Hsp70 Inhibits the Nucleation and Elongation of Tau and Sequesters Tau Aggregates with High Affinity

As a key player of the protein quality control network of the cell, the molecular chaperone Hsp70 inhibits the aggregation of the amyloid protein tau. To date, the mechanism of this inhibition and the tau species targeted by Hsp70 remain unknown. This is partly due to the inherent difficulty of studying amyloid aggregates because of their heterogeneous and transient nature. Here, we used ensemble and single-molecule fluorescence measurements to dissect how Hsp70 counteracts the self-assembly process of the K18 ΔK280 tau variant. We found that Hsp70 blocks the early stages of tau aggregation by suppressing the formation of tau nuclei. Additionally, Hsp70 sequesters oligomers and mature tau fibrils with nanomolar affinity into a protective complex, efficiently neutralizing their ability to damage membranes and seed further tau aggregation. Our results provide novel insights into the molecular mechanisms by which the chaperone Hsp70 counteracts the formation, propagation, and toxicity of tau aggregates.


Table of Contents
Tau was aggregated in the presence of Hsp70 (1:5 Hsp70:tau) for 24 hours and then the sample composition was analysed by TIRF microscopy using ThT to stain beta-sheet rich species. Small ThT-active aggregates detected by this method remain in the supernatant during centrifugation at 16,000 g (20 minutes).

Supporting Figure 2: Fast on-rate of Hsp70 to tau aggregates
To ensure binding of Hsp70 to tau aggregates reached equilibrium conditions by the time of smFRET analysis, the on-rate of Hsp70 to tau was assessed. For this purpose, oligomeric tau was mixed with 1000 nM Hsp70-AF647 and the association was measured by smFRET after indicated periods of time. Complex association had reached maximum values within the experimental dead time (~1 minute), demonstrating fast binding kinetics of Hsp70 to tau oligomers. N=1. The relative toxicity of tau oligomers, untreated (column 1) or incubated with Hsp70 (columns 2 and 3), was tested using a single-vesicle permeabilization assay. Tau was either oligomerized in presence of Hsp70 (column 2), or Hsp70 was added to tau oligomers after the oligomerization process (column 3). Hsp70 neutralizes tau oligomer toxicity by about 50% in either case. Statistical test: One-way ANOVA; *** signifies p ≤ 0.001, **** p ≤ 0.0001.

Testing association and disassociation kinetics
In order to obtain binding saturation curves, first equilibrium or pseudo-equilibrium conditions must be achieved. To ensure the association of Hsp70 to tau aggregates had reached equilibrium conditions by the time of smFRET analysis, the association of Hsp70 to tau aggregates was first measured as a function of time. To perform this experiment, tau-AF488 was first oligomerized for 45 minutes. Note that the fraction of oligomers is around 10% in this sample as determined by smFRET. Then, this oligomeric tau sample was mixed with a high concentration of Hsp70-AF647 (1000 nM) and the association was measured by smFRET after indicated periods of time. Complex association had reached maximum values within the experimental dead time (~1 minute), demonstrating fast binding kinetics of Hsp70 to tau oligomers (see Figure S2). Next, the off-rate of Hsp70 from tau aggregates was assessed. As samples are highly diluted immediately prior to the smFRET analysis (5000-fold), complex dissociation can be a limiting factor for determining binding kinetics. However, association values close to 100% were obtained at high Hsp70 concentrations, indicating that the off-rate of the complex is orders of magnitude lower than the measurement time. Therefore, saturation binding curves were conducted under pseudo-equilibrium conditions.
Fitting kinetic data to obtain relative elongation rates As described in detail previously 1,2 , seeded experiments can be performed to sample the elongation process during aggregation separately from other processes. In the presence of preformed aggregates at high concentrations, the initial aggregation behavior is determined purely by the elongation of existing seeds. Specifically, the initial gradient is given by with M=aggregate mass concentration, k + =elongation rate constant, P 0 = aggregate number concentration and m 0 =initial monomer concentration. At constant monomer concentration m 0 and constant aggregate number concentration P 0 , the initial aggregation gradient can be used directly to extract the elongation rate constant k + . In the seeded experiments performed here, P 0 is constant (as the same seed stock was used), but its values is not known, and therefore relative elongation rate constants ݇ ା ᇱ are shown to allow comparison between the different conditions (Hsp70 concentrations): with n describing the different conditions (1=tau only, etc.). (dM/dt) n was obtained from linear regression of the initial aggregation kinetics (0.5-1.5h of aggregation).

