Introduction
Tau is a microtubule-associated protein mainly found in the axonal part of neurons. Resulting from an alternative splicing mechanism, six major isoforms of tau coexist in the human brain with the presence of either 3 or 4 repeated sequences (named below as 3R-tau or 4R-tau) known as the microtubule-binding regions [
8,
28]. The hyperphosphorylation and deposition of tau proteins in insoluble aggregates inside neurons are a hallmark of around 20 pathologies called tauopathies including the well-known Alzheimer’s disease (AD) [
8,
25]. These aggregates progressively invade the whole neuron and form specific intraneuronal lesions named neurofibrillary degeneration (NFD) ultimately leading to cell death. However, the kinetics of this pathological cascade as well as the exact factors leading to cell death are still poorly understood [
61]. Furthermore, even if tauopathies share common features, this group of pathologies is very heterogeneous with a vast variety of clinical presentations including fronto-temporal dementias (e.g. Pick’s disease (PiD)), movement disorders/parkinsonism (e.g. progressive supranuclear palsy (PSP)) or Alzheimer’s type dementia such as AD or argyrophilic grain disease (AGD) [
39]. This heterogeneity may be explained by strong histopathologic differences and differential laminar and regional brain distribution but also by molecular variations such as isoform composition and posttranslational modifications. For example, in AD, the six isoforms of tau co-aggregate, in contrast to pathologies like PSP or AGD, in which only the 4R-tau isoforms aggregate, and Pick’s disease, where the aggregates are only comprised of 3R-tau [
39]. Adding additional complexity, around 50 autosomal dominant mutations in the tau gene (MAPT) have been reported to promote strong tau aggregation and clinically lead to dramatic fronto-temporal lobar degeneration (formerly FTDP-17 now referred to as genetic FTLD-Tau) [
21,
23].
Likely due to the molecular heterogeneity of tauopathies, different morphologies of lesions can be observed, with mainly flame-shaped neurofibrillary tangles in AD, argyrophilic grains and/or glial lesions in AGD or PSP and Pick bodies in Pick’s disease [
39]. These lesions affect different part of the brain and the pathology evolves differently. Thus, histopathological studies in some sporadic tauopathies such as AD [
6,
15,
24], PSP [
66,
69] and AGD [
55] show that, specifically for each disease, tau lesions appear progressively and hierarchically in the brain along anatomical connections. The mechanisms underlying such evolution had remained unexplained for many years and are still poorly understood [
60]. Increasing evidence both in vitro and in vivo, support the ideas that the evolution across brain areas is the result of the active propagation of NFD within the brain. Indeed, our group and others recently showed that tau assemblies are transferred from cell-to-cell and, by being taken up by a second cell, seed the aggregation of endogenous tau leading to the propagation of tau lesions in the brain [
13,
14,
19,
57,
65] reviewed in [
48].
Interestingly, in these studies, 4R-tau human constructs were always used to observe tau propagation/seeding. In human tauopathies, the spatio-temporal evolution of NFD was also only reported in three sporadic tauopathies in which the 4R-tau isoforms aggregate (AD, AGD and PSP). Conversely, it is still controversial whether 3R-tau can propagate in genetic FTLD-Tau (mutant tau) or Pick’s disease (3R-tau) [
33].
Phosphorylation plays essential roles in tau physiology particularly by controlling its binding with microtubules [
5,
49]. In AD, tau hyperphosphorylation leads to the misfolding of tau protein and its oligomerization in highly structured, insoluble aggregates [
3]. Nowadays, this hypothesis has been widened to all human tauopathies. Tau misfolding is thought to be responsible for the seeding propensity of tau that ultimately becomes aggregated and insoluble [
45]. This sequence of event is however not yet completely clear and some intriguing data obtained with a transgenic mouse model overexpressing mutant tau strongly suggest that the appearance of the epitopes of misfolding (particularly with the antibodies Alz50/MC1) precede hyperphosphorylation (particularly with the antibody AT8) [
29]. Interestingly, in genetic FTLD-Tau (mutant tau), tau proteins show conformational changes even without hyperphosphorylation [
35,
46,
67,
68]. In the present study, we re-explore these issues in human neuropathological samples and experimentally in a rat model, to understand how isoforms and mutations influence tau propensity to misfold and propagate from neuron-to-neuron.
