Tau protein aggregates inhibit the protein-folding and vesicular trafficking arms of the cellular proteostasis network

Tauopathies are a diverse class of neurodegenerative diseases characterized by the formation of insoluble tau aggregates and the loss of cellular function and neuronal death. Tau inclusions have been shown to contain a number of proteins, including molecular chaperones, but the consequences of these entrapments are not well established. Here, using a human cell system for seeding-dependent tau aggregation, we demonstrate that the molecular chaperones heat-shock cognate 71-kDa protein (HSC70)/heat-shock protein 70 (HSP70), HSP90, and J-domain co-chaperones are sequestered by tau aggregates. By employing single-cell analysis of protein-folding and clathrin-mediated endocytosis, we show that both chaperone-dependent cellular activities are significantly impaired by tau aggregation and can be reversed by treatment with small-molecule regulators of heat-shock transcription factor 1 (HSF1) proteostasis that induce the expression of cytosolic chaperones. These results reveal that the sequestration of cytoplasmic molecular chaperones by tau aggregates interferes with two arms of the proteostasis network, likely having profound negative consequences for cellular function.

The cellular pathology of protein misfolding diseases is initiated by highly aggregation-prone proteins that can convert to oligomeric species with a propensity to form aggregates and amyloids (1)(2)(3)(4). To prevent proteotoxic stress, cells have evolved the proteostasis network (PN), 2 a protein quality control system that regulates and balances protein synthesis, folding, transport, and degradation (5)(6)(7). The robustness of the PN determines and maintains a balanced proteome against chal-lenges by exposure to cell stress conditions, the chronic expression of highly aggregation-prone proteins, and during aging (8,9). Nonproductive interactions between protein aggregates and the protein quality control machinery impairs PN function and cellular health (10 -15). A feature observed in many diseases of protein aggregation is the deleterious effect on different arms of the PN (13, 16 -22). For example, in cells expressing either mutant huntingtin, ataxin-1, or SOD-1, the respective protein aggregates have been shown to sequester the chaperone HSC70 resulting in inhibition of clathrin-mediated endocytosis (CME) (23). This suggests that the collapse of the cellular PN occurs when chaperones such as HSC70 are sequestered by these protein aggregates and interfere with chaperone-dependent cellular activities.
Our study examines whether tau likewise interferes with the functional properties of molecular chaperones. Tau is an intrinsically disordered protein that upon mutation and post-translational modifications can adopt alternative non-native states that are highly aggregation-prone (24 -29). Tau aggregation is associated with ϳ20 neurodegenerative diseases, including Alzheimer's disease (AD), frontal temporal dementia with parkinsonism linked to chromosome 17, and Pick's disease (30 -32). Hyperphosphorylated tau can form soluble oligomers and intracellular fibrillar inclusions known as neurofibrillary tangles (NFTs), which contain paired helical filaments of tau (25,33). Tau can generate an altered conformational states that can be propagated by seed-dependent recruitment of native tau (34,35). Tau pathology in AD has been shown to spread by cell-tocell transmission of insoluble tau seeds as well as by amplification of tau inclusions upon seeding between brain regions (34 -43).
At the cellular level, tau aggregation is affected by cytoplasmic chaperones, including HSC70, HSP90, and HSP27 (44 -51). Elevated levels of HSC70 suppress the formation of NFTs by redirecting tau into productive folding pathways to prevent aggregation in tissue culture cells (44). In vitro studies have demonstrated that tau isoforms can interact with a subset of the chaperone network to assist proper folding and prevent misfolding and aggregation (50,(52)(53)(54). The interactions between chaperones with tau aggregates have also been observed in brain tissues from AD patients (44) and are supported by proteomic analysis of NFTs from tissue samples (55,56).
Tau, as for many other highly aggregation-prone proteins in neurodegenerative diseases and other protein conformational diseases, would therefore be predicted to interfere with many cellular processes. To directly examine how sequestration of key PN components by tau aggregates compromises the cellular capacity to detect, respond, and protect against proteotoxicity, we have employed single-cell-based cellular assays that monitor the functional properties of the PN for protein folding and clathrin-mediated vesicular trafficking using the well-established human HEK293 cell model for tauopathy (57). This cellbased tau aggregation model depends upon seeding with tau fibrils and amplification in vivo and provides an opportunity to directly compare different populations of tau-aggregating species within the same experiment as determined by the recognition properties of the well-characterized antibodies MC1 and PG5 (58 -64). PG5 recognizes phosphorylated serine 409 in tau and reacts specifically with the neurofibrillary tangles in AD but not normal brains, whereas MC1 antibody detects an early marker of tau pathology that corresponds to disease-specific conformational modifications of tau. Together, these reagents provide a single-cell approach to examine the interaction of different tau aggregate species with cytoplasmic molecular chaperones and the consequence of these interactions on the functionality of specific arms of the proteostasis network. Our results reveal that tau aggregation inhibits the chaperone-regulated PN processes of clathrin-mediated vesicular trafficking and protein folding in the cytoplasm that can be prevented and/or restored by small molecules that induce the expression of these cytoplasmic chaperones.

