Intraneuronal Tau Misfolding Induced by Extracellular Amyloid-β Oligomers

Abstract

Abnormal folding and aggregation of the microtubule-associated protein, Tau, is a hallmark of several neurodegenerative disorders, including Alzheimer’s disease (AD). Although normal Tau is an intrinsically disordered protein, it does exhibit tertiary structure whereby the N- and C-termini are often in close proximity to each other and to the contiguous microtubule-binding repeat domains that extend C-terminally from the middle of the protein. Unfolding of this paperclip-like conformation might precede formation of toxic Tau oligomers and filaments, like those found in AD brain. While there are many ways to monitor Tau aggregation, methods to monitor changes in Tau folding are not well established. Using full length human 2N4R Tau doubly labeled with the Förster resonance energy transfer (FRET) compatible fluorescent proteins, Venus and Teal, on the N- and C-termini, respectively (Venus-Tau-Teal), intensity and lifetime FRET measurements were able to distinguish folded from unfolded Tau in living cells independently of Tau-Tau intermolecular interactions. When expression was restricted to low levels in which Tau-Tau aggregation was minimized, Venus-Tau-Teal was sensitive to microtubule binding, phosphorylation and pathogenic oligomers. Of particular interest is our finding that amyloid-β oligomers (AβOs) trigger Venus-Tau-Teal unfolding in cultured mouse neurons. We thus provide direct experimental evidence that AβOs convert normally folded Tau into a conformation thought to predominate in toxic Tau aggregates. This finding provides further evidence for a mechanistic connection between Aβ and Tau at seminal stages of AD pathogenesis.

INTRODUCTION

Misfolding and aggregation of the neuron-specific, axon-enriched, microtubule-associated protein, Tau, is a hallmark of Alzheimer’s disease (AD) and a spectrum of non-Alzheimer's Tauopathies [1]. Under normal conditions, Tau plays a role in promotion of microtubule assembly, stabilization of microtubules and regulation of organelle trafficking along microtubules [2-4]. In AD, however, Tau becomes hyperphosphorylated, binds microtubules less effectively and aggregates into oligomers and filaments within neurons [5,6]. While there are many ways to monitor Tau oligomerization and filament formation, methods to monitor earlier conformational changes that may trigger Tau aggregation are less developed. To address this issue we developed a biosensor that detects Tau misfolding in live cells.

Tau contains several important structural domains that mediate its physiological roles, folding and oligomerization. It has 0, 1 or 2 inserts of 29 amino acids each near its N-terminus, a proline-rich region with many potential phosphorylation sites located near the end of the N-terminal half of the protein, and 3 or 4 imperfect, tandem microtubule-binding repeat domains (MTBRs) of 31 or 32 amino acids each located in the C-terminal half [7,8]. Interestingly, the second and third MTBRs are also responsible for Tau-Tau interactions in Tau aggregates [9].

Monomeric Tau is a natively unfolded protein. In vitro, it behaves like a random coil and does not spontaneously form filaments [10-12]. Abnormal Tau aggregation into the straight and paired helical filaments characteristic of Tauopathies is driven by a shift from random coil to a β-sheet structure of regions within the second and third repeat domains [13] Even in its soluble monomeric state, Tau is not entirely devoid of semi-stable structures. In solution, normal Tau was shown to adopt a folded structure in which the N- and C- termini are close together with the C-terminal held closer to the MTBR. This configuration has been called the paperclip [14] or hairpin [15] conformation of Tau.

There is evidence that Tau assembly into filaments depends on an unfolding that abolishes that paperclip or hairpin structure. This form of Tau is recognized by two conformation-dependent antibodies, Alz50 and MC-1, which label Tau in early AD brain [16,17]. Both antibodies recognize similar, but distinct discontinuous epitopes formed by short stretches of amino acids near the N terminal and a region in the MTBR. Additional work has shown that both N- and C-terminal Tau fragments inhibit full length Tau polymerization into filaments [15,18], supporting the model that a conformation change in which movement of the Tau N- and C-termini away from each other is key to the conversion of normal Tau into a toxic, misfolded form.

