Hsp90 activator Aha1 drives production of pathologicaltau aggregatesLindsey B. Sheltona,1, Jeremy D. Bakera,1, Dali Zhenga,1, Leia E. Sullivana, Parth K. Solankia, Jack M. Webstera,Zheying Suna, Jonathan J. Sabbagha, Bryce A. Nordhuesa, John Koren IIIa, Suman Ghoshb, Brian S. J. Blaggb,Laura J. Blaira,2, and Chad A. Dickeya,3

aDepartment of Molecular Medicine and Byrd Alzheimer’s Institute, University of South Florida, Tampa, FL 33613; and bDepartment of Medicinal Chemistry,The University of Kansas, Lawrence, KS 66045

Edited by Manu Sharma, Weill Cornell Medical College, New York, NY, and accepted by Editorial Board Member Gregory A. Petsko July 21, 2017 (received forreview April 27, 2017)

The microtubule-associated protein tau (MAPT, tau) forms neuro-toxic aggregates that promote cognitive deficits in tauopathies,the most common of which is Alzheimer’s disease (AD). The 90-kDaheat shock protein (Hsp90) chaperone system affects the accumu-lation of these toxic tau species, which can be modulated withHsp90 inhibitors. However, many Hsp90 inhibitors are not blood–brain barrier-permeable, and several present associated toxicities.Here, we find that the cochaperone, activator of Hsp90 ATPase ho-molog 1 (Aha1), dramatically increased the production of aggregatedtau. Treatment with an Aha1 inhibitor, KU-177, dramatically reducedthe accumulation of insoluble tau. Aha1 colocalized with tau pathol-ogy in human brain tissue, and this association positively correlatedwith AD progression. Aha1 overexpression in the rTg4510 tau trans-genic mouse model promoted insoluble and oligomeric tau accumu-lation leading to a physiological deficit in cognitive function. Overall,these data demonstrate that Aha1 contributes to tau fibril formationand neurotoxicity through Hsp90. This suggests that therapeuticstargeting Aha1 may reduce toxic tau oligomers and slow or preventneurodegenerative disease progression.

tau oligomers | Aha1 | Alzheimer’s disease | chaperones | Hsp90

The microtubule-associated protein tau (MAPT, tau) accu-mulates and aggregates in a family of neurodegenerative

diseases called tauopathies (1), with the most common being Alz-heimer’s disease (AD) (2). In particular, the pathogenic formationof oligomeric tau species is thought to be a major contributor todisease progression (3). Therefore, strategies aimed at reducingoligomeric tau accumulation could hold therapeutic promise forthese diseases (4).Molecular chaperones, including the 90-kDa heat shock pro-

tein (Hsp90), regulate protein folding, degradation, and accumu-lation (5). Of the proteins regulated by Hsp90, often referred to as“clients,” tau is one of the most thoroughly characterized (6). In thepast decade, Hsp90 emerged as one of the next breakthrough drugtargets for diseases of aging, particularly for neurodegenerativediseases like tauopathies (7). Small molecules inhibiting the ATPaseactivity of Hsp90 showed great promise in preclinical models,prompting the development of a host of clinical leads (8), but thetranslation of this preclinical success into patients has been disap-pointing. Not only have many leads suffered from poor blood–brainbarrier permeability (9), but toxicity has also dampened enthusiasm(10, 11). This has led to the pursuit of Hsp90 cochaperones asdistinct drug targets offering an alternative to Hsp90 (5, 12).Activator of Hsp90 ATPase homolog 1 (Aha1) is the only one

of these cochaperones known to stimulate Hsp90 ATPaseactivity (13). This small 38-kDa cochaperone binds to theN-terminal and middle domains of Hsp90, inducing a partiallyclosed conformation that accelerates the progression of theATPase cycle dramatically (13, 14). Therefore, small moleculestargeting the interaction of Hsp90 with Aha1 could be beneficialin disease by reducing ATPase activity (15, 16). Here, we soughtto determine if Aha1 could facilitate the pathogenesis of tau by

stimulating Hsp90 activity. We determined that Aha1 stimulation ofHsp90 activity can drive tau fibril and oligomer formation, in vitro.Overexpressing Aha1 in a transgenic model of tauopathy increasedneurotoxic oligomeric and insoluble tau. This tau accumulationenhanced both neuron loss and behavioral deficits. Moreover,inhibiting the interaction between Aha1 and Hsp90, using a smallmolecule, reduced insoluble tau accumulation in cultured cells. Ourfindings suggest that targeting Hsp90 cochaperones may enableinhibition of tau aggregation, which could reenergize the trans-lational appeal of the Hsp90 chaperone network as a drug target.