Sequestration of molecular chaperones by tau aggregates
The HEK293 cell system was used for tau seeding amplification because of their low endogenous expression of various tau isoforms, thus rending these cells ideal for the conditional expression of tau (57). Protein quantitation with total tau antibody confirmed that the basal level of tau in HEK293 cells prior to tetracycline induction is 0.70 ng/g total protein, which upon conditional expression of mutant tau P301S achieves a steady-state level of 13.8 ng/g total protein in HEK293 tau P301S cells (Fig. S1, a-f, and quantified in g). By comparison, induced pluripotent stem cell-derived neurons express tau at 1.5 ng/g total protein, and rat cortical neurons express tau at 4.35 ng/g total protein, respectively. The HEK293 cell model for tau seeding and amplification is a two-component system in which the appearance of tau inclusions is absolutely dependent upon the conditional expression of mutant 1N4R tau and simultaneous exposure to Sarkosyl-insoluble tau seeds derived from the brains of TgP301S mice. Tau inclusions were detected using MC1 antibody that has been shown to recognize conformationally altered tau (Fig. S1m, arrowheads) (58 -60) and PG5 antibody that recognize late-stage hyperphosphorylated tau aggregates (Fig. S1m, arrows) (61,62). Neither expression of mutant, WT tau, nor seeding alone resulted in detectable aggregation phenotypes (Fig. S1, h-l). PG5-and MC1-positive cells were observed only following incubation with Sarkosyl-insoluble tau seeds and expression of mutant tau P301S, such that at day 3 post seeding, 6.1% of the tau-expressing cells had aggregates positive for MC1 and 2.6% were positive for both MC1 and PG5 (Fig. S1n).
To examine the effects of tau aggregation on the subcellular localization of molecular chaperones, we examined individual cells by immunostaining (Figs. 1 and 2) with antibodies specific to components of the cytoplasmic Hsp70 machine (HSC70, HSP70, DNAJB1, and HSP110), HSP90 and HSP27. In cells expressing MC1-positive dispersed and diffused tau, the subcellular localization of HSC70 was unaffected (Fig. 1A, panel c, arrowheads), whereas HSC70 localization was altered in PG5positive cells and co-localized with PG5 tau aggregates (Fig. 1A, panels a and c, arrows, overlay enlarged in panel e) that were also MC1-positive (Fig. 1A, panel b, arrows). Inducible HSP70 also co-localized with MC1-and PG5-positive tau aggregates (Fig. 1A, panels h and l, arrows, overlay enlarged in panels j and n) and was associated with elevated levels of HSP70. We quantified the co-localization between different tau conformational species and HSC70 or HSP70 using Pearson correlation coefficient (PCC) (Fig. 1B). For example, PG5-positive tau inclusions and HSC70 exhibited a PCC value of 0.66 indicating a high correlation of co-localization, whereas a lower PCC value of 0.34 was observed for MC1-positive tau species and HSC70. Similarly, using a different antibody that recognizes both HSC70/HSP70, we observed a strong co-localization correlation between PG5-positive tau inclusions and (PCC value of 0.67) (Fig. 1B) relative to MC1 (PCC of 0.07). Similar patterns of co-localization between PG5 tau inclusions and inducible HSP70 (anti HSP70 antibody, clone 4G4) was observed with a PCC value of 0.62.
These observations suggest that MC1-positive tau interacts transiently with HSC70 or HSP70 with a much lower level of sequestration than observed for the PG5/MC1-positive tau inclusions that form stable sequestered interactions with these chaperones. This also suggests that stable interactions with chaperones occurs in the transition between conformationally altered diffuse tau detected by MC1 antibody and the PG5/ MC1-specific tau inclusion species.
In the case of constitutively expressed HSC70, we observed a 40% increase in the levels of this chaperone in MC1-positive cells and a 9% decrease in PG5/MC1-positive cells (p Ͻ 0.1) (Fig. S2A). When using an antibody that specifically recognizes HSP70, we observed a 2-fold increase in HSP70 content in PG5/ MC1 cells compared with negative control cells (Fig. S2B). These results indicate that the expression of inducible HSP70 occurs after the appearance of PG5 tau inclusions.
The effects of tau on the subcellular localization of other cytoplasmic molecular chaperones were also examined in PG5/ MC1-and MC1-positive cells ( Fig. 2A) for DNAJB1 ( Fig. 2A, panels a-e), HSP90␤ ( Fig. 2A, panels f-j), and HSP110 ( Fig. 2A, panels k-o). All three chaperones are sequestered in tau aggregates in PG5/MC1-positive cells ( Fig. 2A, panels c, h, and m, arrows; merged display between PG5 tau aggregates and chaperones in the boxed area enlarged in Fig. 2A, panels e, j, and o) and not altered in MC1-only cells ( Fig. 2A, panels c, h, and m, arrowheads). We then examined the subcellular localization of the inducible chaperones HSP27 and HSP90␣ upon tau aggregation (Fig. 2B). Both HSP27 (Fig. 2B, panels p-t) and HSP90␣ Collapse of proteostasis by tau aggregation (Fig. 2B, panels u-y) were likewise preferentially sequestered by tau aggregates only in PG5/MC1-positive cells (Fig. 2B, panels r and w, arrows, overlay enlarged in panels t and y), with no obvious change in protein expression or subcellular localization detected in the MC1-only cells (Fig. 2B, panels r and w, arrowheads).
Our analysis of chaperone interactions with tau inclusions indicates that multiple classes of constitutive and inducible cytoplasmic chaperones are sequestered by PG5/MC1 inclusions but to a much lesser extent with MC1-only positive inclusions. The balance between chaperone expression and their interactions with misfolded and/or aggregated tau species therefore determines the available pool of chaperones for interaction with other protein clients. Recognizing that chaperones are essential to the functional properties of the PN, we next examined the consequence of sequestration of chaperones on cellular function.

