All chemicals used were reagent grade and aqueous solutions were made with water purified through a Barnstead system. Buffer A used in quantitative DNA–cellulose chromatography experiments with UvsY and its mutants contained 20 mM Tris–HCl (pH 7.4) and variable concentrations of NaCl. Buffer B used in analytical ultracentrifugation experiments contained 20 mM Tris–HCl (pH 7.4), 1 mM MgCl₂ and variable concentrations of NaCl. ssDNA– and dsDNA–cellulose resins were prepared from salmon sperm DNA (Sigma) and Whatman CF-11 cellulose fibers as described (26). Two different batches of ssDNA–cellulose used in these studies contained 1.7 and 2.0 mg of total immobilized DNA per packed milliliter of resin, respectively, as determined by the release of A₂₆₀ units upon boiling (26). By the same method, two different batches of dsDNA–cellulose used in these studies were shown to contain 0.6 and 1.0 mg, respectively, of total immobilized DNA per packed milliliter of resin.

Purification and storage conditions for T4 UvsY, UvsY_K58A and UvsY_K58A,R60A proteins were as described previously (23,27). The purity of each protein stock was >98% based on SDS–PAGE gels stained with Coomassie brilliant blue. All protein stocks were determined to be nuclease-free according to published criteria (28). The concentrations of protein stocks were determined by the absorbance at 280 nm, using an extinction coefficient of 19 180 M⁻¹ cm⁻¹ for UvsY wild-type and mutant proteins (29). HPLC-purified oligonucleotides (dT₄, dT₈, dT₂₄, dT₂₅) were purchased from Operon Biotechnologies and quantified by the absorbance at 260 nm using molar extinction coefficients provided by the manufacturer. The dT₂₄ oligonucleotide was 5′-end labeled with [³²P] using T4 polynucleotide kinase (New England Biolabs), and purified by ethanol precipitation.

The sequence non-specific, intrinsic DNA-binding affinities of UvsY wild-type and mutant proteins for native ssDNA and dsDNA lattices were determined using quantitative DNA–cellulose affinity chromatography (30). Experiments were carried out at very low binding densities of protein on nucleic acid, which allows direct measurements of intrinsic association constants independent of the binding site size and cooperativity parameters of the proteins (31). The ssDNA– and dsDNA–cellulose resins were packed into 2 ml columns and pre-equilibrated with 2 ml buffer A containing different fixed concentrations of NaCl at a flow rate of 40 ml/h. Hexamers (0.67 nmol) of either UvsY, UvsY_K58A or UvsY_K58A,R60A in 2 ml of the same buffer were loaded onto the columns by gravity (UvsY hexamers/total nucleotide residues molar ratio ≈10⁻⁴). The flow was then stopped and the loaded columns allowed to sit for 10–15 min, which was sufficient to achieve an initial binding equilibrium since doubling the loading time had no effect on results. Thereafter, 20 ml of the same buffer followed by 10 ml of buffer A plus 1 M NaCl was used to elute proteins from columns, with a flow rate of 40 ml/h while 1 ml fractions were collected. The protein concentration in each fraction was quantified by intrinsic tryptophan fluorescence (excitation wavelength: 295 nm; emission wavelength: 340 nm) on a Quantamaster QM6 fluorometer (Photon Technology International, South Brunswick, NJ). To obtain the best signal/noise ratio, the excitation/emission bandpass was set at 2 nm/3 nm. Fluorescence data points represented an average of 20 fluorometer readings and were compared to calibration curves to obtain protein concentrations. All chromatography steps and fluorescence measurements were carried out at room temperature (typically 23°C).

