Homologous Pairing in Genetic Recombination THE’PAIRING REACTION CATALYZED BY ESCHERICHIA COLI recA PROTEIN*

Purified recA protein, which is essential for genetic recombination of Escherichia coli, catalyzed ATP-de-pendent homologous pairing of double-stranded DNA and single-stranded fragments to form D-loops. When the double-stranded DNA was nicked circular DNA (form 11) or linear DNA (form 111), the reaction pro-ceeded nearly linearly during 30 min of incubation at 37 “C. When the double-stranded DNA was superhelical (form I), anomalous kinetics was observed. This anom- aly was suppressed by the addition of spermidine without affecting the final yield of D-loops. The formation of D-loops required stoichiometric amounts of recA protein, which were proportional to the concentration of single-stranded DNA but which were not affected by the concentration of double-stranded DNA. With form I1 or 111 DNA as the recipient for the formation of D-loops, the rate of the reaction was greatest when there was one monomer of recA protein/2-3 nucleotide residues of single-stranded DNA; larger amounts of single-stranded DNA inhibited the reaction. The formation of D-loops was half inhibited by 30 m~ NaCl and by 0.6 mM ADP, one of the products of the reaction. The ther- mal stability of D-loops made by recA protein was the same as that of D-loops made by annealing. In addition to pairing linear single strands with duplex DNA, recA protein made joint molecules from single-stranded cir- cular DNA and homologous form I1 or I11 DNA. According

by superhelical DNA (Holloman et al., 1975). To do so, we devised a rapid simple assay which detects the trapping by nitrocellulose of intact duplex DNA by virtue of its attachment to single-stranded DNA . Using this assay, we discovered that purified recA protein will catalyze the formation of D-loops from single-stranded fragments and homologous duplex DNA (Shibata et al., 1979a). The D-loop assay has permitted a rapid characterization of the pairing reaction catalyzed by recA protein, as described here and in other papers (Shibata et al., 1979a and1979b;McEntee et al., 1979;Cunningham et al., 1980). Moreover, the D-loop assay has proven useful in assaying the pairing of a variety of DNA substrates by recA protein because the efficient formation of joint molecules requires that one molecule be single-stranded or partially so (this paper; Cunningham et al., 1980;DasGupta et al., 1980) and because homologous pairing by recA protein does not require superhelical DNA (McEntee et al., 1979;Cunningham et al., 1979).

MATERIALS AND METHODS
The recA protein used in the experiments described here was the DEAE-cellulose fraction V described in the preceding paper . All other methods and special terms are as described there as well. Concentrations of DNA are expressed in moles of nucleotide residues.

