Cloning, Overexpression, Purification, and Characterization of the Escherichia coli RuvC Holliday Junction Resolvase*

The ruvC gene has been cloned into the plasmid pT7-7 under the control of the T7 410 promoter. Following induction with isopropyl-1-thio-P-D-galactopyranoside, the 19-kDa RuvC protein was overexpressed to 2030% of total cell protein. RuvC has been purified to homogene-ity by a simple procedure involving precipitation from the crude lysate, followed by three chromatographic steps. The purified protein resolves synthetic Holliday junctions (60 nucleotides in length) by cleavage at the 3’-side of a phosphate group, to produce nicked duplex DNA. Under the same conditions no cleavage of linear duplex or single-stranded DNA was detected. However, low levels of cleavage were observed with supercoiled form I and single-stranded circular DNA substrates, consistent with the interaction of RuvC with secondary structures. Using synthetic Holliday junctions, we show that RuvC-mediated resolution requires Mg2‘ (10 mm) and exhibits an alkaline pH optimum (pH 9.0). No energy cofactors are needed. When RuvC was analyzed by gel filtration and polyacrylamide gel electrophoresis, monomeric and dimeric forms of the protein were observed. Homologous recombination occurs via an intermediate structure, known as a Holliday

Homologous recombination occurs via an intermediate structure, known as a Holliday junction, in which two DNA molecules are linked by a crossover (1). To complete the recombination event, resolution of the Holliday junction is required to restore the DNA to two discrete molecules. In Escherichia coli, an activity capable of performing Holliday junction resolution in vitro was identified using cell-free extracts (21, and this activity was later found to be absent in ruvC mutants (3).
The ruvC gene is located within the ruv locus, at 41 min on the E. coli chromosome (4-6). The gene forms an operon with orfZ6, which encodes a 26-kDa protein of unknown function. A second operon lying downstream of ruvC encodes two genes, ruvA and ruvB (7,8). The ruvA and ruvB genes are L e dregulated and induced as part of the SOS response to DNA damage (7)(8)(9)(10). Cells carrying mutations in one of the three ruu genes have similar phenotypes, with an increased sensitivity to UV light, ionizing irradiation, and chemical mutagens (4, [11][12][13][14]. In addition, in a recBCsbcA, recBCsbcBC, or recG genetic background, ruu mutants are deficient in homologous recombination (11,(15)(16)(17)(18). In certain cases, the effect of ruu on the phenotype of these multiple mutants is suppressed by recA mutations (151, suggesting that the ruuA, ruvB, a n d ruuC gene products are involved in a late step of recombination and the recombinational repair of damaged DNA.
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
5 To whom correspondence should be addressed.
The RuvA and RuvB proteins have been purified (19-21) a n d shown to promote the branch migration of Holliday junctions in vitro (20,(22)(23)(24)(25)(26). The product of ruvC, the 19-kDa RuvC protein, resolves Holliday junctions in vitro (27). Resolution was demonstrated using: (i) synthetic Holliday junctions; (ii) recombination intermediates made by the RecA protein; and (iii) cruciform structures extruded from supercoiled plasmids (27,281. Cleavage occurred by the introduction of two symmetrical nicks close to the junction point (3,27,29). In the present work, we describe the cloning of the ruvC gene into a high level expression plasmid and detail the purification of RuvC protein. We investigate the substrate specificity and cofactor requirements of the cleavage reaction catalyzed by RuvC and report the general biochemical and physical properties of the purified protein.

Enzymes and Reagents
Restriction enzymes were obtained from New England Biolabs, Am-pliTaq DNA po-lymerase from Perkin-Elmer-Cetus Instruments, calf intestinal phosphatase from Boehringer Mannheim, and T4 polynucleotide kinase from Pharmacia LKB Biotechnology Inc. T4 endonuclease VI1 (30) was a gift of Dr Bonies Kemper (University of Cologne). Terminal transferase, [U-~~PIATP, and [y-32PlATP were obtained from Amersham Corp.

Polymerase Chain Reaction
A standard reaction contained oligonucleotides A and B (100 pmol of each molecule), pGS760 (60 ng), and AmpliTaq DNA polymerase (5 units). 30 cycles were set up as follows: 40 s denaturing at 92 "C, 1 min of annealing at 45 "C, and 2.5 min of extension at 72 "C. The product was isolated by gel purification (Geneclean 11, Stratech).