Chromatography elution profiles are presented in the following format (30,31): log P_ci versus fraction number, in which P_ci denotes the percentage of protein remaining on columns and is calculated by the protein quantity in each fraction and the total protein quantity. This type of plot typically yields a linear relation of data points and the slope is equal to –log(1+k), in which k is a proportionality constant. After obtaining this proportionality constant for each data set, the association constant of protein to DNA at a given NaCl concentration was calculated according to the following equation: (1) where K is the intrinsic association constant of protein to DNA; V_f is the fraction volume; and D_t is the total molar amount of protein-accessible DNA on the column, which was determined as follows (30,31): the 2 ml column resin was taken out and resuspended in 10 ml of DNaseI-cleavage buffer containing 20 mM Tris (pH 7.4), 10 mM MgCl₂ and 100 mM NaCl. DNaseI (2 mg) was then added to this mixture and incubated at 37°C for 4 h. The nucleotides released into solution were considered protein-accessible and determined by phosphate analysis as described (32). Briefly, 10 µl of sample solution was mixed with 30 µl of Mg(NO₃)₂ (10% dissolved in 95% ethanol) and ashed over flame. HCl (0.5N, 0.3 ml) was then added to hydrolyze pyrophosphate formed in the last step. After hydrolysis at 100°C for 15 min, 0.6 ml of ammonium molybdate (0.42% dissolved in 1 N H₂SO₄) and 0.1 ml of 10% ascorbic acid were mixed with the sample and incubated at 45°C for 20 min. Absorbance at 820 nm was then measured and converted to the quantity of nucleotide using a conversion factor of 0.260 absorbance unit = 0.01 µmol of nucleotides. The total amounts of protein-accessible DNA in ssDNA– and dsDNA–cellulose columns were determined to be 7.2 ± 0.5 µmol and 2.3 ± 0.2 µmol of nucleotide residues, respectively. Thus the binding density of UvsY hexamers per protein-accessible nucleotide residue was ≤ 3 × 10⁻⁴ in all experiments.

Sedimentation equilibrium and velocity studies were carried out as previously described (22). All experiments were performed in a Beckman Optima XL-I analytical ultracentrifuge (Beckman Coulter Inc., Fullerton, CA) equipped with both absorbance and Rayleigh interference optics. All experiments utilized a Beckman An50-Ti 8-hole rotor equipped with either two- or six-sector charcoal-filled epon centerpieces and sapphire windows. The rotor temperature was set between 20 and 25°C depending on the experiment. Buffers consisted of buffer B plus variable concentrations of NaCl. All solutions were sterile filtered prior to use.

The stoichiometry of UvsY-oligonucleotide complexes was evaluated in solution using a modification of the meniscus depletion sedimentation equilibrium method (33). Specific rotor conditions are selected that quantitatively remove the desired solute from the radial position immediately adjacent to the solution–air meniscus. For the case of hexameric UvsY protein, meniscus depletion is achieved at equilibrium after 17 h at 18 000 rpm, 22°C, in buffer B plus 300 mM NaCl, using a solution column height of ∼3.5 mm. Solutions containing both protein and oligonucleotide ssDNA were subjected to meniscus depletion sedimentation until equilibrium was achieved. In all cases, oligonucleotide ssDNA (dT₄, dT₈ or dT₂₅; 75 µM nucleotide residues) was present in large stoichiometric excess over UvsY (4.5 µM monomers) assuming a binding site size of n = 4 nucleotide residues/monomer (23). Control experiments were identical except that buffer contained 500 mM NaCl and lower concentrations of solutes (25 µM oligonucleotides and 1.5 µM UvsY). Evaluation of the state of equilibrium was performed as previously described (22). Radial absorbance measurements were collected at two wavelengths (260 and 280 nm) within a narrow radial interval corresponding to the meniscus depleted region. As the molecular weight of the UvsY hexamer (M_r ≈95 000) is very much greater than the largest oligonucleotide used in these studies (dT₂₅, M_r ≈7500), only the protein-containing species are able to achieve meniscus depletion. Because of this, the only observable absorbance within a narrow range immediately adjacent to the solution–air meniscus must result from absorbance due to free ssDNA species residing in the otherwise depleted region. This assumption is reasonable since sedimentation velocity data indicate that the UvsY hexamer is non-dissociable under identical conditions, either by itself or when bound to ssDNA (22). Therefore the UvsY hexamer represents the smallest protein species available in this controlled system.

