Protein Disulfide Isomerase Interacts with Tau Protein and Inhibits Its Fibrillization

Background Tau protein is implicated in the pathogenesis of neurodegenerative disorders such as tauopathies including Alzheimer disease, and Tau fibrillization is thought to be related to neuronal toxicity. Physiological inhibitors of Tau fibrillization hold promise for developing new strategies for treatment of Alzheimer disease. Because protein disulfide isomerase (PDI) is both an enzyme and a chaperone, and implicated in neuroprotection against Alzheimer disease, we want to know whether PDI can prevent Tau fibrillization. In this study, we have investigated the interaction between PDI and Tau protein and the effect of PDI on Tau fibrillization. Methodology/Principal Findings As evidenced by co-immunoprecipitation and confocal laser scanning microscopy, human PDI interacts and co-locates with some endogenous human Tau on the endoplasmic reticulum of undifferentiated SH-SY5Y neuroblastoma cells. The results from isothermal titration calorimetry show that one full-length human PDI binds to one full-length human Tau (or human Tau fragment Tau244–372) monomer with moderate, micromolar affinity at physiological pH and near physiological ionic strength. As revealed by thioflavin T binding assays, Sarkosyl-insoluble SDS-PAGE, and transmission electron microscopy, full-length human PDI remarkably inhibits both steps of nucleation and elongation of Tau244–372 fibrillization in a concentration-dependent manner. Furthermore, we find that two molecules of the a-domain of human PDI interact with one Tau244–372 molecule with sub-micromolar affinity, and inhibit both steps of nucleation and elongation of Tau244–372 fibrillization more strongly than full-length human PDI. Conclusions/Significance We demonstrate for the first time that human PDI binds to Tau protein mainly through its thioredoxin-like catalytic domain a, forming a 1∶1 complex and preventing Tau misfolding. Our findings suggest that PDI could act as a physiological inhibitor of Tau fibrillization, and have applications for developing novel strategies for treatment and early diagnosis of Alzheimer disease.


Introduction
Tau protein is implicated in the pathogenesis of neurodegenerative disorders such as tauopathies including Alzheimer disease, and Tau fibrillization is thought to be related to neuronal toxicity [1][2][3]. There is no efficient treatment available for Alzheimer disease so far, and the mechanism of Alzheimer disease is still unclear [4,5]. Thus on one hand, the characterization of factors regulating Tau fibrillization is of great importance to clarify the etiology of Alzheimer disease and to assist in the establishment of medical treatment [4,5]. On the other hand, physiological inhibitors for Tau fibrillization, such as molecular chaperone heat shock protein 70 (Hsp70) [6,7], hold promise for development of novel strategies for treatment and early diagnosis of Alzheimer disease.
Tau protein does not adopt the compact folded structure typical of most cytosolic proteins because of its hydrophilic character [8]. Instead, Tau protein adopts a natively unfolded or intrinsically disordered structure in solution [9,10]. Tau protein consists of two distinct domains, the projection domain and the microtubule binding domain, and four or three imperfect repeats (R1, R2, R3, and R4) make up the microtubule binding domain [10,11]. Tau 244-372 contains the four-repeat microtubule binding domain forming the core of bundles of filaments in Alzheimer disease, and can form fibrils with the help of heparin in vitro in a relatively short time [4,[12][13][14][15]. Thus it is a frequently used model for Tau fibrillization [16].
The endoplasmic reticulum (ER), a central player in cell physiology, not only is an important organelle for protein folding and protein posttranslational modification, but also acts as a significant intracellular calcium store [17]. The accumulation of misfolded proteins and Ca 2+ influx can cause ER stress in neurons, and severe ER stress can induce apoptosis [18,19]. However, the ER can withstand relatively mild insults through the expression of molecular chaperones such as Hsp70 and protein disulfide isomerase (PDI) [20].
PDI is both an enzyme and a chaperone [21] primarily located at the ER [22], and implicated in neuroprotection against multiple neurodegenerative diseases including Alzheimer disease [23]. PDI is composed of four thioredoxin domains a, b, b', and a', an xlinker between b' and a', and a C-terminal acidic tail [24][25][26][27]. Both the a and a' domains are thioredoxin-like catalytic domains containing a CGHC active site, and are capable of catalyzing simple oxidoreductions and thiol-disulfide exchange reactions [28]. The b' domain provides the main peptide binding site. When PDI catalyzes a complex disulfide isomerization associated with protein folding (involves both thiol/disulfide chemistry and substantive conformational change in the substrate), all the four domains of PDI are required to function synergistically, combining the chaperone activity with the foldase activity [27,[29][30][31].
As an important chaperone induced by ER stress, PDI is believed to accelerate the folding of disulfide-bonded proteins by catalyzing the disulfide interchange reaction [22]. Recently, multiple neurodegenerative diseases including Alzheimer disease are shown to be associated with an upregulation of PDI expression level [20,23,32,33]. Furthermore, PDI and paired helical filament-Tau are co-located in neurofibrillary tangles in the brain of patients with Alzheimer disease [34,35]. Because PDI is both an enzyme and a chaperone, and implicated in neuroprotection against Alzheimer disease as described above, we want to know whether PDI can prevent Tau fibrillization.
In the present study, by using co-immunoprecipitation and confocal laser scanning microscopy and several biophysical methods, such as thioflavin T (ThT) binding assays, isothermal titration calorimetry (ITC), Sarkosyl-insoluble SDS-PAGE, and transmission electron microscopy (TEM), we investigated the interaction of full-length human PDI with human Tau protein as well as the effect of human PDI on the fibrillization of human Tau fragment Tau 244-372 under physiological conditions. We found that PDI interacted with some endogenous Tau protein in the cell lines and showed co-localization in the ER. Our biophysical analysis revealed that one full-length PDI bound to one full-length Tau monomer with micromolar affinity at physiological pH. We also found that PDI (especially thioredoxin-like catalytic domain a) inhibited nucleation and elongation of Tau 244-372 fibrillization. We thus suggest that PDI could act as a physiological inhibitor of Tau fibrillization, protecting ER from the harmful effect of misfolded Tau protein. Our findings can be used to develop novel strategies for treatment of Alzheimer disease.

