Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) Protein-Protein Interaction Inhibitor Reveals a Non-catalytic Role for GAPDH Oligomerization in Cell Death*

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), an important glycolytic enzyme, has a non-catalytic (thus a non-canonical) role in inducing mitochondrial elimination under oxidative stress. We recently demonstrated that phosphorylation of GAPDH by δ protein kinase C (δPKC) inhibits this GAPDH-dependent mitochondrial elimination. δPKC phosphorylation of GAPDH correlates with increased cell injury following oxidative stress, suggesting that inhibiting GAPDH phosphorylation should decrease cell injury. Using rational design, we identified pseudo-GAPDH (ψGAPDH) peptide, an inhibitor of δPKC-mediated GAPDH phosphorylation that does not inhibit the phosphorylation of other δPKC substrates. Unexpectedly, ψGAPDH decreased mitochondrial elimination and increased cardiac damage in an animal model of heart attack. Either treatment with ψGAPDH or direct phosphorylation of GAPDH by δPKC decreased GAPDH tetramerization, which corresponded to reduced GAPDH glycolytic activity in vitro and ex vivo. Taken together, our study identified the potential mechanism by which oxidative stress inhibits the protective GAPDH-mediated elimination of damaged mitochondria. Our study also identified a pharmacological tool, ψGAPDH peptide, with interesting properties. ψGAPDH peptide is an inhibitor of the interaction between δPKC and GAPDH and of the resulting phosphorylation of GAPDH by δPKC. ψGAPDH peptide is also an inhibitor of GAPDH oligomerization and thus an inhibitor of GAPDH glycolytic activity. Finally, we found that ψGAPDH peptide is an inhibitor of the elimination of damaged mitochondria. We discuss how this unique property of increasing cell damage following oxidative stress suggests a potential use for ψGAPDH peptide-based therapy.

is also involved in mitochondrial elimination by mitophagy (1). GAPDH binds to damaged mitochondria, and its phosphorylation by ␦ protein kinase C (␦PKC) under oxidative stress inhibits the removal of the damaged mitochondria through mitophagy (1). We set out to determine whether ␦PKC-mediated GAPDH phosphorylation alone is sufficient to mediate cell death following oxidative stress, using Langendorff, an ex vivo model of heart attack.
We first designed a novel inhibitor that selectively inhibits GAPDH phosphorylation without affecting the phosphorylation of other ␦PKC substrates. To this end, we relied on the rationale that, similar to many other protein kinases, ␦PKC contains a highly conserved catalytic domain and a regulatory domain that keeps the enzyme in an inactive conformation. This inactive conformation is stabilized by several intramolecular auto-inhibitory interactions (2). The first intramolecular interaction identified was the one mediated by the pseudosubstrate site located at the N terminus of the kinase, which mimics the phospho-acceptor sequence in PKC substrates; the pseudosubstrate site binds to the substrate-binding cavity in the catalytic domain of the kinase, thus keeping the enzyme inactive (3,4). In addition to the pseudosubstrate auto-inhibitory site, several protein kinases have a second regulatory site outside of the catalytic site, termed the substrate-docking site (5). The substrate-docking site is a short motif that mediates selective docking of a substrate to its kinase (5,6). We reasoned that if there is a unique GAPDH-docking site on ␦PKC, an inhibitor of this docking site will serve as a specific inhibitor of GAPDH phosphorylation by ␦PKC, without affecting the phosphorylation of other ␦PKC substrates.
The following rationale was used to identify this inhibitor. It is likely that the GAPDH-specific docking site on ␦PKC is masked when ␦PKC is inactive, and a conformational change in ␦PKC to the active state exposes this GAPDH-docking site. Analogous to the pseudosubstrate concept (3), we hypothesized that there is a pseudo-GAPDH-docking site (GAPDH) on ␦PKC that masks the GAPDH-docking site. The GAPDH site on ␦PKC mimics the binding site for ␦PKC on GAPDH, and therefore, it may have a short sequence that is homologous to GAPDH.
Here, we describe how we used this rationale to develop a peptide corresponding to GAPDH, which we termed GAPDH peptide. Our work with GAPDH peptide identified phosphorylation as a novel switch controlling the oligomeric state of GAPDH, modifying its function as a metabolic enzyme and a mediator of mitochondrial elimination.

Animal Care
Animal care and husbandry procedures were in accordance with the National Institutes of Health guidelines. The animal protocols were approved by the Stanford University Institutional Animal Care and Use Committee.

