Targeted redox inhibition of protein phosphatase 1 by Nox4 regulates eIF2α‐mediated stress signaling

Abstract Phosphorylation of translation initiation factor 2α (eIF2α) attenuates global protein synthesis but enhances translation of activating transcription factor 4 (ATF4) and is a crucial evolutionarily conserved adaptive pathway during cellular stresses. The serine–threonine protein phosphatase 1 (PP1) deactivates this pathway whereas prolonging eIF2α phosphorylation enhances cell survival. Here, we show that the reactive oxygen species‐generating NADPH oxidase‐4 (Nox4) is induced downstream of ATF4, binds to a PP1‐targeting subunit GADD34 at the endoplasmic reticulum, and inhibits PP1 activity to increase eIF2α phosphorylation and ATF4 levels. Other PP1 targets distant from the endoplasmic reticulum are unaffected, indicating a spatially confined inhibition of the phosphatase. PP1 inhibition involves metal center oxidation rather than the thiol oxidation that underlies redox inhibition of protein tyrosine phosphatases. We show that this Nox4‐regulated pathway robustly enhances cell survival and has a physiologic role in heart ischemia–reperfusion and acute kidney injury. This work uncovers a novel redox signaling pathway, involving Nox4–GADD34 interaction and a targeted oxidative inactivation of the PP1 metal center, that sustains eIF2α phosphorylation to protect tissues under stress.


MATERIALS AND METHODS
Unless otherwise indicated, all chemicals were purchased from Sigma Aldrich or Calbiochem and were of analytical or higher purity grade.
Cells. Primary cultures of neonatal rat cardiomyocytes were prepared using standard methods (Zhang et al, 2010). Rat H9c2 cardiomyoblasts, HEK293 and U2OS cells were from ATCC. MEFs were prepared from 13.5 day-old embryos of Nox4 -/and littermate WT mice, and immortalized with SV40 large T antigen.
The dichroic mirror in the spinning-disk unit was a Di01-T405/488/568/647 from Semrock Inc. A Sutter instruments filter wheel with Chroma emission filters was used.
Typically, a Z-stack of 11 steps over 3.0 micron was acquired and the maximum projection image was used for display and comparison of expression levels. The same acquisition and image contrast settings were used for control and treated cells.
3D Structural Illumination microscopy (SIM) was performed on a NIKON SIM system equipped with a 100 x 1.49 NA PlanApo oil immersion objective, an Andor EMCCD camera, and 488 and 567 nm diode lasers. Structured illumination image stacks were acquired with a z-distance of 100 nm and with 15 raw images per plane, 5 phases, 3 angles. The structured illumination raw data were computationally reconstructed using the SIM-module in the NIS Elements software (NIKON). Images displayed are reconstructions of one z plane.
Real time-RT PCR. Total RNA was prepared using an RNAase kit (Qiagen), and mRNA expression levels were quantified using specific primers as listed in Table S2.
Quantitative real time PCR was performed with SybrGreen on an Eppendorf PCR thermal cycler. Unless specified, β-actin was used for normalization. The relative fold change was calculated based on the Ct method.
Chromatin immunoprecipitation assay (ChIP). H9c2 cells were cultured with or without tunicamycin in serum-free media for 4 hours. Cells were treated with 1% formaldehyde for 10 min followed by glycine to terminate the cross-linking reaction.
Approximately 7.5 x10 6 cells were lysed with ChIP lysis buffer according to the manufacturer's instructions (SimpleChIP Enzymatic Chromatin IP kitmagnetic beads, Cell Signaling). The chromatin was fragmented by incubation with 2000 gel units micrococcal nuclease per 7.5 x10 6 cells for 20 min at 37°C. The nuclear membrane was broken by 15 sec of gentle sonication and 5 µg of chromatin in 500 µl CHIP buffer was incubated overnight at 4°C with 5 µg normal rabbit IgG (#2729, Cell Signaling) or 5 µg anti-ATF4 Ab (#11815, Cell Signaling). ChIP grade Protein G magnetic beads (#9006) were used to precipitate attached chromatin. Eluted samples were subjected to reverse cross-linking by incubating with 250 µg/ml proteinase K at 65°C for 2 h. DNA was isolated and concentrated using DNA purification spin columns and eluted in 50 µl of elution buffer. The immunoprecipitated DNA was analyzed by semi-quantitative PCR using primers to detect the putative ATF4-binding sites in the Nox4 promoter. ATF4 primers were: forward TGGTCCTGACTTTTCCATCAG, reverse TGGATGTTCGAGAAATTGACTG. PCR was performed under standard conditions for 40 cycles with an annealing temperature of 55°C.
Preparation of membrane fractions. Cells grown in 100-mm dishes were washed with cold PBS and homogenized in lysis buffer (50 mM Tris, pH 7.4, containing 0.1 mM EDTA, 0.1 mM EGTA, protease inhibitor cocktail [Sigma,#P8340], and 2 µg/ml Mg132) by sonication (10 s of 3 cycles at 8 W). After centrifugation at 18,000 g for 15 min to separate mitochondria and nuclei, the supernatant was further centrifuged at 100,000 g for 1 h. The resulting supernatant formed the cytosolic fraction. The pellet containing the membrane-enriched fraction was washed two times with the same lysis buffer to remove any remaining cytosol contamination. Membrane fractions were used to assay Nox activity, PP1 activity or for immunoblotting. The membrane fraction was enriched in the ER marker calnexin while the cytosolic fraction was enriched in GAPDH (Fig. S4A).
The samples were centrifuged at 35,000 g (4 o C, 18 h). Fractions 1-16 (F1-F16) were collected from the base of the column. Each fraction was split into two 200 µl aliquots, one for immunoblotting and the other for immunoprecipitation experiments.
As a control for density gradient separation, a mix of proteins (Gel filtration molecular weight markers, Sigma-Aldrich) was added to the top of the sucrose gradient in a separate tube and centrifuged. The fractions obtained were submitted to SDS-PAGE and proteins were stained with Coomassie Blue.
Immunoprecipitation. Cells grown in 6 well plates were scraped and transferred into tubes, then centrifuged at 2,800 g at 4 o C for 5 min. The cell pellet was resuspended in 200 µl lysis buffer. Samples were briefly sonicated (one 10 s cycle, 8 W). Protein concentration was normalized to 1 μg/μl and immunoprecipitation (IP) was performed using 500 μg of homogenate protein. Protein A/G Sepharose beads (Santa Cruz Technology) were pre-cleared with nonspecific IgG, and samples were then precipitated overnight at 4 o C with specific antibody. Samples with non-specific antibody were used as negative control. The next day, immunoprecipitates were washed 7 times with buffer and resuspended in sample buffer for immunoblotting.
Samples were heated at 95 °C for 5 min. After cooling, reducing agent was added, samples were run on SDS-polyacrylamide gels, and then blotted onto nitrocellulose.

