Hsp multichaperone complex buffers pathologically modified Tau

Alzheimer’s disease is a neurodegenerative disorder in which misfolding and aggregation of pathologically modified Tau is critical for neuronal dysfunction and degeneration. The two central chaperones Hsp70 and Hsp90 coordinate protein homeostasis, but the nature of the interaction of Tau with the Hsp70/Hsp90 machinery has remained enigmatic. Here we show that Tau is a high-affinity substrate of the human Hsp70/Hsp90 machinery. Complex formation involves extensive intermolecular contacts, blocks Tau aggregation and depends on Tau’s aggregation-prone repeat region. The Hsp90 co-chaperone p23 directly binds Tau and stabilizes the multichaperone/substrate complex, whereas the E3 ubiquitin-protein ligase CHIP efficiently disassembles the machinery targeting Tau to proteasomal degradation. Because phosphorylated Tau binds the Hsp70/Hsp90 machinery but is not recognized by Hsp90 alone, the data establish the Hsp70/Hsp90 multichaperone complex as a critical regulator of Tau in neurodegenerative diseases.


Abstract
Alzheimer's disease is a neurodegenerative disorder in which misfolding and aggregation of pathologically modified Tau is critical for neuronal dysfunction and degeneration. The two central chaperones Hsp70 and Hsp90 coordinate protein homeostasis, but the nature of the interaction of Tau with the Hsp70/Hsp90 machinery has remained enigmatic. Here we show that Tau is a high-affinity substrate of the human Hsp70/Hsp90 machinery. Complex formation involves extensive intermolecular contacts, blocks Tau aggregation and depends on Tau's aggregation-prone repeat region. The Hsp90 co-chaperone p23 directly binds Tau and stabilizes the multichaperone/substrate complex, whereas the E3 ubiquitin-protein ligase CHIP efficiently disassembles the machinery targeting Tau to proteasomal degradation. Because phosphorylated Tau binds the Hsp70/Hsp90 machinery but is not recognized by Hsp90 alone, the data establish the Hsp70/Hsp90 multichaperone complex as a critical regulator of Tau in neurodegenerative diseases.
Tauopathies, among which Alzheimer's disease is the most prevalent form, are a class of devastating neurodegenerative disorders in which the microtubule-associated protein Tau misfolds and aggregates into insoluble deposits in the brain of patients. 1,2 Insoluble deposits of hyperphosphorylated Tau are closely associated with neurodegeneration and cognitive impairment. 3 Multiple lines of evidence link the pathogenic aggregation of Tau to changes in protein homeostasis and in particular to Hsp70 and Hsp90, two central members of the most abundant chaperone family of the heat shock proteins. 4,5 Little is known however about the orchestrated activities of molecular chaperones to counteract misfolding and aggregation of pathologically modified Tau.
Hsp70 and Hsp90 work together to promote the native state of globular clients beyond what can be achieved by the individual chaperones. [6][7][8] The joint action by the Hsp70/Hsp90 chaperone complex termed Hsp70/Hsp90 machinery recovers hormone-receptor interactions, regulates the activity of the tumor suppressor protein p53, enables a significant ATP binding blockage, assembles a functional RISC complex and enhances the kinetics of the hepatitis B virus reverse transcriptase. [9][10][11][12] In addition, the Hsp70/Hsp90 chaperone machinery plays a key role in cancer: 13 tumor Hsp90 is predominantly present in multichaperone complexes with Hsp90 displaying a specific high-affinity conformation for small molecules. 12 In vivo, Tau is post-translationally modified at multiple sites and with multiple modifications critically influencing Tau aggregation and Tau-induced neurodegeneration. 2,14,15 Ample evidence exists that hyperphosphorylated Tau accumulates during the development of Alzheimer's disease. 1 Tau phosphorylated at specific sites is actively investigated as biomarker for characterizing early phases of the disease. 16,17 In addition, Tau acetylation is linked to pathogenic aggregation and neurotoxicity, 18 and the reduction of acetylated Tau is neuroprotective. 19 Post-translational modifications of Tau such as phosphorylation and acetylation might further play important roles in determining different Tau aggregate structures (strains) and thus underlie different tauopathies. 20 Here we show that the Hsp multichaperone complex formed by Hsp70, Hop, Hsp90 and p23 specifically regulates pathologically modified Tau and thus is a key target for fighting Tau misfolding and aggregation.

