Identification of a Conserved Oxidation-sensitive Cysteine Residue in the NFI Family of DNA-binding Proteins*

The nuclear factor I (NFI) family of site-specific DNAbinding proteins plays a role in both transcription and adenovirus DNA replication. The DNA binding domain of NFI family members contains 4 cysteine residues (Cys-2, Cys-3, Cys-4, and Cys-5) that are conserved in all NFI proteins. Mutation of the Cys-2, Cys-4, and Cys-5 residues in the human NFI-C protein to several other amino acids abolished DNA binding, while 8 of 10 mutations of the Cys-3 residue had little or no effect on binding. Wild-type NFI-C was inactivated by N-ethylmaleimide in vitro, while the active Cys-3 mutant proteins were resistant to N-ethylmaleimide. Treatment of wild-type NFI in uitro with the oxidizing agent diamide also inactivated DNA binding, and subsequent reduction with dithiothreitol restored binding activity. The active Cys-3 mutant NFI proteins were resistant to diamide-inactivation, indicating that the Cys-3 residue is required for modulation of DNA-binding by oxidation state. These studies indicate that oxidative-inactivation can play an important role in the modifying NFI-DNA-protein interactions. The presence of this nonessential Cys-3 residue in all known NFI proteins raises the possibility that it may function in a manner similar to redox-sensitive cysteine residues found in other site-specific DNA-binding proteins.

The nuclear factor I (NFI)' family of site-specific DNA-binding proteins function in both adenoviral DNA replication and the transcription of a variety of cellular genes. NFI was originally described as a single protein species required for the replication of adenovirus DNA both in vivo and in vitro (1-31, but is now recognized to be a family of proteins expressed from a set of at least four related cellular genes, NFI-A, NFI-B, NFI-C, and NFI-X (4-7). NFI family members bind as dimers (8)(9)(10) to the palindromic sequence TTGGC(N),GCCAA (3,8,(11)(12)(13) in duplex DNA. A highly conserved 220-amino acid NH,terminal domain is present in all NFI family members and is sufficient for dimerization, DNA binding, and the initiation of adenovirus replication i n vitro (4,9,10,14). This conserved DNA binding domain contains a putative lysine-rich helical region (14) and four phylogenetically conserved cysteine residues (151, but is not homologous to any of the well characterized classes of DNA binding domains including the zinc finger, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. * Foundation, Dept. of Cancer Biology NN-1,9500 Euclid Ave., Cleveland, ll To whom correspondence should be addressed: Cleveland Clinic 'The abbreviations used are: NFI, nuclear factor I; NEM, Nethylmaleimide; DTNB, 5,5'-dithiobis-(2-nitrobenzoic acid); Dm, dithiothreitol; GMS, gel mobility shift; PAGE, polyacrylamide gel electrophoresis. OH   leucine zipper, helix-turn-helix, helix-loop-helix, or ring finger domains (16)(17)(18). Although deletional analysis has defined the minimal size of the NH,-terminal domain required for both DNA binding and activation of adenoviral DNA replication (9,10,14,15), little is known about potential regulation of the DNA binding activity of NFI in vivo or in vitro.
Previous studies demonstrated two forms of covalent modification of NFI, N-glycosylation and phosphorylation, that could potentially affect NFI function in vivo. Two of the NFI species purified from human HeLa cells were shown to contain 0-linked carbohydrate with terminal sialic acid residues; however, the DNA binding activity of these modified forms of NFI appeared identical to the non-glycosylated forms of NFI isolated from HeLa cells (19,20). NFI purified from human HeLa cells is a substrate for the dsDNA-dependent protein kinase in vitro (21), although again no differences in the activities of phosphorylated or non-phosphorylated NFI have been reported.
