Oligomerization of Human Gadd45a Protein*

Gadd45a is an 18-kDa acidic protein that is induced by genotoxic and certain other cellular stresses. The exact function of this protein is not known. However, there is evidence for its involvement in growth control, maintenance of genomic stability, DNA repair, cell cycle control, and apoptosis. Consistently, Gadd45a has previously been shown to interact in vitro and/orin vivo with a number of proteins playing central roles in these cellular processes: proliferating cell nuclear antigen, p21Cip1/Waf1, Cdc2-CyclinB complex, MTK1, and histones. Adding to this complexity, we have found that Gadd45a self-associates in solution, both in vitro and when expressed in the cell. Moreover, Gadd45a can complex with the two other members of the Gadd45 family of stress-induced proteins, human Gadd45b (MyD118) and Gadd45g (CR6). Gel-exclusion chromatography, native gel electrophoretic analysis, enzyme-linked immunosorbent assay, and chemical cross-linking showed that recombinant Gadd45a forms dimeric, trimeric, and tetrameric species in vitro, the dimers being the predominant form. Deletion mutant and peptide scanning analyses suggest that Gadd45a has two self-association sites: within N-terminal amino acids 33–61 and within 40 C-terminal amino acids. Despite the low abundance of Gadd45a in the cell, oligomer-forming concentrations can probably be achieved in the foci-like nuclear structures formed by the protein upon overexpression. Evidence for a potential role of Gadd45a self-association in altering DNA accessibility on damaged nucleosomes is presented.

Gadd45a is a member of a mammalian protein family that includes two other homologous proteins, Gadd45b/MyD118 and Gadd45g/CR6 (1)(2)(3)(4). Transcription of the genes encoding these proteins is induced by various kinds of DNA-damaging agents and/or a number of stresses associated with growth arrest. Gadd45a is the only member of the family whose up-regulation has a p53 component. Its gene has a very strong p53-binding site in the third intron (5).
Gadd45a is a relatively small, 18.4 kDa, and highly acidic protein. It is of low abundance in the cell and is thought to localize primarily in the nucleus (6). Gadd45a shares with Gadd45b and Gadd45g this extreme charge characteristic and may act synergistically with them in promoting cell growth arrest upon overexpression in the cell (7). Besides this originally discovered role in suppression of cell growth (1), recent findings suggest specific function(s) for Gadd45a and/or the two other Gadd45 proteins in DNA repair (8 -10), apoptosis (11,12), maintenance of genomic stability (13), and regulation of signaling pathways (4).
The role(s) of Gadd45 proteins in these fundamental cellular processes are supported by several observations made both in vitro and in a mouse model. Studies of the Gadd45aϪ/Ϫ mouse show genomic instability (13), reduced nucleotide excision repair as well as increased levels of mutations and chemically induced tumorigenesis (14). Gadd45a and Gadd45b interact with PCNA, 1 which is an indispensable component of the DNA repair mechanism (8,11,12). This interaction is also suggested to impede UV-induced cellular apoptosis (12). The role of Gadd45a in repair may also be realized through its ability to bind histones and modify accessibility of DNA on damaged chromatin (15). Gadd45a interacts with p21 (Cip1, Wip1, Cdkn1A), a particularly important cell cycle regulator protein (16 -18). Gadd45a forms a specific complex with Cdc2-CyclinB mitosis promoting complex and inhibits its kinase activity, which is crucial for G 2 -M progression (19 -21). In addition Gadd45 proteins seem to directly bind and activate stressresponsive MTK1/MEKK4 kinase (4).
Even though these findings provide important clues to the cellular role(s) of Gadd45a protein, its function(s) remains poorly understood. In characterization of the function of any protein, especially of a protein participating in as multiple interactions as Gadd45a, it is important to establish a potential for self-association. In many cases formation of higher order oligomeric structures contributes crucially to protein functionality and its regulation and extends its ability to interact with other cellular targets (22)(23). Hence, understanding of potential homo-and hetero-interactions among Gadd45 proteins can aid in our understanding of the mechanism(s) by which these proteins carry out their roles in multiple cellular processes.
Here we present evidence that Gadd45a protein can selfassociate both in vitro and in the cell. It was found that two distinct regions of Gadd45a are involved in self-association. These regions at least partially overlap with PCNA binding domains, thus implying functional roles for self-association. As another clue for its functional relevance, it was found that the oligomeric state of Gadd45a may regulate its effect on accessibility of DNA on damaged chromatin. Moreover, Gadd45a is capable of forming hetero-complexes with two other Gadd45 proteins. These observations suggest a potentially complex network of interactions underlying the functions of Gadd45 proteins.

