Functional classification of tauopathy strains reveals the role of protofilament core residues

Abstract

Distinct tau amyloid assemblies underlie diverse tauopathies but defy rapid classification. Cell and animal experiments indicate tau functions as a prion, as different strains propagated in cells cause unique, transmissible neuropathology after inoculation. Strain amplification requires compatibility of the monomer and amyloid template. We used cryo-electron microscopy to study one cell-based yellow fluorescent protein (YFP)-tagged strain, resolving its amyloid nature. We then used sequential alanine (Ala) substitution (scan) within tau repeat domain (RD) to measure incorporation to preexisting tau RD-YFP aggregates. This robustly discriminated strains, defining sequences critical for monomer incorporation. We then created 3R/4R or 4R wild-type RD (amino acids 246 to 408) biosensors. Ala scan of recombinant tau seeds with the Alzheimer's disease (AD) fold matched that of AD homogenate. We scanned 22 brain lysates comprising four tauopathies. This clustered cases by neuropathological syndrome, revealed the role of amino acids in protofilament folds, and allowed strain discrimination based on amino acid requirements for prion replication.

Fig. 1.. DS13 stably propagates amyloids.

(A) Tau RD(LM)-YFP was observed to form stably propagating inclusions in DS13 cells. (B) Low-magnification overview of a single representative grid square imaged using cryo–fluorescence microscopy and (C) cryo-TEM of Tau RD(LM)-YFP inclusions purified from DS13 cells. Notably, the fluorescent signal overlapped with the high-contrast, deposits marked by arrowheads in the cryo-TEM overview. Aggregates were composed of fibrillar assemblies, as seen in (D) a summed projection of 152 central slices from a representative reconstructed cryo–electron tomogram, (inset) zoomed-in view of a single fibril in the tomogram and movie S1. (E) A 2D class average shows the cross-beta repeat of the amyloid fibril core and its corresponding FFT (F) shows a strong peak at 4.7 Å.

Fig. 2.. Incorporation assay overview.

The assay design relies on the expectation that mutating residues in the core of amyloid forming regions of tau will affect aggregation in a strain-dependent manner. The workflow for the incorporation assay consists of transducing cells that stably propagate tau amyloid strains with an arrayed lentiviral library to express various forms of mutant Tau RD(LM)-CFP, and measuring the degree of incorporation of the mutants onto the stable inclusions propagated on Tau RD(LM)-YFP using flow cytometry. a.u., arbitrary unit.

Fig. 3.. Incorporation assay reports a signature of templated aggregation.

(A) Tau RD(LM)-CFP was detectable in DS13 cells at 24 to 48 hours after transduction with lentivirus. Inclusions were apparent at 24 hours and colocalized with endogenous Tau RD(LM)-YFP expressed in cells. Scale bar, 20 μm. (B) Cells that were transduced with Tau RD(LM/2P)-CFP lentivirus lacked inclusions only in the CFP channel. Scale bar, 20 μm. (C) The colocalization of CFP and YFP in inclusions within DS13 cells treated with Tau RD(LM)-CFP lentivirus was detectable via FRET, which increased over time and saturated after 48 hours. Removal of both disease-associated mutations, rendering the WT sequence, also prevented the incorporation into DS13 aggregates. We observed a consistent, intermediate level of FRET when a different disease-associated mutation P301S was used in place of Tau RD(LM)-CFP. (D) Scatter plot of DS13 replicate scans performed 48 hours after transduction with lentivirus. (E) Scatter plot of incorporation assay on DS13 performed at 24 and 48 hours. (F) Line plots of Ala scan performed at 48 and 72 hours of incorporation. Vertical lines demarcate repeat domain boundaries in tau. (G) Line plots of DS19, another strain from our library, and that of DS1 (parental line) seeded with DS19 homogenate before an Ala incorporation scan, were essentially identical. a.u., arbitrary unit. hr, hours.

Fig. 4.. Reliable discrimination of DS strains.

(A) Line plots of dissimilar strains, highlighting usage of diverse repeats in the tau sequence. (B) A heatmap of the incorporation profiles of all DS strains confirms the diversity of strains in our library. (C) Line plots comparing the two most similar Ala scans from our library: DS3 and DS19.

Fig. 5.. Clustering of DS strains based on incorporation assay.