RESULTS
The Time Course of Formation of D-loops-In the standard reaction mixture (legend to Fig. 1; Shibata et al., 1981), the formation of D-loops from form I1 or I11 DNA was roughly proportional to the period of incubation at 37 "C for 30 min or more (Fig. IA). The standard conditions described here, involving the use of 2.0 mM spermidine plus 6.7 mM MgC12, are conditions that we sought after we observed the anomalous kinetics illustrated in Fig. 1B. With superhelical DNA, in the absence of spermidine, complexes trapped by nitrocellulose fiiters were formed within 2 min and then decayed rapidly, giving way to a more regular time course in which the fraction of double-stranded DNA retained by the fiiter gradually reached a stable value. Although complexes formed at 2 min (hereafter called 2-min complexes) decayed during further incubation in the reaction mixture (Fig. lB), paradoxically they had the same resistance to treatment with detergent and the same heat stability as D-loops formed at a later stage of incubation. Moreover, the formation of 2-min complexes depended upon the presence of homologous single-stranded fragments. The decay of 2-min complexes in the reaction mixture ( Fig. 1B) was not due to nicking of superhelical DNA since, as we have observed before (Shibata et al., 1979a), p~ recA protein (fraction V) for the indicated periods at 37 "C in the standard reaction mixture  which contained 31 m M Tris. HC1 (pH 7.5), 6.7 m M MgC12, 2.0 mM spermidine-HCl, 1.3 m M ATP, 1.8 m M dithiothreitol, 88 pg of bovine serum albumin/ml. D-loops were assayed after treatment with 0.5% Sarkosyl for 5 min at 17 "C followed by heating for 4 rnin at 41 "C (0) or 50 "C (0) in 1.5 M NaC1, 0.15 M Na-citrate. B, form I fd [3H]DNA (8.8 p~; more than 76% form I), 12 p~ single-stranded fragments of fd [32P]DNA were incubated with 4.4 p~ recA protein (fraction V) for the indicated time periods at 37 "C. D-loops were assayed as described above. 0, standard reaction mixture (2.0 m M spermidine, 6.7 mM MgC12); A, spermidine was omitted, and 25 mM MgC12 was present during the incubation with recA protein. Because of this abnormality in the time course of formation of D-loops from form I DNA, we have done many experiments with form I1 or I11 DNA, even though spermidine suppresses the kinetic anomaly. Since the time course of the reaction is approximately linear for 30 min with form I1 or 111 DNA, the assay at 20 or 30 min measures the rate of the reaction (Fig.  1).
The Stoichiometric Requirement for recA Protein-As in the renaturation of DNA by helix-destabilizing proteins (Alberts and Frey, 1970;Christiansen and Baldwin, 1977), the formation of D-loops requires stoichiometric amounts of recA protein (Shibata et al., 1979a;McEntee et al., 1979). No Dloops are formed until recA protein reaches some minimal concentration, following which the yield of D-loops rises sharply to some maximal value (Fig. 2). The concentration of single-stranded DNA determines how much recA protein is required, while similar variations in the concentration of double-stranded DNA have no effect on the required amount of protein (Figs. 2 and 3; Shibata et al., 1979b). Fig. 2 shows the effect of single-stranded DNA under the standard condition used in the experiments described here, namely 2.0 mM spermidine plus 6.7 mM MgC12. The data from a number of experiments are summarized in Fig. 3. Estimated from the extrapolated intercept with the abscissa (Fig. 2), the minimal amount of recA protein needed to form D-loops was 1 monomer/&? nucleotide residues of single-stranded DNA when the double-stranded DNA was form I1 and 111 and 1 monomer of recA protein/20 residues when the doublestranded DNA was form I (Fig. 3). The least amount of recA protein needed to form D-loops appears to have been the same whether the reaction mixture contained 2.0 mM sper-midine plus 6.7 mM MgClz or 25 mM MgClz without spermidine, since data obtained under both conditions could be plotted on the same curve (Fig. 3A).
Further stoichiometric relationships are illustrated in Fig.

4.
In these experiments, we used one concentration of form I1 DNA, held the concentrations of recA protein constant at various levels, and measured the synthesis of D-loops at 30 min as a function of the concentration of single-stranded DNA. We determined that the time course of D-loop synthesis was linear, or nearly so, for 30 min for the lowest and highest concentrations of recA protein used. As a function of the concentration of single-stranded DNA, the rate of synthesis of D-loops increased, reached a maximal value, and then  decreased. For each concentration of recA protein studied, from 0.8 to 2.4 p~, the optimal rate of synthesis occurred at a ratio of about 2 nucleotide residues of single-stranded DNA/ monomer of recA protein. This value is about half of what we previously estimated when form I DNA was the recipient molecule (see Fig. 2 and Shibata et al., 1979a and1979b). Consistent with other experiments on the kinetics of formation of D-loops (Radding et al., 1980), the optimal rate of synthesis, i.e. the height of each peak in Fig. 4, was proportional to the concentration of recA protein.
The decreased rate of synthesis at high ratios of singlestranded DNA to recA protein could be due to a relative lack of recA protein or to an inhibitory effect of excess singlestranded DNA. When we started a reaction with an optimal ratio of single-stranded DNA to recA protein and 10 min later added four times as much single-stranded DNA, the rate of formation of D-loops promptly decreased (Fig. 5). This prompt inhibition presumably means either that single strands can inhibit by binding to a second site on recA protein or that complexes of recA protein and DNA dissociate and reassociate more rapidly than we could detect in this experiment.