Tau is a high affinity substrate of the Hsp70/Hsp90 machinery
The core of the Hsp70/Hsp90 chaperone machinery comprises the complex of Hsp70 with Hop and Hsp90 (Fig. 1a). We reconstituted the Hsp70:  Table 1). This is the first evidence that a so called 'client-loading complex', as previously described for globular substrates 21 exists for intrinsically disordered proteins.
The Hsp70:Hop:Hsp90:Tau complex was formed at different absolute protein concentrations ( Supplementary Fig. 1c). Control experiments showed that neither the cochaperone Hsp40, which is known to assist the Hsp70:substrate interaction, 6 nor the addition of nucleotides were required to build up the Hsp70:Hop:Hsp90:Tau complex ( Supplementary   Fig. 1d). From the concentration-dependent intensity increase of the Hsp70:Hop:Hsp90:Tau complex band, we determined the KD = 1.3 ± 0.1 μM (Fig. 1d). The affinity of Tau for the Hsp70:Hop:Hsp90 complex thus exceeds that for binding to the individual chaperones (Fig.   1d).
To define the domains of Tau that bind to the Hsp70/Hsp90 chaperone machinery, we used NMR spectroscopy. 2D 15 N-1 H correlation spectra of Tau's backbone resonances were recorded in the absence and presence of the Hsp70:Hop:Hsp90 complex (Fig. 1e). Increasing concentrations of the Hsp70:Hop:sp90 complex induced changes in the Tau signals. Residues ranging from the proline-rich region P1/P2 to the flanking R' region were strongly broadened along with only few chemical shift perturbations ( Fig. 1f and Supplementary Fig. 1e). Such changes are characteristic for residues interacting in the intermediate-to-slow exchange regime.
To investigate if Tau's repeat region alone is sufficient to evoke the binding to the Hsp70:Hop:Hsp90 complex, we measured the binding of the Tau construct K18, which only contains repeats R1-R4 ( Supplementary Fig. 1f), to the Hsp70/Hsp90 multichaperone complex.
With increasing concentrations of K18 only the level of unbound Hsp70 was reduced, whereas the amount of the Hsp70:Hop:Hsp90 complex remained the same ( Supplementary Fig. 1g).
This suggested that the isolated repeat region of Tau does not evoke an equally strong interaction with the Hsp70/Hsp90 chaperone machinery as full-length Tau.
We then made use of a second, longer Tau construct, termed K32, which comprises the repeats R1-R4 as well as the adjacent regions P2 and R' (Supplementary Fig. 1f). With increasing concentrations of K32, the formation of an Hsp70:Hop:Hsp90:K32 complex was observed ( Supplementary Fig. 1g) along with the reduction of the unbound Hsp70:Hop:Hsp90 complex band, similar to the behavior of full-length Tau (Fig. 1c). Thus, the proline-rich region P2 and the pseudo-repeat R' contribute to a stable Hsp70/Hsp90 chaperone machinery:Tau interaction. In combination, the data show that the central part of Tau is the main interaction domain that associates with the Hsp70:Hop:Hsp90 complex. This interaction domain is predominantly positively charged and contains the hydrophobic repeats of Tau, which play a critical role in the misfolding and pathogenic aggregation of Tau. 2 The co-chaperone p23 stabilizes the Hsp70/Hsp90 machinery:Tau interaction For full functionality of the Hsp70/Hsp90 chaperone machinery, the Hsp90 co-chaperone p23 is required. 22 To determine the importance of p23 for recognition of Tau by the human Hsp70:Hop:Hsp90 complex, we introduced p23 into the assembly reaction ( Supplementary   Fig. 2a-c). In the absence of Tau, the level of the Hsp70:Hop:Hsp90 complex remained unchanged with increasing amounts of p23 ( Fig. 2b and Supplementary Fig. 2b). In contrast, when Tau was present, p23 associated with the Hsp70:Hop:Hsp90:Tau complex as evidenced by the appearance of a high molecular weight band located slightly above the band of the Hsp70:Hop:Hsp90:Tau complex (Fig. 2b). In parallel, the formation of the 5-component Hsp70:Hop:Hsp90:Tau:p23 complex decreased the amount of the ternary Hsp70:Hop:Hsp90 complex (Fig. 2b). The recruitment of p23 into the complex was further supported by the decrease of the amount of free p23 ( Supplementary Fig. 2c, d). Similar to the Tau interaction with the Hsp70:Hop:Hsp90 complex ( Supplementary Fig. 1d), the addition of different nucleotides had no effect on the binding behavior of p23 ( Supplementary Fig. 2e).
To gain insight into the contribution of p23 to the recognition of Tau by the Hsp70/Hsp90 multichaperone complex, we recorded 2D 15 N-1 H NMR correlation spectra of isotopically labeled Tau in the presence of the Hsp70:Hop:Hsp90 complex together with p23 ( Supplementary Fig. 2f). Sequence-specific analysis showed that the peak broadening profiles of Tau bound to the Hsp70/Hsp90 chaperone machinery were highly similar in the absence and presence of p23 (Fig. 2c,  When compared to Tau's interaction profile with Hsp70 and Hsp90 alone (Fig. 2c, lower panel), the same residues of Tau were involved in either interaction. Hsp70 and Hsp90 binding involved four and five major dips/hot spots in the peak intensity profile of Tau, respectively. All five hot spots were also observed in the Hsp70:Hop:Hsp90:Tau and the Hsp70:Hop:Hsp90:Tau:p23 complex. The locally smaller intensity ratios in the presence of p23 (blue bars vs red line in Fig. 2c) suggest that the addition of p23 strengthens the interaction of Tau with the Hsp70/Hsp90 chaperone machinery. NMR experiments of Tau with p23 alone proved that p23 can directly bind to Tau, with most perturbations located in Tau's P2 region

