Reminding Test [FCSRT] Free Recall, and Cog-State One-Card Learning Accuracy) and executive functioning (ie, Digit Symbol Substitution Test and Cog-State One-Back Accuracy) were first adjusted by age, sex, education, race, and ethnicity using the entire data set of 4486 A4 participants.21 These adjusted cognitive values were used in models contrasting A and T groups.
To evaluate the replicability of divergent cortical tau patterns, we examined clinically unimpaired individuals from the ADNI, HABS,22 and Wisconsin data sets23 (eMethods 2 in Supplement 1). This analysis provided a total of 273 clinically unimpaired A+ individuals across all 3 independent cohorts.
The 447 A4 participants (A+ group, 392; normal β-amyloid group, 55) with a mean (SD) age of 71.8 (4.8) years included 239 women (54%). Within the A4 cohort, we identified 36 A+ individuals (9% of all individuals in the A+ group) with asymmetrical cortical tau patterns and/or unusually high levels of cortical tau in comparison with MTL tau. Given the number of ROIs examined, a total of 1023 possible comnbinations across the criteria were possible. It is noteworthy that 23 of the 26 combinations were only present in a single A+T cortical+ individual (Figure 1B). Examination of the ADNI, HABS, and Wisconsin data sets yielded similar A+T cortical+ prevalence rates (ADNI: 12 of 149 [8%]; HABS: 5 of 59 [9%]; and Wisconsin: 7 of 65 [11%]; eTable 1 and eFigure 3 in Supplement 1).
Hierarchical clustering in the A4 cohort revealed 3 A+T cortical+ subtypes (asymmetrical left, precuneus dominant, and asymmetrical right; Figure 1C), and single-subject level tau patterns from the ADNI, HABS, and Wisconsin cohorts showed high similarity with the mean A+T cortical+ cortical maps derived from the A4 cohort (eFigure 4 in Supplement 1). Conjunction maps from the 3 A+T cortical+ subtypes in the A4 cohort showed varying locations in which most individuals within each subtype had elevated signals (inferior temporal for asymmetrical left and right subtypes; precuneus and lateral parietal for precuneus dominant subtype; eFigure 5 in Supplement 1). Further division of the A+T cortical+ asymmetrical right subtype, the largest subtype, was suggestive of 2 right asymmetrical subgroups that showed a similar spatial pattern but varied in the severity of tau PET signal magnitude (eFigure 6 in Supplement 1).
Given numerous studies showing elevated MTL tau among clinically unimpaired A+ individuals,27-29 those in the A+T cortical+ group were contrasted with those in the A+T MTL+ group. Of those who did not meet the criteria for A+T cortical+, 116 clinically unimpaired individuals in the A+ group were classified in the A+T MTL+ group (30% [116 of 392] of all individuals in the A+ group). The A+T MTL+ and A+T cortical+ groups had similar MTL SUVRs but differed across cortical regions (Figure 2; eTable 2 in Supplement 1).
Although global amyloid (eMethods 3 in Supplement 1) significantly differed between the A and T groups (F3,440 = 69.615; P < .001) and amyloid burden was higher in the A+T MTL+ than A−T MTL− (t158.44 = 18.409; P < .001) and A+T MTL− (t199.68 = 4.528; P < .001) groups, amyloid was not significantly different betwee the A+T MTL+ and A+T cortical+ groups (t55.78 = 1.645; P = .11; Figure 3A). Although elevations in tau PET were generally associated with elevations in amyloid PET, these associations were not regionally specific (eFigure 7 in Supplement 1). Examination of markers of neurodegeneration (ie, hippocampal volume and MTL and cortical thickness) showed that greater tau burden in MTL was associated with reduced hippocampal volume and MTL gray matter thickness across both the A+T cortical+ and A+T MTL+ groups. Furthermore, the A+T cortical+ group had reduced thickness in several cortical regions, but these reductions were not directly associated with regional tau SUVRs (eFigure 8 and eTable 3 in Supplement 1).