Effects of Mgz+ and Spermidine on the Pairing Reaction-
In the absence of spermidine, the range of optimal concentration of Mg' + was broader when superhelical DNA was the substrate for the formation of D-loops ( Fig. 6). In 6.7 mM Mg2+, form I1 DNA yielded few D-loops. The addition of 1 to 2 mM spermidine to a reaction mixture containing 6.7 mM Mg2+ restored the synthesis of D-loops with form I1 DNA ( Fig. 7 ) . Similarly, 1.1 mM Mg2+ alone did not support the formation of D-loops with either form I or I1 DNA (Fig. 6), but the addition of 2 mM spermidine restored the reaction to normal levels for both forms of DNA (Table I; Fig. 7). Spermidine alone did not support the formation of D-loops with either form I or I1 DNA (Table I). Thus, spermidine lowered the requirement for Mg2+ to very low levels and reduced the anomolous kinetics noted earlier ( Fig. 1) but did not eliminate the requirement for Mg".
Reversible Formation of D-loops-In experiments on the effect of recA protein on the nucleolytic cleavage of DNA (Williams et al., 1981), we found evidence that recA protein can dissociate D-loops, which suggests that the formation of D-loops by recA protein is a reversible process. To demonstrate directly the reversibility of D-loop formation, we incubated @X174 form I r3'P]DNA with recA protein and purified D-loops made from +X174 form I ["]DNA. The percentage of 'H-and "P-labeled D-loops was measured for 60 min (Fig.   8A). In the presence of recA protein, the 3H-labeled D-loops decreased from 55 to lo%, while 32P-labeled D-loops increased to 5%. In a control from which recA protein was omitted, 3Hlabeled D-loops were stable, and 32P-labeled D-loops were not formed. When this experiment was repeated in the presence of 10 PM single-stranded fragments of +X174 DNA (Fig. 8B), 3H-labeled D-loops were dissociated, and "P-labeled D-loops were formed, as in the previous experiment, but the final yield of D-loops of both labels was higher, at about 30%. In both experiments, which involved the reaction of superhelical DNA in the absence of spermidine, 2-min complexes were observed as described above (Fig. 1B). These experiments show that the formation of D-loops is reversible since 1) D-loops simultaneously dissociated from [3H]DNA and formed in ["PIDNA in the absence of any added single strands, and 2) the addition of single strands decreased the apparent dissociation of Dloops and caused the fraction of D-loops in [32P]DNA and [3H]DNA to approach a higher common equilibrium value.
Inhibition of D-loop Formution-The formation of D-loops catalyzed by recA protein was inhibited by NaC1; 20 mM NaCl inhibited about 20%, and 50 mM NaCl completely inhibited the reaction (Fig. 9). By contrast, the DNA-dependent ATP-

