PGAM5, a Bcl-XL-interacting Protein, Is a Novel Substrate for the Redox-regulated Keap1-dependent Ubiquitin Ligase Complex*

Keap1 is a BTB-Kelch substrate adaptor protein for a Cul3-dependent ubiquitin ligase complex that functions as a sensor for thiol-reactive chemopreventive compounds and oxidative stress. Inhibition of Keap1-dependent ubiquitination of the bZIP transcription factor Nrf2 enables Nrf2 to activate a cyto-protective transcriptional program that counters the damaging effects of oxidative stress. In this report we have identified a member of the phosphoglycerate mutase family, PGAM5, as a novel substrate for Keap1. The N terminus of the PGAM5 protein contains a conserved NXESGE motif that binds to the substrate binding pocket in the Kelch domain of Keap1, whereas the C-terminal PGAM domain binds Bcl-XL. Keap1-dependent ubiquitination of PGAM5 results in proteasome-dependent degradation of PGAM5. Quinone-induced oxidative stress and the chemopreventive agent sulforaphane inhibit Keap1-dependent ubiquitination of PGAM5. The identification of PGAM5 as a novel substrate of Keap1 suggests that Keap1 regulates both transcriptional and post-transcriptional responses of mammalian cells to oxidative stress.

Diverse metabolic activities of eukaryote cells generate reactive oxygen species, which can damage biological macromolecules, including DNA, proteins, and lipids (1). Oxidative damage to biological macromolecules can alter multiple cellular functions and has been implicated in cancer, inflammation, cardiovascular and neurodegenerative diseases, and aging (2)(3)(4)(5)(6). Consequently, eukaryote cells have evolved multiple protective mechanisms against reactive molecules and oxidative stress. A comprehensive analysis of oxidative stress resistance in yeast has identified several core functions, including transcription, protein trafficking, and vacuolar formation that are broadly required for oxidative stress resistance (7).
In mammals, the Keap1 protein has emerged as a major sensor for oxidative stress and electrophilic molecules. Keap1 is a Bric-a-Brac (BTB) 2 -Kelch protein that functions as a substrate adaptor protein for a Cul3-dependent E3 ubiquitin ligase complex. Under basal conditions of cellular redox homeostasis, the N-terminal BTB domain of Keap1 binds to Cul3, whereas the C-terminal Kelch domain of Keap1 binds to the bZIP transcription factor Nrf2 (8 -10). This ubiquitin ligase complex conjugates ubiquitin onto specific lysine residues within the N-terminal Neh2 domain of Nrf2 and targets Nrf2 for degradation by the 26 S proteasome. Cyclical association and dissociation of this E3 ubiquitin ligase complex is required for efficient ubiquitination of Nrf2 and repression of Nrf2-dependent gene expression (11). Reactive electrophilic chemicals and oxidative stress modify one or more cysteine residues located in the N-terminal BTB and central linker domains of Keap1 and disrupt formation of a functional E3 ubiquitin ligase complex (12)(13)(14)(15)(16)(17). Inhibition of Keap1-mediated ubiquitination of Nrf2 results in a marked increase in the steady-state level of Nrf2 and transcription of Nrf2-dependent genes. Induction of Nrf2-dependent genes represents a cytoprotective response to oxidative stress that neutralizes reactive molecules, eliminates damaged macromolecules, and restores cellular redox homeostasis (18 -22).
Structural and functional analyses of Keap1 and Nrf2 have identified an evolutionarily conserved DXETGE motif in the N-terminal Neh2 domain of Nrf2 that binds in a shallow binding pocket on the top face of the Kelch domain of Keap1 (23)(24)(25). Although Nrf2 is the only known substrate for Keap1, the Kelch domain of Keap1 has been reported to bind several other proteins, including Nrf1, prothymosin ␣, and fetal Alz clone 1 (26 -28). In general, substrate adaptors for cullin-dependent ubiquitin ligases often bind several substrates via a common recognition motif (29 -32). For example, ␤-TrCP, a well characterized substrate adaptor for Cul1, recognizes two substrate proteins, IB␣ and ␤-catenin, after these proteins are phosphorylated on two serine residues embedded within a conserved sequence motif (32,33).
In this report we have identified a novel substrate of Keap1 encoded by the PGAM5 gene. The PGAM5 gene encodes two protein isoforms, PGAM5-L and PGAM5-S, which result from alternative splicing. Both PGAM5 isoforms contain an N-terminal region of ϳ100 amino acids, which includes a conserved NXESGE motif that is required for binding to Keap1, and a C-terminal phosphoglycerate mutase (PGAM) domain, which binds to Bcl-X L (34). PGAM5 is ubiquitinated by a Keap1-dependent E3 ubiquitin ligase complex, which targets PGAM5 for proteasome-mediated degradation. Both quinone-induced oxidative stress and sulforaphane inhibit Keap1-dependent ubiquitination and subsequent degradation of PGAM5. Our results establish PGAM5 as a novel substrate of Keap1 and support the notion that Keap1 regulates diverse cellular functions as part of a global response to oxidative stress.

