An autoradiographic evaluation of AV-1451 Tau PET in dementia

Tauopathies constitute a group of neurodegenerative disorders associated with molecular alterations in the microtubule associated protein tau that can be sub-classified by the predominant species of tau that accumulates within neurons and glia [1]. Tau protein promotes tubulin polymerization and acts to stabilize microtubules, which are abundant in axons and important for many neuronal processes, such as axoplasmic transport. Extensive post-translational modifications of tau (e.g., phosphorylation) affect its functional properties, leading to dissociation from microtubules and abnormal accumulation in intracellular inclusions in neurons known as neurofibrillary tangles (NFT). Neuronal accumulation of abnormal tau and its attendant neuronal loss correlate with cognitive deficits in AD [2, 3]. The relative amounts of neuronal and glial accumulation and the neuroanatomical distribution of tau define a range of neurodegenerative tauopathies, including tangle predominant dementia [4, 5], argyrophilic grain disease (AGD) [6], Pick disease (PiD) [7], progressive supranuclear palsy (PSP) [8] and corticobasal degeneration (CBD) [9]. The tau gene (MAPT) has 13 exons, and it undergoes alternative splicing of exons 2, 3 and 10 to generate 6 isoforms [10]. Exon 10 encodes one of four 32-amino acid conserved sequences in the microtubule-binding domain of tau. Tau that includes exon 10 has four repeats (4R tau) in the microtubule binding domain, while tau that excludes exon 10 has three repeats (3R tau) [1]. Variable amino terminal inserts (N) are also present leading to 6 isoforms denoted as 2N4R, 2N3R, 1N4R, 1N3R, 0N4R, and 0N3R, the largest of which being 2N4R with 441 amino acids [1]. The composition of fibrillar tau also differs across tauopathies, related to stoichiometry of the insoluble isoforms that accumulate in the brain, levels of hyperphosphorylation, and the morphology of the fibrillar aggregates (paired helical filaments (PHF) or straight filaments (SF), (reviewed in [11]). Although the specific nature of tau deposits and their neuroanatomical distributions have guided neuropathologic diagnosis of tauopathies, diagnostic accuracy during life remains poor in many of these disorders. In vivo imaging of tau pathology may greatly improve not only diagnostic accuracy, but enhance efforts toward early detection and treatment.

Over the last decade, amyloid imaging using positron emission tomography (PET) scanning has successfully demonstrated evidence of cerebral Aβ deposition in individuals with AD [12] and longitudinal amyloid imaging of neurologically normal individuals shows transition to amyloid positive status in those at risk for AD [13], such that determinations of amyloid status is being integrated into clinical diagnosis of AD [14]. Since NFTs correlate with cognitive decline in AD better than amyloid [15, 16], it is important to better understand the status of tau deposition. As such, recent efforts have focused on developing tau-imaging agents. PET imaging with AV-1451 has shown highly specific binding affinity to tau in AD patients [17, 18]. However, understanding AV-1451 binding to tau in AD, atypical AD and non-AD tauopathies and its relative binding to other pathologic aggregates is also of great importance and is beginning to be elucidated [19]. In addition, anecdotal reports of purported “off-target” binding of AV-1451 have been reported in diseases not considered to be associated with tau accumulation (unpublished data). In patients at our institution, PET imaging with AV-1451 has shown AV-1451 binding prominently in AD, but also in “off-target” sites, and minimally in clinically diagnosed frontotemporal dementia (FTD) and non-AD tauopathies (unpublished data) (Fig. 1). Understanding the relationship between AV-1451 binding and pathology could help define the potential utility of this technology, improve our understanding of the pathologic mechanisms of dementia and direct targeted therapeutic approaches. It is essential, therefore, to determine the specificity of AV-1451 PET and its pathologic correlates. Using autoradiography and immunohistochemistry, our study had three main objectives: 1) to evaluate the binding properties of AV-1451 to AD-tau and compare it to several non-AD tauopathies; 2) to evaluate the specificity of AV-1451 binding in disorders not associated with significant tau deposition such as TDP-43 proteinopathies and α-synucleinopathies; and 3) to investigate the nature and patterns of “off-target” AV-1451 binding.