Ethics statement
All research involving original human work was approved by the Institutional Review Board of the College of Life Sciences, Wuhan University (Wuhan, China), leaded by Dr. Hong-Bing Shu, the Dean of the college, in accordance with the guidelines for the protection of human subjects. Written informed consent for the original human work that produced the plasmid samples was obtained.

Materials
Heparin (average MW = 6 kDa) and ThT were purchased from Sigma-Aldrich (St. Louis, MO). The crowding agent Ficoll 70 was also obtained from Sigma-Aldrich. Dithiothreitol (DTT), Sarkosyl and sodium dodecyl sulfate (SDS) were purchased from Amresco (Solon, OH). Ni Sepharose TM high performance was obtained from GE Healthcare (Uppsala, Sweden), and SP Sepharose Fast Flow was Amersham Biosciences products (Uppsala, Sweden). All other chemicals used were made in China and were of analytical grade.

Plasmids and proteins
All of the recombinant human PDI proteins (full-length human PDI and its fragments abb'xa', bb'x, a'c and a) plasmids were kindly provided by Prof. Lloyd W. Ruddock (Department of Biochemistry, University of Oulu, Oulu, Finland). The cDNA encoding human Tau fragment Tau 244-372 was amplified using the plasmid for human Tau40 (kindly provided by Dr. Michel Goedert) as a template. The PCR-amplified Tau 244-372 was subcloned into pRK172 vector. The sequences of all the constructs were verified by DNA sequencing. Recombinant full-length human Tau and Tau 244-372 were expressed in Escherichia coli and purified to homogeneity by SP-Sepharose chromatography as described [4,36]. Purified Tau protein was analyzed by SDS-PAGE with one band and confirmed by mass spectrometry. The concentrations of full-length human Tau and Tau 244-372 were determined according to their absorbance at 214 nm with a standard curve drawn by bovine serum albumin.
Recombinant full-length human PDI and its fragments were expressed in Escherichia coli BL21 (DE3) pLysS strain and purified to homogeneity by Ni Sepharose chromatography as described [37]. The resulting proteins all included an N-terminal His tag (MHHHHHHM). Protein concentrations of full-length human PDI and its fragments were determined by the Bradford method with bovine serum albumin as a standard.
Cell culture and transfection SH-SY5Y neuroblastoma cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum (FBS), 100 U/ml penicillin, and 100 U/ml streptomycin in 5% CO 2 at 37uC. SH-SY5Y cells were transiently transfected with HA-tagged full-length human PDI in pcDNA 3.1 vector using LipofectamineH 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol and harvested 36 h after transfection.

Co-immunoprecipitation (Co-IP) assays
Approximately 36 h after cell transfection, SH-SY5Y cells were rinsed three times in ice-cold PBS buffer, followed by incubation with 1 ml of ice-cold IP lysis buffer (Beyotime, Haimen, China) containing a serine protease inhibitor PMSF for 30 min at 4uC. Samples were pre-cleared with Protein A+G agarose (Beyotime). Pre-cleared lysates were then incubated with mouse monoclonal anti-HA antibody (Sigma-Aldrich) at 4uC overnight. A 25% slurry of Protein A+G agarose was added into the lysates, incubated for 2 h at 4uC, and washed with ice-cold IP lysis buffer (Beyotime). The pellet was re-suspended in SDS loading buffer, boiled for 10 min, and then centrifugated at 17,000 g for 1 min with an Eppendorf 5810R centrifuge (Eppendorf AG, Hamburg, Germany). The supernatant was removed and separated by 15% SDS-PAGE, and transferred onto a polyvinylidene difluoride membrane (Millipore, Bedford, MA). Membrane blots were probed with mouse anti-Tau (TAU-5) monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and visualized by enhanced chemiluminescent.

Western blot
The separated proteins were transblotted onto a polyvinylidene difluoride membrane at 200 mA for 1 h. The membranes were then blocked with Tris-buffered saline containing 0.05% Tween-20 and 5% dry milk, and incubated sequentially with primary and secondary antibodies dissolved in the same solution for 1 h at room temperature. Endogenous human Tau was recognized by TAU-5 firstly and overexpressed human PDI was recognized by anti-HA antibody secondly. Mouse monoclonal anti-HA antibody (Sigma-Aldrich) was used at 1:5,000 dilution, TAU-5 (Santa Cruz) was used at 1:200 dilution, and horseradish peroxidase-conjugated goat anti-mouse IgG (Beyotime) was used at 1:1,000 dilution. The protein recognized by the antibody was visualized using BeyoECL Plus kit (Beyotime), according to the manufacturer's instructions.