EXPERIMENTAL PROCEDURES
Construction of Recombinant DNA Molecules-The PGAM5-L cDNA (accession number BG030775) and the PGAM5-S cDNA (accession number BC008196) were obtained from ATCC and Open Biosystems, respectively. The PGAM5 expression vectors were constructed by cloning PCR-generated fragments containing the entire open reading frame of either PGAM5-L and PGAM5-S into the BamHI/EcoRI sites of pcDNA3 (Invitrogen) and the NotI/SmaI sites of pFLAG-CMV2 (Sigma). The mutant PGAM5-L cDNA containing alanines in place of Glu-79 and Ser-80 codons were generated by site-specific mutagenesis with standard overlap-extension techniques. Plasmids expressing wild type and the mutant Keap1, chitin binding domain (CBD)-tagged Keap1, and Cul3 proteins have been previously described (10,11). The Myc-Rbx1 expression vector was a gift from Dr. Joan Conaway (35). The Bxl-X L expression vector was a gift from Dr. Richard Youle (36). All the genes used in this study were sequenced in the context of the expression vectors used for the experiments.
Cell Culture, Transfections, and Chemical Reagents-MDA-MB-231, COS1, HEK-293T, and HeLa cells were purchased from the ATCC. Cells were maintained in either Dulbecco's modified Eagle's medium or Eagle's minimal essential medium in the presence of 10% fetal bovine serum. Plasmid DNA transfections were performed with Lipofectamine Plus (Invitrogen) according to the manufacturer's instructions. Clasto-lactacystin ␤-lactone (lactacystin) was purchased from Boston Biochem. Sulforaphane and tert-butylhydroquinone were purchased from Sigma.
Immunofluorescence Assays-HeLa and COS1 cells were grown on glass coverslips on 35-mm plates. Cells were transfected with 1.0 g of the empty vector or the expression vectors for the FLAG-PGAM5 proteins. Cells were fixed with 100% methanol at Ϫ20°C for 10 min. Fixed cells were incubated for 40 min with mouse anti-FLAG M2 antibodies (Sigma) in phosphate-buffered saline (10 mM sodium phosphate (pH 8.0) and 150 mM NaCl) containing 10% (v/v) fetal bovine serum. Coverslips were washed and incubated with fluorescein isothiocyanate-conjugated anti-mouse antibodies (Jackson Laboratories) for another 40 min. Coverslips were washed and incubated with To-pro-3 (Molecular Probes) for another 30 min. Coverslips were washed and mounted on glass slides. Images were obtained with a Bio-Rad Radiance 2000 confocal system coupled to an Olympus IX70 inverted microscope and a digital camera. The images were captured with ImagePro and transferred to Adobe Photoshop for construction of the figure.
For detection of protein expression in total cell lysates, cells were lysed in sample buffer (50 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, 100 mM dithiothreitol (DTT), 0.1% bromphenol blue) at 24 -48 h post-transfection. For immunoprecipitation assays cell extracts were prepared in radioimmune precipitation assay buffer (10 mM sodium phosphate (pH 8.0), 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) containing 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor mixture (Sigma). Soluble cell lysates were incubated either with 2 g of anti-FLAG M2 coupled to agarose (Sigma) or with 2 g of affinity-purified antibodies for 2 h at 4°C followed by incubation at 4°C with protein A-agarose beads (Sigma) for 2 h. Unbound proteins were removed by washing four times with lysis buffer. The immunoprecipitated proteins were eluted in sample buffer by boiling for 5 min, electrophoresed through SDS-polyacrylamide gels, transferred to nitrocellulose membranes, and subjected to immunoblot analysis.
In Vivo Ubiquitination Assay-For detection of ubiquitinated PGAM5-L proteins in vivo, cells were transfected with expression vectors for HA-tagged ubiquitin, FLAG-PGAM5-L, and the indicated proteins. Cells were lysed by boiling in a buffer containing 2% SDS, 10 mM N-ethylmaleimide, 150 mM NaCl, and 10 mM Tris-HCl (pH 8.0). This rapid lysis procedure inactivates cellular ubiquitin hydrolases and, therefore, preserves ubiquitin-substrate protein conjugates present in cells before lysis. Protein-protein interactions, including association of PGAM5-L with Keap1, are also disrupted by this lysis procedure. For immunoprecipitation, these lysates were diluted with 4 volumes of a buffer containing 150 mM NaCl, 10 mM Tris-HCl (pH 8.0), 1 mM DTT, and 1% Triton X-100. The diluted lysates were precleared with protein A-agarose beads (Sigma) and incubated with anti-FLAG M2 agarose (Sigma). HA-ubiquitin conjugates in immunoprecipitated FLAG-PGAM5-L proteins were analyzed by immunoblot analysis with antibodies against the HA epitope.
