Disassembly of Tau fibrils by the human Hsp70 disaggregation machinery generates small seeding-competent species

The accumulation of amyloid Tau aggregates is implicated in Alzheimer's disease (AD) and other tauopathies. Molecular chaperones are known to maintain protein homeostasis. Here, we show that an ATP-dependent human chaperone system disassembles Tau fibrils in vitro. We found that this function is mediated by the core chaperone HSC70, assisted by specific cochaperones, in particular class B J-domain proteins and a heat shock protein 110 (Hsp110)-type nucleotide exchange factor (NEF). The Hsp70 disaggregation machinery processed recombinant fibrils assembled from all six Tau isoforms as well as Sarkosyl-resistant Tau aggregates extracted from cell cultures and human AD brain tissues, demonstrating the ability of the Hsp70 machinery to recognize a broad range of Tau aggregates. However, the chaperone activity released monomeric and small oligomeric Tau species, which induced the aggregation of self-propagating Tau conformers in a Tau cell culture model. We conclude that the activity of the Hsp70 disaggregation machinery is a double-edged sword, as it eliminates Tau amyloids at the cost of generating new seeds.

The accumulation of amyloid Tau aggregates is implicated in Alzheimer's disease and other Tauopathies. Molecular chaperones are known for their function in maintaining protein homeostasis by preventing the formation or promoting the disaggregation of amorphous and amyloid protein aggregates. Here we show that an ATP-dependent human chaperone system disassembles Tau fibrils in vitro. This function is mediated by the core chaperone Hsc70, assisted by specific cochaperones, in particular class B J-domain proteins and an Hsp110-type NEF.
Recombinant fibrils assembled from all six Tau isoforms as well as Sarkosyl-resistant Tau aggregates extracted from cell culture were processed by the Hsp70 disaggregation machinery, demonstrating the ability of this machinery to recognize a broad range of Tau aggregates. Chaperone treatment released monomeric, and small oligomeric Tau species, which induced the aggregation of self-propagating Tau species in a Tau cell culture model. We infer from these results that the activity of the Hsp70 disaggregation machinery is a double-sided sword as it attempts to eliminate Tau amyloids but with the price of generating new seeds. The Hsp70 disaggregase therefore has a crucial function in the Tau propagation cycle, rendering it a potential drug target in Tauopathies.
Amyloid deposits are characteristic of various neurodegenerative diseases, such as Alzheimer's (AD) and Parkinson's disease (PD). Typically, symptoms surface only at advanced age indicating that a buffering system exists that prevents disease onset and amyloid formation earlier in life (Labbadia and Morimoto, 2015).
Disease-associated proteins aggregate into amyloid fibrils characterized by their highly ordered β -sheet structure (Chiti and Dobson, 2017). The monomeric form of these proteins populates conformations susceptible to aggregation, leading to the formation of a variety of assemblies of various molecular weights. Some have seeding propensities that trigger further aggregation into fibrils by templated incorporation of the monomeric form of the constituting protein in a conformation that is compatible with the fibril ends. This templated propagation of the amyloid structure is thought to be the basis for the prion-like spreading of pathological inclusions and toxicity in neurodegenerative diseases (Brundin et al., 2010).
Aggregation of the microtubule-associated protein tau (MAPT/Tau) is implicated in ~20 different diseases termed Tauopathies, with AD being the most common form of dementia (Spillantini and Goedert, 2013). Tau is thus the most frequently aggregating protein in human neurodegenerative diseases. Under physiological conditions, Tau is highly soluble and consists of six alternatively spliced isoforms (Goedert et al., 1989).