ResultsAha1 Enhances Hsp90-Dependent Tau Aggregation. Since Hsp90 hasbeen shown to exacerbate tau fibril formation (17), we screenedfive established Hsp90 cochaperones to determine whether theyhad an inhibitory or stimulatory effect on this process. Recombi-nant P301L tau was incubated with Hsp90 in the presence of ATPwith or without cochaperone proteins, as indicated (Fig. 1A).Aha1 was the only cochaperone to show a significant enhance-ment of tau fibril formation, while CDC37, p23, FKBP51, andFKBP52 were not significantly different from Hsp90 alone. Wethen examined the effects of Hsp90 and Aha1 on tau fibril for-mation over time. We found the most potent inducer of tau fibrilformation was Hsp90 and Aha1 combined (Fig. 1B). Moreover,Aha1 alone did not affect tau aggregation. These results were also

Significance

The accumulation of toxic tau protein, as in Alzheimer’s disease, isregulated by the 90-kDa heat shock protein (Hsp90) chaperonesystem. Inhibition of Hsp90 has been shown to reduce tau levels.However, Hsp90 inhibition can be problematic due to a lack ofblood–brain barrier permeability and established toxicities. Here,we demonstrate that the Hsp90 cochaperone, ATPase homolog 1(Aha1), dramatically increases the production of aggregated tauin vitro and in a mouse model of neurodegenerative disease.Moreover, inhibition of Aha1 reduced tau accumulation in cul-tured cells. These data identify Aha1 as a target for the treatmentof tauopathies.

Author contributions: L.B.S., B.S.J.B., and C.A.D. designed research; L.B.S., J.D.B., D.Z.,L.E.S., P.K.S., Z.S., B.A.N., S.G., and L.J.B. performed research; B.A.N. and B.S.J.B. contrib-uted new reagents/analytic tools; L.B.S., J.D.B., D.Z., J.M.W., J.J.S., and L.J.B. analyzeddata; and L.B.S., J.M.W., J.K., L.J.B., and C.A.D. wrote the paper.

Conflict of interest statement: C.A.D., L.B.S., B.S.J.B., J.K., and L.J.B. are the coinventors forthe following provisional patent application: “The Hsp90 Activator Aha1 Drives Produc-tion of Pathological Tau Aggregates.”

This article is a PNAS Direct Submission. M.S. is a guest editor invited by the EditorialBoard.1L.B.S., J.D.B., and D.Z. contributed equally to this work.2To whom correspondence should be addressed. Email: lblair@health.usf.edu.3Deceased November 25, 2016.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1707039114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1707039114 PNAS | September 5, 2017 | vol. 114 | no. 36 | 9707–9712

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

confirmed using transmission electron microscopy (TEM), whichshows an increase in tau fibrils in the presence of Hsp90 and anexacerbation of fibrils when both Hsp90 and Aha1 are present(Fig. 1C), suggesting that Aha1 could be responsible for the for-mation of toxic tau oligomers and larger aggregates. Additionally,a mutant, Aha1-E67K, which does not bind to Hsp90 (Fig. S1), didnot enhance tau fibril formation (Fig. 1D). Since heparin is aknown tau aggregation inducer, and tau aggregation can bemodulated by DTT, we conducted control experiments to check ifthe aggregation behavior of tau can be affected by Hsp90, Aha1,or their combination in the absence of heparin or DTT. Tau didnot fibrillate under these conditions within the time frame ex-amined (Fig. S2). Moreover, since Aha1 is a known stimulator ofHsp90 ATPase activity (13, 14), we next investigated the effects ofthese proteins on tau aggregation in the absence of ATP. Wefound that ATP was essential for Aha1/Hsp90-mediated tau ag-gregation (Fig. 1E). Together, these data indicate that Aha1 usesATP to enhance Hsp90-mediated tau aggregation.

KU-177 Inhibits Interaction Between Hsp90 and Aha1. There are nocommercially available Aha1-specific inhibitors. We generatednovobiocin analogs designed to bind to both Hsp90 and Aha1(KU-174) or to only Aha1 (KU-177 and KU-308) (Fig. 2A).Immunoprecipitation of Aha1 from PC3-MM2 cells revealedthat Aha1 and Hsp90 complexes were inhibited by KU-308, KU-177, and KU-174 (Fig. 2B). Hsp90-mediated refolding of dena-tured luciferase was inhibited with KU-174 (Fig. 2C), indicatingthat this compound directly inhibits Hsp90, consistent with aprevious report (18). However, both KU-308 and KU-177, whichlack the noviose sugar required for Hsp90 binding (Fig. 2A, red),did not inhibit luciferase refolding (Fig. 2C). This suggests thatthese compounds do not directly inhibit Hsp90, as they wereengineered to specifically bind to Aha1. Because of these char-acteristics, we chose to use KU-177 as our lead compound. Wefurther tested the ability of KU-177 to inhibit the interactionbetween Hsp90 and Aha1 in HEK cells. Consistent with the PC3-MM2 cells, immunoprecipitation of Aha1 revealed that KU-177 inhibited the binding of Aha1 to Hsp90 (Fig. 2D).

KU-177 Inhibits Tau Aggregation in Vitro.We investigated the abilityof KU-177 to inhibit Aha1-mediated tau aggregation. Recombi-nant P301L tau was incubated with Hsp90 alone or withHsp90 and Aha1, and then treated with KU-177 or DMSO as acontrol. KU-177 was able to significantly reduce tau fibril forma-tion compared with the DMSO control (Fig. 3A). KU-177 showeda robust reduction in tau fibril formation, as observed by TEM(Fig. 3B). Inducible HEK (iHEK)-P301L cells transfected withAha1-WT or Aha1-E67K were treated with KU-177 and har-vested to examine soluble and sarkosyl-insoluble tau. We seethat both the mutant Aha1-E67K and the Aha1 inhibitor KU-177 were able to reduce insoluble tau (Fig. 3C). We also notedthat KU-177 increased soluble, phosphorylated tau.Fig. 1. Hsp90 and Aha1 synergize to form tau aggregates. (A) Recombinant