One recently described mechanism for the modulation of the DNA binding activity of proteins is regulation of the intracellular reduction-oxidation (redox) state. A number of site-specific DNA-binding proteins (which possess different DNA-binding structural domains) contain cysteine residues whose oxidation state affects binding ability. Such redox-sensitive cysteine residues have been found to affect DNA binding by the c-Fos (22, 23), c-Jun (22), NFKB/Rel (24, 25), Myb (26), USF (28), and Egr-1 (29) proteins in vitro. In addition, there is increasing evidence that changes in intracellular redox state can affect numerous biological processes, including gene expression (22,(30)(31)(32). Our initial studies indicated that the DNA-binding domain of NFI contains 4 conserved cysteine residues important for DNA binding activity (15,33). The initial rationale for determining of the potential function of these cysteine residues was the observation that the DNA binding activity of NFI was inactivated by incubation with the sulfhydryl modifying agents N-ethylmaleimide (NEM) and 5,5'-dithiobis-(2-nitrobenzoic acid) (DTNB) (15). In the present study, we examine the role of the conserved cysteine residues in NFI-DNA interactions by site-directed mutagenesis, chemical modification, and treatment with the oxidizing agent, diamide. Our findings indicate that one of the four conserved cysteine residues of NFI, although not essential for DNA binding, is sensitive to oxidation in vitro and represents a possible target for the regulation of DNA binding activity in vivo by intracellular redox state.

MATERIALS AND METHODS
Plasmid DNA and Mutagenesis-The pET220 vector encoding the NH,-terminal 220 amino acids of the human NFI-C protein (hNFI-C220) expressed from a T7 promoter was described previously (15). The cloned human NFI-C cDNA was a gift of Dr. R. Tjian. This 220-residue DNA binding domain of hNFI-C/CTFl protein was shown previously to be sufficient for DNA binding activity and stimulation of adenovirus replication but lacks the COOH-terminal proline-rich domain impor-

29949
This is an Open Access article under the CC BY license. tant for activation of transcription in vivo (10). A hexahistidine tag was fused to the N terminus of the hNFI-C220 protein by insertion of a duplex oligonucleotide into the Ne01 site of pET220 present at the initiation methionine of hNFI-C220. The resultant plasmid (pET220H6) expressed a 231-amino acid fusion protein consisting of an NH,-terminal 220 amino acids of hNFI-C/CTFl (4), and four COOH-initiation methionine and six histidines at the amino terminus, the terminal amino acids derived from vector sequences (15). All of the recombinant NFI proteins examined here have this general structure and are referred to in the text as wild-type or mutant NFI-C or hNFI-C220 proteins. In vitro mutagenesis was performed on duplex pET220H6 by the unique site selection (USE) technique (34). An oligonucleotide that converts a unique non-essential BglII site in pET220H6 into a unique PvuI site was used as the selection primer in all mutations. To make mutations at sites encoding each of the cysteine residues of NFI (Cys-1 to Cys-5) (15) (see Fig. l), mutagenic primers were synthesized containing equal mixtures of all 4 bases in the first two positions and an equal mixture of G and C in the third position of each cysteine codon (35). The resulting primer mixture included codons for all the 20 naturally occurring amino acids at the positions encoding cysteines in wild-type NFI. In some cases, equal mixtures of A, G, and C in the first position and of all 4 bases in the second and the third codon positions were used to prevent recovery of wild-type, Phe, Trp, or Tyr mutant alleles. Details of oligonucleotide construction are available upon request. Each mutant NFI protein is named for an amino acid substitution present at a designated cysteine residue with the nomenclature CNX, where CN indicates the cysteine residue modified (Cys-1-Cys-5 positions (see Fig. 1)) and X is the amino acid substitution at that position (i.e. C1S denotes a serine substitution at position Cys-1). A plasmid expressing a mutant NFI protein containing a serine at the Cys-1 position and a valineresidue at the Cys-3 position (pET220H6ClS: C3V) was made by replacement of the single BpullOZI fragment of a PET vector expressing the C1S mutant of NFI with the homologous fragment from a vector expressing the C3V mutant protein. All mutations were confirmed by restriction enzyme mapping and by dideoxy chain termination sequencing (36).