Immunofluorescence microscopy
For staining, SH-SY5Y cells were fixed with 4% paraformaldehyde for 30 min at room temperature, followed by incubation with 0.25% Triton X-100 for 5 min and blocked with 5% bovine serum albumin for 30 min at room temperature. The cells were then incubated with primary antibody, rabbit polyclonal anti-HA antibody (Proteintech Group, Chicago, IL) or TAU-5 (Santa Cruz), at 37uC for 2 h, followed by incubation with fluorescent secondary antibody, Alexa Fluor 488 or Alexa Fluor 350 (Beyotime), at 37uC for 1 h. Finally, the ER of the cells was stained with ER-Tracker Red (Beyotime) at 37uC for 30 min. Images were captured using an Olympus FluoView FV1000 laser scanning confocal microscope (Tokyo, Japan). Green fluorescence was detected using the 488-nm laser line of an Argon laser, blue fluorescence was detected using the 405-nm laser line of a 405-30 Diode laser, and red fluorescence was detected using the 559-nm laser line of the HeNe laser. The polymerization for Tau 244-372 in 96-well plates was set up by a mixture of 10 mM Tau protein, 2.5 mM heparin, 20 mM ThT and 0-20 mM human PDI either in the absence or in the presence of 100 g/l Ficoll 70 in 10 mM HEPES buffer containing 1 mM DTT and 100 mM NaCl (pH 7.4). The reaction components were mixed quickly and immediately read for 10 h at 37uC in SpectraMax M2 microplate reader (Molecular Devices, Sunnyvale, CA) using excitation at 440 nm and emission at 480 nm with a wavelength cut off at 475 nm. Each sample was run in triplicate. Kinetic parameters were determined by fitting ThT fluorescence intensity versus time to a sigmoidal equation [4,[38][39][40]:

Thioflavin T binding assays
where F is the fluorescence intensity, k is the rate constant for the growth of fibrils, and t m is the time to 50% of maximal fluorescence. The initial baseline during the lag time is described by F 0 . The final baseline after the growth phase has ended is described by A+ct. The lag time is calculated as t m 2 2/k.

Seeding experiments
Mature fibrils formed by Tau 244-372 were sonicated on ice using a probe sonicator for 15 s (interval of 3 s, 200 W) to produce the seeds used in the seeding experiments. The kinetics of fibril formation of samples with 0.5% (v/v) preformed seed fibrils were mixed quickly and immediately read for 10 h at 37uC in SpectraMax M2 microplate reader as described above. Each sample was run in triplicate.

Sarkosyl-insoluble Tau SDS-PAGE
The Sarkosyl-insoluble Tau experiments were carried out according to the method as described [4]. Tau polymerization was set up by incubating a mixture of 10 mM Tau 244-372 , 2.5 mM heparin, and 0-20 mM human PDI in 10 mM HEPES buffer containing 1 mM DTT and 100 mM NaCl (pH 7.4) at 37uC without agitation. Aliquots (100 ml) of assembly mixtures were taken out and added into 500 ml of 10 mM HEPES buffer (pH 7.4) containing 100 mM NaCl and 1% Sarkosyl. The mixture was left at room temperature for 30 min and then centrifugated on an Optima LE-80K ultracentrifuge (Beckman Coulter, Fullerton, CA) at 150,000 g for 30 min. The supernatant (sarkosyl-soluble Tau) was removed, and the pellet (sarkosyl-insoluble Tau) was resuspended in 50 ml of SDS sample buffer containing 5% 2mercaptoethanol and subjected to 15% SDS-PAGE. After the electrophoresis the gels were stained with silver.

Transmission electron microscopy
The formation of filaments by human Tau fragment was confirmed by electron microscopy of negatively stained samples. 10 mM Tau 244-372 was incubated with 0-20 mM human PDI in 10 mM HEPES buffer (pH 7.4) containing 100 mM NaCl, 2.5 mM heparin, and 1 mM DTT at 37uC for 600 or 240 min. Sample aliquots of 10 ml were placed on carbon-coated copper grids, and left at room temperature for 1-2 min, rinsed with H 2 O twice, and then stained with 2% (w/v) uranyl acetate for another 1-2 min. The stained samples were examined using an H-8100 (or an H-7000 FA) transmission electron microscope (Hitachi, Tokyo, Japan) operating at 100 kV or an FEI Tecnai G2 20 transmission electron microscope (Hillsboro, OR) operating at 200 kV.

Isothermal titration calorimetry
ITC experiments on the interaction of full-length human PDI and its fragments with human Tau fragment Tau 244-372 (or fulllength human Tau) in the absence of DTT were carried out at 25.0 and 37.0uC using a VP-ITC titration calorimetry (MicroCal, Northampton, MA). Freshly purified human PDI and freshly purified human Tau were dialyzed against 10 mM HEPES buffer (pH 7.4) containing 100 mM NaCl for three times at 4uC. A solution of 8.1 mM PDI (or 3.24 mM PDI) was loaded into the sample cell (1.43 ml), and a solution of 200 mM Tau 244-372 (or 40 mM full-length human Tau) was placed in the injection syringe (300 ml). The first injection (5 ml) of 200 mM Tau 244-372 (or 40 mM full-length human Tau) was followed by 29 injections of 10 ml (or 19 injections of 15 ml). Dilution heats of human Tau were measured by injecting human Tau solution into buffer alone and were subtracted from the experimental curves prior to data analysis. The stirring rate was 300 rpm. The resulting data were fitted to a single set of identical sites model using MicroCal ORIGIN software supplied with the instrument, and the standard molar enthalpy change for the binding,D b H 0 m , the dissociation constant, K d , and the binding stoichiometry, n, were thus obtained. The standard molar free energy change,D b G 0 m , and the standard molar entropy change, D b S 0 m , for the binding reaction were calculated by the fundamental equations of thermodynamics [4,12,13]: All ITC experiments were repeated three times. The experiments were pretty reproducible. Every time human Tau possessed a single set of identical sites for human PDI, although the thermodynamic parameters (K d , D b H 0 m and n) were slightly different in different batches.

