Nitration of Hsp90 induces cell death

We have shown previously that tyrosine-containing peptides prevent peroxynitrite-induced apoptosis in motor neurons and PC12 cells (4, 5), suggesting that tyrosine nitration activates cell-signaling pathways that lead to cell death. To identify potential targets of peroxynitrite able to stimulate cell death upon tyrosine nitration, we first characterized the putative nitrated proteins present in homogenates from PC12 cells after treatment with peroxynitrite by mass spectrometric analysis (Table S1). Four of the identified proteins with the highest scores were actin, tubulin, 70-kDa heat-shock protein (Hsp70), and Hsp90. We delivered these peroxynitrite-treated proteins intracellularly using the membrane-permeating agent Chariot (Active Motif) into PC12 cells and motor neurons before assaying for cell survival. Peroxynitrite-treated Hsp70, actin, and tubulin were nontoxic for motor neurons and PC12 cells (Fig. 1A). In contrast, peroxynitrite-treated Hsp90 induced death in ∼40% of PC12 cells and 60% of motor neurons (Fig. 1 A–D). Nitrated Hsp90 was toxic to the cells only when delivered intracellularly, as evidenced by its lack of effect on cell survival in the absence of Chariot (Fig. 1A). Similarly, no toxicity was observed when Hsp90 was treated with hydrogen peroxide or decomposed peroxynitrite (Fig. 1A) or with peroxynitrite in the presence of uric acid (Fig. 1B). Because uric acid does not react with peroxynitrite directly but efficiently scavenges the nitrating products of peroxynitrite decomposition (20, 21), these results suggest that nitration of Hsp90 is necessary to induce cell death. Equivalent results were obtained by counting all neurons with neurites longer than three soma diameters (Fig. 1C) or by using high-throughput image capture and analysis (Fig. 1D) (4, 22).

Hsp90 constitutes 1–2% of total cytoplasmic protein (18). The previous results strongly suggest that tyrosine nitration activates a gain of function in Hsp90, because the cells still have the endogenous unmodified chaperone after the delivery of the nitrated protein. The proportion of nitrated Hsp90 to the total amount of intracellular Hsp90 in motor neurons and PC12 cells was determined by quantitative dot blot using a monoclonal antibody against nitrated Hsp90 (5) (Fig. 2 A–C). Approximately 4% of Hsp90 was nitrated in motor neurons subjected to trophic factor deprivation. In contrast, none was detected in control cells (Fig. 2D). The proportion of peroxynitrite-treated Hsp90 released using Chariot in Fig. 1 constituted only 2–3% of the endogenous cellular Hsp90 (Fig. 2D). In PC12 cells treated with peroxynitrite, ∼4% of Hsp90 was nitrated (Fig. 2D). Intracellular release of peroxynitrite-treated Hsp90 in these proportions had no effect on proteasome activity (Fig. S1). Because cell death was induced when only a small percentage of total Hsp90 was peroxynitrite-modified Hsp90, we investigated whether this toxic gain of function could result specifically from tyrosine nitration.

Both isoforms of Hsp90 contain 24 tyrosine residues. By quantitative dot blot, we determined that only five of these residues were prone to nitration on the chaperone (Fig. 3 A–C). Mass spectrometry revealed the nitrated tyrosines to be 38, 61, 285, 493, and 605 in Hsp90α and 33, 56, 276, 484, and 596 in Hsp90β (Fig. 3D). The two first tyrosine residues were located in the amino-terminal domain of Hsp90, tyrosine 276 and 484 were located in the middle domain, and tyrosine 596 was located in the carboxy-terminal domain (Fig. 3E).

Functionally, the binding of ATP to the amino-terminal domain of Hsp90 regulates a cycle of conformational changes responsible for the chaperone activity (23). Interestingly, the two tyrosine residues located in the amino-terminal domain, Tyr38 and Tyr61 in Hsp90α and Tyr33 and Tyr56 in Hsp90β, are required for the intrinsic ATPase activity of the chaperone. Tyr61 and Tyr56 are located in the ATPase-active site cleft (Fig. 3E) (24), whereas Tyr38 and Tyr33 are necessary to stabilize the close conformation of the chaperone needed to stimulate ATP hydrolysis (25).