Preparation of Bacterial Extract and Protein Purification-Bacterial extracts containing wild-type and mutant NFI proteins were prepared from Escherichia coli JMlOg(DE3) cells grown in LB medium containing ampicillin (100 pg/ml). Cultures were incubated at 37 "C to A , = 0.4, cooled on ice, induced with 0.2 m~ isopropyl thiogalactoside for 16 h at 17 "C and harvested by centrifugation at 5000 x g for 5 min at 4 "C. Pellets were resuspended in cold buffer L (25 m M Hepes, pH 7.5, 0.35 M NaCl, 10% sucrose, 5 m M Dm, and 1 m M phenylmethylsulfonyl fluoride), and lysozyme was added to 200 pg/ml, followed by incubation on ice for 15 min. Cells were lysed by addition of Nonidet P-40 to 0.1% and incubation on ice for 15 min. Lysates were subjected to centrifugation at 150,000 x g for 1 h at 4 "C, and the resulting supernatants were stored at -80 "C. For the assays shown in Fig. 4, A and B, wild-type and mutant NFI proteins were purified as previously described (1) with the modification that the protein was eluted from phosphocellulose resin with 1.0-1.5 M NaCl. The final purification step was absorption to Ni-nitriloacetic acid-agarose (37) and elution with buffer L containing 150 m M imidazole followed by dialysis against buffer L. Protein concentration was determined by the method of Bradford (38). Protein preparations were analyzed by electrophoresis on 10% SDS-polyacrylamide gels and were determined to be 50-70% pure by silver staining (Bio-Rad) and quantified using bovine serum albumin as a standard.
Gel Mobility Shift (GMS) and WCross-1inkingAssays"Gel mobility shift assays for detection of NFI DNA binding activity were performed with purified recombinant proteins or bacterial extracts as described (6, 15) in the figure legends using the 32P-labeled FIB-2.6 oligonucleotide or a control mutant oligonucleotide FIB-2.6C2. The single point mutation present in FIB-2.6C2 abolishes NFI binding in vitro (13,39). The sulfhydryl-modifying agent NEM and the oxidizing agent diamide were directly added to protein extracts or binding reactions as indicated in the figure legends. All GMS analyses were performed in the linear range of the assay as determined by serial dilution of extracts containing wild-type and mutant NFI proteins. In bacterial extracts, the GMS assay is linear over an -20-fold range and the inactive mutant NFI proteins have a >200-fold reduction in activity when compared to wildtype protein as assessed by either over-exposure of autoradiograms or quantitation of dryed gels using a Molecular Dynamics model 400 Phos-phorImager. UV cross-linking analysis was performed as described earlier (6) with bacterial extracts (20 pg) from control cells or cells expresswith labeled DNA, irradiated for 10 min at 4 "C (-3000 pW/cm2), ing wild-type or mutant NFI proteins. Protein samples were incubated analyzed by SDS-gel electrophoresis, and the labeled covalent NFI-DNA complexes were detected by autoradiography as described elsewhere (6).
SDS-PAGE and Immunodetection-SDS-polyacrylamide gel electrophoresis was performed under reducing conditions on 10-12.5% gels as described previously (40). For immunodetection, proteins were transwashed for 10 min with Tris-buffered saline (20 m M Tris-C1, pH 7.6, 137 ferred to Immobilon-P membrane (Millipore), and the membrane was m M NaCl) containing 0.5% Tween 20 (TBST) and blocked overnight with 5% non-fat dry milk in TBST. The blocked membrane was probed with no-terminal hNFI-C220 peptide acetyl-DEFHPFIEALLC. The mem-an affinity-purified polyclonal rabbit antiserum raised against the amibrane was incubated with primary antibody for 2 h at room temperature and washed three times with TBST, and NFI protein was detected by chemiluminescence using horseradish peroxidase-conjugated goat antirabbit IgG (ECL, Amersham Corp.). Three-fold serial dilutions of extracts containing wild-type NFI protein were used to measure the sensitivity and range of the chemiluminescence assay.