DNA Substrates
The synthetic Holliday junction (4-Xl2) was prepared by annealing four partially complementary oligonucleotides (each approximately 60 nucleotides in length). Duplex DNA was prepared by annealing two complementary oligonucleotides. Annealing was performed as described (331, and the oligonucleotide sequences are listed in the accompanying paper (34). The annealed substrates were 3ZP-end labeled in the common strand (oligonucleotide 2). Labeling was performed using T4 polynucleotide kinase and [y-32PlATP to give a 5'-end label or terminal transferase and [~~-~~Pldideoxy ATP to give a 3'-end label. Singlestranded and replicative form I (RFI)' DNA of 4x174 were purchased from New England Biolabs. 3H-Labeled single-stranded 4x174 DNA was prepared as described (35). Unless stated otherwise amounts of DNA are expressed in mol of nucleotide residues.

Time Course of Induction of RuvC Expression
A 200-ml culture of BL21 (DE3) pLysS pGS775 was grown with aeration at 37 "C in Luna broth containing 100 pg/ml carbenicillin and 50 p g / d chloramphenicol. At a cell density corresponding to A,, = 0.5, synthesis of RuvC protein was induced by the addition of isopropyl-lthio-p-D-galactopyranoside (IFTG) to 0.4 m~. At various times, 1.5-ml samples were removed and the cells pelleted by centrifugation. The cells were resuspended in 100 pl of SDS-sample loading buffer and boiled for 3 min before analysis by SDS-PAGE.

Chromatography
Reactive Blue 4-agarose and single-stranded DNA-cellulose were obtained from Sigma and phosphocellulose (P11) from Whatman. A flow rate of 30 mVh was used with Reactive Blue 4-agarose and phosphocellulose columns, and 20 mVh for the single-stranded DNA-cellulose column. All column chromatography was carried out at 4 "C.

Protein Concentrations
All protein concentrations were determined using a protein assay kit (Bio-Rad) with bovine serum albumin as standard. Amounts of RuvC are expressed in moles of monomeric protein.

Resolution Assay
Reactions (20 pl) containing 5'-32P-labeled junction DNA (150 n~) , cleavage buffer, and RuvC protein were incubated for 30 min at 37 "C. They were stopped by the addition of EDTA to 25 m~, and the DNA was analyzed by neutral or denaturing PAGE followed by autoradiography.

Nonspecific Endonuclease Assays
Single-stranded or RFI 4x174 DNA (76 p) was incubated in cleavage buffer (40 pl) at 37 "C for 15 or 30 min, respectively, in the presence of various amounts of RuvC. Reactions were stopped by the addition of EDTA to 25 m~, and the DNA was analyzed by electrophoresis through 1% agarose.
Analysis of Cleavage %-mini J'-ll?rmini-Reactions (60 pl) containing 5'-32P-end-labeled 4-X12 DNA (approximately 0.15 p) and RuvC (0.67 pd or T4 endonuclease VI1 (300 units) were incubated in cleavage buffer for 30 min at 37 "C. The DNA products were denatured by heating to 100 "C for 2 min and the reactions split into two 30-pl aliquots. To each aliquot we added terminal transferase buffer (solution 1, Amersham 3'-end labeling kit). After a 20-min incubation at room temperature in the presence or absence of terminal transferase (10 units) and 2 m~ dATP (total volume of 50 pl), reactions were stopped by the addition of EDTA to 20 m~, and the 32P-labeled products were analyzed by denaturing PAGE.
Agarose gel electrophoresis was performed using TAE buffer and the DNA visualized by staining with ethidium bromide. For neutral PAGE, samples were electrophoresed through 6% polyacrylamide gels. For denaturing PAGE, the DNA products were heated to 95 "C for 3 min in TBE buffer containing 90% (v/v) formamide. Denatured samples were subjected to electrophoresis through 12% polyacrylamide gels containing 7 M urea. In both cases TBE buffer was used. The DNA was visualized by autoradiography.