This work demonstrates that AV-1451 binding on autoradiography corresponds relatively well to tau pathology demonstrated by immunohistochemistry, with preferential binding to AD-type tau pathology, although the relationship is surprisingly complex. AV-1451 binding is absent or minimal in non-AD primary tauopathies and TDP-43 proteinopathies, and absent in α-synucleinopathies. We also provide evidence for several types of AV-1451 “off-target” binding, which we hypothesize to either be of a pigment-based or mineralization-based etiology.

A few previous studies have described AV-1451 binding characteristics in brain tissue [19, 38, 40], but these prior reports did not include many novel observations that our study includes. First, our study includes IHC capable of detecting differences in AV-1451 binding to tau to predominant forms of early and late neurofibrillary tangle maturity, where CP13 assesses earlier tau pathology and PHF-1 assesses more advanced tangle pathology. We found differences in the association of AV-1451 with tau maturity using both PHF-1 and CP13 immunohistochemical staining. This difference implies that AV-1451 is specific to particular hyperphosphorylated regions or related conformations and therefore has variable binding depending on the maturity of neurofibrillary tangle pathology. CP13 is an antibody to anti phospho-serine 202 tau that phosphorylates at an early time in tau hyperphosphorylation relative to PHF-1 (phospho-serine 396/404 specific) [41]. A demonstration of this effect is observed in Fig. 2 (D,E,F) where more CP13 positive regions are seen relative to PHF-1 and AV-1451 correlates better with PHF-1 findings or a presumably a later hyperphosphorylation time point. A related novel finding is the variable layering of AV-1451 in the grey matter that was often found in AD cases (Fig. 3), which has not been described previously. These findings have important implications to the sensitivity of AV-1451 at different stages of tau maturity and are described in more detail below. Second, we provide evidence for the first time describing AV-1451 binding in PART and tangle predominant dementia. Both neuropathologies may underlie medial temporal lobe changes measured by structural MRI or PET studies. Our study provides the first foundation related to these two pathologies and putative labeling by tau imaging as compared to AD. Third, TDP-43 associated AV-1451 binding is demonstrated as a sporadic finding (similar to the findings of Sander, et al,[38]) and additional patient examples are provided to demonstrate potential tauopathies underlying clinical phenotypes of FTD (Fig. 4) and are therefore additionally possible explanations for AV-1451 positive scans in FTD patients. Fourth, while ref 35 does describe some examples “off target’ binding, the present paper shows additional examples that are not described in the prior literature that include vascular, choroid plexus, basal ganglia, pituitary, and subpial AV-1451 binding.