Mass Spectrometry-Gel slices containing bands A, B, and C as indicated as in Fig. 1A were reduced, alkylated, and digested with trypsin. Tryptic peptides were desalted and subjected to MALDI-TOF mass spectrometry by the MU Proteomics Core using a Voyager DEPro with a 20-Hz 337-nm nitrogen laser (Applied Biosystems, Foster City, CA). The Mascot software package was used to match the mass of the peptides with predicted tryptic peptides generated from the translated human genome. Mowse scores of 74 and 172 were obtained for PGAM5 (band B) in two independent experiments. The cut-off values for statistical significance (p Ͻ 0.05) in these two experiments were 59 and 63, respectively. The matched peptides from the second experiment are shown in Table 1. A Mowse score of 59 was obtained for CH3L1 (Band A) in the first experiment (cut-off value at p Ͻ 0.05 was 59 in this experiment). Details of the mass spectrometry experiments are available upon request.

Identification of PGAM5 as an Interaction Partner of Keap1-
To identify new Keap1-interacting proteins, the human Keap1 protein containing a C-terminal CBD was stably expressed in MDA-MB-231 cells, a human breast carcinoma-derived cell line. The Keap1-CBD protein was isolated from cell lysates using chitin beads, and Keap1-associated proteins were visualized on silver-stained SDS-polyacrylamide gels. Three proteins, ranging in size from 30 to 39 kDa were found to co-purify with the Keap1-CBD protein in a specific and reproducible manner (Fig. 1A, lanes 3 and 4). Tryptic peptides from these three proteins were subjected to MALDI-TOF mass spectrometry. Band A (39 kDa) was tentatively identified as chitinase-3-like protein 1 (CH3L1; NCBI accession number NP_001267), with a borderline Mowse score of 59 in one experiment (the cut-off value for p Ͻ 0.05 was 59; data not shown). Bands B and C (32 and 30 kDa, respectively) were identified as phosphoglycerate mutase 5 (PGAM5; NCBI accession number NP_612642) in two independent experiments. The Mowse scores for band B were 74 and 172, well above the cut-off values for statistical significance of 59 and 63, respectively. The position and sequence of the PGAM5-derived peptides identified by MALDI-TOF MS are given in Table 1.
A data base search revealed the presence of two mRNA transcripts that originate from the PGAM5 gene located on chromosome 12. The proteins encoded by these mRNAs are identical from amino acid 1-239, with the shorter form (PGAM5-S) containing 16 additional C-terminal amino acids and the longer isoform (PGAM5-L) containing 50 additional C-terminal amino acids ( Fig. 2A). Both isoforms of human PGAM5 contain a PGAM domain (pfam00300), which begins at amino acid 98 and extends to the C-terminal end of the longer isoform. The PGAM5 isoforms are distant members of the phosphoglycerate mutase family, with their closest relatives being two proteins, termed STS-1 and STS-2, that do not have phosphoglycerate mutase activity but participate in receptor-mediated signal transduction pathways (37,38) (Fig. 2B). In particular, the phosphohistidine signature motif that is characteristic of PGAM domains with enzymatic activity is poorly conserved in PGAM5 (39) (Fig. 2D). A region within the PGAM domain of PGAM5 has previously been shown to bind to Bcl-X L (34), and the ability of both isoforms of PGAM5 to associate with Bcl-X L was confirmed by coimmunoprecipitation ( Fig. 2E and data not shown).
To confirm association between Keap1 and the two PGAM5 isoforms, untagged or FLAG-tagged versions of both PGAM5-S and PGAM5-L were expressed in HEK-293T cells along with the Keap1-CBD protein. Proteins that bound to chitin beads were analyzed by SDS-PAGE and silver stain. Both PGAM5 isoforms bound to the Keap1-CBD protein (Fig. 1B, lanes [3][4][5][6]. The presence of the N-terminal FLAG tag decreased the mobility of both PGAM5 isoforms but did not perturb their ability to associate with Keap1-CBD (Fig. 1B, lanes 7 and  8). The ectopic untagged PGAM5-L protein co-migrated exactly with the endogenous 32-kDa Keap1-associated protein (Fig. 1B, compare  lanes 1 and 3). In contrast, the FIGURE 2. A, the two isoforms of PGAM5 are depicted with relevant regions of the proteins indicated. The two isoforms are identical up to amino acid 239 and have different C termini as the result of alternative splicing. The NXESGE motif is located between amino acids 77 and 82, and the Bcl-X L interaction region is located between amino acids 125 and 156. The PGAM domain extends from amino acid 98 to the C terminus (amino acid 289) of the PGAM5-L isoform. B, the relationship between 10 human proteins that contain a PGAM domain is depicted by a cladogram generated using the ClustalW program at European Bioinformatics Institute website. The protein sequences were obtained from the Swiss Institute of Bioinformatics website using the following accession numbers (in parentheses): PGAM1 (P18669); PGAM2 (P15259); BPGM (P07738); STS-1 (Q96IG9); STS-2 (P57075); PGAM5 (Q96HS1); PFKFB1 (P16118); PFKFB2 (O60825); PFKFB3 (Q16875); PFKFB4 (Q16877). Several PGAM1-derived pseudogenes, including an expressed pseudogene termed PGAM4 (Q8N0Y7), were not included in this analysis (44). C, mammalian PGAM5 proteins contain a conserved NXE(S/T)GE motif that is similar to the conserved DXETGE motif in the human Nrf2 protein (Q16236). The NCBI accession numbers for the indicated PGAM5 proteins are in parentheses.  a The position and sequence of PGAM5-derived peptides were identified by MALDI-TOF mass spectroscopy. b The calculated mass of the indicated PGAM5-derived peptides. c The observed mass of the indicated PGAM5-derived peptides by mass spectrometry. d The amino acid composition of the indicated peptides was confirmed by tandem mass spectroscopy. These two peptides were only observed in band B of Fig. 1A. All other peptides were observed in both bands B and C. mobility of the ectopic untagged PGAM5-S protein was slightly faster than the endogenous Keap1-associated protein of 30 kDa (Fig. 1B, compare lanes 1 and 5). This result supports the identification of the Keap1-associated protein of 32 kDa as the PGAM5-L isoform and suggests that the Keap1-associated protein of 30 kDa is not the PGAM5-S isoform but is derived from the 32-kDa protein by proteolysis during purification.
