Transmission of Tau Pathology from Human to Rodent Brain: How to Humanise Animal Models for Alzheimer’s Disease Research

Scrapie, a slowly progressive neurologic disease of sheep and goats, was found to be transmissible by injecting healthy animals with extracts from diseased sheep [1]. Ensuing studies on transmissibility of scrapie into mice [2] or of Kuru, a human neurodegenerative disease, into chimpanzees [3-5] strengthened the evidence for the presence of an infectious entity in these disorders. Later, it was shown that infectivity in spongiform encephalopathy, a neurologic disease of cattle, was mainly caused by a proteinaceous agent [5,6]. However, the scientific community initially rejected the possibility of the existence of an infectious agent that is devoid of any genetic information and it was not until 1982 that Prusiner et al. purified and isolated this proteinaceous infectious agent and coined the term “prion” [7].


Spreading of Pathology in Tauopathies -The Inside Story
Misfolded tau protein, a hallmark of human tauopathies, displays prion-like features [31,32]. Tau pathology in AD and some other tauopathies such as argyrophilic grain disease and Pick's disease initially manifests in the entorhinal cortex (EC) and spreads along the hippocampus (HP) to other cortical areas [33][34][35][36][37]. Braak and Braak [33] first showed that in AD there is a characteristic progressive hierarchical distribution of misfolded tau protein in neuroanatomically connected brain areas. More specifically, the initial "pre-tangle" cytoskeletal alterations (i.e., accumulation of hyperphosphorylated tau protein) may spread from the trans-entorhinal region (anteromedial temporal isocortex), the magnocellular nuclei of the basal forebrain and the locus coeruleus to other brain regions [38]. This characteristic spreading is also observed in vivo using Tau-PET scans of AD patients at various Braak stages [39].
In transgenic mice, the tau spreading was successfully recapitulated. The pattern of tau spreading follows a synaptically connected circuit, which is either monosynaptic (as in the hippocampus) or trans-synaptic (as in the cortex) [40]. A transgenic mice model expressing human tau in the EC also recapitulates the trans-synaptic spread of tau pathology [41].
The factors that influence this characteristic spread of tau in the brain are less known. Recently, a brain-wide transcriptomic analysis from healthy individuals revealed an expression signature in regions vulnerable to AD pathology (Braak regions), when compared to non-Braak regions [53]. These regions displayed elevated expression of proteins that co-aggregate with tangles and plaques in AD, such as 14-3-3ζ, 14-3-3Ɛ, Cathepsin D, etc. Moreover, the levels of proteins that prevent aggregation of amyloid and tau, for example, chaperones such as HSPA1A, HSPA1B, HSPA1L and GRP78 were lower in AD-vulnerable regions when compared to non-AD regions. On the other hand, the expression levels of proteins which promote aggregation of tau and amyloid, such as Cdc37, GSK3α, GSK3β and FKBP5 were good predictors of tissue vulnerability to AD and closely mimicked the staging of AD pathogenesis [53]. This study is the first in line to investigate the susceptibility of certain brain regions in correlation with the Braak staging in the brain, necessary raising the question whether transcriptomic levels of certain proteins associated with AD pathogenesis are deregulated, how different the proteomic picture is, especially considering the array of post-translational modifications that are found in AD protein lesions. The post-mortem delay in humans prevents the examination of more delicate pathological features. As a rodent's brain can be preserved within minutes of sacrifice, evaluating transcriptomic and proteomic pattern in animal models recapitulating the spread of tau pathology may provide otherwise unobtainable evidence about region-specific changes in the diseased brain.

Tauons and their Infamous Cousins -The Molecular Abyss
Intrinsically disordered proteins (IDPs) such as tau, β-amyloid, α-synuclein etc. are considered to be key players in the neurodegenerative cascade in human neurodegenerative disorders [54]. These proteins naturally lack characteristic secondary and tertiary structures unless bound to a partner. Another unique property of these proteins is their ability to adopt β-structures which contribute to formation of amyloidlike fibrils [55], alleged to be involved in seeding and aggregation of these proteins [56,57].