AV-1451 autoradiography showed marked binding to multiple cortical and limbic regions in AD, where tau pathology is composed of relatively equimolar 3R and 4R tau [42]. The AV-1451 binding was strong and diffusely positive in cases with high Braak NFT Stages as expected and as previously described [19]. However, we observed different AV-1451 binding patterns within the regions when comparing two phospho-tau antibodies (PHF-1 and CP13) with different profiles on immunoreactivity with tau positive structures in AD that describe early (CP13) and later (PHF-1) tau development. The PHF-1 profile is as follows: neuritic pathology < pretangles < mature tangles > extracellular tangles (with pretangles and extracellular tangles being similarly labeled); while CP13 (labels neuritic pathology ≈ pretangles > mature tangles > > extracellular tangles. The findings depicted in Fig. 2 suggest that early tau deposition, as best shown by CP13, is not represented well in AV-1451 autoradiography (i.e., medial parahippocampal gyrus (red dashed arrows), lateral parahippocampal gyrus (red arrows) and subiculum and CA1 (green dashed arrows) regions, Fig. 2 D-F, J-L). PHF-1 and AV-1451 moderate correspondence is consistent at each successive Braak NFT stage. Although tangle predominant dementia typically has marked medial temporal lobe AD-type tau pathology with Braak NFT stages III or IV, hippocampal neuronal loss is much higher than would be expected in early AD or the mild spectrum of PART [4, 43]. Tau pathology in tangle predominant dementia is primarily composed of extracellular 3R tau positive “ghost” tangles [44]. This shift from a 3R + 4R tau ratio to an overrepresentation of 3R could explain the relatively minimal AV-1451 binding we observed in tangle predominant dementia (Fig. 2 M-O). On the recent paper by Marquie et al. [19], ghost tangle (extracellular) binding of AV-1451 is described on nuclear emulsion autoradiography in AD-tau but the finding is not well demonstrated as no individual IHC and AV1451 comparison images are provided and the technique is therefore unable to confirm our findings. Tangle predominant dementia cases may be the extreme form of PART [20]. There is much controversy surrounding the nature of the tau pathology in PART and whether PART is actually an early stage of AD [45, 46]. Mild PART cases (Fig. 2 J-L) provide supportive evidence that a critical mass of NFTs may be needed for strong AV-1451 binding, as the medial parahippocampal gyrus region has many more NFTs (PHF-1 inset) than the subiculum and also has much stronger AV-1451 binding. As PART is defined by lack of amyloid pathology or minimal amyloid (Thal amyloid phase ≤ 2; possible PART) [37], multi-modal approaches using amyloid PET will be useful – especially as evidence supports Thal amyloid phase ≤ 2 roughly equates to the threshold in established using our PIB data processing pipeline [15]. Knowledge of amyloid burden is of particular importance in AD cases with atypical distribution of tau pathology, as is evidenced by case #6, (Fig. 2 G-I). This case of advanced AD presented as a non-amnestic, svPPA and was not suspected of having AD. AV-1451 binding showed moderate-to-strong binding in cortex, but minimal-moderate binding in hippocampus (green dashed arrow) as the hippocampus was relatively spared of NFT [30]. Of note, off-target binding in the lateral geniculate nucleus was observed in this case, which could confound lower resolution clinical AV-1451 PET scan findings as a result of AV-1451 uptake in a region near the hippocampus.

An interesting pattern of AV-1451 binding was seen in cortical gray matter of advanced AD cases, such as in Fig. 3. AV-1451 was most prominent in deeper cortical layers, corresponding to pyramidal cell layers, and reduced in upper cortical layers, corresponding to molecular layer and granular cell layers. A similar layering effect can be seen in Marquie et al. [19], (Fig. 2, row A, left box) but was not described in that work. This pattern of AV-1451 binding that we observed contrasted with results of tau immunohistochemistry where many tau positive neurites were detected in upper cortical layers. The association of AV-1451 binding to NFTs more than neurites may explain preferential binding to layers with pyramidal neurons as well as the observation of focal, intense AV-1451 binding in CA1 and the subiculum of the hippocampus, reflecting areas with densely packed NFTs. The findings suggest a preference of AV-1451 binding to NFT and less to neurites, but it is not clear if or why such a preference would exist. It could be due to a higher molecular abundance of tau associated with cell bodies compared to neurites. There could be tau conformational differences or differences in filament organization in neurites compared to NFT that correlated with AV-1451 binding. Implications related to maturity of tau may be postulated as well. There are related reports indicating differences in timing of tau pathology with initial pathologic changes in AD occurring in different layers of the entorhinal/parahippocampal cortex [47, 48]. These preferential binding mechanisms remain however speculative and further work is needed to elucidate the pathologic mechanisms of these findings.

