[Frontiers in Bioscience 3, d570-603, June 17, 1998]

	[Frontiers in Bioscience 3, d570-603, June 17, 1998] Reprints PubMed CAVEAT LECTOR
	DNA STRAND EXCHANGE PROTEINS: A BIOCHEMICAL AND PHYSICAL COMPARISON Piero R. Bianco ¹, Robert B. Tracy ^1,2,3and Stephen C. Kowalczykowski ^1,2 1Sections of Microbiology and of Molecular and Cellular Biology, ²Microbiology Graduate Group, University of California, Davis, CA 95616, ³ Norris Comprehensive Cancer Center, USC School of Medicine, 1441 Eastlake Ave., Los Angeles, CA 90033 Received 5/24/98 Accepted 6/11/98 4. MECHANISM OF DNA STRAND EXCHANGE As is evident from the preceding, RecA protein and its analogues mediate a reaction, DNA strand exchange, that is central to homologous recombination. The enzymatic properties of the in vitro strand exchange activity of RecA protein have been studied in great detail (4,28,170) and a summary of the biochemical properties of the DNA strand exchange proteins discussed herein is shown in table 2. Below, we detail the steps of protein-promoted DNA strand exchange. DNA strand exchange catalyzed by RecA protein occurs by a number of kinetically distinct phases that can be subdivided into at least three experimentally distinguishable steps: (1) presynapsis, (2) synapsis and (3) DNA heteroduplex extension (figure 3). Presynapsis is the step where RecA protein assembles onto the ssDNA to form the nucleoprotein species that is active in the homology search. Synapsis is characterized by initially random non-homologous contacts occurring between the presynaptic complex and naked dsDNA (termed coaggregates), the search for homology, homologous pairing, and finally conversion to either paranemic or plectonemic joint molecules (discussed below in the synapsis section). DNA heteroduplex extension, as discussed earlier, is the polar migration of the nascent DNA heteroduplex joint. Figure 3. Kinetic steps of DNA strand exchange. The reaction between circular ssDNA and linear dsDNA is shown. The product is a nicked circle and a displaced single strand of DNA. A similar reaction scheme likely applies to all of the DNA strand exchange proteins discussed but reaction polarity is not indicated; the illustration is adapted from Kowalczykowski (185). The steps shown are: 1-presynapsis; 2 and 3-synapsis; and 4-DNA heteroduplex extension. The green spheres represent the strand exchange protein; the orange squares represent ssDNA binding proteins. DNA strand exchange occurs between two DNA molecules provided that the following criteria are met: (1) one of the DNA molecules contains a region of ssDNA; (2) the ssDNA region occurs at a site homologous to the other dsDNA molecule; and (3) for topological reasons, one of the substrate molecules has an end. A reaction that has become a model for mechanistic studies involves the exchange between circular ssDNA (f; X174 or M13 phage) and linear dsDNA (figure 3). The products are nicked, circular dsDNA and linear ssDNA (171,172). 4.1 Presynapsis (figure 3, Step 1) Presynapsis is the ordered assembly of the RecA protein homologue on ssDNA to produce a nucleoprotein complex that is the active species in DNA strand exchange (filaments formed on dsDNA are shown in figure 5 and discussed in Section V). This process occurs by either of two pathways: first, the single-stranded DNA binding protein (SSB; RPA; UvsY and/or gp32) binds to ssDNA (figure 3, step 1A); this is followed by the subsequent binding of the DNA strand exchange protein (RecA; UvsX or Rad51) and displacement of the single-stranded DNA binding protein to form the presynaptic filament (figure 3, steps 1B and 1C, left panel). In the second pathway, the DNA-strand exchange protein binds to the ssDNA substrate, forming an incomplete nucleoprotein filament due to limitations imposed by DNA secondary structure (step 1A, right panel). The single-stranded DNA binding protein binds, removes the secondary structure and is then displaced by further binding of the DNA strand exchange protein (steps 1B and C). The net result of either of these two pathways, is the formation of a complete nucleoprotein filament that involves direct contact by the primary DNA-binding site of the strand exchange protein, and results in the generation of a contiguous secondary DNA-binding site that is essential for homologous pairing and DNA strand exchange (173,174). 4.1.1. Escherichia coli Filament formation by RecA protein is dependent on DNA length, DNA composition, and the amount of secondary structure within the ssDNA substrate (95,175-178). The ATP-bound form of the protein is needed to constitute the active form of the nucleoprotein filament. ATP greatly stabilizes the complex, and transforms RecA protein into a DNA-binding state with high affinity for ssDNA (120). The RecA nucleoprotein filament is capable of hydrolyzing ATP with a k_cat of 18-30 min^-1, a value that is affected by ssDNA length and sequence (117,177-179). However, DNA serves only as a scaffold to promote filament formation since RecA protein can be activated by high salt concentrations to form both filaments and to hydrolyze ATP (180-182). Though RecA protein is an ATPase, ATP hydrolysis is not essential for DNA strand exchange (64,183,184). Stoichiometric amounts of RecA protein, relative to ssDNA concentrations, are required for filament formation, with maximum DNA strand exchange rates occurring when filaments are formed at a ratio of at least 1 RecA monomer to 3 nucleotides of ssDNA. This stoichiometric requirement indicates the importance of a saturated, contiguous RecA protein-ssDNA filament in the DNA strand exchange process (185). In addition to ATP, ATP-g-S and dATP also support presynaptic filament formation and are capable of transforming RecA protein into the high affinity ssDNA-binding state, the active form of the protein (179,186). This complex, the presynaptic filament, is the physiologically relevant species in RecA protein-mediated cellular processes. Filament assembly onto ssDNA occurs in a 5'®;3' direction (187), which is the same direction as branch migration. The resulting filament has a regular structure and the prominent feature is a large helical groove, with 6.2 monomers per helical turn and each monomer interacting with 3 bases of DNA (figure 5) (188). The SSB protein effects the binding of RecA protein to ssDNA, greatly stimulating the DNA strand exchange process (117,189). The role of SSB protein in presynapsis is to remove secondary structure from ssDNA, which is inhibitory to the formation of the saturated presynaptic complex (117,119-122). This is consistent with the role of the protein as a helix-destabilizing protein and the observation that other helix destabilizing proteins can substitute for SSB protein in DNA strand exchange in vitro (119). 4.1.2. Bacteriophage T4 Presynaptic complex formation by the T4 phage proteins is similar, but features an added complexity, the UvsY protein. During presynapsis, UvsX binds to ssDNA cooperatively with a stoichiometry of one UvsX monomer for every 3 to 5 nucleotides of ssDNA to form a nucleoprotein filament (46,47). In contrast to RecA protein, the binding of ATP does not stabilize the UvsX protein-ssDNA presynaptic filament under steady-state conditions for hydrolysis of ATP; however, the binding of ATP-g-S does (190). This suggests that ATP-g-S increases the equilibrium binding affinity of UvsX protein for ssDNA which is very similar to what is observed for RecA protein (179,191). However, the UvsX protein-ssDNA complex is more vigorous in hydrolyzing ATP (the rate of ATP hydrolysis is 8 to 10-fold higher) (44), making its presynaptic filament more dynamic; this may explain why ATP does not stabilize the UvsX-presynaptic complex, (i.e., the UvsX-ssDNA-ATP complex is short lived) and why an accessory protein like UvsY protein is needed. In addition, UvsX protein is unique among RecA protein analogues studied to date, in that it produces both ADP + P_i and AMP+ PP_i (44). The significance of the AMP production is presently unknown, however. The formation of the UvsX protein presynaptic filament is facilitated by the presence of both gp32 and UvsY protein (a UvsX-dsDNA nucleoprotein filament is shown in figure 5B and is discussed in Section V). Gp32 enhances filament formation by removing secondary structure native to the ssDNA (45,51). UvsY facilitates filament formation by stabilizing the presynaptic complex itself (51,192,193). Although gp32 is the T4 phage counterpart to the E. coli SSB protein, there is no structural homologue of the UvsY protein in the E. coli recombination system; however, the RecF, O, and R proteins may comprise a functional homologue (123,124,194-198). 