EXPERIMENTAL PROCEDURES
Construction of Expression Vectors-Escherichia coli and human constructs expressing Gadd45 proteins with or without N-terminal tag were obtained by inserting PCR amplified fragment containing Gadd45 open reading frames into suitable expression vectors. Primers for PCR amplification of inserts consisted of 15 or 18 nucleotides corresponding to open reading frame sequence flanked with restriction endonuclease sites that were used for cloning. The following combinations of templates containing Gadd45 open reading frames, restriction endonuclease sites, and endonuclease-digested expression vectors were employed. Plasmid pET21-Gadd45a, expressing untagged Gadd45a in E. coli: pET14b-Gadd45a (6), NdeI-and XhoI-flanked primers, NdeI-and XhoItreated pET21a vector (Novagen); pGEX4-Gadd45a plasmid for expression of GST-Gadd45a fusion protein in E. coli: pET14b-Gadd45a, Mun I-and XhoI-flanked primers, EcoRI-and XhoI-treated pGEX4T-1 plasmid (Amersham Pharmacia Biotech); pFLAG-Gadd45a vector, expressing N-terminally FLAG-tagged Gadd45a in human cells: pET14b-Gadd45a, Mun I-and XbaI-flanked primers, EcoRI-and XbaI-treated pcDNA3.1ϩ plasmid, containing FLAG sequence; pHA-Gadd45a and pHA-Gadd45b plasmids expressing N-terminally HA-tagged Gadd45a, Gadd45b, respectively, in human cells: pET14b-Gadd45a and Gadd45b human cDNA (clone number HHCMC72), respectively, Mun I-and XbaI-flanked primers, EcoRI-and XbaI-treated plasmid pcDNA3-HA (24), containing HA sequence. Plasmids pHA-Gadd45g and pcDNA3.1ϩ-FLAG were kindly provided by Drs. K. Smith (Cornell University) and Y. Gao (NIH, NCI), respectively. Expand High Fidelity PCR System (Roche Molecular Biochemicals) was used for PCR amplification. Cloning was carried out according to manufacturer's protocol provided with the Rapid Ligation Kit (Roche Molecular Biochemicals). All restriction endonucleases were from New England Biolabs. PCRamplified inserts and expression vectors were treated with endonucleases and agarose gel-purified before ligation reactions. All expression constructs were verified by sequencing analysis.
Expression and Purification of Gadd45a Protein in E. coli-Competent E. coli B834(DE3)pLysS cells (Novagen) were transfected with pET21-Gadd45a vector for untagged protein. Three-ml mini-cultures of LB supplemented with 100 g/ml carbenicillin and 1% glucose (growth medium) were inoculated with several distinct overnight colonies and grown until an A 600 of 0.5 was reached. Mini-cultures were expanded into 50-ml cultures of fresh growth medium and subsequently into 1-liter cultures. Large scale cultures were grown to an A 600 of 0.5 at 37°C followed by temperature shift to 28°C. Protein was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration 1 mM. After 3 h of induction at 28°C, cells were collected and frozen in liquid nitrogen. For purification, cells were thawed, resuspended in ice-cold lysis buffer (20 mM Tris-Cl, pH 7.4, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM DTT, ϳ20 ml/g of frozen cells), incubated with lysozyme (1 mg/ml, 30 min), and briefly sonicated to complete lysis and shred genomic DNA. All manipulations were on ice. Four M ammonium sulfate, pH 7.0, was slowly added to crude lysate (1 M final concentration). After 1-h incubation on ice, lysate was cleared by centrifugation (SS34 rotor, 12 Krpm, 30 min, 4°C). Cleared lysate was filtered (Millex-GV, 0.22 m, Durapore, low protein-binding syringe filter unit Millipore, SLGV 025LS) and loaded onto HiPrep 16/10 phenyl-Sepharose, high substitution column (Amersham Pharmacia Biotech) equilibrated with buffer A plus 1 M ammonium sulfate. Buffer A composition: 25 mM K ϩ MOPS, pH 7.0, 1 mM EDTA, 10% glycerol, 2 mM DTT. The column was washed with 100 ml of buffer A plus 0.2 M ammonium sulfate followed by 100 ml of buffer A plus 0.1 M ammonium sulfate. Protein was eluted with a linear gradient of 100 ml of buffer A plus 0.1 M ammonium sulfate to 100 ml of buffer A followed by additional 100 ml of buffer A. Fractions containing Gadd45a were pooled and directly loaded onto a HiPrep 16/10 DEAE (Amersham Pharmacia Biotech) column equilibrated with buffer A plus 0.05 M KCl. The column was washed with 100 ml of buffer A plus 0.1 M KCl. Protein was eluted with a linear gradient of 125 ml of buffer A plus 0.1 M KCl to buffer A plus 0.5 M KCl. Fractions, which contained Gadd45a, were pooled and concentrated to Ͻ5 ml using Stir Unit 8200 (Millipore) and a YM10 (Millipore) membrane. The salt content of pooled fractions was ϳ210 -220 mM KCl as determined by conductivity measurement. Concentrated protein was subjected to gel-filtration chromatography on HiPrep 26/60 Sephacryl S-200 (Amersham Pharmacia Biotech) column equilibrated with buffer A plus 0.2 M KCl. Protein was eluted with the same buffer. The concentration of protein in fractions containing the purest Gadd45a was determined spectrophotometrically (25) based on extinction coefficient 14,000 M Ϫ1 cm Ϫ1 calculated using Expasy proteomic tool (www. expasy.ch) and/or by Bradford assay using BSA as a standard. Fractions were aliquoted, frozen, and stored at Ϫ70°C.
