A possible role of the C-terminal domain of the RecA protein. A gateway model for double-stranded DNA binding.

According to the crystal structure, the RecA protein has a domain near the C terminus consisting of amino acid residues 270-328 (from the N terminus). Our model building pointed out the possibility that this domain is a part of “gateway” through which double-stranded DNA finds a path for direct contact with single-stranded DNA within a presynaptic RecA filament in the search for homology. To test this possible function of the domain, we made mutant RecA proteins by site-directed single (or double, in one case) replacement of 2 conserved basic amino acid residues and 5 among 9 nonconserved basic amino acid residues in the domain. Replacement of either of the 2 conserved amino acid residues caused deficiencies in repair of UV-damaged DNA, an in vivo function of RecA protein, whereas the replacement of most (except one) of the tested nonconserved ones gave little or no effect. Purified mutant RecA proteins showed no (or only slight) deficiencies in the formation of presynaptic filaments as assessed by various assays. However, presynaptic filaments of both proteins that had replacement of a conserved amino acid residue had significant defects in binding to and pairing with duplex DNA (secondary binding). These results are consistent with our model that the conserved amino acid residues in the C-terminal domain have a direct role in double-stranded DNA binding and that they constitute a part of a gateway for homologous recognition.

According to the crystal structure, the RecA protein has a domain near the C terminus consisting of amino acid residues 270 -328 (from the N terminus). Our model building pointed out the possibility that this domain is a part of "gateway" through which double-stranded DNA finds a path for direct contact with single-stranded DNA within a presynaptic RecA filament in the search for homology. To test this possible function of the domain, we made mutant RecA proteins by site-directed single (or double, in one case) replacement of 2 conserved basic amino acid residues and 5 among 9 nonconserved basic amino acid residues in the domain. Replacement of either of the 2 conserved amino acid residues caused deficiencies in repair of UV-damaged DNA, an in vivo function of RecA protein, whereas the replacement of most (except one) of the tested nonconserved ones gave little or no effect. Purified mutant RecA proteins showed no (or only slight) deficiencies in the formation of presynaptic filaments as assessed by various assays. However, presynaptic filaments of both proteins that had replacement of a conserved amino acid residue had significant defects in binding to and pairing with duplex DNA (secondary binding). These results are consistent with our model that the conserved amino acid residues in the C-terminal domain have a direct role in doublestranded DNA binding and that they constitute a part of a gateway for homologous recognition.
RecA protein is a key enzyme in homologous genetic recombination and recombinational repair of damaged DNA. RecA protein, which consists of 352 amino acid residues, promotes the formation of heteroduplex joints between single-stranded and double-stranded DNA by homologous pairing and strand exchange in vitro (1,2). In homologous pairing, RecA protein first binds to single-stranded DNA in the presence of ATP ("primary binding"; Ref. 3) and forms helical nucleoprotein filaments called "presynaptic filaments," in which singlestranded DNA is extended to a length 1.5 times longer than the B form DNA (4 -7). Then, presynaptic filaments bind to doublestranded DNA ("secondary binding") and form a three-component complex including single-stranded DNA, double-stranded DNA, and RecA protein (and ATP) without prior homologous alignment. The search for homology between single-stranded and double-stranded DNA occurs in this three-component complex (3,8,9).
We had shown that a chimeric RecA protein of Escherichia coli and Pseudomonas aeruginosa, RecAc38, is defective in homologous pairing of single-stranded DNA with doublestranded DNA because of deficiencies in secondary binding of double-stranded DNA to the mutant protein-single-stranded DNA filaments but is proficient in promoting renaturation of complementary single-stranded DNA (10,11). These observations led us to propose a three-DNA strand binding site model in which each site plays a functionally distinct role, and the characteristics of RecAc38 are explained if we assume that RecAc38 has a deficiency in one of the two sites required for secondary DNA binding (10,11).
From observations on the interaction of RecA protein and chemically modified DNA, Takahashi and co-workers (12) suggested that RecA protein has three DNA binding sites. Others have proposed two symmetric DNA binding sites on a helical RecA protein DNA filament (4). An x-ray crystallographic study and features of mutant RecA proteins suggest that the DNA binding sites are the regions from the 157th to 164th amino acid residues (called loop 1) and from the 195th to 209th amino acid residues (called loop 2), each of which forms a disordered loop (13 and references therein for mutations). This view was supported by photochemical cross-linking experiments (14,15). An oligopeptide containing loop 2 was found to bind both single-stranded and double-stranded DNA (16). Moreover, the oligopeptide was shown to promote homologous pairing of single-stranded DNA and negatively superhelical closed circular double-stranded DNA (form I DNA; Ref. 17). These observations further support the hypothesis that loop 2 is a DNA binding site critical for homologous pairing.