To verify the association of both isoforms of human PGAM5 with Keap1, reciprocal pull-down analyses of the proteins in transfected cells were performed. COS1 cells were transfected with expression vectors for untagged Keap and either of the FLAG-tagged PGAM5 isoforms. Cell lysates were immunoprecipitated with anti-FLAG antibodies, and the resultant immunoprecipitates were probed for the presence of Keap1. The Keap1 protein was present in anti-FLAG immunoprecipitates of both PGAM5 isoforms (Fig. 1C, lanes 2 and 4).
The subcellular localization of both PGAM5 isoforms was determined by indirect immunofluorescence of HeLa cells expressing the FLAG-tagged proteins. The PGAM5-L protein displayed diffuse cytoplasmic staining and a punctate nuclear staining, whereas the PGAM5-S protein displayed a punctate pattern in the cytoplasm (Fig. 3). A similar pattern was observed in COS1 cells (data not shown). Both PGAM isoforms co-localized with the Keap1 protein in the cytoplasm (data not shown). No nuclear staining of Keap1 was observed in our experiments. When co-expressed with Keap1, the cytoplasmic localization of both PGAM5-L and Keap1 shifted from a diffuse pattern to a punctate pattern (data not shown). Keap1 also displayed a punctate cytoplasmic pattern when co-expressed with PGAM5-S (data not shown). In subsequent experiments we focused on the PGAM5-L isoform, as this isoform co-migrated with the 32-kDa protein identified as PGAM5 by mass spectrometry of tryptic peptides and is expressed at higher levels than the PGAM5-S isoform (Fig. 1B).
Identification of the Interaction Interface between Keap1 and pGAM5-L-The Keap1 protein possesses five discrete domains: the N-terminal region (N); an N-terminal Broad complex, Tramtrack, and BTB; a central linker domain (Linker); a C-terminal Kelch repeat domain (Kelch); and a C-terminal short tail (C). To determine which domain(s) in Keap1 is responsible for binding to PGAM5, a series of mutant Keap1 proteins containing a specific deletion of each individual domain was expressed in HEK-293T cells along with FLAG-PGAM5-L proteins. Association of the mutant Keap1 proteins and PGAM5-L was assessed by coimmunoprecipitation. Only the Keap1-⌬Kelch protein did not associate with PGAM5-L (Fig. 4A, lane 5), indicating that the Kelch domain of Keap1 is required for binding to PGAM5-L.
Previous structural and mutational studies have demonstrated that the Kelch domain of Keap1 is a six-bladed ␤-propeller structure that contains a shallow pocket on its top face that binds Nrf2 (23,25,40). The shallow pocket of Keap1 binds an Nrf2-derived peptide containing a highly conserved DXETGE motif. Our previous analysis of more than 30 Keap1 mutant proteins identified seven amino acids in Keap1 whose side chains participate in direct contact with the Nrf2-derived peptide and for which individual alanine substitutions are suf-ficient to disrupt binding of Nrf2 to Keap1 (23). These residues are Tyr-334, Arg-380, Asn-382, Arg-415, Arg-483, Tyr-525, and Tyr-572. The charged residues are located at the base of the binding pocket, whereas the hydrophobic residues line the sides of the binding pocket. A set of mutant CBD-tagged Keap1 proteins were characterized for their ability to bind pGAM5-L. Association of the endogenous PGAM5-L protein with these mutant Keap1 proteins was determined by co-purification on chitin beads followed by SDS-PAGE and visualization by silver stain. This analysis revealed 4 amino acid residues, including Tyr-334, Arg-415, Arg-483, and Tyr-572, for which individual alanine substitutions significantly disrupted the ability of Keap1 to bind PGAM5-L (Fig. 4, B, lanes 2 and 13, and C, lanes  5 and 7).
