Identification of Redox-sensitive Cysteines in GA-binding Protein-α That Regulate DNA Binding and Heterodimerization*

The transcription factor GA-binding protein (GABP) is composed of two subunits, GABPα and GABPβ. The DNA-binding subunit, GABPα, is a member of the Ets family of transcription factors, characterized by the conserved Ets-domain that mediates DNA binding and associates with GABPβ, which lacks a discernible DNA binding domain, through ankyrin repeats in the NH2 terminus of GABPβ. We previously demonstrated that GABP is subject to redox regulation in vitro andin vivo through four COOH-terminal cysteines in GABPα. To determine the roles of individual cysteines in GABP redox regulation, we generated a series of serine substitution mutants by site-directed mutagenesis and identified three redox-sensitive cysteine residues in GABPα (Cys388, Cys401, and Cys421). Sulfhydryl modification of Cys388 and Cys401 inhibits DNA binding by GABPα, whereas, modification of Cys421 has no effect on GABPα DNA binding but inhibits dimerization with GABPβ. The positions of Cys388 and Cys401 within the known Ets-domain structure suggest two very different mechanisms for redox regulation of DNA binding. Sulfhydryl modification of Cys388 could directly interfere with DNA binding or might alter the positioning of the DNA-binding helix 3. Modification of Cys401 may inhibit DNA binding through stabilization of an inhibitory helix similar to that described in the Ets-1 protein. Thus, GABP is regulated through at least two redox-sensitive activities, DNA binding and heterodimerization.

Transcription in eukaryotes depends on at least two groups of proteins, general transcription factors and activator proteins (1,2). While the former direct basal transcription, the later are responsible for cell-, tissue-, and gene-specific expression in response to various stimuli (3,4). Many activator proteins are the end point of complex signal transduction pathways, and their activities are tightly regulated by various post-transcriptional modifications, including phosphorylation, glycosylation, and reduction/oxidation (redox) 1 modification of cysteine residues (3,5,6).
The regulation of activator proteins by redox of reactive cysteine residues has been demonstrated for members of several important transcription factor families including, NFB and AP-1 (for review, see Ref. 6). Both proteins require reducing conditions for DNA binding in vitro; however, in vivo, these factors become activated by oxidative stress-promoting agents (Ref. 6, and references therein). To explain this apparent contradiction, it has been proposed that some effectors of prooxidant conditions may activate either the expression or activity of specific enzymes (thioredoxin, Ref. 1) which maintain NFB and AP-1 in the reduced state (7,8). Other pro-oxidant effectors, such as reactive oxygen species (ROS), activate protein kinases that 1) phosphorylate IB leading to its inactivation and concomitant activation of NFB (6), and 2) phosphorylate Jun and Fos proteins activating AP-1 activity (6,9,10).
Recently we demonstrated that murine GABP is redox-regulated both in vitro and in vivo and that pro-oxidant conditions, in contrast to NFB and AP-1, result in inhibition of GABP DNA binding through COOH-terminal cysteine residues in the GABP␣ subunit (11). GABP is composed of two subunits, GABP␣ and GABP␤, and each subunit provides distinct functions to the complex (12,13). GABP␣ belongs to the Ets-protein family and is characterized by an ϳ85-amino acid region near the COOH terminus (Ets-domain) which is necessary for DNA binding to the sequence (A/C)GGA(A/T)(A/G) (14,15,16). GABP␤ has no discernible DNA binding domain and is unable to bind DNA on its own; however, GABP␤ contains at least one transcription activation domain required for transcription activation through the heteromeric complex (17)(18)(19). In addition, GABP␤ contains two regions involved in protein-protein interactions, four ankyrin repeats at the NH 2 terminus and a leucine-zipper at the COOH terminus, which mediate GABP␣-␤ and GABP␤-␤ interactions, respectively (12,13,17).