P301L tau fibril formation measured by thioflavin T (ThT) fluorescence,comparing the effect of five different recombinant cochaperone proteinswith Hsp90 and ATP (results represent the mean ± SEM, n = 3; ***P < 0.001).ns, not significant. (B) Recombinant P301L tau fibril formation measured byThT fluorescence over a period of 72 h with or without the addition ofHsp90 and Aha1 (results represent the mean ± SEM, n = 3). (C) Representative20,000× TEM images of recombinant P301L tau fibrils formed in the presenceof indicated chaperone proteins with ATP. (Scale bars: 2 μm.) (D) RecombinantP301L tau fibril formation was measured by ThT fluorescence in the presenceATP and chaperones as indicated (results represent the mean ± SEM, n = 3;*P < 0.05). (E) Recombinant P301L tau fibril formation measured by ThTfluorescence with varying mixtures of Hsp90, Aha1, and ATP as indicated(results represent the mean ± SEM, n = 3; ***P < 0.001, *P < 0.05).

Fig. 2. KU-177 inhibits interaction between Hsp90 and Aha1. (A) Chemicalstructure of the novobiocin analogs KU-174, KU-177, and KU-308. The nov-iose sugar moiety (red) is required for Hsp90 binding of novobiocin analogsand is absent in KU-177 and KU-308. The biaryl amide moiety (green) hasbeen shown to interact with Aha1 (18). (B) Immunoprecipitated Aha1 fromPC3-MM2 cells treated with ±10 μM KU-308, KU-177, or KU-174 for 24 h wasanalyzed by Western blot. Without antibody (−Ab) indicates a mock im-munoprecipitation. (C) Comparison of Hsp90-mediated luciferase refoldingactivity in PC3-MM2 cell treated with DMSO or 100, 25, 6.25, 1.56, 0.39, and0.097 μM KU-308, KU-177, or KU-174 for 2 h. The IC50 value for KU-177 isshown (R2 = 0.98). Dose–response curves for KU-308 and KU-177 suggest theIC50 values would be higher than the range of concentrations examined here(KU-308, KU-174: n = 3; KU-177: n = 2). (D) Immunoprecipitated Aha1 fromiHEK cells treated ±10 μM KU-177 for 24 h was analyzed by Western blot.

9708 | www.pnas.org/cgi/doi/10.1073/pnas.1707039114 Shelton et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

Aha1 Colocalization with Tau Tangles Correlates with DiseaseProgression in Human AD Brain. We evaluated postmortem hu-man brain samples from patients with AD or healthy age-matched controls for Aha1 localization in relation to tau tangles(Fig. 4A). We found a significant increase in the amount ofcolocalization between Aha1 and tau tangles as shown by immu-nofluorescence (pS396/404, PHF1) in AD samples compared withcontrols (Fig. 4B). There was a positive correlation betweenAha1 immunofluorescence intensity and tau Braak staging (Fig.4C). This suggests a role for Aha1 in pathological tau progression.

Aha1 Overexpression in rTg4510 Mice Increased Oligomeric andInsoluble Tau Species. Five-month-old rTg4510 mice received bi-lateral hippocampal injections of adenoassociated virus serotype9 (AAV9)-Aha1 (n = 9) or AAV9-mCherry (n = 8) (Fig. 5A).Immunohistochemical staining revealed that Aha1 was overexpressedthroughout the hippocampus (Fig. 5B). Aha1 overexpression signif-icantly increased monomeric and multimeric sarkosyl-insoluble tau inthe hippocampus (Fig. 6); sarkosyl-insoluble tau was not detected inwild-type mice (Fig. S3). Insoluble phosphorylated tau was not sig-nificantly increased. Aha1 overexpression also increased toxic T22-tau oligomer levels (17) both in individual mouse samples (Fig. 7 Aand B) and in pooled samples from each treatment group (Fig. 7 Cand D). This increase of T22-tau oligomers in Aha1-overexpressingmice was further confirmed using immunohistochemistry (Fig. 7 Eand F) and semidenaturing Western blotting (Fig. 7 G and H).

Aha1 Overexpression in rTg4510 Mice Leads to Neuronal Loss andCognitive Impairments. Using unbiased stereology, rTg4510 miceoverexpressing Aha1 showed a significant reduction in hippo-campal CA1 neurons compared with mCherry controls (Fig. 8 Aand B). Learning and memory were evaluated in mice injectedwith AAV9-Aha1 (n = 9) and AAV9-mCherry (n = 8) using the2-d radial arm water maze (RAWM). Animals overexpressingAha1 made significantly more errors in locating the submergedescape platform compared with mCherry-overexpressing litter-mates, demonstrating a memory recall deficit (Fig. 8C). Overall,these data demonstrate that Aha1 enhances Hsp90-mediated tauaggregation. This interaction results in increased oligomeric andinsoluble tau concomitant with neuronal loss and memory deficits.