Inhibition of clathrin-mediated endocytosis in cells expressing tau aggregates
HSC70-dependent CME was monitored using a single-cellbased assay that quantifies transferrin uptake by the transferrin receptor as a functional measure of CME. The seeding-depen-Figure 1. Tau inclusions sequester major cytosolic molecular chaperones of the HSP70 family. HEK293 tau P301S cells were induced to express tau P301S mutant protein and incubated with insoluble tau seeds. Tau inclusions were detected by immunostaining cells with PG5 (panels a, f, and k) and MC1 antibodies (panels b, g, and l) 3 days post seeding. A, molecular chaperones in the cytosol were detected by immunofluorescence microscopy. The cellular localization of chaperones HSC70 (panels a-e) and inducible HSP70 (panels f-j, anti-HSP70/HSC70 antibody; k-n, anti-HSP70 antibody, clone 4G4) was examined in three cell populations: negative cells, MC1-positive (arrowheads) and PG5/MC1-positive (arrows) cells. The anti HSP70 antibody 4G4 is not compatible with MC1, and thus co-staining was conducted only with PG5 and 4G4 antibodies. Constitutively expressed HSC70 and inducible HSP70 were sequestered to the PG5/MC1positive tau inclusions (panels c, h, and l, arrows), whereas sequestration was not observed in cells with dispersed/diffuse MC1-positive tau (panels c, h, and l, arrowheads). The boxed areas (panels c, h, and l) are enlarged to highlight the co-localization between PG5-positive tau aggregates and chaperones that appear yellow in the overlay (panels e, j, and n). Images shown are maximal projections of a z-series of fluorescence confocal slices through the entire cell volume (scale bar, 10 m) and are representative of three independent experiments. B, co-localization of tau aggregates with chaperone indicated was quantified using PCC. Individual cells were selected as regions of interest and analyzed using the Coloc2 plugin of the Fiji ImageJ software. Pearson's R value (above threshold) was measured per cell (n Ն25) and dot-plotted with mean and S.D. p values were determined using a two-tailed unpaired Student's t test; ****, p Ͻ 0.0001.