Peptide Synthesis
In Brief-Peptides were synthesized on solid support using a fully automated microwave peptide synthesizer (Liberty, CEM Corp.) (8) using the Fmoc 3 /tert-butyl protocol. The final cleavage and side chain deprotection were done manually. Peptides were analyzed by analytical reverse-phase high pressure liquid chromatography (RP-HPLC) (Shimadzu) and matrix-assisted laser desorption/ionization (MALDI) mass spectrometry (MS) and purified by preparative RP-HPLC (Shimadzu).
Peptides were chemically synthesized using Liberty Microwave Peptide Synthesizer (CEM Corp.) on solid support with an additional module of Discover (CEM Corp.) equipped with fiber-optic temperature probe for controlling the microwave power delivery following the Fmoc/tert-butyl method in a 30-ml Teflon reaction vessel. Each deprotection and coupling reaction was performed with microwave energy and nitrogen bubbling.
The synthesis products were purified by preparative RP-HPLC (Shimadzu LC-20 equipped with CBM-20A system controller, SPD-20A detector, CTO-20A column oven, 2ϫ LC-6AD solvent delivery unit, and FRC-10A fraction collector from Shimadzu), using an XBridge Prep OBD C18 5 m (19 mm ϫ 150 mm) column (Waters) at 10 ml/min. The solvent systems used were A (H 2 O with 0.1% TFA) and B (CH 3 CN with 0.1% TFA). For separation, a linear gradient of 5-50% B in 45 min was applied, and the detection was at 215 nm.
Peptides were conjugated to TAT(47-57) (YGRKKR RQRRR) carrier peptide through an amide bond to the N terminus of the cargo peptide, as part of the solid-phase peptide synthesis and included a GG spacer between TAT and the cargo. ␦V1-1 was synthesized and conjugated to TAT as previously described (10). Note that TAT(47-57)-based delivery of a variety of peptide cargoes into cells has been used now for over 25 years and delivery of the cargo across biological membranes and into subcellular organelles has been confirmed (11)(12)(13). The peptides were dissolved in water or saline.

Interaction of GAPDH/␦PKC, in Vitro
To determine the level of GAPDH binding to ␦PKC in the presence of GAPDH peptide, 200 ng of recombinant ␦PKC (Invitrogen) were incubated with or without the indicated peptides (1 M) for 10 min at 4°C, prior to adding 200 ng of recombinant GST-GAPDH (Abnova, Taiwan) for 20 min at 37°C. GAPDH was immunoprecipitated using GAPDH antibody (mAb 6C5CA, Advanced ImmunoChemical Inc., CA) and ␦PKC binding to GAPDH was determined with ␦PKC antibody (C-17, Santa Cruz Biotechnology) and then with HRP-conjugated secondary antibody using Western blot analysis.

GAPDH Phosphorylation by ␦PKC and GAPDH Activity, in Vitro
To determine the level of GAPDH phosphorylation in the presence of GAPDH peptide, 200 ng of recombinant ␦PKC protein (Invitrogen) was incubated with or without the peptides (1 M) for 10 min at 4°C and then 200 ng of recombinant GST-GAPDH (Abnova, Taiwan) was added for 20 min at 37°C in kinase buffer (40 l) (20 mM Tris-HCl, 20 mM MgCl 2 , 1 M DTT, 25 M ATP, and 1 mM CaCl 2 ) in the presence of the PKC activators, phosphatidylserine (1.25 g) and 1,2-dioleoyl snglycerol (0.04 g). The kinase assay was terminated by adding Laemmli buffer containing 5% SDS, and the samples were loaded on a SDS-10% polyacrylamide gel, and the levels of phosphorylated GAPDH protein were determined using antiphosphothreonine (9381S and 2351S, Cell Signaling) and anti-phosphoserine PKC substrate (2261L, Cell Signaling) antibodies and then with HRP-conjugated secondary antibody using Western blot analysis. The nitrocellulose was also reprobed using GAPDH antibody (mAb 6C5CA, Advanced ImmunoChemical Inc.) to confirm that equal amounts of GAPDH were used.
GAPDH activity was determined using a KDalert GAPDH assay kit as instructed by the manufacturer (Applied Biosystems) using the same reagent as above. Briefly, ␦PKC protein (200 ng, Invitrogen) was incubated without or with GAPDH peptide (10 M) for 10 min at 4°C, and then GST-GAPDH (200 ng, Abnova, Taiwan) was added for 20 min at 37°C in kinase buffer (40 l) in the presence of the PKC activators. Samples were combined with the KDalert master mixture in 96-well plates and read at 690 nm using an UV-visible plate reader. GAPDH activity was calculated and expressed as units of GAPDH/mg of total protein.

Cell Culture, Viability Assay, Measurements of Mitochondrial Content, and siRNA Treatment
Cardiac H9C2 cells were maintained in modified Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal bovine serum (FBS) and 1% (v/v) penicillin/streptomycin. Cultured H9C2 cells in black 96-well plates were treated without or with GAPDH peptide (1 M), 15 min prior to treatment with H 2 O 2 (1 mM). After 1 h of treatment, cell death was quantified by measurement of the activity of lactate dehydrogenase (LDH) released from damaged cells into the supernatant using a cytotoxicity detection kit (Roche Diagnostics).

Animal Studies
All treatments were performed between 9:00 a.m. and 4:00 p.m. by observers blinded to the treatment groups. Male rats were housed in a temperature-and light-controlled room for at least 3 days before use. All animals were randomized and assigned to testing groups to generate biological replicates for each group.