Measurement of ROS.
Nox activity (NADPH-stimulated ROS generation) was measured in membrane fractions preapared as described above, using HPLC-based Miklós Geiszt (Department of Physiology, Semmelweis University, Budapest, Hungary). The respective C199S mutant probes for HyPer-ER and HyPerRed, which are ROS-insensitive, were used as negative controls to exclude changes in pH. Cells were co-transfected with HyPer-ER and HyPerRed and kept in phenol red-free medium supplemented with 2 mM glutamine and antibiotics for 48 hours before treatment with tunicamycin (2 µg/ml for 4 hours) or control vehicle. Imaging was performed at 37° C / 5% CO 2 on an inverted Nikon Ti-E microscope equipped with a Yokogawa CSU-X1 spinning-disk confocal unit, an Andor Neo sCMOS camera and a Sutter filter wheel. A 60x Plan Apo VC NA 1.40 Nikon objective was used. HyPer-ER fluorescence emission was monitored at 525/50 nm following excitation at 405 nm and 488 nm, and the ratio of fluorescence intensity was quantified. HyPerRed fluorescence emission was monitored at 647/75 nm following excitation at 560/40 nm. Extracellular H 2 O 2 (200 nM) was added as a positive control and the HyPer-ER and HyPerRed signals acquired simultaneously. NIS Elements v.4.0 software (Nikon) was used for image analysis. Images were background-subtracted and thresholded.
Changes in HyPer-ER fluorescence ratio (R) or HyPerRed fluorescence intensity (F) between the indicated time-points or treatments were quantified. The resulting images were displayed in pseudocolor.