The nature of signaling between mitochondria and the nucleus affecting the regulation of nuclear-encoded mitochondrial genes is not known, but likely candidates include reactive oxygen species (for review, see Ref. 27). Approximately 1-2% of the oxygen consumed during respiration is only partially reduced, forming reactive oxygen species including superoxide, hydrogen peroxide, and hydroxyl radicals (27,28). The inhibi-tion of the terminal steps of the respiratory chain, including cytochrome C oxidase and ATPase (complexes IV and V), by drugs or due to mutation, leads to increased production of superoxide and hydrogen peroxide (27,28). ROSs may directly oxidize redox-sensitive sulfhydryl groups in the COOH-terminal portion of GABP␣ and therefore inhibit GABP␣ DNA binding and GABP␣-dependent transcription, providing a mechanism linking the activity of the respiratory chain with the expression of its components.
To characterize the mechanisms of redox sensitivity of the GABP␣ protein, we performed site directed mutagenesis of cysteine residues in the GABP␣ DNA-binding and dimerization domains. We demonstrated that two cysteine residues in the Ets DNA binding domain, Cys 388 and Cys 401 , are sensitive to redox changes affecting GABP␣ DNA binding while Cys 421 in the GABP␣/GABP␤ dimerization domain confers redox sensitivity to GABP␣-␤ complex formation.

EXPERIMENTAL PROCEDURES
Cloning and Expression of Recombinant Proteins in Escherichia coli-The nucleotide sequence coding for GABP␣, GABP␤, and GABP␣ c (C-terminal region containing the Ets and the GABP␣/␤ dimerization domains) were amplified by PCR from the cDNAs kindly provided by C. C. Thompson, Carnegie Institute of Washington (17), and cloned into pET15b (Novagen) as described previously (11). The individual rGABP proteins were expressed in E. coli BL21 strain and purified (Ն90% purity estimated by Coomassie Brilliant Blue R-250 staining) by nickel chelating chromatography as described previously (11).
Site Directed Mutagenesis-Site-directed mutagenesis was performed using the Transformer site directed mutagenesis kit (CLONTECH) according to the manufacturer recommendations or as described by Kunkel,et al. (29). The following primers were synthesized by the University of Missouri DNA Core Facility on an Applied Biosystems DNA Synthesizer Model 380B and used to substitute the corresponding cysteines to serines (substituted serine codons are underlined): 5Ј-GCTCGAGACTCGATATCTTGGGTT for Cys 338 , 5Ј-GGACATGATTTCGAAA GTTCAAGG for Cys 388 , 5Ј-TACAAATTTGTT-TCTGACTTGAAGACT for Cys 401 , 5Ј-CTGGTCATAGAGTCTGAACAGAA-GAAA for Cys 421 . The sequences of all mutants were confirmed by restriction analysis as well as by DNA sequencing (Sequenase 2.0 kit, U. S. Biochemical Corp.).
Treatment of Recombinant GABP Proteins with Sulfhydryl Modifying Reagents-The sulfhydryl modifiers N-ethylmaleimide (NEM) (Sigma) or 5,5Ј-dithiobis(2-nitrobenzoic acid) (DTNB, Ellman's reagent) (Pierce) (31) were added to concentrated solutions of pre-reduced (1.0 mM DTT) rGABP proteins (ϳ0.3 mg/ml) and then diluted to a final DTT concentration of 20 M. The prereduced proteins were then treated with NEM (250 -500 M) or DTNB (1 mM) for 20 min on ice, and the DNAbinding properties of the treated proteins were analyzed by EMSA. To prevent possible nonspecific modification of ⑀-amino group of lysines or imidazole rings of histidine residue by NEM, all modification experiments were performed at low NEM concentrations, at low temperature, and at pH 7.1, conditions which are unfavorable to the modification of lysines and histidines (30).