4.1.3. S. cerevisiae Presynaptic filament formation by Rad51 protein results in a stoichiometric nucleoprotein filament that resembles the filament formed by both the RecA and UvsX proteins (figure 5C, discussed in Section V) (63). This complex is the active species in DNA strand exchange (63). Optimal reaction rates are achieved with a stoichiometry of 3 nucleotides per Rad51 monomer (63,199). Binding to ssDNA is stabilized by either ATP (59) or the non-hydrolyzable analogues of ATP, ATP-g-S and AMP-PNP (64). Rad51 protein filament formation is stimulated by the RPA heterotrimer (63,65), which is the eukaryotic analogue of the E. coli SSB and T4 gp32 proteins. In contrast to both the RecA and UvsX proteins, very little ATP is hydrolyzed by the Rad51 nucleoprotein complex (k_cat = 0.73 min^-1) during presynapsis or during DNA strand exchange (64); interestingly, some of the enzymatic characteristics of Rad51 protein more closely parallel those of a mutant RecA protein, RecA K72R, which has a reduced level of ATP hydrolysis (~ 600-850-fold reduced to a k_cat of 0.032min^-1) yet nonetheless, is able to promote DNA strand exchange. 4.2 Synapsis (figure 3, Steps 2, 3) Once the presynaptic filament has assembled on ssDNA, synapsis ensues. In this stage of the reaction, a dsDNA molecule must be bound to the filament, homology to the ssDNA within the filament located within the dsDNA, and a plectonemic heteroduplex joint formed. For this phase of DNA strand exchange to occur, a second DNA molecule must bind in a sequence-independent fashion to the secondary DNA binding site of the nucleoprotein filament; presumably, the close approach of this second DNA molecule is facilitated by shielding of the negatively charged phosphodiester backbones of the two DNA molecules by the DNA strand exchange protein. Once dsDNA is bound, the search for homology takes place. The search is rapid, and requires that the binding of dsDNA to the secondary site be both weak and transient. The recognition of homology takes place when the ssDNA within the presynaptic filament hydrogen bonds, via non-Watson-Crick base pairing, to either the major or minor groove of the bound dsDNA, in a mechanism that does not require triplex intermediate formation. This alignment provides a signal to the strand exchange protein that homology has been located; base switching occurs, and the heteroduplex dsDNA occupies the primary site whereas the displaced strand occupies the secondary site. If the location of homology occurs in the center of the dsDNA, a paranemic joint forms; if this occurs at the dsDNA molecule ends, a plectonemic joint forms. Due to their inherent instability, paranemic joints must be converted to plectonemic joints in order to survive and become recombinant DNA structures. 4.2.1. Escherichia coli Once an active presynaptic complex forms, the complex can then pair the ssDNA within the filament to dsDNA. On the basis of probability, this initial contact must be at a site of non-homology. Indeed, RecA protein is capable of forming a complex between ss- and dsDNA, or between ssDNA molecules with little or no sequence complementarity (173,200-203). These complexes, called coaggregates, contain many DNA molecules arranged together in a three-dimensional network that can be sedimented in a low speed centrifuge (204,205). They form rapidly, and kinetically precede joint molecule formation. Thus, it has been inferred that coaggregates, or a related complex, of RecA protein and ss- and dsDNA are intermediates on the pathway of the strand exchange reaction (204,205). Although RecA protein binds ssDNA in a sequence-independent manner (though compositional preferences exist (95,176,206,207)), the problem of the homology search is analogous to the problem that sequence-specific DNA binding proteins face in locating their target sequence (208). However, for the case of RecA protein and its analogues, the scale of the search problem is much larger: the binding protein is the entire nucleoprotein filament (which can consist of thousands of protein monomers, depending on the ssDNA length), and the target is the unique complementary sequence within the entire genome. Although the entire nucleoprotein filament is involved in the search for homology, the minimum length of homology required for recognition is as low as 15 nucleotides in vitro (95), which is somewhat less than that needed in vivo, where homologous recombination requires minimally, about 23-40 bp of homology (209,210). For homologous alignment to occur, a second DNA molecule must bind to the secondary DNA binding site of the RecA protein filament. This binding, although not sequence specific, does have a hierarchy: optimal binding occurs with oligopyrimidinic DNA while poorer binding is observed with oligoadenylic DNA (173). This hierarchy of binding is not due to the sequence of the ssDNA present in the primary site, but rather is due to an intrinsic property of RecA protein (173). Thus, the proposal that homologous recognition occurred by a novel form of non-Watson-Crick base-pairing called "self-recognition" (211,212) is not substantiated (173,174). How then does RecA protein "sense" when homology has been located? Though many details of the homology search remain unknown, several steps are clear. During the search, the binding of a second DNA molecule to the nucleoprotein filament must, by necessity, be both weak and transient, to facilitate a rapid search. When the homologous locus is found, the dsDNA is unwound, and one of the strands of the duplex pairs with its complement (which is bound to the primary, original DNA-binding site of the pairing protein). The other strand is displaced into the secondary DNA-binding site, where it is bound more tightly than the duplex DNA. During the search for homology, the DNA duplex remains base-paired; it is only when homology is located, that strand switching occurs. The recognition of homology between the ssDNA within the filament and either the minor groove of dsDNA (213-215), or the major groove of the dsDNA, provides the signal to RecA protein that homology has been located (173). Immediately after base-pair switching, the heteroduplex dsDNA product would occupy the primary site, whereas the displaced ssDNA would occupy the secondary site; there is evidence to support this scenario (174,214,216,217). Once the two DNA molecules are homologously aligned, RecA protein then catalyzes the nascent exchange of strands. In the prototypical in vitro reaction depicted in figure 3, this requires a local denaturation of the dsDNA molecule and the subsequent exchange of the identical single strands of DNA. These steps may be simultaneous or separated in time, but the result is the production of an intermediate known as a joint molecule. Two types of joint molecules may form, depending on the topological constraints of the DNA: either paranemic or plectonemic. A paranemic joint is one in which the individual complementary strands do not intertwine, producing a molecule that is base-paired but not topologically linked; whereas, a plectonemic joint is one in which the incoming single strand is intertwined around its complement as in native dsDNA. In the reaction displayed in figure 3, paranemic joints will form in the interior of the duplex substrate, and plectonemic joints will form at the ends of the duplex substrate. An experimental distinction between paranemic and plectonemic joints is that paranemic joints require RecA protein (or its counterpart) for stability whereas plectonemic do not. Since paranemic joints are statistically more probable than plectonemic, paranemic molecules are likely intermediates on the reaction pathway to the formation of more stable plectonemic joint molecules (218,219). Both the type and efficiency of joint molecule formation are affected by SSB protein. If SSB protein is omitted, the displaced ssDNA is used by RecA protein to form a second joint molecule with another dsDNA molecule; this results in the formation of complex, homology-dependent networks of joint molecules. In the presence of SSB protein, RecA protein does not form these networks, since SSB protein prevents re-invasion events by binding to the ssDNA displaced from the joint molecules; in addition, the yield of joint molecules is greater (220). However, under reaction conditions where RecA protein is better able to compete with SSB protein (i.e. in the presence of the volume excluding agents polyethylene glycol or polyvinyl alcohol; or when dATP is used as cofactor), networks are readily formed (186,221,222). Thus SSB protein acts both at the pre- and post-synaptic steps of DNA strand exchange. 