Recently it was shown that a major factor that contributes to formation of amyloid-like fibrils is the presence of segments with high fibrillisation propensity (HFP) that form "steric zippers", or self-complementary beta sheets, that serve as the backbone for the nascent aggregates [58]. These regions are particularly enriched with valine (V), isoleucine (I), alanine (A) and serine (S). Aggregation of tau protein into PHF is driven by its transition to β sheet structure [59]. Based on "PASTA 2.0", an advanced algorithm for prediction of amyloid-like structural aggregation depending on aggregation-prone regions (http://protein.bio.unipd.it/pasta2/) [60], we observed that tau protein (2N4R; longest tau isoform, 441aa) has 17.69% of sequences with potential for beta-strand formation (Table 1). Moreover, several HFP sequences are distributed along the length of tau protein, mainly in the microtubule binding repeat domain (MTBR), indicating a propensity for the formation of prion-like motifs in tau protein. For example, aa306 VQIVYK 311 (based on single letter amino acid code) in the third MTBR of tau is identical to a segment globally classified as having high beta-sheet aggregation propensity [58]. This sequence is known to be involved in assembly of PHF [61]. Another HFP sequence, aa274 KVQIIN 279 , is also present in the microtubule binding region of tau. Post-translational modifications lead to the structural changes of HFP segments that may allow oligomerisation of tau molecules. Recently, independent studies on PHFs and straight filaments (SF) from AD conclusively demonstrated that the 3 rd repeat and 4 th repeat binding domains of tau occupy the common core of these filaments, with mixture of 1 st and 2 nd repeats [62,63]. These are arranged as eight beta-sheet structures with N-terminal ends formed by aa306 VQIVYK 311 sequence which complement with aa 373-378 residues. In addition, the purified fraction of sarcosyl-insoluble tau was successful in seeding aggregation in vitro [63]. Put together, this evidence strongly implicates HFP sequences in seeding of aggregation and highlights their potential for propagation of tau pathology.
It is proposed that these disease-modified tau proteins termed "tauons" can undergo self-assembly and aggregate healthy tau protein in a non-saturable fashion and spread through the brain, resulting in infectivity and disease progression [12]. Tau is plagued by a wide range of pathological post-translational modifications, such as truncation, hyper phosphorylation and ubiquitination, which promote structural misfolding of the protein [12,19,20,[64][65][66][67]. Along with the presence of HFP sequences which allow tau's aggregation into beta-sheets this means that pathologically altered protein tau is an elegant agent for initiation and propagation of disease pathogenesis in tauopathies.

Experimental Seeding of Tau Protein Lead to Progression of Pathology to Defined Brain Areas
An early study on transmissibility of AD was performed on hamsters which were inoculated with a buffy coat preparation from AD patients [68]. These animals developed CJD-like histopathological alterations. However, later studies strived to mimic the spreading of tau pathology by direct injection of diseased brain extracts from animal models, AD or other tauopathies [69][70][71][72], mainly into the hippocampus, a region highly vulnerable to tau pathology. The initial study after injection of insoluble tau (hyper phosphorylated tau) from AD brain into wild type rats showed accumulation of amyloid beta, ubiquitin and α1antichymotrypsin along with accumulation of tau filaments in these animals [69]. However, it was uncertain whether insoluble tau alone contributed to the formation of filaments, or whether the accumulation of these proteins was due to the presence of other proteins. The research groups of Boluda et al. were the first who clearly demonstrated the ability of misfolded tau to induce tau neurodegeneration [71]. The study showed that the injection of brain extract from tau P301S mice line into ALZ17 mice which express a single isoform of human wild type tau induced full-blown tau pathology. Prion-like properties of human misfolded tau were observed also after inoculation of purified tau oligomers from human AD into C57BL/6 mice. However, due to the lack of comprehensive immunohistochemical staining this study does not represent the final proof of the infectivity of tau oligomers [27].