The highly conserved DXETGE motif in Nrf2 is required for the association of Nrf2 with Keap1 (23-25). The side chains of the glutamate residues contact multiple residues in Keap1, including Arg-380, Asn-382, Arg-415, and Arg-483. The side chains of the aspartate and threonine residues, on the other hand, interact with the backbone of the Nrf2-derived peptide and stabilize a ␤-turn region that enables the Nrf2-derived peptide to fit into the binding pocket on Keap1. Alanine substitution of the threonine residue within the DXETGE motif markedly reduces binding of Nrf2 to Keap1 and enables the mutant Nrf2 protein to escape Keap1-mediated repression (23). In contrast, serine substitution of this threonine residue did not disrupt the binding of Nrf2 to Keap1, suggesting that the hydroxyl group of the serine residue is an effective substitute for the hydroxyl group of the threonine residue. A NXE(S/T)GE motif, amino acids 77-82 in the human PGAM5 isoforms, is strongly conserved in mammals (Fig. 2C). To determine whether the NXESGE motif in PGAM5-L is required for binding to Keap1, a mutant PGAM5-L protein was constructed in which two alanine substitutions were introduced in place of Glu-79 and Ser-80. FLAG-tagged versions of the wild-type and PGAM5-L-E79A/S80A mutant proteins were expressed in HEK-293T cells along with Keap1-CBD, and association with Keap1 was assessed by immunoblot after affinity purification with anti-FLAG beads. The mutant PGAM5-L-E79A/S80A protein did not bind to Keap1 (Fig. 4D, lane 5). The cytoplasmic localization (Fig. 3M) and association of PGAM5-L with Bcl-X L (Fig.  4D) were not altered by the E79A/S80A mutation. Taken together, these results indicate that four amino acids within the substrate binding pocket on the top face of the Kelch domain of Keap1 form critical contacts with the NXESGE motif in PGAM5-L and support the notion that Keap1 binds PGAM-L in a manner that is very similar to the way in which Keap1 binds Nrf2.

Keap1 Targets PGAM5-L for Ubiquitin-dependent
Degradation-Binding of Nrf2 to Keap1 results in the ubiquitination of specific lysine residues in Nrf2 by a Cul3-dependent ubiquitin ligase complex, leading to the proteasome-dependent degradation of Nrf2 (8 -10). Several experimental approaches were used to determine whether PGAM5-L is also a substrate for Keap1-dependent ubiquitination. In one set of experiments an in vivo ubiquitination analysis was performed in which expression vectors for FLAG-PGAM5-L, Keap1, and HA-ubiquitin were transfected into COS1 cells. Cell lysates were col-lected under strongly denaturing conditions to eliminate non-covalent protein-protein interactions, and the presence of HA-ubiquitin conjugation onto the FLAG-PGAM5-L protein was measured by anti-HA immunoblot analysis of anti-FLAG immunoprecipitates. A low level of ubiquitin conjugation was observed in the absence of coexpressed Keap1 (Fig. 5A, lane 2). Coexpression of Keap1 resulted in a marked increase in ubiquitin conjugation onto PGAM5-L, including a distinct band with apparent molecular weight of 40 kDa and a prominent ladder comprising multiple bands of higher apparent molecular weight (Fig. 5A, lane 3). Ubiquitin conjugation onto the mutant PGAM5-L-E79A/S80A protein was not observed (Fig. 5A, lane 4), consistent with the inability of the mutant PGAM5-L protein to associate with Keap1 (Fig. 4D, lane 5). Mutant Keap1 proteins defective in binding to PGAM5-L were likewise defective in directing ubiquitin conjugation of PGAM5-L (data not shown).
In a second set of experiments, the ability of Keap1 to recruit PGAM5-L into a Cul3-dependent ubiquitin ligase complex capable of targeting PGAM5-L for ubiquitination was determined. Expression vectors for FLAG-tagged PGAM5-L, Keap1, HA-Cul3, and Myc-Rbx1 were cotransfected into COS1 cells. Cell lysates were immunoprecipitated with anti-FLAG antibodies and analyzed for the presence of Keap1, Cul3, and Rbx1 by immunoblotting with specific antibodies. Coimmunoprecipitation of either Cul3 and Rbx1 was markedly enhanced in the presence of co-expressed Keap1 (Fig. 5B, compare lanes 2 and 3), consistent with the notion that Keap1 bridges the interaction between PGAM5-L and the Cul3⅐Rbx1 complex. Furthermore, co-expression of Cul3 and Rbx1 markedly increased Keap1-dependent ubiquitination of PGAM5-L (Fig. 5C, compare lanes 2-4). Finally, the quaternary PGAM5-L⅐Keap1-CBD⅐Cul3⅐Rbx1 complex was purified from transfected cells using chitin beads, and the ability of this complex to support ubiquitin conjugation onto PGAM5-L was determined. Ubiquitin conjugation onto PGAM5-L was readily observed in the presence of E1 activating enzyme, the UbcH5 E2 conjugating enzyme, ubiquitin, and ATP (Fig. 5D, lane 2). Ubiquitin conjuga- Lysates from four 60-mmdiameter dishes were pooled for each sample and incubated with chitin beads. After washing the chitin beads were incubated with E1, E2-UbcH5a, ubiquitin, and ATP. The E1 enzyme was omitted from one sample (lane 1). Subsequently, the chitin beads were pelleted and washed, and proteins that were eluted from the beads under denaturing conditions were split into two sets of samples. One set was immunoprecipitated with anti-FLAG M2 agarose beads and then analyzed by immunoblotting with anti-ubiquitin antibodies (top panel). The other set was subjected to immunoblot analysis with the indicated antibodies (bottom four panels). DECEMBER 8, 2006 • VOLUME 281 • NUMBER 49 tion onto PGAM5-L was not observed with reactions carried out in the absence of the E1 ubiquitin-activating enzyme (Fig. 5D, lane 1), confirming that conjugation of ubiquitin onto PGAM5-L occurred in vitro. A low level of ubiquitin conjugation onto PGAM5-L was observed in the absence of the Cul3⅐Rbx1 subcomplex (Fig. 5D, lane 3), presumably due to copurification of trace amount of the endogenous Cu3⅐Rbx1 subcomplex with ectopically expressed Keap1-CBD protein.