Molecular mass markers for gel-filtration chromatography were run on the column under the same conditions in duplicate for calibration. Elution of dextran blue was used to determine the void volume of the column. Elution volumes and partition coefficients of the standards and samples were calculated according to Amersham Pharmacia Biotech protocol.
Native Gels-Laemmli discontinuous system without SDS was used for native gel separations. Mobilities of proteins were determined in 8, 10, 12, and 14% gels. Protein concentrations were adjusted to ϳ1 mg/ml with loading buffer. Ten or 20 g of protein was loaded onto gel. Relative electrophoretic mobilities (R f ) were calculated relative to the mobility of bromphenol blue. Calibration markers for native gels were from Sigma. Calibration curves were obtained and molecular masses calculated according to Refs. 26 and 27 and Sigma protocol (technical bulletin No. MKR-137).
Chemical Cross-linking in Solution-Dimethylsuberimidate (DMS) and disuccinimidyl glutarate chemical cross-linkers were from Pierce. Solutions of these cross-linkers were prepared and reactions carried out according to manufacturer's recommendations. In the case of DMS, a stock solution of reagent (25 mg/ml) was prepared in 20 mM HEPES, pH 7.9. Reactions were conducted for 30 min at 25°C in 20-l volume of 20 mM HEPES, pH 7.9, at 10 M protein, and 5 mg/ml DMS and varying KCl and DTT concentrations (Fig. 3). Reactions were stopped by adding 2 l of 1 M Tris-Cl, pH 6.8. Reactions were adjusted to 1 ϫ SDS loading buffer, boiled, and resolved by 12% SDS-PAGE. Protein products were detected by Coomassie Blue G-250 staining. Unstained protein molecular mass markers for denaturing gels (broad range, Bio-Rad) were run along the samples on the same gels.
ELISA Assays-Microtiter plates (Immulon II, Dynatech) were coated with higly purified Gadd45a protein by incubating wells with 50 l of protein diluted in PBS (2 g/ml) for 1 h at room temperature. After washing with PBST (PBS with 0.05% Tween 20) and blocking with 4% BSA, purified GST-Gadd45a (or GST in control) protein 2-fold serially diluted in 20 mM HEPES, pH 7.9, 150 mM NaCl, 1 mM DTT, 0.1 mM EDTA was added to the wells. Following 2-h incubation at 25°C and extensive washing with PBST, anti-GST monoclonal Ab (Covance Resarch Products) diluted 1:1000 in 1% BSA/PBS was applied to the wells for 1 h. Following reactions, plates were washed and incubated with horseradish peroxidase (HRP)-conjugated anti-mouse sheep Ab (Amersham Pharmacia Biotech) diluted 1:10000 in 1% BSA/PBS. Following 1-h incubation and washing, 0.4 mg/ml solution of o-phenylendiamine (Sigma) in citric-phosphate buffer, pH 5.0, containing 0.02% hydrogen peroxide was added for color development at 37°C. Reactions were stopped in 30 min with 2 M sulfuric acid, and optical density at 490 nm was measured using a plate reader (Molecular Dynamics).
In the case of peptide scan analysis, the same basic steps were followed except that different Gadd45a-derived peptides were immobilized on microtiter plates. The synthesis and sequence of peptides as well as details of ELISA procedure were described previously (21).
Transfections, Immunoblot, and Immunoprecipitation Analyses-Transfections were carried out with Effectene reagent according to manufacturer's protocol (Qiagen). For immunoanalyses, cell cultures were washed twice in ice-cold PBS, lysed on ice in ice-cold RIPA buffer (50 mM Tris-Cl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 20 mM ␤-glycerolphosphate; 0.6 ml per P100 plate) supplemented with protease inhibitor mixture (for mammalian cell and tissue extracts, Sigma). Lysates were cleared by centrifugation, and total protein concentration was determined by Bradford assay kit (Bio-Rad) using BSA as a standard. For immunoblot, an aliquot of lysate was supplemented with SDS loading buffer to 1ϫ final concentration, and denatured by boiling. Twenty-five to 100 g of total protein were typically loaded onto premade 12% SDS-PAGE (Novex) and electrophoresed. After electrophoresis, proteins were transferred onto polyvinylidene difluoride membrane (Millipore). Membranes were blocked in 5% dry milk in PBST and proteins analyzed with specific primary Abs diluted in 5% dry milk followed by extensive washing in PBS and incubation with secondary HRP-conjugated Ab diluted 1:5000. Following extensive washing in PBS, blots were developed with ECL reagents (PerkinElmer Life Sciences).
For immunoprecipitation, 0.5 or 1 mg of cleared total lysates were incubated with affinity matrix anti-FLAG(M2)-agarose (Sigma), anti-HA-agarose (Covance Research Products), or anti-c-myc-agarose (Santa Cruz), for 2 h or overnight at 4°C. Isolated immunocomplexes were washed five times in RIPA buffer and subjected to immunoblot analysis as described above.