4.2.2. Bacteriophage T4 Some aspects of synapsis are similar for UvsX protein and RecA protein: 1) UvsX protein coaggregates non-homologous ssDNA and dsDNA (47); and 2) UvsX protein catalyzes formation of approximately equal numbers of paranemic and plectonemic joint molecules (223). Differences also exist between the two proteins. First, while both RecA protein and UvsX protein form homology-dependent DNA networks in the absence of a single-stranded DNA-binding protein, UvsX protein also forms networks even in the presence of gp32, but this may simply reflect an increased ability of UvsX to displace gp32 (50). Second, since UvsX protein, unlike RecA protein, binds readily to either ssDNA or dsDNA under normal strand exchange conditions (46), joint molecule formation is reduced due to UvsX protein binding to the dsDNA (223); this behavior more closely resembles that of Rad51 protein (see below). This limitation is overcome in vitro by using limiting amounts of UvsX protein relative to the DNA (45). 4.2.3. S. cerevisiae As for UvsX protein, joint molecule formation is inhibited if Rad51 protein is allowed to bind dsDNA and can be avoided by assembling presynaptic filaments using exactly stoichiometric amounts of Rad51 protein, or by using excess dsDNA (63). This inhibition of joint molecule formation results from the rapid and kinetically stable binding of Rad51 protein to dsDNA, a situation that is very different from the behavior of RecA protein. Rad51 protein initiates joint molecule formation if the linear dsDNA contains a single-stranded overhang at least 2 nucleotides in length that is complementary to the ssDNA within the presynaptic filament (199). Both joint molecules and nicked circle products are reported to form more efficiently with dsDNA containing a 3'-overhang than with a 5'-overhang (199). 4.3 Branch migration (figure 3, Step 4) Once the plectonemic joint has formed, the branch migration phase of DNA strand exchange commences. During this phase, the nascent, heteroduplex joint is extended until complete exchange of single strands of DNA occurs, resulting in a nicked, double-stranded circle. Though kinetically distinct, branch migration may not be a mechanistically separate step, but rather may represent a continuation of plectonemic joint molecule formation (219,224). 4.3.1. Escherichia coli For RecA protein, branch migration proceeds in a 5'®;3' direction relative to the incoming single strand (the same direction as RecA protein polymerization), at a rate of 2-10 bp sec^-1 (171,225), requires ATP hydrolysis (19,226) and induces torsional stress in the dsDNA (227,228). Branch migration is relatively tolerant of nucleotide sequence mismatches (229,230) and DNA lesions (231,232), although the reaction is inhibited 60% by heterologous insertions of 140 bp within the dsDNA substrate and by insertions of 1000 nucleotides within the ssDNA substrate (230,231). The ability of RecA protein to not only promote DNA heteroduplex extension, but also to extend the DNA heteroduplex beyond limited regions of sequence non-homology (heterology), may serve a valuable biological function, despite the existence of specialized branch migration helicases like the RuvAB proteins (233). For this reason, the mechanism by which RecA protein facilitates the bypass of DNA sequence heterology is an important issue. Several studies established the key requirements for this heterology-bypass (234,235) (D.A. Kitchell and S.C.K., unpublished observations). RecA protein can facilitate bypass through a heterology as large as 140 bp within the dsDNA substrate (230,231), provided that these mismatched sequences are beyond the point at which DNA strand exchange initiates (i.e., the heterology needs to be beyond the 5'-end of the strand being displaced from the dsDNA). However, RecA protein is unable to bypass an insert of only 22 base pairs that is at the distal end of the dsDNA (234). In all cases, the ability of RecA protein to promote DNA heteroduplex extension progressively diminishes as the length of homology distal to the heterology decreases. Furthermore, when the length of the heterology increases, the amount of distal homology required to bypass this heterology increases, apparently to compensate for the longer heterologous distance that must be traversed. Thus, to bypass a heterology, DNA strand exchange must initiate in a region of DNA sequence homology, and there must be additional homology beyond the heterology. ATP-g-S could not substitute for ATP in these bypass experiments, demonstrating that ATP hydrolysis was necessary, presumably to permit some dissociation/redistribution of the RecA nucleoprotein filament. Several mechanisms were proposed to explain these important characteristics. One envisioned that once the heterology was encountered, lateral slippage of the exchanging strand within the RecA nucleoprotein filament would permit re-alignment of the homologous sequences (234). A second mechanism envisioned that facilitated rotation of the DNA substrates about each other mediated by the RecA protein filament in an ATP-hydrolysis coupled mechanism, provided the "motor" that rotated the DNA past the heterology; the specific details of this model were elegantly elaborated previously to which the reader is referred for details (235). Here we would like to present an alternative idea, based on these collective data and unpublished considerations (D.A. Kitchell and S.C.K., unpublished observations). Figure 4 shows how topological stress, resulting from local unwinding of dsDNA that is constrained in a closed topological domain, can facilitate bypass through a heterology. The basic premise of the topological model is that, homologous pairing involves intermediates (the synaptic complexes) in which the dsDNA is highly unwound. Furthermore, in the presence of ATP, the RecA nucleoprotein filament that participates in DNA strand exchange is dynamic, and it is frequently binding and releasing dsDNA in its quest for proper DNA sequence alignment. Electron microscopy shows that the RecA nucleoprotein filament can envelop kilobase pair-lengths of homologous dsDNA (216), and that the binding of dsDNA by the RecA nucleoprotein filament produces a topological unwinding of the dsDNA from the canonical 10.4 bp/turn to 18.6 bp/turn (236-241,242 (figure 4)). Thus, in the static situation, when the RecA nucleoprotein filament envelops dsDNA, the dsDNA including the heterology is extensively unwound by up to 8.2 bp/turn resulting in 18.6 bp of dsDNA/filament turn versus 10.4 bp turn in B-form dsDNA (figure 4C). However, in the dynamic case, in the presence of ATP, if the RecA nucleoprotein filament were to release a medial region of the dsDNA while the two flanking regions of the dsDNA remained bound to the filament, then the binding energy that maintained the bound dsDNA in its unwound form would be absent, and the released dsDNA segment would be both topologically constrained and highly underwound (figure 4D, left side); in fact, the release of dsDNA from just 2 turns of the filament would result in 16.4 bp of unwinding. Thus, the transient release of dsDNA from 20 turns of the filament would produce a domain that is unwound by 164 bp relative to B-form DNA. If this unwinding occurred in a region of dsDNA heterology, then there would be sufficient, locally trapped unwinding to induce transient strand separation of the dsDNA (figure 4D, right side), and to bypass the heterology in a kinetically stochastic way. With the region of dsDNA heterology locally melted, bypass would readily occur upon subsequent re-polymerization of the filament (figure 4E); however, if a nick were located in the region of heterology, then the superhelical strain is released and no bypass occurs. This model explains the key experimental requirements for bypass (the requirement for increasingly longer lengths of homology distal to the heterology, the need for ATP, and the inhibitory effect of a dsDNA nick), and it lends insight into how RecA protein, and presumably all of the other DNA strand exchange proteins (since they all extend dsDNA when bound) can mediate bypass of short regions of DNA sequence heterology without the need for specialized branch migrations proteins. Figure 4. Model for heterology bypass by RecA protein. A pairing event involving ssDNA (drawn in blue) and dsDNA (drawn in red/pink) containing a region of heterology, is shown. The spheres represent RecA protein; the large black arrow indicates the direction of migration of the heteroduplex joint. Details of this model are discussed in the text. 4.3.2. Bacteriophage T4 UvsX protein, like RecA protein, catalyzes branch migration such that DNA strand displacement is 5�®;3� relative to either the incoming presynaptic filament or the displaced strand (45,50). However, UvsX protein catalyzes branch migration at a rate of 12-15 bp sec^-1 (50,151), which is faster than the rate catalyzed by RecA protein (see above). 