Full-length human PDI Interacted with some endogenous human Tau in vitro
In this study, we confirmed the interaction of PDI with endogenous Tau in SH-SY5Y cells by co-immunoprecipitation analysis. As shown in Fig. 1, anti-HA antibody was able to pull down endogenous human Tau, which was detected by immunoblotting with TAU-5, and overexpressed human PDI was also detected by Western blot with anti-HA antibody. By contrast, coimmunoprecipitated endogenous human Tau was not detected when non-specific IgG antibody was used in immunoprecipitation ( Fig. 1). HA-tagged full-length human PDI co-immunoprecipated with endogenous human Tau, indicating that human PDI and human Tau interacted in SH-SY5Y cells.
PDI has been reported to be localized specifically in the lumen of the ER [22]. At first, we performed control experiments to observe the localization of endogenous human Tau in SH-SY5Y cells by using double immunofluorescence. Confocal microscopy image of endogenous human Tau in SH-SY5Y cells in the absence of overexpressed human PDI is shown in Figure S1. Clearly, endogenous human Tau was mainly localized in the nucleus of undifferentiated SH-SY5Y cells and the remaining endogenous human Tau was localized in the cytosol (Fig. S1B), which is similar to the previous observations [41,42]. Superimposing the two colors blue and red (Merge) resulted in a magenta signal with a blue core ( Fig. S1D), indicating that endogenous human Tau was localized both in the nucleus and on the ER of undifferentiated SH-SY5Y cells in the absence of overexpressed human PDI. Immunocytochemical assays were then used to observe the subcellular locations of overexpressed PDI as well as its co-location with cellular endogenous Tau in SH-SY5Y cells. Fig. 2 shows confocal microscopy images of immunofluorescence staining for endogenous human Tau (blue) and human PDI (green) in SH-SY5Y cells with TAU-5 and anti-HA antibody, respectively, as well as fluorescence staining for the ER (red) of the cells with ER-Tracker Red. Superimposing the two colors blue and red (Merge) resulted in a strong magenta signal with a blue core (Fig. 2D), indicating that endogenous human Tau was localized both in the nucleus and on the ER of undifferentiated SH-SY5Y cells when human PDI was overexpressed. Superimposing the three colors green, blue, and red (Merge) resulted in a strong light cyan signal (Fig. 2D), indicating co-localization of full-length human PDI with some endogenous human Tau on the ER of undifferentiated SH-SY5Y cells. Therefore, as evidenced by co-immunoprecipitation and confocal laser scanning microscopy (Figs. 1, 2, and S1), human PDI interacted and co-located with some endogenous human Tau on the ER of undifferentiated SH-SY5Y cells.

Thermodynamics of the binding of human PDI to human Tau
ITC provides a direct route to the complete thermodynamic characterization of non-covalent, equilibrium interactions [4,12,13,43], and DTT concentrations as low as 1 mM can cause severe baseline artifacts due to background oxidation during the titration. Therefore ITC was used to measure the binding affinity of human PDI to human Tau monomer in the absence of DTT. ITC profiles for the binding of full-length human PDI and its fragments to human Tau fragment Tau  for 37.0uC). The calorimetric data were best fit to a model assuming a single set of identical sites. The thermodynamic parameters for the binding of full-length human PDI and its fragments to Tau 244-372 are summarized in Table 1. As shown in Table 1, at physiological pH, one full-length human PDI bound to one Tau 244-372 molecule with a dissociation constant of 3.46 mM at 25.0uC and 4.83 mM at 37.0uC. The binding affinity of the four thioredoxin domains abb'xa' of human PDI to Tau 244-372 monomer was moderately lower than that of full-length human PDI, with a dissociation constant of 7.35 mM, but the binding affinity of the a-domain of human PDI to Tau 244-372 monomer was significantly higher than that of full-length human PDI, with a dissociation constant of 0.48 mM at 25.0uC and 1.34 mM at 37.0uC (Table 1). No binding reaction for domain combinations bb'x and a'c with Tau 244-372 monomer was detected by ITC ( Fig.  S2A-2D and Table 1), demonstrating that the a-domain of human  PDI is very important for the interaction of human PDI with human Tau at physiological pH.
So far, our ITC experiments were conducted with human Tau fragment Tau 244-372 , but since endogenous human Tau is fulllength human Tau it is important to demonstrate the thermodynamic behavior of full-length human Tau bound by full-length human PDI. Tau 244-372 encompasses the four microtubule binding repeats [4,[12][13][14][15][16], which does exclude a number of domains that could potentially interact with PDI. ITC profiles for the binding of full-length human PDI to full-length human Tau at 25.0uC are also shown in Fig. 3. Fig. 3D shows raw ITC curves resulting from the injections of full-length human Tau into a solution of full-length human PDI. The titration curves show that full-length human PDI binding to full-length human Tau is also exothermic, resulting in negative peaks in the plots of power versus time. Fig. 3H shows the plots of the heat evolved per mole of fulllength human Tau added, corrected for the heat of full-length human Tau dilution, against the molar ratio of human Tau to human PDI. The thermodynamic parameters for the interaction between full-length human PDI and full-length human Tau (Table 1) were obtained by fitting the data to a single set of identical sites model, indicating that one full-length human PDI bound to one full-length human Tau molecule with a moderate affinity in the absence of DTT. The binding affinity of full-length human PDI to full-length human Tau resembles that of full-length human PDI to Tau 244-372 ( Table 1), suggesting that full-length human PDI binds to full-length human Tau mainly through the segment 244-372 of human Tau at physiological pH.
As shown in Table 1, the formation of full-length human PDI-Tau complex was driven by moderately favorable enthalpy decreases but with a large (or moderately) unfavorable entropy decrease for full-length human Tau (or Tau 244-372 ). However, the formation of PDI domain combination abb'xa'-Tau complex was driven by moderately favorable increases in entropy in combina-  tion with moderately favorable enthalpy decreases, and the formation of the a-domain of human PDI-Tau complex was driven by large favorable increases in entropy in combination with moderately favorable enthalpy decreases (Table 1). Our data indicate that deletion of the entire acidic C-terminal domain of human PDI has significant effect on the ''driving force'' for the formation of human PDI-Tau complex at physiological pH, although this flexible C-terminal tail has been determined to be not essential for chaperone and enzymatic activities of human PDI [44].
Taken together, our ITC data demonstrated that one full-length human PDI bound to one full-length human Tau (or human Tau fragment Tau 244-372 ) monomer with moderate, micromolar affinity at physiological pH and near physiological ionic strength, and that two molecules of the a-domain of human PDI interacted with one Tau 244-372 molecule with sub-micromolar affinity. Our data also indicated that at physiological pH, human PDI bound to Tau protein mainly through its thioredoxin-like catalytic domain a, forming a 1:1 complex.

