Protein disulfide isomerase dissolves and detoxifies oligomeric assemblies of amyloid beta peptide

Abstract

Aggregated amyloid beta peptide (Aβ) contributes to Alzheimer's disease through neurotoxic effects and a prion‐like mode of transmission. We report that protein disulfide isomerase (PDI) exhibits disaggregase activity against oligomeric but not fibrillar forms of Aβ. PDI did not bind monomeric Aβ, indicating its highly effective inhibition of fibril formation occurs through reversal of early‐stage oligomers rather than prevention of the initial aggregate. Cells exposed to both PDI and oligomeric Aβ were protected from Aβ‐induced toxicity. An S‐nitrosylated form of PDI that is associated with neurodegeneration could not bind to oligomeric Aβ, thereby eliminating its neuroprotective disaggregase activity. Our observations suggest PDI could be used both physiologically and therapeutically to dissolve the oligomeric forms of Aβ.

The neuroprotective chaperone protein disulfide isomerase (PDI) specifically recognizes aggregated forms of the Aβ peptide. Its interaction with oligomeric but not fibrillar Aβ results in dissolution and detoxification of the aggregate. Contact with oligomeric Aβ induced the partial unfolding of PDI, a phenomenon that is functionally linked to its disaggregase activity.

Abbreviations

AD, Alzheimer's disease

Aβ, amyloid beta peptide

BSA, bovine serum albumin

CT, cholera toxin

EDC, 1‐ethyl‐3‐[3‐dimethylaminopropyl]carbodiimide hydrochloride

ELISA, enzyme‐linked immunosorbent assay

FTIR, Fourier transform infrared.

HFIP, 1,1,1,3,3,3‐Hexafluoro‐2‐propanol

HRP, horseradish peroxidase.

NHS, N‐hydroxysuccinimide

PDI, protein disulfide isomerase

Q3R, quercetin‐3‐rutinoside

RU, resonance units

SNO‐PDI, S‐nitrosylated PDI

SPR, surface plasmon resonance

ThT, Thioflavin‐T

Alzheimer's disease (AD) is a neurodegenerative disorder characterized by the deposition of extracellular amyloid plaques in the brain [1, 2]. Amyloid beta (Aβ) is a major component of the senile plaque. Proteolytic nicking of a membrane‐anchored amyloid precursor protein by β‐ and γ‐secretases generates the soluble Aβ peptide [3, 4]. Aβ is an aggregation‐prone peptide that progresses from monomers to a heterogeneous collection of oligomers, including intermediate species such as protofibrils, and then to insoluble fibrils which form the senile plaque [5, 6]. In the human brain, this peptide occurs in multiple forms that vary in the number of amino acid residues. The 40‐ and 42‐residue peptides (Aβ_1‐40, Aβ_1‐42) are the most prevalent forms found in AD brains, with the early, oligomeric forms of aggregated Aβ_1‐42 representing a major toxic species of the protein [2, 7]. Early and intermediate forms of aggregated Aβ may also self‐propagate by a prion‐like “seed” mechanism that stimulates further aggregation and disease progression [8, 9]. Several therapeutic strategies have accordingly focused on the clearance of oligomeric Aβ as a potential treatment for AD [2, 10, 11, 12].

Protein disulfide isomerase (PDI) can prevent the aggregation of misfolded proteins, including amyloid‐forming proteins such as Aβ or α‐synuclein [13, 14, 15, 16]. Its overexpression during cellular stress is thought to represent a neuroprotective mechanism [17, 18, 19], and loss of PDI function through S‐nitrosylation is frequently linked to amyloid formation and neurodegeneration [20, 21, 22]. Furthermore, PDI—often specifically identified as S‐nitrosylated PDI (SNO‐PDI)—is embedded in amyloid plaques isolated from brains afflicted with AD or Parkinson's disease [23, 24, 25]. These collective observations suggest PDI plays a physiological role in preventing amyloid formation and could possibly be used as a therapeutic to limit amyloid‐induced neurodegeneration.

We previously reported that PDI can dissolve nascent but not mature amyloid fibrils of α‐synuclein [26]. Here, we examined the potential disaggregase activity of PDI against Aβ. PDI recognized aggregated but not monomeric Aβ and could prevent fibril formation at extremely low sub‐stoichiometric molar ratios of PDI : Aβ. Contact with an early‐stage aggregate of Aβ led to the partial unfolding of PDI, a phenomenon that is functionally linked to its disaggregase activity [27, 28]. Consistent with this observation, we found that PDI could dissolve and detoxify the early, oligomeric forms of Aβ. It could not, however, break down Thioflavin‐T (ThT)‐positive fibrils of Aβ. SNO‐PDI could not bind to aggregated Aβ, which eliminated its neuroprotective disaggregation of oligomeric Aβ. Our work reports a new physiological mechanism for the reversal of early‐stage Aβ aggregation by PDI. This process suggests a novel therapeutic approach to AD treatment in which PDI selectively targets oligomeric Aβ for disaggregation, resulting in the loss of both toxicity and prion‐like transmission.