Multiple Cysteines in the COOH Terminus of GABP␣ Are
Involved in Redox Regulation of GABP DNA Binding-Previously, we demonstrated that GABP␣ DNA binding and transcription activation functions are redox regulated both in vivo and in vitro and that COOH-terminal cysteine residues in GABP␣ are important for this regulation (11). We have shown that both full-length GABP␣ and a COOH-terminal truncation protein, GABP␣ c , containing only the DNA binding domain and the GABP␣␤ dimerization domain, require the presence of reducing agents (DTT) for DNA binding and are inhibited by treatment with oxidized glutathione (GSSG) or NEM (11), implicating the COOH-terminal cysteine residues in redox regulation of GABP DNA-binding. In contrast, GABP␤ was found to have no detectable role in redox regulation of GABP DNAbinding activity. GABP␣ contains nine cysteine residues, five in the NH 2 -terminal two thirds of the protein (Cys 22 , Cys 37 , Cys 61 , Cys 69 , and Cys 223 ), two in the Ets domain (Cys 338 and Cys 388 ), and two in the GABP␣␤ dimerization domain (Cys 401 and Cys 421 ) (Fig. 1). The latter four residues are present in GABP␣ c and are the putative targets for redox regulation of GABP.
To identify which of the four COOH-terminal cysteines are involved in GABP redox regulation, a set of GABP␣ c mutants, each with a single Cys 3 Ser substitution was constructed, and the sensitivity of each mutant protein to NEM treatment was determined. The mutant proteins designated ␣ c SCCC, ␣ c CSCC, ␣ c CCSC, and ␣ c CCCS (described in Fig. 1) were expressed in E. coli, purified as described under "Experimental Procedures," and their requirement for DTT and sensitivity to sulfhydryl modifiers was tested by EMSA assays. If only a single cysteine residue was necessary for redox regulation, the substitution of this residue to serine would be expected to render the mutant protein redox insensitive and resistant to sulfhydryl modification. However, all of the singly substituted GABP␣ c mutants bound DNA as heterotetramers in the presence of rGABP␤ with comparable activity (Fig. 2) and all required reduction with DTT for DNA binding. 2 DNA-binding activity of the mutant proteins in the absence of rGABP␤ was nearly undetectable, consistent with our previous observations in which wild-type GABP␣ c also bound DNA poorly in the absence of rGABP␤ (11). Treatment with NEM inhibited DNA binding of all four singly substituted mutant proteins although ␣ c CCSC consistently retained some residual DNA-binding activity following NEM treatment (Fig. 2, lanes 7 and 8). These results suggest that multiple cysteines are redox sensitive and participate in regulation of GABP␣ DNA binding.
Cys 388 , Cys 401 , and Cys 421 Are Targets for Redox Regulation of rGABP␣ c DNA Binding in Vitro-Since single Cys 3 Ser substitutions of COOH-terminal cysteines in GABP␣ c failed to confer complete resistance to sulfhydryl modifiers, we generated a set of GABP␣ c triple Cys 3 Ser substitution mutants, resulting in mutant GABP␣ c proteins containing only one of the four cysteine residues designated ␣ c CSSS, ␣ c SCSS, ␣ c SSCS, and ␣ c SSSC (Fig. 1). In addition, a mutant lacking all four cysteines, GABP␣ c Q (␣ c SSSS), was also generated. GABP␣ c and the mutant proteins were treated with NEM or DTNB, and the DNA-binding activity of the modified proteins was measured by EMSA assay in the presence of rGABP␤ (Fig.  3). As expected, the ␣ c CCCC protein was completely inhibited by NEM treatment (Fig. 3, lanes 1 and 2), whereas the ␣ c SSSS (Q) mutant, lacking any cysteine residues was unaffected by NEM treatment (Fig. 3, lanes 3 and 4). Treatment of ␣ c SCSS (Fig. 3, lanes 7 and 8) and ␣ c SSCS (Fig. 3, lanes 9 and 10) with NEM, or DTNB, 2 nearly completely abolished DNA binding, demonstrating the importance of Cys 388 and Cys 401 in redox sensitivity of GABP␣ c DNA binding. Treatment of ␣ c CSSS (Fig.  3, lanes 5 and 6) with NEM had no significant effect on DNA binding, suggesting that Cys 338 is not likely to be important in GABP␣ redox sensitivity. Surprisingly, although Cys 421 lies outside the known DNA binding domain, NEM modification partially inhibited ␣ c SSSC DNA binding (Fig. 3, lanes 11 and  12). Since DNA binding of the truncated rGABP␣ c proteins can be detected by EMSA only when complexed with rGABP␤, interference with the GABP␣ c -GABP␤ interaction would affect the apparent DNA binding of the GABP␣ c mutant proteins. Thus, it is possible that modification of Cys 421 prevents ␣ c SSSC/GABP␤ complex formation rather than affecting the intrinsic DNA-binding activity.