There was a significant age difference between the A and T groups (F3,440 = 7.833; P = .003; Figure 3B) such that the A+T MTL+ group was significantly older than the A–T MTL– group (t107.41 = –5.032; P < .001), the A+T MTL– group (t241.18 = –2.704; P = .004), and the A+T cortical+ group (t61.867 = –2.597; P = .03). There was no significant age difference between individuals within A+T cortical+ subtypes (F2,33 = 1.523; P = .23). Given the younger age of individuals in the A+T cortical+ group, we additionally examined the associations of age with global amyloid and regional tau. Age was not significantly associated with global amyloid (B = 0.003, SE = 0.003; P = .29); this association did not differ between the A+T cortical+ and A+T MTL+ groups (B = 0.006, SE = 0.006; P = .33; Figure 4A; eMethods 3 in Supplement 1). In contrast, the association between age and tau SUVR significantly differed between the A+T MTL+ and A+T cortical+ groups in medial parietal (B = –0.016, SE = –0.004; P < .001), lateral parietal (B = –0.017, SE = –0.004; P < .001), and occipital regions (B = –0.007, SE = –0.004; P = .04), such that younger age was associated with higher tau SUVR in the A+T cortical+ group (Figure 4B and C).
The A and T groups showed episodic memory differences in Logical Memory Delayed Recall (F3,440 = 7.165; P < .001) and Cog-State One Card Learning Accuracy (F3,419 = 4.002; P = .008) as well as trend-level differences in FCSRT Free Recall (F3,440 = 3.408; P = .08; Figure 3C; eTable 4 in Supplement 1). Post hoc comparisons indicated that the A+T cortical+ group had comparable performance to the A+T MTL+ group on Logical Memory Delayed Recall (t54.34 = 1.631; P = .11) and Cog-State One Card Learning Accuracy (t54.61 = 1.594; P = .11). The A and T groups showed executive functioning differences in Digit Symbol (F3,440 = 5.726; P = .02) and Cog-State One-Back Accuracy (F3,418 = 3.183; P = .02), with worse performance among the A+T cortical+ group than the A+T MTL+ group (Digit Symbol: t51.01 = 3.257; P = .001; Cog-State One-Back Accuracy: t56.93 = 2.247; P = .02; Figure 3C; eTable 4 in Supplement 1).
Across 4 cohorts of individuals with preclinical AD, approximately 9% showed asymmetrical cortical tau patterns and/or unusually high levels of cortical tau in comparison with MTL tau (A+T cortical+ group). The A+T cortical+ group was younger, had reduced cortical thickness, and had worse executive function than those with preclinical AD and tau elevation in MTL (A+T MTL+ group). These findings provide evidence for divergent cortical tau PET patterns in preclinical AD and suggest that distinct mechanisms may be associated with the spatial distribution of early tau spread. Our results also have implications for accurate disease staging and monitoring as well as ramifications for disease-specific therapeutic trials.31
Investigation of tau PET spatial heterogeneity in preclinical AD has previously been limited by sample size because only 20% to 30% of clinically unimpaired individuals in their 70s are A+,32 and less than half of A+ individuals will have an elevated tau signal, depending on the criterion used to define tau positivity.1,4,17 The comparatively smaller data sets of clinically unimpaired A+ individuals17 and multicohort approaches that are not specifically focused on clinically unimpaired A+ individuals33 have thus far been unable to characterize tau heterogeneity in preclinical AD. This finding is highlighted by our examination of 3 additional aging cohorts that included 149, 59, and 65 clinically unimpaired A+ individuals, yielding only 12, 5, and 7 individuals in the A+T cortical+ group, respectively. Characterization of potential tau PET subtypes during the mild cognitive impairment stage is also likely biased against the presence of divergent cortical tau distributions given the field’s focus on amnestic mild cognitive impairment that may preclude atypical phenotypes associated with impairments in nonmemory domains, such as visuospatial and language ability.11,12 Thus, the large sample of clinically unimpaired A+ individuals with tau PET in the A4 Study (n = 392) has resulted in an unprecedented opportunity to explore early tau PET variation, allowing for characterization of the 9% (n = 36) of clinically unimpaired individuals in the A+ group with divergent cortical tau patterns. This prevalence is somewhat consistent with the 5% of clinically unimpaired and mildly cognitively impaired who showed tau origins outside of the rhinal cortex.34