4.3.3. S. cerevisiae In contrast to RecA and UvsX proteins, Rad51 protein promotes strand exchange with the opposite polarity (i.e. 3�®;5� relative to the incoming single strand (63)) at a rate of 0.5 bp sec^-1, and has no requirement for ATP hydrolysis at any stage (64). In addition, the rate of formation of both joint molecule intermediates and products in the reaction is significantly reduced compared to that of either the RecA (2-10 bp sec^-1) or UvsX proteins (12-15 bp sec^-1) (62-64). Also, this stage of the reaction is inhibited by a 458 bp heterologous insertion in the dsDNA substrate (63). 4.4. ATP hydrolysis in DNA strand exchange DNA strand exchange requires the presence of a nucleoside triphosphate cofactor, usually ATP. Under standard conditions, ATP hydrolysis coincides with the pairing and exchange of strands of DNA. Initially, it was thought that ATP hydrolysis was a requirement for DNA strand exchange. It is now known that neither the hydrolysis of ATP (183,226) nor the presence of a high-energy phosphate bond (184) are necessary for DNA strand exchange to occur. It is the formation of the high-affinity ssDNA-binding state of RecA protein that is necessary for DNA strand exchange; the formation of which is brought about by ATP binding to RecA protein bound to DNA. The binding of the non-hydrolyzable analogue of ATP, ATP-g-S, or the non-covalent complex of ADP-AlF^-₄, can also induce the high-affinity DNA-binding state of RecA protein. Although the hydrolysis of ATP is not required for the exchange of DNA strands, it is required at phases of DNA strand exchange that require the dissociation of RecA protein that is induced by ADP, the product of ATP hydrolysis. 4.4.1. Escherichia coli RecA protein is a DNA-dependent ATPase with a single active site for the binding and hydrolysis of ATP and other nucleoside triphosphates (17,206,243-245). The enzyme is able to hydrolyze ATP with either ss- or dsDNA as cofactor, resulting in a k_catwith ssDNA of 18 - 30 min^-1 and with dsDNA of 19 to 22 min^-1 (178,246,247). The active species in ATP hydrolysis is the nucleoprotein filament, with ATP being hydrolyzed uniformly throughout the filament and with no detectable enhancement at filament ends (117,177,248). Hydrolysis of ATP results in one or more conformational changes in RecA protein. ATP and ADP serve to modulate RecA between the "high-affinity" DNA-binding and "low-affinity" DNA-binding states, respectively (179,191). It may not be not the hydrolysis per se that modulates the two states, but rather the release of P_i, which is extremely rapid (249). This proposal is supported by experiments utilizing the ground state analogue, ADP-AlF^-₄, which maintains RecA protein in the high-affinity ssDNA-binding state in the absence of a high-energy bond (184,250). Although ATP hydrolysis occurs throughout DNA strand exchange promoted by RecA protein, it is not required for all of the stages (19,184,219,226,251-254). ATP hydrolysis is not required for the exchange of DNA strands (184,250,255). However, ATP hydrolysis is required to dissociate RecA protein from the heteroduplex products once the reaction is complete (226,252,256), to facilitate the bypass of structural barriers such as heterologous sequences (253,254), and to maintain reaction polarity (125,257). In addition, ATP hydrolysis is required for DNA strand exchange involving four strands (258). Wild-type RecA protein is able to promote strand exchange using the non-hydrolyzable analogue ATP-g-S (226). In addition, a mutant RecA protein, K72R, in which ATP hydrolysis is reduced by approximately 600-850-fold, is able to promote homologous pairing and exchange of up to 1.5 kilobase pairs of DNA (183). In both of these situations, extension of the heteroduplex is blocked (i.e. joint molecule intermediates form but are not converted to gapped or nicked dsDNA products) and the reaction is bi-directional (257), which is in contrast to that of the ATP reaction, which has a distinct polarity (171). However, a nucleoside triphosphate is not an essential component of the pairing and exchange process, because a complex of ADP and AlF^-₄activates all of the activities of RecA protein (184,250,255). 4.4.2. Bacteriophage T4 UvsX protein, in contrast to RecA protein, has an absolute requirement for ATP hydrolysis in DNA strand exchange (259). The rate of ATP hydrolysis by UvsX protein is 8- to 10-fold higher than that of RecA protein, with a k_cat of 145 - 240 min^-1 (44), and this activity is inhibited by both ADP and ATP-g-S (259). In the presence of 2 mM ATP- g-S, UvsX protein is unable to promote strand exchange; however at 16 and 80 mM ATP-g-S, UvsX protein can form intermediates (259). This may simply reflect the situation that existed for RecA protein prior to the realization that the ATP-g-S reaction had a unique reaction optimum (184). Substitution of ATP-g-S for ATP allows UvsX protein to bind ssDNA, unwind dsDNA (259), and promote DNA strand annealing (45). Addition of ATP-g-S to an ongoing DNA strand exchange reaction, initiated with ATP, causes a transient increase in the rate of branch migration (50). This transient stimulation, which occurs due to inhibition of ATP hydrolysis, is attributed to a kinetic stabilization of the active filamentous form of UvsX protein. The high turnover of the UvsX protein nucleoprotein filament appears to limit the steady-state rate of branch migration (259). 4.4.3. S. cerevisiae Positioned at the opposite end of the ATP hydrolysis spectrum is the eukaryotic Rad51 protein. Although Rad51 protein does not require ATP hydrolysis at any step of DNA strand exchange in vitro or in vivo (64), it does hydrolyze ATP with a k_cat of 0.73 min^-1 (64). Rad51 protein is able to promote both intermediate formation and directional DNA strand exchange using either ATP-g-S or AMP-PNP, two non-hydrolyzable analogues of ATP (64). In addition, a Rad51 mutant protein, K191R, (the analogue of the RecA K72R mutant protein), is defective for ATPase activity in vitro, yet it can promote polar DNA strand exchange (64). Interestingly, the mutant Rad51 protein is phenotypically normal when expressed in yeast cells, demonstrating that ATP hydrolysis is unnecessary for Rad51 protein function (64). This raises the question of why such a large difference exists between these three proteins in the rates of ATP hydrolysis during DNA strand exchange? Since ATP hydrolysis is not needed for DNA strand exchange, at least two possible explanations exist. First, there may be a relationship between the requirement for ATP hydrolysis and the genetic complexity of the organism. Such a relationship would exist if the presynaptic filament was required to persist for the longer times needed to find DNA sequence homology in the more complex organism: those with the most complex genome would require the more stable presynaptic complex which would exist if the ATP turnover were lower. In accord with this idea, T4 phage which has the least complex genome, has the highest rate of ATP hydrolysis; E. coli which has an intermediate complexity, has a corresponding intermediate rate of ATP hydrolysis; and S. cerevisiae, which has a more complex genome, has little or no ATP hydrolysis. Alternatively, the requirement for ATP hydrolysis may be related to the length of the organism�s cell cycle: T4 undergoes rapid growth after infection; E. coli has a doubling time of 20 to 30 minutes; whereas S. cerevisiae has a doubling time ranging from 90 to 140 minutes. Since disassembly of the presynaptic complex is also presumably important to a cell's metabolism, the cells with the fastest growth rate should have a pairing protein with the highest ATP turnover. Both views may be correct: since it should take a longer time to identify homology in a more complex genome, it is reasonable to expect that the presynaptic filament should be kinetically more stable. Furthermore, since ATP hydrolysis is also used to remove the DNA strand exchange protein from the DNA heteroduplex product, the more slowly growing organism may simply have the luxury of waiting for a slower dissociation process.