The localization and structural function of these tyrosine residues suggest that nitration should inhibit Hsp90 intrinsic ATPase activity. Hsp90 ATPase activity was determined using a coupled enzyme system that provides the high sensitivity needed to measure the inorganic phosphate produced by the ATP hydrolysis catalyzed by Hsp90 (26). Treatment with peroxynitrite inhibited more than 80% of geldanamycin-inhabitable Hsp90 intrinsic ATPase activity (Fig. 3F).

Two of the other three tyrosine residues prone to nitration are located in the middle domain, which also is the intermediate client-binding domain. The effect of nitration on Hsp90’s ability to interact with other proteins and prevent unfolding was determined by the citrate synthase inactivation and aggregation assay (27). Peroxynitrite treatment reduced Hsp90 capacity to prevent heat-induced unfolding of citrate synthase by only 50% (Fig. 3G), suggesting that nitrated Hsp90 still retains the ability to interact with other proteins.

In addition to tyrosine residues, peroxynitrite can cause oxidative modifications to other amino acids such as cysteine, methionine, histidine, and tryptophan (28–31) and also can lead to lysine formylation (31). To investigate whether tyrosine nitration is required for the induction of Hsp90 toxicity or other oxidative modifications are responsible for the conversion of the chaperone into a toxic protein, the five tyrosine residues prone to nitration were replaced by phenylalanine (5xPhe-Hsp90). Phenylalanine is the most conservative substitution for tyrosine and also is resistant to nitration (29). The replacement was made in the highly conserved sequence of human Hsp90β (Fig. S2A), the more abundant isoform in normal cells (16, 17). Recombinant wild-type Hsp90β was nontoxic to motor neurons when delivered with Chariot at the concentrations previously described for purified Hsp90 but became toxic when treated with peroxynitrite (Fig. 4A). In contrast, 5xPhe-Hsp90 remained nontoxic to motor neurons after incubation with peroxynitrite (Fig. 4A), demonstrating that tyrosine nitration is necessary for peroxynitrite-treated Hsp90 toxic gain of function.

The previous results reveal that nitration of Hsp90 is necessary for the toxic activity of the chaperone. However, oxidation of other residues on Hsp90 also may play a role in the toxic gain of function. To determine the relevance of tyrosine nitration in the conversion of Hsp90 into a toxic protein, we performed recombinant expression using site-specific unnatural amino acid replacement to encode nitrotyrosine genetically in the positions prone to nitration (32, 33). Five different Hsp90β proteins were produced containing a single nitrotyrosine at position 33, 56, 276, 484, or 596 as the sole oxidative modification on the protein. The incorporation of nitrotyrosine in these positions of Hsp90 was verified by Western blot (Fig. 4B) and mass spectrometry (Fig. S2 B–G). The intracellular delivery of the recombinant protein with nitrotyrosine in either position 33 or 56 induced motor neuron death, reproducing the effects of peroxynitrite on Hsp90 (Fig. 4C). In contrast, the release of Hsp90 with nitrotyrosine in position 276, 484, or 596 had no effect on motor neuron survival, but these proteins became toxic after incubation with peroxynitrite (Fig. 4C).

The role of tyrosine 33 and 56 on the toxic gain of function was confirmed by intracellular delivery of Hsp90 with tyrosine at positions 33 or 56 and phenylalanine in the other four positions. Both proteins became toxic to motor neurons after treatment with peroxynitrite (Fig. 4E). Conversely, peroxynitrite-treated Hsp90 in which tyrosine 33 and 56 had been replaced by phenylalanine had no effect on motor neuron survival (Fig. 4E). These results demonstrate that nitration of tyrosine 33 or 56 in the amino terminal domain of Hsp90 is both necessary and sufficient to induce the toxic gain of function responsible for motor neuron death.