Immunofluorecsence-Cells stably expressing FLAG-Gadd45a protein were used for immunofluorescence analysis. To obtain these cells, RKO human colorectal carcinoma cells were co-transfected with pFLAG-Gadd45a plasmid and plasmid carrying puromycin resistance in 10:1 ratio. For control cell lines, pcDNA3.1(ϩ)-FLAG vector was employed instead of pFLAG-Gadd45a. After 1 week selection in medium with 2 g/ml puromycin, clones expressing pFLAG-Gadd45a protein were determined by immunoblot analysis of total lysates using anti-FLAG monoclonal Ab. Expression of Gadd45a protein in FLAGpositive clones was verified by immunoblot with anti-Gadd45a Ab. For immunofluorescence, cells were seeded onto coverslips 24 h before fixation in 95% ethanol:5% acetic acid for 10 min at 25°C. Cellular localization of pFLAG-Gadd45a was probed with anti-FLAG monoclonal Ab (1:100 dilution) and visualized with Cy3-conjugated donkey anti-mouse Ab (Jackson ImmunoResearch) diluted 1:600. Slides were mounted in 4,6-diamidino-2-phenylindole-containing medium (Vectashield), and immunofluorescence was examined on an Olympus AX70 microscope.
Nucleosome Core Particles Preparation and T4 Endo Protection Assay-Nucleosome core particles isolated from chicken erythrocytes were a kind gift from Dr. Yu. Postnikov (NCI, NIH). One g of core particles was labeled for 10 min at 30°C in 10 l of polynucleotide kinase reaction buffer (Stratagene) with 5 Ci of [␥-32 P]ATP (Amersham Pharmacia Biotech) and 5 units of T4 polynucleotidekinase (Stratagene). The reaction was stopped by the addition of 1 l of 100 mM EDTA. Glycerol was added to 2% final concentration and the reaction electrophoresed on pre-run 5% PAGE. Core particles and free DNA, separated by electrophoresis, were isolated from crushed gel slices by overnight incubation in 20 mM HEPES, pH 7.9, 1 mM EDTA, 5% glycerol, and 100 g/ml BSA. Extracted molecules were cleaned of gel material by centrifugation through filter columns (Millipore) and dialyzed against T4 UV endo buffer (20 mM Tris-Cl, pH 6.8, 100 mM NaCl, 5 mM EDTA, 5% glycerol) in Microdialyzer System 100 using 50-kDa cut-off membranes (Pierce). Dialysis was according to Pierce protocol. Electrophoresis in 5% polyacrylamide showed that core particles obtained in this way contained less than 5% of free DNA (data not shown).
The T4 UV endo protection assay on UV-damaged nucleosomes has been described previously (15). Briefly, core particles and free DNA were UV-irradiated, and about 20 ng of the material was incubated with 0.5 l of T4 UV endo stock supplied by Dr. D. Yarosh (AGI Dermatics) with or without Gadd45a protein for 30 min at 37°C. The final reaction volume was 5 l. Reactions were stopped by addition of 20 l of 90% formamide in TBE buffer (89 mM Tris base, 89 mM boric acid, 2 mM EDTA, pH 8.3) containing electrophoretic dyes. Samples were boiled for 10 min and 5 l electrophoresed on pre-run 6% PAGE containing 7 M urea. Gels were dried and autoradiographed.
To enrich the oligomeric fraction in Gadd45a samples used in the T4 protection assay (Gadd45o), highly purified protein was incubated on ice for 15 min with 10 M NiCl 2 followed by dialysis to T4 reaction buffer using a Microdialyzer and 5-kDa cut-off membranes (Pierce). Gadd45a samples with preferentially monomeric species (Gadd45m) were treated in the same way except that NiCl 2 exposure was omitted.

RESULTS
Gel-filtration Chromatography-Gel-filtration chromatography was used as a final step in the purification of Gadd45 (see "Experimental Procedures"). Protein was already about 90% pure after the two preceding chromatographic steps. Chromatography on Sephacryl S200 clearly separates most of remaining higher molecular mass contaminants. The resulting protein was estimated to be more than 99% pure. Gadd45a was eluted from the column as a single peak, although with some "tailing" in the front, suggesting the presence of higher molecular mass species. Molecular mass standards for gel filtration chromatography (Amersham Pharmacia Biotech) were run through the column to obtain a calibration plot for determination of protein partition coefficients (Fig. 1). Comparison of the Gadd45a peaks from three purification runs with this calibration plot yielded an apparent molecular mass estimate of 30 Ϯ 4 kDa. 2 The molecular mass of Gadd45a calculated from its amino acid sequence is 18.4 kDa. This suggested that under the conditions of chromatography Gadd45a existed as a dimer or as monomer and dimer in rapid equilibrium.