Confocal smFRET data analysis
Studying the influence of Hsp70 on the oligomerisation of K18 ∆K280 tau by smFRET Data analysis. The collected photon traces were analysed using a custom-written Igor Pro 6.22 (Wavemetrics) script as previously described 3 . To analyse the photon time traces the average auto-fluorescence for each channel (0.62 counts/bin for the donor channel and 0.59 counts/bin for the acceptor channel) was subtracted and the data was corrected for the spectral cross-talk from the donor channel into the acceptor channel (2 %). A threshold of 10 counts/bin for the donor and the acceptor channel was used to separate the photon burst-events from background as determined using dual-labelled DNA duplex. The number of donor events and oligomer events recorded per second (r D and r C , respectively) allows the calculation of the association quotient Q representing the fraction of dual-labelled molecules (aggregates) 4 : The ratio of the coincident event rate to the donor burst rate is divided by two to account for the acceptor bursts rate which cannot be measured directly. Therefore, it is assumed to be identical to the donor burst rate. Furthermore, the fraction of oligomers Q was further corrected for the efficiency of detection of coincidence fluorescence previously determined to be 25% using dual-labelled 40 bp-dsDNA 4 . Finally, the Q-value was corrected for the fact that oligomeric species containing either only donor or only acceptor molecules are not detectable by smFRET. Thus, a correction factor derived from the encounter probabilities according to Pascal's triangle was applied [5][6][7] .
The FRET-efficiencies (proximity ratios) are calculated according to where I A and I DA correspond to the acceptor fluorescence intensity and to the donor fluorescence intensity in the presence of acceptor, respectively. γ is an instrument specific correction factor which accounts for different quantum yields and detection efficiencies of the donor and acceptor (0.65 here). FRET efficiencies were calculated for all events and combined in a FRET efficiency histogram 7 .
The approximate number of monomers per oligomer event can be determined based on the average intensity from a blue monomer I Dmonomer according to the following equation Species occupying more than one bin e.g. due to increased size were excluded from analysis as they were assumed to be fibrillar events 7 .

Measuring binding affinities of tau species to Hsp70
The collected photon traces were analysed as described above with the following adaptions. The association Q was calculated from the rate of donor events per second, r d , and the rate of coincident FRET events per second, r c : The Q-values obtained were then plotted as a function of total Hsp70 concentration to obtain saturation binding curves. To correct for ligand depletion and to account for the fact that the association was measured as a function of total Hsp70 concentration rather than the concentration of free Hsp70, fits were performed using the following where A is the total concentration of aggregates, B is the total concentration of Hsp70 and K d is the dissociation constant.
The apparent aggregate size was determined from the average donor monomer intensity, I ୈ ౣౣ౨ , and the sum of the acceptor fluorescence intensity, I A , and to the donor fluorescence intensity in the presence of acceptor, I DA , respectively: As mentioned above, γ is an instrument specific correction factor which accounts for different quantum yields and detection efficiencies of the donor and acceptor (0.65 here). To extract the binding curves of Hsp70 to tau aggregates of different apparent oligomer sizes, donor events were binned into the indicated size ranges and their association was calculated as described above. For the fibrillar sample, a minimum threshold of 10 monomers per aggregates was applied. Binding stoichiometry analysis. Data analysis was performed using ImageJ 9 ; the stacks were first averaged over the 100 frames for each channel and the "Find Maxima" command (based on a plugin contributed by Michael Schmid) was used to detect spots present in the AF405, AF488, and AF647 channels. A standard noise tolerance level was set to differentiate the spots from the background (a change in signal of 100 fluorescent counts was used as a threshold).

TIRF microscope and data analysis
The pixel co-ordinates and intensities for the spots identified in the three channels were imported into Igor Pro (Wavemetrics) for further analysis. Oligomers were first differentiated from monomeric protein by selecting only those puncta in the AF647 channel that also had a corresponding spot in the AF488 channel (within three pixels of each other). The size of the oligomers was approximated according to equation 9: where I G,oligomer and I r,oligomer are the intensities of the coincident spot in the AF488 and AF647 channel, respectively, and I G,monomer and I r,monomer are the mean intensities of the non-coincident spots in the AF488 and AF647 channels, respectively. Similar analysis was then performed between the oligomeric puncta and the AF405 spots to determine which oligomers were bound by Hsp70. The approximate number of Hsp70 molecules present per oligomer was calculated using the following equation: approximate number of HSP70 molecules = ୍ ా,ౙౙ ୍ ా,ౙౙ Where I B,coinc corresponds to the intensity of each coincident spot, and I B,noncoinc is the mean intensity from the noncoincident spots. The stoichiometries were then calculated by dividing the approximate number of Hsp70 molecules by the approximate size of the oligomer.