Alkylation of Cys 388 and Cys 401 Inhibits the Intrinsic DNAbinding Activity of Full-length GABP␣-Because full-length rGABP␣ is able to bind DNA in the absence of GABP␤, we generated a set of triply substituted proteins in the context of the full-length GABP␣ protein, as well as a mutant (GABP␣Q) lacking all four COOH-terminal cysteines (Fig. 1). All fulllength triple mutants, designated ␣CSSS, ␣SCSS, ␣SSCS, and ␣SSSC, and GABP␣Q (SSSS) were able to bind DNA in the absence of GABP␤, although the DNA-binding activity of most of the mutant proteins was reduced (10 -40%) relative to the wild-type protein. 2 Since the full-length GABP␣ protein contains five additional cysteines, the wild-type and GABP␣Q proteins were treated with DTNB to confirm that redox sensitivity of the full-length GABP␣ protein was exclusively mediated through the four COOH-terminal cysteines (Fig. 4). DTNB is widely used as a highly specific, reversible, sulfhydryl modifier, which can be removed from conjugated cysteines by reduction with DTT (30,31). DTNB treatment completely inhibited GABP␣ DNA binding (Fig. 4, lane 2), whereas GABP␣Q was totally unaffected (Fig. 4, lane 5). The inhibitory effect of DTNB on GABP␣ was completely reversed by the subsequent addition excess of DTT (Fig. 4, lane 3) demonstrating that DTNB inhibition of GABP␣ DNA binding was sulfhydryl-specific. These experiments demonstrate that the COOH-terminal cysteine residues (Cys 338 , Cys 388 , Cys 401 , and Cys 421 ) are necessary and sufficient for redox regulation of GABP DNA binding in vitro and confirm that the remaining NH 2 -terminal cysteine residues play no detectable role in such regulation.
To determine the effect of sulfhydryl modification of individual cysteines on the intrinsic DNA-binding activity of GABP␣, the wild-type and mutant proteins were treated with NEM and analyzed by EMSA assays in the presence and absence of GABP␤. As expected, NEM treatment of wild-type GABP␣ nearly completely inhibited its DNA-binding activity in the presence and absence of GABP␤ (Fig. 5A, lanes 1-4). In contrast, NEM treatment of GABP␣Q had no effect on DNA binding (Fig. 5A, lanes 5-8), and the presence of GABP␤ had no effect on DNA binding of NEM-treated GABP␣ or GABP␣Q proteins (Fig. 5A, lanes 4 and 8). The consistency of the results of NEM and DTNB modification and the resistance of GABP␣Q to NEM treatment (Fig. 5, lanes 5-8), even at high concentrations (5 mM) 2 , demonstrate that inhibition of GABP␣ DNA binding by these reagents is not the result of nonspecific reactions of NEM (or DTNB) with other amino acids such as lysine or histidine.