We used previously established methods to capture asymmetrical patterns as reported in atypical AD phenotypes, such as posterior cortical atrophy and logopenic variant primary progressive aphasia,11 as well as relative differences between cortical and MTL burden as reported in postmortem-defined hippocampal-sparing AD.9 The selected participants in the A+T cortical+ group could be further divided into 3 subtypes (asymmetrical left, precuneus, and asymmetrical right) that are similar to some of the subtypes recently identified by Vogel and colleagues33 in a large multicohort study examining clinically unimpaired individuals, those with mild cognitive impairment, and those with AD, as well as smaller studies focused on atypical AD patterns.10-13 Two of our subtypes show lateralization, similar to the lateral temporal subtype identified by Vogel and colleagues,33 which showed left asymmetry in the primary analyses but right asymmetry during replication. Our asymmetrical left subtype is also similar to descriptions of logopenic variant primary progressive aphasia.14,15,35,36 Posterior cortical atrophy presentations,14,15,35,36 dysexecutive presentations of early-onset AD,17 early-onset AD with PSEN1-E280A mutation,37 and the posterior subtype identified by Vogel and colleagues33 all include a precuneus component, similar to our A+T cortical+ precuneus subtype. The inverse age associations with multiple cortical regions within the A+T cortical+ group draw additional parallels with patterns noted in early-onset clinical cohorts.12 The tau PET patterns from individual participants in the A+T cortical+ group in the ADNI, HABS, and Wisconsin data sets converge with the subtypes in the A4 Study (eFigure 4 in Supplement 1). Taken together, it is clear that heterogeneity in tau PET patterns exists among cohorts with preclinical AD more frequently than currently appreciated.
The postmortem literature may provide an incomplete assessment of early tau aggregation. Traditional Braak staging does not consider the burden of tau pathology in the neocortex (ie, Braak stage V is defined by the presence of neurofibrillary tangles in the association cortex, and Braak stage VI is defined by the presence of neurofibrillary tangles in the primary cortex) or the heterogeneity of tau pathology in different association cortices. In addition, postmortem assessment does not always sample key regions, such as the precuneus. Thus, while divergent cortical tau PET patterns do not violate current Braak staging, Braak staging may be too vague to describe patterns in the cortex. With tau PET, previous work in the Mayo Clinic Study of Aging showed that approximately 40%, approximately 25%, approximately 20%, and approximately 15% of clinically unimpaired A+ individuals had elevations in the inferior temporal, precuneus, inferior parietal, and superior parietal, respectively; hemispheric differences were not reported.17 Together with our results (eFigure 6 in Supplement 1), the inferior temporal, lateral parietal, and precuneus may be especially vulnerable areas. Whether this vulnerability reflects distinct patterns of tau spread across interconnected regions and/or regional differences that may affect the local replication rate of tau seeds is unknown.38
We demonstrated that tau PET was associated with neurodegeneration within the MTL and that this association did not differ between the A+T MTL+ and A+T cortical+ groups. In contrast, tau and neurodegeneration were not significantly associated in most cortical regions, but the A+T cortical+ group had reduced cortical thickness in the right lateral temporal, left medial parietal, and right lateral parietal regions. The lack of association between regional tau and thickness beyond the MTL could reflect cortical tangle deposition emerging later than MTL deposition, as the predominant model of AD describes a temporal order of amyloid deposition, emergence of tau pathology, and then neurodegeneration.39 Our findings also raise questions related to differential cortical vulnerability and resistance to AD pathophysiology. Reduced cortical thickness in the temporal and parietal regions among the A+T cortical+ group compared with the A+T MTL+ group, as well as the lack of association between tau and cortical thickness in these regions, could imply that developmental, functional, or neurovascular differences in these regions may have a causal role in determining resistance or vulnerability to future tau deposition. Longitudinal magnetic resonance imaging data may help establish whether reductions in thickness reflect a neurodegenerative process that may result from tau elevations and/or represent a preexisting vulnerability.
