A fluorescent marker shows increased uptake of disease-related tau protein by retinal immune cells in Alzheimer's patients and mice

One of the most characteristic changes in Alzheimer’s disease (AD) is the accumulation of aggregated amyloid beta (Aβ), which deposits in the brain as extracellular plaques. The plaques vary in size and morphology and can be classified into at least 12 different types, of which diffuse and dense-core plaques are the most common [1]. A subset of these dense-core plaques contains dystrophic neurites — abnormal and swollen neuronal processes that contain aggregated hyperphosphorylated tau (p-tau). Such plaques, called cored neuritic Aβ plaques, are closely associated with neuronal loss and cognitive decline in AD [2–4] and the presence of neuritic plaques is today required for neuropathological diagnosis of the disease [4]. Intraneuronal accumulation and aggregation of p-tau are also found in neurofibrillary tangles (NFTs) and neurophil threads (NTs). These structures are spread in a specific pattern throughout the AD brain and can also be observed in other neurodegenerative disorders such as progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), and frontotemporal dementia (FTD) [5]. The Aβ and tau pathology in AD is accompanied by neuroinflammation, seen as activation of astrocytes and microglia. The activated glial cells, in particular activated microglia, are often found encircling neuritic plaques and are currently considered to be an additive hallmark valuable to consider at neuropathological evaluations [1, 6]. Since microglia engulf and degrade both Aβ [7] and p-tau [8, 9], it has been suggested that the function of neuritic plaque-associated microglia is to engulf and degrade the plaque-forming Aβ and p-tau-containing dystrophic neurites.

Interestingly, a growing body of research indicates that the hallmark AD changes, including Aβ deposits [10–13] and p-tau accumulation [10, 13–17], are present in the retinas of both AD patients and transgenic mouse models of the disease. Also, activated microglia [13, 18–20] that have engulfed Aβ have been identified, but the number of microglia engaged in Aβ phagocytosis appears to decrease in patients with mild cognitive impairment and AD compared to controls [13], indicating a reduced capability to remove Aβ as the disease progresses. However, it remains unexplored whether retinal microglia also remove p-tau and how this capability is affected in AD. Notably, these retinal alterations are suggested to occur simultaneously or even before the equivalent changes are detected in the brain [12, 14]. This, in combination with the fact that the retina has the same embryological origin as the brain [21] and is peripherally located (and thereby more accessible compared with the brain), makes the retina highly interesting from a diagnostic perspective. Hence, research on non-invasive imaging methods enabling monitoring of retina’s characteristic AD changes is encouraged and ongoing. In the current study, we aim to determine if a thiophene-based ligand called bTVBT2 can be used to capture AD characteristic changes in the retina. Such thiophene-based ligands produce a conformation-dependent fluorescent illumination when they bind to protein aggregate formations [22]. This property becomes useful in neuropathological evaluation of brain tissue, as these ligands surpass traditional immunostaining in terms of revealing and detecting, for example, Aβ and p-tau aggregates. The ligands are relatively small and peripherally administered ligands have been shown to cross the mouse blood–brain barrier (BBB) and bind to aggregates in vivo [22]. Given the resemblance between the BBB and the blood-retina barrier (BRB) [23], it is plausible that the ligands can also reach and bind to amyloid aggregates in the retina. Thiophene-based ligands may thus be of potential use in the development of future retinal imaging methods. However, although there are numerous commercial and non-commercial thiophene-based ligands available, it is important to point out that their binding is dependent on their molecular structure and can therefore vary in terms of specificity. While, for example, pFTAA, an anionic pentameric oligothiophene, binds to several different types of amyloid aggregates (including Aβ, p-tau, and prions) [22, 24–26], the recently developed ligand bTVBT2 appears to specifically bind to aggregated tau pathologies, and not Aβ, found in human AD brain tissue Sects. [27]. This makes the latter ligand interesting from an AD perspective and could be useful when investigating microglial uptake and degradation of p-tau aggregates in retina. In the current study, we investigated if bTVBT2 binds to structures within microglia in the human retina, if the amount of bTVBT2-containing microglia is altered in AD retina compared to non-demented controls, and if the amount correlates with tauopathy in the brain as well as retinal levels of high molecular weight Aβ (representing levels of highly aggregated Aβ). We also investigated if the presence of bTVBT2-bound structures in microglia is related to disease progression by analyzing retinae from the AD knock-in mouse model App^{NL−F/NL−F} at ages before and after the formation of Aβ plaques and dystrophic neurites.