RESULTS
Construction of RuvC Overexpression Plasmid-In previous studies, we purified RuvC protein from plasmid pGS762 in which the ruvC gene was under control of the lac promoter (27).
Induction of E. coli FB800 carrying pGS762 by IPTG led to low level overexpression of RuvC protein ( 3 4 % of total cellular protein). To facilitate greater overexpression of RuvC, the ruvC gene was cloned into the plasmid pT7-7. A clone was generated by site-directed mutagenesis of ruvC, in which the upstream and downstream sequences were altered. Two oligonucleotides were constructed as primers for a polymerase chain reaction (Fig. 1B). Oligonucleotide A corresponds to 34 bases of the 5' sequence of ruvC. It differs from the wild-type sequence at 6 residues giving an EcoRI restriction site and an improved ribosome binding site. Oligonucleotide B corresponds to 22 bases of the 3' sequence of ruvC and differs from the wild-type sequence at 2 residues giving an HindIII site. Using the two oligonucleotides a polymerase chain reaction product was generated from the plasmid pGS760 (ruvC+) (5). The product was digested with EcoRI and HindIII and inserted into the plasmid pT7-7 (32). The resulting plasmid, pGS775, carries the ruvC gene under the control of the T7 410 promoter (Fig. lA) and is able to restore complete UV resistance to the four available ruvC mutants. Confirmation that pGS775 carried the wild-type ruvC gene was obtained by DNA sequencing. 2 Plasmid pGS775 was transformed into BL21 (DE31 pLysS. A time course of the induction of pGS775 following IPTG treatment is shown in Fig. 1C. Overexpression of the 19-kDa RuvC protein was apparent after 60 min of induction ( Purification of RuvC Protein-A summary of the purification procedure is shown in Fig. 2 A . Two 1-liter cultures of E. coli strain BL21 (DE3) pLysS carrying pGS775 (ruuC') were grow, with aeration, at 37 "C in Luna broth containing 100 pdml carbenicillin and 50 pg/ml chloramphenicol. At a cell density corresponding to Asso = 0.5, IPTG was added to 0.4 m~ and incubation continued for 3.5 h. After chilling on ice, induced cells ( Fig. 2   essary to prevent precipitation of RuvC. Cells were lysed by three rounds of freezing and thawing, and the lysate was centrifuged for 60 min at 38,000 rpm in a Beckman Ti-45 rotor. The clear supernatant (36 ml) was dialyzed for 3 h against R buffer supplemented with 0.1 M KCI. During dialysis a heavy precipitate formed. The precipitate was collected by centrifugation at 15,000 rpm for 10 min, washed with R buffer containing 0.1 M KCI, and recentrifuged. The pellet was resuspended by addition of 90 ml of R buffer containing 0.5 M KCl. SDS-PAGE of the resuspended precipitate (fraction I) showed it was more than 90% pure (Fig. 2 B , lane d ) . This fraction is sufficiently pure for physical analysis of RuvC.
To remove minor contaminants and obtain protein for biochemical analysis, RuvC was purified further by three chromatographic steps. Fraction I (90 ml; approximately 0.6 mg of proteidml) was applied to a Reactive Blue 4-agarose column (1.6 x 14.0 cm, 28-ml bed volume) equilibrated with R buffer containing 0.5 M KCI. The column was eluted with a 400-ml gradient of 0.5-1.75 M KC1 in R b~f f e r .~ Fractions eluted from the column were assayed in two ways; (i) the peak of RuvC protein was identified by SDS-PAGE or (ii) the peak of resolvase activity was determined using 32P-labeled synthetic Holliday junctions. Resolution resulted in the formation of nicked duplex DNA, as detected by neutral or denaturing For the purification described here the lyophilized powder form of Reactive Blue 4-agarose (Sigma, R8754) was used. In subsequent purifications, the matrix was purchased as a suspension (Sigma, R2507) to which RuvC binds with a lower affinity. This makes it necessary to dialyze fraction I against R buffer containing 0.3 m~ KC1 before loading and use a gradient of 0.3-1.25 M KC1 in R buffer. PAGE followed by autoradiography (29).
The peak of RuvC protein eluting from Reactive Blue 4-agarose (0.6-0.7 M KCI) was pooled and dialyzed against R buffer containing 0.2 M KC1 for 3 h (fraction 11). At this stage, RuvC appeared as a single band by SDS-PAGE (Fig. 2, lane e ) . However, the presence of contaminating nuclease activity was revealed by a loss of end label during the resolution assay (data not shown). Fraction I1 (50 ml; approximately 0.5 mg of proteid ml) was then applied to a phosphocellulose column (1.0 x 12.7 cm, 10-ml bed volume) and eluted with a 200-ml gradient of 0.2-1.0 M KC1 in R buffer. SDS-PAGE revealed a peak of protein in fractions 14-17 (Fig. 3A, lanes h-k) which coeluted with the peak of resolvase activity (Fig. 3B, lanes t-w). Although the protein appeared homogeneous a t this stage, overexposure of the autoradiogram shown in Fig. 3B revealed trace amounts of a nonspecific nuclease activity in fractions 16-19, overlapping the RuvC peak. Rather than accept a considerable loss of RuvC by discarding the contaminated fractions, a third chromatographic step was used to remove the nonspecific nuclease.