RESULTS
Our previous studies indicated that the DNA binding activity of NFI was inactivated by the sulfhydryl alkylating agent NEM and that Ser substitutions at the Cys-2, Cys-3, Cys-4, and Cys-5 positions of hNFI-C220 inactivated DNA binding while Ser substitution at Cys-1 had no effect on binding (15, 33). However, recently we observed that cells expressing the C3S mutant NFI protein had a somewhat slower growth rate in liquid cultures and a smaller colony size at 37 "C when compared to those expressing the wild-type protein. In addition, we determined that 5-10-fold more active NFI protein was present in soluble extracts of E. coli when isopropyl-1-thio-P-o-galactopyranoside induction was performed at 17 "C rather than the 37 "C temperature used earlier These results suggested that low temperature induction of NFI protein might reduce any potential toxicity of the wild-type and mutant NFI proteins in E. coli. In the current study we have used this low temperature induction system to examine the effects of multiple mutations at the 4 conserved Cys residues present in the hNFI-C220 protein. Our initial aim was to determine whether any other amino acids could substitute for the conserved Cys residues of hNFI-C220 with retention of DNA binding activity. Several amino acid substitutions were made a t each conserved Cys position by site-directed mutagenesis, and the DNA binding activity of NFI in extracts containing wild-type and mutant NFI proteins was measured using a gel mobility shift assay (6,15). As summarized in Fig. 1, extracts containing recombinant hNFI-C220 with substitutions at the Cys-2 (Ser, Trp, Phe, Gly, Leu, Val, Ile, His), Cys-4 (Ser, Trp, Arg, Phe, Gly, Leu, Val, His, Ala, Asp, Tyr, Met), and Cys-5 (Ser, Trp, Gly, Leu, His, Ala, Tyr, Thr, Ile, Lys, Pro) positions were devoid of DNAbinding activity (binding <-0.1% of wild-type NFI-containing extracts). However, DNA binding activity was detected in 8 of 11 extracts expressing NFI proteins with substitutions at the Cys-3 position (Figs. 1 and 2, lanes 5-12, active Ser, Gly, Leu, Val, Thr, Ile, Ala, and His substitutions, respectively; lanes 13 and 14, inactive Arg and Trp mutants, respectively). As shown previously, E. coli extracts expressing wild-type or mutant NFI proteins generated a single specific protein-DNA complex with the FIB-2.6 nucleotide ( Fig. 2 A , lane 4 , denoted by an arrow at the left) and showed no binding of the mutant oligonucleotide FIB-2.6C2 ( Fig. 2 A , lane 2 ) . The level of DNA binding activity was similar in extracts containing either wild-type NFI (Fig. M,  lane 4 ) or mutant NFI proteins with serine, leucine, valine, threonine, isoleucine, alanine, or histidine substitutions at the Cys-3 position (Fig. 2 A , lanes 5,7,8,9,10, and 11, respectively). However, substitution of the Cys-3 position with glycine yielded extracts with reduced DNA binding activity ( Fig. 2
To ensure that similar quantities of wild-type or mutants proteins were used to assess DNA binding activity, immunodetection of NFI-C protein was performed after SDS-PAGE analysis of bacterial extracts. Equivalent quantities of NFI-C were present in extracts containing either wild-type ( Fig. 2 B , lane 2, NFI band denoted by the arrow at the right) or mutant NFI-C proteins ( Fig. 2 B , lanes [3][4][5][6][7][8][9][10][11][12]; including extracts containing the inactive arginine (Fig. 2B, lane 11) and tryptophan (Fig. 2B, lane 12) mutants. Similar levels of immunoreactive NFI protein were also present in extracts containing the inactive Cys-2, Cys-4, and Cys-5 mutant proteins (not shown). Titration experiments indicated that less than a %fold difference in NFI protein levels could be readily detected by this immunoassay (not shown). Control extracts from cells containing the parent vector lacking an NFI coding region contained only low levels of a number of nonspecific stained bands (Fig. 2B, lane 1).
We had previously shown that the DNA binding activity of NFI was inactivated by incubation with the sulfhydryl modifying agents NEM and DTNB (15). The finding that the conserved Cys-3 residue of NFI-C could be substituted with a variety of different amino acids with complete retention of DNA binding activity, led us to test the NEM sensitivity of the active Cys-3 mutants. While the wild-type NFI protein was completely inactivated by NEM (Fig. 3, lane 2 versus 11, all the active Cys-3 mutants were resistant to NEM-inactivation (Fig.  3, lanes 3-18). The active Cys-3 mutants were also resistant to inactivation by the sulfhydryl modifying agent DTNB (not shown). These data indicate that the cysteine residue present at the Cys-3 position is the only readily accessible alkylationsensitive sulfhydryl residue involved in NFI-DNA interaction.