Similar to prions, "tauon" strains are specific for different tauopathies [70,72,73]. Misfolded tau isolated from different tauopathies, such as argyrophilic grain disease (AGD), tangle-only dementia (TD), corticobasal degeneration (CBD) and progressive supranuclear palsy (PSP) also induced and drove tau pathology in injected animals [70,72]. The morphology of tau inclusions was diverse and identical to their respective human counterparts. For example, injection of AD or TD material induced the typical neurofibrillary pathology to these diseases [70,72]. On the other hand, CBD tau induced glial tau pathology in the white matter tracts and hippocampus [72]. Inoculation of AGD into ALZ17 mice induced small spherical or comma-shaped argyrophilic grains identical to human AGD [70]. Moreover, seeding of tau from different tauopathies into the PS19 mouse line also affected the specific cell types seen in these human tauopathies i.e., neuronal pathology was initiated by AD extracts and predominant oligodendroglial pathology by material from CBD [72]. Furthermore, injection of material from other tauopathies (AGD and PSP) into ALZ17 mice also resulted in lesions similar to these human pathological diseases. However, formation or propagation of classical Pick bodies was not observed after injection of homogenates from Pick's disease [70,72]. One of the reasons for this selective infectivity could be the absence of a proper "sparring partner" in the inoculated organism, analogous to the species barrier to prion transmission. We showed that injection of brain extract from the transgenic rat SHR24 line, which expresses 3R form of human truncated tau, into the SHR72 transgenic rat line, which expresses 4R human truncated tau, developed tau pathology at the site of injection, but no propagation of pathology to other regions was observed [28].
However, administration of misfolded tau from SHR72 transgenic rats not only induced pathology in cortical neurons in these animals, but also induced neurofibrillary tangles in adjacent brain areas. Since the Alz17 line used by Clavaguera et al. [70] and Boluda et al. [71] mainly expresses human four-repeat tau protein, the three-repeat tau inclusions from PiD lacked the substrate necessary for the propagation of pathology. These results clearly show that pathologically altered tau protein requires a proper substrate in the target organism in order to manifest its infectious and spreading properties i.e., 3R tau or 4R tau for the respective tauopathies. The tau isoforms from diseased human or rodent brains differ in their specificity and infectivity, which accounts for the pathological differences between 3R and 4R tauopathies.
Most studies on the tau spreading phenomenon have been performed on transgenic rodent models; however, few studies have investigated the spreading of misfolded tau in wild type mice. Wild type animals injected with tau extracts from human P301S mice resulted in Gallyas silver and AT100 positive threads and coiled bodies at the site of injection [71], but no NFTs or progression of disease were observed at 6 and 12 months post-injection. In a later study, injection of up to 8 µg of tau protein induced tau pathology, as well as spreading in non-transgenic mice [74]. Furthermore, neuropil threads and oligodendroglial coiled bodies were observed in wild type mice following injection of AD, TD or PSP brain extracts; only astroglial tau inclusions positive for AT8 or AT100 were observed post injection of CBD brain extracts [70]. A recent study clearly demonstrated that by injecting pathological tau extracted from AD, PSP and CBD patients into different brain regions of wild-type mice, endogenous mouse tau aggregates were inducted and propagated [75]. Interestingly, PSP and CBD tau strains induced astroglial and oligodendroglial tau inclusions as was shown by Clavaguera et al. [70]. These findings suggest that higher levels of pathological tau injected into the brain of wild type mice may break the protective barrier of the endogenous mouse tau proteome.
Recently, it was shown that synaptic terminals release tau upon depolarisation, however, the levels of released tau were higher in synaptic terminals from AD brains [91,92]. Calafate et al. [92] demonstrated that hippocampal neurons expressing human truncated tau K18 (aaQ244-E372) with the P301L mutation develop intracellular tau aggregates on exposure to K18P301L fibrils aggregated with synthetic heparin. When cultured in a microfluidic chamber, these cells unstimulated right hemisphere [92,93]. The depolarisation of synaptic terminals from AD brains is enhanced in comparison to control human synapses. Moreover, AD synaptic terminals contain tau oligomers [94], C-terminally truncated tau along with 20 kDa tau fragment [91], which may be released after activation. In addition, tau is reported as a key component of tunnelling nanotubes (TNT) which are tubular structures that facilitate cell-cell communication under stress [95]. In addition, extracellular tau increases the formation of these tunnelling nanotubes, therefore aggravating transfer of tau between neurons [95].
The presence of extracellular tau in the CSF of AD patients and transgenic models strongly advocates that at least a portion of extracellular tau (N terminal fragments) may exit the brain [90,96]. It was recently shown that mTor activity is involved in tau secretion into the extracellular space [97]. On the other hand, tau can be uptake by neuronal and non-neuronal cells. HEK293 cells expressing aggregation prone tau species produced fibrillary tau aggregated which were released into extracellular medium and was seeded in neighbouring cells [77]. Moreover, HEK293 cells expressing human mutant P301L tau develop intracellular tau aggregates and co-culture of these cells with hippocampal neurons induced tau pathology in these neurons [93], indicating a possibility transmission of tau pathology from nonneuronal donor to neuronal cell. Interestingly, the infectivity of prions and α-synuclein between distal non-neuronal cells and the brain has also been reported [98][99][100]. It would be interesting to study whether similar infectivity of tau is observed in vivo in transgenic models or AD.