Studies using electron microscopy and neutron scattering on presynaptic filaments have located single-stranded DNA inside RecA protein filaments (18,19). In secondary binding, doublestranded DNA outside presynaptic filaments has to find a "gateway" to contact single-stranded DNA directly and to search for homologous sequences. Following this model, the sites for secondary binding are not really "sites" but, rather, "paths." Each strand of double-stranded DNA has multiple contact with amino acid residues along each path that begins at the gateways. By comparing a helical filament structure of presynaptic filaments obtained by the electron microscopic studies and a spiral structure of free or ADP-bound RecA protein obtained by an x-ray crystallographic analysis, one can assume that presynaptic filaments have a spiral structure similar (but extended) to that of free or ADP-bound RecA protein (20,21). In the crystal structure of RecA filaments, there is a cleft between adjacent RecA monomers, which connects the inside and outside of the RecA helical filaments ( Fig. 1; see ref. 13). A domain consisting of amino acid residues 270 -328 near the C terminus is located at the entrance of the cleft (Fig. 1). Recent photochemical cross-linking experiments suggested that a region (amino acid residues 233-243) facing the cleft and another region (residues 257-280) adjacent to and including a part of the C-terminal domain are directly involved in DNA binding (15,22). Double-stranded DNA can be located in this cleft without serious clashes with RecA monomers and can be apposed to the single-stranded DNA without any topological problems. 1 The epitope of an anti-RecA monoclonal IgG, ARM191, which inhibits the binding of double-stranded DNA to RecA protein without affecting single-stranded DNAdependent ATPase activity of the protein (23), was mapped in the C-terminal domain (24).
Few mutant RecA proteins with an amino acid substitution in the C-terminal domain have been characterized in vitro.
Thus, considering the model-building study described above and the epitope of ARM191, we constructed by site-directed mutagenesis six mutant RecA proteins that have a single (or double, in one case) substitution for basic amino acid residues in the C-terminal domain, two of them at either of two well conserved basic amino acid residues (allowing substitution of Arg for Lys and vice versa) among bacterial RecA proteins and others at a nonconserved or less conserved Lys or Arg residue, as references (25). Features of these mutant RecA proteins are consistent with the proposed gateway for double-stranded DNA described above.

MATERIALS AND METHODS
DNA Substrates-Negatively superhelical double-stranded DNA produced in vivo (form I DNA) and phage circular single-stranded DNA of Escherichia coli phage M13 (6407 bases), M13mp19 (7249 bases) and X174am3 (5386 bases) were prepared as described (26,27). M13Gori1 form I DNA was prepared as described (28). Linear double-stranded DNA of M13Gori1 with termini homologous to M13 was formed by treatment with BamHI (see Ref. 28) and labeled with [␥-32 P]ATP by using T4 polynucleotide kinase (New England Biolabs; conditions described in the manual). 35 S-Labeled single-stranded DNA fragments were synthesized by Klenow fragment in vitro by the following method. An M13 primer (M4; purchased from Takara Shuzo Ltd., Kyoto, Japan) and M13mp19 circular single-stranded DNA were annealed at 1:1 (in DNA molecules), and the primer (1 M) was extended by Klenow fragment (2 units/l; purchased from Takara Shuzo Ltd.) at 37°C for 20 min in a reaction mixture (23 l) containing 7 mM Tris-HCl (pH 7.5), 0.1 mM EDTA, 20 1 A. Sarai and H. Kurumizaka, unpublished model building study. mM NaCl, 7 mM MgCl 2 , 29 M dATP, 29 M dTTP, 1.5 M dGTP, 20 M 2,3Ј-dideoxy-GTP, and 0.87 M [␣-35 S]dCTP (specific activity, 37 MBq/ nmol). After 20 min, dATP, dCTP, dTTP, and dGTP (258 M each) were added to complete the reaction, and the incubation was continued at 37°C for 20 min. Then, the reaction was stopped by heating at 68°C for 10 min, and free nucleotides were removed by ethanol precipitation twice. The DNA substrates synthesized were denatured by boiling for 10 min and were immediately frozen by dipping in liquid N 2 . The sizes of the 35 S-labeled single-stranded DNA fragments ranged between 700 and 1000 bases, as estimated by agarose gel electrophoresis.
Single-stranded oligodeoxyribonucleotides were prepared as described (29,30). Their sequences (5Ј 3 3Ј) were: Oligodeoxyribonucleotides were labeled with [␥-32 P]ATP by using T4 polynucleotide kinase (New England Biolabs; conditions described in the manual). The amounts of DNA and oligodeoxyribonucleotides were expressed in moles of nucleotide residues.
Site-directed Mutagenesis-A 2.5-kilobase pair Escherichia coli DNA fragment carrying the recA gene was cloned into a BamHI site of the pUC119 plasmid vector. The plasmid vector carrying the recA gene, pUC119-recA, was amplified in E. coli dut Ϫ ung Ϫ strain (CJ236) to isolate uracil-containing single-stranded DNA (31). Primers that were used for the mutagenesis were: K280/282N primer, 5Ј-GAC CTG GGC GTA AAC GAG AAT CTG ATC GAG AAA GC-3Ј; K286N primer, 5Ј-G ATC GAG AAC GCA GGC GCG T-3Ј; K297N primer, 5Ј-A GGT GAG AAC ATC GGT CAG G-3Ј; K302N primer, 5Ј-GT CAG GGT AAC GCG AAT GCG-3Ј; K317N primer, 5Ј-A ACC GCG AAC GAG ATC GAG A-3Ј; and R324Q primer, 5Ј-C GAG AAG AAA GTA CAG GAG TTG CTG CTG AG-3Ј. Bold letters indicate the residues that were replaced for the mutagenesis. The primers, which were phosphorylated at the 5Ј-ends, were annealed to uracil-containing single-stranded pUC119-recA. The complementary strand of pUC119-recA was synthesized by using T4 DNA polymerase (Takara Shuzo Ltd.) and was ligated with E. coli DNA ligase at 25°C for 2 h in a buffer containing 50 mM Tris-HCl (pH 8.0), 60 mM ammonium acetate, 5 mM MgCl 2 , 5 mM dithiothreitol, 1 mM NAD, and 0.5 mM each of dATP, dTTP, dCTP, and dGTP. Then, doublestranded pUC119-recA was amplified in E. coli BMH71-18 (mutS), followed by amplification in E. coli MV1184. The plasmid DNAs carrying the mutant recA genes were isolated from the transformants and were sequenced to confirm the presence of a defined mutation and the absence of other mutations.