We analyzed NFD by immunohistochemistry in different brains areas from genetic FTLD-Tau (3 different mutations) and AD patients (at different Braak stages) using either conformation-dependent or phospho-dependent antibodies. Conformational changes might occur before hyperphosphorylation only in genetic FTLD-Tau patients and not in AD. To further explore these observations, we used a rat model of tauopathies [
10,
19], to examine the pathophysiological propagation of tau using different species, 3R or 4R, mutant or wild-type (WT). As previously described, 4R-tau propagates physiologically and pathologically from neuron-to-neuron to distant brain areas [
19]. Interestingly, when mutant or 3R-tau constructs are used, the transfer of non-pathological species of tau remains functional but tau pathology does not spread and stays in the vicinity of the initiation site. Early conformational changes as indicated by the MC-1 immunoreactivity might facilitate aggregation and neurodegeneration rather than propagation. Therefore, our results suggest that different tau species may encounter different misfolding processes that could explain such differences.
Discussion
The data in the present study reveal new aspects of the propagation of tau proteins by demonstrating that different species of tau have different behaviors in terms of pathological spreading and folding properties. Since the discovery of tau protein as the principal component of NFD [
7,
26], around 20 pathologies involving aggregated tau were described with various time-courses, lesions and involved tau species [
39]. This group includes pathologies with aggregated 3R-tau, 4R-tau, or both isoforms. Rare fronto-temporal dementia cases also involve mutation in the tau gene [
23]. These latter proteins are pro-aggregative [
4,
12] often showing strong tau pathology when overexpressed and therefore are widely used in the modelling of tau pathologies [
18]. The in vivo studies dealing with tau propagation are also mainly based on the overexpression or injection of mutant 4R-tau proteins [
1,
13,
14,
32,
41,
63]. However, in view of the neuroanatomical and biochemical differences among human tauopathies [
2,
4,
22], one may ask if the different tau species trigger the same pathological mechanisms.
Here, we studied the misfolding and hyperphosphorylation status of tau proteins in human brains with AD or with MAPT mutations. In all groups, we found a majority of tangles showing both misfolding and hyperphosphorylation and also a high number of neurons showing only hyperphosphorylation. More interestingly though, the misfolding-only neurons were most prevalent in mutant tau brains (Fig.
1). This result doesn’t show per se that tau misfolding precedes hyperphosphorylation in MAPT-mutant patients and the contrary in AD patients but seems to indicate that the mutant-tau may have different folding properties compared to WT tau.
To model such differences, we took advantage of lentiviral technology to induce the accumulation of 3R or 4R WT or mutant tau, and to investigate the propagation of both the protein and the pathology. We show that regardless of the isoform or mutation, all tau proteins are capable of long-distance propagation through the brain (Fig.
2) consistent with the existence of a cell-to-cell transfer mechanism as previously suggested [
14,
19,
41]. Our study indicates that most of the tau species, travel in a non-pathological, non-phosphorylated, non-misfolded state. Indeed, we clearly see transfer for every species studied (Fig.
2) but only some of them show long distance pathological epitopes presence (Fig.
3). It is clear in this model that at least part of tau cell-to-cell transfer is physiological, as tau does not seem to be either misfolded or hyperphosphorylated. This finding is in line with numerous studies showing that secreted tau is mostly monomeric and non-phosphorylated [
11,
17,
37,
38,
47,
50‐
52,
56,
58,
62,
64,
70].
We also show differences in tau pathology between species of tau in the human brains (Fig.
1). Therefore, we wondered if, in the rat model, the propagation of pathological epitopes is impacted by the species. Indeed, all constructs trigger the development of tau pathology in the hippocampus of rats, but these pathologies evolve in a different manner in the whole brain. 4R-tau leads to a strong long-distance spreading of tau pathology when 3R-tau or mutant-tau mediated-pathologies stay in the vicinity of the pathology initiation site and don’t spread in long-distant brain areas (Fig.