PGAM5, a Novel Substrate for Keap1
To determine whether the ubiquitination of PGAM5-L by Keap1 alters the stability of the PGAM5-L protein, the steadystate levels of the wild-type and E79A/S80A mutant PGAM5-L proteins in the absence and presence of Keap1 were compared. Both proteins were expressed at comparable levels in HeLa cells in the absence of Keap1 (Fig. 6A, compare lanes 1 and 4). However, co-expression of Keap1 markedly decreased steady-state levels of the wild type but not the mutant PGAM5-L protein (Fig. 6A, compare lanes 2 and 3 with lanes 5 and 6). The half-life of the PGAM5-L protein in the absence or presence of Keap1 was determined in a cycloheximide pulse-chase experiment. Transfected HeLa cells were treated with cycloheximide to block protein synthesis, and the level of the PGAM5-L protein was determined at time points after the addition of cycloheximide. The half-life of the ectopically expressed PGAM5-L protein was 75 h in the absence of co-expressed Keap1 but was reduced to 6 h in the presence of coexpressed Keap1 (Figs. 6, B  and C).
To confirm that Keap1 targets PGAM5-L for degradation by the 26 S proteasome, the ability of the proteasome-specific inhibitor, lactacystin, to block Keap1-dependent degradation and increase steady-state levels of PGAM5-L was determined. HeLa cells transfected with expression vectors for PGAM5-L and Keap1 were exposed to either Me 2 SO or lactacystin followed by cycloheximide to block protein synthesis. The level of PGAM5-L was determined at 0, 2, 6, and 8 h after cycloheximide treatment. As before, the level of PGAM5-L was reduced in the presence of cycloheximide (Fig. 6D, lanes 1-4). However, levels of PGAM5-L did not decrease in the presence of lactacystin (Fig. 6D, lanes 5-8). Taken together these experiments demonstrate that Keap1-mediated ubiquitination of PGAM5-L protein targets the PGAM5-L protein for degradation by the proteasome.

Keap1-dependent Ubiquitination of pGAM5-L Is Blocked by Sulforaphane and Quinone-induced Oxidative Stress-Several
reports have demonstrated that oxidative stress and thiol-reactive chemopreventive compounds block the ability of Keap1 to target Nrf2 for ubiquitin-dependent degradation (10,17,41). As a result, steady-state levels of Nrf2 are increased, and transcription of Nrf2-dependent target genes is increased. Because Keap1 binds the PGAM5-L protein and targets the PGAM5-L protein for ubiquitin-dependent degradation, the ability of both overt quinone-induced oxidative stress and sulforaphane, a chemopreventive isothiocyanate compound, to inhibit Keap1dependent ubiquitination of PGAM5-L was determined. Both quinone-induced oxidative stress and sulforaphane markedly diminished Keap1-dependent ubiquitination of PGAM5-L (Fig. 7A, lanes 3 and 4) and increased steady-state levels of PGAM5-L (Fig. 7B, lanes 3-8). Treatment of cells with either lactacystin or sulforaphane also increased the amount of the endogenous PGAM5 that was associated with Keap1-CBD (Fig. 7C).  ; lanes 2, 3, 5, and 6). Total cell lysates were subjected to immunoblot analysis with anti-FLAG and anti-tubulin antibodies. B, 35-mm-diameter dishes of HeLa cells were transfected with an expression vector for FLAG-PGAM5-L (0.8 g) along with an empty vector (lanes 1-6) or an expression vector for Keap1 (0.2 g, lanes 7-12). Cells were treated with 50 g/ml cycloheximide. Total cell lysates were collected at the indicated time points after cycloheximide treatment and subjected to immunoblot analysis with anti-FLAG and anti-tubulin antibodies. C, the relative intensities of FLAG-PGAM5-L were quantified by using the Multi Gauge software (FUJIFILM) and plotted on a semi-log graph. The amount of FLAG-PGAM5-L present at the beginning of cycloheximide treatment was set at 100. Inhibition of Keap1-dependent ubiquitination of Nrf2 by electrophilic molecules and oxidative stress has been suggested to occur through modification of one or more cysteine residues in Keap1 (10, 12, 14 -17). In previous work we have identified a single cysteine residue, Cys-151, located in the BTB domain of Keap1 that is required for inhibition of Keap1-dependent ubiquitination of Nrf2 by both quinone-induced oxidative stress and sulforaphane (10,17). Direct modification of Cys-151 in Keap1 by iodoacetamide in vitro has been demonstrated (15). In agreement with the importance of Cys-151 for inhibition of Keap1-dependent substrate ubiquitination, the ability of the Keap1-C151S mutant protein to target PGAM5-L for ubiquitin conjugation (Fig. 7A, lanes 5-7) was resistant to inhibition by either oxidative stress or sulforaphane.