Materials and methods

Materials

Protease inhibitor cocktail, H31L21 rabbit anti‐Aβ monoclonal antibody, AMY‐33 mouse anti‐Aβ monoclonal antibody, and 1‐ethyl‐3‐[3‐dimethylaminopropyl]carbodiimide hydrochloride (EDC) were purchased from ThermoFisher Scientific (Waltham, MA, USA). Horseradish peroxidase (HRP)‐labeled AMY‐33 was generated with the HRP conjugation kit from Abcam (Cambridge, UK), while the HRP‐conjugated goat anti‐rabbit IgG antibody was purchased from Jackson ImmunoResearch (West Grove, PA, USA). 1,1,1,3,3,3‐Hexafluoro‐2‐propanol (HFIP) was purchased from Sigma‐Aldrich (St. Louis, MO, USA). N‐hydroxysuccinimide (NHS) was purchased from Pierce Biotechnology Inc. (Rockford, IL, USA). ThT was purchased from Anaspec Inc. (Fremont, CA, USA).

Lyophilized Aβ_1‐42 (Innovagen, Lund, Sweden) was dissolved in HFIP at 222 μm concentration and aliquoted to microcentrifuge tubes for evaporation of the HFIP. Dried samples were stored at −20 °C and reconstituted for use in 25 mm sodium phosphate (pH 7.4) with 25 mm NaCl (NaCl/phosphate buffer) unless otherwise noted.

Adherent PC12 cells (PC‐12 Adh, catalog #CRL‐1721.1, RRID:CVCL_F659) were obtained from the American Type Culture Collection (Manassas, VA, USA) and maintained in Ham's F‐12 K medium (Gibco Life Technologies, Carlsbad, CA, USA) supplemented with 2.5% fetal bovine serum (R&D Systems, Minneapolis, MN, USA), 15% horse serum (Corning, Corning NY, USA), and antibiotic‐antimycotic (Gibco Life Technologies, Waltham, MA, USA). They were split to 96‐well plates at 80% confluency for toxicity assays in serum‐free medium.

Production of recombinant PDI, SNO‐PDI, and ¹³C‐labeled PDI

Human PDI with an N‐terminal His₆ tag was purified from Escherichia coli BL21(DE3)pLysS transformed with the pOLR130 plasmid [29]. A 5 mL overnight starter culture in M9 minimal medium was expanded to 500 mL and grown at 37 °C until an OD₆₀₀ of ~ 0.7 was reached. PDI expression was then induced with IPTG and continued growth at 25 °C for 18 h. Cell pellets collected after centrifugation were frozen at −80 °C before resuspension in 5 mL of lysis buffer [50 mm Tris (pH 8.0), 150 mm NaCl, 0.1% Triton X‐100, 1% deoxycholic acid, 100 μg·mL⁻¹ lysozyme, and protease inhibitor cocktail] and incubation for 30 min at room temperature with shaking. The suspension was then sonicated on ice with a Branson Digital Sonifier, using 60 s intervals between three bursts. After centrifugation, the clarified lysate was added to washed TALON metal affinity resin (Takara Bio, San Jose, CA) and incubated with rotation for 18 h at 4 °C. The resin was then pelleted, resuspended in wash buffer [50 mm Tris (pH 7.4), 300 mm NaCl, and 5 mm imidazole] with rocking for 10 min, and transferred to a gravity column. PDI was eluted from the column with sequential 2 mL volumes of 10 mm sodium borate (pH 7.4) containing increasing concentrations of imidazole: 10, 20, 40, 60, 100, and 300 mm twice. Samples from each imidazole elution were resolved by SDS/PAGE and visualized with Coomassie stain to assess sample purity. The appropriate fractions were pooled and dialyzed against 10 mm sodium borate buffer (pH 7.4). Finally, single‐use aliquots were transferred to microcentrifuge tubes and frozen overnight at −80 °C before lyophilization and long‐term storage at −80 °C.

Uniformly ¹³C‐labeled PDI was purified in an identical manner, except uniformly ¹³C‐labeled ¹³C₆‐D‐glucose (Cambridge Isotope Laboratories, Andover, MA, USA) was used as the sole carbon source in the minimal medium.

To generate SNO‐PDI, a 100 mm solution of S‐nitrosocysteine was generated by mixing 100 mm L‐cysteine and 100 mm NaNO₂ in 5% (v/v) 10 N HCl. The solution was shielded from light and stored at −80 °C. A final concentration of 1 mm S‐nitrosocysteine was added to 1 mg·mL⁻¹ of PDI for 30 min at room temperature for the S‐nitrosylation of PDI. Single‐use aliquots of SNO‐PDI were stored at −80 °C.

Detection of Aβ fibrils by ThT fluorescence

Working stocks of 50 μm Aβ were passed through 0.45 μm syringe filters that had been pre‐wet with NaCl/phosphate buffer. The Aβ stocks were then supplemented with ThT (20 μm final concentration) and vortexed for 5 s. PDI, SNO‐PDI, BSA, or an equivalent volume of buffer alone were subsequently added to separate Aβ preparations as indicated. All preparations were aliquoted in 150 μL volumes to the wells of a Corning 96‐well black‐walled plate that was placed on a LabDoctor orbital shaker (MidSci, Fenton, MO, USA) at 37 °C with 500 rpm shaking. Fluorescence measurements were recorded using a BioTek (Winooski, VT, USA) Synergy2 plate reader with 440 nm excitation and 482 nm emission wavelengths. Measurements taken at 0 h were background‐subtracted from all subsequent readings, which were then expressed as percentages of the end‐point ThT signal from aggregated Aβ alone.