TABLE I Partial replacement of Mg2' by spermidine in thepazring reaction
D-loops were made by incubation at 37 "C for 30 min and assayed as described in Shibata et al., 1981. The reaction mixture contained 2 mM spermidine (see Fig. 1A). The concentration of double-stranded DNA was 4.4 p~ and that of single-stranded fragments was 6 p~. Fd form 11, I11 DNA was 61% form I1 and 38% form 111; fd form I DNA contained more than 70% form I. : 6 0 h -" " " " " Q " " " " " " ~ " " " " " " _ B 4 0 0 I 0 -" " " _ " _ " " " " " " " " "  inhibited the reaction about two-thirds (Fig. 9). On the other hand, 1.2 mM AMP had no effect on the reaction (Fig. 9). During the incubation of 8.8 PM fd form I DNA and 12 PM fd single-stranded fragments at 37 "C for 30 min in the presence of 1.8 PM recA protein under standard conditions, 45% of ATP was hydrolyzed. Since this corresponds to 0.57 ,UM ADP, the  (1981) and spread for electron microscopy as described before (Shibata et al., 1979a). Micrographs were magnified 10 to 20 times and molecules were measured with a mechanical map measurer (Keuffel and Esser). Because it was not possible in all cases to distinguish the single-stranded arm from the double-stranded arm of a D-loop and since the two arms were not very different in length, we measured both arms and took the average. To exclude Dloops whose size was limited by the length of the single- pairing reaction is presumably inhibited by one of the products of the reaction.
The Structure of D-loops-Electron micrographs of D-loops made by recA protein (Shibata et al., 1979a;Cunningham et al., 1980) have shown that they have the same appearance as D-loops made by the uncatalyzed reaction at high temperatures . Previously, we have observed that the average length of the heteroduplex region in D-loops made by annealing is the length expected on the basis of the average superhelix density of the recipient duplex molecule Wiegand et al., 1977). A similar relationship appears to hold for D-loops made in superhelical DNA by recA protein. In superhelical fd DNA, the mean size of Dloops made by recA protein was about 400 nucleotides long, corresponding to 40 superhelical turns (Fig. 10). By contrast, the size of D-loops in nicked circular duplex fd DNA showed a broader distribution, as one would expect (Fig. 10). The observed narrow distribution of sizes of D-loops made by recA protein in superhelical DNA supports the conclusion that recA protein has no topoisomerase activity under the conditions studied (Cunningham et al., 1979). RecA protein appears neither to increase nor decrease the superhelix density of closed circular DNA in which D-loops are made.
The stability of D-loops diluted in 1.5 M NaCl, 0.15 M Nacitrate was examined at various temperatures after treatment with Sarkosyl to remove recA protein. D-loops made from form I DNA by recA protein were as stable as form I1 or I11 DNA itself; the latter melted above 80 "C (Fig. llB), and Dloops in form I DNA dissociated above 80 "C (Fig. 1lA). This behavior is identical with that of D-loops made in form I DNA by annealing . D-loops made from form I1 or I11 DNA were much less stable; in 1.5 M NaC1, 0.15 M Na-citrate, they dissociated at temperatures above 25 "C ( Fig.  11A). D-loops made from form I DNA by recA protein and then nicked by pancreatic DNase had exactly the same stability as D-loops made from form I1 or I11 DNA by recA protein (Fig. 1L4). This behavior is also the same as that of D-loops made by annealing . We have   (fraction V) plus the other components of the standard reaction mixture. Mixtures were incubated for 30 min (form I DNA) or 60 min (other than form I DNA) at 37 "C. Complexes formed during incubation were measured by the D-loop assay after treatment with 0.5% Sarkosyl for 5 min at 17 "C followed by heat treatment for 4 min at 25 or 65 "C in 1.5 M NaC1, 0.15 M Na-citrate. The index of heat stability is the ratio of double-stranded [3H]DNA in complexes after heat treatment at 65 "C to that after treatment at 25 "C. Examination of the preparation of circular single-stranded DNA by electron microscopy showed that 90% of molecules were circular. The preparation of form I DNA contained 89% form I and 11% of forms I1 and 111; the preparation of form I1 DNA contained 89% form 11, 7%  used the relative stability of joint molecules at 65" versus 25 "C as a criterion for characterizing products of the pairing reaction (see below and Table 11).