To test whether Tau does, indeed, convert between folded and unfolded states in living cells, we developed and characterized a full length Tau biosensor labeled at its N- and C-termini with Venus and Teal fluorescent proteins respectively (Venus-Tau-Teal). Because Teal and Venus constitute an effective donor-acceptor pair for Forster resonance energy transfer (FRET), Venus-Tau-Teal allowed us to visualize normally folded (paperclip/hairpin) and unfolded Tau in live cells that expressed the biosensor. Venus-Tau-Teal is operationally analogous to a previously described FRET biosensor from another group based on 0N4R Tau labeled with CFP and YFP [19]. A key distinction between the present study and the earlier publication is that our approach ensured that intramolecular FRET signals, an indicator of the conformation of individual Tau molecules, was minimally contaminated by intermolecular FRET signals or FRET-inhibiting effects due to Tau-Tau aggregation.

By expressing Venus-Tau-Teal in CV-1 African green monkey kidney fibroblasts and primary mouse cortical neurons we obtained evidence that microtubule-associated Tau has a folded, paperclip/hairpin-like conformation that can be modulated by drugs affecting microtubule binding or Tau phosphorylation. Most importantly, we found that amyloid-β oligomers (AβOs) cause Tau to unfold and adopt a conformation associated with toxic Tau aggregates. These observations provide direct evidence that AβOs control Tau conformation, and thereby constitute yet another example of how Aβ and Tau work together to drive AD pathogenesis [20].

MATERIALS AND METHODS

Cell culture

Cell culture reagents were from Gibco/Invitrogen unless specified otherwise. Primary cortical neurons were isolated from wild type (C57/Bl6) mouse embryos aged approximately 18 days as previously described [21,22], except that phenol red-free Neurobasal medium was used exclusively. CV-1 African green monkey kidney cells (ATCC catalog # CCL-70) were maintained in Dulbeco's Modified Eagle's Medium (DMEM) with 10% fetal bovine serum and 50 μg/ml gentamycin, and were dissociated with TyrpLE Express (Thermo Fisher) for subculturing. For live imaging, CV-1 cells and neurons were plated onto 14 mm #1 glass-bottom MatTek dishes.

Fluorescent fusion proteins

Expression vectors for singly and doubly labeled Tau are in the background vector pCSC-SP-PW-NepX (pBOB-NEPX; from Inder Verma, Addgene plasmid #12340). Human 2N4R Tau coding sequence was inserted between the Age1 and HpaI restriction sites by standard restriction digest techniques. cDNAs for the fluorescent proteins, mTFP (Teal) [23] and Venus [24] were then inserted at the N-terminal Age1 site or the C-terminal HpaI site, and confirmed for sequence accuracy and direction of insertion. The expected spectral properties and sizes of all fluorescent fusion proteins were confirmed by spectral imaging and western blots using mouse monoclonal Tau5 antibody [25]. Constructs were then inserted into lentivirus with the packaging vectors pMD2.G ( from Didier Trono, Addgene plasmid # 12259) and psPAX2 ( from Didier Trono, Addgene plasmid # 12260) using the lentiviral protocol provided by Thermo Fisher for Lipofectamine 3000 production of lentivirus (Thermo Fisher L3000015.)

Wild type neurons were transduced with lentivirus between day 8 and day 10 in vitro. CV-1 cells were infected at approximately 60% confluence to allow for easier imaging of individual cells. Expression was monitored until the fluorescence was visible on an EVOS FL microscope (ThermoFisher). Cells were ready for imaging approximately 2 days post-transduction for CV-1 cells and approximately 4 days for neurons.