Since sensitivity to NEM is a common feature of a number of proteins whose DNA binding activity is affected by oxidation (22,24,27,28), we examined the sensitivity of the DNA binding activity of wild-type and mutant NFI proteins to oxidation by the chemical oxidizing agent diamide (22,41). Treatment of wild-type NFI with 1 mM diamide abolished DNA binding activity ( Fig. 4.4 . . substitution mutants of NFI-C220. Bacterial extracts (20 pg) were analyzed by electrophoresis on a 12.58 SDS-polyacrylamide gel, transferred to a nitrocellulose membrane, and probed with rabbit polyclonal antipeptide antisera; hNFI-C220 was detected by chemiluminescence. Lane 1, control extract containing no hNFI-C220; lane 2, extract containing wild-type hNFLC220; lanes 3-12, extracts containing Cys-3 mutant hNFI-C220 proteins with the specific substitutions indicated of hNFI-(2220. Molecular weight mass are shown on the left (in above. The arrow marked NFI at the right indicates the position kilodaltons).
by the arrow marked A at the left). Although inactivation of NFI by NEM is irreversible (data not shown), the inactivation by diamide was reversed by subsequent addition of a 10-fold molar excess of DTT (Fig. 4.4, lane 3). Inactivation by diamide and reactivation by excess DTT was also seen with a mutant NFI protein containing a serine residue at the non-essential Cys-1 position (Fig. 4.4, labeled CIS, lanes 4-6). However, diamide did not inactivate mutant NFI proteins containing a valine residue at the Cys-3 position (Fig. 4.4, labeled C3V, lanes [7][8][9] or the double mutant with a serine residue at the Cys-1 position and a valine residue at the Cys-3 position (Fig. 4.4, labeled ClS:C3V, lanes [10][11][12]. Although the C3V and ClS,C3V proteins were resistant to inactivation by diamide, the mobility of the NFI-DNA complexes formed with these proteins was increased after treatment with diamide ( Fig. 4 A ,  lanes 8 and 11, denoted by the lower arrow marked B at left). This faster mobility suggests either an increased negative charge or a decreased size of the NFI.DNA complex after treatment with diamide (see "Discussion"). When wild-type NFI was incubated with DNA prior to diamide treatment, the DNA binding activity was partially protected from diamide inactivation (Fig. 4B, lane 5 versus lane 2). Again, the mobility of the DNAprotected, diamide-treated NFI.DNA complex was faster than the mobility of the untreated NFI.DNA complex and was similar in mobility to the NFI.DNA complex detected after diamide treatment of the C3V mutant protein (Fig. 4B, lanes 4-6 versus  1 and 3).