Ex Vivo Rat Heart Model of Myocardial Infarction-induced Ischemia and Reperfusion Injury
Wistar male rats, 4 -6 weeks old, were treated with different peptides or control, and the levels of phosphorylation, infarct size, and CK release were measured, as described previously (14,15). We excluded hearts if they met one of the following criteria: 1) time to perfusion prior to ischemia and reperfusion protocol exceeded 3 min; 2) coronary flow was outside the range of 9 -15 ml/min; or 3) heart rate was below 240 beats/ min, or severe arrhythmia was observed.
An ex vivo model of acute ischemic heart injury was carried out as described previously (10). Briefly, Wistar rats (250 -275 g) purchased from Charles River were heparinized (1000 units/ kg; intraperitoneal injection) and then anesthetized with Beuthanasia-D (100 mg/kg intraperitoneal injection). Hearts were rapidly excised and then perfused with an oxygenated Krebs-Henseleit buffer (120 mM NaCl, 5.8 mM KCl, 25 mM NaHCO 3 , 1.2 mM NaH 2 PO 4 , 1.2 mM MgSO 4 , 1 mM CaCl 2 , and 10 mM dextrose, pH 7.4) at 37°C in a Langendorff coronary perfusion system. A constant coronary flow rate of 10 ml/min was used. Hearts were submerged into a heat-jacketed organ bath at 37°C. After 10 min of equilibration, the hearts were subjected to 30 min of global ischemia and 60 min of reperfusion. The hearts were perfused with 1 M peptides as indicated for 20 min after the ischemia. Normoxic control hearts were subjected to 90 min of perfusion in the absence of ischemia. Coronary effluent was collected to determine CK release during the first 30 min of the reperfusion period. At the end of the reperfusion period, hearts were sliced into 1-mm-thick transverse sections and incubated in triphenyltetrazolium chloride solution (1% in phosphate buffer, pH 7.4) at 37°C for 15 min. Infarct size was expressed as a percentage of the risk zone (equivalent to total muscle mass). Infarct size and CK release were used to assess cardiac damage, as described previously (10) . Docking of these substrates to the kinase increases the access of the catalytic site to the substrates, leading to their phosphorylation. In inactive ␦PKC (left), the GAPDH-docking site on ␦PKC interacts with another ␦PKC sequence, GIYQGF, which mimics the kinase-binding site on GAPDH (blue, GIVEGL). B, peptide corresponding to this GAPDH-like sequence on ␦PKC, GAPDH sequence, is a competitive inhibitor for docking to and phosphorylation of GAPDH by ␦PKC, without affecting docking and phosphorylation of other PKC substrates (e.g. substrate XX). C, sequence alignment of human ␦PKC and GAPDH identified a short sequence of homology, GIYQGF/GIVEGL. D, GIVEGL (light blue) in GAPDH (Protein Data Bank code 1U8F) is exposed and is located right near the ␦PKC phosphorylation site (Thr-246, yellow). E, conservation of GIYQGF sequence in ␦PKC, and F, conservation of GIVEGL in GAPDH in a variety of species. Symbols below the letters are * for identity and ∧ for homology. G, GIYQGF sequence is found in the V3 domain of ␦PKC. GIYQGF sequence is not present in other PKC isozymes, including PKC, the most homologous PKC isozyme to ␦PKC, and ⑀PKC, which is also a member of the novel PKC isozymes, which include ␦PKC (25). H, GIYQGF sequence is found in 15 human proteins (GAPDH, ␦PKC, and 13 other proteins, including four from the ADT family). A heat map of the GIYQGF conservation in orthologs of these proteins shows GIYQGF conservation only in ␦PKC and GAPDH.