Purification of E. coli RecA Protein and Mutant RecA Proteins-E. coli RecA protein was extensively purified as described (27). The mutant recA genes were inserted into a multicopy plasmid (pKK223-3) and were expressed under the control of the tac promoter in a ⌬recA derivative of E. coli AB1157 in which an N-terminal region of the recA gene E. coli cells that had a plasmid for overexpression of a mutant recA gene were grown in 3 liters of L broth supplemented with ampicillin (100 g/ml) at 37°C. At the midlog phase of cell growth, the mutant recA gene was induced by the addition of 1 mM isopropyl-1-thio-␤-Dgalactopyranoside, followed by incubation at 37°C for 4 h. Then, cells were collected and frozen by dipping in liquid N 2 . Subsequent procedures for purification of mutant RecA proteins were the same as those for wild-type RecA protein, except that gel filtration was omitted. All fractions that contained a mutant RecA protein were tested to confirm the absence of detectable endonuclease activity. The fractions of mutant RecA proteins (ϳ5 g) from the final chromatography were analyzed by electrophoresis through a 10% polyacrylamide gel in the presence of 0.1% SDS followed by staining with Coomassie Blue R-250. We found no detectable protein bands other than a band of RecA protein in the fractions used in this study. Thus, the purity of the mutant RecA protein preparations used in this study was 98 -99% or more.
The concentration of wild-type RecA protein was determined by the Folin phenol reagent method described by Lowry et al. (32) using bovine serum albumin as a standard. The value obtained by this method coincides (within 10%) with that determined by the absorbance at 277 nm using E 277 nm 1% ϭ 6.33 (7). The concentrations of mutant RecA proteins were determined by the use of Coomassie Blue G-250 (a protein assay kit of Bio-Rad; Ref. 33 35 S-labeled DNA fragments (ϳ1 M) were incubated in a buffer containing 31 mM Tris-HCl (pH 7.5), 13 mM MgCl 2 , 1.3 mM ATP, 1.8 mM dithiothreitol and 88 g of bovine serum albumin/ml at 37°C. The reaction was initiated by the addition of RecA protein. After 10 min of the incubation, the reaction was terminated by chilling at 0°C, and RecA protein was removed from DNA by treatment with 0.5% SDS, 40 mM EDTA, and proteinase K (240 g/ml), followed by incubation at 37°C for 10 min. The products of homologous pairing (D loops) were separated from unreacted DNA substrates by electrophoresis through a 1% agarose gel. The labeled single-stranded DNA fragments incorporated into D loops and those unreacted were quantitated with a BAS-2000 image analyzer (Fuji). Fractions (percentages) of 35 S-labeled fragments incorporated into D loops were calculated. We compared this homologous paring assay using labeled single-stranded DNA fragments and doublestranded DNA with the conventional D loop filter assay (see Ref. 27;34). In both assays, homologous pairing promoted by wild-type RecA protein showed similar kinetic parameters in relation to concentrations of RecA protein, time course, and the dissociation phase of D loops. 2 Although unlabeled template DNA, complementary to labeled singlestranded DNA fragments, remained in the preparation of the labeled fragments, we did not observe any products of the unlabeled template DNA. Thus, we concluded that the effect of the unlabeled template DNA on the assay of homologous pairing can be ignored.
Assay for Homologous Pairing of Circular Single-stranded DNA and Full-length Linear Double-stranded DNA by RecA Protein-M13 singlestranded DNA (3.0 M) was incubated with RecA protein at 37°C for 7 min in a buffer containing 31 mM Tris-HCl (pH 7.5), 1 mM MgCl 2 , 1.3 mM ATP, 1.8 mM dithiothreitol, 88 g of bovine serum albumin/ml and an ATP-regenerating system consisting of 8.0 mM phosphocreatine and 10 units/ml creatine phosphokinase. Then, the MgCl 2 concentration was increased to 13 mM, and the incubation was continued. Three min after the addition, the reaction was initiated by the addition of 4.0 M M13Gori1 double-stranded 32 P-labeled DNA that had been linearized by the treatment with BamHI (generating homologous termini). After 30 min of the incubation, the reaction was terminated by chilling at 0°C, and RecA protein was removed from DNA by the treatment with 1.1% sarkosyl and 12 mM EDTA at 0°C. The products of homologous pairing (joint molecules) were separated from unreacted DNA substrates by electrophoresis through a 0.9% agarose gel. The labeled joint molecules and the unreacted double-stranded DNA were analyzed with a BAS-2000 image analyzer. Fractions (percentages) of 32 P-labeled joint molecules were calculated.