DiscussionIn this study, we identified the Hsp90 cochaperone Aha1 as apotential therapeutic target for the treatment of tauopathies.Our data suggest that Aha1 increased tau fibril formation,

resulting in insoluble tau accumulation by stimulating Hsp90ATPase activity. Expression of Aha1 not only increased insolubletau levels but also significantly increased T22 immunoreactivetau oligomers. This increase in pathological tau levels manifestedin neuronal loss and cognitive deficits. Furthermore, we dem-onstrated that the Aha1 inhibitor KU-177 reduced the accu-mulation of insoluble P301L tau in cultured cells. This suggeststhat Aha1 may be a promising target for the development oftherapeutics directed toward reducing tau aggregation.Previous work has focused on Hsp90 as a therapeutic target to

reduce the toxic load of amyloidogenic proteins in cells (19).However, this endeavor has been challenging as Hsp90 has manyclient proteins within the cell and inhibiting this chaperone canlead to many pleiotropic effects (10, 20). Compounds that targetspecific Hsp90 cochaperones (12) are being investigated for theirpotential to be less toxic as well as more specific (5). Targetingthe Hsp90/p23 and Hsp90/CDC37 complexes with celastrol an-alogs (21–24) or withanolides (25–27) has been investigated.However, these compounds still bind Hsp90 and have effects sim-ilar to Hsp90 inhibitors (27, 28). Alternatively, small- molecule in-hibitors of Hsp90/HOP complexes disrupt this complex by bindingdirectly to HOP (29). One of these compounds, C9, was shown tohave anticancer effects similar to direct Hsp90 inhibition, withoutinducing heat shock response (30). Until recently, there wereno known small-molecule inhibitors of Aha1. Ghosh et al. (18)

Fig. 3. KU-177 inhibits Aha1 enhancement of Hsp90-mediated tau aggre-gation. (A) Recombinant P301L tau fibril formation measured by thioflavin T(ThT) fluorescence, comparing the effect of 10 μM KU-177 or DMSO on taufibril formation (results represent the mean ± SEM, n = 3; **P < 0.01, *P <0.05). (B) Representative 20,000× TEM images of recombinant P301L tau fi-brils formed with KU-177 or DMSO control. (Scale bars: 2 μm.) (C) iHEK-P301Lcells transfected with Aha1-WT, Aha1-E67K, or empty vector were treatedwith 10 μM KU-177 or DMSO and harvested, and soluble and sarkosyl-insoluble fractions were then prepared. Blots were probed by antibodiesas indicated.

Fig. 4. Human AD samples show colocalization between Aha1 and tautangles. (A) Tissue samples from the medial temporal gyrus of patients atBraak stage 2, 5, or 6 were stained for Aha1 (red), pS396/404 tau tangles(green), and neuronal Nissl (Neurotrace, blue), and then imaged using con-focal microscopy; images were taken at a magnification of 60×. (Scale bars:20 μm.) Representative, control sections lacking primary antibody are shownon the Far Right. (B) Quantification of colocalization between Aha1 andphosphorylated tau tangles (pS396/404) (results represent the mean Pear-son’s correlation coefficient ± SEM, n = 10 images; ***P < 0.001). (C) Scatterplot of the intensity of Aha1 fluorescence and Braak staging (results repre-sent the mean fluorescence intensity ± SEM; Braak stage 2: n = 10 images,Braak stage 5: n = 14 images, Braak stage 6: n = 9 images; ***P < 0.001).RFU, relative fluorescence units.

Shelton et al. PNAS | September 5, 2017 | vol. 114 | no. 36 | 9709

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

identified compounds that bind to either Hsp90 or Aha1 basedon the novobiocin scaffold. More recently, two additional Aha1/Hsp90 inhibitors were identified (31). These compounds dem-onstrated protection against pathologies related to cystic fibrosis,but it is still unclear if these inhibitors bind directly toHsp90 or Aha1.Here, we demonstrated that the Aha1-binding inhibitor KU-

177 reduced Hsp90/Aha1-mediated toxic tau accumulation. Fur-ther studies will be required to determine the pharmacokinetics,brain distribution, and efficacy of KU-177 and future classes ofAha1 inhibitors. Collectively, this study identified a role for Aha1in the progression of tauopathies. This suggests inhibition of Aha1may prevent or reverse the accumulation of pathogenic tau.

Materials and MethodsAntibodies. The following antibodies were used: anti-Aha1 antibodies (SMC-172D, StressMarq; ab83036 for immunoprecipitation, Abcam), anti-Hsp90α(SMC-149B; StressMarq), anti-GAPDH (60004-1-Ig; Proteintech), anti-NeuN(MAB377B; Millipore), H-150 anti-tau (sc-5587; Santa Cruz Biotechnology),and anti-tau pT231 (55313-025; Anaspec). PHF1 anti-tau (pS396/404) was akind gift from Peter Davies, Feinstein Institute for Medical Research, Man-hasset, NY. T22 anti-tau oligomer was a kind gift from Rakez Kayed, Uni-versity of Texas Medical Branch, Galveston, TX.

Plasmids and Viral Vectors. Aha1 WT and Aha1 E67K expression plasmidswere generated in our laboratory using the pCMV6 backbone. AAV9-Aha1and AAV9-mCherry were generated in our laboratory for murine genetherapy studies.

Protein Expression. The details of the expression and purification ofrecombinant human P301L tau, Aha1, Aha1 E67K, p23, FKBP51, FKBP52, andCDC37 are described in SI Materials and Methods. Hsp90α protein was a kindgift from Johannes Buchner, Technical University of Munich, Munich.