It binds and supports the assembly of microtubules that are vital for axonal transport in neurons (Mandelkow and Mandelkow, 2012). Under pathological conditions, the affinity of Tau to microtubules is reduced, either by disease-associated mutations or hyperphosphorylation (Ballatore et al., 2007;Biernat et al., 1993;Katsinelos et al., 2018). Detached Tau then forms aggregates in the cytoplasm that eventually evolve into fibrillar inclusions in affected neurons or glia cells (Spillantini and Goedert, 2013).
In healthy cells, the homeostasis of Tau and other proteins is tightly controlled by a protein quality control network, including molecular chaperones (Klaips et al., 2018;Labbadia and Morimoto, 2015;Wentink et al., 2019). This quality control system protects the proteome by regulating the synthesis, folding, and trafficking of native proteins to their subcellular destination as well as the refolding and degradation of misfolded species. As such, molecular chaperones act at every step of the amyloid formation and clearance process (Balchin et al., 2016;Kampinga and Craig, 2010;Rosenzweig et al., 2019;Saibil, 2013;Wentink et al., 2019).
Numerous studies have linked molecular chaperone action to Tau aggregation both in vitro and in vivo. It was shown that Hsp70 family members, several J-domain protein co-chaperones, Hsp60, and the small Hsp Hsp27 delay Tau fibril formation (Abisambra et al., 2010;Mok et al., 2018;Patterson et al., 2011;Voss et al., 2012).
Moreover, Hsp70 chaperones interact with oligomeric Tau and prevent further aggregation into fibrils (Patterson et al., 2011). So far, only inefficient disassembly of preformed Tau fibrils by Hsp70 activity has been observed (Patterson et al., 2011;Voss et al., 2012). Our work and other studies demonstrated that Hsp70 disaggregation activity strongly relies on co-chaperone action (Kampinga and Craig, 2010;Rosenzweig et al., 2019 other chaperone systems or handover to the degradation machinery (Bracher and Verghese, 2015). To date, the correct composition of chaperones and co-chaperone combinations that efficiently dissolves Tau fibrils is unknown.
Here we demonstrate that the human Hsp70 disaggregation machinery, referred to hereafter as "Hsp70 disaggregase", has the capacity to disassemble amyloid Tau fibrils in vitro. The Hsp70 disaggregase is an ATP-dependent chaperone system, which is comprised of the constitutively expressed Hsp70 family member Hsc70, the J-domain proteins DnaJB1 or DnaJB4, and Apg2, an Hsp110-type NEF. Both, recombinant fibrils of all six Tau isoforms and Sarkosyl-resistant Tau aggregates extracted from a cell culture model, could be processed by this chaperone system.
We further show that class B J-domain proteins are essential for this activity and that there is partial redundancy within this class of chaperones, while class A J-domain proteins were not able to support Hsp70 disaggregase function. The disaggregation reaction produced monomeric Tau as well as small oligomers. Tau species liberated by the disaggregation reaction were seeding-competent and induced self- propagating Tau species in a biosensor cell line for Tau aggregation.
This study shows that Hsp70 disaggregase activity can be extended to the most prevalent neurodegenerative diseases involving Tau. Although almost completely depolymerized to monomers, the fraction of Tau, which was released from amyloid fibrils by chaperone action, was still seeding-competent. As the generation of seeding-competent species might boost the prion-like propagation of amyloid Tau aggregates, it needs to be examined whether chaperone-mediated Tau disaggregation may exacerbate the associated neurotoxicity in vivo.