Western Blotting

Western blots for Supplemental Figure 1 were run using Bio-Rad 10% mini-PROTEAN pre-cast gels in Tris/Glycine SDS running buffer. Boiled samples were run alongside precision Plus Protein ladder (1610374). Staining for Tau utilized mouse monoclonal Tau5 antibody [25] and mouse monoclonal anti-GFP (NeuroMab clone N86/38; also recognizes Venus) with secondary Licor antibodies: IRDye 800CW Donkey anti-rabbit IgG (926-32213) and IRDye 680RD Donkey anti-mouse IgG (926-68072). Protein phosphorylation relied on rabbit polyclonal anti-Tau(pS262) (Anaspec AS-54973) and rabbit polyclonal anti-phosphoSerine/Threonine (Cell Signaling 9381) primary antibodies. Mouse monoclonal anti-Tau oligomer antibodies were kindly provided by Dr. Nick Kanaan of Michigan Sate University (TOC1), and Rakez Kayed of the University of Texas Medical Branch at Galveston (TOMA-1, TTC-35 and TTC-99).

Imaging

For CV-1 cells, standard DMEM was replaced with phenol red-free DMEM prior to imaging. FRET imaging was done on a Zeiss 780 confocal/ NLO/FLIM microscope. Excitation for intensity FRET experiments used an Argon-ion laser (458 nm and 514 nm excitation; 455-500 nm and 526-579 emission). Confocal FRET images were acquired with a GaAsP (Gallium Arsenide Phosphide) detector with approximately 40% quantum efficiency. 2-photon FLIM-FRET images were acquired by exciting the donor (Teal) with a Coherent Chameleon Vision-II Ti:Sapphire laser tuned to 820 nm. Becker & Hickl (Germany). A time correlated single photon counting (TCSPC) board (SPC-150) was used with a high-sensitivity hybrid GaAsP detector (HPM-100-40, 300-650 nm). Images were acquired with a 40X oil objective (1.30 NA). All live imaging was done on a temperature controlled stage at 37° C with humidified 5% CO₂/95% air gas flow to maintain pH and humidity.

Tau and Aβ oligomer preparation

Tau oligomers were prepared as described previously [22,26]. 2N4R Tau was brought to 4 μM in 100 mM Tris, 0.1 mM EDTA, 150 μM Tris(2-carboxyethyl)phosphine (TCEP; Life Technologies) and treated overnight with 50 μM benzophenone-4-maleimide (B4M; Sigma-Aldrich) followed by 5 mM dithiothreitol (DTT; Roche) to inactivate B4M, and dialyzed into 100 mM Tris, 0.1 mM EDTA and 5 mM DTT. A portion was removed and flash frozen for monomer treatments, and the remaining protein was aggregated overnight in the presence of 150 μM arachidonic acid (AA). The oligomers were treated for 5 minutes with UV light at 254 nm (Spectroline model EF-180), then flash frozen in small aliquots to be used immediately after thawing.

Amyloid-β oligomers (AβOs) were prepared from lyophilized synthetic Aβ(1–42) (AnaSpec), dissolved in HFIP (1,1,1,3,3,3-hexafluoro-2-propanol; Sigma- Aldrich Co.) to 1 mM and evaporated overnight at room temperature. The dried powder was resuspended in DMSO to 5 mM and sonicated for 10 minutes in a water bath. The peptide was then diluted to 100 μM in Neurobasal media, and incubated for 48 hrs at 4° C with rocking. Prior to use, oligomers were spun briefly to remove large oligomers and fibrils, bringing the concentration in solution to approximately 50 μM.

Experimental Perturbation of Cells

Cells were imaged prior to any medium additions and then nocodazole, okadaic acid (OA) or taxol (all from abcam), were added to a final concentration of 1 μM while the cultures remained in place on the microscope stage. One hour later the same fields of view were imaged again.

Aβ and Tau oligomer treatments were too long to be done on stage without compromising the health of the cells, so parallel coverslips were imaged for treated and control cultures for these conditions. For AβO treatments, neuron medium (Neurobasal + B27) was replaced with B27-free Neurobasal, and 1 hour later freshly prepared AβOs (see above) were added to the medium to a final concentration of approximately 1.5 μM total Aβ. Cells were imaged 6 hours later. For Tau oligomer treatments, freshly thawed aliquots (see above) were diluted into medium to a final concentration of 250 nM total Tau 18 hours before imaging.