TEM. Ten microliters of protein samples was adsorbed onto square meshcopper grids (EMS300-Cu) for 60 s and washed twice with 10 μL of deionizedwater, and excess water was removed by wicking with filter paper. Sampleswere negatively stained with 1% uranyl acetate for 30 s and dried overnight.Grids were viewed using a JEOL 1400 Digital Transmission Electron Micro-scope, and images were captured with a Gatan Orius wide-field camera.Fields shown are representative.

Thioflavin T Fluorescence Assay. Tenmicromolar P301L tau was incubatedwith400 nM indicated chaperone in 100 μM sodium acetate (pH 7.0) buffer with2 mM DTT, 2.5 μM heparin (3,000 Da), and 10 μM thioflavin T in 100-μLvolumes in a 96-well, black, clear-bottom plate (07-200-525; Fisher) for 3 d at37 °C. Fluorescence was read at 440 nm excitation and 482 nm emission in a

BioTek Synergy H1 plate reader at indicated time points. All conditions wereperformed at least in duplicate.

Cell Culture and Transfection. The iHEK-P301L cells (32) and luciferase-expressing PC3-MM2 cells (32) were cultured in DMEM supplemented with10% FBS and 1% penicillin/streptomycin (Invitrogen). Inducible cells wereincubated with 3 μg of tetracycline for 72 h. Forty-eight hours before har-vest, transfections were performed with 2.5 μL of Lipofectamine 2000(Invitrogen) per 1 μg of DNA, which was incubated in serum-free Opti-MEMfor 5 min before adding the mixture drop-wise to the cells. KU-177 wasadded 24 h before harvest at indicated concentrations. Cells were harvestedin Tsaio TBS buffer [50 mM Tris base, 274 mM NaCl, 5 mM KCl (pH 8.0)]containing protease inhibitors. Samples were prepared as previously de-scribed (33) to obtain soluble (S1) and sarkosyl-insoluble (P3) fractions.Coimmunoprecipitation. Coimmunoprecipitation of Hsp90α with Aha1 fromPC3-MM2 and iHEK-P301L cells incubated with the indicated compounds for24 h was performed as previously described (18).Luciferase refolding assay. Compound dissolved in DMSO at the indicatedconcentrations or a DMSO control was evaluated in a luciferase refoldingassay in PC3-MM2 cells as previously described (34), and dose–response curvesof the luminescence signal relative to DMSO control were generated usingGraphPad Prism 5.0.

Human Tissue Processing. Brain tissue samples from themedial temporal gyrusof patients with Braak stage 2, 5, or 6 were provided by the University ofCalifornia Alzheimer’s Disease Research Center and the Institute for MemoryImpairments and Neurological Disorders. Samples were fixed in 4% para-formaldehyde overnight; sucrose gradients up to 30% were then used, andtissue was sectioned on a sliding microtome at 25-μm-thick sections. Sectionswere stored at 4 °C in Dulbecco’s PBS supplemented with 0.065% sodiumazide until they were used for immunohistochemistry.

Animal Studies and Tissue Processing. The rTg4510 (Jackson Laboratories) andnontransgenic control mice received bilateral stereotaxic hippocampal(X = ±3.6, Y = −3.5, Z = +2.68) injections of AAV9 vector (miniature CMV +

Fig. 6. Aha1 overexpression in rTg4510 mice leads to increases in insolubletau species. (A) Western blot analysis of soluble and sarkosyl-insolublefractions from hippocampal tissue of rTg4510 mice expressing eitherAAV9-Aha1 or AAV9-mCherry. Six representative samples from AAV9-Aha1–and AAV9-mCherry–injected mice are shown. (B) Quantification of Westernblots of sarkosyl-insoluble total (amino acids 1–150), pS396/404, andpT231 tau (results represent the mean ± SEM relative to the level of mo-nomeric tau in AAV9-mCherry–injected mice (mCherry, n = 8; Aha1, n = 9;*P < 0.05, **P < 0.01). ns, not significant.

Fig. 5. Viral transduction leads to sustained overexpression of Aha1 in thehippocampus of rTg4510 mice. (A) Characteristic phenotype of rTg4510 tautransgenic mouse model along with experimental design time points.(B) Representative images of brain sections showing viral expression ofAha1 protein in AAV9-injected Aha1 and mCherry control littermates. (Scalebars: whole slice, 1,000 μm; Inset, 250 μm.)

9710 | www.pnas.org/cgi/doi/10.1073/pnas.1707039114 Shelton et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

chicken β-actin; 1012) at 5 mo of age [n = 20 (10 transgenic [seven male, threefemale] and 10 nontransgenic [seven male, three female]); n = 19 (ninetransgenic [six male, three female] for Aha1, 10 nontransgenic [six male,four female] for mCherry)]. Each injection delivered 2 μL of AAV9 particles.At 7 mo of age, the mice were used for behavioral testing using the RAWMtask. Upon completion of the RAWM task, the brains were harvested aftercardiac perfusion with 0.9% saline. The right hemisphere from each mousewas dissected, and the hippocampus was then snap-frozen and stored at−80 °C until processed as previously described (33) to obtain S1 andP3 fractions. The left hemisphere from each mouse was fixed in 4% para-formaldehyde overnight, and sucrose gradients up to 30% were then used,with 25-μm-thick tissue sections were generated using a sliding microtomefor general histochemical staining and 50-μm sections generated for stere-ology studies. Sections were stored at 4 °C in Dulbecco’s PBS supplementedwith 0.065% sodium azide until they were used for immunohistochemistry.