We observed a wide variation in AV-1451 binding in non-AD neurodegenerative tauopathies. Prior work with other tau-PET probes has shown that THK-523 [49] does not bind to non-AD tau and PBB-3 [50] does bind to tau from non-AD tauopathies namely Pick’s disease, PSP and CBD. Marquie et al [19], present that AV-1451 does not bind to non-AD tauopathies or TDP-43 but Sanders, et al, have data that suggest that AV-1451 does bind to non-AD tauopathies (greatest in PiD and FTDP-17) and additionally to FTLD-TDP Types A and C (1/3 cases for each type). Our findings for AV-1451 align best with those of Sanders, et al.. We have not tested the other tau-PET probes but it is clear that differences in tau specificity of various probes will make clinical utilization complicated. Different methods of investigation such as fluourescence with PBB-3 [50] make direct comparison between these reports problematic. Head-to-head testing of the different probes would be helpful. The intensity of AV-1451 uptake that we see in different tauopathies could be a reflection of the different tau isoforms or structural differences in tau filaments. Higuchi and colleagues have reported that AD and several other diseases (tangle predominant dementia, Down syndrome, Guam-Parkinson-dementia complex, atypical CJD with NFTs, Niemann-Pick disease type C, and FTDP-17 with R406W MAPT mutation) have morphologically similar neurofibrillary structures composed of 95 % PHFs and minimal straight filaments (SFs) [51]. In contrast, PiD and PSP have neurofibrillary structures that contain more SFs and fewer PHFs. Morphologically different ribbon-like tau filaments are typical of FTDP-17 and CBD; these filaments are wider and have irregular periodicity of 90 to 130 nm [51]. In addition, while all tau isoforms are present in AD, 3R isoforms predominate in PiD and 4R isoforms predominate in CBD and PSP. Either morphological or isoform differences could be implicated in the different intensity of AV-1451 binding in these disorders. The tendency of AV-1451 to show greater intensity of localization in AD than PiD, PSP or CBD would suggest that AV-1451 has greater affinity for PHF-rich neurofibrillary structures as opposed to SF-rich structures (PiD and PSP), as recently suggested [19]. Our data adds to these findings by demonstrating strong binding in the R406W MAPT mutation case, which was also seen by Sander et al., corresponding to moderate-severe PHF-1 immunoreactive tau pathology. Filament specificity is supported by our observations of poor correspondence between ARG and IHC in N279K mutation carriers (prevalence of mixed twisted ribbon and straight filaments) and P301L mutation carriers (predominantly narrow twisted filaments), compared to R406W mutation carriers (mix of paired helical filaments and straight filaments). The different isoforms may also play a role in the different specificity. The strength of AV-1451 binding with respect to tau isoforms that preferential accumulate in various neurodegenerative disorders was as follows: 3R + 4R tau (e.g., AD) > 3R tau (e.g., Pick disease) or 4R tau and also matches the pattern seen by Sander et al. Our findings suggest that disorders with tau pathology have sufficient differences with respect to binding to PET ligands that they may need to be tailored to particular tau species.

We found no consistent evidence for correspondence of AV-1451 with TDP-43 pathology. In the FTLD-TDP cases, the amygdala and frontal lobe tissue showed no AV-1451 binding although these areas had TDP-43 pathology on adjacent sections. On the other hand, we did find some AV-1451 binding greater than background in regions of TDP-43 pathology in two of the three FTLD-TDP cases in frontal, parahippocampal and temporal cortices. These results may represent genuine, weak co-localization with TDP-43, although we cannot confirm this, but there may be other unexplained pathologic causes. One potential explanation relates to severe degeneration of subjacent white matter in many cases of FTLD-TDP, which may lack background or basal-level radiotracer uptake, thus enhancing the contrast of signal between gray and white matter. Data from clinically imaged FTLD patients with high likelihood of TDP-43 pathology, such as those with svPPA, will be necessary to evaluate the frequency of this finding, with eventual confirmation depending upon postmortem studies. There have been anecdotal reports (unpublished) of AV-1451 PET uptake in svPPA cases. We demonstrate that cases with the clinical phenotype of svPPA may have AV-1451 PET positive scans and tau pathology due to unsuspected AD pathology and draw attention to the consideration of this and other tau pathologies as potential causes of such AV-1451 findings in svPPA [52, 53]. In any case, these data suggest that there is minimal-to-no AV-1451 binding in TDP-43 proteinopathies and that any observed binding is at levels much less than that in AD. The other major group of non-tauopathies is associated with α-synuclein deposition. We found no evidence for AV-1451 binding to regions with dense α-synuclein pathology on adjacent sections in either MSA or LBD.