[Frontiers in Bioscience 3, d570-603, June 17, 1998]
Reprints
PubMed
CAVEAT LECTOR

DNA STRAND EXCHANGE PROTEINS: A BIOCHEMICAL AND PHYSICAL COMPARISON

Piero R. Bianco ¹, Robert B. Tracy ^1,2,3and Stephen C. Kowalczykowski ^1,2

1Sections of Microbiology and of Molecular and Cellular Biology, ²Microbiology Graduate Group, University of California, Davis, CA 95616, ³ Norris Comprehensive Cancer Center, USC School of Medicine, 1441 Eastlake Ave., Los Angeles, CA 90033

Received 5/24/98 Accepted 6/11/98

4. MECHANISM OF DNA STRAND EXCHANGE

As is evident from the preceding, RecA protein and its analogues mediate a reaction, DNA strand exchange, that is central to homologous recombination. The enzymatic properties of the in vitro strand exchange activity of RecA protein have been studied in great detail (4,28,170) and a summary of the biochemical properties of the DNA strand exchange proteins discussed herein is shown in table 2. Below, we detail the steps of protein-promoted DNA strand exchange.

DNA strand exchange catalyzed by RecA protein occurs by a number of kinetically distinct phases that can be subdivided into at least three experimentally distinguishable steps: (1) presynapsis, (2) synapsis and (3) DNA heteroduplex extension (figure 3). Presynapsis is the step where RecA protein assembles onto the ssDNA to form the nucleoprotein species that is active in the homology search. Synapsis is characterized by initially random non-homologous contacts occurring between the presynaptic complex and naked dsDNA (termed coaggregates), the search for homology, homologous pairing, and finally conversion to either paranemic or plectonemic joint molecules (discussed below in the synapsis section). DNA heteroduplex extension, as discussed earlier, is the polar migration of the nascent DNA heteroduplex joint.

Figure 3. Kinetic steps of DNA strand exchange. The reaction between circular ssDNA and linear dsDNA is shown. The product is a nicked circle and a displaced single strand of DNA. A similar reaction scheme likely applies to all of the DNA strand exchange proteins discussed but reaction polarity is not indicated; the illustration is adapted from Kowalczykowski (185). The steps shown are: 1-presynapsis; 2 and 3-synapsis; and 4-DNA heteroduplex extension. The green spheres represent the strand exchange protein; the orange squares represent ssDNA binding proteins.

DNA strand exchange occurs between two DNA molecules provided that the following criteria are met: (1) one of the DNA molecules contains a region of ssDNA; (2) the ssDNA region occurs at a site homologous to the other dsDNA molecule; and (3) for topological reasons, one of the substrate molecules has an end. A reaction that has become a model for mechanistic studies involves the exchange between circular ssDNA (f; X174 or M13 phage) and linear dsDNA (figure 3). The products are nicked, circular dsDNA and linear ssDNA (171,172).

4.1 Presynapsis (figure 3, Step 1)

Presynapsis is the ordered assembly of the RecA protein homologue on ssDNA to produce a nucleoprotein complex that is the active species in DNA strand exchange (filaments formed on dsDNA are shown in figure 5 and discussed in Section V). This process occurs by either of two pathways: first, the single-stranded DNA binding protein (SSB; RPA; UvsY and/or gp32) binds to ssDNA (figure 3, step 1A); this is followed by the subsequent binding of the DNA strand exchange protein (RecA; UvsX or Rad51) and displacement of the single-stranded DNA binding protein to form the presynaptic filament (figure 3, steps 1B and 1C, left panel). In the second pathway, the DNA-strand exchange protein binds to the ssDNA substrate, forming an incomplete nucleoprotein filament due to limitations imposed by DNA secondary structure (step 1A, right panel). The single-stranded DNA binding protein binds, removes the secondary structure and is then displaced by further binding of the DNA strand exchange protein (steps 1B and C). The net result of either of these two pathways, is the formation of a complete nucleoprotein filament that involves direct contact by the primary DNA-binding site of the strand exchange protein, and results in the generation of a contiguous secondary DNA-binding site that is essential for homologous pairing and DNA strand exchange (173,174).

4.1.1. Escherichia coli

Filament formation by RecA protein is dependent on DNA length, DNA composition, and the amount of secondary structure within the ssDNA substrate (95,175-178). The ATP-bound form of the protein is needed to constitute the active form of the nucleoprotein filament. ATP greatly stabilizes the complex, and transforms RecA protein into a DNA-binding state with high affinity for ssDNA (120). The RecA nucleoprotein filament is capable of hydrolyzing ATP with a k_cat of 18-30 min^-1, a value that is affected by ssDNA length and sequence (117,177-179). However, DNA serves only as a scaffold to promote filament formation since RecA protein can be activated by high salt concentrations to form both filaments and to hydrolyze ATP (180-182). Though RecA protein is an ATPase, ATP hydrolysis is not essential for DNA strand exchange (64,183,184). Stoichiometric amounts of RecA protein, relative to ssDNA concentrations, are required for filament formation, with maximum DNA strand exchange rates occurring when filaments are formed at a ratio of at least 1 RecA monomer to 3 nucleotides of ssDNA. This stoichiometric requirement indicates the importance of a saturated, contiguous RecA protein-ssDNA filament in the DNA strand exchange process (185).

In addition to ATP, ATP-g-S and dATP also support presynaptic filament formation and are capable of transforming RecA protein into the high affinity ssDNA-binding state, the active form of the protein (179,186). This complex, the presynaptic filament, is the physiologically relevant species in RecA protein-mediated cellular processes. Filament assembly onto ssDNA occurs in a 5'®;3' direction (187), which is the same direction as branch migration. The resulting filament has a regular structure and the prominent feature is a large helical groove, with 6.2 monomers per helical turn and each monomer interacting with 3 bases of DNA (figure 5) (188).

The SSB protein effects the binding of RecA protein to ssDNA, greatly stimulating the DNA strand exchange process (117,189). The role of SSB protein in presynapsis is to remove secondary structure from ssDNA, which is inhibitory to the formation of the saturated presynaptic complex (117,119-122). This is consistent with the role of the protein as a helix-destabilizing protein and the observation that other helix destabilizing proteins can substitute for SSB protein in DNA strand exchange in vitro (119).

4.1.2. Bacteriophage T4

Presynaptic complex formation by the T4 phage proteins is similar, but features an added complexity, the UvsY protein. During presynapsis, UvsX binds to ssDNA cooperatively with a stoichiometry of one UvsX monomer for every 3 to 5 nucleotides of ssDNA to form a nucleoprotein filament (46,47). In contrast to RecA protein, the binding of ATP does not stabilize the UvsX protein-ssDNA presynaptic filament under steady-state conditions for hydrolysis of ATP; however, the binding of ATP-g-S does (190). This suggests that ATP-g-S increases the equilibrium binding affinity of UvsX protein for ssDNA which is very similar to what is observed for RecA protein (179,191). However, the UvsX protein-ssDNA complex is more vigorous in hydrolyzing ATP (the rate of ATP hydrolysis is 8 to 10-fold higher) (44), making its presynaptic filament more dynamic; this may explain why ATP does not stabilize the UvsX-presynaptic complex, (i.e., the UvsX-ssDNA-ATP complex is short lived) and why an accessory protein like UvsY protein is needed. In addition, UvsX protein is unique among RecA protein analogues studied to date, in that it produces both ADP + P_i and AMP+ PP_i (44). The significance of the AMP production is presently unknown, however.