The human Hsp70 disaggregase disassembles recombinant Tau fibrils in vitro
To investigate whether Tau fibrils can be disassembled by the human Hsp70 disaggregase, we performed in vitro disaggregation assays (Gao et al., 2015) (Fig.   1A). Recombinant 1N3R Tau, harboring one N-terminal insertion and three microtubule binding repeats (Fig. S1), was assembled into fibrils and fibril formation was verified by negative stain transmission electron microscopy (TEM) (Fig. 1B). Tau fibrils were treated with the human Hsp70 disaggregation machinery (Hsc70, DnaJB1, Apg2) and subsequently centrifuged in order to separate larger fibrils from liberated smaller oligomers and monomers. The amount of Tau in supernatant (S) and pellet (P) fractions was analyzed by SDS-PAGE and immunoblotting (Fig. 1C). In the presence of the three chaperones and ATP more than 40% of 1N3R Tau was detected in the supernatant fraction ( Fig. 1C and 1D). In contrast, in the absence of ATP, the chaperone mix did not confer any significant disaggregation activity ( Fig. 1C and 1D). The three components of the Hsp70 disaggregation machinery were also added individually and in all possible combinations to test their respective contribution to fibril disassembly ( Fig. 1E and 1F). Treatment with single or pairwise combinations of chaperones did not promote a shift of Tau to the supernatant fraction, except Hsc70 together with DnaJB1, which resulted in relocation of ~28% of Tau to the supernatant ( Fig. 1E and 1F). Yet, the combination of all three chaperones was required for a most efficient disaggregation reaction leading to ~43% disassembly ( Fig. 1E and 1F), indicating a critical role of the HSP110-type NEF in Tau fibril disaggregation. In conclusion, the human Hsc70/DnaJB1/Apg2 disaggregation machinery efficiently disassembles a significant fraction of Tau fibrils in vitro.

All six Tau isoforms can be disassembled by the disaggregation machinery
Human Tau has six different isoforms that are generated by alternative splicing (Goedert et al., 1989) (Fig. S1). Whereas all isoforms were found in aggregates isolated from AD patients' brains, there are also isoform-specific Tauopathies where amyloid deposits consist exclusively of either 3R or 4R Tau isoforms (Spillantini and Goedert, 2013 To test whether all six isoforms are substrates for the disaggregation machinery, recombinant fibrils of the other five Tau isoforms (0N3R, 2N3R, 0N4R, 1N4R, and 2N4R) ( Fig. 2A) were subjected to disaggregation and the reaction products were analyzed by differential centrifugation (Fig. 2B). Similar to 1N3R fibrils, fibrils formed by all other Tau isoforms could be disassembled by the human disaggregation machinery, although with varying efficiencies ( Fig. 2B and 2C). In general, all 3R isoforms displayed higher disaggregation rates compared to their 4R counterparts, with 0N3R Tau fibrils being most efficiently disassembled (53%) and 0N4R fibrils being most resistant to chaperone mediated disaggregation (11%). Overall these results show that the Hsp70 disaggregase exhibits disaggregation activity towards all Tau variants and is not limited to fibrils assembled from a certain Tau isoform. aggregates that were formed in cells (Fig. 3A). The Sarkosyl-resistant material was extracted from the seeded cells and subjected to in vitro disaggregation assays. In the presence of Hsc70, DnaJB1, Apg2, and ATP 30% of the TauP301S-Venus was recovered in the supernatant fraction following centrifugation ( Fig. 3B and 3C). This result demonstrates that the human Hsp70 disaggregation machinery is capable of disassembling Sarkosyl-insoluble Tau extracted from a human cell line.

Class B J-domain proteins mediate disaggregation
Hsp70 substrate specificity is mediated by J-domain proteins that recognize chaperone clients and deliver them to Hsp70 (Kampinga and Craig, 2010). Humans encode more than 40 different J-domain proteins, subdivided into structural classes A, B, and C, with distinct substrate specificities and cellular localization (Kampinga and Craig, 2010;Rosenzweig et al., 2019). Several class A and B J-domain proteins are differentially regulated in the brain both during aging and in neurodegenerative diseases (Brehme et al., 2014). In particular, the class A co-chaperone DnaJA1 as well as the class B co-chaperone DnaJB4 are upregulated in patients with AD, PD, and Huntington's disease (HD) compared to age-matched controls (Brehme et al., 2014). These findings point to a potential role of these co-chaperones in regulating proteostasis in the context of protein misfolding diseases.
Therefore, we investigated whether these J-domain proteins could also serve in a complex with Hsc70 and Apg2 to disaggregate Tau fibrils (Fig. 4A)

