Regional Distribution,Asymmetry,and Clinical Correlates of Tau Uptake on [ 18 F]AV-1451 PET in Atypical Alzheimer’s Disease

INTRODUCTION

Alzheimer’s disease (AD) can have many different clinical presentations, despite a common underlying pathology of amyloid-β (Aβ) and tau deposition in the brain. Pathological studies have shown that there is little association between location and density of Aβ deposition and cortical atrophy or clinical symptoms in AD; the location and density of tau neurofibrillary tangles at autopsy, on the other hand, have shown stronger correlations with both atrophy and symptoms [1–4]. Until recently, investigating the distribution of tau deposition has only been possible postmortem, but PET ligands, such as [¹⁸F]AV-1451, have now been developed that can detect tau proteins in vivo in patients with AD [5, 6]. This technology can thus be used to investigate the relationship between tau deposition and clinical presentation during life in AD.

The two most commonly recognized atypical clinical variants of AD are posterior cortical atrophy (PCA) [7, 8] and logopenic progressive aphasia (lvPPA) [9]. Posterior cortical atrophy is characterized by difficulties with visuospatial, visuoperceptive, and praxic skills; these deficits can also be accompanied by Balint’s or Gerstmann’s syndromes [7, 11]. In contrast, lvPPA typically presents with language difficulties, including impaired single-word retrieval, difficulties with sentence repetition, and phonological errors, some of which are believed to be caused by impairment in the phonological loop of working memory [12–14]. The majority of patients diagnosed with either PCA or lvPPA have underlying AD pathology or show Aβ deposition on PET [15–18].

These two disorders are associated with different, although somewhat overlapping, patterns of neurodegeneration as measured on volumetric MRI and [¹⁸F] fluorodeoxyglucose PET. The main regions of involvement in PCA include the occipital, parietal, and occipito-temporal regions [17, 19], while those affected in lvPPA include the posterior portions of the superior and middle temporal lobe, the inferior parietal lobe, and the temporo-parietal junction [20, 21]. Patterns of neurodegeneration are usually asymmetric in lvPPA, involving the left hemisphere greater than right, and can also be asymmetric in PCA [22], more commonly involving the right hemisphere. Given the close association between tau deposition in the brain and the presence of neurodegeneration, one could hypothesize that patterns of tau uptake on [¹⁸F]AV-1451 PET would also differ between these two atypical AD syndromes. Understanding how tau deposition differs between lvPPA and PCA will increase understanding of the biological mechanisms underlying clinical variability in AD and may help improve recognition and diagnosis of these syndromes.

Therefore, the goal of this study was to assess and compare the regional patterns of tau uptake using [¹⁸F]AV-1451 PET in a cohort of PCA and lvPPA subjects. We also aimed to investigate whether asymmetry in tau uptake differed across the syndromes, and determine whether regional tau uptake correlates with clinical performance in these groups.

METHODS

RESULTS

Group comparisons of tau uptake

Patients versus controls

The PCA group showed increased uptake bilaterally in the occipital, temporal and parietal lobes, with less striking uptake also observed in the frontal lobes (Figs. 1 and 2). In contrast, increased tau uptake was observed throughout the left hemisphere of the brain in lvPPA, particularly involving the lateral temporal lobe, but also the medial and lateral parietal and frontal lobes, with less striking involvement of the occipital lobe compared to controls (Figs. 1 and 2). Similar, although less severe, patterns of uptake were observed in the right hemisphere. The highest AUROCs discriminating PCA from controls were observed for left and right lateral occipital regions, while the highest AUROCs discriminating lvPPA from controls were observed for left lateral temporal and parietal regions (Fig. 3).

OPEN IN VIEWER

Fig.1

Voxel-level comparisons of tau PET uptake in PCA, lvPPA, and controls. Results are shown at p < 0.001 corrected for multiple comparisons using the family wise error correction.

OPEN IN VIEWER

Fig.2

Box-plots showing regional tau SUVRs for PCA, lvPPA, and controls. Boxes represent median, 25th and 75th quartiles, and whiskers extend to 1.5 x the quartiles. Subject data points that are beyond the whiskers are shown.

OPEN IN VIEWER

Fig.3

Plots showing AUROC values depicting how well tau SUVR in each region differentiates PCA from controls, and lvPPA from controls. AUROCs are shown separately for the left and right hemisphere. *indicates subcortical regions.

Increased tau uptake was also observed in left and right medial temporal regions, including hippocampus, parahippocampal gyrus, entorhinal cortex and amygdala, and temporal pole in both PCA and lvPPA compared to controls (Fig. 2), although these regions tended to have lower SUVRs compared to cortical regions (Fig. 2). The only regions that did not show bilaterally significant differences in tau uptake between both patient groups and controls were the thalamus and basal ganglia (Fig. 2).