Thresholding for intramolecular FRET

Images were taken in parallel on the same system for intensity (confocal) FRET and 2-photon lifetime FRET (FLIM). It was determined that intensities below 750 arbitrary units of acceptor intensity had minimal contamination of intermolecular FRET (see Figure 3 and corresponding text). Using the parallel images we determined that this intensity range corresponded to a photon count of lower than approximately 500 on the 2-photon detector so cells were selected to be primarily within this expression range. All images then had single pixel regions of interest selected above background (varied slightly by experiment; usually 30-50 photons) and below a photon count of 500. Doing so excluded any cells or regions of cells that had a biosensor concentration that might have yielded significant intermolecular FRET according to the analysis described in Figure 3.

Fig. 3.

Venus-Tau-Teal is superior to Teal-Tau-Venus for detecting intramolecular FRET uncompromised by aggregation effects. Venus-Tau-Teal and Teal-Tau-Venus both target to microtubules and show differing dependence of FRET efficiency (E%) on fluorescent fusion protein concentration. Venus-Tau-Teal, but not Teal-Tau-Venus, shows a broad range of E% independent of increasing acceptor (Venus) intensity (up to 750 arbitrary intensity units). FRET within that Venus intensity range was therefore judged to be predominantly intramolecular, and Venus-Tau-Teal was used as the Tau conformation biosensor for all subsequent experiments. To minimize FRET signal contamination caused by Venus-Tau-Teal aggregation, we restricted subsequent observations to the low end of the range in which E% is independent of Venus intensity (correlated to a photon count of less than ~500 arbitrary intensity units). Bar graphs represent the merged data of 6 fields of view per experiment, repeated in 4 biological replicates with their standard errors of the mean. Statistical significance was determined by one-way ANOVA with Tukey’s post-hoc test for multiple comparisons. Standard errors calculated by Prism column statistics.

FRET / FLIM analysis

Intensity FRET analysis was done largely in ImageJ (https://imagej.nih.gov/ij/) via a PFRET plugin written at the University of Virginia's Keck Center for Cellular imaging. This plugin allows subtraction of background and removal of spectral bleedthrough. After these corrections the result is a calculated E% (efficiency of energy transfer) for each ROI [27,28].

FLIM images were first analyzed by Becker & Hickl SPCImage software (https://www.becker-hickl.com/) Curve fitting procedures done as described previously [29,30]. Briefly, an exponential decay curve was fitted such that the Chi² was approximately 1 (with a two-component analysis of the donor, incomplete exponential decay and a measured instrument response function). These data were exported in a series of asc files. 2-photon images were used to designate single pixel regions of interest within the previously determined intensity range. A Fiji (https://fiji.sc/) macro written by us was then used to measure pixel intensity values in the selected ROIs and generate Excel (Microsoft) based result files for each image, which included a calculated lifetime value for each pixel/ROI. Some additional manual filtering of the results to remove outliers was performed including any ROI’s with a Chi² below 0.5 or above 2.0 (indicating the ROI was not well fitted by the decay curve). Lifetimes below 500 ps were also disregarded. These very low lifetimes were often seen outside of cells in the background, associated with very large Chi² values and were not in the reasonable range for the fluorophore (possibly due to dying cells or degraded biosensors). Lifetimes were plotted as a frequency distribution histogram to determine the peak and distribution of the lifetime for each image/condition.

For each figure (3-6) a single “experiment” constitutes 6-9 fields of view per condition, with 1-2 cells per field of view. The number of ROI’s analyzed per field of view/ experiment varied significantly based on how many were removed by thresholdeding (as background pixels, too bright, poor Chi², but most fields of view were in the range of 5-10 thousand analyzed ROI’s after thresholding.

Fig. 6.