Western Blot Analysis and Two-dimensional Gel Analysis
Rat hearts following the ischemia and reperfusion protocol were homogenized in buffer A (210 mM mannitol, 70 mM sucrose, 5 mM MOPS, and 1 mM EDTA) in the presence of protease inhibitor and phosphatase inhibitor mixtures (Sigma) followed by isolation of the mitochondrial fraction.
Heart mitochondria were isolated as described previously (16). Briefly, cardiac samples from a remote area were minced and homogenized in isolation buffer (300 mM sucrose, 10 mM Hepes, 2 mM EGTA, pH 7.2, 4°C) containing 0.1 mg/ml type I protease (bovine pancreas) to release mitochondria from within the muscle fibers and later washed in the same buffer in the presence of 1 mg/ml bovine serum albumin. The suspension was homogenized in a 40-ml tissue grinder and centrifuged at 800 ϫ g for 5 min. The resulting supernatant was centrifuged at 9,500 ϫ g for 10 min. The mitochondrial pellet was washed, resuspended in isolation buffer, and submitted to a new centrifugation (10,000 ϫ g for 10 min). The mitochondrial pellet was washed, and the final pellet was resuspended in a minimal volume of isolation buffer.
For protein phosphorylation, cardiac fractions were resuspended in buffer A (210 mM mannitol, 70 mM sucrose, 5 mM MOPS, and 1 mM EDTA). Pyruvate dehydrogenase kinase (PDK) phosphorylation was determined using two-dimensional isoelectric focusing/SDS-PAGE; the samples were homogenized in buffer consisting of 7 M urea, 2 M thiourea, and 4% CHAPS in the presence of protease inhibitor and phosphatase inhibitor mixtures (Sigma). Supernatants were subjected to a first dimensional separation by an IPGphor isoelectric focus power supply using pre-cast Immobiline DryStrip pI 3-10 strips according to the manufacturer's instruction manual (Amersham Biosciences). 10% SDS-gel electrophoresis and Western blotting were carried out using anti-PDK2 (Abgent), anti-phosphothreonine (9381S and 2351S Cell Signaling), and anti-phosphoserine PKC substrate (2261L Cell Signaling) antibodies and then with HRP-conjugated secondary antibody. Phosphatase treatment confirmed that the leftward shift in pyruvate dehydrogenase kinase mobility is due to phosphorylation (16).
Phosphorylation of HSP27, MARCKS, STAT, GAPDH, and troponin I in the ex vivo model were determined on onedimensional SDS-PAGE using anti-phospho-HSP27 (04-447, EMD Millipore) and anti-HSP27 (ADI-SPA-800-F, Enzo Life Sciences), anti-phospho-MARCKS (2741, Cell Signaling) and anti-MARCKS (6455, Santa Cruz Biotechnology), anti-phospho-STAT (8826S, Cell Signaling) and anti-STAT (346, Santa Cruz Biotechnology), anti-phosphotroponin I (4004, Cell Signaling) and anti-troponin I (31655, Santa Cruz Biotechnology) antibodies. The levels of phosphorylated substrates were normalized to total substrate and presented as a ratio to the phosphorylation in hearts subjected to ischemia and reperfusion in the presence of control treatment. Data are representative of four independent experiments.
Phosphorylation of GAPDH in the ex vivo heart attack model was determined after immunoprecipitation. Samples were incubated with GAPDH antibody (Mab6C5, Advanced Immunochemical) in buffer containing 10 mM Tris-base, pH 7.4, 150 mM NaCl, 0.1% Triton X-100, 5 mM EDTA, pH 8, and protease inhibitor overnight at 4°C with gentle agitation. Protein A/G beads were then added, and the mixture was incubated for 2 h at 4°C. The mixture was centrifuged for 1 min at 800 ϫ g, and the immunoprecipitates were washed three times with buffer, analyzed by 10% SDS-PAGE, followed by Western blot using anti-phosphothreonine (9381S and 2351S, Cell Signaling), anti-phosphoserine PKC substrate (2261L, Cell Signaling), and anti-GAPDH (Mab6C5, Advanced Immunochemical) antibodies and then with HRP-conjugated secondary antibody. The levels of phosphorylated substrate were normalized to total substrate and presented as a ratio to the phosphorylation in hearts subjected to ischemia and reperfusion in the presence of control peptide. Data are representative of four independent experiments.

ATP Level Determination
Following the ex vivo ischemia and reperfusion experiment, 100 mg of heart tissue was homogenized in 1% TCA (500 l). The lysate was spun to remove debris, and the pH was adjusted to pH 7.4. 10 l of the lysate were used in a 200-l assay for quantitative determination of ATP levels with recombinant firefly luciferase and its substrate, D-luciferin, according to the manufacturer's protocol (Invitrogen). Data are representative of six independent experiments.

Determination of GAPDH Activity in Samples from Whole Hearts Subjected to Ischemia and Reperfusion ex Vivo
GAPDH activity was determined using a KDalert GAPDH assay kit as instructed by the manufacturer (Applied Biosystems). Briefly, samples were combined with the KDalert master mixture in 96-well plates and read at 690 nm using a UV-visible plate reader. GAPDH activity was calculated and expressed as units of GAPDH/mg of total protein. Data are representative of at least four independent experiments.

Determination of GAPDH Oligomerization
GAPDH oligomerization was analyzed by native gel electrophoresis. The samples were loaded on a 6 or 10% native gel, and the levels of tetramer, trimer, dimer, and monomer GAPDH protein were determined using GAPDH antibody (mAb 6C5CA, Advanced ImmunoChemical Inc.) and then with HRPconjugated secondary antibody.

Statistical Analysis
Data are provided as individual points and means Ϯ S.E.; the number of independent experiments performed is provided in each data set. Data were tested for significance by using the two-tailed unpaired Student's t test. Differences were considered statistically significant when p values were Ͻ0.05. Sample sizes were estimated based on previous experience with similar assays and the effect of size observed in preliminary experiments. All samples were identical prior to allocation of treatments, and analysis was carried out by an observer blinded to the experimental conditions. The intensity of the spots on the gels were measured using ImageJ from National Institutes of Health (17).