The peak fractions from phosphocellulose (eluting between 0.5-0.7 M KCI) were pooled and glycerol added to 50% to allow overnight storage at -20 "C without freezing. The following day the glycerol was removed by dialysis for 3 h against R buffer containing 50 mM KCI, giving fraction I11 (Fig. 2B, Fig. B, lane g). Fraction IV was homogeneous as determined by SDS-PAGE followed by silver staining (Fig. 2C, lane i ) . After fast freezing on dry ice/ethanol the protein was stored in aliquots a t -70 "C. Working stocks were stored a t -20 "C. The total yield of RuvC protein was 6.3 mg in 10 ml. The pure RuvC appears to be unusually stable. After incubation at 37 "C for 14 h, we were unable to detect degradation of the protein by SDS-PAGE, and there was little or no loss of activity when tested in the resolution assay (data not shown).
Substrate Specificity of Purified RuvC-To determine whether RuvC possesses any nonspecific nuclease activity, various amounts of the protein were incubated with supercoiled RFI 4x174 duplex DNA. Reaction products were analyzed by agarose gel electrophoresis followed by ethidium bromide staining. At RuvC concentrations of up to 1.4 p~ (i.e. 1 pg of RuvC/pg of RFI DNA) we were unable to detect cleavage of the substrate (Fig. 4A, lunes b d ) . However, at a RuvC concentration of 4.2 PM, which conesponds to 1 RuvC monomer/l8 nucleotides, we observed the appearance of trace amounts of linear DNA product (Fig. 4A, lune e).
To determine whether RuvC protein possesses singlestranded endonuclease activity, varying amounts of RuvC were incubated with single-stranded circular 4x174 DNA. This DNA substrate provides a sensitive assay for endonuclease activity since the introduction of two nicks/DNA molecule (5,382  (lanes a d and m-p), 19  m linear duplex DNA (lanes e 4 and q-t), or 9.5 m single-stranded linear DNA (lanes i-l and u-x) were incubated with the indicated amounts of RuvC. 5'-32P-End-labeled oligonucleotide 2 was common to all three substrates. After 60 min at 37 "C, the reactions were stopped with EDTA and the labeled products visualized by autoradiography following neutral (panel A ) or denaturing (panel B ) PAGE. nucleotides) results in loss of the substrate band as detected by agarose gel electrophoresis followed by ethidium bromide staining. At RuvC concentrations below 0.5 PM, we observed no detectable cleavage of the substrate DNA (Fig. 4B, lanes i d ) . However, at concentrations above 0.8 p (? 1 RuvC monomer/95 nucleotides), we observed loss of the substrate, characteristic of single-stranded DNA endonuclease activity (Fig. 4B, lanes 1-n). Since the gel assay is incapable of determining the extent of DNA digestion, reactions containing uniformly 3H-labeled single-stranded 4x174 DNA were analyzed for the release of acid-soluble counts following incubation with RuvC. At concentrations of RuvC which promoted the complete loss of the single-stranded circular DNA, as measured by gel electrophoresis (Fig. 4B, lanes m-n), we observed that less than 0.1% of the total 3H-labeled DNA was made acid-soluble (data not shown). These experiments indicate that RuvC does not possess any exonuclease activity and interacts with the singlestranded circular DNA to introduce a small number of endonucleolytic incisions.
To investigate the specificity of the endonuclease activity of RuvC protein, we compared its activity on a synthetic Holliday junction (4-X12), linear duplex DNA and a single-stranded oligonucleotide (Fig. 5). All substrates were of the same length and contained a common DNA strand (5'-32P-end labeled oligonucleotide 2). Oligonucleotide 2 was designed such that it possessed no internal complementary sequences and so could not form any secondary structure. Reactions contained an equal number of substrate molecules, and increasing amounts of RuvC. The DNA products were analyzed by both neutral (Fig. 5 A ) and denaturing (Fig. 5B) PAGE. By neutral PAGE, we observed that RuvC cleaved the junction to give products which migrated in the duplex position (Fig. 5A, lanes c and d ) . Denaturing PAGE of the same samples revealed that this occurred by specific nicking of the DNA resulting in a single labeled cleavage fragment (Fig. 5B, lanes o and p ) . Under identical conditions and RuvC concentrations, we were unable to detect cleavage or degradation of the duplex or single-stranded DNA by neutral or denaturing PAGE (Fig. 5, lanes e 4 and q-x). Fig. 5 indicates that RuvC has a high specificity for the junction DNA, with little or no activity on double-or singlestranded DNA. The apparent incongruity of these results with those presented in Fig. 4 can be explained in terms of secondary structure. Previously, Iwasaki et a1 (28) showed that RuvC cleaves the cruciform structure in pUC4 by the introduction of nicks at the base of the cruciform to produce linear duplex DNA. Since RFI 4x174 DNA contains a palindromic sequence that can extrude into a cruciform (37), the small amount of RFI 4x174 DNA cleaved may be a result of interaction of RuvC with this structure. Cleavage of single-stranded circular 6x174 DNA could also be structure-specific since interactions between regions of complementary DNA sequence could lead to the formation of junction-like structures.