Under our previous condition for assaying D-loops, which included heating at 50 "C for 4 min in 1.5 M NaC1,0.15 M Nacitrate, more than two-thirds of D-loops made with form I1 or I11 DNA dissociated. This explains in part why we previously observed a low yield of D-loops with form I1 DNA (Shibata et al., 1979a;Cunningham et al., 1979). In our assays for D-loops, we nonetheless heated the product of the reaction at 41 "C for 4 min (except when stated otherwise) in 1.5 M NaC1, 0.15 M Na-citrate where one-third of D-loops of form I1 or I11 DNA dissociated, since we usually obtained more reproducible results than without heating or by heating at lower temperatures.
When the product of the reaction was treated with Sarkosyl at a temperature above 17 "C, a significant fraction of D-loops made with form I1 DNA dissociated. Under the conditions used in the experiments described in this paper, including incubation with 0.5% Sarkosyl at 17 "C for 5 min, recA protein appears to be inactivated and detached from DNA, since almost all of the ternary complexes of double-stranded DNA, recA protein, and single-stranded DNA formed in the presence of adenosine 5'-0-(3-thiotriphosphate) dissociated when treated in the same way (see Shibata et al., 1979b).
Substrate Specificity of the Pairing Reaction-The observation that circular single-stranded phage DNA stimulates recA protein to unwind duplex DNA partially (Cunningham et al., 1979) prompted us to explore the specificity of the pairing reaction. Elsewhere, we and others have reported that recA protein will make joint molecules from closed circular duplex DNA and circular duplex DNA with a gap in one strand (Cunningham et al., 1980;Cassuto et al., 1980). To characterize further the pairing activity of recA protein, we studied the possible reaction of circular single-stranded DNA with several kinds of duplex DNA (Table 11).
According to examination by gel electrophoresis, our preparation of single-stranded phage DNA consisted predominantly of circular DNA (data not shown), and according to examination by electron microscopy, about 90% of singlestranded molecules were circular (DasGupta et al., 1980). As measured by the D-loop assay, circular single-stranded DNA does not form a detectable fraction of stable joint molecules with form I DNA (Shibata et al., 1979a). The preparation of circular single-stranded DNA used in the present experiments made a small fraction of complexes with a preparation of form I DNA, and these complexes had the same thermal stability as D-loops made with form I DNA ( Table 11). Some of these at least are attributable to contamination of the singlestranded circular DNA with 10% of linear single strands.
When we reacted the circular single-stranded DNA with form I1 or I11 DNA, the yield of complexes exceeded the amounts made by fragments of single-stranded DNA ( Table   11). The yield of complexes made by circular single-stranded DNA relative to complexes made by linear single-stranded fragments was four times greater for form I1 or I11 DNA than for form I DNA. The complexes made by circular DNA were also more stable than complexes formed by single-stranded fragments that were about 600 nucleotides long. Circular single-stranded DNA also paired with fragments of doublestranded DNA made by digestion of fd duplex DNA by endonuclease R Hae 111. Complexes formed with these restriction fragments had the relative heat stability of D-loops in form I DNA, which is very similar to the stability of duplex DNA itself (Table 11, Fig. 11, and Wiegand et al., 1977).
Controls with circular single-stranded DNA of 4x174 showed that the formation of all of the complexes described required homologous DNA (Table 11).
These observations suggested that recA protein stably pairs circular single-stranded DNA with either nicked circular duplex DNA or linear duplex DNA. We have confirmed that inference and characterized the structures of the products by electron microscopy, as described elsewhere (DasGupta et al.,