The Hsp70/Hsp90 multichaperone complex blocks Tau aggregation
To investigate if the Hsp70/Hsp90 multichaperone complex regulates the pathogenic aggregation of Tau, we monitored Tau amyloid fibrillization kinetics in the absence and presence of the Hsp70:Hop:Hsp90 complex ( Fig. 2e-g). So far, studies investigating the impact of molecular chaperones on Tau aggregation were complicated by the need of highly negatively charged co-factors to convert Tau into amyloid fibrils. We recently overcame this bottleneck and established a co-factor-free assay in which full-length Tau efficiently aggregates into amyloid fibrils. 23 We performed the co-factor-free aggregation assay using Tau with individual or combined components of the Hsp70/Hsp90 multichaperone complex. We quantified the % delay of Tau aggregation from the midpoint of the growth phase of the Tau fibrillization curves and determined % inhibition of Tau aggregation from the maximal fluorescence of Thioflavin-T (ThT). In the conditions of the co-factor-free aggregation assay, only Tau but none of the chaperones or co-chaperones formed ThT-positive amyloid fibrils ( Supplementary Fig. 2g).
Addition of substoichiometric concentrations of the Hsp70 or Hsp90 delayed Tau aggregation but did not decrease the magnitude of the ThT-signal at the end of the aggregation assay ( Strong inhibition was likewise observed when p23 was present ( Supplementary Fig. 2g-h).