Collapse of proteostasis by tau aggregation
dent amplification of the tau aggregate assay provided a number of important controls indicating that tetracycline-induced expression of either WT (Fig. S3, b and e) or mutant tau (Fig. S3, c and f) does not affect the kinetics of transferrin uptake compared with mock-treated cells (Fig. S3, a and d). Likewise, that incubation with insoluble tau seeds without conditional expression of mutant tau P301S or with WT tau expression also had no effect on CME (Fig. S3, and quantified in g).
The effect of tau aggregation on CME was demonstrated in tau P301S-expressing cells following seeding. PG5/MC1positive cells (Fig. 3a, arrows) and MC1-positive cells (Fig. 3b, arrowheads) both showed reduced levels of transferrin uptake (Fig. 3c). In MC1 cells, transferrin uptake declined by 60% compared with control cells, and this corresponded to a 49% decline in PG5/MC1-positive cells (Fig. 3i). As expected, cells that were negative for both PG5 and MC1 showed the same level of A, cellular localization of other major constitutively expressed chaperones and co-chaperones such as DNAJB1 (panels a-e), HSP90␤ (panels f-j), and HSP110 (panels k-o) was examined. B, cellular localization of the inducible chaperones HSP27 (panels p-t) and HSP90␣ (panels u-y) was examined. Cells exhibiting only MC1 staining do not exhibit sequestration of chaperones (arrowheads), whereas constitutive and inducible chaperones are sequestered by tau inclusions (arrows). The overlay between tau aggregates and chaperones in the boxed area is enlarged and displays co-localization as a yellow colored overlay (panels e, j, o, t, and y). Images shown are representative of three independent experiments. Scale bar, 10 m.

Collapse of proteostasis by tau aggregation
transferrin uptake as untreated control cells (Fig. S3g, P301Stet-S group). This suggests that MC1 inclusions inhibit CME to a level similar to that in PG5-positive cells despite the lower levels of chaperone interaction. The inhibition of CME in PG5/ MC1 cells is associated with the appearance of tau aggregates as the levels of total tau remained unaltered from PG5-negative control cells (Fig. 3, e-h) as detected by the anti-total Tau (clone HT7) antibody (Fig. 3f, arrows). Furthermore, to demonstrate that CME inhibition was not a result of fluctuation in the levels of cell-surface-localized transferrin receptors, we performed transferrin-loading assays at 4°C to show that equivalent levels of transferrin binding to surface receptors were detected in PG5/MC1-positive cells and MC1-positive cells relative to negative control cells (Fig. S3, j and n).

Effects of tau aggregation on cytoplasmic protein folding
Molecular chaperones have a central role in regulating protein conformation by co-and post-translation folding to the native state and a holding activity that maintains unfolded intermediates on-pathway to their functional folded states (65)(66)(67). To further demonstrate how tau-mediated sequestration of chaperones affects proteostasis in the cytoplasm, we examined the protein-folding arm of the PN using a folding sensor that had been developed using firefly luciferase whose folding in vivo was known to be dependent upon HSC70 and HSP90 (67,68). We therefore employed the metastable protein-folding sensor EGFP-firefly luciferase DM (fLucDM-EGFP) mutant (69) delivered by lentivirus that was conditionally expressed with tau P301S protein (Fig. S4). Whereas the EGFP protein (Fig. 4A, panels a-d) does not co-localize with PG5/MC1 tau inclusions (Fig. 4, A, panel c, arrows, and B, panels i and j), the fLucDM-EGFP folding sensor (Fig. 4A, panels e-h) co-localized with PG5/MC1-positive inclusions (Fig. 4, A, panel g, arrows, and B, panels k and l). In contrast, sequestration or aggregation of the fLucDM-EGFP-folding sensor in MC1-positive cells was not observed (Fig. 4A, panel g, arrowheads). The co-localization between PG5 tau inclusions and fLucDM-EGFP gave a PCC value of 0.61 indicating strong level of interaction (Fig. 4, B, panels k and l, and quantified in C) relative to the PCC value of Ϫ0.32 for PG5-positive tau inclusions and EGFP protein (Fig. 4,  B, panels i and j, and C). MC1-positive tau species also showed poor correlation with the folding sensor (PCC value of 0.06). These results suggest that changes in tau conformation detected by MC1 and inhibition of CME precedes the inhibitory effects of tau aggregates (PG5/MC1-positive) on protein folding.