Trophic factor deprivation-induced motor neuron apoptosis requires activation of the Fas death receptor (34, 35). The activation of Fas leads to the activation of two parallel pathways. One involves the activation of Fas-associated protein with death domain FADD and caspase 8, followed by release of cytochrome c from the mitochondria. The other pathway has been shown to be exclusive for motor neurons and involves activation of the Fas-associated protein, death-associated protein 6 (DAXX), which leads to the activation of apoptosis signal-regulating kinase 1 (Ask1) and p38 MAPK and the expression of the neuronal isoform of nitric oxide synthase (NOS), which in turn leads to production of nitric oxide and peroxynitrite (3, 34, 36, 37).

To investigate the role of the Fas pathway in the motor neuron death induced by the intracellular delivery of peroxynitrite-treated Hsp90, the cells were cultured for 24 h in the presence of the chimeric fusion protein Fas:Fc, which acts as a Fas ligand decoy competing with Fas. In the presence of Fas:Fc, motor neurons were completely protected from cell death induced by peroxynitrite-treated Hsp90 (Fig. 5A), suggesting that the Fas pathway is activated downstream of the toxic protein.

We next examined the role of the two pathways downstream of the Fas receptor, the DAXX and FADD components of the Fas pathway. After intracellular delivery of peroxynitrite-treated Hsp90, motor neurons were incubated with inhibitors of caspase 8, 9, and 3 (IETD-fmk, LEHD-fmk, and DEVD-fmk, respectively, from Sigma-Aldrich, St. Louis, MO), which are activated downstream of FADD. All the inhibitors completely prevented motor neuron death (Fig. 5B). Conversely, incubation of motor neurons with either the p38 inhibitor SB253580 (5 μM) or the NOS inhibitor l-NG-nitroarginine methyl ester (100 μM) after intracellular delivery of nitrated Hsp90 was not protective (Fig. S3 A and B). These results suggest that the DAXX component of the Fas pathway is not activated downstream of peroxynitrite-treated Hsp90.

Trophic factor deprivation in motor neurons causes inactivation of the phosphatidylinositol 3-kinase (PI3K)/Akt pathway with the subsequent dephosphorylation and translocation to the nucleus of the transcription factor Forkhead box O3a (FOXO3a), which induces the expression of Fas ligand (FasL), activating the Fas pathway (35). There were no differences in the expression of Fas receptor or FasL in motor neurons up to 16 h after intracellular delivery of either the unmodified Hsp90 or the peroxynitrite-treated chaperone (Fig. S3 C and D).

Several mechanisms control FasL expression in the cell membrane. Mobilization of lysosomes containing FasL from the Golgi network to the membrane is one such mechanism (38, 39). To investigate the possible role of the mobilization of FasL to the plasma membrane in peroxynitrite-treated Hsp90-dependent activation of the Fas pathway, motor neurons were stained with antibodies to FasL. Peroxynitrite-treated Hsp90 induced the translocation of FasL immunofluorescence from a perinuclear vesicular distribution to the plasma membrane (Fig. 5C). Further evidence for this mechanism was obtained by using N-ethylmaleimide (NEM), the inhibitor of the NEM-sensitive factor, to prevent the last steps of exocytosis (40, 41). The incubation with NEM completely prevented motor neuron death induced by peroxynitrite-treated Hsp90 (Fig. 5D). Membrane fusion during exocytosis is calcium dependent (42). The intracellular calcium chelator 1,2-Bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetrakis(acetoxymethyl ester) (BAPTA-AM) decreased the mobilization of FasL to the cell membrane (Fig. S3E) and prevented the death of motor neurons after the intracellular delivery of peroxynitrite-treated Hsp90 (Fig. 5D).

In neurons and in the immune system, the ATP-gated ion channel P2X7 is responsible for calcium influx upon activation and can initiate the early steps of exocytosis (43, 44). The P2X7 receptor is present in motor neurons (Fig. S3F), and the cells incubated with the P2X7 inhibitors Brilliant Blue G (BBG) and KN-62 were protected from apoptosis induced by peroxynitrite-treated Hsp90 (Fig. 5E). Similar results were obtained by knocking down the expression of the P2X7 receptor in motor neurons with shRNA before the intracellular delivery of the peroxynitrite-treated chaperone (Fig. 5F and Fig. S3 G and H). There were no changes in P2X7 expression upon delivery of peroxynitrite-treated Hsp90 (Fig. S3I).