Tau disaggregation yields monomeric and small oligomeric seedingcompetent species
The products of the disaggregation reaction could consist of multiple protein species, such as monomers, small oligomers and other fibril fragments with intermediate lengths. Furthermore, it is not yet clear whether chaperone-mediated disaggregation of amyloid fibrils is advantageous or disadvantageous. A complete resolubilization and refolding of Tau into monomers is considered beneficial, while the production and accumulation of fibrillar intermediates is considered disadvantageous, as the latter could contribute to the propagation of Tau aggregates through the continuous production of new seeds. Therefore, it is important to analyze the products of the disaggregation reaction more closely.
In order to monitor the disaggregation dynamics, we determined the quantity of To further characterize these low molecular weight Tau species, we next subjected Tau fibrils with and without chaperone treatment to sequential centrifugation ( Fig.   6A). In untreated 1N4R Tau fibril preparations about 6% of Tau was found in the 20 000 g supernatant (Fig. 2B). However, these Tau species sedimented at 337 000 g while the Tau material that was released by disaggregation remained in the supernatant hinting towards a smaller particle size of these species (Fig. 6B). Hence, the 337 000 g supernatant contains Tau species specifically produced by the action of the Hsp70 disaggregation machinery, which are not present without chaperone treatment. Tris-acetate-SDS-PAGE revealed that the disaggregated material contained monomeric as well as oligomeric Tau species with an apparent molecular weight compatible with that of dimeric and tetrameric Tau (Fig. 6C). The latter migrated as distinct bands with apparent molecular weights of 100 kDa and 200 kDa in the gel and could not be dissolved by incubating in 2% SDS at 22 °C or 95 °C (Fig.   6C).
Finally, we evaluated the seeding propensity of this fraction of Tau released by the were not only seeding-competent but were also able to induce self-propagating aggregate species.
In conclusion, Tau fibril disassembly by the human Hsp70 disaggregation machinery did liberate monomeric, as well as dimeric and tetrameric Tau species, which were seeding-competent and induced self-propagating Tau conformers in a HEK293 cell culture model for Tau aggregation.