3). These observations using mutant tau confirm our previous data obtained with the P301L mutation [
19], located in the 2nd tau repeat, and extend our conclusions to another genetic FTLD-Tau mutation located in the 3rd tau repeat [
16]. Given the reproducibility of our results among several cohorts of animals, as well as our observations in human brains, such differences between WT-4R-tau and 3R-tau or mutant-tau are likely to be due to intrinsic properties of mutant tau proteins. First, both mutant-tau and 3R-tau are known to induce better fibrillogenesis than WT-4R-tau [
12,
59] probably due to conformational changes in the protein when tau protein is mutated [
2,
20,
22]. For 3R-tau, the presence of a single cysteine in its sequence allows for the formation of intermolecular bridges initiating tau conversion/aggregation. Conversely, the two cysteines (C291, C322) present in 4R tau mostly drive the formation of intramolecular bridges, potentially slowing-down the process of oligomerization and subsequent aggregation [
59].
Indeed, in this study, we suggest that 3R-, 4R- or mutant-tau support different types of pathological conversion. The classical view regarding the pathological conversion of tau proteins from a disordered state to insoluble, ordered and hyperphosphorylated aggregates suggests that tau becomes hyperphosphorylated inducing first the misfolding of the protein, and then its oligomerization. This hypothesis is further supported by the early appearance during the pathology of certain epitopes of phosphorylation [
3,
43]. However, recently, Diamond’s team showed in a transgenic mouse model overexpressing the mutant P301S-tau that the seeding propensity of tau proteins is the first detectable indicator of tau pathology, before misfolding (MC1 antibody) and then hyperphosphorylation (AT8) [
29]. We also previously reported the precocious appearance of tau misfolding epitopes and not hyperphosphorylation when mutant tau was overexpressed (As early as 2 months post lentiviral vectors injection, see [
10]). By contrast, it is obvious that when overexpressing WT tau, hyperphosphorylation occurs first at the initiation site [
10] but also in distant regions ([
19], Fig.
3). Here, we confirm that when mutant-tau proteins accumulate, tau may first acquire misfolding properties (Figs.
1 and
3). This supports the existence of an intrinsic misfolding in mutant-tau proteins leading to the early appearance of a misfolding epitope and to the formation of fibrils with different structures [
2,
20,
22]. Interestingly, the structure of these fibrils is transmissible to other tau species [
20,
22]. This prominent misfolding might be the cause of the higher toxicity and neurodegeneration reported when mutant tau is expressed compared to WT [
10,
30].
The phosphorylation state of tau may be a key player in the propagation processes. We know that tau is retained within axons due to its binding with microtubules which is highly dependent on phosphorylation [
40]. Given that 3R-tau and mutant-tau show weaker binding to microtubules compared to 4R-WT-tau [
9,
16,
27], it is very unlikely for 3R-phospho-tau and mutant-phospho-tau to stay in the axons. We hypothesize that they relocalize to the soma and therefore be less available for trans-synaptic transfer than 4R-tau. These results also suggest the presence of different tau species within the same individual (e.g. non-phosphorylated tau, phosphorylated tau, misfolded tau, truncated tau, dimers, oligomers, polymers, seeding-competent tau) that act differentially for tau transfer and pathological propagation. Most of them can be called “tau pathological species” but it is rather difficult to clearly identify the role of each in the pathology. In further studies, these different tau species should be analyzed independently to understand the part of each in the pathophysiological processes such as tau pathology spreading, misfolding or aggregation.
To conclude, the mechanisms of tau propagation and cell-to-cell transfer are highly dependent on tau species and our study is the first to identify this differential propensity of tau isoforms/mutations to mediate tau pathology spreading. This observation probably relies on intrinsic differences between tau species such as folding. These characteristics are consistent with what is observed in the human tauopathies and could explain their phenotypic specificities. This study also highlights the fundamental difference between tau physiological cell-to-cell transfer and tau pathological propagation. Those two mechanisms probably involve different species of tau that behave differently in the brain. This concept must be carefully taken into account and addressed in further studies dealing with tau propagation.
Acknowledgements
We are grateful from LabEx DISTALZ, Association France Alzheimer, Fondation Plan Alzheimer, ANR Spreadtau, CNRS, DN2M VicTaur for funding this work. We also wanted to thank Lille NeuroBank and the Massachusetts Alzheimer’s Disease Research Center for providing human tissues and Dr. Peter Davis for the gift of Alz50 and MC1 antibodies. Simon Dujardin is supported by a grant from the Alzheimer’s Association (2018-AARF-591935). Massachusetts Alzheimer’s Disease Research Center is supported by a grant from National Institute of Health grant number P50 AG005134.