Rational Design of a Peptide Based on Homology between
␦PKC and GAPDH-We hypothesized that when ␦PKC is inactive, its GAPDH-docking site may be occupied by a pseudo-GAPDH (GAPDH) site, a GAPDH-like sequence that mimics the ␦PKC-binding site on GAPDH (Fig. 1A, left panel)   rect, upon ␦PKC activation, a conformational change in ␦PKC dissociates the GAPDH site from the GAPDH-docking site to expose the GAPDH-docking site for interaction with GAPDH (Fig. 1A, right panel). A peptide corresponding to the sequence of GAPDH is expected to selectively inhibit GAPDH docking to ␦PKC, and therefore it will diminish ␦PKC/GAPDH binding and the phosphorylation of GAPDH by ␦PKC (Fig. 1B). We also expect that GAPDH will not affect the binding and phosphorylation of other ␦PKC substrates (e.g. substrate xx, Fig. 1B). Similar to the pseudosubstrate site on ␦PKC that mimics and has a similar sequence to the phospho-acceptor site on the substrate, the GAPDH-docking site on ␦PKC may have a similar sequence to GAPDH. Using a sequence homology search, Lalign (18), we searched for such a site and identified a six-amino acid stretch in GAPDH (GIVEGL; amino acids 169 -174) that is 83% homologous to a sequence in ␦PKC (GIYQGF, amino acids 311-316; Fig. 1C). The GIVEGL sequence in the GAPDH structure is located in an exposed region in the monomeric enzyme, potentially available for protein-protein interaction (Fig. 1D). Interestingly, the GIVEGL sequence in GAPDH is adjacent to the main phosphorylation site by ␦PKC, Thr-246 (1) (Fig. 1D).
Sequences that are critical for protein-protein interactions are expected to be conserved in evolution (19 -22). We therefore determined the evolutionary conservation of these short homologous sequences in GAPDH and ␦PKC. The GIYQGF sequence in ␦PKC and the GIVEGL sequence in GAPDH are 83% conserved in mammalian, avian, fish, and lizards (Fig. 1, E  and F). We reasoned that if these sequences are specific for only ␦PKC-GAPDH interaction, then they are unlikely to be conserved in other proteins. Indeed, the GIYQGF sequence in ␦PKC is not present in other PKC isozymes, including PKC, the most similar PKC isozyme to ␦PKC (Fig. 1G) (23). Next, we searched for GIYQGF sequence in other proteins in the human genome. Thirteen other proteins were identified with 83% homology to GIYQGF. However, the GIYQGF sequence was not conserved in evolution in any of these proteins (Fig. 1H). Altogether, these results suggest a functionally important and selective role of GIYQGF/GIVEGL for ␦PKC and GAPDH interaction.
GAPDH Peptide Specifically Inhibits GAPDH Phosphorylation by ␦PKC, in Vitro and Ex Vivo-We synthesized the peptide corresponding to the GAPDH site on ␦PKC, GIYQGF, and we tested its specificity as an inhibitor of the protein-protein interaction between ␦PKC and GAPDH. As a control, we synthesized a peptide with one amino acid substitution (Gln substitute to Ala, GIYAGF), and referred to it as control (Ctrl peptide). As expected, GAPDH peptide (1 M) blocked ␦PKC binding to GAPDH, as determined by inhibition of co-immunoprecipitation ( Fig. 2A). GAPDH peptide also inhibited ␦PKC-mediated GAPDH phosphorylation by over 70%, in vitro (Fig. 2, B and C), although the control peptide did not inhibit ␦PKC-GAPDH interaction or the phosphorylation of GAPDH by ␦PKC. Thus, the GAPDH peptide inhibited the interaction between GAPDH and ␦PKC and therefore inhibited GAPDH phosphorylation by ␦PKC in vitro.
If GAPDH inhibits the docking of GAPDH to ␦PKC, it should be a selective inhibitor of phosphorylation of only GAPDH and should not affect the phosphorylation of other ␦PKC substrates. To test this prediction, we used an ex vivo Langendorff model of heart attack, an oxidative stress model that we previously showed results in ␦PKC activation and phosphorylation of many protein substrates (16,24). In this model, rat hearts are subjected to 30 min of ischemia, followed by 60 min of reperfusion (ischemia and reperfusion protocol; Fig.  3A), mimicking a heart attack. GAPDH peptide was conjugated to TAT (47)(48)(49)(50)(51)(52)(53)(54)(55)(56)(57), to enable entry of the peptide into cells (11)(12)(13), and was added to the hearts during the reperfusion period, when ␦PKC activation is maximal (15).
GAPDH  stress correlates with inhibition of GAPDH-mediated mitochondrial elimination, as well as a 70% increase in cell damage (1). We therefore predicted that GAPDH peptide treatment should increase mitochondrial elimination and reduce cell damage.
We therefore determined the effect of GAPDH peptide treatment on the response of the heart to ischemia and reperfusion injury, using the Langendorff isolated heart model (ex vivo; Fig. 4A). Unexpectedly, treatment with 1 M GAPDH peptide increased, rather than decreased, cardiac injury following ischemia and reperfusion, as measured by infarct size (Fig. 4,  B and C). Furthermore, when measuring the release of CK into the perfusate, a measure of cardiac myocyte lysis, an increase in cardiac injury was observed relative to TAT (vehicle) and nonpeptide controls (Fig. 4D). In addition, GAPDH peptide increased oxidative stress as measured by 4HNE, a marker of lipid peroxidation (Fig. 4E).
GAPDH Peptide Treatment Does Not Induce Mitochondrial Elimination in the Myocardial Infarction Model-During cardiac ischemia and reperfusion injury, excessive dysfunctional and damaged mitochondria are removed by processes of mitochondrial elimination such as autophagy and mitophagy (30 -32). This selective removal is a protective process that prevents the release of cytotoxic substances and reactive oxygen species from the damaged mitochondria, thus preventing damage to the healthy/functional mitochondria and neighboring cells and enabling the recovery of ATP generation in the tissue (33). Previously, we have shown that GAPDH mediates mitochondrial elimination by mitophagy after cardiac ischemia and reperfusion injury and that inhibition of GAPDH phosphorylation by ␦PKC increases GAPDH-mediated elimination of damaged mitochondria (1).