Buffer Requirements of RuuC--To determine the buffer requirements of RuvC, cleavage of the synthetic Holliday junction was assayed under various reaction conditions. The results are shown in Fig. 6. RuvC activity was stimulated by alkaline pH, with optimum resolution occurring at pH 9.0 (Fig. 6A). The amount of cleavage obtained at more physiological pH (7.5) was about 17% of that observed at the optimum. RuvC showed a requirement for divalent cations (Mg2+) with an optimum at 5-10 mM (Fig. 6B). M$+ could be replaced by Mn2+ but not by Cu2+, Zn2+, Co2+, or Ca2+.4 Resolvase activity was inhibited by salt, with 50% inhibition observed at 85 m M KC1 (Fig. 6C). We failed to observe any stimulation of activity by inclusion ofATP, spermidine or potassium glutamate (data not shown).
Characterization of the Nick Introduced by RuuC--To determine the nature of the 3"terminal group created by RuvC a t the site of resolution, the cleavage products of a 5'-32P-end labeled junction were heat denatured and treated with terminal transferase. Terminal transferase catalyzes the addition of deoxynucleotides to the 3'-end of DNA substrates only if a hydroxyl group is present. Upon incubation of the cleavage products (Fig. 7A, lane c ) with terminal transferase, a ladder of DNA fragments was produced (lane d ) , demonstrating the presence of a 3"hydroxyl group at the site of RuvC resolution. As a control, similar reactions were carried out using cleavage fragments produced by T4 endonuclease VII. This enzyme cleaved 4-X12 DNA at a number of sites throughout the homologous core to give three major and several minor products (Fig. 7A, lane e). Treatment with terminal transferase again resulted in a ladder of DNA fragments (Fig. 7A, lane f ).
To determine the nature of the group at the 5'-terminus of the nick, the junction was labeled at the 3'-end of oligonucleotide 2. After incubation with RuvC, the DNA products were heat denatured and treated with calf intestinal phosphatase, which removes 5"phosphate groups from DNA. Phosphatase treatment of a DNAmolecule possessing a 5"phosphate should result in a shift of mobility, as detected by denaturing PAGE, because of the removal of a charged group. As seen in Fig. 7B, treatment of the resolution products with calf intestinal phosphatase resulted in a slight reduction in the mobility of the R. Shah, R. J. Bennett, and S. C. West, unpublished observations.  h and j ) . To confirm that this mobility shift was caused by the loss of a phosphate group, the DNA was treated further with T4 polynucleotide kinase to restore the terminal phosphate. The migration of the DNA fragment returned to the original position (Fig. 7B, lane K).
These results indicate that RuvC cleaves DNA at the 3'-side of the phosphate group to generate a nick with 5"phosphate and 3'-hydroxyl termini. Elsewhere, we show that the nicked duplex resolution products formed by RuvC can be ligated efficiently by E. coli DNA ligase (29).
Physical Analysis of RuuC-To determine the native molecular weight of RuvC, approximately 30 pg of purified protein was applied to a gel filtration fast protein liquid chromatography column in R buffer supplemented with 10 ~l l~ MgC12 and 150 mM NaCl. Measurement of the absorbance at 280 nm resulted in the elution profile seen in Fig. 8 A . By comparison with protein standards analyzed in parallel (Fig. 8B), it was determined that the protein peak corresponded to a molecular mass of around 14 kDa. Upon SDS-PAGE of fractions collected during gel filtration, we observed that the elution of RuvC protein coincided exactly with the peak seen in the absorbance profile (data not shown). Since the predicted mass of RuvC is 18.7 kDa (5, 61, the gel filtration data indicate that RuvC is monomeric under these conditions. SDS-PAGE analysis of RuvC led to the observation of multiple bands on a gel, as seen in Fig. 9. The three major bands have been termed I, 11, and 111. When denatured by dilution in 2% SDS and 10 mM D m , RuvC ran as a single band (11) of about 20 kDa (Fig. 9, lane a ) . However, after dilution in R buffer lacking DTT, we observed two additional major bands (Fig. 9,  lane b ) . Band I corresponds to a molecular mass of around 39 kDa, consistent with the predicted mass of a RuvC dimer. Band I11 ran slightly ahead of the RuvC monomer, at around 17 kDa, presumably because of incomplete unfolding. We also observed several minor bands (70 and 56 kDa and a doublet at 36 kDa). When RuvC was diluted in the same buffer containing 1 l~l~ DTT, we observed the unfolded and partially unfolded forms of monomeric RuvC, with only a trace of the 39-kDa dimer (band I) (Fig. 9, lane c). From these results we conclude that in the presence of a reducing agent, the majority of the RuvC protein is monomeric. However, in its absence, RuvC protein is found in both monomeric and dimeric states.