The formation of the UvsX protein presynaptic filament is facilitated by the presence of both gp32 and UvsY protein (a UvsX-dsDNA nucleoprotein filament is shown in figure 5B and is discussed in Section V). Gp32 enhances filament formation by removing secondary structure native to the ssDNA (45,51). UvsY facilitates filament formation by stabilizing the presynaptic complex itself (51,192,193). Although gp32 is the T4 phage counterpart to the E. coli SSB protein, there is no structural homologue of the UvsY protein in the E. coli recombination system; however, the RecF, O, and R proteins may comprise a functional homologue (123,124,194-198).

4.1.3. S. cerevisiae

Presynaptic filament formation by Rad51 protein results in a stoichiometric nucleoprotein filament that resembles the filament formed by both the RecA and UvsX proteins (figure 5C, discussed in Section V) (63). This complex is the active species in DNA strand exchange (63). Optimal reaction rates are achieved with a stoichiometry of 3 nucleotides per Rad51 monomer (63,199). Binding to ssDNA is stabilized by either ATP (59) or the non-hydrolyzable analogues of ATP, ATP-g-S and AMP-PNP (64). Rad51 protein filament formation is stimulated by the RPA heterotrimer (63,65), which is the eukaryotic analogue of the E. coli SSB and T4 gp32 proteins. In contrast to both the RecA and UvsX proteins, very little ATP is hydrolyzed by the Rad51 nucleoprotein complex (k_cat = 0.73 min^-1) during presynapsis or during DNA strand exchange (64); interestingly, some of the enzymatic characteristics of Rad51 protein more closely parallel those of a mutant RecA protein, RecA K72R, which has a reduced level of ATP hydrolysis (~ 600-850-fold reduced to a k_cat of 0.032min^-1) yet nonetheless, is able to promote DNA strand exchange.

4.2 Synapsis (figure 3, Steps 2, 3)

Once the presynaptic filament has assembled on ssDNA, synapsis ensues. In this stage of the reaction, a dsDNA molecule must be bound to the filament, homology to the ssDNA within the filament located within the dsDNA, and a plectonemic heteroduplex joint formed. For this phase of DNA strand exchange to occur, a second DNA molecule must bind in a sequence-independent fashion to the secondary DNA binding site of the nucleoprotein filament; presumably, the close approach of this second DNA molecule is facilitated by shielding of the negatively charged phosphodiester backbones of the two DNA molecules by the DNA strand exchange protein. Once dsDNA is bound, the search for homology takes place. The search is rapid, and requires that the binding of dsDNA to the secondary site be both weak and transient. The recognition of homology takes place when the ssDNA within the presynaptic filament hydrogen bonds, via non-Watson-Crick base pairing, to either the major or minor groove of the bound dsDNA, in a mechanism that does not require triplex intermediate formation. This alignment provides a signal to the strand exchange protein that homology has been located; base switching occurs, and the heteroduplex dsDNA occupies the primary site whereas the displaced strand occupies the secondary site. If the location of homology occurs in the center of the dsDNA, a paranemic joint forms; if this occurs at the dsDNA molecule ends, a plectonemic joint forms. Due to their inherent instability, paranemic joints must be converted to plectonemic joints in order to survive and become recombinant DNA structures.

4.2.1. Escherichia coli

Once an active presynaptic complex forms, the complex can then pair the ssDNA within the filament to dsDNA. On the basis of probability, this initial contact must be at a site of non-homology. Indeed, RecA protein is capable of forming a complex between ss- and dsDNA, or between ssDNA molecules with little or no sequence complementarity (173,200-203). These complexes, called coaggregates, contain many DNA molecules arranged together in a three-dimensional network that can be sedimented in a low speed centrifuge (204,205). They form rapidly, and kinetically precede joint molecule formation. Thus, it has been inferred that coaggregates, or a related complex, of RecA protein and ss- and dsDNA are intermediates on the pathway of the strand exchange reaction (204,205).

Although RecA protein binds ssDNA in a sequence-independent manner (though compositional preferences exist (95,176,206,207)), the problem of the homology search is analogous to the problem that sequence-specific DNA binding proteins face in locating their target sequence (208). However, for the case of RecA protein and its analogues, the scale of the search problem is much larger: the binding protein is the entire nucleoprotein filament (which can consist of thousands of protein monomers, depending on the ssDNA length), and the target is the unique complementary sequence within the entire genome. Although the entire nucleoprotein filament is involved in the search for homology, the minimum length of homology required for recognition is as low as 15 nucleotides in vitro (95), which is somewhat less than that needed in vivo, where homologous recombination requires minimally, about 23-40 bp of homology (209,210).

For homologous alignment to occur, a second DNA molecule must bind to the secondary DNA binding site of the RecA protein filament. This binding, although not sequence specific, does have a hierarchy: optimal binding occurs with oligopyrimidinic DNA while poorer binding is observed with oligoadenylic DNA (173). This hierarchy of binding is not due to the sequence of the ssDNA present in the primary site, but rather is due to an intrinsic property of RecA protein (173). Thus, the proposal that homologous recognition occurred by a novel form of non-Watson-Crick base-pairing called "self-recognition" (211,212) is not substantiated (173,174).

How then does RecA protein "sense" when homology has been located? Though many details of the homology search remain unknown, several steps are clear. During the search, the binding of a second DNA molecule to the nucleoprotein filament must, by necessity, be both weak and transient, to facilitate a rapid search. When the homologous locus is found, the dsDNA is unwound, and one of the strands of the duplex pairs with its complement (which is bound to the primary, original DNA-binding site of the pairing protein). The other strand is displaced into the secondary DNA-binding site, where it is bound more tightly than the duplex DNA. During the search for homology, the DNA duplex remains base-paired; it is only when homology is located, that strand switching occurs. The recognition of homology between the ssDNA within the filament and either the minor groove of dsDNA (213-215), or the major groove of the dsDNA, provides the signal to RecA protein that homology has been located (173). Immediately after base-pair switching, the heteroduplex dsDNA product would occupy the primary site, whereas the displaced ssDNA would occupy the secondary site; there is evidence to support this scenario (174,214,216,217).

Once the two DNA molecules are homologously aligned, RecA protein then catalyzes the nascent exchange of strands. In the prototypical in vitro reaction depicted in figure 3, this requires a local denaturation of the dsDNA molecule and the subsequent exchange of the identical single strands of DNA. These steps may be simultaneous or separated in time, but the result is the production of an intermediate known as a joint molecule. Two types of joint molecules may form, depending on the topological constraints of the DNA: either paranemic or plectonemic. A paranemic joint is one in which the individual complementary strands do not intertwine, producing a molecule that is base-paired but not topologically linked; whereas, a plectonemic joint is one in which the incoming single strand is intertwined around its complement as in native dsDNA. In the reaction displayed in figure 3, paranemic joints will form in the interior of the duplex substrate, and plectonemic joints will form at the ends of the duplex substrate. An experimental distinction between paranemic and plectonemic joints is that paranemic joints require RecA protein (or its counterpart) for stability whereas plectonemic do not. Since paranemic joints are statistically more probable than plectonemic, paranemic molecules are likely intermediates on the reaction pathway to the formation of more stable plectonemic joint molecules (218,219).

Both the type and efficiency of joint molecule formation are affected by SSB protein. If SSB protein is omitted, the displaced ssDNA is used by RecA protein to form a second joint molecule with another dsDNA molecule; this results in the formation of complex, homology-dependent networks of joint molecules. In the presence of SSB protein, RecA protein does not form these networks, since SSB protein prevents re-invasion events by binding to the ssDNA displaced from the joint molecules; in addition, the yield of joint molecules is greater (220). However, under reaction conditions where RecA protein is better able to compete with SSB protein (i.e. in the presence of the volume excluding agents polyethylene glycol or polyvinyl alcohol; or when dATP is used as cofactor), networks are readily formed (186,221,222). Thus SSB protein acts both at the pre- and post-synaptic steps of DNA strand exchange.