Enhancing chaperone levels reverses the tau inhibitory effects on clathrin-mediated endocytosis
If the sequestration of cytoplasmic chaperones by tau aggregates interferes with chaperone-dependent cellular activities, this would suggest that these cellular activities could be restored by enhancing the levels of these chaperones. We therefore examined the effects of small-molecule proteostasis regu-

Collapse of proteostasis by tau aggregation
lators that activate HSF1 and consequently the expression of cytosolic chaperones (70). Three small molecules (A3, C1, and F1) previously shown to activate HSF1 (70) and the expression of HSC70/HSP70 in HEK293 tau P301S cells by 2-3-fold (Fig.  S5, a-d, and quantified in e) were employed. Using a protocol in which tau P301S cells were seeded to induce tau aggregation for 2 days followed by addition of either A3, C1, or F1 for 24 h, we examined three cell populations (negative, PG5/MC1positive, and MC1-positive only) (Fig. 5, e-p) for transferrin uptake relative to vehicle-alone-treated cells (Fig. 5, a-d). All three proteostasis regulators either partially or completely restored CME in the PG5/MC1 cells and MC1 cells (Fig. 5, g, k  and o, arrows and arrowheads, compared with c, and quantified in q). It was notable that although CME was restored, the level of tau aggregates was not notably affected as measured by the number of PG5/MC1-positive cells (Fig. S5f, 24 h). However, by extending the treatment period of F1 to 48 h, we observed a 54% reduction in tau aggregate-containing cells (Fig. S5f, 48 h).
These results indicate that enhancing the levels of chaperones in this protocol can prevent the decline in chaperone-dependent activities and thus alleviate cellular toxicity from tau aggregation.
As a complement to small-molecule proteostasis regulators, we employed a conditional genetic switch to increase the levels of HSC70. HEK293 tau P301S cells stably expressing V5-HSC70 under the control of the cumate conditional promoter A, HEK293 tau P301S cells were infected with lentivirus-encoding EGFP (panels a-d) or firefly luciferase DM-EGFP-folding sensor (panels e-h). Cells were induced with tetracycline to express both tau P301S and fLucDM-EGFP proteins and subsequently incubated with Sarkosyl-insoluble tau seeds. Three days following seeding, cells were immunostained to identify MC1-(panels b and f, arrowheads) and PG5/MC1 (panels a and e, arrows)-positive cells and to examine for the subcellular localization of the luciferase-folding sensor (panels c and g). PG5-positive tau inclusions sequester fLucDM-EGFP (panel g, arrows) but not EGFP proteins (panel c, arrows) indicating the changing cellular folding capacity for conformationally challenged proteins. Scale bar, 10 m. B, PG5-positive tau inclusions sequester fLucDM-EGFP (panel k is the overlay image of panels e and g; the boxed area is enlarged in panel l to highlight the co-localization) but not EGFP proteins (panel i, enlarged in j) indicating the changing cellular folding capacity for conformationally challenged proteins. C, co-localization between tau species and fLucDM-EGFP was quantified using PCC. Pearson's R value (above threshold) was measured per cell (n Ն25) and presented as a dot plot with mean and S.D. p values were determined using a two-tailed unpaired Student's t test, **, p Ͻ 0.01; ****, p Ͻ 0.0001.

Collapse of proteostasis by tau aggregation
were established (Fig. 6, A and B) for the conditional expression of V5-tagged HSC70 (Fig. 6A, panel b versus panel d, anti-V5 antibody) (Fig. 6B, panel f versus panel h, anti-HSC70/HSP70 antibody). Tetracycline-induced tau P301S-expressing cells were seeded to induce tau aggregation, and cumate was added at day 2 post seeding to induce HSC70 expression for 24 h.

Collapse of proteostasis by tau aggregation
Transferrin uptake was quantified in cells with and without tau aggregation (Fig. 6C). Cumate-induced expression of HSC70 alone does not affect transferrin uptake in control cells but restores transferrin uptake to unaffected levels in MC1-only cells and partially in PG5/MC1-positive cells (32.4% compared with 55.8% inhibition with or without V5-HSC70 expression).
To further support the deleterious consequences of chaperone competition, we performed experiments in which chaperone levels were reduced to determine whether tau aggregation and the inhibition of CME were exacerbated. This was accomplished by expression of siRNA to the HSC70 (HSPA8) in HEK293 tau P301S cells for 48 h after which the cytosolic level of HSC70 was reduced by 3.6-fold (Fig. S6A, compare panels a to c, and quantified in panel e). siRNA treatment was conducted 24 h after the cells were seeded with tau seeds at a 1:10 ratio. The results showed that control siRNA-transfected cells were 7% MC1-only positive and 11% PG5/MC1-positive, whereas cells treated with HSPA8 siRNA showed a significant reduction of MC1-only cells to 0.3% and an increased number (23%) of PG5-positive cells (Fig. S6A, panel f). This further supports the importance of HSC70 in the transition between the MC1 and PG5 conformations of tau. Following HSC70 genetic knockdown, transferrin uptake in cells without tau inclusions was reduced by 28% indicating the sensitivity of CME to cellular HSC70 (Fig. S6B, panel o) (23). CME in PG5-positive cells was further reduced compared with the control siRNA-treated cells (39% versus 53% reduction) (Fig. S6B, panel i versus panel m,  arrows, and quantified in panel o).