Surface plasmon resonance (SPR) binding assays

Working stocks of 50 μm Aβ were reconstituted in 10 mm MES buffer (pH 7.0) containing 200 mm NaCl. Samples aliquoted to microcentrifuge tubes were incubated at 37 °C for 0 h (monomer), 4 h (oligomer), or 24 h (fibril) with 500 rpm agitation using a LabDoctor orbital shaker.

Both analyte and reference channels of a Nicoya (Kitchener, ON, Canada) OpenSPR carboxy sensor were conditioned with a 250 μL injection of 10 mm HCl at a flow rate of 150 μL·min⁻¹. All subsequent steps used 250 μL injections and a flow rate of 20 μL·min⁻¹ for both channels unless otherwise stated. After conditioning, the sensor was activated with an injection of 200 mm EDC and 100 mm NHS. 50 μm samples of monomeric, oligomeric, or fibrillar Aβ were then perfused over the analyte channel alone for coupling to the activated sensor surface. This was followed by two blocking injections of 2.5% BSA in phosphate‐buffered saline (pH 7.4) with 0.05% Tween‐20. The time course was initiated with an injection of 100 μg·mL⁻¹ of PDI for approximately 250 s. To ensure monomeric Aβ had been captured on the sensor, the PDI perfusion was followed with an injection of the AMY‐33 anti‐Aβ antibody at a 1 : 1000 dilution.

Detection of oligomeric Aβ by enzyme‐linked immunosorbent assay (ELISA)

A working stock of 50 μm Aβ was passed through a 0.45 μm syringe filter that had been pre‐wet with NaCl/phosphate buffer. Samples were then added to microcentrifuge tubes and incubated at 37 °C for 2 h with 500 rpm agitation using a LabDoctor orbital shaker. After 2 h of agitation, the samples were supplemented with various concentrations of PDI, 5 μm SNO‐PDI, 5 μm BSA, or an equivalent volume of buffer alone. The agitation then continued for another 2 h before sample processing for ELISA as described above. Samples of Aβ that had undergone 0 or 2 h of agitation were prepared in parallel for synchronous addition to the ELISA plate with samples that had undergone 4 h of agitation.

The AMY‐33 anti‐Aβ monoclonal antibody was placed in 0.2 m Na₂HCO₃ (pH 9.6) at a 2 μg·mL⁻¹ concentration and added in 100 μL volumes to the wells of a 96‐well plate for an overnight incubation at 4 °C. The plate was then washed three times with PBST and blocked with 1% BSA in PBS for 1 h at room temperature. After removal of the block, Aβ samples were added to the plate in 150 μL volumes for a 2 h incubation at room temperature. This was followed by 3 washes with PBST and the addition of 100 μL HRP‐conjugated AMY‐33 antibody (500 ng·mL⁻¹ in PBS with 1% BSA) for 1 h at room temperature. Extensive washing with PBST then preceded the addition of 100 μL TMB substrate and, after about 5 min, 100 μL of 20% H₂SO₄ stop solution. Absorbance was read at 450 nm on a BioTek Synergy 2 plate reader. Four replicate wells were used for each condition. Values obtained from antibody‐coated wells incubated with monomeric Aβ were background‐subtracted from all experimental values, which were then expressed as percentages of the Aβ sample that had undergone 2 h of aggregation.

Detection of oligomeric Aβ by differential centrifugation

A working stock of 50 μm Aβ was vortexed for 1 min before transfer to a capped 4 × 4 mm quartz cuvette (Starna Cells, Atascadero, CA, USA) and rotation in a 50 mL conical tube on a Fisherbrand multi‐purpose rotator at speed 6. After 2 h of rotation, the samples were supplemented with 5 μm PDI, 5 μm SNO‐PDI, or an equivalent volume of buffer alone. The rotation then continued for another 2 h. Each sample was subsequently transferred to a microcentrifuge tube and spun at 16 000×  g for 10 min. Pellet fractions were resuspended in 100 μL of Laemmli SDS‐sample buffer, and 20 μL of each supernatant fraction was resuspended in 20 μL of Laemmli SDS‐sample buffer. One‐tenth of each sample was resolved by SDS/PAGE with 10% polyacrylamide gels and probed by western blot using the rabbit anti‐Aβ monoclonal antibody H31L21 at a 1 : 5000 dilution for an overnight incubation at 4 °C, followed by a 30 min room temperature incubation with a 1 : 20 000 dilution of an HRP‐conjugated goat anti‐rabbit IgG antibody.