Modification of the ␣CSSS protein with NEM had little or no effect on DNA-binding activity (Fig. 5B, lanes 1-4), which is consistent with the results obtained with the ␣ c CSSS protein (Fig. 3), and suggests that Cys 338 plays no detectable role in redox sensitivity of GABP␣ DNA binding. The DNA-binding activities of the ␣SCSS (Fig. 5B, lanes 5-8) and ␣SSCS proteins (Fig. 5C, lanes 1-4), were inhibited by treatment with NEM, also consistent with our results obtained with the truncated GABP␣ c mutant proteins. Similar results were obtained with DTNB treatment. 2 Therefore, modification of either Cys 388 or Cys 401 is sufficient to inhibit DNA binding, either directly or by affecting the conformation of the GABP␣ DNA binding domain.
The DNA-binding activity of the ␣SSSC monomer complex was not affected by NEM treatment (Fig. 5C, lanes 5-8), even at high concentrations (up to 5 mM), 2 demonstrating that Cys 421 plays no role in redox sensitivity of the intrinsic DNAbinding activity of GABP␣. However, in the presence of GABP␤, NEM-treated ␣SSSC protein failed to form the GABP␣/GABP␤ dimer complex (Fig. 5C, lane 8), even when GABP␤ was added in excess. 2 Cys 421 lies within the region of GABP␣ previously implicated in GABP␣-␤ dimerization (17). Modulation of the ability of the ␣SSSC protein to form GABP␣-␤ complexes by sulfhydryl modifiers is consistent with the notion that Cys 421 lies within the GABP␣/GABP␤ dimer-ization interface. Thus, these observations demonstrate that redox regulation of GABP may occur on at least two levels, 1) the intrinsic GABP␣ DNA-binding activity and 2) GABP␣-␤ dimerization.

DISCUSSION
In this report, we have investigated the roles of the four COOH-terminal cysteine residues of GABP␣ in redox sensitivity of GABP DNA binding in vitro. Using Cys 3 Ser substitution mutants in the context of the full-length protein and a truncated protein (GABP␣ c ) containing only the DNA binding and dimerization domains, we demonstrate that redox regulation of GABP occurs at two levels, the activity of the GABP␣ DNA binding domain and heterodimerization of GABP␣ with the GABP␤ subunit.
GABP␣ Cys 338 Is Insensitive to Redox Modification-Sulfhydryl modifiers have little or no effect on the DNA-binding activity of ␣CSSS, suggesting that Cys 338 is insensitive to redox changes and is not involved in redox regulation of GABP DNA binding. These findings are in contrast with published reports implicating Cys 394 (Ala 335 in GABP␣) in redox sensitivity of the v-Ets DNA binding domain (32). While the three-dimensional structure of the GABP␣ Ets-domain has not been reported, the high level of homology in primary structure and high similarity of the published three-dimensional structures of PU.1, Ets-1, and Fli-1 proteins allow us to use these structures to estimate the molecular environment of residues of interest in GABP␣. Analysis of the tertiary structures of several Ets-proteins (Ets-1, Fli-1, and PU.1) (33)(34)(35)(36)(37)(38) revealed that the local environment of Cys 350 in Ets-1 (Cys 394 in vEts) (Fig. 6A) is likely to be very different from that of Cys 338 in GABP␣ and may explain the observed difference in their redox sensitivity. In Fli-1 (33,34) (lanes 1 and 2), rGABP␣ c Q (lanes 3 and 4), rGABP␣ c SCCC ( lanes 5 and 6), rGABP␣ c CSCC (lanes 7 and 8), rGABP␣ c CCSC (lanes 9 and 10), and rGABP␣ c CCCS (lanes 11 and 12) proteins (0.01 g each) were treated with 250 M NEM as described under "Experimental Procedures." A slight excess of rGABP␤ 1 protein (0.015 g) was then added, and DNA binding of the heteromeric complexes containing treated and untreated rGABP␣ c proteins was measured by EMSA analysis using the dPEA3-0 probe containing two adjacent PEA3/EBS sites. NEM treatment is indicated by ϩ. The (␣ c ) 2 ␤ 2 tetramer complex is the predominant complex formed with all rGABP␣ c proteins bound to the dPEA3-0 probe.

FIG. 3. The effect of NEM treatment on GABP␣ c Cys 3 Ser mutants containing only one of the four COOH-terminal cysteine residues.