Small-molecule proteostasis regulators reduce the sequestration of luciferase-folding sensor by tau inclusions
To further demonstrate that reduction in the cytoplasmic pool of chaperones by sequestration in inclusions can be overcome by increasing the levels of cytoplasmic chaperones, we examined whether small-molecule proteostasis regulators that activate HSF1 and expression of cytoplasmic chaperones can rescue the misfolding and subsequent sequestration of luciferase by tau inclusions. HEK293 tau P301S cells expressing the fLucDM-EGFP sensor were treated with compound A3 for 24 h after seeding, and the levels of luciferase co-localization with the PG5 tau inclusions were then measured. In DMSOtreated cells, fLucDM-EGFP showed strong co-localization with PG5 tau inclusions with an average PCC of 0.63 (Fig. 7 a-e, arrows, and quantified in k) that was reduced to PPC of 0.29 in cells treated with A3 (Fig. 7, h-j, arrows, and quantified in k). Likewise, the co-localization between MC1 tau and luciferase was 0.28 in DMSO-treated cells that reduced to PCC of 0.18 in A3-treated cells.
These studies reveal that chaperone competition by tau aggregation has selective inhibitory effects on vesicular trafficking and the protein-folding arms of proteostasis that can be reversed, and moreover, the different classes of protein conformation states can interfere with certain arms of the PN.

Discussion
Tau aggregation has profound consequences on diverse cellular events (31, 71), and here we demonstrate that tau-dependent sequestration of cytoplasmic molecular chaperones leads  (panels a and b) and by the increased level of total HSC70/HSP70 (panels e-h). Scale bar, 10 m. C, HEK293 tau P301S-V5 HSC70 cells were seeded with insoluble tau. Two days post seeding, cells were treated with or without cumate for an additional 24 h. Transferrin uptake in the cell population stained by pathological tau-specific PG5 and MC1 antibodies was quantified as described.

Collapse of proteostasis by tau aggregation
to chaperone mislocalization and the functional inhibition of chaperone-dependent events of protein folding and clathrinmediated endocytosis. Whereas tau inhibitory effects on protein folding will undoubtedly have broad negative consequences on protein function and cellular health, the loss of CME is particularly relevant for neuronal signaling and receptor activity regulated by ligand-stimulated internalization through clathrin-coated vesicles (72)(73)(74). Failure in endocytic internalization is associated with the abnormal accumulation of activated receptors at the cell surface affecting downstream signaling events such as memory consolidation and long-term depression (75)(76)(77). Moreover, persistent signaling from glutamate receptors at the plasma membrane results in excitotoxicity and eventual neuronal cell death (78,79), and CME is essential for the retrieval of synaptic vesicles after neurotransmitter release (80 -83).
Among the many client interactions for HSC70, interaction with tau directs productive folding and prevents aggregation (44). Upon dissociation of tau from microtubules, HSC70 reengages with tau to facilitate folding into a conformation rec-ognized by the MC1 antibody. Although this serves as a protective mechanism for WT tau, this interaction may also promote protein aggregation when mutant tau persists in a more unstructured state (49). Our data show that the appearance of the MC1 conformational state of tau is associated with CME failure that is detected before the appearance of PG5-positive tau inclusions and the stable sequestration of HSC70. The timing of these cellular events suggests that the inhibition of CME is an early cellular event of tauopathy and precedes failure of other cellular processes such as chaperone-dependent protein folding. This is consistent with observations that tau oligomers corresponding to the MC1 tau conformational species initiate cellular toxicity. Because the MC1 conformational states of tau correspond to the transition of chaperone sequestration, we suggest that chaperone-dependent cellular processes are not equally sensitive to aggregation-induced failure and that CME, in particular, is more at risk by comparison with protein folding.
Inhibition of the protein folding arm of the PN by tau aggregation would further exacerbate the consequence of tau aggregation and accelerate misfolding and aggregation of other  a and f, arrows) and to examine the subcellular localization of the luciferase folding sensor (b and g). The boxed areas in the superimposed images between PG5 and fLucDM-EGFP (c and h) are enlarged to show co-localization between PG5-positive tau aggregates and chaperones, which appears yellow in the overlay (e and j). Images are maximal projections of a z-series of fluorescence confocal slices through the entire cell volume (scale bar, 10 m) and are representative of three independent experiments. k, co-localization between tau species and fLucDM-EGFP was quantified using PCC. Pearson's R value (above threshold) was measured per cell (n Ն40) and presented as a dot plot with mean and S.D. p values were determined using a two-tailed unpaired Student's t test, *, p Ͻ 0.1; ****, p Ͻ 0.0001.