Venus-Tau-Teal is sensitive to AβOs. Primary cortical mouse brain neurons were treated with vehicle or Aβ_1-42 oligomers at a total Aβ_1-42 concentration of approximately 1.5 μM for 6 hours. Middle panels show the results of a single experiment with distributions compared by the Kolmogorov-Smirnov test. Right panels show a summary of the peak lifetime results of 4 separate experiments.

Statistical Analysis

GraphPad Prism 7 software was used to analyze each data set by Kolmogorov-Smirnov test or one-way analysis of variance with a Tukey post-hoc test or t-test as appropriate.

RESULTS

Teal and Venus were chosen as FRET donor and acceptor, respectively, because of their superior fluorescence properties compared to other fluorescent protein pairs with similar excitation and emission spectra [23,24,31]. When Venus and Teal are in close enough proximity to cause a FRET event (approximately 10 nm or less), energy transfer efficiency rises and donor fluorescence lifetime declines with decreasing distance between the fluorophores. To establish the optimal structure for a Tau folding biosensor, we designed Venus-Tau-Teal, Teal-Tau-Venus and four singly tagged fusion proteins: Venus-Tau, Teal-Tau, Tau-Venus and Tau-Teal (Fig. 1). The singly tagged proteins were necessary for calculating spectral bleedthrough for intensity FRET efficiency (E%) measurements [28] and the fluorescence lifetime of unquenched donor (Teal) for fluorescence lifetime imaging (FLIM) [31]. When expressed in CV-1 African green monkey kidney fibroblasts and imaged by confocal microscopy, all singly (Fig. 2) and doubly (Fig. 3) labeled fusion proteins localized to microtubules, indicating that coupling Teal, Venus or both to Tau does not obviously impair Tau's microtubule-binding activity.

Fig. 1.

Fluorescent Tau fusion proteins. Illustrated here are the fluorescent fusion proteins of human 2N4R Tau used in this study. A) Venus-Tau-Teal was compared with B) Teal-Tau-Venus for discriminating intramolecular from intermolecular FRET (see Fig. 3), and the singly labeled fluorescent fusion proteins in B) were used as standards for FRET efficiency (E%) and FLIM experiments. R1, R2, R3 and R4 signify the microtubule-binding repeat domains of Tau.

Fig. 2.

Expression of singly labeled fluorescent Tau fusion proteins in CV-1 fibroblasts. All such fluorescent fusion proteins target to microtubules. These proteins were used for subtracting spectral bleedthrough for FRET efficiency (E%) measurements and for calculation of unquenched donor (Teal) fluorescence lifetime for FLIM experiments.

Teal-Tau-Venus and Venus-Tau-Teal are capable, in principle, of producing intramolecular FRET. In addition, each protein is theoretically capable of aggregating to form oligomers or filaments resulting in intermolecular FRET because of favorable orientations of Teal and Venus on adjacent fluorescent fusion proteins. Intermolecular FRET could also be generated by unfolding coupled with a parallel arrangement of individual fluorescent fusion proteins within aggregates. To use a doubly tagged protein as a FRET biosensor for the folding state of Tau, we therefore had to identify conditions in which measured FRET signals are unaffected by fluorescent fusion protein aggregation. This was accomplished by analyzing Teal-Tau-Venus and Venus-Teal-Tau at various expression levels in CV-1 cells.

More specifically, we sought to establish an expression range in which FRET efficiency (E%) is independent of expression level and therefore is dominated by intramolecular FRET. Since an intermolecular FRET signal would be dependent on the amount of biosensor expressed, a significant effect of acceptor intensity on calculated E% would indicate the FRET signal includes a substantial intermolecular component. Venus-Tau-Teal expression levels were determined by measuring the fluorescence intensity of the acceptor fluorochrome, Venus, using the 514 nm argon laser line to excite Venus without exciting Teal. For each pixel of a measured intensity, an E% was calculated and the data were binned into expression levels, as illustrated in Figure 3. At low levels of biosensor expression, (bins up to 500-750 arbitrary units of acceptor intensity), the increasing biosensor concentration did not have a significant effect on E%. At concentration ranges higher than this level, Venus-Tau-Teal yielded a concentration-dependent decrease in E%, indicating an intermolecular component to the signal.