Discussion
It is well established that the cellular network of molecular chaperones assists in all aspects of protein quality control, from folding of newly synthesized peptides to the disassembly of protein aggregates and degradation of terminally misfolded proteins (Klaips et al., 2018;Labbadia and Morimoto, 2015;Wentink et al., 2019).
Chaperones thereby affect many disease states, which is why chaperone-based therapies could be a promising treatment approach. However, it remains poorly understood to what extent chaperones are capable of disassembling already existing amyloids, given their high thermodynamic stability (Baldwin et al., 2011). Only for α synuclein and HTTExon1Q 48 it has been established that the human Hsp70 machinery is able to disassemble preformed fibrils in vitro (Gao et al., 2015;Scior et al., 2018). Here we investigate the broader role of this machinery in amyloid biology by testing its potential to process aggregates of amyloidogenic Tau isolated from cells or produced in vitro and by characterizing more precisely the products of chaperone-mediated Tau disaggregation. This is particularly important because Tau aggregation is central to the most prevalent human neurodegenerative diseases, including AD, and also plays a role in traumatic brain injuries (McKee et al., 2013;Spillantini and Goedert, 2013).
We show that the Hsp70 disaggregation machinery is capable of disassembling in vitro aggregated Tau amyloid fibrils as well as Sarkosyl-resistant Tau aggregates formed in cells. Tau disaggregation resulted in a rapid accumulation of low molecular weight Tau species ( Fig. 5C and 6C). Further characterization of the liberated Tau pool revealed that it contained mostly monomeric and also some oligomeric species with apparent molecular weights of ~100 kDa and ~200 kDa, compatible with that of dimeric and tetrameric Tau, respectively. Intriguingly, this material was still seedingcompetent as it induced longitudinal self-propagating aggregates of a stably expressed full-length Tau reporter in a HEK293 cell culture model, implying that chaperone-mediated Tau disaggregation is not per se beneficial, but may be involved in the prion-like propagation of Tau pathology.
Amyloid fibrils share a common core structure consisting of a characteristic β -sheet rich conformation (Jahn et al., 2010). Although exhibiting a very similar architecture, fibrils composed of α -synuclein, HTTExon1Q 48 , and Tau will display different surface properties, as they do not share any sequence homology (Knowles et al., 2014;Melki, 2018;Tycko, 2015). Nevertheless, the Hsp70 disaggregation machinery is able to disassemble amyloid fibrils composed of each of these proteins albeit with varying efficiencies, highlighting the versatility of this chaperone system to process various amyloid substrates.
Still, despite overall structural similarity, conformational variations of the amyloid structure formed by a given protein including Tau are known to exist and to affect the pathology of the associated disease (Knowles et al., 2014;Melki, 2018;Tycko, 2015). This variability could make some fibrils more resistant to chaperone action.
We indeed observed differences in disaggregation efficiencies between the six distinct Tau isoforms. Fibrils comprised of 0N4R Tau were most resistant to disaggregation resulting in only 10% disaggregated material compared to up to 64% obtained with the other isoforms. The 0N4R Tau isoform remained also intact after the addition of the Hsp70 disaggregation complex in another study (Mok et al., 2018).
Fibrils assembled from 3R Tau isoforms were disassembled to a greater extent than those made of the 4R isoforms ( Fig. 1D and 2C). In vitro assembled fibrils from 2N3R and 2N4R Tau vary in their architecture (Zhang et al., 2019). The ordered core of 2N3R Tau fibrils comprises the R3 repeat of two parallel Tau molecules, whereas 2N4R Tau fibrils adopt several conformations with a core consisting of R2 and R3 β strands of the same molecule (Zhang et al., 2019). Intriguingly, these differences in fibril architecture may lead to a different stability of 3R and 4R Tau fibrils, which may explain their varying susceptibility to the human Hsp70 disaggregation machinery.
Alternatively, or additionally, the kinetics of chaperone binding may vary between isoforms, leading to different disaggregation efficiencies.
We never observed a disaggregation efficiency greater than 64%. This could be due to a mixture of different Tau conformers in our fibril preparation that is subjected to disaggregation. A subset thereof might be readily disaggregated while other fibril types might be completely resistant. However, this is unlikely as we could not detect the disappearance of a certain type of fibril after chaperone treatment by TEM.
Alternatively, the amyloid equilibrium might prevent disaggregation completion.
Amyloid fibrils exist in equilibrium with monomeric species in a solution with the amyloid state highly favored. Disaggregation produces both monomeric and oligomeric species that have the ability to reaggregate over the course of the disaggregation reaction. The exact percentage of disaggregation likely depends on the kinetics of disaggregation in relation to the kinetics of (re-)aggregation and seeding by disaggregation products, which in turn is dependent on the respective fibril type. This ultimately leads to a conformation-specific disaggregation efficiency until the equilibrium is reached.
J-domain protein co-chaperones are known to confer substrate specificity to the Hsp70 machinery (Kampinga and Craig, 2010). Our data revealed that the class B Jdomain proteins, DnaJB1 and DnaJB4, enabled Hsc70 to disaggregate Tau fibrils. In contrast, both of the two major cytosolic class A J-domain proteins we tested did not mediate Tau disaggregation. This is consistent with our past observations with α synuclein fibrils (Gao et al., 2015).  6D). This implies that low molecular weight species or even monomeric Tau might be seeding-competent, which is in line with a recent study showing that fibril-derived Tau monomers exhibit seeding activity (Mirbaha et al., 2018). Hence, monomeric or small oligomeric Tau liberated from fibrils by chaperone action might still maintain a seeding-competent conformation that is different than naïve monomeric Tau. aggregates, thereby decreasing their amplification and toxicity. Hence, it is tempting to speculate that the chaperone-mediated disaggregation of Tau plays a similar role in the prion-like propagation of Tau pathology throughout the brain.
Chaperone activity is commonly believed to decline during aging and in the context of neurodegenerative diseases. However, this view is too simplified. Rather, a recent study revealed that an imbalance occurs, where individual members are deregulated, with some being up-and others downregulated in the aging or diseased human brain (Brehme et al., 2014). Intriguingly, DnaJB4 gets induced in AD, PD, and HD. Since it promoted disaggregation as efficiently as DnaJB1 in our study (Fig. 4)