Assay for Homologous Pairing of Single-stranded DNA and Duplex Oligodeoxyribonucleotide-M13 circular single-stranded DNA (10 M) was incubated for 8 min with 5.0 M RecA protein at 37°C in a buffer containing 33 mM PIPES-NaOH 3 (pH 7.0), 1.0 mM magnesium acetate, 2.0 mM dithiothreitol, 100 g/ml bovine serum albumin, 1.2 mM ATP, and an ATP-regenerating system consisting of 8.0 mM phosphocreatine and 10 units/ml creatine phosphokinase. After the concentration of Mg-acetate was increased to 15 mM and 0.10 M of preannealed ϩ 33-mer oligodeoxyribonucleotide-Ϫ 33-mer oligodeoxyribonucleotide duplex in which the plus strand was 32 P-labeled and was added to initiate the reaction. Aliquots were taken at the indicated times and deproteinized by treatment with 0.5% SDS, 20 mM EDTA, and 200 g/ml proteinase K for 15 min at 37°C. Aliquots were subjected to electrophoresis in a 15% polyacrylamide gel under nondenaturing conditions. A Molecular Dynamics PhosphorImager was used to measure the release of the plus strand from the oligodeoxyribonucleotide duplex.
Assay for Homologous Pairing of Single-stranded DNA and Proximal Hairpin Oligodeoxyribonucleotide-M13 circular single-stranded DNA (2.0 M) was incubated for 8 min with 4.0 M RecA protein at 37°C in the presence of a buffer containing 33 mM Tris-HCl (pH 7.5), 1.0 mM MgCl 2 , 2.0 mM dithiothreitol, 100 g/ml bovine serum albumin, 1.2 mM ATP, and an ATP-regenerating system consisting of 8.0 mM phosphocreatine and 10 units/ml creatine phosphokinase. After the concentration of MgCl 2 was raised to 15 mM and the reaction mixture was incubated another 12 min, 0.045 M proximal hairpin oligodeoxyribonucleotide labeled with 32 P at 5Ј-termini was added to initiate the reaction. Aliquots were taken at the indicated times and deproteinized by treatment with 0.5% SDS, 20 mM EDTA, and 200 g/ml proteinase K for 15 min at 37°C. Samples were subjected to electrophoresis in 0.5% agarose. Hairpin-M13 homologous joint molecules were quantitated by means of a PhosphorImager.
Assay for Single-stranded DNA-dependent ATPase Activity of RecA Protein-[ 14 C]ATP (1.3 mM) was incubated with the indicated amounts of RecA protein and M13mp19 circular single-stranded DNA at 37°C for 45 min in a buffer containing 31 mM Tris-HCl (pH 7.5), 13 mM MgCl 2 , 1.8 mM dithiothreitol and 88 g of bovine serum albumin/ml. After the termination of the reaction by chilling at 0°C, unlabeled ATP, ADP, and AMP (final concentration, 1.5 mM each) were added to each sample, and the sample was subjected to thin layer chromatography as described (27). The amounts of 14 C-labeled ATP, ADP, and AMP were determined with a radio-TLC analyzer (Raytest RITA 68000) or a BAS-2000 image analyzer, and the amounts of ATP hydrolyzed by the reaction were calculated. When mutant RecA proteins were tested for competition with single-stranded DNA-binding protein (SSB; purchased from Pharmacia Biotech Inc.) in the binding to single-stranded DNA, each mutant RecA protein was incubated with single-stranded DNA and 1.3 mM [ 14 C]ATP for 2 min prior to the addition of SSB, followed by the incubation at 37°C for 28 min.
Assay for Removal of the Intramolecular Base Pairing by RecA Protein-The extent of intramolecular base pairing in single-stranded DNA was estimated from an enhancement of the fluorescence emitted from ethidium bromide on the intercalation of the dye into regions of secondary structure in the single-stranded DNA (10). M13mp19 circular single-stranded DNA (2.0 M) and 60 nM ethidium bromide were dissolved in a buffer containing 31 mM Tris-HCl (pH 7.5), 13 mM MgCl 2 , 1.3 mM ATP, 1.8 mM dithiothreitol and 88 g of bovine serum albumin/ml at 37°C. The reaction was started by the addition of 0.60 M RecA protein in the presence of 4 mM phosphocreatine and 5 units of creatine phosphokinase/ml as an ATP-regenerating system. The fluorescence was recorded at 600 nm with the excitation at 530 nm. Ethidium bromide (60 nM) was present throughout the reaction, but this concentration of ethidium bromide was far below a level that causes a detectable effect on the activities of RecA protein (see Ref. 10). 4 Under these conditions, the enhancement of fluorescence is caused solely by the interaction of ethidium bromide with a double-stranded region, and the binding of RecA protein to double-stranded DNA per se did not cause a decrease in the fluorescence (see Ref. 10). 4 Assay for the Binding of RecA Protein to Single-stranded 83-mer Oligodeoxyribonucleotide-32 P-Labeled single-stranded 83-mer oligodeoxyribonucleotide (1.2 M) and the indicated amounts of RecA protein were incubated in a reaction mixture containing 33 mM PIPES-NaOH (pH 7.0), 13 mM magnesium acetate, 1.2 mM ATP␥S, 2.0 mM dithiothreitol, and 100 g of bovine serum albumin/ml at 37°C for 10 min. After the incubation, samples were diluted 10-fold with a dilution buffer containing 30 mM PIPES-NaOH (pH 7.0) and 15 mM magnesium acetate. 20-l aliquots were subjected to gel electrophoresis in 0.55% agarose in the presence of 40 mM Tris acetate (pH 7.3) and 15 mM magnesium acetate. Material that was shifted from the position of free oligodeoxyribonucleotide was quantitated by use of a PhosphorImager.