Hydrodynamic modeling was performed using directly measured sedimentation and frictional coefficients (s and f, respectively) for all species (22). All hydrodynamic parameters were corrected to 20°C and zero concentration of sedimenting solute in water (34). Molecular masses of protein and nucleoprotein complexes were calculated from primary sequence data for UvsY (35) and oligonucleotides (Operon Technologies). The partial specific volume, , of the UvsY–ssDNA complex was calculated using a estimate for ssDNA of 0.55 cm³/g (T. Laue, personal communication) and weight-averaging the contribution of both components in the complex. This method makes the assumption that the of each component does not change upon association. Theoretical frictional coefficients, f₀, were calculated as described, using volumetric determinants obtained from macromolecular crystallography data (34,36). Calculation of f₀ depends on R_p, the ‘equivalent radius’, which is the expected radius of a rigid, impermeable and incompressible sphere having the identical volume as the species of interest. R_p and f₀ values were calculated from Equations (2) and (3), respectively: (2) (3) where η is the viscosity parameter. Note that f₀ calculated in this manner corresponds to the predicted minimum value of the frictional coefficient, and therefore results in a maximum estimation of the sedimentation coefficient, s, according to the Svedberg relationship: (4) where ρ equals density and N₀ is Avogadro’s number. The value of f/f₀ gives a rough approximation of the asymmetry of the solute species, indicating the deviation from an ideal sphere.

Structural models of associated oligomers were calculated using hydrodynamic simulation and the relationship between predicted sedimentation parameters of monomers and their oligomeric assemblies (37–40). Here the monomeric sedimentation coefficient is related to the n-state of oligomeric assembly by Equations (5) and (6): (5) where (6) In this relationship, s⁰_n is the sedimentation coefficient predicted for a rigid structure containing n identical spherical units, each with radius R₀ and sedimentation coefficient s⁰₁. R_ij is the radial distance between units, and therefore α_ij is determined by the assumed geometry of unit assembly. By designing various models that describe a state of n-assembly, theoretical sedimentation parameters can be determined.

5′-[³²P]-labeled dT₂₄ (2 µM nucleotides) was incubated with or without 0.5 µM UvsY at ambient temperature, in buffer A plus 300 mM NaCl for 30 min to achieve equilibrium. The above reaction solutions (20 µl) were spotted onto a paraffin wax sheet (ParaFilm) and subjected to UV exposure for the indicated times. UV light (254 nm) was applied maintaining ∼1 cm between the liquid drop and the UV source (hand-held UV lamp, UVP Inc.). After UV crosslinking, samples were brought to 0.02% SDS and heated to 100°C to fully disrupt non-covalent structures. Crosslinked species were separated by SDS–PAGE and visualized by autoradiography.

To establish the relationships between UvsY–ssDNA and UvsY–dsDNA interactions under native DNA conditions, we measured the binding of UvsY to isolated sites on relaxed, unmodified dsDNA and ssDNA molecules, using the quantitative DNA–cellulose chromatography method (30,31). Measurements were made at very low binding densities of protein on DNA (≤3 × 10⁻⁴ UvsY hexamers per protein-accessible nucleotide residue; see ‘Materials and Methods’ section). For proteins that bind to DNA with a cooperativity parameter ω, binding to isolated sites is observed when the binding density ν ≤ 1/10ω (31). UvsY binds non-cooperatively to ssDNA (ω ≈1) (23). Therefore the experimental conditions are consistent with UvsY binding to isolated, not contiguous sites on DNA, and the binding constants measured are intrinsic affinity values that do not depend on cooperativity or binding site size (31). The intrinsic affinity parameters of UvsY–dsDNA and UvsY–ssDNA interactions were measured as functions of salt. These studies provide the first quantitative details of UvsY–dsDNA and UvsY–ssDNA interactions under native DNA conditions, and allow direct comparisons of both under identical solution conditions.

The elution profiles of UvsY from dsDNA–cellulose and from ssDNA–cellulose are shown in Figure 1A and B, respectively, and the derived intrinsic association constants (K_ds and K_ss values) are listed in Table 1. The results demonstrate that UvsY binds to dsDNA strongly under relatively low-salt conditions (K_ds ≈10⁵–10⁷ M⁻¹ in 50–100 mM NaCl), however UvsY–dsDNA binding affinity decreases dramatically with increasing salt concentration, becoming quite weak at 200 mM NaCl and undetectable by this method above 300 mM NaCl (Figure 1A, Table 1). Figure 1.