In an elegant study by Holmes et al. [101], it was shown that the uptake and propagation of tau is mediated by heparin sulphate proteoglycans via micropinocytosis. Increased tau propagation was observed following down regulation of amphiphysin 2 (BIN1), a regulator of clathrin-mediated endocytosis [102]. The sizes of tau fibrils or aggregates also play a crucial role in the process of active endocytosis. Aggregated full-length tau or truncated tau forms are readily endocytosed by neuronal and non-neuronal cells; tau monomers, long fibrils or filaments are not [78,81,87]. Put together, the ability to seed, induce aggregation and propagate tau pathology in specific cells and cell compartments depends on tau conformation, its post-translational modification/s and size and involves a complex of known and unknown secretory and uptake mechanisms.

The Role of Misfolded Tau in the Seeding and Propagation
Various forms of tau protein like fibrils [87,103,104], filaments [28,72,105,106] and oligomers [27] or monomers [107,108] display different seeding properties despite the difference in the origin of the seeding tau protein. Studies also showed differences in propagation (brain areas affected, distance of spreading) after injection of recombinant tau [87,103,104] or native forms [105,107]. Interestingly, tau from human tauopathies brain extracts induced tau pathology mimicking the parent neurodegenerative diseases (Table 2). Moreover, in most of these cases tau pathology in the form of NFTs, NTs and argyrophilic grains were observed. On the contrary, few studies only demonstrated presence of hyper phosphorylated tau inclusions after seeding of tau from human AD without formation of NFTs [27,72,74,106]. In these studies, spreading of tau pathology from site of injection was observed. The only exception was extract from Pick's disease, which induced tau pathology only at the site of injection [70].
Similarly, various forms of pathological tau protein extracted from rodent brain induced tau pathology when injected into the rodent brains ( Table 2). Majority of the studies showed the presence of mature NFTs and NTs in these models [28,71,107]. In a first rat study, we also showed the formation of NFTs in cortical areas and striatum after injection of insoluble tau isolated from two transgenic rat lines expressing human truncated tau [28]. All these studies suggest that the misfolded tau from transgenic rodent models display similar seeding potency as human misfolded tau.
In contrast to tau from rodent and human brain extracts, various forms of synthetic tau oligomers or filaments did not induce NFTs or NTs; however, in all the cases the presence of hyper phosphorylated tau was reported [74,103,104,[108][109][110]. Only one study showed NFT like tau pathology and presence of conformationally modified tau protein in the tau inclusions [103]. Thus, synthetic tau does not fully recapitulate the seeding and spreading potency of human or rodent pathological tau [111][112][113][114].
Finally, Sanders et al. [72] and Kaufman et al. [115] elegantly demonstrated the presence of different tau strains in the brain of human tauopathies. Isolated tau strains from human tauopathies induced different forms of pathological aggregates in HEK293 cells. These aggregates were able to induce AT8 positive tangle-like pathology if they were injected into the mouse brain. Contralateral spreading of tau pathology was also reported in both cases implicating the seeding and propagation potency of various strains of misfolded tau. To sum up, the 'tauons' retain their pathological potential even in cell cultures and successfully seed and transmit pathology when reintroduced in brains.
Although the propagation of tau pathology in the brain is observed, less is known about the intrinsic property of tau protein responsible for its seeding and propagation. Recently it was shown that the conformation of tau protein determines the templating potency and assembly into filaments. Interestingly, in vitro heparin induced synthetic tau aggregates displayed higher stability to guanidium hydrochloride solubility than native sarcosyl-insoluble tau aggregates from TgP301S mice. This was attributed to the difference in conformations between the two preparations [112]. However, the insoluble tau aggregates from TgP301S mice displayed higher potency for seeding when compared to in vitro heparin generated tau aggregates. Several other groups also report a weak seeding potency of soluble tau [71,109], suggesting that insoluble tau predominates as the infectious component of the tau proteome in spreading of tau pathology. Apparently, the misfolded conformation is encoded more stably in insoluble and/or aggregated tau protein and aggregates find it easier to convert and aggregate healthy tau protein than soluble pathological tau molecules do.