First, we confirmed that GAPDH treatment caused an inhibition of GAPDH phosphorylation (Fig. 5, A and B). We also found that GAPDH translocation to the mitochondria was completely blocked by GAPDH treatment (Fig. 5, C and D). GAPDH peptide treatment inhibited mitochondrial elimination as assessed by measuring the levels of GRP75 and aconitase (two mitochondrial matrix proteins) and TOM20 (a mitochondrial outer membrane protein) relative to a cytosolic protein, enolase (Fig. 5, E-J). These unexpected data suggest that mitochondrial elimination by ␦PKC under oxidative stress is impaired by the GAPDH peptide, thus contributing to the increased cell damage described above.
Because mitochondrial elimination can also occur by activating the canonical autophagy process, we determined the levels of mitochondrial Parkin, which marks damaged mitochondria for autophagy. We found an increase in Parkin association with the mitochondrial fraction (Fig. 5, K and L) following simulated ischemia in the presence of GAPDH, indicating that there was an increase in mitochondrial damage (34).
GAPDH Peptide Treatment Does Not Improve ATP Levels Following Ischemia and Reperfusion Injury-To confirm that the increase in mitochondrial content does not correlate with an increase in functional mitochondria, we measured ATP levels after ischemia and reperfusion in the Langendorff ex vivo model of heart attack. Even an hour after the return of oxygen and nutrients to the heart by reperfusion, ATP levels FIGURE 6. GAPDH peptide treatment inhibits GAPDH protein activity and tetramerization, in vitro and ex vivo. Measurement of GAPDH activity ex vivo (A) (n Ն4) in total heart extracts and in vitro (no peptide ϭ NP) (B) (n Ն4). C and D, GAPDH peptide significantly reduced the tetramer formation of GAPDH enzyme, when GAPDH samples were separated on non-denaturing gel electrophoresis, in vitro (n Ն4). E and F, phosphorylation of GAPDH by ␦PKC significantly reduced the tetramer formation of GAPDH enzyme, when GAPDH samples were separated on non-denaturing gel electrophoresis (n Ն4). G and H, treatment of whole heart ex vivo with GAPDH peptide significantly reduced the tetramer formation of GAPDH enzyme (n Ն4). I and J, GAPDH peptide treatment of whole heart ex vivo significantly reduced the tetramer formation of GAPDH enzyme at the mitochondria (n Ն4). Statistical analysis was performed using two-tailed unpaired Student's t test. ****, p Ͻ 0.001; ***, p Ͻ 0.005; *, p Ͻ 0.05 compared with control.
were ϳ30% lower than the levels under normoxic conditions, and these were not altered by GAPDH peptide treatment (Fig. 5M). Together, these data suggest that GAPDH peptide inhibited mitochondrial elimination following oxidative stress, but these additional mitochondria are unlikely to be functional.
GAPDH Peptide Treatment Reduces GAPDH Activity and GAPDH Tetramerization, in Vitro and Ex Vivo-Following ischemia and reperfusion injury, GAPDH phosphorylation by ␦PKC increases (1), and GAPDH activity is reduced (35). We therefore expected that treatment with GAPDH peptide, which inhibits ␦PKC phosphorylation of GAPDH (Figs. 2, B and  C, and 3, B and C), should increase GAPDH activity in the myocardial infarction model. Surprisingly, relative to TATtreated hearts, GAPDH treatment of hearts subjected to ischemia and reperfusion, a model of myocardial infarction, decreased GAPDH activity further, by over 25% (Fig. 6A). Because GAPDH is homologous to the GAPDH sequence, we then determined whether the inhibition of GAPDH activity is due to a direct effect of this peptide on GAPDH. Indeed, in the presence of GAPDH peptide (and in the absence of ␦PKC), the activity of recombinant GAPDH decreased by 80% in vitro, as compared with vehicle control (Fig. 6B).
The most active form of GAPDH is a tetramer (36,37). Using a native gel, we therefore determined the oligomerization state of recombinant GAPDH with and without the treatment of GAPDH peptide in vitro. After a 30-min incubation with GAPDH peptide, the tetrameric form of GAPDH was almost eliminated as compared with control-treated GAPDH, in vitro (Fig. 6, C and D). A similar decline in the tetrameric form of GAPDH was observed after in vitro phosphorylation of GAPDH by ␦PKC (Fig. 6, E and F). Finally, ischemia and reperfusion injury in the presence of GAPDH peptide also greatly reduced the level of tetrameric GAPDH relative to tetrameric GAPDH from ischemic hearts treated with control peptide, both in the cytosol (Fig. 6, G and H) and at the mitochondria (Fig. 6, I and J).
We then examined the crystal structure of GAPDH. Tetrameric GAPDH is a dimer of dimers with three non-equivalent P, Q, and R interfaces (Fig. 7A). The R axis marks the dimerdimer interface; the P axis is the interface with the most contacts between two monomeric subunits, and these contacts are highly conserved in evolution (38,39). Relevant to our study, we found that the sequence from which the GAPDH peptide was derived as well as the phosphorylation site Thr-246 are both found on the P axis (Fig. 7B). At the dimerization interface between subunits O and P (Fig. 7A), Glu-179 (the fourth amino acid in the ␦PKC-like sequence, GIVEGL) in subunit O makes contact with Asn-304 in subunit P, and Leu-181 sits in a hydrophobic pocket on subunit P, stabilizing the dimer formation (Fig. 7C). These interactions may explain how the presence of GAPDH peptide inhibits GAPDH oligomerization, leading to a decrease in GAPDH activity.