Analysis of Gadd45a Electrophoretic Mobility in Native Polyacrylamide Gels (Fergusson's Analysis)-Analysis of the rela- tive electrophoretic mobilities (R f ) of the globular proteins in nondenaturing gels provides a sensitive technique to determine the aggregation state of the protein and estimate molecular masses of the oligomeric species (26,27). At least four bands can be resolved when Gadd45a at 1 mg/ml (starting concentration) is electrophoresed through such gels ( Fig. 2A). In contrast, only one band can be detected when the same amount of the protein is run under denaturing conditions (Fig. 2B). Thus, the additional bands in native conditions may represent oligomers of Gadd45a containing from two to four subunits. To estimate molecular masses of these presumed oligomers, molecular mass protein standards for native gel (Sigma) were electrophoresed along with the Gadd45a in duplicate through native gels of differing acrylamide percentage (%T). E. coli UvrB protein (28) was also used in some experiments. The log(R f ) values were plotted against %T. In all cases the plots were found to be linear (r 2 Ͼ0.97, data not shown). As a part of analysis, the linear plots for all four Gadd45a species would intercept log(R f )-axis in the same point (data not shown) if extrapolated to %T equal 0. This means that all Gadd45a species have the same charge per subunit, thus confirming that they are oligomers and not charge isomers (26,27). In further analysis, the logs of the retardation coefficients, represented by the slopes of R f versus %T graphs, for molecular mass standards were plotted against logs of their molecular masses, thus producing the molecular mass calibration plot (26,27). Comparison of the log(retardation coefficient) for Gadd45a species I-IV with this calibration plot (Fig. 2C) allowed us to estimate their molecular masses (Table I). 3 Comparison of thus obtained molecular masses with those calculated from amino acid composition for oligomers of Gadd45a suggests that I and II are most probably monomer and dimer, while III and IV contain three and four subunits, respectively, although additional higher subunit contents cannot be ruled out.
Chemical Cross-linking of Gadd45a Protein in Solution-The hydrodynamic methods (gel filtration and gel electrophoresis) strongly suggest a potential for Gadd45a self-association. These methods utilize movement of the protein through porous medium. Some artifacts can be generated due to the interaction with this medium. To rule out such a possibility and to further study Gadd45a self-association, chemical cross-linking of the protein in solution was examined with denaturing ( Fig. 3) gel electrophoresis. Such cross-linking was, indeed, achieved with a number of bifunctional reagents that react primarily with lysines, DMS and disuccinimidyl glutarate (Pierce). These reagents yielded a similar pattern of cross-linking. The results obtained with DMS are presented in Fig. 3. Incubation of Gadd45a with DMS resulted in the appearance of covalently linked oligomers of the protein, which were resolved by SDS-PAGE under denaturing conditions (Fig. 3). Comparison of the mobilities of the cross-linked species with those of molecular mass standards for denaturing gels (Bio-Rad) reveals them to be dimers and trimers, the former being the major species at this protein:cross-linker ratio. Increase in salt (KCl) concentration from 25 to 300 mM did not alter the yield of cross-linked species, implying the hydrophobic nature of oligomerization. Likewise, a 10-fold increase in DTT concentration had no effect on the extent and pattern of cross-linking (Fig. 3). Human Gadd45a protein has 5 cysteines. The absence of DTT effect argues strongly against disulfide formation-mediated nonspecific aggregation of the protein. Reducing Gadd45a concentration to 1 M resulted in the same distribution of covalently linked species, although higher order oligomers were predominant at higher cross-linker:protein ratio (data not shown).
Estimation of the Overall Affinity of Gadd45a Self-association-An ELISA-based immunoassay (29) was used to roughly estimate the overall affinity of Gadd45a self-association. This estimation was obtained to determine whether association occurred at physiologically relevant concentrations. Highly purified Gadd45a protein coated onto a 96-well microtiter plate was incubated with glutathione-agarose purified GST-Gadd45a at different concentrations. After washing, the amount of bound GST-Gadd45a was assessed by sequential incubations of the wells with anti-GST monoclonal Ab and HRP-conjugated sec-ondary Ab intermittent with extensive washing. Following another washing and incubation with HRP colorgenic substrate o-phenylendiamine, plates were read and a binding curve generated (Fig. 4). GST protein was used to monitor for possible nonspecific binding in this assay. It was found to be insignificant even at high concentrations (data not shown). The concentration, ϳ2.5 M, of GST-Gadd45a resulting in 50% of maximal binding was taken as an estimate of apparent K d of self-association. It should be emphasized that this value can constitute only a rough estimate, because self-association of GST-Gadd45a in solution as well as potential masking of binding sites by immobilization in the well were not taken into account. In addition, some oligomerization might be expected for the protein bound to the plate. All these effects probably contributed to an underestimate of the actual affinity for self-association.
Gadd45a Forms Foci-like Structures in Cell Nucleus-The levels of cellular Gadd45a can be appreciably induced by various types of stress (1). This increase may promote oligomerization. To determine the effect of increased levels of Gadd45a on its distribution in the cell, cell lines stably expressing FLAGtagged Gadd45a were generated. The levels of Gadd45a were increased about 10-fold in these cell lines (data not shown), which is comparable with stress-induced up-regulation. Highly specific anti-FLAG Ab was used for immunofluorescence analysis of these cells. It was found that the pattern of Gadd45a distribution forms foci in these cells (Fig. 5). Such compartmentalization of Gadd45a upon stress may produce concentrations sufficient for its oligomerization. This result provides yet another argument in favor of the possibility of Gadd45a oligomerization in vivo.