We had previously shown that NFI can be covalently crosslinked to DNA by UV irradiation (6). Since a specific Cys residue of the bacteriophage fd gene 5 DNA-binding protein is known to be the site of cross-linking the gene 5 protein to DNA (42), we determined whether mutation of the Cys-3 position of NFI affected the cross-linking of NFI to DNA. A covalent crosslinked NFI.DNA complex was detected after UV irradiation of reactions containing wild-type hNFI-C220 and the FIB-2.6 oligonucleotide (Fig. 5, lane 8, denoted by the arrow marked NFI DNA at the right). The apparent size of the NFI.DNA complex detected by SDS-PAGE (-45 kDa) is consistent with the combined sizes of the hNFI-C220 protein (-30 kDa, Fig. 2B, lanes [2][3][4][5][6][7][8][9][10][11][12] and the 26-bp oligonucleotide . No complex was detected in the absence of protein extract (Fig. 5, lanes 1-41, in the absence of UV irradiation (Fig. 5, lane 7) or with the mutant FIB-C2 oligonucleotide (Fig. 5, lanes 5 and 6 ) . Similar quantities of cross-linked NFI.DNA complex were seen with both wild-type NFI and mutant NFI proteins where the Cys-3 position was replaced by an alanine, leucine, or valine residue (Fig. 5, lanes 8, 12, 14, and 16, respectively). As seen with wild-type NFI, no cross-linking was detected when the C2 oligonucleotide was incubated with the mutant proteins (Fig. 5,  lanes 11,13, and 15) or in the absence of UV irradiation (Fig. 5, lane 9, and data not shown). Thus, NFI proteins with alanine, leucine, or valine substitutions at the Cys-3 position form UVinduced cross-links with DNA with an efficiency similar to that of wild-type hNFI-C220. DISCUSSION These findings indicate that of the 4 cysteine residues in hNFI-C which are completely conserved between known NFI family members (15), the Cys-2, Cys-4, and Cys-5 residues are intolerant of amino acid substitutions (Figs. 1 and 2), while the Cys-3 position may play a role in redox regulation of NFI DNA binding activity (Fig. 4). The complete intolerance of the Cys-2, Cys-4, and Cysd residues to substitution indicate an essential role for these residues in some aspect of NFI-DNA interaction. Although our lack of information on the structure of the NFI DNA binding domain prevents us from determining the precise role of these essential residues, our findings indicate a number of possibilities. Since the relatively conservative substitutions of serine or alanine a t these positions completely abolished DNA binding activity (Fig. 11, it appears that even minor changes in the charge or size of the residues a t these positions disrupts NFI function. This finding, in combination with our observation that all of the active Cys-3 substitution mutants are resistant to inactivation by NEM and diamide (Figs. 3 and  4), suggests that the Cys-2, Cys-4, and Cysd residues may be inaccessible to modification with NEM or diamide and thus may be buried in the core of the NFI protein. It is possible, for example, that two of these residues may participate in an intramolecular disulfide bond in NFI that is required for DNA binding activity, although the known resistance of the DNA binding activity of NFI to the presence 50-100 mM DTT (not shown) would indicate that such a disulfide bond would have to be unusually resistant to reduction. Another possibility is that these residues operate together for some essential function in NFI, such as the coordination of a metal ion needed for DNA binding. However, no such essential metal ion has been demonstrated for NFI-DNA interaction (11,151, and DNA binding by NFI is reduced by less than 50% in the absence of MgC1, and in the presence of 10 mM EDTA. 3 Since NFI has been shown to interact with DNA as a dimer 03-10), it will be important to determine which potential step in NFI function (i.e. initial subunit folding, subunit dimerization, or dimer-DNA interaction) is affected by mutation of the Cys-2, Cys-4, and Cys-5 residues.
The ability of the majority of NFI Cys-3 substitution mutants to bind DNA indicates that this cysteine residue is unlikely to be directly involved in NFI-DNA interactions. However, Arg or Trp substitution at the Cys-3 position of NFI abolished DNA binding activity (Figs. 1 and 21, and these two residues produce the largest side-chain substitutions generated in this study. In addition, the sensitivity of wild-type NFI to inactivation by NEM (Fig. 3), diamide (Fig. 4), and DTNB ( E ) , and the resistance of Cys-3 substitution mutants to such inactivation indicates that the addition of large chemical adducts a t the Cys-3 position can abolish DNA binding by NFI. Given these results, it is perhaps surprising that DNA