The presence of human PDI influenced Tau aggregation
The enhanced fluorescence emission of the dye ThT has been frequently used for monitoring the kinetics of amyloid fibril formation of Tau protein and prion protein, which is a specific marker for the b-sheet conformation of fibril structures [4,[13][14][15]39,40]. Neuronal cells normally have a reducing environment maintained by an excess of glutathione [1,4,11,12]. In this study, DTT, a strong reducing agent, was used to mimic the reducing environment present in normal neuronal cells and block the formation of an intramolecular disulfide bond. In this study, human PDI at physiological concentrations was incubated separately with physiological concentrations of recombinant human Tau fragment Tau 244-372 at physiological pH and in the presence of 1 mM DTT, and the effects of full-length human PDI and its fragments on kinetics of Tau 244-372 fibrillization were examined by ThT binding assays (Fig. 4), as a function of PDI concentration. As shown in Fig. 4, A and B, the addition of 10-20 mM full-length human PDI or domain combination abb'xa' did inhibit Tau 244-372 fibrillization at physiological pH, accompanied by a remarkable decline of the maximum ThT intensity. When the concentration of PDI went higher, the kinetic curves of Tau 244-372 fibrillization went down gradually, with a longer lag phase.
To elucidate the mechanisms of the inhibitory effect of human PDI on Tau 244-372 fibrillization, a sigmoidal equation [4,[38][39][40] was used to fit the kinetic data, yielding three kinetic parameters k, t m , and the lag time, which are summarized in Table 2. As shown in Table 2, full-length human PDI and its fragments abb'xa' remarkably inhibited both steps of nucleation and elongation of Tau 244-372 fibrillization in a concentration-dependent manner, resulting in a lag time of 242 min (or 247 min) in the presence of 20 mM full-length human PDI (or 20 mM abb'xa'), which is 30% increased compared to that in the absence of PDI (184 min), and a rate constant for the growth of fibrils of 22.4610 23 min 21 (or 23.3610 23 min 21 ), which is 40% decreased compared to that in the absence of PDI (36.7610 23 min 21 ). The addition of 20 mM full-length human PDI or domain combination abb'xa' significantly decreased the amount of Tau 244-372 filaments represented by the maximum ThT intensity (Fig. 4, A and B). Furthermore, we found that the a-domain of human PDI inhibited both steps of nucleation and elongation of Tau 244-372 fibrillization more strongly than full-length human PDI (Table 2 and Fig. 4D), resulting in a lag time of 253 min in the presence of 20 mM domain a, which is 40% increased compared to that in the absence of PDI, and a rate constant for the growth of fibrils of 17.6610 23 min 21 , which is 52% decreased compared to that in the absence of PDI. The addition of 20 mM domain a strongly decreased the amount of Tau 244-372 filaments represented by the maximum ThT intensity (Fig. 4D), but domain combinations bb'x and a'c of human PDI had no obvious inhibitory effect on the amount of fibrils and fibrillization kinetics of Tau 244-372 in reducing conditions (Table 2 and Fig. 4C), demonstrating that the a-domain of human PDI is very important for the inhibitory effect of human PDI on Tau fibrillization at physiological pH. Our data indicated that human PDI prevented abnormal Tau aggregation mainly through its thioredoxin-like catalytic domain a.
In order to check whether human PDI has an inhibitory effect on Tau fibrillization in the presence of seeds, we carried out seeding experiments. As shown in Figs. S4 and 4, in the absence of PDI, the addition of seeds significantly reduced the lag time of Tau 244-372 fibrillization. The addition of 20 mM full-length human PDI remarkably inhibited both steps of nucleation and elongation Table 1. Thermodynamic parameters for the binding of full-length human PDI and its fragments to human Tau fragment Tau 244-372 (or full-length human Tau) as determined by ITC at 25.0uC. Thermodynamic parameters, K d , D b H 0 m and n, were determined using a single set of identical sites model. The standard molar binding free energy (D b G 0 m ) and the standard molar binding entropy (D b S 0 m ) for the binding reaction were calculated using Equations 2 and 3 respectively. The buffer used was 10 mM HEPES buffer (pH 7.4) containing 100 mM NaCl. Errors shown are standard errors of the mean. NB, no binding observed in the present conditions. a Thermodynamic parameters for the binding of full-length human PDI and its a-domain to Tau 244-372 as determined by ITC at 37.0uC. b Thermodynamic parameters for the binding of full-length human PDI to full-length human Tau as determined by ITC at 25.0uC. doi:10.1371/journal.pone.0076657.t001 of seed-induced Tau 244-372 fibrillization (Fig. S4). Furthermore, we found that the a-domain of human PDI (20 mM) not only inhibited both steps of seed-induced Tau 244-372 fibrillization more strongly than full-length human PDI, but also significantly decreased the amount of Tau 244-372 filaments represented by the maximum ThT intensity (Fig. S4). Clearly, human PDI did have an inhibitory effect on Tau fibrillization in the presence of seeds. This is probably due to the binding between Tau seeds and human PDI, so that the amyloid elongation via lateral association of Tau seeds/ monomer was blocked.
Ficoll 70 is widely accepted as a perfect model for the principal crowding components in living cells where the folding and misfolding of proteins take place, because its interaction with proteins can be described using pure excluded-volume models [40,45]. In this study, the effect of such a macromolecular crowding agent on human Tau filament formation suppressed by human PDI under reducing conditions was examined by ThT binding assays ( Table 2). As shown in Table 2, the presence of 100 g/l Ficoll 70 in the reaction systems significantly accelerated amyloid fibril formation of Tau 244-372 in the absence of PDI, almost without a lag phase. However, 20 mM full-length human PDI strongly inhibited both steps of nucleation and elongation of Tau 244-372 fibrillization in the presence of 100 g/l Ficoll 70, resulting in a lag time of 44.8 min, which is infinite increase compared with that in the absence of PDI (0 min), and a rate constant for the growth of fibrils of 31.4610 23 min 21 , which is 43% decreased compared to that in the absence of PDI (54.9610 23 min 21 ) ( Table 2). The above results indicated that the inhibitory effect of human PDI on the nucleation step of Tau fibrillization under crowded physiological conditions is much stronger than that in dilute solutions.