Cytotoxicity assay

A working stock of 50 μm Aβ was vortexed for 1 min and transferred to a 4 × 4 mm quartz cuvette that was capped and placed in a 50 mL conical tube for 2 h of circular rotation using a Fisherbrand multi‐purpose rotator at speed 6. This was followed by the addition of 5 μm PDI, 5 μm SNO‐PDI, or an equivalent volume of buffer alone. Samples were then allowed to rotate for another 2 h before their addition to PC12 cells at final protein concentrations of 10 μm Aβ and 1 μm PDI or SNO‐PDI in tissue culture medium. An MTS cell viability assay was conducted after 50 h of incubation with Aβ or, as a positive control for extensive cell death, after 10 min of incubation in medium containing 10% DMSO. Cells incubated in medium alone were used to establish maximum cell viability.

Isotope‐edited Fourier transform infrared (FTIR) spectroscopy

Lyophilized Aβ was resuspended to 50 μm in D₂O‐based buffer containing 20 mm sodium borate (pD 7.4). To initiate aggregation, Aβ was vortexed for 1 min, transferred to a 4 × 4 mm quartz cuvette that was capped and rotated continuously on a Fisherbrand multi‐purpose rotator at speed 6. After 0 and 4 h of aggregation, the sample was divided into two 75 μL volumes. One was used directly to record the FTIR spectrum of Aβ alone, and the other was used to reconstitute a purified, lyophilized sample of uniformly ¹³C‐labeled PDI at a final concentration of 5 μm. FTIR spectra of Aβ and/or uniformly ¹³C‐labeled PDI were recorded at 25 °C using a FT/IR‐4X FTIR spectrometer (Jasco, Tokyo, Japan). Spectra were smoothed with a 15‐point Savitzky–Golay filter and baseline‐corrected from 1700 to 1574 cm⁻¹. Peak area of the combined protein spectra was normalized to the spectral sum of the two proteins measured separately. Three independent experiments were conducted and averaged. Wavenumbers for spectral peaks in the replicate experiments did not vary by more than 3 cm⁻¹.

Results

PDI inhibits the formation of Aβ fibrils

Aβ aggregation was initially monitored with ThT, a fluorophore that undergoes fluorescence enhancement and red shift upon contact with amyloid fibrils [30]. Aβ_1‐42 was used for our studies, as this variant represents a toxic form of the peptide that is associated with AD [31, 32]. A 1 : 10 molar ratio of PDI : Aβ completely blocked the formation of Aβ fibrils (Fig. 1A). Titration experiments demonstrated that a 1 : 1000 molar ratio of PDI : Aβ strongly inhibited the aggregation of Aβ, with partial inhibition still occurring at a 1 : 2000 molar ratio of PDI : Aβ (Fig. 1B). These observations were consistent with a previous report on the potent inhibition of Aβ aggregation by PDI [15]. An irrelevant protein (bovine serum albumin; BSA) had no effect on Aβ fibrillization at a 1 : 10 molar ratio of BSA : Aβ. Thus, the dose‐dependent disruption of Aβ fibrillization was specific to PDI.

Fig. 1.

Inhibition of Aβ aggregation by sub‐stoichiometric quantities of PDI. To initiate fibril formation at time 0, a solution containing Aβ and ThT was vortexed for 5 s and aliquoted in 150 μL volumes to a 96‐well plate that was subsequently placed at 37 °C with shaking. (A) Fluorescence measurements were taken with a plate reader at the indicated time points. Circles represent Aβ alone; squares represent Aβ co‐incubated with a 1 : 10 molar ratio of PDI : Aβ. Values were expressed as percentages of the maximal signal from Aβ alone. (B) Fluorescence measurements were taken after a 36 h co‐incubation with the indicated molar ratios of PDI : Aβ or a 1 : 10 molar ratio of BSA : Aβ. Values for PDI‐treated samples were expressed as percentages of the signal recorded for untreated Aβ. Data represent the means ± standard errors of at least 4 experiments with 4–6 replicate wells per condition. (C) Two identical samples of Aβ were allowed to aggregate in parallel for 10 h before a ThT measurement was taken. At this point, one sample was left untreated (gray bars). PDI was added to the other sample at a 1 : 10 molar ratio of PDI : Aβ (black bars). Both samples were incubated with shaking for an additional 26 h before a second ThT measurement was taken. Values were expressed as percentages of the maximal signal from Aβ alone. Data represent the averages ± standard deviations of 3 independent experiments with five replicate samples per condition.

Although PDI effectively and efficiently inhibits Aβ aggregation, we found it could not break down existing ThT‐positive amyloid fibrils (Fig. 1C). For this experiment, ThT measurements were taken 10 h after the initiation of Aβ aggregation. PDI was then added to the sample, and aggregation was allowed to continue for another 26 h before a second ThT measurement was recorded. PDI did not eliminate the ThT signal present at 10 h, but it did prevent further aggregation from occurring. PDI can thus inhibit fibril formation, but it cannot dissolve existing Aβ fibrils. This observation was consistent with a previous report documenting the inability of PDI to dissolve end‐stage fibrils of Aβ [15].