Molecular interactions in the human Tau:Hsp70/Hsp90 chaperone complex
To gain insight into the molecular interactions that stabilize the Hsp70/Hsp90 chaperone machinery:Tau complex, we recorded methyl-transverse relaxation optimized spectroscopy (TROSY) NMR spectra 24 26 or one in which a single Hsp70 molecule is bound to each of the two Hsp90 molecules in the dimer as observed in a recent cryo-EM structure of a client-loading complex containing a Hsp702:Hop1:Hsp902 machinery. 21 Notably, in the same cryo-EM study, a stoichiometry of Hsp701:Hop1:Hsp902 was also observed. 21 In agreement with the high molecular weight of the multichaperone complexes, most Hsp90 signals were broadened beyond detection upon the addition of Tau and p23 (Fig. 4a).
Only the signals of Hsp90's charged linker persisted indicating that this region remains highly mobile within the Hsp70/Hsp90 chaperone machinery (Fig. 4b). For residue I224 located in the charged linker, we observed peak splitting (Fig. 4a) suggesting that the charged linker of interaction demonstrated by NMR (Fig. 2d), the majority of p23 crosslinks were found with the P2 region of Tau (Fig. 4c, d and Supplementary Fig. 3e). Notably, the P2 region was shown to be the additional binding site for Hsp90 that is not involved in Hsp70 binding (Fig. 2c).
To estimate the stoichiometry of the complex of Tau with Hsp70, Hop and Hsp90, we used DSS-crosslinking and density gradient centrifugation followed by size exclusion chromatography (Fig. 4e, f). For the Hsp70:Hop:Hsp90 complex in the absence of Tau, we obtained an elution volume corresponding to a molecular mass of 411 kDa (Fig. 4e, f). In the presence of Tau, however, the estimated mass increased to 739 kDa ( Fig. 4e, f), i.e. much larger than the molecular weights expected for a Hsp702:Hop1:Hsp902:Tau1 complex (425 kDa) and a Hsp701:Hop1:Hsp902:Tau1 complex (354 kDa). However, given the broad nature of the complex peak in SEC it cannot be excluded that multiple stoichiometries ranging from ~400 kDa to ~700 kDa may be formed (corresponding to elution volumes ~11 mL to ~9 mL). This is

Proteins can be guided towards proteasomal degradation via a direct interaction of Hsp70 or
Hsp90 with the E3-ubiquitin ligase CHIP (carboxyl terminus of Hsc70-interacting protein). 28,29 CHIP is the functional counterpart of Hop competing for the binding to Hsp70 and Hsp90 via the TPR domain, and by that ensuring a continuous cycle of machinery buildup and breakdown

The Hsp70/Hsp90 multichaperone complex binds pathologically modified Tau
To gain insight into the molecular underpinnings of the interaction of the Hsp70/Hsp90 multichaperone complex with pathologically modified Tau, we phosphorylated recombinant Tau in vitro by two different kinases (Cdk2 and MARK2) and also prepared separate samples in which the protein was acetylated with either the acetyltransferases CBP or p300. The phosphorylated (PTau) and acetylated Tau (AcTau) proteins remained monomeric in solution as evidenced by hydrodynamic radii in the range from 6.2 -7.0 nm ( Supplementary Fig. 4a). 32 Native PAGE analysis showed that the proteins were heterogeneously modified Cdk2 generates a phosphorylation pattern in Tau that is overlapping with that observed in Alzheimer's disease (Supplementary Table 4, Fig. 6a). 33 In particular, LC-MS/MS confirmed the phosphorylation of epitopes, which are recognized by the monoclonal antibodies AT8, AT180 and PHF1 detecting phosphorylated Tau species in Alzheimer's disease. 34 When we titrated the Hsp70/Hsp90 multichaperone complex with increasing concentrations of PTau Cdk2 , we observed the formation of Hsp70:Hop:Hsp90:PTau Cdk2 and Hsp70:Hop:Hsp90:PTau Cdk2 :p23 complexes comparable to those formed by unmodified Tau (Fig. 6b, and Supplementary Fig. 4b, c). In addition, we found that both PTau Cdk2 and unmodified Tau interact with unbound Hsp70 (Supplementary Fig. 4d). In contrast, PTau Cdk2 displayed a decreased affinity for Hsp90 when compared to unmodified Tau (Fig. 6b, c and Supplementary Fig. 4d). The decreased affinity of PTau Cdk2 for Hsp90 was further supported by NMR spectroscopy of PTau Cdk2 displaying only minor signal perturbations upon Hsp90 addition when compared to unmodified Tau (Fig. 6d), or compared to PTau Cdk2 titrated with Hsp70:Hop:Hsp90 (Fig. 6e). This is also corroborated by ThT-based co-factor-free PTau Cdk2 aggregation assays showing that PTau Cdk2 fibrillization is inhibited more effectively by Hsp70:Hop:Hsp90:p23 than by Hsp90 alone (Fig. 6f-g). Interestingly, Hsp90 alone accelerates PTau Cdk2 fibrillization while keeping a modest inhibitory effect (Fig. 6f-h). The addition of p23 to the Hsp70:Hop:Hsp90 machinery is crucial in enhancing the inhibitory effect on PTau Cdk2 fibrillization ( Fig. 6f-g), but only slightly delays fibril formation (Fig. 6h). Overall, the finding that PTau Cdk2 efficiently interacts with the Hsp70/Hsp90 multichaperone complex, but not with Hsp90 alone (Fig. 6), suggests that the assembly of Hsp90 into the Hsp70/Hsp90 multichaperone complex is important for Tau retention. Hsp90-mediated chaperoning pathways -including Tau degradation induced by the Hsp90/Tau/CHIP interaction (Fig. 5)might be no longer accessible, potentially prohibiting or delaying the degradation of pathologically modified Tau during disease.