The A and T subtypes described here likely reflect a combination of distinct phenotypes and staging, similar to a recently proposed 2-dimensional (typicality and severity) framework of heterogeneity.40 The A+T cortical+ group, in particular, may consist of individuals with divergent cortical tau phenotypes as well as individuals with the expected MTL burden who are at a more advanced stage. The combination of typicality and severity within the A+T cortical+ group is further demonstrated by the division of the A+T cortical+ asymmetrical right subtype, which showed that some participants had especially high cortical SUVRs (eFigure 5 in Supplement 1), suggesting that these individuals may reflect a more advanced stage of asymmetrical right tau deposition. In contrast, the asymmetrical left subtype could not be further subdivided when examining a 6-cluster solution. This may be because high left hemispheric tau burden that results in language deficits is more easily detectable during neuropsychological assessment because most neuropsychological tests are language-based and ubiquitous language deficits may more readily contribute to impairment in other domains, including memory. This possibility would result in individuals with high left hemispheric tau burden being excluded from the present sample of clinically unimpaired participants. In contrast, high right hemispheric tau burden associated with visuospatial deficits41 may be less likely to be detected if eligibility screening included minimal visuospatial assessment.21
Both memory and executive functioning were lower in the A+T cortical+ group, whereas those in the A+T MTL+ group showed reductions only in memory, although scores were still within normal limits and did not warrant a mild cognitive impairment diagnosis. Although MTL tau has been consistently associated with memory deficits even among clinically unimpaired individuals,27,42 a meta-analysis on cognition among clinically unimpaired A+ individuals also showed cognitive impairment in multiple domains beyond episodic memory, including visuospatial function, processing speed, and executive function.43 Incorporating additional tests that capture change in nonmemory domains during the preclinical AD stage is needed to understand the effect of heterogeneity on early tau spatial patterns.
This study has some limitations. First, all A and T groups were defined using arbitrary cutoffs despite a continuum of SUVRs in both the MTL and cortex. Second, the mean age across all the examined preclinical cohorts was in the late 60s to mid-70s, and it is known that individuals with atypical clinical presentations of AD often have symptom onset before 65 years of age. Thus, the prevalence of the tau PET subtypes identified in our work is likely higher among individuals with preclinical AD who are in their 50s and 60s. Third, the A4 Study design precludes large samples of individuals in the A–T MTL+ and A–T cortical+ groups because only 12% were A–. Fourth, our study included only cross-sectional tau PET and cognitive data, and cognitive data were limited because visuospatial and language domains were not assessed. Longitudinal follow-up, in terms of both repeated tau PET and cognitive assessments, is needed to understand the longitudinal time course of divergent cortical tau patterns, how this is associated with future clinical symptoms, and the degree to which subtypes reflect different stages of disease progression. Fifth, because it is possible that [18F]flortaucipir is binding to alternative forms of tau, glial tau aggregates, non-AD tau isoforms, or non–AD-related processes,44-46 and because recent postmortem to imaging comparisons have called [18F]flortaucipir sensitivity at pre–Braak V stages into question,47 some of the A+T cortical+ cases identified here may instead reflect other nontau processes. Sixth, the data sets examined here are primarily of highly educated non-Hispanic White individuals, limiting the generalizability of these findings.
This cross-sectional study found that 9% of clinically unimpaired A+ individuals had a divergent cortical tau PET pattern, which could further be grouped into multiple preclinical AD subtypes. Participants with preclinical AD and cortical tau were younger, had reduced cortical thickness, and showed worse executive functioning performance compared with participants with preclinical AD and MTL tau PET elevations. Tau PET staging procedures, especially those based on neuropathology, that assume a uniform trajectory across individuals with preclinical AD may be insufficient for disease monitoring, enrollment into therapeutic trials, and attempts to understand the initiating mechanisms driving tau spread.
Accepted for Publication: February 3, 2022.
Published Online: April 18, 2022. doi:10.1001/jamaneurol.2022.0676
Corresponding Author: Christina B. Young, PhD, Department of Neurology and Neurological Sciences, Stanford University School of Medicine, 453 Quarry Rd, Palo Alto, CA 94304 (firstname.lastname@example.org).
Author Contributions: Dr Young and Ms Cody had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Young, Sperling, Mormino.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Young, Mormino.
Critical revision of the manuscript for important intellectual content: Young, Winer, Younes, Cody, Betthauser, Johnson, Schultz, Sperling, Greicius, Cobos, Poston.
Statistical analysis: Young, Cody, Mormino.
Obtained funding: Young, Sperling, Poston, Mormino.
Administrative, technical, or material support: Betthauser, Johnson, Schultz, Greicius, Poston, Mormino.
Supervision: Sperling, Poston, Mormino.