Together, these results suggest that peroxynitrite-treated Hsp90 stimulates the P2X7 receptor to induce an influx of calcium that in turn mobilizes FasL-containing intracellular vesicles to transfer FasL to the cellular membrane and thereby activates cell death. Immunoprecipitation studies of P2X7 with nitrated Hsp90 further support the association of P2X7 receptors in nitrated Hsp90-induced motor neuron apoptosis (Fig. 5G).

We sought to characterize the relevance of nitrated Hsp90 in vivo. Protein nitration in motor neurons correlates with the progressive degeneration in ALS in human patients as well as in SOD-transgenic animals, whether measured by immunohistology (45–49) or by mass spectrometry (50, 51). The presence of nitrated Hsp90 in human and animal pathology was investigated using a monoclonal antibody we developed against nitrated Hsp90 (5). This antibody specifically recognizes nitrotyrosine-56 and thus identifies one of the toxic forms of nitrated Hsp90 (Fig. S4 A–C). Immunohistology with this antibody revealed intensely stained motor neurons for nitrated Hsp90 in spinal cords from sporadic ALS patients (Fig. 6 A and B) and disease-affected ALS SOD-mutant mice (Fig. 6C).

Spinal cord trauma is another condition in which the presence of nitrotyrosine is well documented (52–54). Nitrated Hsp90 immunoreactivity also was found in the spinal cords of rats 6 and 24 h following experimental spinal cord contusion injury, but the staining was absent in uninjured animals (Fig. 6D). These results show that nitrated Hsp90 is present in vivo both in the cell type more affected in conditions of chronic degeneration, such as ALS, and after acute damage of the nervous tissue.

Peroxynitrite was synthesized by rapidly mixing acidified hydrogen peroxide with sodium nitrite and quenching the reaction with sodium hydroxide as previously described (70). Residual hydrogen peroxide was eliminated with manganese dioxide, and the peroxynitrite concentration was determined spectrophotometrically using ɛ₃₀₂ = 1,700 M⁻¹⋅cm⁻¹. The concentration of peroxynitrite was measured immediately before each experiment.

PC12 cells were cultured on collagen-coated dishes in RPMI medium (Gibco/Invitrogen), supplemented with 10% (vol/vol) horse serum, 5% (vol/vol) FetalClone II (HyClone), and antibiotics. The treatment with peroxynitrite was performed as previously described (71). Briefly, the cells were plated in the same medium at a density of 1.44 × 10⁵ per 35-mm dish and incubated overnight. Following three washings with warm PBS, the cultures were incubated with peroxynitrite (0.5 mM) for 3 min in 1 mL of 50 mM phosphate buffer saline. The buffer was replaced by complete RPMI medium.

The immunoprecipitation of nitrated proteins in PC12 cell (2 x10⁷) homogenates was performed using polyclonal nitrotyrosine antibodies linked to AminoLink gel (Pierce) immediately after incubation with 0.5 mM peroxynitrite as previously described (5). Briefly, after separation of the proteins by SDS/PAGE [12% (vol/vol)], the proteins were transferred to nitrocellulose membranes for Western blot analysis or stained in gel using GelCode Blue (Pierce). Stained gels were use for mass spectrometry analysis (MALDI-TOF) of trypsinized peptides at the University of Alabama at Birmingham Mass Spectrometry Facility. The protein was identified by analysis of the tryptic fragments and comparison with the National Center for Biotechnology Information (NCBI) database. The peptide masses were entered into Mascot database search engine (www.matrixscience.com/), and the NCBI database was searched to identify the protein. Molecular weight search (MOWSE) scores above 73 were considered statistically significant. For the coimmunoprecipitation of Hsp90 and P2X7, homogenates from PC12 cells (1 mg/mL) were incubated with recombinant human Hsp90-myc-tag or with Hsp90-myc-tag with nitrotyrosine at position 56 (purified as described below) for 1 h at 4 °C. The immunoprecipitation of Hsp90-myc was performed using an anti-myc antibody conjugated to Sepharose beads (Cell Signaling Technology) according to the manufacturer’s instructions. The membrane was blotted for P2X7 (1:500; Chemicon) and myc-tag (1:2,000; Cell Signaling).