In contrast, the Teal-Tau-Venus version of the biosensor did not yield a low Venus intensity range in which E% was independent of intensity. We therefore decided to use Venus-Tau-Teal for all further experiments. To minimize potential interference of intramolecular FRET caused by Venus-Tau-Teal aggregation, we also restricted our data collection for subsequent experiments to cells and regions of interest (ROI’s) in which the Venus intensity was less than this determined range of intensity.

Venus-Tau-Teal is sensitive to microtubule loss and protein phosphatase inhibition

To test whether Venus-Tau-Teal responds to cellular perturbations, we first studied its properties under conditions in which microtubules were manipulated. A 1 hour, 1 μM nocodazole treatment of CV-1 cells caused extensive microtubule depolymerization, a concomitant loss of Venus-Tau-Teal association with microtubules and a shift towards longer Teal fluorescence lifetime (Fig. 4A-D). This lifetime increase is indicative of decreased FRET efficiency, implying that Tau is predominantly in the paperclip/hairpin conformation when bound to microtubules and adopts a more open conformation when dissociated from microtubules. Treatment of CV-1 cells with the microtubule-stabilizing drug, taxol, which competes with Tau for MT binding [32], yielded lesser effects on Teal lifetime (Supplementary Figure 2.) In three of four experiments taxol caused a small but statistically significant increase in Teal lifetime which is consistent with a more open tau conformation.

Fig. 4.

Modulation of fluorescence lifetime by perturbation of microtubules and protein phosphorylation. CV-1 cells expressing Venus-Tau-Teal were treated on stage for 1 hour with 1 μM Nocodazole or Okadaic Acid (OA). Middle panels show the results of a single experiment with distributions compared by the Kolmogorov-Smirnov test. Right panels show a summary of the peak lifetime results of 4 separate experiments.

We also tested the effects of overall protein phosphorylation on Venus-Tau-Teal lifetime. To do so we utilized okadaic acid (OA), a broad spectrum protein phosphatase inhibitor that preferentially inhibits protein phosphatase 2A and has been shown to cause several AD-like neuropathologies in vitro and in vivo [33]. Treatment of CV-1 cells for 1 hour with 1 μM okadaic acid increased phosphorylation overall and specifically at p262 of Venus-Tau-Teal (Supplementary Figure 3), and caused a shortening of Teal fluorescence lifetime (Fig. 4.) While this decrease in Teal lifetime might be due to phosphate accumulation on tau, for which 85 phosphorylation sites have been identified (http://cnr.iop.kcl.ac.uk/hangerlab/tautable), increased phosphorylation of other proteins cannot be ignored as contributing factors.

Additionally, because phosphorylation of Tau can induce its oligomerization [34], the possibility that the decrease in Teal lifetime induced by OA is due to intermolecular FRET, rather than increased intramolecular FRET should not be ignored. While our experimental design minimizes intermolecular FRET under basal conditions (see Figure 3), local oligomerization might cause an increase in intermolecular FRET at low overall Venus-Tau-Teal concentrations that would otherwise detect predominantly intramolecular FRET. To shed further light on the fluorescence lifetime decrease caused by OA, CV-1 cells treated with OA were analyzed by western blotting with the Tau oligomer antibodies, TOC1, TOMA-1, TTC-35 and TTC-99. None of those antibodies detected oligomeric Tau either before or after exposing CV-1 cells to OA (data not shown). While these results imply that OA does not induce Tau oligomerization, we cannot eliminate the possibilities that the fluorescent tags on the Tau biosensor blocked binding of anti-Tau oligomer antibodes to Venus-Tau-Teal that did oligomerize, or that the biosensors in OA-treated CV-1 cells formed specific types of oligomers that were not recognized by any of the anti-Tau oligomer antibodies that we tested.