PDI recognizes the aggregated forms of Aβ

Binding assays involving SPR were used to examine which forms of Aβ are recognized by PDI (Fig. 2). Monomeric or aggregated samples of Aβ were chemically coupled to a SPR sensor, and a baseline measurement of 0 resonance units (RU) was established for the Aβ‐coated sensor. PDI was then injected over the sensor at time 0. We found that PDI did not bind to monomeric Aβ and confirmed that Aβ was on the sensor through the elevated RU signal resulting from injection of an antibody against Aβ (Fig. 2A). It is possible that the lack of PDI binding to an Aβ monomer was due to steric hindrance of the immobilized protein, but we believe this is unlikely when considering (a) SPR is a standard technique to detect protein–protein interactions; (b) the immobilized Aβ monomer was accessible to the AMY‐33 monoclonal antibody; and (c) PDI produced strong RU responses for fibrillar Aβ (Fig. 2B) and oligomeric Aβ (Fig. 2C) with the same binding conditions used for monomeric Aβ. Binding to oligomeric Aβ was blocked when PDI was inactivated by S‐nitrosylation, exposed to the intramolecular cross‐linker EDC, or treated with the PDI inhibitor quercetin‐3‐rutinoside (Q3R) (Fig. 2D). These collective observations documented the specificity of PDI binding to aggregated forms of Aβ. The lack of interaction between EDC‐treated PDI and Aβ further suggested that PDI requires a degree of conformational flexibility for binding to oligomeric Aβ.

Fig. 2.

PDI specifically recognizes aggregated Aβ. Amide coupling anchored (A) monomeric Aβ, (B) a 24 h fibrillar aggregate of Aβ, or (C, D) a 2 h oligomeric aggregate of Aβ to a SPR sensor slide. A baseline measurement corresponding to the mass of sensor‐bound Aβ established the 0 RU signal. The time course was then initiated by injecting PDI over the Aβ‐coated slide. (A) PDI was removed from the perfusion buffer after 400 s and was replaced with an antibody against Aβ. (B–D) PDI was removed from the perfusion buffer after ~ 250 s. In panel (D), oligomeric Aβ was exposed to additional PDI samples that had been S‐nitrosylated (SNO), treated with the intramolecular cross‐linker EDC, or incubated with the PDI inhibitor Q3R. One of two representative experiments is shown for each panel.

PDI breaks down the oligomeric form of Aβ

PDI blocks amyloid formation but does not bind to the Aβ monomer and does not dissolve amyloid fibrils. We therefore predicted that PDI breaks down the oligomeric, ThT‐negative aggregates of Aβ that form early in the aggregation process. This possibility was examined with an ELISA to identify the early‐stage aggregates of Aβ that are not detected by ThT [33]. The ELISA uses a monoclonal antibody that only recognizes one site on each molecule of Aβ. This antibody is adsorbed to a plate where it can capture Aβ. The same antibody, conjugated to HRP, is then added in soluble form to the Aβ‐coated plate. If Aβ is present on the plate as a monomer, it will not be recognized by the HRP‐conjugated antibody because the single antibody recognition site for Aβ is already occupied by the plate‐adsorbed antibody. However, an oligomer or larger aggregate of Aβ will have additional sites for antibody recognition that will result in a positive signal. The ELISA can thus detect the transition of aggregated Aβ to monomeric Aβ (i.e., disaggregation) through loss of the initial signal generated by oligomeric Aβ. With this assay, we found that PDI can dissolve oligomeric Aβ when added 2 h after the onset of aggregation (Fig. 3A). Substantial dissolution was accomplished with a molar ratio of PDI : Aβ as low as 1 : 1000, while a minor disaggregation effect was still detected with a 1 : 5000 molar ratio of PDI : Aβ. SNO‐PDI exhibited weak disaggregase activity and BSA exhibited no disaggregase activity when present at 1 : 10 molar ratios with Aβ. A functional form of PDI was therefore required for the effective reversal of Aβ oligomerization.

Fig. 3.

PDI reverses the oligomerization of Aβ. (A) After 2 h of aggregation, samples of Aβ were supplemented with PDI or SNO‐PDI at the indicated molar ratios. An additional condition involved the addition of BSA to aggregated Aβ at a 1 : 10 molar ratio of BSA : Aβ. Aggregation then continued for another 2 h before sample processing by ELISA. Values from the 4 h samples were expressed as percentages of an Aβ sample that had undergone 2 h of aggregation. Data represent the means ± standard errors of at least 4 independent experiments with four replicate samples per condition. For reference, a 24 h fibrillar aggregate of Aβ produced an ELISA signal that was 6–9 fold greater than the 2 h signal (n = 2, not shown). (B) Samples of Aβ were allowed to aggregate for 0 h (lane 1), 2 h (lane 2), or 4 h (lanes 3–5). PDI or SNO‐PDI was added to the samples in lanes 4 and 5, respectively, after the first 2 h of aggregation. All samples were separated into pellet and supernatant fractions with a 10 min 16 000× g spin at 4 °C. Using the rabbit anti‐Aβ monoclonal antibody H31L21, the fractions were probed by SDS/PAGE with western blot for the presence of Aβ. Signals from both the pellet and supernatant fractions were detected around the 4.5 kDa mobility expected for Aβ_1‐42. One of two representative experiments is shown.