Assay for the Formation of Coaggregate by RecA Protein-M13mp19 closed circular double-stranded 3 H-labeled DNA (form I; 2.0 M) and X174 circular single-stranded DNA (2.0 M) were incubated with the indicated amounts of RecA protein at 37°C in a buffer containing 31 mM Tris-HCl (pH 7.5), 13 mM MgCl 2 , 1.3 mM ATP, 1.8 mM dithiothreitol, 88 g of bovine serum albumin/ml, and an ATP-regenerating system consisting of 4 mM phosphocreatine and 5 units of creatine phosphokinase/ ml. After 20 min of the incubation, the amount of 3 H-labeled DNA integrated into coaggregates was determined by means of a centrifugation assay (9). RecA protein was removed from DNA by treatment with 0.5% SDS, 40 mM EDTA, and proteinase K (240 g/ml), followed by incubation at 30°C for 10 min. The products of annealing were separated by gel electrophoresis in 1% agarose, and the labeled single-stranded DNA fragments that were annealed with M13mp19 circular single-stranded DNA were quantitated with a BAS-2000 image analyzer.

Construction of Mutant RecA Proteins That Have Amino Acid Substitutions for Lys or Arg Residues
We introduced amino acid substitutions in the domain consisting of amino acid residues 270 -328, a candidate for the gateway for secondary binding of double-stranded DNA to presynaptic filaments of RecA protein and single-stranded DNA. In this study, we looked at basic amino acid residues (Lys or Arg) that could directly interact with negatively charged sugar phosphate backbones of DNA. This domain contains two well conserved (among bacterial RecA proteins) basic amino acid residues (Lys 286 and Lys 302 , allowing replacement by Arg), two less conserved ones (Lys 297 and Arg 324 ), and seven nonconserved ones (25). We constructed by site-directed mutagenesis mutant RecA proteins that have single (or double, in one case) replacement of the basic amino acid residues by polar ones to minimize the change of hydrophilic environments caused by the replacements; i.e . RecAK286N, RecAK297N, RecAK302N, and RecAK317N have single substitution of Asn for Lys residues at positions 286, 297, 302, and 317, respectively. RecAR324Q has a substitution of the Gln for the Arg residue at position 324. RecAK280/282N has double substitution of Asn for two Lys residues at positions 280 and 282.
The mutant recA genes were tested for UV sensitivity in vivo (Fig. 2). In this test, the wild-type recA gene and the mutant alleles were expressed exclusively from a multicopy plasmid under the control of the tac promoter (pKK223-3). Thus surviving cell fraction curve). RecA proteins with the mutations at conserved Lys residues (RecAK286N or RecAK302N) were 20-and 11-fold, respectively, less effective than wild-type RecA protein, indicating significant deficiencies in repair of UV-irradiated DNA caused by the mutations (Fig. 2). Single replacement of less conserved Lys 297 (RecAK297N) caused some sensitivity to UV irradiation compared with wild-type protein (7fold; Fig. 2). Replacement of nonconserved Lys 317 (RecAK317N) and less conserved Arg 324 (RecAR324Q) caused no significant defect in UV sensitivity (Fig. 2). Double replacement of nonconserved Lys 280 and Lys 282 caused a slight defect (4-fold).
Each mutant RecA protein was overexpressed in ⌬recA cells that have a mutant recA gene on a multicopy plasmid under the control of the induced tac promoter and extensively purified. RecAR324Q was unstable; it lost activity during purification and was thus excluded from this study.

Homologous Pairing of Single-stranded DNA with Doublestranded DNA by the Mutant RecA Proteins
Formation of D loops-RecA protein promotes joint molecule formation in vitro through ATP-dependent homologous pairing of various combinations of single-stranded and doublestranded DNA. Some other proteins unrelated to RecA protein also promote joint molecule formation from linear doublestranded and single-stranded DNA, but they promote it through ATP-independent reactions (35)(36)(37)(38)(39)(40). As far as we know, joint molecule formation from closed circular doublestranded DNA and single-stranded DNA fragments (D loop formation) is catalyzed only by RecA protein or its analogs through an ATP-dependent reaction (17). 5 The only exceptions are RecO protein of E. coli and a 20-amino acid peptide derived from RecA protein that were reported to promote ATP-independent homologous pairing of single-stranded DNA and superhelical double-stranded DNA in vitro (17,41). Thus, this combination of substrates is suitable to characterize RecA protein-specific homologous pairing. We examined the ability of the mutant RecA proteins to promote homologous pairing of 35 S-labeled single-stranded DNA fragments and unlabeled negatively superhelical closed circular double-stranded DNA (form I DNA) to form D loops (Fig. 3). The D loops produced by these reactions were analyzed by agarose gel electrophoresis ( Fig. 3; see Ref. 2). Under the conditions used in the current study, wild-type RecA protein increased the amount of D loops to the maximum level after 5-10 min of incubation and then decreased (Fig. 3F). The decrease of D loops is an ATP hydrolysisdependent reaction promoted by RecA protein and was suggested to be a variation of strand exchange (23,42,43).