In the current study, we used bTVBT2 to detect p-tau aggregates inside retinal microglia in AD patients and App^{NL−F/NL−F} knock-in mice. The specificity of the ligand was evaluated by staining AD hippocampal samples which showed that, consistent with previous studies [27], bTVBT2 binds to aggregated tau pathologies such as NTs and NFTs, but it is most pronounced in dystrophic neurites. It is interesting to note that staining against different p-tau variants yielded varying percentages of overlap, where the PHF-1 staining gave rise to the highest percentage. This result may indicate that the bTVBT2 binds aggregates in the dystrophic neurites to a higher degree formed by tau phosphorylated at Ser396/404 compared to tau phosphorylated at other sites we examined. Importantly, dystrophic neurites also contain other proteins including the APP and lysosomal proteins [31]. Thus, although we found no overlap between the ligand and Aβ, IAPP, and pTDP-43, we cannot entirely exclude the possibility that bTVBT2 also binds to other amyloid protein aggregates within the dystrophic neurites. Nevertheless, and most importantly for our study, the bTVBT2 inside the AD retinal microglia was positive for both PHF-1 and AT8, two well-known markers for tauopathy in the brain. Surprisingly, no co-localization was found between retinal p-tau181 and bTVBT2. However, although this marker also detects brain tauopathy, recent studies suggest that p-tau181 has a physiological role in the retina [17]. Indeed, our p-tau181 staining yielded a robust immunoreactivity pattern resembling neuronal processes, which supports this idea. These p-tau181 results stand in contrast to the AT8 staining (which showed binding to tau phosphorylated at Ser202 and Thr205), which yielded much less immunoreactivity and were foremost found as aggregates co-localizing with bTVBT2. Hence, AT8 appears to be a more reliant marker for AD-specific tauopathy in the retina, which also has been suggested in recent studies [17].

We further found no co-localization between bTVBT2 and Aβ, IAPP, or pTDP-43, and the percentage of bTVBT2-containing microglia, and the retinal levels of high molecular weight Aβ (theoretically corresponding to aggregated Aβ) did not correlate. Hence, although uptake of Aβ by retinal microglia has been described [13], we conclude that the bTVBT2-positive aggregates taken up by microglia are most likely formed by tau and not by Aβ. This conclusion finds support in previous in vitro and in vivo mouse studies that have shown the internalization of soluble and insoluble tau by microglia [8, 9].

The bTVBT2-positive microglia density in cases with high Aβ load was higher compared to cases with low Aβ load. This finding could be a result of the previously shown increase in p-tau (and thereby the increased need of removal) in the AD retina [10, 13–16]. This increase in p-tau uptake contrasts with the previous study demonstrating reduced microglial Aβ uptake in the AD retina [13], but while the latter is suggested to be due to a reduced capacity to remove Aβ (and thereby contributing to the accumulation and deposition of Aβ in brain and retina), microglial p-tau engulfment may contribute to AD pathology in a different manner. While microglia take up p-tau to degrade it [9], they also re-release p-tau in exosomes [32, 33]. Such exosomes have been found to a greater extent in cerebrospinal fluid and blood of AD patients [33, 34] and it has therefore been proposed that microglia help to propagate tauopathy by engulfing, re-packaging, and releasing tau aggregates [35]. Support for this idea was found in a study demonstrating the impediment of tauopathy after depleting microglia or inhibiting tau microglial release in tauopathy mouse models [32]. Therefore, it may be that the enhanced density of bTVBT2-containing microglia seen in our study contributes to the increased presence of p-tau in the AD retina. It should however be noted that the staining pattern of retinal p-tau in previous studies is rather dispersed and structures such as NTs and NFTs have not been reported [17]. Our bTVBT2 staining also did not capture classical NT or NFT structures, and hence, we speculate that the aggregation of p-tau occurs once it is taken up by microglia.