Gadd45a Self-associates and Complexes with Gadd45b and Gadd45g in the Cell-The self-association of the Gadd45a protein as well as its potential interaction with two other members of Gadd45 protein family, Gadd45b and Gadd45g, was further studied using in vivo co-precipitation experiments. RKO cells were transiently transfected with plasmids that express a FLAG-tagged and HA-tagged Gadd45a protein. FLAG-tagged (or HA-tagged) protein was immunoprecipitated, and co-precipitation of HA-tagged (or FLAG-tagged) Gadd45a was probed with anti-HA (or anti-FLAG, respectively) Ab (Fig. 6A). Both FLAG and HA immuniprecipitation resulted in HA-and FLAG-tagged Gadd45a co-precipitation, respectively (Fig. 6A,  upper panels). The same results were obtained when myctagged and FLAG-tagged Gadd45a expressing constructs were  used in co-precipitation experiments (Fig. 7A). These results provide evidence that Gadd45a can self-associate in mammalian cells.
Using a similar approach, the interaction of Gadd45 with the two other human Gadd45-related proteins, Gadd45b and Gadd45g, was investigated in RKO cells. These cells were cotransfected with plasmids expressing myc-tagged Gadd45a and either HA-tagged Gadd45b or HA-tagged Gadd45g. Tag-specific immunoprecipitation showed that Gadd45a co-precipitates with both Gadd45b and Gadd45g (Fig. 6B, upper panels). It was also found by the same approach that FLAG-tagged Gadd45b and HA-tagged Gadd45g co-precipitate when co-expressed in RKO cells (data not shown). These findings indicate that all three members of the Gadd45 protein family can form homo-and/or hetero-complexes in the cell.
Immunoblotting analysis of tagged-protein levels in total lysates, used for immunoprecipitation, was employed as a control for their comparable expression (Fig. 6, A and B, lower  panels).
Identification of the Self-Association Domain(s) of Gadd45a Protein-To map the Gadd45a domain(s) involved in self-association, a series of myc-tagged Gadd45a deletion mutants was employed. RKO cells were co-transfected with plasmids expressing myc-tagged Gadd45a deletion mutants and FLAGtagged full-length Gadd45a. The lysates from transfected cells were immunoprecipitated with anti-FLAG Abs and analyzed by immunoblotting with anti-myc Ab (Fig. 7A). The expression levels of myc-tagged Gadd45a proteins were analyzed after immunoprecipitation (upper panel) or in lysates (lower panel). The lysate levels were found to be abundant and comparable in all samples (Fig. 7A, lower panel). Full-length myc-Gadd45a and myc-tagged C-terminal deletion mutant Gadd45a-(1-71) were able to interact with full-length FLAG-tagged Gadd45a. N-terminal deletion Gadd45a-(74 -165) was also able to efficiently interact with FLAG-tagged Gadd45a. In contrast, the deletion mutant Gadd45a-(48 -132), lacking both N-terminal and C-terminal portions of the protein, was significantly deficient in the binding (Fig. 7A, upper panel). This suggests the existence of two self-association domains of Gadd45, one within the N-terminal region and another near the C-terminal region.
A series of overlapping peptides spanning through the entire sequence of Gadd45a (21) were used to further delineate interacting domains. Gadd45a peptides were immobilized onto microtiter plates and incubated with purified GST-Gadd45a protein. The amount of bound GST-Gadd45a was assessed with anti-GST Abs as described above (see "ELISA Assays"). Incubation of peptides with GST protein alone was used as the control for specificity and background. It was found that Nterminal peptides 33-52 and 42-61 as well as C-terminal peptides 113-132, 129 -148, and 145-165 bound GST-Gadd45 two to three times stronger than neighboring peptides or peptides in the central region (Fig. 7B). The absence of a correlation between binding and the presence of cysteines in peptide provides another argument against nonspecific disulfide formation-driven interaction.
Oligomerization State of Gadd45a Protein Influences Its Ability to Protect Cyclobutane Pyrimidine Dimers in in Vitro UV-irradiated Nucleosomes from T4 UV Endonuclease Cleavage-It was previously found that Gadd45a can bind nucleosomes in vitro and protect cyclobutane pyrimidine dimers in nucleosomes from T4 UV endo digestion (15). These effects, however, were observed only with Gadd45a purified without usage of nickel-chelating chromatography, although nickel column-purified protein was similarily active in other nucleosome related assays (Ref. 15; data not shown). Consequently, it was found that small amounts of nickel result in a significantly increased level of oligomerization of Gadd45a, which is persistent even after nickel removal 4 ; these oligomers appear indistinguishable from spontaneous oligomers. Total nucleosomes obtained from chicken erythrocytes were UV-irradiated and incubated with Gadd45a preparations having different levels of oligomerization followed by incubation with T4 UV endo. The unirradiated nucleosomes were used as a control. The T4 UV endo digestion of total UV-irradiated nucleosomes resulted in a decrease of full-length nucleosomal DNA and the appearance of a diffuse smear of lower molecular mass DNA fragments (Fig.  8). Total nucleosomes contain DNA of the same length but different sequence, and digestion of these DNAs produces DNA fragments of many possible lengths. The electrophoretic mobility of these fragments can additionally be influenced by their sequence, thus resulting in a smear-like pattern instead of distinct bands. It was found that an increased level of Gadd45a oligomerization significantly reduced the protection of nucleosomes (Fig. 8). Consistent with previous observations, there was no protective effect in the case of DNA devoid of nucleosomes regardless of the Gadd45a oligomerization level. Thus it appears that oligomerization of Gadd45a protein abolishes its nucleosomal damage protective effect on T4 UV endo digestion. The effect of nickel on Gadd45a other than increase of its oligomerization cannot be ruled out at present. However, the ability of nickel-exposed protein to efficiently promote chromatin relaxation by eukayotic topoisomerase I (Ref. 15; data not shown) lends support to the above suggestion.