Pre-reduced rGABP␣ c WT (lanes 1 and 2), rGABP␣ c Q (lanes 3 and 4), rGABP␣ c CSSS ( lanes 5 and 6), rGABP␣ c SCSS (lanes 7 and 8), rGABP␣ c SSCS (lanes 9 and 10), and rGABP␣ c SSSC (lanes 11 and 12) proteins (0.01 g each) were treated with 250 M NEM as described under "Experimental Procedures." A slight excess of rGABP␤ 1 protein (0.015 g) was then added, and DNA binding of heteromeric complexes containing treated and untreated rGABP␣ c proteins was measured by EMSA analysis using the dPEA3-0 probe. Treatment with NEM is indicated by ϩ. The position of the predominant (␣ c ) 2 (␤) 2 complexes are indicated by an arrow on the left. the position of Cys 394 in vEts (Cys 350 in Ets-1) is within the loop between ␣-helix 1 and ␤-strand 1, is not involved in formation of the central hydrophobic core, and is likely to be solventexposed (Fig. 6A) (33)(34)(35)(36)(37)(38). Most Ets proteins contain either a hydrophobic or an aromatic residue in the position that is equivalent to Cys 338 in GABP␣, suggesting the involvement of residues in this position in formation of the central hydrophobic core. Thus, it is likely that the sulfhydryl of Cys 338 in GABP␣ is at least partially buried within this hydrophobic core and is inaccessible to modification.
Modification of GABP␣ Cys 388 Directly Interferes with DNA Binding-GABP␣ Cys 388 is located in the region corresponding to ␤-strand 3 of the Ets-1 (35,36), Fli-1 (33,34), and PU.1 Ets domains (37,38) (Fig. 6B). This residue is not highly conserved in ETS proteins; although, neither bulky aromatic nor negatively charged residues have been found in this position in any of the known ETS proteins (16,38,39). Substitution in v-Ets of the histidine equivalent to His 403 in human Ets-1 (Cys 388 in GABP␣) to aspartic acid results in a virus whose transforming activity is temperature-sensitive (40). Similarly, substitution of Arg 74 in Elk-1 (Cys 388 in GABP␣) to aspartic acid leads to the inactivation of DNA binding (41). Furthermore, substitution of Lys 404 in Ets-1 or Lys 245 in PU.1, which correspond to Lys 389 adjacent to Cys 388 in GABP␣, completely abolished DNA binding of these proteins (40,42). Lysine in the position corresponding to Lys 389 in GABP␣ (Lys 245 in PU.1) is absolutely conserved in all known Ets-proteins and is directly involved in contacts with the phosphate backbone 5Ј to the core GGAA sequence in the PU.1-DNA complex crystal structure (Fig. 6B) (37,38). Similarly, in Fli-1, the equivalent residue was determined by NMR to be within 4 Å of the bound DNA by intermolecular nuclear Overhauser effects (33). Therefore, since Cys 388 is located adjacent to Lys 389 , it is likely to be in close proximity to bound DNA, suggesting that sulfhydryl oxidation to sulfenic (RSOH) or sulfinic (RSO 2 H) acids at this site could directly interfere with DNA binding through electrostatic repulsion as has been proposed for redox regulation of Fos and Jun proteins (Ref. 6, and references therein). This is consistent with the fact that, as described above, substitution to aspartic acid of residues in v-Ets and Elk-1 proteins analogous to Cys 388 inhibits transformation or DNA binding, respectively. Alternatively, modification by glutathione disulfide (GSSG), which we previously demonstrated to inhibit GABP DNA binding (11)  nal ␣-helix 4 (36). Based on the published NMR structure (35,36), in Ets-1, Cys 416 (Cys 401 in GABP␣) is solvent-exposed and accessible for modification. The precise role of this region in DNA binding is not known. However, substitution of either of the adjacent residues in Ets-1 (Val 415 or Asp 417 ) significantly decreases DNA binding, and substitution of the conserved residue, Phe 414 to leucine, in ␤-strand 4 abolishes DNA binding (42). Residues analogous to GABP␣ Cys401 in Ets-1, PU.1, and Fli-1 do not form any direct contacts with DNA in the threedimensional structures of these proteins. Furthermore, a mutant GABP␣ protein truncated at residue 400 is competent for DNA binding, suggesting that residues beyond Val 400 , including Cys 401 , are not directly involved in GABP␣ DNA binding. 2 Thus, it is likely that mutations or modification of residues in this region results in some structural alteration rather than directly interfering with DNA binding.