Salt-dependent elution of UvsY protein from dsDNA– and ssDNA–cellulose columns. Elution profiles are presented in the format of Log P_ci versus fraction number as described in ‘Materials and Methods’ section. UvsY (4 nmol) was loaded onto and eluted from dsDNA–cellulose columns (A) or ssDNA–cellulose columns (B) in buffer A plus various concentrations of NaCl as denoted in the figure. The fractions were quantified by intrinsic tryptophan fluorescence (excitation 295 nm, emission 340 nm) and calibrated against known concentrations of UvsY. All data points are averaged values from three independent experiments and standard deviations are shown as error bars. At low NaCl concentrations, the error bars are too small to be visualized on the plots.

Missense UvsY_K58A and double missense UvsY_K58A,R60A mutant proteins have been characterized in our laboratory, and shown qualitatively to have decreased ssDNA-binding activity while retaining overall hexameric structure (18,27). To quantify their defects, the binding of both mutants to dsDNA– and ssDNA–cellulose columns was measured (Figures 2 and 3), and the derived intrinsic association constants are listed in Table 1. Figure 2A and B shows the elution profiles of UvsY_K58A from dsDNA– and ssDNA–cellulose columns, respectively. Figure 3A and B shows the elution profiles of UvsY_K58A,R60A from dsDNA– and ssDNA–cellulose columns, respectively. The intrinsic affinities of both UvsY mutants for both lattices are dramatically reduced compared to wild-type UvsY. K_ds values of both mutants can only be accurately determined at low salt (50 mM NaCl), and are at least 10⁴-fold lower than wild-type (Table 1). UvsY_K58A exhibits ∼5-fold higher affinity for both ssDNA and dsDNA than does UvsY_K58A,R60A in experiments performed at 50 mM NaCl. At the same salt concentration, the ratio K_ss/K_ds ≈14 for UvsY_K58A and ≈10 for UvsY_K58A,R60A. Therefore in both mutants, the intrinsic affinities for ssDNA are higher than those for dsDNA, which follows the same trend as wild-type UvsY. The UvsY_K58A mutant exhibits detectable binding to ssDNA–cellulose at salt concentrations up to 200 mM NaCl, but its K_ss value is projected to be at least 10³-fold lower than wild-type at 200 mM NaCl (Table 1; see also Figure 4). These observations are consistent with the weak binding of these mutants to etheno-modified ssDNA and to stretched dsDNA that was previously reported, and may explain their partial defects in stabilizing UvsX–ssDNA complexes and in stimulating UvsX enzymatic activities (13,18,24,27). Figure 2.

Salt-dependent elution of UvsY_K58A single mutant protein from dsDNA– and ssDNA–cellulose columns. UvsY_K58A (4 nmol) was loaded onto and eluted from dsDNA–cellulose columns (A) or ssDNA–cellulose columns (B) in buffer A plus various concentrations of NaCl as denoted in the figure. All other experimental details were identical to those reported in Figure 1.

Salt effects on K_ss and K_ds for UvsY–DNA interactions were determined from the slopes of plots of logK_ss or logK_ds versus log[NaCl] (Figure 4, Table 2). The slope dlog K/dlog[NaCl] is related to the number of ionic interactions involved in protein–DNA binding (41–44). The magnitude of dlog K_ss/dlog[NaCl] for UvsY–ssDNA interactions is very large, –13.9, indicative of highly electrostatic binding. In contrast, the magnitude of dlog K_ss/dlog[NaCl] for UvsY_K58A–ssDNA interactions is only –1.7, demonstrating a major disruption of electrostatic binding determinants caused by the loss of this single basic residue within each subunit of a UvsY hexamer (Figure 4, Table 2). Interestingly, the magnitude of dlog K_ds/dlog[NaCl] for UvsY–dsDNA interactions is –6.6 (Figure 4, Table 2), or approximately half that observed with the ssDNA lattice. Therefore UvsY exhibits different binding modes for ssDNA and dsDNA, involving large differences in the electrostatic component of binding. Implications for UvsY structure and function are discussed below (see ‘Discussion’ section). Table 2.