Concluding Remarks
Based on evidence from in vitro and in vivo studies, the spread of tau pathology displays the following characteristic features: 1) Pathological misfolded tau from a diseased brain can induce tau pathology in the brain of transgenic and wild type animals 2) Spreading is a dynamic process which is time and dose-dependent and propagates through synaptic connectivity 3) Specific tauopathy brain extracts induced cell tau pathology in injected brains and the induced pathology is analogous in appearance to the parent tauopathy (AD -neuronal lesions, CBD, PSP -astroglial and oligodendroglial lesions etc.) 4) A high amount of tau, of either exogenous or endogenous origin, seems to be essential for induction and spread of tau pathology in mouse models  Brain extracts from rodent models [28] 2 months rat lines SHR72 and SHR24  5) The spreading of tau pathology in the brain of rodent models is a combined process involving injected human/rodent tau and misfolding and aggregation of endogenous tau 6) Most importantly, the induction and propagation of misfolded tau requires the presence of proper substrate, or "sparring partner" for template-mediated conversion, namely 3R or 4R tau for respective tauopathies respectively.

100-175 d PI
In conclusion, the variable infectious potential of tau strains depends on the origin of the seed protein, its post-translational, conformational and morphological state. Moreover, the difference in ability to induce a pathological process confirms the existence of various tau strains or "tauons" in diseased brains [12,113]. Major questions that yet remain to be thoroughly understood are the mode of propagation of pathology and how tau aggregates from one neuron drive aggregation and fibrillation in another neuron. Although some mechanisms of secretion have been identified (synaptic, exosomal, or vesicular) and a few other mechanisms of secretion and spreading have been speculated, a clear picture of the role of various cell compartments and cell types in propagating tau pathology is missing.
Although prion-like transmission of tau protein is widely being investigated in mouse models, some argue that tau protein cannot behave like a genuine prion [114]. This is mainly due to the fact that prion diseases are transmitted from animals to humans or humans to humans. Interestingly, cadaver-derived human growth hormone containing neurodegenerative disease-associated proteins such as tau, Aβ and α-synuclein when injected into humans did not show any risk for Alzheimer's disease or Parkinson's disease in these recipients indicating the lack of transmissivity properties in these extracts [115,116]. This may also be due to the lack of potential seeding strains of tau or other proteins in these extracts or due to the lack of required incubation period for the disorders to manifest and mature. Therefore, elaborate studies are required to thoroughly address the discrepancies in human to human transmission of tauopathies using similar approaches. However, documented in vitro and in vivo evidence on cell lines and rodent models universally supports the seeding and propagation of misfolded tau. Future studies will define the nature of the tau seed in different tauopathies and delineate the various models of tau transmission in the diseased brain.

Using Future Applications of Humanise Models
The usefulness of the above findings becomes clear when one considers the root cause of many development failures in AD drug research: inappropriate models. To illustrate, some drug research in AD is conducted on toxin-based models; in these models, toxic lesions to the hippocampus produce a phenotype dominated by learning disability, overtly AD-like. Yet, the brains of these models contain neither tau nor amyloid pathology and thus are utterly unsuitable for the development of disease-modifying therapies that would halt the progression of these pathologies. In contrast, models based on the spreading of injected material are very faithful to the human disorder: the pathology is induced and driven by genuine pathological tau from human tauopathies, progresses as in humans and the conformationbased nature of tau protein seeds means that the pathology induced in the animals retains important molecular characteristics. Models with such human-like pathology constitute an excellent tool for the study of therapeutic and diagnostic candidates, with increased predictive value for subsequent human studies and therapeutic development. Similarly, generating genuine tau pathology in animals makes them excellent screening platforms for novel tau PET ligands and may be helpful in selection of tau ligands dedicated for various human tauopathies.
A potential shortcoming of spreading-based models is the increased variability, which arises from the interaction of inter-individual variability of the model animals and inter-individual variability of donors of tau seeding material; this has to be overcome through rigorous characterisation of the injected tau seeds and through an increased understanding of what makes a particular tau strain a strong seed.
To sum up, beside the scientific value, the rodent models of humanlike tau spreading open up previously unavailable vistas for the effective preclinical development of novel drugs and diagnostic tools.