The amounts of D loops at 10 min of incubation with various amounts of wild-type and mutant RecA proteins are shown in Fig. 3, A-E, and those after various times of incubation are shown in Fig. 3F. The residual activities of RecAK286N in promoting D loop formation were 3-10% and 10% of the wildtype level when compared at a limiting amount (1 M) and at an excessive amount (3.0 M) of protein, respectively (Fig. 3, A and  F), and those of RecAK302N were 34 -44% and 41%, respectively (Fig. 3, C and F). On the other hand, the residual activities of RecAK280/282N were 58 and 69%, respectively, but unlike the cases of RecAK286N and RecAK302N, the amount of D loops appeared to approach the wild-type level with the addition of more protein (Fig. 3B). RecAK297N and RecAK317N are fully active (Fig. 3, D and B). Thus, the replacement of the conserved Lys residues caused significant defects in promoting homologous pairing, whereas the single replacement of less conserved or nonconserved Lys or Arg residue caused no deficiencies in D loop formation. Double replacement of nonconserved Lys residues showed some additive effects of slight deficiencies. formed by the mutant RecA proteins after incubation for 30 min. As shown in Fig. 4A, RecAK286N and RecAK302N showed a clear defect in the formation of the joint molecules.
We did a similar test by use of a double-stranded oligodeoxyribonucleotide and circular single-stranded DNA. The reactions involving an ordinary duplex oligodeoxyribonucleotide are shown in Fig. 4B, and those involving a proximal hairpin oligodeoxyribonucleotide are shown in Fig. 4C. In the former case, the assay for strand exchange measured release of the labeled plus strand after deproteinization. In the reaction involving an ordinary duplex oligodeoxyribonucleotide, both RecAK286N and RecAK302N showed the same results as in the case of D loop formation shown in Fig. 3. In the reaction involving a proximal hairpin oligodeoxyribonucleotide, triplex joints formed by RecA protein are stable following deproteinization (29). RecAK286N and RecAK302N showed the same level of strong deficiencies (Fig. 4B). We did not study further the variation in the deficiency of RecAK302N caused by changing DNA substrates.

The Mutant RecA Proteins had no Deficiency in Unfolding of Single-stranded DNA by the Formation of Presynaptic Filaments
Next, we looked for steps leading to defective D loop formation mediated by mutant RecA proteins, especially RecAK286N and RecAK302N. In homologous pairing, RecA protein first binds to single-stranded DNA in the presence of ATP and forms presynaptic filaments. Since presynaptic filaments are the first intermediates for homologous pairing, defects of mutant RecA proteins in this step would affect the subsequent pairing steps.
First, we tested the ability of the mutant RecA proteins to hydrolyze ATP in the presence of excess single-stranded DNA (44,45), since the binding of ATP and subsequent activation of the protein are crucial for RecA protein to promote homologous pairing and strand exchange. None of the five mutant RecA proteins isolated in this study showed defects in singlestranded DNA-dependent ATP hydrolysis (Fig. 5A). These results indicate that the mutant RecA proteins are proficient in activation of the protein through binding of both singlestranded DNA and ATP.
The most sensitive assay for deficiencies in binding of mutant RecA protein to single-stranded DNA would be an assay using competition by SSB. When SSB is added to a reaction mixture before the addition of RecA protein, SSB completely inhibits the binding of RecA protein to single-stranded DNA. On the other hand, SSB added after RecA protein is preincubated with single-stranded DNA does not inhibit binding of RecA protein to single-stranded DNA. Instead, it stimulates various activities of RecA protein, especially when singlestranded DNA is limited (46 -51). Therefore, we tested competition by SSB in single-stranded DNA binding of mutant RecA proteins by monitoring single-stranded DNA-dependent ATP hydrolysis. We incubated each mutant RecA protein, while and then added sufficient SSB to the reaction, followed by incubation at 37°C and measurements of the ADP formed. In a control experiment, the SSB preparation used in this study completely inhibited the single-stranded DNA-dependent ATP hydrolysis by wild-type RecA protein when added before the addition of RecA protein (data not shown). In the case of wildtype RecA protein, SSB inhibited ATP hydrolysis when the amounts of RecA protein were limited but stimulated it when RecA protein was added in excess (Fig. 5E). RecAK297N,  RecA317N, and RecAK280/282N showed almost the same behavior as wild-type RecA protein in this test (Fig. 5C and data not shown). ATP hydrolysis by RecAK286N was partially inhibited by SSB even when the mutant protein was added in excess (Fig. 5B). RecAK302N was more sensitive to the inhibitory effects of SSB than wild-type protein, but the inhibition was overcome by addition of 2-fold more mutant protein than wild-type protein (Fig. 5, D versus E). This test assesses the overall ability of RecA protein to bind to single-stranded DNA but does not tell us which feature of the binding is affected by the mutation.