Salt effects on the intrinsic association constants of UvsY–DNA interactions

Previous sedimentation studies demonstrated that UvsY exists as a monodisperse hexamer of 95 kDa at salt concentrations ≥200 mM NaCl (22). The hexameric structure of UvsY protein suggests that it could wrap ssDNA around itself through contacts of multiple subunits with polynucleotide. To test this multi-subunit binding hypothesis, we performed meniscus depletion sedimentation equilibrium experiments to measure the stoichiometries of small oligonucleotides (dT₄, dT₈ and dT₂₄) bound to UvsY. Results of one such experiment are shown in Figure 5. In all cases, control experiments conducted at the non-permissive salt concentration of 500 mM NaCl resulted in no co-sedimentation of oligonucleotides with UvsY protein (Figure 5A), consistent with previous results (22). In contrast, when experiments were performed at the permissive salt concentration of 300 mM NaCl, meniscus depletion of the oligonucleotide dT₂₄ by co-sedimentation with UvsY was evident (Figure 5B). The loss of DNA absorbance at the meniscus allowed us to calculate a binding stoichiometry of 1.2 dT₂₄ molecules per UvsY hexamer, a value consistent with previous measurements of 1:1 stoichiometric interactions between UvsY hexamers and dT₂₄ or dT₂₅ oligos [(22), H. Xu, H. Beernink and S. Morrical, unpublished results]. Similar meniscus depletion studies performed with dT₄ and dT₈ oligonucleotides in 300 mM NaCl (data not shown) yielded stoichiometries of 5.3 and 2.7 oligos/hexamer, respectively. Given a solution average binding site size of n = 4 nt residues per UvsY monomer (23), the meniscus depletion data provide clear evidence that multiple subunits within the UvsY hexamer can interact with their binding sites on ssDNA simultaneously. The most dramatic example is given by the dT₄ oligo, which at 5.3 oligos/hexamer approaches the theoretical maximum of six oligos/hexamer. Therefore with a short enough oligo, it appears that all subunits of the UvsY hexamer retain the ability to interact with ssDNA. Figure 5.

Meniscus depletion of dT₂₄ oligonucleotide by UvsY protein. Meniscus depletion sedimentation equilibrium experiments were carried out as described in ‘Materials and Methods’ section. (A) Depletion experiments conducted under non-DNA-binding conditions (buffer B plus 500 mM NaCl) show no co-sedimentation of oligonucleotide dT₂₄ with UvsY. (B) Similar experiments conducted under permissive conditions for DNA binding (buffer B plus 300 mM NaCl) show a ∼1: 1 stoichiometry of UvsY₆:dT₂₄. In both experiments, oligonucleotide dT₂₄ alone (squares), UvsY alone (circle) or both (triangles) were centrifuged to equilibrium in a Beckman Optima XL-I analytical ultracentrifuge while scanning the absorbance at 260 and 280 nm. Co-sedimentation of dT₂₄ with UvsY under permissive binding conditions lowers absorbance at the meniscus, allowing calculation of complex stoichiometry.