PCA versus lvPPA

Comparing PCA and lvPPA directly revealed disease specific regions of tau deposition. On the voxel-maps, PCA showed increased tau uptake throughout the occipital lobe, bilaterally, but more so in the right hemisphere, compared to lvPPA (Fig. 1). lvPPA showed increased tau uptake in the left temporal lobe, particularly the temporal pole, and medial prefrontal cortex compared to PCA (Fig. 1).

In the ROI analysis, PCA showed significantly greater uptake bilaterally in the cuneus, calcarine, and lingual regions, as well as in right inferior, middle, and superior occipital lobe (FDR adjusted p < 0.05), compared to lvPPA. Conversely, lvPPA showed greater tau uptake in the left temporal pole (FDR adjusted p = 0.03) compared to PCA. These results were similar when using PVC tau-PET images, with the addition of the left basal ganglia uptake being significantly higher in lvPPA than PCA.

Group comparisons of grey matter volume

The PCA group showed volume loss bilaterally in the medial and lateral occipital and parietal lobes, lateral temporal lobes, posterior hippocampus, and parahippocampal gyrus compared to controls (Fig. 4). The lvPPA group showed volume loss predominantly in the left lateral temporal lobe, with less severe volume loss observed in the right lateral temporal lobe. On direct comparison, PCA showed greater volume loss in the right occipital lobe and fusiform gyrus compared to lvPPA. No regions showed greater loss in lvPPA compared to PCA.

OPEN IN VIEWER

Fig.4

Voxel-level comparisons of grey matter volume loss in PCA, lvPPA, and controls. Results are shown at p < 0.001 corrected for multiple comparisons using the family wise error correction.

DISCUSSION

We have shown that both PCA and lvPPA are characterized by widespread cortical tau deposition, although there was evidence for differences between the syndromes, with PCA showing greater uptake in the occipital lobe and lvPPA showing greater uptake in some left frontal and temporal regions; sub-cortical regions were less affected in both diseases. Both diseases were asymmetric in terms of tau uptake; however, lvPPA is strikingly left-predominant, while PCA is more variable. We also found evidence that regional tau deposition is associated with specific clinical features in these atypical AD syndromes.

The regional distribution of tau uptake we observed in both PCA and lvPPA fit well with the previously described patterns of neurodegeneration associated with these syndromes, as well as the distribution of tau pathology observed at autopsy [1, 39]. In line with studies that have looked at autopsy data, we found that tau uptake was greater in the cortex compared to the medial temporal lobes in both syndromes [40, 41], possibly reflecting the fact that non-memory symptoms dominate the clinical presentation in both syndromes. Another consistent feature was that PCA and lvPPA showed less elevated tau uptake in the thalamus and basal ganglia; a feature that could help differentiate these AD syndromes from other neurodegenerative diseases that target these regions, such as progressive supranuclear palsy [42]. Although patterns of tau uptake overlapped to a relatively large degree in lvPPA and PCA, we did find regions that differed between the syndromes. The most striking difference was the finding that PCA had greater tau uptake throughout the right occipital regions compared to lvPPA. Two previous group-level tau-PET studies have compared PCA and lvPPA, with both finding greater tau uptake in the medial parietal lobe in PCA [4, 43], and one finding greater involvement of the occipital lobe [4]. Discrepancies between these studies and with our findings may be due to the small number of subjects in the previous studies, which had sample sizes of n = 7 or less. We also found evidence that lvPPA has greater tau uptake compared to PCA in frontal and temporal regions, consistent with one previous study [43]. These group differences provide a potential biological explanation for the clinical differences observed between these two AD syndromes; a hypothesis that is supported by the correlations we observed between tau uptake and clinical performance that are discussed below.