4.2.2. Bacteriophage T4

Some aspects of synapsis are similar for UvsX protein and RecA protein: 1) UvsX protein coaggregates non-homologous ssDNA and dsDNA (47); and 2) UvsX protein catalyzes formation of approximately equal numbers of paranemic and plectonemic joint molecules (223). Differences also exist between the two proteins. First, while both RecA protein and UvsX protein form homology-dependent DNA networks in the absence of a single-stranded DNA-binding protein, UvsX protein also forms networks even in the presence of gp32, but this may simply reflect an increased ability of UvsX to displace gp32 (50). Second, since UvsX protein, unlike RecA protein, binds readily to either ssDNA or dsDNA under normal strand exchange conditions (46), joint molecule formation is reduced due to UvsX protein binding to the dsDNA (223); this behavior more closely resembles that of Rad51 protein (see below). This limitation is overcome in vitro by using limiting amounts of UvsX protein relative to the DNA (45).

4.2.3. S. cerevisiae

As for UvsX protein, joint molecule formation is inhibited if Rad51 protein is allowed to bind dsDNA and can be avoided by assembling presynaptic filaments using exactly stoichiometric amounts of Rad51 protein, or by using excess dsDNA (63). This inhibition of joint molecule formation results from the rapid and kinetically stable binding of Rad51 protein to dsDNA, a situation that is very different from the behavior of RecA protein.

Rad51 protein initiates joint molecule formation if the linear dsDNA contains a single-stranded overhang at least 2 nucleotides in length that is complementary to the ssDNA within the presynaptic filament (199). Both joint molecules and nicked circle products are reported to form more efficiently with dsDNA containing a 3'-overhang than with a 5'-overhang (199).

4.3 Branch migration (figure 3, Step 4)

Once the plectonemic joint has formed, the branch migration phase of DNA strand exchange commences. During this phase, the nascent, heteroduplex joint is extended until complete exchange of single strands of DNA occurs, resulting in a nicked, double-stranded circle. Though kinetically distinct, branch migration may not be a mechanistically separate step, but rather may represent a continuation of plectonemic joint molecule formation (219,224).

4.3.1. Escherichia coli

For RecA protein, branch migration proceeds in a 5'®;3' direction relative to the incoming single strand (the same direction as RecA protein polymerization), at a rate of 2-10 bp sec^-1 (171,225), requires ATP hydrolysis (19,226) and induces torsional stress in the dsDNA (227,228). Branch migration is relatively tolerant of nucleotide sequence mismatches (229,230) and DNA lesions (231,232), although the reaction is inhibited 60% by heterologous insertions of 140 bp within the dsDNA substrate and by insertions of 1000 nucleotides within the ssDNA substrate (230,231).

The ability of RecA protein to not only promote DNA heteroduplex extension, but also to extend the DNA heteroduplex beyond limited regions of sequence non-homology (heterology), may serve a valuable biological function, despite the existence of specialized branch migration helicases like the RuvAB proteins (233). For this reason, the mechanism by which RecA protein facilitates the bypass of DNA sequence heterology is an important issue. Several studies established the key requirements for this heterology-bypass (234,235) (D.A. Kitchell and S.C.K., unpublished observations). RecA protein can facilitate bypass through a heterology as large as 140 bp within the dsDNA substrate (230,231), provided that these mismatched sequences are beyond the point at which DNA strand exchange initiates (i.e., the heterology needs to be beyond the 5'-end of the strand being displaced from the dsDNA). However, RecA protein is unable to bypass an insert of only 22 base pairs that is at the distal end of the dsDNA (234). In all cases, the ability of RecA protein to promote DNA heteroduplex extension progressively diminishes as the length of homology distal to the heterology decreases. Furthermore, when the length of the heterology increases, the amount of distal homology required to bypass this heterology increases, apparently to compensate for the longer heterologous distance that must be traversed. Thus, to bypass a heterology, DNA strand exchange must initiate in a region of DNA sequence homology, and there must be additional homology beyond the heterology. ATP-g-S could not substitute for ATP in these bypass experiments, demonstrating that ATP hydrolysis was necessary, presumably to permit some dissociation/redistribution of the RecA nucleoprotein filament.

Several mechanisms were proposed to explain these important characteristics. One envisioned that once the heterology was encountered, lateral slippage of the exchanging strand within the RecA nucleoprotein filament would permit re-alignment of the homologous sequences (234). A second mechanism envisioned that facilitated rotation of the DNA substrates about each other mediated by the RecA protein filament in an ATP-hydrolysis coupled mechanism, provided the "motor" that rotated the DNA past the heterology; the specific details of this model were elegantly elaborated previously to which the reader is referred for details (235). Here we would like to present an alternative idea, based on these collective data and unpublished considerations (D.A. Kitchell and S.C.K., unpublished observations).

Figure 4 shows how topological stress, resulting from local unwinding of dsDNA that is constrained in a closed topological domain, can facilitate bypass through a heterology. The basic premise of the topological model is that, homologous pairing involves intermediates (the synaptic complexes) in which the dsDNA is highly unwound. Furthermore, in the presence of ATP, the RecA nucleoprotein filament that participates in DNA strand exchange is dynamic, and it is frequently binding and releasing dsDNA in its quest for proper DNA sequence alignment. Electron microscopy shows that the RecA nucleoprotein filament can envelop kilobase pair-lengths of homologous dsDNA (216), and that the binding of dsDNA by the RecA nucleoprotein filament produces a topological unwinding of the dsDNA from the canonical 10.4 bp/turn to 18.6 bp/turn (236-241,242 (figure 4)). Thus, in the static situation, when the RecA nucleoprotein filament envelops dsDNA, the dsDNA including the heterology is extensively unwound by up to 8.2 bp/turn resulting in 18.6 bp of dsDNA/filament turn versus 10.4 bp turn in B-form dsDNA (figure 4C). However, in the dynamic case, in the presence of ATP, if the RecA nucleoprotein filament were to release a medial region of the dsDNA while the two flanking regions of the dsDNA remained bound to the filament, then the binding energy that maintained the bound dsDNA in its unwound form would be absent, and the released dsDNA segment would be both topologically constrained and highly underwound (figure 4D, left side); in fact, the release of dsDNA from just 2 turns of the filament would result in 16.4 bp of unwinding. Thus, the transient release of dsDNA from 20 turns of the filament would produce a domain that is unwound by 164 bp relative to B-form DNA. If this unwinding occurred in a region of dsDNA heterology, then there would be sufficient, locally trapped unwinding to induce transient strand separation of the dsDNA (figure 4D, right side), and to bypass the heterology in a kinetically stochastic way. With the region of dsDNA heterology locally melted, bypass would readily occur upon subsequent re-polymerization of the filament (figure 4E); however, if a nick were located in the region of heterology, then the superhelical strain is released and no bypass occurs. This model explains the key experimental requirements for bypass (the requirement for increasingly longer lengths of homology distal to the heterology, the need for ATP, and the inhibitory effect of a dsDNA nick), and it lends insight into how RecA protein, and presumably all of the other DNA strand exchange proteins (since they all extend dsDNA when bound) can mediate bypass of short regions of DNA sequence heterology without the need for specialized branch migrations proteins.

Figure 4. Model for heterology bypass by RecA protein. A pairing event involving ssDNA (drawn in blue) and dsDNA (drawn in red/pink) containing a region of heterology, is shown. The spheres represent RecA protein; the large black arrow indicates the direction of migration of the heteroduplex joint. Details of this model are discussed in the text.

4.3.2. Bacteriophage T4

UvsX protein, like RecA protein, catalyzes branch migration such that DNA strand displacement is 5�®;3� relative to either the incoming presynaptic filament or the displaced strand (45,50). However, UvsX protein catalyzes branch migration at a rate of 12-15 bp sec^-1 (50,151), which is faster than the rate catalyzed by RecA protein (see above).