Sedimentation velocity data were used to estimate the overall apparent shape of the UvsY hexamer and of the hexamer–ssDNA complex from f / f ₀, the hydrodynamic frictional ratio. Experimental frictional coefficients (f-values) for UvsY₆ and UvsY₆–ssDNA were calculated from published s⁰₂₀, w values (22), and are listed in Table 3. When compared with the theoretical frictional coefficient, f₀, calculated for a rigid, incompressible sphere of identical molecular mass and partial specific volume, it appears that the overall shape of UvsY deviates markedly from a spheroid, as f / f ₀ = 1.2 for both UvsY₆ and UvsY₆–dT₂₅ complexes. It is not possible to model UvsY oligomeric shapes directly from the sedimentation properties of monomers, because dissociation of UvsY hexamers into monomers requires chaotropic reagents (22). Therefore we estimated the oligomeric shape of UvsY indirectly by comparing the experimental f / f ₀ values to those of theoretical oligomers (36–40). Figure 7 presents a series of theoretical models and the corresponding frictional ratios for hexameric associations of UvsY monomers based on this method. The method assumes that each UvsY subunit is globular and approximately spherical in shape. Based on this assumption, the observed frictional ratio of UvsY₆ ± dT₂₅ (f / f ₀ = 1.2) is best approximated by ring-like structures such as the hexagonal model in Figure 7C (f / f ₀ = 1.17) or the open lockwasher model in Figure 7A (f / f ₀ = 1.20). Other reasonable hexameric arrangements shown in Figure 7 are not consistent with the experimental data. Independent biochemical evidence exists for structural asymmetry within the UvsY hexamer: UvsY contains one cysteine residue per monomer (Cys-110) (35) and therefore six per hexamer. However, covalent labeling of Cys-110 with excess 6-iodoacetamidofluorescein (6-IAF) under native hexamer conditions reproducibly yields a stoichiometry of one fluorescein moiety incorporated per hexamer (J. Liu, H. Xu and S. Morrical, unpublished results), suggesting that Cys-110 is only accessible to reagent at one end of an asymmetric structure. This finding is consistent either with a stable open lockwasher arrangement or with an equilibrium between open and closed forms of the hexagonal ring. Table 3.

Hydrodynamic parameters of UvsY₆ hexamers in the presence and absence of dT₂₅ oligonucleotide

The data demonstrate that UvsY has a strong thermodynamic preference for binding to relaxed ssDNA over relaxed dsDNA lattices (Figure 1, Table 1). The affinity differential is a result of ssDNA wrapping by UvsY. Results demonstrate that ssDNA is wrapped around hexamers of UvsY protein, since ssDNA contacts multiple subunits within each hexamer under monodisperse conditions (Figures 5 and 6) (22). The hydrodynamic and biochemical properties of UvsY hexamers are consistent with a ring- or lockwasher-like arrangement of subunits (Table 3; Figure 7A and C), suggesting that UvsY may wrap ssDNA into ring- or screw-like structures.

DNA stretching studies provide additional evidence for ssDNA wrapping by UvsY at the single-molecule level (24). Force spectroscopy measurements indicate that UvsY strongly wraps ssDNA that is created by the exposure of stretched dsDNA molecules to glyoxal. Presumably UvsY cannot wrap dsDNA because its persistence length is much higher than that of ssDNA (45). ssDNA wrapping occurs at low stretching forces where the DNA is relatively relaxed, whereas wrapping is suppressed when the DNA is under tension. This suppression of wrapping leads to the loss of preferential binding to ssDNA under the conditions of the DNA stretching experiments—in fact, UvsY binds tighter to dsDNA than to ssDNA under high stretching forces (24). In contrast, under relaxed conditions where ssDNA wrapping occurs, UvsY exhibits ∼1000-fold higher affinity for ssDNA than for dsDNA (Figure 1, Table 1). Further evidence that ssDNA wrapping promotes high-affinity UvsY–ssDNA binding comes from the observation that a monomeric form of UvsY exhibits a 10⁴-fold reduced affinity for ssDNA while retaining the same binding site size (4 nt residues per monomer) on ssDNA that is observed in hexameric UvsY (46).

Results indicate that UvsY uses different electrostatic binding modes for ssDNA and dsDNA, based on the large difference in the magnitude of the salt effect on K_ss versus K_ds (Figure 4, Table 2). Theoretically the magnitude of the salt effect on the affinity parameter for a protein–DNA interaction is defined as (7) where m′ equals the number of contacts between negatively charged DNA phosphate residues and positively charged protein amino acid residues, a is the number of anions displaced from the protein upon binding to DNA, and Ψ is a partitioning constant equal to 0.71 for ssDNA or 0.88 for dsDNA (41–44). In this study, we used the quantitative ssDNA–cellulose method to measure UvsY interactions with isolated sites on native ssDNA, and found a large salt effect on K_ss (dlog K_ss/dlog[NaCl] = –13.9). Previously we studied UvsY–ssDNA interactions by titrating etheno-modified ssDNA (εDNA) with protein at different salt concentrations, and found a much smaller salt effect (dlog K_ss/dlog[NaCl] = −6.8) (23). The latter value is similar to the salt effect on dsDNA binding (dlog K_ds/dlog[NaCl] = −6.6) determined from dsDNA–cellulose binding data in this study (Figure 4, Table 2). Possible explanations for the method-dependent difference in salt effects on UvsY–ssDNA interactions include:

The K58A and R60A mutations neutralize basic residues within a conserved LKARLDY motif found in the N-terminal domain of UvsY that is important for DNA binding (18,27). The motif is conserved in several other DNA-binding proteins including human ERCC2 and yeast Rad3 proteins (48). By attenuating potential electrostatic contacts with DNA, the single mutant UvsY_K58A is expected to have lower affinity compared to wild-type and the affinity of double mutant UvsY_{K58A, R60A} should be even lower. This is exactly what is observed in this study. The magnitude of the salt-dependence of K_ss observed for UvsY_R58A–ssDNA interactions is also lower than wild-type (dlog K_ss/dlog[NaCl] = −1.7 versus −13.9, respectively), consistent with a loss of electrostatic protein–DNA contacts. The large magnitude of this change is surprising considering the single amino acid change, however, and indicates that the K58A (and presumably R60A) mutations compromise a larger electrostatic network involved in UvsY–DNA contacts and/or in the wrapping of ssDNA around UvsY hexamers.

The results of this study suggest that UvsY could play an important role in the selective assembly of presynaptic filaments on ssDNA in the presence of excess dsDNA, a partitioning that is essential for recombination transactions both in vivo and in vitro. Driving this partitioning is the >1000-fold intrinsic affinity advantage of UvsY–ssDNA over UvsY–dsDNA interactions at physiologically relevant ionic strengths (Table 1). Like many recombinases of the RecA/Rad51 family, the T4 UvsX protein binds strongly to dsDNA as well as to ssDNA (8,49) (R. Maher and S. Morrical, manuscript in preparation). In the case of eukaryotic Rad51, pre-bound dsDNA inhibits DNA strand exchange, thus the translocation/remodeling activities of Rad54 protein are needed to disrupt inappropriate Rad51–dsDNA contacts and to promote productive filament assembly on ssDNA (50). In the T4 recombination system, the need for a Rad54-like function may be circumvented by UvsY, which by virtue of its strong affinity preference for ssDNA may selectively recruit UvsX to and nucleate filament formation on ssDNA. Reinforcing this partitioning is the fact that UvsY and nucleoside triphosphate increase UvsX–ssDNA binding affinity synergistically (18). The combination of factors would therefore result in stringent discrimination in favor of presynaptic filament assembly on ssDNA.

The ability of UvsY to bind to ssDNA in either extended or wrapped conformations has important implications for its recombination mediator functions. As a classical recombination mediator protein, UvsY promotes the loading of UvsX recombinase onto pre-existing Gp32–ssDNA complexes to form presynaptic filaments with concomitant displacement of Gp32 from ssDNA. Ensemble, single-molecule and mutagenesis data are consistent with a model in which UvsY induces changes in ssDNA structure that destabilize Gp32–ssDNA interactions and stabilize UvsX–ssDNA interactions (17,18,24,25). Previous studies of the UvsY_K58A and UvsY_K58A,R60A mutants, which bind weakly to ssDNA, established that UvsY–ssDNA interactions are the major factor controlling UvsY stabilization of UvsX–ssDNA interactions, while UvsY–UvsX protein–protein interactions play a minor role in stabilizing UvsX–ssDNA interactions (18,27).

