Signature of an aggregation-prone conformation of tau

The self-assembly of the microtubule associated tau protein into fibrillar cell inclusions is linked to a number of devastating neurodegenerative disorders collectively known as tauopathies. The mechanism by which tau self-assembles into pathological entities is a matter of much debate, largely due to the lack of direct experimental insights into the earliest stages of aggregation. We present pulsed double electron-electron resonance measurements of two key fibril-forming regions of tau, PHF6 and PHF6*, in transient as aggregation happens. By monitoring the end-to-end distance distribution of these segments as a function of aggregation time, we show that the PHF6(*) regions dramatically extend to distances commensurate with extended β-strand structures within the earliest stages of aggregation, well before fibril formation. Combined with simulations, our experiments show that the extended β-strand conformational state of PHF6(*) is readily populated under aggregating conditions, constituting a defining signature of aggregation-prone tau, and as such, a possible target for therapeutic interventions.

It was found that each of the tau peptides and proteins used in this study form fibrillar aggregates in the presence of heparin when aggregated by the addition of heparin at a 4:1 ratio of tau to heparin, which is the typical condition used to induce tau aggregation 1 . Both tau peptides R2/12 G273C/L284C and R2/14 G272C/S285C with the nitroxide spin probe MTSL attached form fibrillar aggregates as shown by TEM in Fig. S1 (a) and (b), respectively. Also, the two tau187 mutants used in this study, G272C/S285C and G303C/S316C, labeled with MTSL are shown to aggregate as shown in Fig. S1 (c) and (d), respectively. It is worth noting that TEM images generated with these two mutants do not always generate neat fibrils but at times clumpy fibrillar aggregates showing a certain level of disorder. However, ThT staining assays (Fig. S3) in addition to CW EPR lineshape analysis (Fig. S4) verify that the majority of tau forms -sheet ordered structures. Therefore, the variants studied here are considered valid model systems of the tau protein.
Supplementary Figure S2. DEER data (17.3 GHz) and reconstructed distances for various magnetic dilutions. The data were fitted using L-curve Tikhonov regularization followed by the method of maximum entropy for the cases of 1:15 (blue), 1:40 (red), and 1:80 (magenta) magnetic dilution of MTSL spin labeled by diamagnetically labeled G272C/S285C variants of tau187.
These produced similar results, differing only in the extent of noise. Each sample was flash frozen 20 minutes after adding the required amount of 11 kDa heparin for a 4:1 ratio of tau:heparin.
Samples with a magnetic dilution of 1:15 were made using 50 M MTSL-labeled and 750 M analogue labeled proteins, 1:40 samples were made using 25 M MTSL and 1 mM analogue labeled proteins, and 1:80 samples were made using 5 M MTSL and 400 M analogue labeled proteins. Fibril formation, as measured by fluorescence spectroscopy of ThT-active fibrils, shows an increase in ThT activity for all mutants over a 12-16 hour period after aggregation is induced by the addition of heparin (see Fig. S3). This provides further evidence that all tau variants used in this study do successfully aggregate and form amyloid fibrils. The trends of ThT activity as a function of aggregation time also show that the formation and maturation of ThT-active -sheet fibrils takes several hours from tau187 proteins, but occurs more rapidly with the 12-or 14-mer tau peptides. However, ThT experiments are conducted at lower tau concentrations compared to the DEER experiment. Therefore, CW-EPR combined with line-shape analysis was completed on tau187 labeled with MTSL at site 322 (Fig. S4). It has been previously shown that parallel - The primary echo decays clearly demonstrate significant shortening of the Tm of tau samples that contain heparin, as compared to tau samples without heparin as seen in Fig. S5. This

Supplementary
indicates that in all cases the presence of heparin leads to a more occluded or partially buried spin-6 label environment 5 , as would be found at inter-or intra-tau interfaces. However, there is no significant decrease in water exposure as probed by the nitroxide moiety tethered to the peptide, as judged by the magnitude of the ESEEM effect. This indicates that the spin labels may experience weak tertiary contacts, but do not become entirely buried. Note that soft pulses of 40 and 80 ns lengths were used to suppress local spin concentration effects to the echo decays 6 Table 1 (presented in the main text). Based on the increase in the content of long distances, manifested as a build-up at the long-distance edge of the distributions, the estimates from Gaussians fittings of DEER distance data strongly support the S-to-S* transition model for both Δtau187 and tau peptides. The Gaussian envelopes present better matches for the DEER data 8 of Δtau187 mutants than the DEER data for tau peptides, whose P(r)'s are more complex, particularly for R2/14 peptides. The use of a harder, 16 ns pump pulse, led to a visibly narrow artifact peak at 2.3 nm superimposed to the DEER-derived P(r) distribution (see Fig. S5). This is a well-known and well-understood artifact peak caused by electron-deuterium nuclear hyperfine couplings leading to electron spin echo envelope modulation (ESEEM) at the 2  suggesting that long-range couplings in fibrils do contribute to non-linear baselines. Hence, the 2nd degree polynomial is a better option to fit the baseline of the time-domain DEER data from tau after heparin addition. However, the difference between these two baseline correction methods is still minor, partly due to a sufficiently long recording time of >3 s, reducing the uncertainty.
We thus reconstructed the distance distribution, P(r), of all heparin-treated samples of Δtau187 (as shown in Fig. S6) by applying a 2 nd order polynomial background fit to their baselines to avoid the destabilizing effect of unwanted intermolecular contributions. For all tau187 samples without heparin and all peptide samples, P(r) was reconstructed after subtracting a linear baseline on a semi-log scale from the time-domain DEER data.

Thiovlavin T (ThT) Fluorescence
ThT fluorescence was measured using a Tecan M220 Infinite Pro plate reader using an excitation wavelength of 450 nm and an emission wavelength of 488 nm. Samples were prepared by mixing 25 M tau with 6.25 M heparin supplemented with 6.7 M ThT.

Continuous-wave EPR and lineshape analysis
CW-EPR measurements were completed on tau187 labeled at site 322C with MTSL at X-band (9.8 GHz) using a Bruker EMX spectrometer with attached dielectric cavity (ER3123D).
A 4 L sample comprised of 800 M tau with 200 M heparin was loaded into a quartz capillary (Vitrocom) and placed inside the EPR cavity. EPR spectra were recorded while using ~6 mW of microwave power, 0.3 G of modulation amplitude and a sweep width of 150 G.
CW-EPR data were simulated using the MultiComponent software of Dr. Christian Altenbach (University of California, Los Angeles). In each spectrum, the A tensors were set to Axx = 6.2, Ayy = 5.9 and Azz = 36.4 while the g values were fixed to gxx = 2.0078, gyy = 2.0058, and gzz = 2.0022. The rotational correlation time of the mobile component was determined from EPR before addition of heparin to the sample. The Heisenberg spin exchange frequency () was allowed to vary in order to obtain a single-line component. Simulation of each EPR spectrum was then completed by fitting the following parameters: rotational correlation time of the immobile and sheet components, order parameter S and the relative populations of each of the three components.