DISCUSSION
Determining the oligomeric state of Gadd45a protein is important for understanding the molecular mechanism(s) by which this protein carries out its multiple functional interactions. Using a variety of approaches, it was shown that Gadd45a is capable of self-association forming stable and reproducible oligomeric structures both in vitro and when expressed in the cell. Gel filtration, electrophoresis, and chemical cross-linking data strongly suggest that under in vitro physiological solution conditions dimers are the predominant oligomeric form of Gadd45a protein, although oligomers of higher order can also be detected by these methods.
The self-association seems to be driven largely by hydrophobic interactions, because neither the extent nor the pattern of cross-linking are changed by a 10-fold increase in salt concentration. Consistently, it was found that Gadd45a is strongly retained by hydrophobic matrix, phenyl-Sepharose. This property provided a very efficient step in chromatographic purification of Gadd45a protein. It also indicates the presence of hydrophobic patch(es) on the surface of the protein, which is accessible for interaction with other proteins and cellular components. Gadd45a is overall negatively charged under physiological conditions (pK 4.2) (7), which hints of involvement of the protein in ionic interactions. The presence of surface-accessible hydrophobic domain(s) extends the spectrum of potential interactions of Gadd45a protein. As a potential reflection of its hydrophobicity, it was found that Gadd45a is only partially extractable from the cell by non-ionic detergent-based RIPAlike lysis buffers, even after increase of ionic strength (data not shown). A significant part of cellular Gadd45a remained bound to the insoluble nuclear matrix/chromatin fraction. This strong binding can be mediated in vivo by hydrophobic interactions.
It has been shown that a number of cellular proteins involved in DNA processing form foci-like structures in the nucleus, particularly in response to cell stress associated with DNA damage. These structures may constitute repair sites at which specialized proteins assembled on chromatin/nuclear matrix carry out functional activities (30 -33). Gadd45a was found to also form foci-like structures in the nucleus (Fig. 5). This observation together with the abovementioned low extractability of Gadd45a suggests that Gadd45a may be a component of such repair assemblies. These foci are found in cells stably transfected with a Gadd45a-expressing construct even in the absence of exogenous stress. However, the increased level of Gadd45a protein in these cells seems to be itself "stressful" for the cells judging from their slow growth (data not shown). Moreover, this level, estimated to be about 10-fold higher than in control cells, is comparable with the levels of the protein induced in this type of cells by methylmethane sulfonate, x-ray, and UV radiation treatment (6,9). 5 Therefore, these cells may to some extent mimic the situation in normal cells upon exogenous stress. In agreement with the latter suggestion, the foci-like pattern of Gadd45a localization was not qualitatively changed after genotoxic stress, although the number of foci appeared to be increased (preliminary experiments with UV radiation and methylmethane sulfonate, data not shown). Compartmentalization of Gadd45a molecules in foci structures would be expected to further drive oligomerization through the increase of effective concentration.
An important issue is whether Gadd45a can oligomerize at physiologically relevant concentrations. For example, a K d of association in the millimolar range would be unlikely to be achieved in vivo. To address this issue, an ELISA approach (29) gave an estimated K d of ϳ2.5 M, although the actual value is probably less for the reasons already discussed. However, even this approximate result increases the confidence that self-association occurs in vivo. Indeed, we estimate that there are ϳ40,000 copies of Gadd45a per RKO cell, 6 6. Association of Gadd45a protein with itself and the two other members of the Gadd45 protein family, Gadd45b and Gadd45g, in mammalian cells. RKO cells were transiently co-transfected with: FLAG-Gadd45a and HA-Gadd45a constructs (A). Either FLAG-Gadd45a (left panels) or HA-Gadd45a (right panels) was immunoprecipitated from total lysates of transfected cells with anti-tag immunomatrix (IP), and the presence of HA-Gadd45a or FLAG-Gadd45a was probed by immunoblotting (IB) with the respective anti-tag Ab. The amounts of HA-(left) and FLAG-Gadd45a (right) proteins in total lysates detected by immunobloting are shown in the lower panels. B, myc-Gadd45a and either HA-Gadd45b or HA-Gadd45g constructs. Samples were immunoprecipitated and immunoblotted as described above using anti-myc and anti-HA immunomatrix and Abs. The amounts of mycand HA-tagged Gadd45 proteins in total lysates are shown in the lower panels. For controls, co-expression with FLAG, HA, and myc vectors was employed.
spheroid-shaped cell. Taking into account that Gadd45a is predominantly a nuclear protein, and that the RKO nucleus is estimated to be at least 5-fold smaller in volume, the nuclear concentration of Gadd45a falls into the hundreds of nanomolar to micromolar range. Further compartmentalization in Gadd45a foci would easily bring its concentration into the micromolar range.