Material and Methods
All chemicals were purchased from Sigma-Aldrich or Carl Roth unless stated otherwise.
Monomeric Tau was used as a control in several experiments of this study. Aliquots were stored at -80 °C. In order to remove aggregates that might have formed during the freeze-thaw process, samples were centrifuged at least at 100 000 g immediately before using the supernatant for any experiment.
Subsequently, the tag was cleaved off by Ulp-1 digest and both the His6-Sumo-tag and the His tagged Ulp1 were removed by a second Ni 2+ affinity purification step. The proteins were further purified by size exclusion chromatography on a Superdex200 16/60 column (GE Healthcare).
The human DnaJB4 DNA sequence (The ORFeome Collaboration (2016), DKFZ) was cloned into a pCool6 vector with an N-terminal His6-Sumo tag generated previously (Ho et al., 2019) and expressed at 16 °C. Further purification steps were performed following the protocol stated above. Aliquoted proteins were stored at -80 °C.

ThT disaggregation assay
For kinetic analysis of the disaggregation reaction, 50 µl samples were prepared as described above, but including 20 µM Thioflavin T (ThT

Negative stain electron microscopy
Tau fibrils alone or treated with chaperones in the absence or presence of ATP and ATP regenerating system were diluted in PBS or disaggregation buffer, respectively and pipetted onto carbon-coated copper grids (Plano GmbH). Samples were allowed to absorb for 1 min before washing twice with 10 µl water for 1 min. Negative stain was achieved by incubation with 2% (w/v) aqueous uranyl acetate for 1 min. Excess solution was removed by blotting the grids carefully on filter paper before imaging on an EM-900 or an EM-910 electron microscope (Zeiss) with an accelerating voltage of 80 kV.

Cell culture
The HEK293 cell line expressing 0N4R TauP301S

Generation of enriched seeded pool of TauP301S-Venus HEK293 cells
The naïve TauP301S

In vivo seeding assay
To test the seeding capacity of Tau liberated by the disaggregation machinery, disaggregation reactions were performed as described above. In order to obtain the fraction of Tau that was liberated by chaperone action, differential centrifugation steps first at 20 000 g and then at 337 000 g were performed, both for 30 min, 4 °C.
After ultracentrifugation only the upper two thirds of the supernatant were carefully collected thus avoiding disturbing the pelleted material. Samples of all fractions were subjected to SDS-PAGE and immunoblotting to confirm successful differential centrifugation.
The seeding assay with the biosensor TauP301S In order to monitor the propagation of TauP301S-Venus foci over time, the cells were passaged for 27 days (6 passages) after seeding and imaged regularly with a Leica DM IL LED system equipped with a HI PLAN I Phase 2 40x/0.50 (Leica) objective lens.

Statistical analysis
The statistical analysis was performed using GraphPad Prism (GraphPad Software,       (A) Experimental set-up of the in vivo seeding assay. Differential centrifugation of 20 000 g followed by ultracentrifugation at 337 000 g was applied in order to isolate the Tau material which was liberated by the action of the Hsp70 disaggregation machinery. The 337 000 g S fractions were tested for their seeding capacity in a HEK293 cell culture model for Tau aggregation.
(B) Tau levels in the S and P fractions that were collected during the differential centrifugation steps shown in (A) were analyzed by immunoblotting.