binding by NFI partially protects the protein from inactivation by diamide (Fig. 4B), NEM or DTNB (15). It may be that the Cys-3 residue is located close to the DNA in the NFI.DNA complex but is not directly involved in protein-DNA interactions. This appears to be true S. Bandyopadhyay and R. Gronostajski, unpublished data. for the oxidation-sensitive cysteine residues detected in the c-Fos and c-Jun proteins, which are located near the basic DNA binding domains of these proteins but do not appear to directly contact DNA (22). Alternatively, DNA binding by NFI may induce a conformational change in the protein that reduces the accessibility of the Cys-3 residue to modification even though the Cys-3 position is quite distant from the actually DNA binding interface. Another possibility is that the Cys-3 residue in wild-type NFI does indeed interact directly with DNA, but that mutation of the residue has little or no effect on DNA binding affinity. Such an effect was seen in the bovine papilloma virus type 1 E2 protein, where conversion of a cysteine residue that was shown to directly contact DNA (43) to a serine residue had only a small effect on DNA binding affinity (27). Further analysis of the structure and function of the DNA binding domain of NFI will be required to distinguish between these possibilities. The DNA binding activity of the active C3S mutant protein was somewhat surprising, since we had previously failed to detect binding activity in a C3S mutant of hNFI-C220 (15). We attribute the former inability to detect activity in extracts of this mutant to two potential causes. First, the C3S mutation appears to modestly increase the toxicity of the NFI gene product in E. coli as assessed by a reduced growth rate in liquid cultures and a smaller colony size following transformation of plasmids containing this mutation. In addition, the induction conditions used previously (37 "C for 3 h) yield levels of active NFI that are 5-10-fold lower than those obtained using the conditions described here (17 "C for 16 h). Induction of NFI expression at 17 "C appears to significantly reduce the toxicity seen a t 37 "C, and we have determined that both the new C3 mutants described above and the former C3S mutant protein possess levels of DNA binding activity comparable to that of wild-type NFI. We are currently determining whether muta-tions at the Cys-3 position affect the temperature sensitivity, pH dependence, or other parameters of the DNA binding activity of hNFI-C.
The ability of NFI proteins containing alanine, isoleucine, and valine substitutions at the Cys-3 position to cross-link to DNA after UV irradiation (Fig. 5) demonstrates a second property of NFI that is unaffected by mutagenesis a t this position. Although the precise residueb) of NFI that cross-link to DNA have not been determined, previous studies with other DNAbinding proteins indicate that a variety of amino acids can participate in such cross-linking reactions. While a specific cysteine residue of the bacteriophage fd gene 5 protein cross-links to single-stranded DNA (Cys-33) (42) after UV treatment, specific phenylalanine residues on the bacteriophage T4 gene 32 protein (Phe-183) (441, the E. coli SSB protein (Phe-60) (45), and the adenovirus DNA-binding protein (Phe-418) (46) crosslink to DNA under similar conditions (47,481. While UV crosslinks have not been identified between alanine residues of proteins and specific bases of native DNA, an alanine residue of GCN4 (Ala-238) cross-links to 5-bromo-2'-deoxyuridine-substituted DNA when a 5-bromouridine base replaces a thymine base a t position +3 of a GCN4 binding site (49). In addition, although no studies have been reported which examine the relative ability of different natural amino acids a t the same position of a protein to cross-link to DNA after UV irradiation, it is likely that the efficiency of cross-linking is related to both the proximity of an amino acid residue to DNA and its chemical reactivity (47,48). Thus, the equivalent degree of cross-linking seen after UV irradiation of wild-type and mutant NFI.DNA complexes suggests that the Cys-3 position is not the site of UV cross-linking, but is not conclusive. However, the similar efficiency of both DNA binding ability, as measured in the gel mobility shift assay (Fig. 2), and UV cross-linking (Fig. 5) of the various Cys-3 mutant and wild-type NFI proteins, indicates that a large change in the overall structure of the DNA-protein complex due to these mutations is unlikely. It will be necessary to map the specific amino acid residueb) involved in NFI-DNA cross-linking to more fully interpret these data.