GAPDH Peptide Treatment Following Oxidative Stress in Cultured H9C2 Cells Recapitulates the Data Observed in Whole
Hearts under Oxidative Stress-To confirm our data, we also used H9C2 cardiac cells in culture exposed to H 2 O 2 . H 2 O 2 is a stressor that induces reactive oxygen species, thus leading to cell death. Cell death was measured by the release of LDH, a cytoplasmic enzyme, to the cell culture medium (Fig. 8A). As in the heart attack model, treatment with GAPDH peptide (1 M) under these conditions increased (rather than decreased) LDH release (Fig. 8A). Mitochondrial elimination was also inhibited by GAPDH treatment as determined by measuring the cellular levels of aconitase and TOM20 (Fig. 8, B-E).
We then determined the effect of GAPDH on GAPDH polymerization and found that cellular tetrameric GAPDH was almost eliminated (Fig. 8, F and G). GAPDH has been implicated in other functions, including change in transcription under oxidative stress, following its translocation into the nucleus (40,41). Although the time scale for changes in gene expression is too long to affect cell death measured in this study, we determined GAPDH translocation into the nucleus (42). We found that GAPDH also inhibited GAPDH translocation into the nucleus (Fig. 8, H and I), indicating that tetramerization of GAPDH is also required for its non-canonical role in regulating gene expression in the nucleus.
Finally, to confirm that GAPDH selectively affects GAPDH-mediated functions, we knocked down GAPDH in H9C2 cells, then exposed them to H 2 O 2 , and determined the effect of GAPDH treatment in the presence of pyruvate as a carbon source. In cells with GAPDH siRNA, GAPDH protein levels were reduced by 50% (Fig. 8, J and K), and the ϳ50% increase in mitochondrial content (which occurred even under basal conditions) was not increased by GAPDH treatment following simulated ischemia; if anything, there was a trend toward a declined in mitochondrial content (Fig. 8, L-M). Consistent with the lack of GAPDH in cells treated with GAPDH siRNA, cell viability in these cells also was unaffected by GAPDH as compared with TAT (Fig. 8N). Finally, treatment with scrambled siRNA did not affect any of these measurements (Fig. 8, J-N).