DISCUSSION
The ability of Keap1 to function as a redox-sensitive substrate adaptor protein for a Cul3-based E3 ubiquitin ligase has emerged as a major regulatory mechanism that governs downstream cellular responses to oxidative stress. In the present work we have identified a member of the phosphoglycerate mutase family, PGAM5, as an interaction partner for Keap1. Human cells express two isoforms of human PGAM5 proteins, PGAM5-L and PGAM5-S, both of which bind to Keap1. We provide several lines of evidence that the PGAM5-L protein is a substrate of a Keap1-dependent ubiquitin ligase complex. First, PGAM5-L contains a conserved NXE(S/T)GE motif that is required for binding to Keap1. Because PGAM5-L and PGAM5-S only differ at their C termini, it is likely that PGAM5-S also uses this motif to bind Keap1. Second, co-expression of Keap1 with PGAM-L in cells markedly increases ubiquitin conjugation onto PGAM5-L, and Keap1-dependent ubiquitination of PGAM5-L by a Cul3⅐Rbx1-dependent E3 ubiquitin ligase complex can be reconstituted in vitro. Third, co-expression of Keap1 decreases steady-state levels of PGAM5-L, which are recovered by inhibition of the 26 S proteasome. Finally, Keap1-dependent ubiquitination of PGAM5-L is impaired by both quinone-induced oxidative stress and sulforaphane, resulting in elevated levels of PGAM5-L. Mutation of a single cysteine residue in the BTBdomain of Keap1, Cys-151, confers resistance to inhibition by All the cells were further treated with cycloheximide (50 g/ml) for six additional hours before lysis to block new protein synthesis. Cells were lysed in radioimmune precipitation assay buffer, and the lysates were incubated with chitin beads, pelleted by centrifugation (3000 ϫ g), and washed 3 times in the lysis buffer. Proteins that remained associated with the chitin beads were analyzed by SDS-PAGE and visualized by silver stain (top panel). Total cell lysates were analyzed by immunoblotting with anti-tubulin antibodies (bottom panel). DECEMBER 8, 2006 • VOLUME 281 • NUMBER 49 oxidative stress and sulforaphane of Keap1-dependent ubiquitination of PGAM5-L Taken together, these results indicate that PGAM5-L joins Nrf2 on a short list of bona fide substrates for a Keap1-dependent ubiquitin ligase complex.

PGAM5, a Novel Substrate for Keap1
The NXE(S/T)GE motif present in PGAM5 is similar in sequence to the conserved DXETGE motif in Nrf2 that binds Keap1 (24). The crystal structure of a Nrf2-derived peptide containing the DXETGE motif bound to the Kelch domain of Keap1 reveals that the hydroxyl residue of the threonine residue stabilizes a type I ␤-turn that enables the two glutamate side chains to contact multiple residues in Keap1 (23,25). A serine residue, but not an alanine residue, can substitute for the threonine residue in the Nrf2-derived DXETGE motif (23). It is likely that the serine residue in the NXESGE motif in PGAM5-L forms a similar type I ␤-turn that fits into the binding pocket in Keap1 and enables the side-chains of the flanking glutamate residues to contact specific residues in Keap1. In support of this notion, two arginine residues, Arg-415 and Arg-483, which participate in direct contact with the corresponding glutamate residues in Nrf2 (23,25), are required for binding of PGAM5-L. Taken together, our mutational data indicate that PGAM5-L binds to Keap1 in a manner that is similar, although not identical, to the way in which Nrf2 binds to Keap1. In particular, the number of critical contacts formed between Nrf2 and Keap1 appear to be greater than the number of critical contacts formed between PGAM5-L and Keap1. Two other proteins, prothymosin ␣ and fetal Alz clone 1, also interact with Keap1 in a way that suggests these two proteins may also be substrates of Keap1 (27,42). However, neither prothymosin ␣ nor fetal Alz clone 1 has an (D/N)XE(S/T)GE-like motif, suggesting that the binding pocket in the Kelch domain of Keap1 may recognize several different binding motifs.