Western Blot and Dot Blot Analysis. Cell and mouse brain tissues samples wereanalyzed by Western blot using 4–15% SDS gradient gels (BioRad). Antibodydilutions were 1:1,000 unless otherwise stated, and all secondary antibodieswere used at 1:1,000 (Southern Biotech). Blots were developed using ECL(Pierce) on a LAS-4000 mini imager (GE Healthcare). For dot blots, proteinswere applied onto a wet nitrocellulose membrane and dried by vacuum.Dried membranes were blocked and developed as described above.

Semidenaturing Western Blot. Tissue was homogenized using sonication, andthe low-speed spin fraction was collected after centrifugation at 13,000 × gfor 15 min. Samples were then mixed with 2× Laemmli sample buffer (Bio-Rad) containing 2.1% SDS, and run on a blot using 4–15% SDS gradient gels(BioRad) without boiling the samples or adding β-mercaptoethanol.

RAWM. The RAWM task was performed as previously described (35). Briefly, acircular black tank with a six-arm metal insert was filled with water, anda platform was submerged 1 cm below the surface of the water at the end ofa designated goal arm. Animals were permitted 60 s to locate the platform,during which time an observer blinded to treatment manually scored thenumber of errors. An error was defined as an entry into an incorrect arm orthe absence of an arm choice within 15 s. Mice were trained over 2 d with12 trials per day, which were divided into four blocks of three trials each.Average errors were calculated for each mouse on day 1 and day 2. Groupswere evaluated separately each day with one-way ANOVA, using a leastsignificant difference test to compare groups.

Immunohistochemistry. All immunohistochemistry was done using free-floating sections as previously described (36). Human tissue was stained aspreviously described using immunofluorescent secondary antibodies (17, 37).A detailed description of the immunohistochemistry methods can be foundin SI Materials and Methods. Sections stained for stereology were blockedand permeabilized as described above and incubated overnight at roomtemperature with biotinylated anti-NeuN (1:3,000). Following washes, avidin–biotin complex (ABC) conjugation, and peroxidase development, tissue wasmounted on charged glass slides and allowed to dry overnight. A 0.05%Cresyl violet counterstain was applied to slides and then briefly and quicklydestained with 0.3% acetic acid in water before dehydration.

Microscopy. Bright-field–stained tissue was imaged using a Plan-Apochromat(PLAN-APO) 20×/0.88 objective on a Zeiss Axioscan.Z1 slide scanner. Braintissue immunofluorescently stained was imaged using a Leica TCS SP2 forimage analysis. A Zeiss LSM 880 AxioObserver laser scanning confocal mi-croscope was used for representative images. Fields of view were selected inthe cortex based on tau-positive staining. A 63×/1.40 PLAN APO oil objectivewas used to take a minimum of ten 1-μm Z-stacked images with Argon (for

Fig. 8. Aha1 overexpression in rTg4510 mice leads to cognitive impair-ments. (A) Representative images of NeuN-stained neurons in theCA1 region of the hippocampus (brown) counterstained with Cresyl violet(purple) from AAV9-mCherry– and AAV9-Aha1–injected mice (Insets). (Scalebars: 100 μm.) (B) Quantification of unbiased stereology (results representthe mean ± SEM; mCherry, n = 7; Aha1, n = 8; ***P = 0.0003). (C) RAWMwasperformed on AAV9-Aha1 and AAV9-mCherry rTg4510 (Tg) and WT litter-mates as indicated. Average errors from day 1 (training) and day 2 (memory)are shown (results represent the mean ± SEM; n ≥ 9; *P < 0.05).

Fig. 7. Aha1 overexpression in rTg4510 mice leads to increases in patho-logical tau species. (A) Dot blot of hippocampal tissue of individual miceshown in triplicate probed by T22. (B) Quantification of dot blot (resultsrepresent the mean ± SEM; mCherry, n = 8; Aha1, n = 8; **P < 0.01). (C) Dotblot of pooled hippocampal tissue shown in triplicate probed by T22.(D) Quantification of dot blot (results represent the mean ± SEM of triplicatesamples taken from the pooled fractions; n = 3; **P < 0.05). (E) Represen-tative images of brain tissue slices stained with T22 from AAV9-mCherry–and AAV9-Aha1–injected mice. (Scale bars: whole slice, 1,000 μm; Inset,250 μm.) (F) Quantification of the T22-positive area in the hippocampal fieldof view (Inset from E) (results represent the mean ± SEM; mCherry, n = 8;Aha1, n = 9; *P < 0.05). (G) Samples from AAV9-Aha1 and AAV9-mCherrymice were run on a semidenaturing gel and probed by T22 (1:500, approx-imately 75 kDa) along with other antibodies as indicated. (H) Quantificationof T22 Western blot (results represent the mean ± SEM; mCherry, n = 6;Aha1, n = 7; ***P < 0.001).

Shelton et al. PNAS | September 5, 2017 | vol. 114 | no. 36 | 9711

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

tau-positive signal in green), and Red HeNe (for Aha1-positive and Neuro-trace signal in red).