Venus-Tau-Teal responds to pathological oligomers of Tau and Aβ

Extracellular Tau oligomers have been found to cause aggregation of intracellular Tau, accumulation of endogenous Tau in the somatodendritic compartment, alteration of fast axonal transport and synaptotoxicity [22,35,36]. Extracellular AβOs cause a separate set of adverse neuronal responses, such as impaired synaptic activity, ectopic neuronal cell cycle re-entry, which is a prelude to massive neuron death in AD [37], inhibition of nutrient-induced mitochondrial activity, and disruption of normal axon initial segment function [21,38-44]. Because the new results presented here so far indicate that Venus-Tau-Teal can serve as a biosensor for conversions between the compact, paperclip/hairpin and unfolded conformations of Tau, we next tested if extracellular oligomers of Tau or Aβ can alter Venus-Tau-Teal conformation.

After exposure of primary mouse cortical neurons to extracellular human 2N4R Tau oligomers for 18 hours, the Teal fluorescence lifetime of Venus-Tau-Teal shortened (Figure 5), consistent with prior evidence that extracellular aggregated Tau causes intracellular tau to aggregate. [22,35]. In contrast, when primary mouse cortical neurons were exposed to AβOs for 6 hours, the Teal fluorescence lifetime of Venus-Tau-Teal increased (Figure 6). This result indicates that AβOs induce Tau unfolding from the paperclip/hairpin conformation.

Fig. 5.

Venus-Tau-Teal is sensitive to Tau oligomers. Primary cortical mouse brain neurons were treated with human 2N4R Tau monomers or oligomers at a total Tau concentration of 250 nM for 18 hours. Middle panels show the results of a single experiment with distributions compared by the Kolmogorov-Smirnov test. Right panels show a summary of the peak lifetime results of 4 separate experiments.

DISCUSSION

Tau misfolding and aggregation underlie the pathogenesis of AD and non-Alzheimer's tauopathies, such as progressive supranuclear palsy, Pick's disease, Parkinson's disease, Huntington's disease and many others. Several methods are well established for detecting aggregated tau, including binding of Congo red [45], and a variety of antibodies specific for oligomeric or fibrillar Tau [46,47], and PET imaging that can detect neurofibrillary tangles in live patients [48]. In contrast, methods to detect Tau conformational changes that are thought to precede and promote Tau aggregation are far more limited. Two monoclonal antibodies, Alz50 and MC-1, recognize similar discontinuous epitopes that comprise regions of Tau near its N-terminal and within its microtubule-binding repeat region [16,17]. The Alz50 and MC-1 epitopes are infrequently detected in normal brain, and Alz50 and MC-1 immunoreactivity are thought to represent a seminal step in the conversion of normal Tau to pathogenic Tau. The utility of Alz50 and MC-1 is limited, however, to examination of fixed cells and tissues.

Here we describe a new fluorescence-based biosensor, Venus-Tau-Teal, that can detect Tau conformational changes in live cells. Venus-Tau-Teal can discriminate the Tau paperclip/hairpin conformation, in which the N- and C-termini are located in close proximity to each other and to the microtubule-binding repeat region [14,15], from an unfolded conformation in which the Tau N- and C-termini have dissociated. By expressing this biosensor in CV-1 cell fibroblasts and primary mouse cortical neurons, we gathered evidence that the paperclip/hairpin conformation predominates on microtubule-associated Tau, and is sensitive to a variety of experimental perturbations.

This study represents a refinement of prior work that describes a similar biosensor, ECFP-Tau-EYFP, based on human 0N4R Tau [19]. While our study focuses on a different isoform of the 6 that are expressed in the CNS, there is precedence for Tau isoform-specific toxicity [22,49] and drug response[50]. Accordingly, further work with additional biosensors for other Tau isoforms will be required to assess how widely the results reported here apply to isoforms other than 2N4R Tau.