Next, we examined a specific feature of the formation of presynaptic filaments by the mutant RecA proteins. The secondary structure of single-stranded DNA prevents homologous pairing. An important role of presynaptic filament formation is the removal of the secondary structure of single-stranded DNA (unfolding; Refs. 5, 6, 47, and 50) and the extension of singlestranded DNA to ϳ1.5 times the length of B form DNA (6). To assess the ability of the mutant RecA proteins to unfold singlestranded DNA, we measured the fluorescence emitted by intercalated ethidium bromide in intramolecular double-stranded regions of phage M13mp19 circular single-stranded DNA. When RecA protein unfolds base pairing, intercalated ethidium bromide is removed, resulting in a decrease in fluorescence emitted (see Ref. 10). 4 This assay is sensitive enough to detect a partial defect of RecA430 protein (a mutation in loop 2, Gly 204 is replaced by Ser; Ref. 52) under standard conditions for homologous pairing (without both SSB and salt); i.e. the maximum level of unfolding accomplished by RecA430 was 85% of that by wild-type RecA protein even at saturating levels (data not shown). The significance of this result will be described under "Discussion." We tested the newly isolated mutant RecA proteins for unfolding activity. As shown in Fig. 6, none of the five mutant RecA proteins tested showed detectable deficiency in unfolding of intramolecular base pairs of single-stranded DNA.
We also tested RecAK286N, RecAK302N, and RecAK297N for binding to a single-stranded 83-mer oligodeoxyribonucleotide. RecA protein binds to the oligomer much less stably than to large single-stranded DNAs such as M13 phage DNA. None of the mutant RecA protein showed detectable defects in binding to the oligomer (data not shown).

The Non-sequence-specific Binding of Duplex DNA to Presynaptic Filaments
Following the formation of presynaptic filaments, two sequential stages lead to homologous pairing: the non-sequencespecific binding of double-stranded DNA to the filaments (secondary binding) and a search for homology between two DNA molecules in the three-component complex formed by secondary binding. We examined secondary binding of doublestranded DNA to presynaptic filaments by assaying for the formation of "coaggregates" (8,9). Since coaggregates are detected by low speed centrifugation, the level of signals obtained by this assay depends on the size of aggregates, and thus, active but small coaggregates formed by partially defective mutant RecA protein might not give a signal in this assay. Thus, it should be noted that the absence of a signal indicates a defect in secondary binding but does not indicate the complete deficiency.
The results are shown in Fig. 7. For RecAK286N and RecAK302N no signal for the formation of coaggregates was detected, indicating that these mutant RecA proteins are significantly defective in secondary double-stranded DNA binding. In the case of RecAK280/282N, about 2-fold more protein was required for the same level of signals as wild-type protein.
Considering the characteristics of the assay, RecAK280/282N has a slight defect in secondary binding.

The Recognition of Complementary Sequences
Presynaptic filaments containing single-stranded DNA interact with a naked complementary single strands to form double-stranded DNA ("renaturation of complementary strands"; Refs. 53-55). We tested the ability of RecAK286N, RecAK302N, and RecAK297N to promote duplex formation from complementary single-stranded DNAs. The results showed that RecAK302N and RecAK297N are proficient in renaturation activity (Fig. 8, C and B, respectively). On the other hand, RecAK286N showed partial but significant deficiencies in renaturation (Fig. 8A). The deficiency of RecAK286N in renaturation is in contrast to RecAc38 and RecAK302N, which are proficient (Ref. 10 and Fig. 8), indicating that RecA protein-mediated (i.e. not spontaneous) recognition of the complementarity between single-stranded DNA in presynaptic filaments and a strand of duplex DNA is a part of the mechanism for homologous pairing. DISCUSSION In this study, we characterized two mutant RecA proteins, RecAK286N and RecAK302N, which have a single substitution for either of two well conserved basic amino acid residues in the domain consisting of residues 270 -328 near the C terminus of the RecA polypeptide. As controls, we fully or partially characterized three other mutant RecA proteins: RecAK297N, which has a single substitution for a less-conserved Lys residue in the domain; and RecAK317N and RecAK280/282N, which have single or double substitutions for nonconserved Lys residues. RecAK286N and RecAK302N lose 90% and more than one-half, respectively, of the wild-type activity for homologous pairing of double-stranded DNA with single-stranded DNA and are significantly defective in secondary binding of double-stranded DNA to the RecA single-stranded DNA filaments, without detectable defects in unfolding of single-stranded DNA by presynaptic filament formation and ATP hydrolysis. RecAK280/282N partially loses (about one-third) homologous pairing activity and secondary DNA binding, but both defects were overcome by adding larger amounts of protein. On the other hand, RecAK297N and RecAK317N retain full activity for homologous pairing. These results indicate that the effect of replacement of either Lys 286 or Lys 302 is not a general effect of a loss in the positive charge of the domain but is specific to each substitution.