Recombinant PP1 expression and purification.
A pCW vector expressing the untagged  isoform of the catalytic subunit of human PP1 (Alessi et al, 1993) was obtained from the MRC Protein Phosphorylation Unit (Dundee, UK). Protein expression and purification was carried out essentially as described (Barford and Keller, 1994;Egloff et al, 1995). Transformed E.coli DH5 cells were grown in Luria-Bertani (LB) medium supplemented with 2 mM MnCl 2 and 100 μg.ml −1 ampicillin at 30 °C until OD600 reached approximately 0.25. Protein expression was induced with 0.5 mM IPTG. Cells were harvested by centrifugation at 5000 g for 15 min at 4 °C and resuspended in buffer A (50 mM imidazole, 0.5 mM EDTA, 0.5 mM EGTA, 100 mM NaCl, 10% glycerol, 2 mM -mercaptoethanol, 2 mM MnCl 2 , pH 7.5) supplemented with Complete EDTA-free protease inhibitor cocktail (Roche), lysozyme (0.01 mg/ml) and DNAse (0.05 mg/ml). Cell lysis was accomplished by sonication or using a cell disruptor (Constant Systems Ltd). Insoluble material was sedimented by centrifugation at 19500 g for 1 h at 4 °C and the supernatant filtered using 0.22 μm prior to loading on a 5 mL heparin column equilibrated with buffer A.
PP1 was eluted using a 100 ml gradient to 50% buffer A supplemented with 1M NaCl. Fractions were analysed on a 12% SDS-PAGE gel and those containing PP1 were pooled and diluted 10-fold with buffer C (50 mM imidazole, 0.5 mM EDTA, 0.5 mM EGTA, 10% glycerol, 5 mM -mercaptoethanol, 2 mM MnCl 2 , pH 7.2) for injection in a HiTrapQ HP (GE Healhcare) column. PP1 was eluted using a gradient to 40% buffer C supplemented with 1M NaCl. PP1 was further purified by sizeexclusion chromatography (SEC) using a Superdex 75 16/60 (GE Healthcare) column equilibrated with SEC buffer (50 mM imidazole, 0.5 mM EDTA, 0.5 mM EGTA, 300 mM NaCl, 10% glycerol, 5 mM -mercaptoethanol, 2 mM MnCl 2 , pH 7.5) for downstream applications. PP1 mutations (PP1 N124D and PP1 D64N) were introduced using the Q5  Site-Directed Mutagenesis Kit (New England Biolabs). All constructs were verified by sequencing. Expression and purification of PP1 variants were carried out as for wild-type PP1.  Table S3. For assessment of cellular phosphatase activity, cell membrane extracts (Hubbard et al, 1990) were incubated with phosphopeptide substrate (0.1 mM) in the presence or absence of okadaic acid (10 nM), which does not inhibit PP1 at this concentration (Ishihara et al, 1989), and then phosphatase activity was estimated as described above. For each sample, incubation without the phosphopeptide substrate was used as a blank. PP1 activity was taken as the okadaic acid-resistant fraction and was normalized by protein content. Calyculin A (60 nM) (which inhibits both PP1 and PP2a) (Ishihara et al, 1989) was used as a control to confirm total PP activity. In some experiments, ascorbate (0.5 mM) was added to cells for 30 min before cell lysis.
Electron paramagnetic resonance spectroscopy (EPR). EPR was used to measure ascorbyl radical generation and to assess the PP1 metal redox status. For ascorbyl detection, EPR spectra were recorded at room temperature in a Magnatech Miniscope MS2000 spectrometer. The instrument conditions were: microwave power 50 mW, modulation amplitude 1 Gauss (G), scan time 328 ms, with a gain of 9 × 10 2 .
All spectra were the accumulation of 4 scans and were recorded 5 min after addition of H 2 O 2 . EPR instrument conditions were calibrated with 4-hydroxy-2, 2, 6, 6tetramethyl-1-piperidinyloxy (Tempol). The reaction was carried out in 0.1 mM Tris buffer at pH 7.0 and 37 o C under the different conditions described in Figure legends, and was transferred to a 50 µl flat cell immediately after the addition of ascorbate.
The two line spectrum was consistent with an ascorbyl radical with a hyperfine splitting constant (a H = 1.8G) (Monteiro et al, 2007), as generated using the positive control ascorbate and H 2 O 2 .
EPR at low temperature is a method to detect chemical species with unpaired electrons and is used for studies of transitional metal ion complexes in proteins (Cammack and Cooper, 1993;Ubbink et al, 2002). We used a Bruker EMX 300 spectrometer with a 3mm cavity and a helium cooling system. Purified PP1 (5 mg/ml) was studied at baseline and after treatment with H 2 O 2 (1 mM) in TrisHCl buffer pH 7.2 at 37 o C. The reaction mixture was transferred to a flat cell and frozen in liquid nitrogen. Spectrometer conditions were: temperature, 4 K; microwave frequency, 9.66 GHz; modulation amplitude, 2 G at 100 kHz; microwave power, 20mW.
Cell viability. Cells were plated in 24-well plates at 70% confluence and Nox4 levels were manipulated as described in the Figure legends. Cells were then exposed to tunicamycin (2 µg/ml), guanabenz or clonidine (both dissolved in PBS), or salubrinal After 20 min equilibration, global ischemia was initiated for 25 min and the hearts were then reperfused for 100 min. Hearts were weighed, frozen and cut into 1 mm thick slices. Viable tissue was stained red with 1% 2,3,5-triphenyl-tetrazolium chloride (TTC) in phosphate buffer; sections were then immersed in formalin and scanned.
The infarcted area was calculated as a proportion of the total left ventricular area using Image J Software. For immunobloting studies, hearts were reperfused for 30 min following ischemia and snap frozen for subsequent analyses. Some animals were injected with guanabenz (1.8 mg/kg body weight) 24 h prior to heart perfusion.
The hearts of these animals were perfused with modified KH buffer containing 0.5 µM guanabenz.
To induce ER stress-related AKI, animals were treated with tunicamycin (3 mg/kg/day i.p. for two days) (Zinszner et al, 1998). Some animals were pre-treated with guanabenz (1.8 mg/kg ip). After sacrifice, serum was collected and the plasma urea concentration was measured using a commercial Kit (Bioassay Systems).
Kidneys were harvested for immunoblotting or were fixed and paraffin-embedded to assess apoptosis using TUNEL staining (Millipore S7110 Kit).
Statistics. Data are presented as mean±SEM. Comparisons among groups were undertaken by Student's t test or one-way ANOVA, as appropriate. Kaplan Meier analysis was used to compare survival. Statistical analyses were performed on GraphPad-Prism (GraphPad-Software, San Diego, Ca). P<0.05 was considered significant.