The PDI‐driven disaggregation of oligomeric Aβ was confirmed with an assay that uses centrifugation to separate soluble Aβ from aggregated Aβ [15, 34] (Fig. 3B). As with the ELISA, Aβ was allowed to aggregate for 2 h before the addition of PDI and continued incubation for another 2 h. A centrifugal spin then pelleted the aggregated forms of Aβ while leaving soluble Aβ in the supernatant. Both fractions were probed by western blot for the presence of Aβ. A minimal amount of Aβ was found in the pellet fraction before the onset of aggregation (lane 1), and there was a clear increase in the quantity of pelleted Aβ after 2 h of aggregation (lane 2). This pelleted material was still present after 4 h of aggregation when only Aβ was present in the buffer (lane 3). However, when PDI was added 2 h after the onset of aggregation, the amount of aggregated Aβ at the 4 h time point had reverted to the starting level seen at 0 h (lane 4). SNO‐PDI did not reduce the quantity of aggregated Aβ when added after 2 h of aggregation (lane 5). A functional form of PDI was therefore necessary to dissolve the early‐stage aggregates of Aβ.

PDI detoxifies oligomeric Aβ

The early oligomeric species of Aβ are more toxic than the Aβ fibrils that form after longer intervals of aggregation [35], so we examined the potential neuroprotective property of PDI by mixing it with a 2 h aggregate of Aβ for another 2 h of incubation before application to PC12 cells (Fig. 4). A 50 h exposure to oligomeric Aβ reduced cell viability to an average of 62% of the untreated control value, which was consistent with other PC12‐based toxicity assays [35, 36, 37, 38]. Cells incubated with a combination of Aβ and PDI exhibited a higher level of survival (86%) that represented a statistically significant difference from the cells incubated with Aβ alone. Viability after incubation with Aβ and PDI was also statistically higher than cell survival after exposure to Aβ and SNO‐PDI (75%). The extent of cell viability after incubation with Aβ and SNO‐PDI was higher than cells incubated with Aβ alone, but this was not a statistically significant difference and likely reflected the incomplete inhibition of PDI function as documented in our ELISA‐based disaggregation assay (Fig. 3A). PDI, but not SNO‐PDI, could thus dissolve and detoxify the oligomeric form of Aβ.

Fig. 4.

PDI attenuates the toxicity of oligomeric Aβ. Aβ was allowed to aggregate for 2 h before the addition of PDI or SNO‐PDI at a 1 : 10 molar ratio of PDI : Aβ. Samples were then allowed to aggregate for another 2 h before addition to PC12 cells at a final Aβ concentration of 10 μm. Cell viability was assessed with an MTS assay after 50 h of incubation, with results expressed as percentages of the value recorded for untreated control cells. Cells treated with 10% DMSO for 10 min were used as a control for extensive cell death. Data are presented as box‐and‐whisker plots representing 12 technical replicates from three independent experiments. Asterisks denote statistically significant differences at *p ≤ 0.05 or **p ≤ 0.01; ns = not significant. The DMSO control was significantly different (p < 0.0001) from all Aβ conditions. Data analysis was performed using one‐way ANOVA with the Dunnett test for multiple comparisons.

PDI unfolds upon contact with aggregated Aβ

We have proposed that the substrate‐induced unfolding of PDI allows it to function as a disaggregase: the expanded size of PDI would push between two or more proteins in a multimeric complex, thereby acting as a wedge to disperse the individual components of an aggregate [26, 28]. This represents a unique function for conditional disorder, a well‐established phenomenon in which a loss of structure occurs upon ligand binding to a protein [39, 40, 41]. Isotope‐edited FTIR spectroscopy was accordingly used to monitor the structural changes in both PDI and Aβ that result from their interaction (Fig. 5). With this technique, the spectrum of a uniformly ¹³C‐labeled protein can be distinguished from the spectrum of its unlabeled binding partner [42, 43]. Spectra for the individual proteins are recorded and used to generate a predicted spectrum for the combined sum of both proteins. This theoretical spectrum is then compared to the measured spectrum of both proteins together. Differences between the predicted and measured spectra are indicative of structural changes in one or both proteins that result from their interaction. Those changes can often be attributed to defined structural features in either the labeled or unlabeled protein, as each type of secondary structure is prominent at different wavenumbers (Table S1).

Fig. 5.

PDI unfolds upon contact with aggregated Aβ. FTIR spectra are shown for samples containing uniformly ¹³C‐labeled PDI (black lines) and unlabeled Aβ. Aβ was allowed to aggregate for 4 h before mixing with ¹³C‐PDI at a 1 : 10 molar ratio ¹³C‐PDI : Aβ. Spectra were recorded at (A) 5 min, (B) 30 min, and (C) 60 min after mixing the proteins. Each panel shows the spectra of Aβ alone (gray lines), ¹³C‐PDI alone (black lines), the physical combination of both proteins (blue lines), and the predicted spectral summation of the two proteins (dotted lines). The spectra of the combined samples were normalized to have the same area (integrated intensity) as the sum of the individual Aβ and ¹³C‐PDI spectra. All measurements were taken at 25 °C in D₂0‐based 20 mm sodium borate buffer (pD 7.4). Spectra represent the average of three independent experiments for each condition.