The amount of Tau filaments present in the solution measured by centrifugation assays
ThT fluorescence is not perfectly specific for amyloid fibrils and, depending on the particular protein and experimental conditions, assays may render both false positive (spectroscopic change upon binding to non-fibrillar material) and false negative results (its fluorescence not being affected by some amyloid fibrils) [14]. Considering this, we investigated the correlation between the spectroscopic signal monitored and the amount of Tau filaments present in the solution measured by centrifugation assays. In order to semi-quantify the increase/decrease of Tau filaments in the presence of human PDI, we carried out Sarkosyl-insoluble SDS-PAGE experiments after centrifugation assays. As shown in Fig. 5,  A, B, and E, a clear band corresponding to Sarkosyl-insoluble Tau 244-372 filaments was observed when Tau 244-372 was incubated in the absence of a physiological inhibitor human PDI for 200 min, while the Sarkosyl-insoluble Tau band was clearly observed when Tau 244-372 was incubated with 20 mM full-length human PDI for a remarkably longer time (300 min) or incubated with 20 mM domain a for a much longer time (400 min). Furthermore, when Tau 244-372 was incubated for 300/400/ 600 min, the intensity of the Sarkosyl-insoluble Tau band in the presence of 20 mM full-length human PDI was remarkably lower than that in the absence of PDI, and the intensity of the Sarkosylinsoluble Tau band in the presence of 20 mM domain a was much lower than that in the absence of PDI (Fig. 5, A, B, and E). All Sarkosyl-insoluble SDS-PAGE experiments were repeated three times in order to get a more quantitative result. Fig. S5 shows the quantitative characterization of time-dependent SDS-PAGE of Sarkosyl-insoluble Tau 244-372 in the presence of human PDI, compared with that in the absence of PDI. As shown in Fig. S5, A, B, and C, the addition of 20 mM full-length human PDI or 20 mM the a-domain resulted in a lag time of Tau 244-372 fibrillization remarkably longer than that in the absence of PDI. These findings further support the observations mentioned above that the addition of 20 mM full-length human PDI did inhibit Tau 244-372 fibrillization at physiological pH, and that the a-domain of human PDI inhibited Tau 244-372 fibrillization more strongly than fulllength human PDI. However, domain combinations bb'x and a'c of human PDI had no obvious inhibitory effect on Tau 244-372 fibrillization under reducing conditions (Fig. 5, C and D), further demonstrating that the a-domain of human PDI is important for the inhibitory effect of human PDI on Tau fibrillization at physiological pH.
Characterization of the morphology of Tau protein aggregates formed in the presence of human PDI TEM was employed to characterize the morphology of Tau protein aggregates formed in the absence and in the presence of human PDI. Fig. 6 shows TEM images of Tau 244-372 samples incubated with 20 mM human PDI under reducing conditions and at 37uC for 600 min (the final stage). For Tau 244-372, the addition of 1 mM full-length human PDI or 1 mM the a-domain had no significant effect on the morphology of Tau samples, and long and branched fibrils were observed in all samples (Fig. 6, A, B, M, and  N). In the presence of 10 mM full-length human PDI or 10 mM the a-domain, however, some short filaments were observed when Tau samples were incubated for 600 min (Fig. 6, C and O). The amount of fibrils formed by Tau 244-372 in the presence of 10 mM full-length human PDI or 10 mM the a-domain (Fig. 6, C and O) appears to be markedly less than that in the absence of PDI (Fig. 6, A and M) on the same time scale. In the presence of 20 mM fulllength human PDI (or 20 mM the a-domain), only a few short amyloid fibrils (or almost no fibrils) were observed when Tau samples were incubated for 600 min (Fig. 6, D and P). As expected, for Tau 244-372, the addition of 10-20 mM domain combination bb'x or 10-20 mM a'c had no significant effect on the morphology of Tau samples, and long and branched fibrils were   Aliquots were taken at 0, 120, 200, 250, 300, 400, and 600 min, respectively, and then incubated with HEPES buffer containing 1% Sarkosyl followed by centrifuging at 150,000 g for 30 min. Pellets were re-suspended with sample buffer containing 5% 2-mercaptoethanol and subjected to 15% SDS-PAGE. Gels were stained with silver. doi:10.1371/journal.pone.0076657.g005 observed in all samples (Fig. 6, E-L). Furthermore, we semiquantitatively analyzed the inhibitory effect of PDI on Tau fibrillization by using TEM. Fig. 7 shows TEM images of Tau 244-372 samples incubated with 20 mM human PDI under reducing conditions and at 37uC for 240 min (the early stage). In the absence of PDI, some short amyloid fibrils accompanied by a few long fibrils were observed when Tau samples were incubated for 240 min (Fig. 7A). In the presence of 20 mM full-length human PDI, however, some shorter and thinner protofibrils were observed when Tau samples were incubated for 240 min (Fig. 7B). In the presence of 20 mM the a-domain, no fibrils were observed when Tau samples were incubated for 240 min (Fig. 7C). These findings once again support the observations that the addition of 10-20 mM full-length human PDI did inhibit Tau 244-372 fibrillization at physiological pH, and that the a-domain of human PDI inhibited Tau 244-372 fibrillization more strongly than full-length human PDI.