DISCUSSION
In this paper we describe the construction of a plasmid in which the ruvC gene was placed under the control of the powerful T7 410 promoter. Following induction, RuvC protein was expressed to 20-30% of total cell protein. By lysing the cells in the presence of 1 M salt we were able to keep most of the RuvC in solution and allow its isolation from the crude lysate. The first step in the purification procedure, precipitation of RuvC by dialysis of the crude lysate to low salt, yields protein that is > 90% pure. At this stage, after a very simple and quick procedure, the protein is sufficiently pure for physical analysis. Further purification through three chromatographic steps yielded approximately 6 mg of > 99% pure RuvC from 2 liters of culture. The method described here provides a significant improvement over our previous purification scheme, which utilized a comparatively poor expression system (27), and that of Iwasaki et al. (28), in which over 70% of the RuvC was not recovered from the cells.
Using small synthetic DNA molecules, made by annealing oligonucleotides, we investigated the substrate specificity of RuvC. We found that the purified protein cleaves synthetic Holliday junctions to form nicked duplex products. At similar RuvC concentrations, we were unable to detect cleavage of duplex or single-stranded DNA substrates. The resolution activity was optimal at alkaline pH and required divalent cations (5-10 mM Mg2+). RuvC resolved the synthetic junctions by cleavage at the 3'-side of a phosphate, leaving 5"phosphate and 3'-hydroxyl groups at the incision termini. The resulting nicks may therefore act as substrates for DNA ligase (29).
In addition to the cleavage of Holliday junctions, we observed that RuvC cleaved RFI 4x174 DNA with low efficiency to produce linear duplex DNA. Since 4x174 DNA contains short palindromic sequences capable of cruciform extrusion (37), it is likely that linearization results from resolution of a cruciform structure as described previously (28). Similarly, the fragmentation of single-stranded circular 4x174 DNA by RuvC is likely to be caused by cleavage of junction-like structures formed by the interaction of complementary sequences. In the accompanying paper, we describe the substrate specificity of RuvC protein in further detail (34).
The subunit structure of RuvC protein has been investigated by gel filtration and PAGE. Upon gel filtration, the 19-kDa RuvC protein eluted as a single peak corresponding to a molecular mass of about 14 kDa. These results contrast with those presented by Iwasaki et al. (28), who reported that RuvC behaves as a dimer upon gel filtration. However, the former studies were performed a t higher protein concentrations than those used in the experiment of Fig. 8. When we applied RuvC protein to an SDS gel without prior addition of a reducing agent (or boiling), we observed the presence of a band (39 kDa) that corresponded to the expected mass of a RuvC dimer. This band was not observed when a reducing agent was added. From these data we suggest that the native form of RuvC is a dimer that can be easily dissociated into the monomeric form. Consistent with this proposal, analyses of RuvC-Holliday junction complexes by band shift assays indicate that the bound form of RuvC protein is d i m e r i~.~