Besides being based on a different Tau isoform, the Venus-Tau-Teal biosensor incorporates a superior FRET donor-acceptor pair [23,24,31] compared to ECFP-Tau-EYFP. Most critically, we defined, and exclusively relied on experimental conditions in which intramolecular FRET was minimally contaminated by intermolecular FRET and other possible complications caused by biosensor aggregation.

While most of the data presented here focused on the characterization of Venus-Tau-Teal (Figs 1-4), our goal from the start was to develop a Tau conformation biosensor that can discriminate normally folded from pathologically folded Tau. We therefore included in the study experiments that monitored Venus-Tau-Teal responses to pathogenic extracellular oligomers made from Tau or Aβ, each of which disrupts multiple aspects of neuronal homeostasis [21,22,36,38-44,51]. It is important to note that oligomers of Tau and Aβ are very heterogeneous in terms of their subunit composition and organization, and knowledge of their structures at the atomic resolution level is lacking. With this limitation in mind, in agreement with the work of Di Primio and colleagues[19], we found that extracellular Tau oligomers cause increased biosensor FRET (Fig. 5), presumably due to intraneuronal Tau aggregation [22,35] in a manner that reinforces Teal-Venus proximity.

Extracellular AβOs caused the opposite response: lengthened Teal fluorescence lifetime of Venus-Tau-Teal, indicative of Tau unfolding from the paperclip/hairpin conformation (Fig 6). One mechanism by which this might occur involves site-specific Tau phosphorylation by multiple protein kinases activated by AβOs. We have shown that AβOs induce ectopic neuronal cell cycle re-entry, which ironically leads to neuron death, by a mechanism that requires Tau phosphorylation at Y18, S262, S409 and S416 by fyn, mTORC1 (probably indirectly through S6 kinase), protein kinase A and CaMKII, respectively [21,44]. It is therefore possible that phosphorylation at some or all of those sites provokes the conformational change from compact and folded to unfolded.

Regardless of what the mechanism may be, the finding that AβOs cause Tau unfolding emphasizes that biochemical effects of AβOs on Tau, such as phosphorylation, are matched by changes in the physical structure of Tau. It follows naturally that detection of molecular species that block or reduce AβO-induced Tau unfolding might aid discovery of new diagnostic biomarkers and disease-modifying drugs for AD.

Supplementary Material

Supp. Fig. 1

Supplementary Figure 1. Venus-Tau-Teal is not significantly proteolyzed in cultured neurons. Primary mouse cortical neurons that expressed Venus-Tau-Teal were analyzed by western blotting with antibodies to Tau (Tau5) and GFP/Teal (N86).

Supp. Fig. 2

Supplementary Figure 2. Modulation of fluorescence lifetime by perturbation of microtubules with taxol. Venus-Tau-Teal expressing CV-1 cells were treated on stage for 1 hour with 1 μM taxol. Pre- and post-treatment lifetime distributions analyzed in Prism7 with the Kolmogrov-Smirnov test (a nonparametric test that compares a distribution to a reference distribution) shows a histogram of a single experiment (left side). Peak lifetime bar graphs (right side) show multiple experimental replicates. Statistically significant, but small magnitude lifetime increases were observed in 3 of 4 experiments, each of which includes data from 6-9 fields of view. Error bars indicate standard errors of the mean.

Supp. Fig. 3

Supplementary Figure 3. Okadaic Acid treatment of CV-1 cells drives an increase in tau phosphorylation at S262 of Venus-Tau-Teal and an increase in pan-S/T phosphorylation. Western blots (left) and quantification normalized to total Tau (right) are shown. The pS/T blot shows proteins in the electrophoretic mobility range of tau. Statistical significance was assessed by t-test. Error bars indicate standard errors of the mean for n=3 replicates.

Supp. Fig. 4

Supplementary Figure 4. Analysis of Tau and Aβ oligomerization by western blotting. Tau (2N4R) and Aβ_1-42 monomers and oligomers were prepared as described in the Materials and Methods section.