Discussion
The combined data establish the Hsp70/Hsp90 multichaperone complex as a critical regulator of Tau homeostasis in neurodegenerative disorders. We reveal that the Hsp70/Hsp90 machinery has a critical holding function for Tau and likely other intrinsically disordered proteins. This holding function is specific for the Hsp70/Hsp90 machinery and does neither depend on energy consumption nor the interaction with nucleotides.
A key finding of our study is that pathologically phosphorylated Tau is efficiently chaperoned by the Hsp70/Hsp90 machinery, but can only weakly bind to Hsp90 alone.
Previously, Hsp90 inhibition as well as p23 silencing have been described to decrease the levels of phosphorylated Tau in vivo, 35 a finding that can now be ascribed to phosphorylated Tau molecules incorporated into the p23-stabilized Hsp70/Hsp90 multichaperone complex.
Thus, despite its intended protective role to inhibit Tau aggregation, the Hsp70/Hsp90 chaperone machinery might simultaneously bear a harmful holding function, thereby fatally increasing the levels of pathologically modified Tau in vivo.
In contrast to previous findings suggesting that Hop and p23 would not be present together in one complex, 36,37 we found that p23 associates with the Hsp70:Hop:Hsp90 complex in the presence of Tau. In addition, we observed enhanced/more stable binding of Tau to the multichaperone complex when p23 was present. Lower amounts of unbound Hsp70:Hop:Hsp90 machinery upon addition of p23 and Tau versus Tau alone were detected by native page and SEC ( Fig. 2b and Fig. 4e). In addition, higher CHIP concentrations were required to dissociate comparable amounts of the Hsp70:Hop:Hsp90:Tau:p23 when compared to the Hsp70:Hop:Hsp90:Tau complex (Fig. 5b). The combined data indicate that p23 strengthens the binding of Tau to the multichaperone complex.
So far, the Hsp70/Hsp90 multichaperone machinery is suggested as a promoter of protein folding and activation. 31 For globular proteins, the chaperones are required to obtain distinct nucleotide states for client processing. 36,38,39 . Our finding that nucleotides are not required for Tau chaperoning is however in agreement with previous results that, for the substrate Tau, the affinity for Hsp90 remained unaffected with or without ATPyS. 40 Moreover, we found that the Tau interaction with the Hsp70/Hsp90 chaperone machinery was independent of the co-chaperone Hsp40 ( Supplementary Fig. 1d), which is known to assist for the Hsp70:substrate interaction by stabilizing the chaperone's ADP-state. 6 Hence, with regard to Tau as an intrinsically disordered protein, the interaction with the Hsp70/Hsp90 chaperone machinery might be because of protein holding rather than protein folding, which in turn does not necessarily require energy. Instead of the event of ATP hydrolysis that is reported to complete the Hsp90 action for other substrates, 8 additional co-chaperones such as Aha1 or PPIases could, similarly to what has been shown with CHIP (Fig. 5), regulate the dynamic assembly and disassembly of the Hsp70/Hsp90 chaperone machinery:Tau complex.
The identification of the Hsp70/Hsp90 multichaperone complex as a critical regulator of Tau homeostasis is consistent with the importance of multichaperone assemblies for the regulation of protein homeostasis in neurodegeneration. Eukaryotic Hsp70 and Hsp90 are known to act through an orchestrated interplay, forming distinct machinery assemblies to accomplish diverse, though target-oriented activities. 41 Accordingly, the synergistic function of multiple chaperones in a disaggregation machinery is required for the disassembly of Tau amyloid fibrils. 42 Moreover, the Hsp70/Hsp90 multichaperone complex may also play a pivotal role in Parkinson's disease given that α-synuclein interacts with multiple chaperones including Hsp70 and Hsp90 in vitro and in mammalian cells. 43 The identification of the Tau:Hsp70/Hsp90 machinery complex provides novel therapeutic opportunities to counteract Tau aggregation and neurodegeneration. Indeed, a distinct Hsp90 conformation with high affinity for small molecules was observed in an Hsp70/Hsp90 multichaperone complex that drives tumor selectivity. 12 Further taking into account that pathologically modified Tau is a high affinity substrate of the Hsp70/Hsp90 chaperone machinery, we suggest that targeting Hsp90 and other components of the Hsp70/Hsp90 multichaperone complex might be a powerful approach to specifically target the most toxic Tau species.