Imaging Analysis. Bright-field image analysis was performed using NearCYTEsoftware (www.nearcyte.org) as previously described (17). This program wasused to outline regions of interest, and thresholds were then set manuallyuntil all of the user-determined positive cells were selected with as littlenonspecific area selected as possible. Using the batch process option, thearea positive ratio was automatically calculated for each slide.

Fluorescent image analysis was performed using ImageJ (NIH). Backgroundwas subtracted from the red channel using theGaussian Blur tool (radius= 50 μm),and the new blurred image was then subtracted from the original image. Thered channel was also despeckled before image analysis. Both channels were setto a consistent threshold, and colocalization between the red and greenchannels was then quantified with a Pearson’s coefficient. The intensity of redfluorescence was also measured to make a scatter plot showing levels ofAha1 in relation to Braak staging.

Stereology.Neurons were stained with anti-NeuN and Cresyl violet, and thosepositive for both were counted in the CA1 of the hippocampus. A comput-erized stereological system, connected to a Leica DM4000Bmicroscopewith aPrior motorized stage, was used to outline the area using distinct landmarksin the brain at a magnification of 4× (37, 38). Every eighth section was slicedat 50 μm to be used for stereology, and only sections containing hippocampi(as determined by analyzer) were counted (mCherry: n = 7 animals, ap-proximately nine sections per animal, approximately five reference pointsper section; Aha1: n = 8 animals, approximately nine sections per animal,approximately five reference points per section). After the initial analysis,the mCherry control group was reanalyzed at a higher stringency level.Neurons were counted in this region by using randomly designated areas inthe computer-generated grid using a 100× oil immersion lens. Neurons were

counted when they were located within the 3D dissectors or touching theinclusion lines, and the top 1 μm and bottom 1 μm of tissue were excluded.After analysis of all tissue, the number of neurons per animal was multipliedby 4.5 to reflect the total number of neurons throughout the hippocampus.

Statistical Analysis. To compare two groups, a t test was used. Groups largerthan two were evaluated using one-way ANOVA with Dunnett’s multiplecomparison test. P values below 0.05 were considered significant.

Study Approval. All studies were carried out following the guidelines set bythe University of South Florida’s Institutional Animal Care and Use Com-mittee in accordance with the Association for Assessment and Accreditationof Laboratory Animal Care International regulations. All human tissue wasacquired under approved Institutional Review Board protocols for the Uni-versity of California, Irvine. Patient samples were deidentified and approvedfor studies of this nature, with written informed consent provided to use thetissue for research purposes.

ACKNOWLEDGMENTS. We thank Dr. Peter Davies for the PHF1 antiphos-phorylated tau (pS396/404) antibody. We also thank Dr. Rakez Kayed forproviding the T22 anti-tau oligomer antibody. We acknowledge Dr. PeterMouton for his stereology expertise, Dr. Vladamir Uversky for insightful edits,and Dr. Andrew Lesniak for providing the NearCYTE software. We also thankDrs. Nicole Berchtold and Carl Cotman for access to human tissue samples fromthe University of California, Irvine, Alzheimer’s Disease Research Center, whichis funded by NIH/National Institute on Aging Grant P50 AG16573. This workwas supported by NIH Grants NS073899 and MH103848 and by Veteran’sHealth Administration Grants BX001637 and BX002475. This material is theresult of work supported with resources and the use of facilities at the JamesA. Haley Veterans’ Hospital. The contents of this publication do not repre-sent the views of the Department of Veterans Affairs or the United StatesGovernment.

1. Lee VM, Goedert M, Trojanowski JQ (2001) Neurodegenerative tauopathies. AnnuRev Neurosci 24:1121–1159.

2. Ballatore C, Lee VM-Y, Trojanowski JQ (2007) Tau-mediated neurodegeneration inAlzheimer’s disease and related disorders. Nat Rev Neurosci 8:663–672.

3. Caughey B, Lansbury PT (2003) Protofibrils, pores, fibrils, and neurodegeneration:Separating the responsible protein aggregates from the innocent bystanders. AnnuRev Neurosci 26:267–298.

4. Sahara N, et al. (2007) Molecular chaperone-mediated tau protein metabolismcounteracts the formation of granular tau oligomers in human brain. J Neurosci Res85:3098–3108.

5. Blair LJ, Sabbagh JJ, Dickey CA (2014) Targeting Hsp90 and its co-chaperones to treatAlzheimer’s disease. Expert Opin Ther Targets 18:1219–1232.

6. Karagöz GE, et al. (2014) Hsp90-Tau complex reveals molecular basis for specificity inchaperone action. Cell 156:963–974.

7. Pratt WB, Gestwicki JE, Osawa Y, Lieberman AP (2015) Targeting Hsp90/Hsp70-basedprotein quality control for treatment of adult onset neurodegenerative diseases.Annu Rev Pharmacol Toxicol 55:353–371.

8. Neckers L, Workman P (2012) Hsp90 molecular chaperone inhibitors: Are we thereyet? Clin Cancer Res 18:64–76.

9. Inda C, Bolaender A, Wang T, Gandu SR, Koren J, 3rd (2016) Stressing out Hsp90 inneurotoxic proteinopathies. Curr Top Med Chem 16:2829–2838.

10. Butler LM, Ferraldeschi R, Armstrong HK, Centenera MM, Workman P (2015) Maxi-mizing the therapeutic potential of HSP90 inhibitors. Mol Cancer Res 13:1445–1451.