(A) Clustergram-based comparison of all strains in the tau RD(LM)-YFP library. Broadly, there were two very distinct sets of strains; one that encompassed DS7 and DS18 and a second group that encompassed the rest. Within this larger group, there were smaller subclusters that showed very high intra-group correlation, such as DS2, DS3, DS10, and DS19, highlighted in dark green. (B) Clustergram of the correlation coefficients for each pairwise comparison of mutated positions. This was filtered to only show correlations deemed significant (threshold at R = 0.5 based on permutation test). Short range correlations were abundant especially within 274 to 294, 306 to 321, and 339 to 363.

Fig. 6.. A WT-tau incorporation assay reports on amyloid cores.

(A) We adapted the incorporation assay for use with a wider range of samples. WT-Tau RD-Cer/Rub–expressing cells were seeded with test samples to induce aggregation, followed by transduction with a WT-Tau RD-mEOS3.2 lentiviral library. After incubation, flow cytometry was used for readout. (B) Confocal images show Tau RD(3R/4Rext)-Cer/Rub and RD(4Rext)-Cer/Rub seeded with AD brain homogenate. After 48 hours, the cells were transduced with either WT-Tau RD(4Rext)-mEOS3.2 or anti-aggregation RD(4Rext/2P)-mEOS3.2. Aggregation was observed in both cell lines, but only WT-mEOS3.2 colocalized with Cer, not the 2P-mEOS3.2. Scale bar, 20 μm. (C) Tau fibrils were prepared under various conditions, as in Lövestam et al., and detailed in Materials and Methods. The preparations were characterized by cryo-EM and cross sections of the reconstructions are shown. Tau (266 to 391) fibrils resembled filaments “44a” and “45a” from Lövestam et al. [highlighted in purple, using the R2 and part of R3 domains (see also, fig. S9, D and E)]. Tau (287 to 391) fibrils (green) and Tau (297 to 391) fibrils (pink) formed protofilaments with the AD fold, with four and three different arrangements, respectively. (D) Tau RD(4Rext)-Cer/Rub were treated with tau (287 to 391) fibrils and the resulting Ala scan plotted in green. In pink, the incorporation profile of tau (297 to 391) fibrils, which was performed in Tau RD (3R/4Rext)-CFP/Rub biosensors. (E) Tau RD(4Rext)-Cer/Rub was treated with tau (266 to 391) fibrils and the resulting Ala scan plotted in purple. Shown in pink is the incorporation profile of tau (297 to 391) fibrils, which was performed in Tau RD (3R/4Rext)-Cer/Rub biosensors as shown in (D).

Fig. 7.. Recombinant fibrils with the AD fold mimic seeding activity of AD homogenates.

(A) Average WT Ala scan incorporation values from samples that were composed of the AD fold reproduce patterns from AD brain homogenate. The incorporation profile of the in vitro fibrils correlated well (B) to that of AD brain homogenates. (C) Mapping of the incorporation assay values on the AD protofilament from PHF1 illustrates that the strongest hits were in the core of the amyloid, corresponding to residues present in the interacting faces β sheets. Values representing no change in incorporation are in green, while positions in magenta were strongly affected by mutation.

Fig. 8.. Ala scan signature overlaps with amyloid core and identifies tauopathies.

The average incorporation profile with 95% confidence interval of various tauopathy brain homogenates next to their corresponding mapped values on published models of each disease protofilament. (A and E) AD (n = 8) mapped on PDB:5O3L, (B and F) CBD (n = 7) mapped on PDB:6TJX, (C and G) PSP (n = 5) mapped on PDB:7P65, and (D and H) CTE (n = 2) mapped on PDB:6NWP. Green and magenta represent high and low incorporation values, respectively.

Fig. 9.. The incorporation signature clusters individuals by disease and uncovers long-range residue correlations for aggregation.

(A) Clustered heatmap of individual cases naturally subdivides into groups that correspond to neuropathological diagnosis. (B) Correlation matrix of individual replicates from the 23 cases in (A) using values between 274 and 380, along with histograms of correlations between Ala scans shuffled by residue and grouped (or not) by disease. (C) Clustergram-based comparison of the weighted correlation between residues across all treatments reveals modules involved in short and long correlations throughout the repeat domains. Highlighted in red are regions discussed in the main text.