Collapse of proteostasis by tau aggregation
structurally challenged proteins with missense mutations, errors, and structural instabilities (15,84,85). We posit that these events, initiated by tau aggregation, would lead to a vicious cycle of self-propagated protein damage that contributes to proteostasis collapse. Our data further show that tau inclusions associate with a number of cytoplasmic molecular chaperones, which over time would be predicted to interfere with a multitude of cellular processes. Although in this study we have only focused on two specific HSC70-dependent cellular activities of protein folding and CME, our results suggest a much broader negative consequence of global chaperone sequestration.
A basis for proteostasis collapse with implications for aging and protein conformational disease has been observed in Caenorhabditis elegans whereby the regulation of the heatshock response in adult animals is repressed at reproductive maturity by affecting chromatin structure and HSF1 binding at genes encoding molecular chaperones (86). The programmed decline of the heat-shock response and stress-induced expression of molecular chaperones suggests that maintaining and enhancing chaperone expression and function in aging could have beneficial consequences on proteome health. Here, we show that the decline of the chaperone-regulated arms of the PN in tau-expressing cells could be compensated by small-molecule proteostasis regulators that enhanced expression of cytoplasmic chaperones and by a genetic approach using conditional expression of HSC70. Moreover, these experiments suggest that restoration of clathrin-mediated endocytosis and therefore recycling of receptors important to neuronal signaling could prevent neuronal dysfunction in tau-associated diseases.

DNA plasmids and antibodies
To generate the lentiviral expression plasmid encoding the firefly luciferase folding sensor, the blunt-end fragment was released from pCIneoFlucDM-EGFP (69) following Eco53KI and PuvII digestion and ligated to the pENTR11-Dual selection vector (Invitrogen) that was linearized by XmnI and EcoRV. Gateway LR reaction was performed to translocate the fLucDM-EGFP-encoding fragment into the pLenti CMV/TO Puro DEST (670-1) vector (Addgene 17293) (87). The ORF of human HSC70 with a V5 tag at the N terminus was PCR-amplified using pcDNA5/FRT/TO V5 HSPA8 (Addgene 19514) (88), and the PCR fragment was cloned into the SparQ All-in-One Cumate Switch-inducible Lentivector (System Biosciences, QM800A-1) using the NheI and BstBI restriction sites.

Tau seeding and aggregation assay and transferrin assay
Mock Flp-In TM 293 T-REx cells and T-REx TM HEK293-derived tau 1N4R WT and tau 1N4R P301S cells (57) were maintained at 37°C and 5% CO 2 in DMEM complete medium supplemented with 10% heat-inactivated fetal bovine serum (Gibco, 16140071) together with 1% penicillin/streptomycin/ glutamine. Cells were plated at 5 ϫ 10 5 cells/well in 12-well plates and induced with 1 g/ml tetracycline for 24 h to express exogenous tau protein. Cells were then incubated with Sarkosyl-insoluble tau P301S seeds (57, 89) at a 1:40 dilution in Opti-MEM for 3 h. After removing seeds, cells were incubated in complete medium containing tetracycline for an additional 24 h. Cells were transferred to tetracycline-free medium for another 48 h before fixation and immunostaining to detect pathological tau conformation.
For the transferrin assay, 48 h after treatment with insoluble tau seeds, the cells were replated at 3 ϫ 10 5 cells/well onto fibronectin-coated coverslips in a 12-well plate. Transferrin internalization was performed 24 h post replating. Cells were placed in pre-warmed serum-free DMEM for 60 min and then incubated with Alexa 555-conjugated transferrin at 16.7 g/ml (Life Technologies, Inc.) in pre-warmed DMEM, 0.1% BSA for 15 min. The plates were placed on ice, and the cells washed with ice-cold PBS after removing the media. Cells were acid-washed by a 5-min incubation with ice-cold acid wash buffer (0.1 M glycine, pH 2.5, 150 mM NaCl) to remove surfacebound transferrin. Cells were then fixed with 4% paraformaldehyde in PBS for 20 min and immunostained with antibodies recognizing total tau (HT7) and pathological tau species (PG5 and MC1).