Rat embryo motor neurons were purified using a combination of cushion centrifugation and immunoaffinity and then were cultured in Neurobasal medium enriched with glutamine, glutamate, β-mercaptoethanol, and B27 supplement (Gibco-Invitrogen) as previously described (2, 34, 36, 72) in the presence or absence of brain-derived neurotrophic factor (BDNF; 1 ng/mL), glial-derived neurotrophic factor (GDNF; 0.1 ng/mL), and cardiotrophin 1 (CT-1, 10 ng/mL). More than 95% of the cells were immunoreactive for p75 neurotrophin receptor and Islet1, two early markers of motor neurons (73–75).

Rat liver Hsp90 was purified by a combination of ion exchange, gel filtration, and hydroxyapatite chromatography. Fractions were collected and analyzed with SDS/PAGE. The protein was more than 90% pure in all the preparations. Detailed methods are provided in SI Materials and Methods.

The human Hsp90β sequence was optimized for bacterial expression (GenScript USA) and cloned in a pBad/Myc-Hisx6 vector (Invitrogen). The replacement of two, four, and five tyrosine residues by phenylalanine (Y33F+Y56F, 4xPhe(Y56), and 5xPhe-Hsp90, respectively) was performed on the optimized wild-type human Hsp90β sequence (GenScript). The proteins were expressed after induction of the bacterial culture with 0.2% l-arabinose for 2 h at 37 °C and 220 rpm. Recombinant wild-type and mutant Hsp90β were purified from the bacterial culture using the Ni-NTA purification system (Invitrogen) according to the manufacturer’s instructions.

The incorporation of a nitrotyrosine residue in a selected position of recombinant human Hsp90β was performed as previously described (32) but with a newly selected orthogonal nitrotyrosine-bearing suppressor tRNA and engineered tRNA synthase pair. Briefly the codon codifying for each of the five tyrosine residues in the human Hsp90β sequence optimized for bacterial expression was replaced by an Amber stop codon (TAG), which was recognized by the orthogonal nitrotyrosine-bearing suppressor tRNA and tRNA synthase. The modified proteins were expressed and purified as described above for recombinant human Hsp90β.

For the intracellular delivery of Hsp90, Hsp70, actin, and tubulin, 200 µL of a mixture containing 10 μg of the protein in 100 µL PBS and 4 μL of the lipophilic Chariot permeation agent in 100 µL of H₂O was incubated for 30 min at room temperature, according to the manufacturer’s instructions (Active Motif). A pellet corresponding to 1 × 10⁶ PC12 cells was resuspended gently in the protein/Chariot mixture, and 400 µL of RPMI 1640 was added. After 1 h incubation at 37 °C in 5% CO₂/air, 1.5 mL of RPMI medium supplemented with 10% horse serum, 5% FetalClone II, and antibiotics was added, and the cells were incubated in the same conditions for an additional hour. After incubation, 4 × 10⁴ cells per 35-mm dish were plated in RPMI medium supplemented with 10% horse serum, 5% FetalClone II, and antibiotics to reach a final volume of 2 mL Viability was determined using fluorescein diacetate/propidium iodide after 24 h in culture (Molecular Probes) (71, 76, 77).

For the intracellular delivery of Hsp90, Hsp70, actin, and tubulin, 100 µL of a mixture containing 1 μg of the protein in 50 µL PBS and 1 μL of the lipophilic permeation agent Chariot solution in 50 µL of H₂O was incubated for 30 min at room temperature, as described above. About 1,500 motor neurons were plated in 100 µL of Neurobasal medium (Invitrogen/Gibco) per well in four-well plates precoated with polyornithine/lamin. The protein/Chariot mixture was added to each well, incubated for 1 h at 37 °C in a humidified CO₂ incubator, and then supplements plus trophic factors (BDNF, GDNF, and CT-1) in 800 µL of the complete neurobasal medium were added to reach a final volume of 1 mL The cultures were incubated for an additional 24 h; then motor neuron survival was determined by counting all neurons with neurites longer than three soma diameters as previously described (2) or by high-throughput image capture and analysis in 96-well plates using a Runner microscope (Trophos) as reported previously (4). For high-throughput image capture, motor neurons plated in black 96-well plates coated with poly-d-ornithine and laminin (Greiner Bio-one) were treated with calcein-AM (Molecular Probes-Invitrogen) in L15 medium (Invitrogen) for 1 h at 37 °C. Image analysis was performed using the Tina software (Trophos) (4, 22). The survival of motor neurons cultured in the presence of trophic factors or trophic factors plus Chariot was considered 100% and was used to standardize the experiments.