The COOH-terminal 24 residues of the Ets-1 protein, including ␣-helix 4, have been implicated in the regulation of DNA binding through interaction with an NH 2 -terminal inhibitory domain (44). In the Ets-1 NMR structure, ␣-helix 4 lies antiparallel to and forms several contacts with ␣-helix 1 from the Ets-domain and forms additional contacts with the NH 2 -terminal inhibitory domain. The Ets-1 ␣-helix 4 is well defined locally although it is not positioned precisely within the tertiary fold (36), suggesting significant conformational mobility of this putative inhibitory helix. Therefore, mutations or modifications in the loop between ␤-strand 4 and ␣-helix 4, or in ␣-helix 4 itself, that increase the stability of interactions with the Ets-domain may result in inhibition of DNA binding. This putative inhibitory domain is highly conserved in GABP␣ (59% identity, 88% similarity) (Fig. 6C), and Cys 401 is located within the loop region between ␤-strand 4 and ␣-helix 4 (45,46,47). Our data suggest that redox modification of Cys 401 inhibits GABP␣ DNA binding, possibly by facilitating the interaction between the inhibitory ␣-helix 4 and the DNA binding domain. The Ets-2, PEA3, and ERM proteins also contain cysteines in the position analogous to Cys 401 in GABP␣, suggesting that these Ets proteins may also be redox-sensitive.
Modification of GABP␣ Cys 421 Inhibits GABP␣-␤ Dimerization-Deletion of the 54 COOH-terminal residues of GABP␣ abolishes GABP␣-GABP␤ dimerization while not affecting DNA binding (17,48). 2 Consistent with this result, the modification of Cys 421 in the COOH-terminal portion of GABP␣ does not directly affect the DNA binding of monomeric GABP␣; however, modification of Cys 421 inhibits GABP␣-GABP␤ dimerization. This result strongly suggests that Cys 421 lies within the GABP␣-GABP␤ dimerization interface and that modification of this residue prevents interaction with GABP␤. Recently the redox regulation of intersubunit interactions in the CAAT-binding protein, NF-Y, has been reported (49). Under non-reducing conditions, the NF-YB subunit fails to associate with NF-YC, which causes a substantial decrease in the DNA-binding activity of NF-Y. In contrast, disruption of GABP␣-GABP␤ complexes does not significantly diminish DNA binding of the GABP␣ subunit. Since GABP␣ lacks any discernible transactivation domain (50), transcription activation by GABP depends on the transactivation domain provided by GABP␤ (50). Thus, GABP␣ bound to DNA in the absence of GABP␤ may function as a low-affinity repressor.
GABP has now been demonstrated to be regulated through at least three pathways: by phosphorylation (51), by redox modification of DNA binding (Ref. 11, and this study), and by redox modification of inter-subunit interactions (this study). The importance of GABP(NRF2) for the regulation of genes involved in energy production (cytochrome oxidase subunits IV, Vb, and VII, and ATPase ␤ subunit) and control of mitochondrial transcription and replication (mitochondrial transcription factor 1), together with the wide distribution of this factor, suggests that GABP may function as a transcriptional sensor of the redox and energy states of mammalian cells. Stable expression of a non-redox-sensitive mutant, such as GABP␣Q, in a mammalian cell line will be useful to further characterize redox regulation of GABP and establish its importance in cellular homeostasis.