11. Hong DS, et al. (2013) Targeting the molecular chaperone heat shock protein 90(HSP90): Lessons learned and future directions. Cancer Treat Rev 39:375–387.

12. Gaali S, et al. (2015) Selective inhibitors of the FK506-binding protein 51 by inducedfit. Nat Chem Biol 11:33–37.

13. Wolmarans A, Lee B, Spyracopoulos L, LaPointe P (2016) The mechanism of Hsp90ATPase stimulation by Aha1. Sci Rep 6:33179.

14. Li J, Richter K, Reinstein J, Buchner J (2013) Integration of the accelerator Aha1 in theHsp90 co-chaperone cycle. Nat Struct Mol Biol 20:326–331.

15. Okayama S, et al. (2014) p53 protein regulates Hsp90 ATPase activity and therebyWntsignaling by modulating Aha1 expression. J Biol Chem 289:6513–6525.

16. Wang X, et al. (2006) Hsp90 cochaperone Aha1 downregulation rescues misfolding ofCFTR in cystic fibrosis. Cell 127:803–815.

17. Blair LJ, et al. (2013) Accelerated neurodegeneration through chaperone-mediatedoligomerization of tau. J Clin Invest 123:4158–4169.

18. Ghosh S, et al. (2015) Hsp90 C-terminal inhibitors exhibit antimigratory activity bydisrupting the Hsp90α/Aha1 complex in PC3-MM2 cells. ACS Chem Biol 10:577–590.

19. Zhao H, Michaelis ML, Blagg BSJ (2012) Hsp90 modulation for the treatment ofAlzheimer’s disease. Adv Pharmacol 64:1–25.

20. Schulz R, Dobbelstein M, Moll UM (2012) HSP90 inhibitor antagonizing MIF: Thespecifics of pleiotropic cancer drug candidates. Oncoimmunology 1:1425–1426.

21. Jiang F, et al. (2016) Optimization and biological evaluation of celastrol derivatives asHsp90-Cdc37 interaction disruptors with improved druglike properties. Bioorg MedChem 24:5431–5439.

22. Zhang T, et al. (2008) A novel Hsp90 inhibitor to disrupt Hsp90/Cdc37 complex againstpancreatic cancer cells. Mol Cancer Ther 7:162–170.

23. Patwardhan CA, et al. (2013) Gedunin inactivates the co-chaperone p23 proteincausing cancer cell death by apoptosis. J Biol Chem 288:7313–7325.

24. Chadli A, et al. (2010) Celastrol inhibits Hsp90 chaperoning of steroid receptors byinducing fibrillization of the Co-chaperone p23. J Biol Chem 285:4224–4231.

25. Gu M, et al. (2014) Structure-activity relationship (SAR) of withanolides to inhibitHsp90 for its activity in pancreatic cancer cells. Invest New Drugs 32:68–74.

26. Yu Y, et al. (2010) Withaferin A targets heat shock protein 90 in pancreatic cancercells. Biochem Pharmacol 79:542–551.

27. Sinadinos C, et al. (2013) Low endogenous and chemical induced heat shock proteininduction in a 0N3Rtau-expressing Drosophila larval model of Alzheimer’s disease.J Alzheimers Dis 33:1117–1133.

28. Westerheide SD, et al. (2004) Celastrols as inducers of the heat shock response andcytoprotection. J Biol Chem 279:56053–56060.

29. Yi F, Regan L (2008) A novel class of small molecule inhibitors of Hsp90. ACS Chem Biol3:645–654.

30. Pimienta G, Herbert KM, Regan L (2011) A compound that inhibits the HOP-Hsp90 complex formation and has unique killing effects in breast cancer cell lines.Mol Pharm 8:2252–2261.

31. Ihrig V, Obermann WMJ (2017) Identifying inhibitors of the Hsp90-Aha1 proteincomplex, a potential target to drug cystic fibrosis, by alpha technology. SLAS Discov22:923–928.

32. Ghosh S, et al. (2016) Diverging novobiocin anti-cancer activity from neuroprotectiveactivity through modification of the amide tail. ACS Med Chem Lett 7:813–818.

33. Ramsden M, et al. (2005) Age-dependent neurofibrillary tangle formation, neuronloss, and memory impairment in a mouse model of human tauopathy (P301L).J Neurosci 25:10637–10647.

34. Hall JA, et al. (2016) Novobiocin analogues that inhibit the MAPK pathway. J MedChem 59:925–933.

35. Alamed J, Wilcock DM, Diamond DM, Gordon MN, Morgan D (2006) Two-day radial-arm water maze learning and memory task; robust resolution of amyloid-relatedmemory deficits in transgenic mice. Nat Protoc 1:1671–1679.

36. Dickey C, et al. (2009) Aging analysis reveals slowed tau turnover and enhanced stressresponse in a mouse model of tauopathy. Am J Pathol 174:228–238.

37. Abisambra JF, et al. (2013) Tau accumulation activates the unfolded protein re-sponse by impairing endoplasmic reticulum-associated degradation. J Neurosci 33:9498–9507.

38. Mouton PR, Pakkenberg B, Gundersen HJ, Price DL (1994) Absolute number and sizeof pigmented locus coeruleus neurons in young and aged individuals. J ChemNeuroanat7:185–190.

9712 | www.pnas.org/cgi/doi/10.1073/pnas.1707039114 Shelton et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020