Characterization of the effect of human PDI on Tau fibrillization at different stages
We then checked whether the inhibitory effect of human PDI on Tau fibrillization was dependent on the incubation time or different compositions of Tau protein. We first added 20 mM fulllength human PDI at time zero when Tau 244-372 was pure monomers. As shown in Fig. 8A, full-length human PDI remarkably inhibited both steps of nucleation and elongation of Tau 244-372 fibrillization under such conditions. We then added 20 mM full-length human PDI after incubation for 190 s, 10 min, and 20 min, when Tau 244-372 was oligomers/protofibrils, fibrils, and mature fibrils, respectively. As shown in Fig. 8, B-D, full-length human PDI added at 190 s in the lag phase, 10 min in the growth phase, or at 20 min in the final equilibrium phase, did not inhibit fibril formation of Tau 244-372 under reducing conditions. Particularly, when human PDI was added after incubation of human Tau for 190 s, which there should be some Tau oligomers (seeds) formed in solution, no inhibition effect was observed (Fig. 8B). The above results suggested that the inhibitory effect on fibrillization kinetics of Tau by full-length human PDI was due to its direct interaction with Tau protein. Such an interaction was solidly demonstrated by co-immunoprecipitation ( Fig. 1), confocal laser scanning microscopy (Fig. 2), and isothermal titration calorimetry (Figs. 3 and S3). It is this interaction that decelerated the nucleation of Tau aggregation. The reason why PDI reduced Tau seeds fibrillization but not Tau oligomers fibrillization is unknown.