Conflict of Interest Disclosures: Dr Young reported receiving grants from the Alzheimer’s Association during the conduct of the study. Ms Cody reported receiving grants from the National Institutes of Health (NIH) during the conduct of the study. Dr Johnson reported receiving grants from the NIH during the conduct of the study; and personal fees from Roche Diagnostics outside the submitted work. Dr Schultz reported receiving grants from the NIH during the conduct of the study; and personal fees from NervGen, Qynapse, Biogen, and Janssen outside the submitted work. Dr Sperling reported receiving grants from the National Institute on Aging and the Alzheimer’s Association during the conduct of the study; and personal fees from AC Immune, Janssen, Genentech, Ionis, NerGen, Shionogi, and Oligmerix outside the submitted work. Dr Greicius reported receiving grants from the NIH during the conduct of the study. Dr Poston reported receiving grants from the NIH/National Institute of Neurological Disorders and Stroke during the conduct of the study; and grants from MJFF, Alzheimer’s Drug Discovery Foundation, and Lewy Body Dementia Association outside the submitted work. Dr Mormino reported receiving grants from the NIH during the conduct of the study; and personal fees from Eli Lilly, Roche, and NeuroTrack; and grants from the Alzheimer’s Association outside the submitted work. No other disclosures were reported.
Funding/Support: This study was supported by grants P30AG066515 (Drs Greicius, Poston, and Mormino), F32AG074625 (Dr Winer), K01AG051718 (Dr Mormino), R01AG027161 (Dr Johnson), R01AG021155 (Dr Johnson), RF1AG027161 (Dr Johnson), S10OD025245, P30AG062715, U54HD090256, UL1TR002373, P01AG036694 (Dr Sperling), and P50AG005134 (Dr Sperling) from the NIH; grants AARFD-21-849349 (Dr Young) and AARF-19-614533 (Dr Betthauser) from the Alzheimer’s Association; and the Ben Barres Early Career Acceleration Award from the Chen Zuckerberg Initiative (grant 199150; Dr Cobos).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Group Information: The members of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the Harvard Aging Brain Study are listed in Supplement 2.
Additional Information: The A4 Study is a secondary prevention trial in preclinical Alzheimer disease, aiming to slow cognitive decline associated with brain amyloid accumulation in clinically normal older individuals. The A4 Study is funded by a public-private-philanthropic partnership, including funding from the National Institutes of Health–National Institute on Aging, Eli Lilly and Company, Alzheimer’s Association, Accelerating Medicines Partnership, GHR Foundation, an anonymous foundation and additional private donors, with in-kind support from Avid and Cogstate. The companion observational Longitudinal Evaluation of Amyloid Risk and Neurodegeneration (LEARN) Study is funded by the Alzheimer’s Association and GHR Foundation. The A4 and LEARN Studies are led by Dr Sperling at Brigham and Women’s Hospital, Harvard Medical School and Dr Paul Aisen at the Alzheimer’s Therapeutic Research Institute (ATRI), University of Southern California. The A4 and LEARN Studies are coordinated by ATRI at the University of Southern California, and the data are made available through the Laboratory for Neuro Imaging at the University of Southern California. The participants screening for the A4 Study provided permission to share their deidentified data in order to advance the quest to find a successful treatment for Alzheimer disease. We would like to acknowledge the dedication of all the participants, the site personnel, and all of the partnership team members who continue to make the A4 and LEARN Studies possible. The complete A4 Study Team list is available at a4study.org/a4-study-team. Data collection and sharing for this project was funded by the ADNI (National Institutes of Health grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12–2–0012). The 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 Co, CereSpir Inc, Cogstate, Eisai Inc, Elan Pharmaceuticals Inc, Eli Lilly and Co, 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 Corp, Pfizer Inc, Piramal Imaging, Servier, Takeda Pharmaceutical Co, and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.
Data used in preparation of this article were obtained from the ADNI database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf. Data used in preparation of this article were obtained from the Harvard Aging Brain Study (HABS; P01AG036694; https://habs.mgh.harvard.edu/). The HABS was launched in 2010, funded by the National Institute on Aging, and is led by principal investigators Reisa A. Sperling, MD, and Keith Johnson, MD, at Massachusetts General Hospital/Harvard Medical School.