Discussion
Kinases have evolved at least two mechanisms to optimize substrate specificity and affinity. The first involves coordinated binding of a substrate and its kinase onto scaffolding or adaptor proteins (43). The second involves a direct interaction between the kinase and its substrates, mediated by sequences outside the catalytic site on the enzyme, at what is referred to as the substrate-docking site (5). These latter interactions have often been found to increase the affinity of kinases for their substrates (44,45). Such docking sites are prevalent in a number of serine/ threonine kinases, including JNKs, cyclin-dependent kinases, and MAPKs (46 -53), and peptides derived from these docking sequences can be used as inhibitors of the phosphorylation of the corresponding substrate. For example, Lee et al. (54) identified a six-amino acid substrate-docking site on C-terminal Src kinase (Csk), and a peptide mimicking the docking site inhib-ited Csk phosphorylation of Src (IC 50 ϭ 21 M) and only moderately inhibited Src kinase activity on other substrates.
These kinase-substrate interaction sites, and the inhibitors derived from them, are often identified by extensive mutagenesis analysis and co-crystallization studies. Here, we used a simple and fast rational approach to identify the interaction site between GAPDH and ␦PKC, and we showed that a sequence derived from this site is an inhibitor of GAPDH docking to ␦PKC. This inhibitor, GAPDH peptide, is selective for binding to and phosphorylation of GAPDH by ␦PKC; it does not affect the phosphorylation of five other ␦PKC substrates in the same preparation (Figs. 2 and 3).
We previously showed that inhibition of GAPDH phosphorylation by ␦PKC correlates with increased elimination of damaged mitochondria and increased cytoprotection under oxidative stress (1). Unexpectedly, although GAPDH peptide inhibited ␦PKC-mediated GAPDH phosphorylation, mitochondrial elimination did not increase; ATP levels did not increase, and cytoprotection was not achieved (Figs. 4, 5, and 8). These unexpected results are likely attributed to the fact that GAPDH directly decreased the activity of GAPDH by reducing the levels of the tetrameric active form of GAPDH (Fig. 6). Therefore, our identification of a novel GAPDH inhibitor elucidated the role of GAPDH tetramerization in glycolysis and in mitochondrial elimination. However, it is still to be determined whether this effect is only due to the reduced activity of GAPDH or possibly by direct action of ␦PKC on mitochondrial function. Currently, we are developing additional ␦PKC-specific inhibitors for other ␦PKC substrates using the same rational design approach that may enable us to address this important possibility.
Both the ␦PKC-mediated phosphorylation site on GAPDH (Thr-246) and the potential GAPDH-␦PKC interaction site (amino acids 169 -174 in GAPDH) are adjacent to each other (Fig. 1D), and both are at the interface between monomers in the tetrameric structure (Fig. 7). Therefore, our data are consistent with a dynamic oligomerization of GAPDH, in which either GAPDH peptide or phosphorylation of Thr-246 by ␦PKC interferes with monomer-monomer interaction, thus decreasing the levels of the catalytically competent GAPDH tetramer. Our data are also consistent with tetrameric GAPDH as a mediator of mitochondrial elimination, as decreased tetramerization by both GAPDH and Thr-246 phosphorylation resulted in decreased mitochondrial elimination (Figs. 5 and 8) (1) and increased cell injury (Figs. 4 and 8).
In addition to its role in glycolysis, GAPDH plays a critical role in nuclear events. GAPDH translocation to the nucleus has  4). The level of TOM20 is measured by the level of the protein divided by the level of ␤-actin. F and G, GAPDH peptide significantly reduced the tetramer formation of GAPDH enzyme in H9C2 cells after oxidative stress (samples were separated on non-denaturing gel electrophoresis; n ϭ 2). H and I, GAPDH peptide (1 M) treatment reduced nuclear translocation of GAPDH following oxidative stress in culture, as measured by immunofluorescence; I provides quantitation of the immunofluorescence data. J and K, siRNA knockdown reduced GAPDH levels by ϳ50% (n ϭ 4). L and M, GAPDH peptide had no effect on mitochondrial content in GAPDH-silenced H9C2 cells after oxidative stress (n ϭ 4). N, GAPDH did not increase cell death (as measured by LDH release) in cells subjected to oxidative stress, when GAPDH was silenced (n ϭ 6). Statistical analysis was performed using two-tailed unpaired Student's t test. **, p Ͻ 0.01; *, p Ͻ 0.05 compared with control.
been associated with change in gene expression, leading to apoptosis in a number of cell systems; blocking this translocation was found to be sufficient to reduce cytotoxicity (55,56). Transfection with GAPDH-specific siRNA resulted in substantial reduction of GAPDH expression to Ͻ55% of control as shown in Fig. 8, J and K. GAPDH silencing had no effect on cell viability but showed a general increase in mitochondrial content. GAPDH peptide had no effect in silenced H9C2 cells. Our data also suggest that although nuclear translocation of GAPDH is essential for initiating apoptosis, translocation is regulated by the GAPDH oligomerization state and that blocking with GAPDH significantly reduced nuclear GAPDH. These data indicate that the oligomerization state of GAPDH is critical for at least two non-canonical functions of GAPDH outside glycolysis.
GAPDH is a vital glycolytic enzyme that is involved in DNA repair (57), tRNA export (58), membrane fusion and transport (59), cytoskeletal dynamics (60), cell death (55), as well as in elimination of damaged mitochondria following oxidative stress (1). GAPDH has been implicated in various human diseases; increased nuclear localization of GAPDH correlates with increased neuronal apoptosis in Parkinson disease (61), Huntington disease (62), and Alzheimer disease (63), and increased glycolytic activity correlates with cancer cell survival (64). GAPDH is also an attractive target for drugs against protozoan parasites, whose bloodstream forms depend solely on glycolysis for energy (65). Therefore, an inhibitor of GAPDH glycolytic activity that also increases oxidative stress-induced injury by damaged mitochondria, such as GAPDH, may be an attractive drug lead for cancer as well as an anti-parasitic drug.
Author Contributions-N. Q. and D. M. R. conceived the experiments, analyzed the data, and wrote the article, which was read by all the authors. N. Q., A. U. J., and J. C. B. F. performed the experiments, and A. D. C. analyzed the GAPDH crystal structure.