Although both Nrf2 and PGAM5-L are substrates for Keap1dependent ubiquitination and are subsequently degraded by the 26 S proteasome, PGAM5-L is a more abundant protein with a longer half-life than Nrf2. Keap1-dependent ubiquitination of Nrf2 results in a rapid turnover of Nrf2, with a half-life of ϳ30 min under basal conditions (17). In contrast, PGAM5-L has a half-life under basal conditions of ϳ6 h when co-expressed with Keap1. This marked difference in the rate of protein turnover between Nrf2 and PGAM5-L likely reflects the rate at which multiubiquitination chains are added to Nrf2 or PGAM5-L. Several factors, including the affinity of the respective substrate proteins for Keap1 and the spatial orientation of the targeted lysine residues in the substrate, will contribute to the rate at which a given substrate acquires multiubiquitination chains. The importance of processivity of ubiquitin ligase complexes for determining the relative rates of substrate degradation has been suggested by a recent study of the anaphase-promoting complex. The anaphase-promoting complex (APC) has several different substrates that are degraded in a temporal manner during progression through mitosis and G 1 , and the processivity of APC toward the different substrates has been suggested to determine the temporal order of substrate degradation (43).
PGAM5 is a member of the phosphoglycerate mutase superfamily. In humans, the phosphoglycerate mutase superfamily consists of at least 10 distinct protein-encoding genes defined by the presence of the evolutionarily conserved PGAM domain (pfam00300). These genes can be divided into three groups based on sequence similarities and known functions (Fig. 2B). One group includes two genes that encode enzymes with known phosphoglycerate mutase activity (PGAM1 and PGAM2) and one gene that encodes an enzyme with diphosphoglycerate mutase activity (BPGM). This group also includes several pseudogenes, at least one of which is expressed (44). A second group of four genes encode isozymes of the bifunctional 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase enzyme (PFKFB). This group includes the TIGAR protein, encoded by p53-inducible gene, that may contribute to dysregulation of glycolysis and apoptosis in cancer cells (45). A third group of proteins includes two related proteins, termed STS-1 and STS-2, and PGAM5. STS-1 and STS-2, which associate with the Cbl ubiquitin ligase, contain a ubiquitin-associated motif and an SH3 domain in addition to a C-terminal PGAM domain and have been implicated in regulation of T-cell receptor signaling and endocytosis of the receptor-tyrosine kinases (37,38).
A common feature of phosphoglycerate mutase, diphosphoglycerate mutase, and the 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase enzymes is the use of a phosphohistidine intermediate to mediate the transfer of phosphate to carbon atoms of their respective substrates (39). This catalytic histidine residue is embedded within a highly conserved signature motif of (L/I/V/M)XRHG(E/Q)XXXN (Prosite PS00175) (39). The catalytic histidine residue is present in the proteins encoded by the third group of PGAM-domain containing proteins, although the C-terminal Asn residue in this signature motif, which is part of the active site of PGAM1 (39), has not been conserved in STS-1, STS-2, and PGAM5 (Fig. 2D). The PGAM domains in STS-1 and STS-2 do not appear to have phosphotransfer activity toward phosphoglycerate (43,46). Rather, the PGAM domains in STS-1 and STS-2 function as protein-protein interaction domains that enable the STS proteins to form dimers (38). Preliminary results suggest that the PGAM5 isoforms also form dimers. 3 Because the phosphohistidine signature motif of PGAM5 is the most divergent of all PGAM-domain containing proteins, it is likely that the PGAM5 proteins, like STS-1 and STS-2, will function as a mediators of proteinprotein interactions rather than as a metabolic enzymes.
Our results confirm a previous report that identified PGAM5 as a Bcl-X L -binding protein using in vitro mRNA display and peptide binding analysis (34). Bcl-X L is a member of the Bcl-2 protein family that contains both anti-apoptotic (Bcl-2, Bcl-X L , etc.) and pro-apoptotic members (Bax, Bak, Bad, etc.). A balance between the actions of anti-apoptotic and pro-apoptotic members is essential to maintain tissue homeostasis in multicellular organisms. The formation of homodimeric and heterodimeric complexes between Bcl-2 family members, mediated in large part by the conserved BH3 domain, regulates the activity of the Bcl-2 proteins. For example, BH3-derived peptides from the BH3-only proteins BID or BIM are able to induce oligomerization of the pro-apoptotic BAX or BAK proteins, with subsequent induction of apoptosis (47). These BH3-de-rived peptides bind in a hydrophobic cleft located on the surface of the Bcl-2 family proteins. A PGAM5-derived peptide, amino acids 125-156, has been shown to bind Bcl-X L (34). However, binding of the PGAM5-derived peptide to Bcl-X L was not competed by a peptide derived from the BH3 domain of Bak, although a Bim-derived BH3 peptide was a very effective competitor for the Bak-derived BH3 peptide (34). This result suggests that the PGAM5 proteins bind outside of the groove in Bcl-X L that binds BH3 domains and are not likely to directly compete with BH3-only proteins for binding to Bcl-X L . Although it remains to be determined if binding of PGAM5 to Bcl-X L will modulate the balance between cell survival and death in response to changes in redox homeostasis, the PGAM5 proteins provide an intriguing molecular link between oxidative stress and apoptosis.
In summary, we have identified PGAM5 as a novel substrate for a redox-regulated Keap1-dependent ubiquitin ligase. The possibility that PGAM5 proteins may modulate oxidative stress-induced apoptosis is under active investigation by our laboratory.