Discussion
A hallmark of tauopathies including Alzheimer disease is accumulation of misfolded Tau protein within neurons, forming neurofibrillary tangles and leading to cellular dysfunction and cell death [1][2][3]. Amyloid fibrils associated with Alzheimer disease [46][47][48][49][50] can be considered biologically relevant failures of cellular quality control mechanisms including molecular chaperones, proteolysis, autophagy, and proteasomes [51,52]. As an important molecular chaperone induced by ER stress, PDI is found to colocate with neurofibrillary tangles in Alzheimer disease patient brain tissue [34,35] and is believed to prevent the neurotoxicity associated with ER stress and protein misfolding [20,23]. However, the connection between PDI function and Tau protein misfolding during ER stress remains enigmatic. Our Co-IP and confocal laser scanning microscopy data provided direct evidence for the interaction between human PDI and human Tau in neuronal cells. We demonstrated for the first time that human PDI interacted and co-located with some endogenous human Tau on the ER of SH-SY5Y cells. Our data here clearly showed that at physiological pH, one human PDI molecule bound to one human Tau molecule with moderate, micromolar affinity, and thus remarkably inhibited both steps of nucleation and elongation of Tau protein misfolding in a concentration-dependent manner. Therefore, human PDI formed a 1:1 complex with human Tau protein and prevented abnormal Tau aggregation under physiological conditions.
In this paper, DTT was not used due to its background noise. We then discussed the possible influence of the absence of DTT toward ITC measurement results. We have previously demonstrated that the extended Tau monomers are analyzed in our ITC measurements, which are made in the absence of DTT [13]. It is  possible that the binding affinity of human PDI to Tau monomer in the presence of DTT is higher than that of human PDI to Tau monomer in the absence of DTT.
During neurodegenerative disorders and cerebral ischaemia, the accumulation of immature and denatured proteins or misfolded proteins in the ER results in a cellular condition known as ER stress [19,20,53], which triggers an adaptive program called the unfolded protein response [23,54,55]. An upregulation of PDI during ER stress [56][57][58] associated with multiple neurodegenerative diseases including Alzheimer disease [20,23,32,33,58,59] represents an adaptive response to protect neuronal cells. PDI family members have been reported to prevent abnormal aggregation of several proteins/peptides associated with neurodegenerative diseases through a direct interaction between PDI and these amyloidogenic proteins, including prion proteins [60,61], asynuclein [62], amyloid b peptides [63], FUS [64], and copper, zinc superoxide dismutase [59]. Our results here indicate that human PDI, a key component of the protein quality control system in the ER [22,62], can directly interact with human Tau on the ER, and could counteract the toxicity caused by abnormal Tau aggregation in the ER by binding to Tau protein and inhibiting Tau fibrillization. These results, coupled with some previous studies [23,[59][60][61][62][63][64], highlight the potent effectiveness of PDI in preventing abnormal aggregation of amyloidogenic proteins at the various stages of their aggregation pathway. Therefore, PDI family members can reduce general ER stress levels in protein misfolding disorders as a part of the unfolded protein response [23,54,55], prevent the misfolding of amyloidogenic proteins through a direct interaction [23,[59][60][61][62][63][64], and prevent neurotoxicity associated with ER stress and protein misfolding [20,23,33,[59][60][61].
Both domains a and a' of human PDI contain an active site CGHC that is responsible for the oxidoreductase activity [27,28]. Isolated a and a' domains have been shown to effectively catalyze the introduction of disulfides into protein substrates [65]. It has been reported that domain a' is very important for human PDI to inhibit a-synuclein fibril formation [62]. Very recently, it has been found that non-native disulfides are formed early in the folding pathway and can trigger misfolding, but the a-domain of human PDI favors native disulfides by catalyzing oxidation at a late stage of folding [66]. Furthermore, the molecular flexibility of the adomain of PDI is demonstrated to be essential for the enzymatic activity in vitro and in vivo [67]. In the present study, we found that two molecules of the a-domain of human PDI interacted with one Tau molecule with sub-micromolar affinity, and inhibited both steps of nucleation and elongation of Tau fibrillization more strongly than full-length human PDI. In other words, the adomain is very important for the inhibitory effect of human PDI on Tau fibrillization under physiological reducing conditions. Our results indicated that human PDI bound to human Tau mainly through its thioredoxin-like catalytic domain a, forming a 1:1 complex and preventing abnormal Tau aggregation. We had hoped to purify domain a' to homogeneity by using Ni Sepharose chromatography, but unfortunately we did not succeed. Although no binding reaction for domain combination a'c with Tau monomer was detected by ITC and a'c had no obvious inhibitory effect on Tau fibrillization under physiological reducing condi- tions, the possibility that domain a' interacts with human Tau and inhibits Tau fibrillization to some extent cannot be excluded.
Very recently, it has been demonstrated that operational plasticity enables molecular chaperone Hsp104 to disaggregate diverse amyloid fibrils, including those formed by full-length human Tau and Tau 244-372 [68]. Our data here suggest that human PDI could act as a physiological inhibitor of Tau fibrillization, protecting ER from the harmful effect of misfolded Tau protein. Our findings are highly relevant to the emerging viewpoint that PDI action is intimately involved in the prevention of in vivo amyloid fibril formation [23,[59][60][61][62][63][64]. Physiological inhibitors for Tau fibrillization, such as molecular chaperones Hsp70 [6,7] and PDI (this work), and molecular chaperones disaggregating Tau filaments, such as Hsp104 [68], hold promise for development of novel strategies for treatment and early diagnosis of Alzheimer disease.
In conclusion we have shown that: (i) full-length human PDI Interacted with endogenous human Tau in SH-SY5Y cells; (ii) fulllength human PDI co-located with some endogenous human Tau on the ER of SH-SY5Y cells; (iii) at physiological pH, one human PDI bound to one human Tau monomer with moderate, micromolar affinity and mainly through its thioredoxin-like catalytic domain a; (iv) human PDI prevented abnormal Tau aggregation also mainly through its thioredoxin-like catalytic domain a; (v) the inhibitory effect on fibrillization kinetics of human Tau by human PDI was mainly due to its direct interaction with Tau protein. The above findings can be used to develop novel strategies for treatment of Alzheimer disease. This study provides clues to understand the role of PDI in preventing Tau protein misfolding during ER stress, and offers new insights into pathogenesis, placing ER stress onto a generic pathophysiology for Alzheimer disease and holding the promise to help solve Alzheimer disease. Figure S1 The localization of endogenous human Tau in SH-SY5Y cells detected by double immunofluorescencecontrol experiments. Confocal microscopy image of immunofluorescence staining for endogenous human Tau (blue) in SH-SY5Y cells with TAU-5 (B) in the absence of overexpressed human PDI (C). Endoplasmic reticulum (ER) (red) was stained with ER-Tracker Red (A). Double labeling of SH-SY5Y cells (Merge) resulted in a magenta signal (D). The scale bars represent 10 mm. (DOC) Figure S2 ITC profiles for the binding of human PDI fragments to Tau 244-372 at 25.06C. The top panels represent the raw data for sequential injections of 214 mM Tau 244-372 into 8.1 mM human PDI fragment bb'x (A) and sequential injections of 336 mM Tau 244-372 into 16.2 mM human PDI fragment a'c (B) in 10 mM HEPES buffer (pH 7.4) containing 100 mM NaCl, respectively. The first injection (5 ml) was followed by 29 injections of 10 ml. The bottom panels (C and D) show the plots of the heat evolved (kcal) per mole of Tau 244-372 added, corrected for the heat of Tau 244-372 dilution, against the molar ratio of Tau 244-372 to human PDI fragment. The data (solid squares) were too small to be fitted, indicating that no binding was observed in the conditions used. (DOC) Figure S3 ITC profiles for the binding of full-length human PDI and its a-domain to Tau 244-372 at 37.06C. The top panels represent the raw data for sequential injections of 200 mM Tau 244-372 into 8.1 mM full-length human PDI (A) and its fragment a (B) in 10 mM HEPES buffer (pH 7.4) containing 100 mM NaCl, respectively. The first injection (5 ml) was followed by 29 injections of 10 ml. The bottom panels (C and D) show the plots of the heat evolved (kcal) per mole of Tau 244-372 added, corrected for the heat of Tau 244-372 dilution, against the molar ratio of Tau 244-372 to human PDI fragment. The data (solid squares) were best fitted to a single set of identical sites model and the solid lines represent the best fit.