References
Alessi DR, Street AJ, Cohen P, Cohen PT (1993) Inhibitor-2 functions like a chaperone to fold three expressed isoforms of mammalian protein phosphatase-1 into a conformation with the specificity and regulatory properties of the native enzyme. Eur

D.
Nox activity increased after tunicamycin treatment but was substantially reduced by the knockdown of Nox4. Nox activity was measured in membrane fractions isolated after 4h of Tn treatment, using HPLC-based detection of the dihydroethidium (DHE) oxidation products, 2-hydroxyethidium (EOH) and ethidium (E). Inset shows Nox4 protein levels.

E.
Effects of shRNA-mediated knockdown of Nox4 in H9c2 cells on the tunicamycin (Tn, 2 µg/ml)-induced changes in nuclear levels of ATF4 (red bar graph) and ATF6 (black bar graph); mRNA levels of Xbp1-s (blue bar graph); and protein levels of Grp78 (green bar graph). Representative blots and gels for this experiment are shown in Fig 1C. All data are mean ± SEM of n=3 per group except panel D, which is n=4 per group. *, significant compared to baseline. #, significant comparing Ad.shNox4 versus corresponding control (Ad.Ctl). Values above bar graphs denote the level of significance.  antibody or with non-specific rabbit IgG, as indicated. Purified DNA was analyzed using primers specific for the rat Nox4 promoter comprising the putative ATF4 binding sites (see schematic).
All data are mean ± SEM of 3 experiments/group apart from ATF6 protein levels in panel A which were n=4/group. *, significant compared to baseline. #, significant comparing siNox4, Ad.Nox4 or siATF4 versus corresponding controls. Values above bar graphs denote the level of significance.    The maximum likelihood estimate for coordinates' uncertainty ( x ) derived from crystallographic refinement is 0.08 Å and 0.11 Å for the ascobate-treated (reduced) and H 2 O 2 -treated (oxidized) structures, respectively. The standard uncertainty on metal-ligand (M-L) distances is  d(M-L) = 2 (1/2)  x . As there are two PP1 molecules in the a.u., each distance is determined twice resulting in the standard estimate of the mean (s.e.m.) to be (s.e.m. d M-L ) = (2 (1/2)  x ) / 2 (1/2) =  x . Thus, s.e.m d (M-L)reduced = 0.08 Å and s.e.m d (M-L)oxidized = 0.11 Å.
To evaluate whether a correlation exists between the change in metal-ligand coordination distance defined as  = d (M-L)oxidized -d (M-L)reduced determined by theoretical and X-ray experimental methods we plotted  X-ray against  theory for each of the twelve (M-L) distances. A perfect correlation would result in a straight line of unitary slope. Errors for  X-ray are calculated as  X-ray = [( x(oxidised) ) 2 + ( x(reduced) ) 2 ] (1/2) . The plot is shown in Fig. S5f.
There is a good correlation between theory and experiment with 75% of the  values (black circles) lying on the diagonal within error whilst three values (red circles) can be considered outliers. The availability of X-ray data at higher resolution and the use of a more complex description of protein restraints in the theoretical calculations might improve the agreement even further. The correlation is statistically significant and quantified by a Spearman  coefficient of 0.706 for all twelve  values with a two-tailed p-value of 0.0124.
The plot shows that most points lie on the lower-left quadrant. This implies a contraction of the average (M-L) distance upon metal oxidation. We tested whether the contraction of the average (M-L) distance observed experimentally is statistically significant. A Wilcoxon matched-pairs signed rank test gives a one-tailed p-value of 0.023 indicating statistical significance (p<0.05).