For each FTIR experiment shown in Fig. 5, measurements were recorded for uniformly ¹³C‐labeled PDI alone (black lines), Aβ alone (gray lines), and their combination at a 1 : 10 molar ratio of ¹³C‐PDI : Aβ (blue lines). In addition, a predicted spectrum for the combination of ¹³C‐PDI and Aβ was generated from the summation of the spectra for each individual protein (dotted lines). The uniform ¹³C‐labeling of PDI produced an isotope‐induced spectral downshift of ~ 45 cm⁻¹ [42, 43] that resolved its amide I band from that of unlabeled Aβ, which was allowed to aggregate for 4 h before the addition of ¹³C‐PDI. FTIR spectra were recorded at 5, 30, and 60 min post‐mixing. FTIR spectra of the individual proteins were also recorded at 5, 30, and 60 min after the initial 4 h incubation period.

As shown in Fig. 5, the spectrum of ¹³C‐PDI displayed a main peak centered around 1600 cm⁻¹, assigned to overlapping signals from α‐helix (1612–1606 cm⁻¹), irregular structure (1595 cm⁻¹), and β‐sheet (1590 cm⁻¹). The amide I spectra of early‐stage Aβ aggregates exhibited two prominent spectral features with peaks centered at 1672 cm⁻¹ and 1639 cm⁻¹, assigned respectively to (a) β‐turns and (b) overlapping signals from irregular (1647–1640 cm⁻¹) and β‐sheet (1636–1630 cm⁻¹) structures. Earlier FTIR data on Aβ_1‐42 identified main peaks at 1672 cm⁻¹ and 1647 cm⁻¹ during the initial stages (< 1 h) of incubation in buffer, indicating β‐turn and irregular conformations [44]. The latter component gradually shifted to the characteristic intermolecular β‐sheet region of 1627 cm⁻¹ upon incubation for 48 h and beyond [44]. These results are consistent with experimental and computational studies that reported the Aβ_1‐42 monomer is rich in irregular and turn secondary structures before the onset of fibrillogenesis, with four turns per monomer [45, 46]. The amide I components in the spectral regions of unlabeled Aβ and ¹³C‐PDI are revealed more clearly in the second derivative spectra of the combined samples (Fig. S1).

Immediately after mixing ¹³C‐PDI with a 4 h aggregate of Aβ, the spectrum of the combined sample displayed peaks at 1672 and 1628 cm⁻¹ in the Aβ region (Fig. 5A). The spectral sum of the two individual proteins exhibited similar features, thus indicating there was no change to the structure of aggregated Aβ upon its initial contact with ¹³C‐PDI. In the ¹³C‐PDI region, the spectrum of the combined proteins likewise showed a peak at the same wavenumber as the predicted spectrum. This indicated a lack of structural change in PDI.

Remarkably, spectra collected 30 min (Fig. 5B) and 60 min (Fig. 5C) after mixing¹³C‐PDI with a 4 h aggregate of Aβ differed significantly from the corresponding theoretical summation of the individual spectra. The measured spectrum of the combined proteins showed a sharp peak at 1632 cm⁻¹ with a shoulder in the 1650–1643 cm⁻¹ region (blue lines) in place of the broad spectral feature between 1657 and 1628 cm⁻¹ seen in the predicted spectral sum of the two proteins (dotted lines). These features can be more readily discerned in Fig. S1. In conjunction with the increased signal intensity at 1672 cm⁻¹, these spectral changes suggest Aβ transitioned from a combination of α‐helix, irregular, and intermolecular β‐sheet structures to an intramolecular β‐sheet structure. In the ¹³C‐PDI region, the predicted shoulder at 1612–1606 cm⁻¹ assigned to α‐helix strongly decreased while the signal around 1595 cm⁻¹ assigned to irregular and β‐sheet structures slightly downshifted without a compensatory increase in intensity. This result may suggest a transition in PDI from α‐helix to unordered structure, which has considerably lower absorptivity compared to α‐helix and β‐sheet structures [42]. An unordered structure with greater solvent exposure would also enhance hydrogen‐deuterium exchange, thus producing the observed spectral downshift from the initial 1600 cm⁻¹ peak for ¹³C‐PDI. Collectively, our FTIR analysis is consistent with a model where PDI undergoes partial unfolding during its interaction with Aβ oligomers and facilitates oligomer disassembly into lower order aggregates or monomers containing intramolecular H‐bonding. Appearance of this effect only at the 30 and 60 min timepoints but not immediately following combination of the two proteins suggests that the protein–protein interaction and ensuing conformational change is a slow process. Our isotope‐edited FTIR spectroscopy experiments thus suggest that PDI undergoes time‐dependent partial unfolding after binding to early‐stage Aβ assemblies, and that this interaction facilitates a conformational transition in Aβ from intermolecular to intramolecular β‐sheet enriched species.

Discussion

Chaperones that function through conditional disorder and chaperones that function as a disaggregase have been described [39, 47, 48, 49], but neither property has previously been attributed to PDI. Furthermore, no disaggregase is known to act by conditional disorder. We have established the general phenomenon of conditional disorder as a disaggregase mechanism for the PDI‐driven disassembly of the multimeric cholera toxin (CT) [27, 28] and have since documented the disaggregation of nascent α‐synuclein fibrils by PDI [26]. Yet PDI does not function as a disaggregase against all substrates—it will, for example, inhibit but not reverse the aggregation of malate dehydrogenase [26] in a process that does not involve the unfolding of PDI (H. Burress and K. Teter, unpublished observations). Here, we recorded the partial unfolding of PDI that occurs upon contact with oligomeric Aβ. This phenomenon of conditional disorder corresponded to the dissolution and detoxification of oligomeric Aβ by PDI.