4.3.3. S. cerevisiae

In contrast to RecA and UvsX proteins, Rad51 protein promotes strand exchange with the opposite polarity (i.e. 3�®;5� relative to the incoming single strand (63)) at a rate of 0.5 bp sec^-1, and has no requirement for ATP hydrolysis at any stage (64). In addition, the rate of formation of both joint molecule intermediates and products in the reaction is significantly reduced compared to that of either the RecA (2-10 bp sec^-1) or UvsX proteins (12-15 bp sec^-1) (62-64). Also, this stage of the reaction is inhibited by a 458 bp heterologous insertion in the dsDNA substrate (63).

4.4. ATP hydrolysis in DNA strand exchange

DNA strand exchange requires the presence of a nucleoside triphosphate cofactor, usually ATP. Under standard conditions, ATP hydrolysis coincides with the pairing and exchange of strands of DNA. Initially, it was thought that ATP hydrolysis was a requirement for DNA strand exchange. It is now known that neither the hydrolysis of ATP (183,226) nor the presence of a high-energy phosphate bond (184) are necessary for DNA strand exchange to occur. It is the formation of the high-affinity ssDNA-binding state of RecA protein that is necessary for DNA strand exchange; the formation of which is brought about by ATP binding to RecA protein bound to DNA. The binding of the non-hydrolyzable analogue of ATP, ATP-g-S, or the non-covalent complex of ADP-AlF^-₄, can also induce the high-affinity DNA-binding state of RecA protein. Although the hydrolysis of ATP is not required for the exchange of DNA strands, it is required at phases of DNA strand exchange that require the dissociation of RecA protein that is induced by ADP, the product of ATP hydrolysis.

4.4.1. Escherichia coli

RecA protein is a DNA-dependent ATPase with a single active site for the binding and hydrolysis of ATP and other nucleoside triphosphates (17,206,243-245). The enzyme is able to hydrolyze ATP with either ss- or dsDNA as cofactor, resulting in a k_catwith ssDNA of 18 - 30 min^-1 and with dsDNA of 19 to 22 min^-1 (178,246,247). The active species in ATP hydrolysis is the nucleoprotein filament, with ATP being hydrolyzed uniformly throughout the filament and with no detectable enhancement at filament ends (117,177,248). Hydrolysis of ATP results in one or more conformational changes in RecA protein. ATP and ADP serve to modulate RecA between the "high-affinity" DNA-binding and "low-affinity" DNA-binding states, respectively (179,191). It may not be not the hydrolysis per se that modulates the two states, but rather the release of P_i, which is extremely rapid (249). This proposal is supported by experiments utilizing the ground state analogue, ADP-AlF^-₄, which maintains RecA protein in the high-affinity ssDNA-binding state in the absence of a high-energy bond (184,250).

Although ATP hydrolysis occurs throughout DNA strand exchange promoted by RecA protein, it is not required for all of the stages (19,184,219,226,251-254). ATP hydrolysis is not required for the exchange of DNA strands (184,250,255). However, ATP hydrolysis is required to dissociate RecA protein from the heteroduplex products once the reaction is complete (226,252,256), to facilitate the bypass of structural barriers such as heterologous sequences (253,254), and to maintain reaction polarity (125,257). In addition, ATP hydrolysis is required for DNA strand exchange involving four strands (258).

Wild-type RecA protein is able to promote strand exchange using the non-hydrolyzable analogue ATP-g-S (226). In addition, a mutant RecA protein, K72R, in which ATP hydrolysis is reduced by approximately 600-850-fold, is able to promote homologous pairing and exchange of up to 1.5 kilobase pairs of DNA (183). In both of these situations, extension of the heteroduplex is blocked (i.e. joint molecule intermediates form but are not converted to gapped or nicked dsDNA products) and the reaction is bi-directional (257), which is in contrast to that of the ATP reaction, which has a distinct polarity (171). However, a nucleoside triphosphate is not an essential component of the pairing and exchange process, because a complex of ADP and AlF^-₄activates all of the activities of RecA protein (184,250,255).

4.4.2. Bacteriophage T4

UvsX protein, in contrast to RecA protein, has an absolute requirement for ATP hydrolysis in DNA strand exchange (259). The rate of ATP hydrolysis by UvsX protein is 8- to 10-fold higher than that of RecA protein, with a k_cat of 145 - 240 min^-1 (44), and this activity is inhibited by both ADP and ATP-g-S (259). In the presence of 2 mM ATP- g-S, UvsX protein is unable to promote strand exchange; however at 16 and 80 mM ATP-g-S, UvsX protein can form intermediates (259). This may simply reflect the situation that existed for RecA protein prior to the realization that the ATP-g-S reaction had a unique reaction optimum (184). Substitution of ATP-g-S for ATP allows UvsX protein to bind ssDNA, unwind dsDNA (259), and promote DNA strand annealing (45). Addition of ATP-g-S to an ongoing DNA strand exchange reaction, initiated with ATP, causes a transient increase in the rate of branch migration (50). This transient stimulation, which occurs due to inhibition of ATP hydrolysis, is attributed to a kinetic stabilization of the active filamentous form of UvsX protein. The high turnover of the UvsX protein nucleoprotein filament appears to limit the steady-state rate of branch migration (259).

4.4.3. S. cerevisiae

Positioned at the opposite end of the ATP hydrolysis spectrum is the eukaryotic Rad51 protein. Although Rad51 protein does not require ATP hydrolysis at any step of DNA strand exchange in vitro or in vivo (64), it does hydrolyze ATP with a k_cat of 0.73 min^-1 (64). Rad51 protein is able to promote both intermediate formation and directional DNA strand exchange using either ATP-g-S or AMP-PNP, two non-hydrolyzable analogues of ATP (64). In addition, a Rad51 mutant protein, K191R, (the analogue of the RecA K72R mutant protein), is defective for ATPase activity in vitro, yet it can promote polar DNA strand exchange (64). Interestingly, the mutant Rad51 protein is phenotypically normal when expressed in yeast cells, demonstrating that ATP hydrolysis is unnecessary for Rad51 protein function (64).

This raises the question of why such a large difference exists between these three proteins in the rates of ATP hydrolysis during DNA strand exchange? Since ATP hydrolysis is not needed for DNA strand exchange, at least two possible explanations exist. First, there may be a relationship between the requirement for ATP hydrolysis and the genetic complexity of the organism. Such a relationship would exist if the presynaptic filament was required to persist for the longer times needed to find DNA sequence homology in the more complex organism: those with the most complex genome would require the more stable presynaptic complex which would exist if the ATP turnover were lower. In accord with this idea, T4 phage which has the least complex genome, has the highest rate of ATP hydrolysis; E. coli which has an intermediate complexity, has a corresponding intermediate rate of ATP hydrolysis; and S. cerevisiae, which has a more complex genome, has little or no ATP hydrolysis. Alternatively, the requirement for ATP hydrolysis may be related to the length of the organism�s cell cycle: T4 undergoes rapid growth after infection; E. coli has a doubling time of 20 to 30 minutes; whereas S. cerevisiae has a doubling time ranging from 90 to 140 minutes. Since disassembly of the presynaptic complex is also presumably important to a cell's metabolism, the cells with the fastest growth rate should have a pairing protein with the highest ATP turnover. Both views may be correct: since it should take a longer time to identify homology in a more complex genome, it is reasonable to expect that the presynaptic filament should be kinetically more stable. Furthermore, since ATP hydrolysis is also used to remove the DNA strand exchange protein from the DNA heteroduplex product, the more slowly growing organism may simply have the luxury of waiting for a slower dissociation process.