Protein Production
Full-length human DNA of Hsp40 (DNAJB4), Hsp70 (HSPA1A), Hop (STIP1), Hsp90 (HSP90AB1) and p23 (PTGES3) were cloned into the pET28a vector (Novagen) each with an N-terminal The band intensities Ix were quantified with ImageJ v1.52n and v1.53e and normalized using the equation Ix = (Im -Ibg) / Iref , where Im is the mean intensity of the band of interest, Ibg is the mean intensity of the background and Iref is the mean intensity of the reference band.
Errors represent the standard deviation from three independent experiments.

Affinity determination
The affinity of Tau for the Hsp70/Hsp90 chaperone machinery was determined based on the , where ∆ , ∆ and ∆ are the changes in proton, nitrogen and carbon chemical shifts [ppm], respectively. Intensity plots represent the ratio of I/I0, where I is the peak intensity of the titration point and I0 are the corresponding peak intensities in the reference spectra. 48 Error bars were calculated from the signal-to-noise ratio.

Aggregation assay
We followed the Tau aggregation assay as described recently 23 in presence and absence of the crosslinked peptide spectrum matches were taken into account.
Crosslinked sites were plotted as networks with xiView 50 and mapped onto structures using Xlink Analyzer 51 combined with UCSF Chimera. 52 Intramolecular crosslinks were affirmed as valid according to the crosslinker length with a distance threshold set to 29.4 Å for DSS (= 11.4 Å crosslinker arm + 2 x 6 Å for the two lysine side chains + 6 Å accounting for backbone flexibility) and to 17 Å for EDC (= 0 Å crosslinker arm + 6 Å for the two lysine side chain + 5 Å for the carboxylic acid side chain + 6 Å accounting for backbone flexibility). 53

Molecular weight determination
After crosslinking with DSS and density gradient centrifugation, the complexes were analyzed by size exclusion chromatography using an SD200 10 300 column (GE Healthcare) equilibrated  54 and MS/MS spectra were searched against the human Tau protein sequence using the Andromeda search engine. 55 We set carbamidomethyl cysteine as a fixed modification. For variable modification, we used deamidated asparagine, oxidated methionine, and phosphorylated serine, threonine, and tyrosine. We enabled the LFQ and match between runs, while other settings maintained as default. 56 Phosphopeptides that were identified with 1% of false discovery rate (FDR) were further processed with Perseus to filter out decoy peptides and contaminants. 57 Phosphopeptide intensities were log2 transformed and normalized according to the intensity sum of all detected Tau peptides.

Dynamic light scattering (DLS)
Dynamic light scattering (DynaPro NanoStar; Wyatt) was used to determine the hydrodynamic radii of unmodified, acetylated and phosphorylated Tau

Data availability
All MS raw files were deposited to the ProteomeXchange Consortium (www.proteomexchange.org) via the PRIDE1 partner repository with the dataset identifier PXD032037. 58 All PDB codes cited (5fwk, 5aqz, 4po2, 1elw, 1elr, 1ejf) are publicly available in the PDB. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments
We   Supplementary Fig. 2g). g Delay of Tau aggregation by Hsp70 or Hsp90 (* p≤0.033, ** p≤0.0021, *** p≤0.0002, **** p≤0.0001) determined from the ThT curves in panel e. The % delay cannot be accurately calculated for Tau:70Hop90/ Tau:70Hop90p23 because the ThT curves for these systems are almost flat, i.e. not sigmoidal. A lower limit for the % delay in these systems is ~200%, which is the % delay of Tau aggregation in the presence of Hsp70.