Two regions of Gadd45a protein were found in this study to be implicated in self-association. One is localized within Nterminal amino acids 33-61 and another within the C-terminal amino acids 129 -165. The N-terminal binding site is enriched in hydrophobic and non-charged amino acids and, hence, may contribute crucially to the observed hydrophobic nature of oligomerization. It is not clear whether these two sites form a single binding surface in the protein. However, removal of one of the sites in deletion mutants does not abolish the capability of the other site to mediate self-association in vitro and in the cell.
It was previously shown that Gadd45a protein can interact in vitro and/or in vivo with a number of other cellular proteins: p21, PCNA, cdc2-cyclinB complex, histones, and MTK1/ MEKK4 kinase. Delineation of Gadd45a regions, which are involved in binding of these proteins and self-association domains, shows that the latter overlaps with PCNA binding sites (Fig. 9). The physiological significance of the overlap remains to be determined. However, it may imply a potentially regulatory role for Gadd45a oligomerization. In an attractive model, the formation of oligomers would render Gadd45a incapable of interacting with one of these proteins while promoting interaction with another. Stress to the cell induces Gadd45a levels and, hence, should influence the oligomeric state of the protein.
Consequently, it may result in altering the pattern of Gadd45a interactions in the cell. Because different stresses up-regulate Gadd45a to different extents (34), the oligomeric status of the protein may be stress-dependent. Hence, cellular interactions of Gadd45a, which are necessary to adequately respond to  (1-165). The total lystaes of transfected cells were immunoprecipitated with anti-FLAG matrix (anti-Flag IP), and the presence of deletion mutants in precipitate was analyzed (upper panel) by immunoblotting with anti-myc Ab (anti-myc IB). The abundant expression of myc-tagged Gadd45a proteins in total lysates was confirmed by immunoblotting with anti-myc Abs (lower panel). The cotransfection of FLAG vector and myctagged Gadd45a construct 1-165 was used as control. B, binding of GST-Gadd45a protein to Gadd45a peptides was determined by ELISA assay. Peptides spanning the entire sequence of Gadd45a protein were immobilized on a microtiter plate and incubated with full size Gadd45a-GST fusion protein. The amount of bound Gadd45a-GST fusion protein was assessed with anti-GST Ab. specific stresses, can be regulated by adjusting its oligomerization status. Taking into account structural and functional similarities between all three Gadd45 proteins (24), it is highly probable that Gadd45b and Gadd45g are also capable of efficient oligomerization. Together with the evidence of heterocomplexing of these proteins (Fig. 6B), this property of Gadd45 proteins significantly extends the spectrum of possible responses to a multiplicity of stresses and stimuli to the cell. As an example of potential cooperation of Gadd45a and Gadd45b proteins, it was found that they synergistically suppressed cell growth after overexpression (7). Also, tissue and stress/stimulus specificity of Gadd45 proteins distribution and induction (35) may further contribute to the specificity of cellular response.
As evidence of oligomerization dependence of Gadd45a interactions, it was found that self-association of the protein influences accessibility of DNA on damaged nucleosomes (Fig. 8). It was shown previously that Gadd45a protects cyclobutane pyrimidine dimers induced by UV irradiation of isolated nucleosomes from digestion by T4 UV endo (15). This effect suggested that Gadd45a may preferentially recognize and bind DNA-damaged sites in chromatin, thus influencing their accessibility to DNA processing proteins, e.g. DNA repair machinery. The protective effect was significantly reduced in a Gadd45a preparation enriched in oligomers, so that cyclobutane pyrimidine dimers are no longer efficiently protected from digestion (Fig. 8). Hence, the mode of interaction of Gadd45a with damaged chromatin is dependent on its oligomeric status.
In summary we show here that Gadd45a protein is capable of self-assembly and interaction with two other Gadd45 proteins both in vitro and in the cell. These interactions may constitute a basis for regulation and tuning of the functional activities of Gadd45 proteins in growth arrest, DNA repair, and other stress-induced processes. FIG. 8. Oligomerization state of Gadd45a protein influences its protection of UV-irradiated nucleosomes from T4 UV endo digestion. Nucleosomes (left) or free DNA (right) were UVirradiated (254 nm, 7500 Jm Ϫ2 ) and incubated in the presence or absence of samples of highly purified Gadd45a protein, which contain predominantly monomeric (Gadd45a,m) or oligomeric (Gadd45a,o) species. One-hundred and 300 g of protein were used in reactions with T4 UV endo and UV-irradiated nucleosomes, while 300 g were used in the rest of reactions containing Gadd45a. Native gel electrophoresis shows distribution of monomers and oligomers in the above samples of Gadd45a (lower panel).
FIG. 9. Domains of Gadd45a protein implicated in its interaction with other cellular proteins and self-association. Regions of Gadd45a implicated in interaction with PCNA (12 and 16), p21 (18), Cdc2 (21), and core histones (15) are shown along with oligomerization domains identified in the current study.