Establishing stable-expression cell lines using lentivirus
T-Rex TM HEK293 tau 1N4R P301S cells were infected with lentivirus-encoding luciferase-folding sensor fLucDM-EGFP or V5-HSPA8 for inducible protein expression. Three days after infection, cells were transferred to medium containing 2 g/ml puromycin and maintained until puromycin selection was complete. Stable single clones for the proteostasis reporter lines were further selected by dilution cloning. The resulting reporter cell lines were maintained in complete DMEM supplemented with 200 g/ml Zeocin TM , 5 g/ml blasticidin, and 2 g/ml puromycin.

Small-molecule regulators of PN treatment
HEK293 tau P301S cells were induced to express tau P301S protein, and HEK293 tau P301S-fLucDM-EGFP cells were induced to express both tau P301S and fLucDM-EGFP protein for 24 h. Cells were seeded with Sarkosyl-insoluble tau seeds. One day post tau seeding, cells were replated at 2 ϫ 10 5 cells/ well in a 12-well plate on fibronectin-coated coverslips and grown for an additional 24 h. Small-molecule PN regulators at the indicated concentrations premixed with fresh complete medium were added to cells that were then incubated for 24 h. Cells were then fixed and immunostained to test the induction of HSC70/HSP70 protein expression. Transferrin uptake assay was performed as described.

Tunable expression of human HSC70 protein
The HEK293 tau P301S stable cell line encoding V5-tagged HSC70 protein under the control of the cumate-inducible promoter was generated. Seeding and aggregation assay was performed as described. Two days post seeding, 300 g/ml cumate (System Biosciences, QM150A-1) was added to the cells that were then incubated for an additional 24 h to induce V5-tagged human HSC70 expression. Transferrin uptake was performed. Cells were also immunostained with anti-V5 and -HSC70/ HSP70 antibodies.

Indirect immunofluorescence, microscopy, and image analysis
Cells were fixed with 4% paraformaldehyde in PBS for 20 min. Cells were permeabilized in PBS containing 0.2% Triton X-100 and 1% BSA for 10 min. Cells were blocked with Odyssey Blocking Buffer (LI-COR Biosciences, 927-40000) for 1 h and sequentially incubated with primary antibodies and Alexa Fluor-conjugated secondary antibodies in Odyssey Blocking Buffer for 1 h at room temperature with PBS washes after each antibody treatment. Coverslips were mounted using Fluoromount-G (Southern Biotech).
Image acquisition was conducted as described (23). Figs. S2 and 6S and Fig. 7 were captured with a Leica TCS SP8 confocal microscope, and all other images were obtained using a Leica SP5 II laser-scanning confocal microscope. For image presentation, maximal projections of the z-series were created, and brightness was adjusted across the entire image using LAS AF lite software. Images shown are representative of three independent experiments.
For quantification of transferrin internalization or endogenous chaperone expression levels, raw 16-bit z-series were converted to sum slice projections using ImageJ. Individual cells were outlined as regions of interest (ROI), and the total integrated fluorescence intensities were measured. The mean intensities of three ROIs outside of the cell were also measured to serve as background intensity. The corrected total cell fluorescence (CTCF) was calculated using the following formula: CTCF ϭ integrated density of selected cell Ϫ (area of selected cell ϫ mean fluorescence of background readings). CTCF of the individual cell population was plotted using GraphPad Prism software with 1 ϫ 10 3 fluorescence unit per cell. For transferrin analysis, it was empirically determined that the cells with the 5% highest and lowest fluorescence intensities reliably represented mechanically broken or dead cells, respectively, and these were excluded from the analyses (23).
For co-localization analysis, PCC was calculated from regions of interest identified from confocal image stacks using the Fiji ImageJ co-localization plugin Coloc2 with a calculated point spread function at 3 and 10 Costes shuffling iterations.