Off-target AV-1451 binding patterns in the basal ganglia corresponding with Prussian blue staining, (Fig. 9 q, r), led us to postulate that AV-1451 may associate with iron. We also observed high AV-1451 binding in the substantia nigra. Thus, we performed two observational experiments to test this hypothesis: 1) the substantia nigra, another region known to accumulate iron, and 2) a case with brain iron accumulation who was genetically predisposed to abnormally accumulate iron. The strong binding of AV-1451 to the substantia nigra in all cases, including cases with no tau pathology, suggests a potential binding to iron. Iron binds to high affinity sites of neuromelanin to form stable complexes that protect against iron overload [54]. Experimental iron overload of neuromelanin results in low affinity iron to exacerbate neurotoxicity through oxidative processes. Intrinsic melanin binding may be an alternative explanation for substantia nigra AV-1451 signal as postulated by others [19]. Although significant Prussian blue staining was observed in the globus pallidus of the NBIA case, we did not observe a matching pattern of AV-1451 binding. Minimal AV-1451 binding in the red nucleus was seen corresponding to PHF-1 (Fig. 5 g, h, red arrows), which reflects a region susceptible to neuronal loss in CBD and PSP and related tau pathology. This binding was minimal in comparison to AV-1451 binding in the substantia nigra, which could make specific, tau-involved AV1451 signal identification in regions adjacent to the substantia nigra challenging.

Other “off-target” AV-1451 binding sites as seen in Fig. 9 have implications that vary with respect to minor non-specific binding to interfering with interpretation. Potentially the most clinically relevant “off-target” binding regions are the substantia nigra (for confounding of red nucleus tau binding in PSP), the lateral geniculate binding in hippocampal Braak tangle staging, subpial melanin-containing structures, and the choroid plexus or vascular structures when closely associated with the hippocampus or other Braak-important regions. “Off-target” binding in the pituitary and binding to scalp or calcification may be less important. “Off-target” binding may provide clues regarding the molecular target of the AV-1451 ligand. To this end, we sought to bin the “off-target” examples based on their similarities. Within in the pigment-related category, we included neuromelanin-containing neurons, lipofuscin-filled lateral geniculate neurons, and subpial melanin-containing structures. There is evidence to suggest that melanin interacts with amyloidin (functional amyloid) [55], suggesting that perhaps neuromelanin may be recognized based on its structure and not necessarily protein sequence. AV-1451 binding to calcifications in choroid plexus and in areas known to accumulate iron was binned into a mineralization-related category. Contrary to clinical findings suggesting strong AV-1451 signal in choroid, our autopsy study revealed variations – perhaps suggesting there is a dynamic process that we are not able to assess in the postmortem tissue. Inspection of the choroid plexus using thioflavin-S microscopy reveals fluorescent Biondi ring-like tangles [56]. We observed Biondi bodies regardless of the strength of AV-1451 binding, but interestingly they have been found to contain melanin at the ultrastructural level.

In conclusion, we demonstrate the relative specificity of AV-1451 to 3R + 4R tau over other isoforms of tau. We demonstrate minimal binding in association with TDP-43 pathologically involved regions and no association with α-synuclein pathologies. Our findings suggest that multi-modal tau PET and amyloid PET studies may be useful for differentiating atypical AD with non-amnestic presentations from other proteinopathies. These data suggest that patients with strongly positive AV-1451 scans and svPPA could indeed have tau pathology (probably underlying AD-tau) while patients with minimally positive AV-1451 scans more likely have TDP pathology. AV-1451 was found to bind preferentially to NFT compared to neurites. Moreover, AV-1451 had a predilection for mature tangles over extracellular “ghost” tangles. These observations, along with our findings of variable AV-1451 binding in tangle predominant dementia, suggest that variable signal of AV-1451 may be seen with tau disease progression due to variations in tau location and isoforms. A limitation of this work includes the small number of individual cases for each disease. The lack of clinical PET AV-1451 imaging in these cases prevents direct comparisons of the minimal areas of binding and off-target binding with PET imaging which would be needed to make more accurate correlations of the findings.

All participants provided written consent with approval of the Mayo Clinic Foundation and Olmsted Medical Center Institutional Review Boards.

All participants provided written consent to allow publication of anonymized participant data.