Functions of the C-Terminal Domain of the RecA Protein-Since the C-terminal domain (residues 270 -328) of the RecA protein was shown to be involved in filament-filament interactions by an x-ray crystallographic study (13) and by a mutational study (56), a possible explanation for defective D loop formation by RecAK302N and RecAK286N would be that these mutations stabilize bundles of RecA filaments assumed to be an inactive storage form (13), resulting in the prevention of both activation of the protein and secondary binding of doublestranded DNA. If this explanation is correct, one can predict that: (i) the mutations cause reduction in single-stranded DNAdependent ATPase activity of the RecA protein; and (ii) nucleoprotein filaments of these mutant RecA proteins are also defective in their interactions with external single-stranded DNA. However, our results are inconsistent with this prediction: (i) both RecAK286N and RecAK302N are proficient in single-stranded DNA-dependent ATPase activity (Fig. 5A); and (ii) RecAK302N is proficient in renaturation with external complementary single-stranded DNA. Thus, stabilized bundle formation is unlikely to be a major cause of defective D loop formation, especially in the case of RecAK302N but also in the case of RecAK286N.
RecAK297N showed a deficiency in repair of UV-irradiated DNA in vivo but no defects in vitro. We assume that the in vivo deficiency would be caused by an activity of the RecA protein not involved in homologous pairing, such as coprotease activity, but we did not study this possibility further.
Gateway for Secondary DNA Binding-Previous electron microscopic studies showed that the RecA protein can include three strands within the filament (57,58), and one can assume that sites for both single-stranded and double-stranded DNA binding are located inside the RecA filaments. On the other hand, when one supposes an initial contact of double-stranded DNA with presynaptic filaments in which single-stranded DNA is located inside the RecA filaments (18,19), one would assume that RecA filaments have a gateway through which the doublestranded DNA finds a path for direct contact with the singlestranded DNA to search for homology (see the Introduction). Double-stranded DNA would be first trapped by amino acid residues that are located outside the filament and at the gateway, and then be led into the filament by successive binding to residues facing the gateway and to those inside the filament. The major aim of this study was to explore amino acid residues involved in the gateway by assuming further that mutations of such amino acid residues result in deficiencies in secondary double-stranded DNA binding to presynaptic filaments with little or no defect in primary binding of the protein to singlestranded DNA.
Since both RecAK302N and RecAK286N are partially defective in competition with SSB for binding to single-stranded DNA (Fig. 5, D and B), and since secondary binding has been shown to be dependent on the ATP-dependent primary singlestranded DNA binding (3,8,9), one might explain the deficiency of these mutant RecA proteins in homologous pairing by enhanced defects in secondary DNA binding induced by a minor defect in primary binding to single-stranded DNA. We cannot exclude this possibility, but we assume that the deficiency in primary DNA binding is not a major cause of the deficiency in homologous pairing by the mutant RecA proteins for two reasons. First, although the binding of RecA430 protein to single-stranded DNA is completely inhibited by SSB even when SSB is added after preincubating the mutant protein with single-stranded DNA (59), the deficiency of RecA430 protein in homologous pairing is only partial; the defects were recovered by adding approximately 4-fold more mutant protein than wild-type protein (60). On the other hand, RecA430 showed a partial deficiency in unfolding single-stranded DNA (this study). Thus, although the partial deficiency in unfolding by RecA430 could be a cause of the partial deficiency in homologous pairing by this mutant protein, it is unlikely that the partial deficiencies of mutant proteins detected by competition with SSB are major causes of the deficiencies of RecAK286N and RecAK302N in homologous pairing. Second, partial defects in primary single-stranded binding of a mutant RecA protein are expected to be overcome by adding larger amounts of the protein, as is the case for RecA430, as described above. However, addition of excess RecAK286N or RecAK302N did not overcome their deficiency in homologous pairing (Fig. 3).
Although RecAK286N and RecAK302N are proficient in unfolding of single-stranded DNA (Fig. 6), single-stranded DNAdependent ATP hydrolysis (Fig. 5A) and binding to a singlestranded 83-mer oligodeoxyribonucleotide (see "Results"), both are defective in non-sequence-specific binding of doublestranded DNA by presynaptic filaments. Considering this result and the above discussion, we assume that a deficiency in secondary double-stranded DNA binding is a major cause of defective homologous pairing by RecAK286N and RecAK302N.
RecAK302N and RecAK286N have substitutions for amino acid residues near the opening of a cleft between adjacent RecA monomers ( Fig. 1), which were revealed by an x-ray crystallographic study of free and ADP-bound RecA protein (13). Thus, the locations of these two amino acid residues and biochemical characteristics of these mutant RecA proteins are consistent with the assumption that this cleft is a gateway for the binding of double-stranded DNA, and also that Lys 286 and Lys 302 have a direct role in secondary double-stranded DNA binding.
Since the structure of the RecA nucleoprotein filaments has not been solved yet, the model presented here is based on the crystal structure of RecA protein alone. Moreover, we understand that the mutational approach used in this study is not sufficient to discriminate between direct roles and indirect roles for a particular residue in DNA binding. Structural studies on wild-type or mutant RecA proteins and their DNA complexes are required for evaluation of the effects of each amino acid substitution and for the identification of the essential interactions between amino acid residues of the RecA polypeptide and DNA in the search for homology between singlestranded and double-stranded DNA. We are currently trying to test this model by use of nuclear magnetic resonance spectroscopy. Preliminary results of this approach suggest that amino acid residues near Lys 302 directly interact with doublestranded DNA.