The increase in density of bTVBT2-containing microglia correlated significantly with Braak NFT stages, indicating that this event co-occurs with enhanced tauopathy in the brain. The correlation was, however, only modest and the density of bTVBT2-containing microglia was not higher in controls (NFT Braak stage 3 and below) compared to AD patients (NFT Braak stage 4 and above). This could be because NFT Braak stages represent stages of NFT spreading, while bTVBT2 foremost captures aggregates in dystrophic neurites (which are foremost found in Aβ plaques). This would explain why a difference in the density of bTVBT2-containing microglia only was detected when the cohort was stratified on Aβ scores. From this perspective, it would be of interest to investigate a potential correlation with a neuropathological evaluation according to CERAD, which scores neuritic plaques throughout the brain [4]. Unfortunately, the collection of retinas was performed before this scoring system was implemented at the NBB and hence, we are unable to investigate such a correlation.

When analyzing the retina of App^{NL−F/NL−F} knock-in mice, we noted that as many as approximately 50% of the retinal microglia contained bTVBT2 at 3 months of age. At this age, neither Aβ plaques nor dystrophic neurites have been formed [30], and since a similar percentage was found in 12- to 18-month-old WT mice, we speculate that the uptake and degradation of tau in mice retina is an ongoing process unrelated to AD pathology. The increase of bTVBT2-containing microglia in 9- and 12-month-old App^{NL−F/NL−F} knock-in mice may reflect, in similarity to the human AD retina, an augmented AD pathology, which in the brain is evident from 6 months of age in these mice.

Besides the limitations mentioned above, there are additional important limitations to point out. One of these limitations is the case-cohort size. Very few brain banks collect and store postmortem eyes in the manner required to perform the flat-mount staining and the most common way to analyze retinas involves cross-sectioning. This method gives many times more sections to analyze, but in turn, it makes it more difficult to retrieve important information such as structural differences and morphological changes. We therefore would like to encourage the collection and storage of retinas in both ways to enable further studies on larger cohorts. Another limitation is the method of retina sample collection. Previous studies on retinal p-tau and Aβ aggregates in AD patients have mostly been conducted on samples from either flat-mount entire human retinas or cross-sections of the nasal-temporal part of the retina [11, 13, 17]. In contrast, our study was performed on flat-mount samples of a part of the inferior nasal retina. Since previous studies suggest that AD pathology may differ between regions of the retina [15], we cannot exclude the possibility that microglia in other parts of the retina may exhibit a different aggregate pattern. Finally, there are several different transgenic AD mouse models available today, but each model mimics only certain aspects of the AD pathology seen in humans. Hence, when choosing a model for our experiments, we needed to consider both the relevance and as well as the strengths and weaknesses of different models. Since the purpose of our transgenic AD mouse study was to investigate if the microglial uptake of p-tau is associated with AD pathology, we needed a model with a slow AD pathology progression. This is to ensure that we could collect fully matured retinas both before and after visible signs of AD pathology. We also needed a model that demonstrates dystrophic neurites, the p-tau structures we know the ligand binds to. Our model of choice, the App^{NL−F/NL−F} knock-in mice, shows both Aβ plaques and dystrophic neurites at 6 months of age, which successively and regardless of sex [36] increases over time much like the AD pathology progression seen in humans. Hence, the model fulfills the two most important criteria for our study. However, the App^{NL−F/NL−F} knock-in mice lack other AD characteristic pathologies including tau tangles (although different p-tau variants are present in both the brain and retina). In addition, only the 3-month-old App^{NL−F/NL−F} knock-in mice group and WT group contained males. Although sex differences in regard of AD pathology have not previously been shown in App^{NL−F/NL−F} knock-in mice [36], studies have shown that microglia density, activation, and antigen presentation (but not phagocytosis) in the brain differs between female and male mice [37]. We can therefore not exclude the possibility that microglial uptake in the retina also is affected by sex. Hence, our results should be interpreted with care and viewed foremost as results supporting the idea that microglia p-tau uptake is increased along with AD pathology.

Our study shows that p-tau aggregates can be found within the retinal microglia in AD patients and App^{NL−F/NL−F} knock-in mice and that the ligand bTVBT2 can effectively detect these aggregates. We also show that the density of bTVBT2-containing microglia increases alongside brain AD pathology and we propose the microglial event as an event implicated in retinal tauopathy. These findings thereby not only support previous findings demonstrating the presence of tauopathy in the AD retina, but further suggest a specific role of microglia, an event which could potentially be monitored using fluorescent ligands and imaging techniques in the future.