Past studies have shown asymmetric patterns of FDG hypometabolism and grey matter atrophy in PCA and lvPPA cohorts. Cerebral atrophy and FDG hypometabolism in PCA has shown right-sided asymmetry, with the occipital lobe being the most affected [17, 44]. lvPPA, on the other hand, has been shown to have left-predominant asymmetry in both atrophy and FDG hypometabolism, primarily in the left posterior temporal, lateral temporal, inferior parietal, and inferior frontal regions [13, 45]. In this study, we have shown that both PCA and lvPPA can have asymmetric patterns of tau uptake. Tau uptake was relatively bilateral in the PCA cohort, especially in the occipital regions; nonetheless, PCA patients did commonly show right or left predominance. lvPPA showed left-sided asymmetry in all lobar regions, most prominently in the frontal, parietal, and occipital regions, with the majority of subjects showing left-sided asymmetric patterns. The temporal lobes also showed significant asymmetry but had the highest proportion of subjects with bilateral patterns. We could hypothesize that the lower degree of asymmetry in the temporal lobes may reflect the fact that tau deposition spreads to the right temporal lobe before other regions in the right hemisphere in lvPPA; a hypothesis that will need to be tested in longitudinal studies. These tau-PET findings of hemispheric asymmetry are particularly valuable because autopsy studies are typically performed with only one hemisphere, and hence do not allow the assessment of asymmetry in tau pathology. The fact that we looked at tau in both hemispheres, and that it seems to match asymmetry in grey matter atrophy and FDG hypometabolism, supports an association between tau deposition and cerebral atrophy [1–3, 46].

We also found evidence for a relationship between the degree of tau deposition and age at the time of the PET scan. Tau uptake in frontal and parietal lobes, bilaterally, showed a strong negative correlation with age, with weaker correlations in the occipital and temporal regions, such that younger patients had increased tau deposition. This mirrors findings in typical AD, where patients with early onset AD have increased cortical tau uptake compared to their late onset counterparts, specifically in parietal regions, which has been hypothesized to be a result of increased neuro-inflammation in the younger patients [47]. Our findings demonstrate that this age effect is present even within a relatively young cohort of only atypical AD subjects.

Strong negative correlations were also observed between tau uptake and clinical features in atypical AD suggesting a close relationship between tau deposition and clinical presentation. Tau uptake bilaterally in the frontal and temporal lobe was associated with deficits in sentence repetition, as well as in the left parietal region, likely reflecting disruption of the language network [48]. Tau uptake in both temporal lobes was also associated with poorer naming, concordant with the role of the temporal lobes in anomia and semantic processing [49]. Tau uptake bilaterally in the occipital lobe was related to severity of both simultanagnosia and visuoperceptual function, features which have previously been associated with dysfunction in the occipital lobe [4, 50–52]. Given that these tests evaluate clinical symptoms related to PCA and were negatively correlated with tau in the occipital regions, which was the most affected in this cohort, these exams may be good predictors of tau deposition. We did not observe any correlations between general cognitive ability measured on the MoCA and tau-PET uptake, perhaps suggesting that the MoCA is not anatomically specific.

Although this study was not designed to assess the topographic similarity in patterns of tau uptake and atrophy in our cohort, we did find that the pattern of tau uptake was similar to the pattern of grey matter loss in both PCA and lvPPA. Grey matter atrophy was, however, notably less severe and widespread than tau uptake, as others have noted [43]. For example, we observed elevated tau uptake, but very little atrophy, in the frontal lobes in both PCA and lvPPA, with atrophy in lvPPA mainly restricted to the lateral temporal lobes. This may have a biological explanation; possibly reflecting the fact that tau deposition likely precedes atrophy. However, it is unclear whether the same statistical threshold is appropriate for these different types of neuroimaging data, with different underlying properties, and hence one should be cautious in drawing conclusions from comparisons of thresholded maps.

Strengths of our study were that all subjects underwent identical clinical and imaging assessments, and we corrected the data for partial volume effects and showed that this correction did not significantly affect the findings. A limitation, however, was the relatively small number of subjects. While non-parametric statistical methods were used in the ROI-based analysis, the voxel-based analysis utilized parametric statistical tests which may have been biased by the small sample size. Another weakness was that the PCA cohort had a trend for longer disease duration than the lvPPA cohort (FDR p = 0.07). It is possible that we may have observed more regions with greater tau uptake in lvPPA compared to PCA, had the two disease groups had more similar durations. In addition, while tau-PET is relatively specific for AD neurofibrillary tangles, we have shown that tau-PET signal can be seen in off-target sites and other tauopathies, and these findings may need to be validated by autopsy studies [53]. All of our subjects showed Aβ deposition supporting an underlying AD pathology; future studies will be needed to determine whether tau PET uptake differs in PCA and lvPPA subjects without Aβ deposition.

In summary, this study demonstrates that tau is densely and diffusely present throughout the majority of cortical regions in both PCA and lvPPA. Even so, the findings of this study show that tau-PET scans have potential to distinguish different types of atypical AD variants. Particularly, PCA and lvPPA can be distinguished by the amount of tau uptake in the occipital regions, where PCA subjects have increased tau deposition compared to lvPPA subjects, particularly in the right hemisphere. We also provide evidence that tau deposition is closely related to clinical presentation in AD and that age plays an important role in determining the degree of cortical tau deposition.