The transduction of motor neurons with lentiviral particles expressing P2X7 shRNA, control shRNA, or coGFP (Santa Cruz Biotechnology) was performed at multiplicity of infection 20. The lentiviral particles were added to the culture medium 2 h after plating the motor neurons. The cells were incubated for additional 84 h before the intracellular delivery of the proteins. The efficiency of transduction was 90–95%.

The proteasome chymotrypsin-like activity was determined in PC12 cells using a cell-based luminescent assay (Promega Corp.) according to the manufacturer’s instructions.

RNA was extracted from 50,000 motor neurons using TRIzol (Invitrogen) and then was purified using PureLink RNA micro kit (Invitrogen) according to the manufacturer’s instructions. The cDNA was synthesized using the SuperScript III First Strand Synthesis Supermix kit (Invitrogen). The PCRs for P2X7 and β-actin were performed using Platinum PCR SuperMix High Fidelity (Invitrogen) and the following primer sequences: P2X7, sense (5′-AAGGGAAAGAAGCCCCACGG-3′) and antisense (5′-CCGCTTTTCCATGCCATTTT-3′); β-actin, sense (5′-CCCTAAGGCCAACCGTGAA-3′) and antisense (5′-GCCATCTCTTGCTCGAAGTC-3′).

Western blots were performed as previously described (5, 6). PC12 cells incubated with peroxynitrite were harvested and rinsed and process for Western blotting. All Western blots were visualized and the bands quantified using the Odyssey System (Li-Cor Biosciences).

Dot-blot quantification of nitrated Hsp90 was performed by harvesting total protein from 30,000 motor neurons or from 40,000 PC12 cells plated in a 35-mm dish for 24 h. Proteins were harvested, and samples containing equal amounts of protein were blotted on a PVDF membrane by gravity flow using a Bio-Blot Microfiltration apparatus (Bio-Rad) as previously described (78). Slot blots of Hsp90 peptides synthesized to contain nitrotyrosine at selected positions were performed by blotting the indicated amounts of peptide on a PVDF membrane as described for dot blot. Membranes then were processed as described for Western blotting.

The intrinsic ATPase activity of Hsp90 was determined as previously described for the optimized Hsp90 ATPase assay without modifications (26). Geldanamycin (5 µM), a specific inhibitor of Hsp90 ATPase activity, was used to verify ATP hydrolysis by the chaperone. The citrate synthase aggregation assay was performed as previously described (27).

Spinal cord sections from ALS patients and the G93A animal model of ALS were collected as free-floating sections and processed for immunohistochemistry as previously described (79). The spinal cord contusion injury in rats was performed as previously described (80). Mice and rats were housed and used following the guidelines of the Weill Cornell Medical College and Rutgers, The State University of New Jersey, respectively. Human samples were obtained and used according with the guidlines of the Weill Cornell Medical College. Following fixation, spinal cord tissue sections were paraffin embedded and processed for immunofluorescence. The sections were incubated with the antibodies anti-nitrotyrosine (1:500), anti–microtubule-associated protein 2 (anti-MAP2) (1:500, Chemicon), and anti-nitrated Hsp90 (1:100). Fluorescence images were acquired using a Zeiss Axiovert 200 epifluorescence microscope. To study the mobilization of FasL to the plasma membrane, motor neurons were culture for 16 h in the presence or absence of 5 μM BAPTA-AM after intracellular delivery of nitrated Hsp90. Cultures were fixed with paraformaldehyde plus glutaraldehyde (Sigma) on ice for 20 min before the staining was performed, as described previously (2). The cells then were incubated with anti FasL (1:250, Santa Cruz Biotechnology).

For statistical analyses, ANOVA was used with all groups compared by Bonferroni test. All tests were performed using the program Prism (GraphPad Software Inc.).