The finding that all of the active Cys-3 mutants of NFI are resistant to inactivation by both NEM (Fig. 2) or diamide oxidation (Fig. 4)? suggests that NFI may be subject to redox regulation in vitro and in vivo. Redox regulation of DNA binding activity was first demonstrated with the OxyR protein of E. coli, where oxidation of the protein changes its footprint on DNA and is required for transcriptional activation by OxyR on specific promoter sites (50,51). Subsequent studies in eukaryotes have demonstrated that the DNA binding activities of a number of site-specific DNA-binding proteins, including c-Fos and c-Jun (22,23,52), NFKB/Rel(24,25), BPV-1 E2 (27), USF (28), and Myb (26), are inactivated by oxidation of specific cysteine residues in vitro. These proteins contain a wide variety of different DNA-binding structural domains, which include the basic leucine zipper domain (c-Fos and c-Jun), the helix-loophelix domain (USF), an a-helix, dimeric P-barrel domain (BPVE2) (431, and two as yet uncharacterized DNA binding domains (Myb and NFKB/Rel). For each of these DNA-binding proteins, a specific cysteine residue or residues is sensitive to oxidation by diamide or other oxidizing and alkylating agents in vitro, and mutation of these sensitive residues generates proteins that are resistant to oxidative inactivation or NEM (22,24,(26)(27)(28)53). These properties, sensitivity to oxidation/ alkylation in vitro and subsequent resistance to inactivation following mutation, are similar to those described here for NFI. However, there are numerous intracellular proteins that share similar properties for which there is no evidence of redox regulation. Thus, additional studies on oxidative inactivation and the characteristics of oxidation-resistant mutants of NFI in vivo are essential to determine whether the transcriptional activation properties of NFI family members are modified by intracellular redox state.
Although there appear to be common features in the oxidative inactivation of several DNA-binding proteins, there are also distinct differences in the apparent mechanism of inactivation between the different proteins. Inter-and intramolecular disulfide bonds have been shown to form during oxidative inactivation of USF (28) and NFKB (24) and have been proposed to interfere with DNA binding by these proteins. However, interor intramolecular disulfide bonds have not been detected in the oxidative inactivation of c-Fos/c-Jun heterodimers or the BPV-1 E2 protein, and several groups have proposed that inactivation of these proteins may be mediated by the oxidation of cysteine sulfhydryl residues (-SH) to sulfenic acid residues (-SOH) (22,27,52,54). Reversible oxidation of a specific cysteine residue to sulfenic acid has also been proposed for the reversible modification of the OxyR protein in E. coli (50) Although frequently unstable, stabilized sulfenic acid residues have now been detected in a number of flavoproteins and oxidized enzymes and represent a reversible oxidation state of cysteine residues (see Ref. 54 for a review). While the mechanism for stabilization of sulfenic acid residues on proteins is unknown, one model that has been proposed is that such oxidized residues may be stabilized by a highly basic local environment surrounding the oxidation-sensitive cysteine residues of a number of proteins ( Fig. 6) (22,27,50,54). It is of interest that the oxidation sensitive Cys-3 residue of hNFI-C is located within such a potentially highly basic environment (Fig. 6, hNFI-C). In addition, although the mechanism of oxidativeinactivation of NFI is currently unknown, we have failed to detect any evidence by SDS-PAGE for interor intramolecular disulfide bond formation during oxidative-inactivation of wildtype NFI.'Also, the faster mobility of NFI-DNAcomplexes after treatment with diamide is more consistent with a small change in the charge of the complex than with covalent dimer formation. Similar increases in the mobility of NFI.DNA complexes were seen previously after treatment of the complexes with DTNB (15). Since mutation of the conserved Cys-2, Cys-4, and Cys-5 residues inactivates DNA binding, it will be difficult to determine by mutagenesis whether any of these residues participate in interor intramolecular disulfide linkages during oxidative inactivation of NFI. Thus, further biochemical analysis of the mechanism for oxidative inactivation of NFI in vitro is needed to resolve this issue.
In addition to studies on oxidative inactivation in vitro, there is growing evidence that redox regulation may play an important role in gene expression and cellular physiology in vivo. For DNA seen during apoptosis, we propose that the oxidativeinactivation of cellular DNA-binding proteins (which might normally protect DNA from degradation), may also be important in this process. In this regard it is intriguing that the bifunctional protein, HAPl/Ref-1, participates in both the enzymatic reduction of oxidized proteins and the endonucleolytic cleavage and repair of DNA damaged by reactive oxygen species (60-61). To test this model, it will be necessary t o determine whether oxidative inactivation of NFI occurs in vivo under physiological or pathological conditions and whether expression of oxidation-resistant mutants of NFI or other DNAbinding proteins affect apoptosis.