Activation of recombination activities and repressor cleavage by single-stranded DNA
All entries in the table are the number of nucleotide residues in single-stranded DNA/monomer of recA protein.

Reaction
Formation of D-loops" ing of dou-of repres-  (Cunningham et al., 1980), show that the D-loop assay can be used to measure a variety of joint molecules that can be made by recA protein.

DISCUSSION
The energy of superhelix formation drives the uncatalyzed formation of D-loops by stabilizing the product. The energy of superhelix formation is not sufficient, however. The uncatalyzed reaction has a temperature threshold, which can be rationalized by supposing that heat partially unwinds the duplex DNA, thereby permitting it to pair with a third strand Radding et al., 1977). Using the energy of ATP, recA protein is able to promote this pairing reaction at 37 "C, possibly by means of its ability to partially unwind duplex DNA (Cunningham et al., 1979;Shibata et al., 1981).
RecA protein makes D-loops in duplex DNA whether it is superhelical or not (Cunningham et al., 1979;McEntee et al., 1979). Nonetheless, as one might expect, there are distinct differences between superhelical and nonsuperhelical DNA as substrates for recA protein: 1) although the amount of recA protein needed is independent of the amount of duplex DNA ( Fig. 3) but rather is a function of the concentration of singlestranded DNA, the kind of duplex DNA affects the requirement for recA protein; about half as much recA protein is needed for a given amount of single-stranded DNA when the recipient molecule is superhelical ( Fig. 3; Table 111). 2) The time course of D-loop synthesis with form I DNA showed an anomaly that was suppressed by spermidine (Fig. 1). 3) In the absence of spermidine, form I DNA formed D-loops a t lower concentrations of MgClz than form I1 (Fig. 6). Elsewhere we have reported that gapped circular molecules preferentially pair with superhelical DNA uersus nicked circular DNA (Cunningham et al., 1980). In addition, form I DNA in 1-2 mM MgClz was a good cofactor for the ATPase activity of recA protein, whereas form I1 DNA was not .
These observations suggest that the energy of superhelix formation can help to make D-loops, but they also indicate that superhelicity may significantly alter the mechanism of the reaction (Fig. 1B).
In the presence of 1 mM MgC12 and single-stranded DNA, or form I DNA, recA protein hydrolyzes ATP without spermidine . Under this condition, neither the pairing reaction (Fig. 6) nor the formation of ternary complexes occurs.2 Tke pairing reaction was inhibited com-T. Shibata and R. P. Cunningham, unpublished data.
pletely by 70 mM NaCl (Fig. 9), but ATP hydrolysis was only slightly affected by even 100 mM NaC1.' Thus, the hydrolysis of ATP can be uncoupled from the complete reaction. Despite the functional dissimilarity between the proteolytic inactivation of a molecule of repressor and the homologous pairing of DNA molecules, striking similarities exist between the requirements for three relevant reactions in uitro, namely cleavage of repressor, pairing of single-and double-stranded DNA, and unwinding of double-stranded DNA: 1) all three reactions require single-stranded DNA as a cofactor, and the single-stranded DNA governs the properties of all three reactions in similar ways. At a limiting concentration of recA protein, increasing amounts of single-stranded DNA fiist stimulate the reactions and then inhibit Shibata et al., 1979b;Fig. 4). Optimal reactions occur at ratios of 2-8 nucleotide residues of single-stranded DNA/monomer of recA protein (Table 111). Similarly, none of the three reactions occurs until a certain minimal amount of recA protein is present ( Fig. 2; Craig and Roberts, 1980;Shibata et al., 1979a and1979b;McEntee et al., 1979).3 The minimal amount of recA protein required is proportional to the concentration of single-stranded DNA, both for synthesis of Dloops (Fig. 3) and for inactivation of repressor  and corresponds to 1 molecule of protein monomer/8 to 20 nucleotide residues (Table 111). 2 ) A31 three reactions require ATP as a cofactor Shibata et al., 1979a;McEntee et al., 1979;Cunningham et al., 1979). In the cases of the cleavage of repressor and the unwinding of the double helix, adenosine 5'-0-(3-thiotriphosphate), an analog of ATP which competitively inhibits the ATPase activity of recA protein (Shibata et al., 1979b), is much more effective than ATP (Roberts et al., 1979;Cunningham et al., 1979). 3) In the cases of cleavage of repressor and unwinding of duplex DNA, single-stranded DNA can be replaced by oligonucleotides. In both cases, the reaction requires a larger molar amount of oligonucleotides than that of single-stranded DNA.'.'' These observations show that in uitro a common mechanism activates recA protein as a specific protease and as a DNA enzyme. By analogy, we infer that in uivo a common mechanism activates the direct participation of recA protein in the recombination and repair of DNA and its indirect participation via the derepression of other genes (Witkin 1976;Oishi et al., 1979;McPartland et al., 1980;Little et al., 1980;Kenyon and Walker, 1980). The relative extent of direct and indirect action of recA protein in various aspects of ' J. W. Roberts, personal communication.

The Pairing Reaction Promoted by RecA Protein
recombination and repair remains to be determined. The experiments reported here and elsewhere (Cunningham et al., 1980;DasGupta et al., 1980;Cassuto et al., 1980) have revealed a broader specificity of recA protein in the synthesis of joint molecules than indicated by the formation of D-loops. RecA protein will form joint molecules of circular single strands and either linear or nicked circular duplex DNA (Table 11; this paper; DasGupta et al., 1980), as well as joint molecules of gapped circular DNA and form I DNA (Cunningham et al., 1980;Cassuto et al., 1980). Stable complexes that were detectable by the D-loop assay did not form between superhelical DNA and circular single-stranded DNA (Table 11; this paper; DasGupta et al., 1980) or between superhelical DNA and nicked circular DNA (Cunningham et al., 1980). The substrates that formed joint molecules are summarized on the right of Fig. 12. All of these combinations produced joint molecules with approximately equal efficiencies. From these observations, we deduce a rule that describes the specificity of the reaction. RecA protein will catalyze the formation of joint molecules when one of them is singlestranded, or partially so, and when either one has a free end. We interpret the requirement for single-stranded DNA to reflect not only the need for a donor strand but also the specific role that single-stranded DNA plays as an effector that stimulates recA protein to bind and unwind duplex DNA, as described above. We interpret the need for a free end as topologic and thermodynamic; in the absence of topoisomerase activity (cf Cunningham et al., 1979), a free end is necessary to form a heteroduplex region that has the normal Watson-Crick duplex structure ( Fig. 12 step b). In Fig. 12, we have presented a model that relates all of these observations, and elsewhere we have discussed further inferences derived from the electron microscopic study of the various joint molecules (DasGupta et al., 1980).