Three distinct techniques directly documented the dissolution of oligomeric Aβ by PDI: (a) an ELISA using the same monoclonal antibody for both capture and detection of Aβ; (b) differential centrifugation to separate soluble Aβ from the aggregated forms of Aβ; and (c) isotope‐edited FTIR spectroscopy to monitor the structural changes in Aβ that accompanied its conversion from the oligomeric to monomeric state. The consistency of our observations across these methods, along with controls demonstrating the specificity of interactions between a functional form of PDI and aggregated Aβ, strengthens our conclusion that PDI exhibits a disaggregase activity against oligomeric Aβ. This activity is consistent with the neuroprotective role of PDI [17, 20, 21, 22] that, as shown here, can neutralize the toxicity of oligomeric Aβ.

We hypothesize that the substrate‐induced unfolding of PDI allows it to function as a chaperone with disaggregase activity: by unfolding in the presence of an aggregation‐prone substrate, PDI would act as a wedge to dislodge individual proteins from the aggregate. The ability of PDI to repeatedly cycle between folded and unfolded states [28, 50] allows it to effectively disrupt Aβ aggregation at extremely low sub‐stoichiometric molar ratios [15] (Figs 1B and 3A). This ATP‐independent event occurs by a physical rather than enzymatic process, as the unfolding of PDI and resulting increase in entropy would displace several monomers from a single aggregate, leading to an overall decrease in the free energy of the system.

A functional correlation between the substrate‐induced unfolding of PDI and its ability to break apart multimeric complexes was established for the interaction between PDI and CT: conditions that blocked the unfolding of PDI also blocked the PDI‐driven disassembly of CT [27, 28]. In those studies, we found that EDC and Q3R did not prevent PDI binding to CT but did inhibit the subsequent unfolding of PDI. Here, we found that PDI shifts to a partially unfolded state upon contact with oligomeric Aβ and dissolves those neurotoxic aggregates. At this time, however, we cannot provide additional functional correlations between PDI unfolding and its dissolution of oligomeric Aβ because EDC, Q3R, and S‐nitrosylation all inhibit the binding of PDI to oligomeric Aβ.

The disaggregase activity of PDI is effective against oligomeric assemblies of Aβ but not ThT‐positive fibrils of Aβ. Likewise, nascent but not mature amyloid fibrils of α‐synuclein are broken down by PDI [26]. The relatively inert end‐point amyloid fibrils are thought to act as a neuroprotective sink for the sequestration of toxic oligomers [4, 11], so the selective targeting of early‐stage aggregates by PDI is consistent with its role in preventing neurodegeneration. Our results have further shown that SNO‐PDI loses its ability to dissolve and detoxify oligomeric Aβ. Nitrosative stress could thus contribute to neurodegeneration [51] through the S‐nitrosylation and inactivation of PDI, thereby eliminating a barrier to the accumulation of toxic Aβ oligomers.

PDI is mainly found in the endoplasmic reticulum but also functions at other locations [52, 53]. For example, under conditions of cellular stress, PDI moves from the endoplasmic reticulum to the cytosol where it could interact with the cytosolic pool of Aβ [54, 55, 56]. It is also released into the bloodstream where it has a well‐established role in thrombosis [57]. Extracellular PDI can act as a chaperone as well [58], and it is considered a biomarker for tau pathology when detected in the cerebrospinal fluid [59]. Furthermore, PDI is packaged into neuroprotective macrosomes that are released by microglia [60] and is secreted by astrocytes to dissolve extracellular aggregates of α‐synuclein [61]. These observations suggest PDI could disrupt both intracellular and extracellular toxic oligomers of Aβ—either as a normal physiological process or as a therapeutic agent. This event would both detoxify Aβ and prevent its prion‐like transmission to other neurons. Our work thus documents a new physiological mechanism for the reversal of early‐stage Aβ aggregation that could serve as the foundation for a novel therapeutic strategy to halt the progression of Alzheimer's disease.

Author contributions

KT contributed to conceptualization. AM, AS, MCZ‐R, BAE, HL, JK, MT, SAT, KT contributed to data analysis. AM, AS, MCZ‐R, BAE, HL, JK, MT contributed to investigation. AM, AS, MCZ‐R contributed to methodology. SAT, KT contributed to supervision. AM, AS, SAT, KT contributed to writing—original draft. AM, AS, MCZ‐R, BAE, HL, MT, SAT, KT contributed to writing—review and editing.

Supporting information

Table S1. Approximate amide I band wavenumber (cm⁻¹) ranges of common secondary structures for unlabeled and uniformly ¹³C‐labeled proteins in D₂O‐based buffers.

Fig. S1. Second derivative FTIR spectra of the ¹³C‐PDI + Aβ samples from Fig. 5.

Data accessibility

Data supporting the findings of this study are available in the main figures and/or Supporting Information of the article. Original experiments are available from the corresponding author (kteter@mail.ucf.edu) upon reasonable request.