Many nucleic acid pathways channel their DNA or RNA substrates through a series of hand-off transactions in which a nucleic acid intermediate is passed directly from one protein in the pathway to the next without ever being released as free nucleic acid (51–54). The advantages of this strategy include: potential cytotoxic effects of the nucleic acid intermediate are minimized by keeping it sequestered; also the nucleic acid is protected from inappropriate degradation or side reactions. Evidence suggests that ssDNA in recombination pathways is similarly channeled (25). Based on the DNA-binding properties of the three protein players, we propose that the loading of UvsX recombinase onto Gp32–ssDNA complexes proceeds through a series of at least two sequential hand-off steps (Figure 8) (12,24,25): Gp32 first binds to ssDNA and converts it into an extended form resembling the mechanically stretched DNA that was sampled in single-molecule experiments. The first hand-off occurs as UvsY recognizes and binds to the extended ssDNA, then converts it into a wrapped form that destabilizes Gp32–ssDNA interactions. The UvsY-wrapped ssDNA complex is postulated to be in equilibrium between ‘closed’ and ‘open’ states (25). The closed form is the one that destabilizes Gp32–ssDNA interactions, and it is inaccessible to UvsX. The open form is optimal for UvsX–ssDNA interactions. The second hand-off occurs as the ATP-bound form of UvsX recognizes and binds to the open UvsY–ssDNA structure, nucleating or propagating a UvsX–ssDNA filament while inducing Gp32 to leave. Conceivably UvsX loading/Gp32 displacement could occur either in a stepwise fashion (Figure 8A) or in a concerted fashion (Figure 8B) depending on solution variables including salt concentration (24,25). Superimposed on the underlying hand-off mechanism are effects involving UvsX cooperativity (55) and a nucleotide exchange factor activity of UvsY (25) that have not been fully explored. Also, linking ATP hydrolysis by UvsX to the hand-off scheme creates dynamic instability in presynaptic filaments that may be linked to filament turnover, translocation, or DNA strand exchange activity (12,13). Figure 8.

A double hand-off model for the mechanism of mediator protein UvsY in T4 presynaptic filament assembly. Adapted with modifications from ref. (25). UvsY protein facilitates the loading of UvsX recombinase onto ssDNA and the concomitant displacement of Gp32 ssDNA-binding protein from ssDNA. The figure shows UvsX loading and Gp32 displacement from the perspective of a single UvsY hexamer, as if looking down the helical axis of a nascent presynaptic filament. The cooperative binding of Gp32 to ssDNA extends the polynucleotide lattice. The first hand-off occurs as hexameric UvsY recognizes and binds to the extended ssDNA (step 1), then converts it into a wrapped conformation(s) (steps 2 and 3), destabilizing Gp32–ssDNA interactions in the process. The UvsY-wrapped ssDNA complex is postulated to be in equilibrium between ‘closed’ and ‘open’ conformations (step 3), the latter of which is recognized by the ATP-bound form of UvsX protein to nucleate presynaptic filament assembly (step 4) while displacing Gp32. (A) Steps 3 and 4 constitute a step-wise mechanism for Gp32 displacement and UvsX loading by UvsY, which may occur under low-salt conditions. (B) Under high-salt conditions UvsY does not displace Gp32 from ssDNA directly, so filament assembly likely occurs by a concerted mechanism in which synergistic action of UvsY and ATP-bound UvsX is required to displace Gp32. See text for details.

Previously, recombination mediator proteins have been thought of primarily as factors that facilitate the exchange of bound SSB proteins for recombinase on ssDNA (10). Results of this study suggest an additional function for RMPs: that of recruiting recombinases onto ssDNA while minimizing inappropriate recombinase–dsDNA interactions during the presynaptic phase of homologous recombination. The universal role of RMPs in promoting recombinase–ssDNA assembly suggests that different RMPs may share common mechanistic features. It is noteworthy that, like UvsY, many other well-characterized RMPs have oligomeric structures with multiple ssDNA-binding sites. Bacterial RMPs include the RecO and RecR proteins which function together as hetero-oligomers (56,57). The RecR protein of Deinococcus radiodurans crystallizes as a tetrameric ring, exhibits both ssDNA- and dsDNA-binding activity, and interacts with RecO, whereas RecO itself contains multiple DNA-binding sites (58,59). Eukaryotic RMPs include the Rad52 proteins, the human ortholog of which appears to function as a heptameric ring, and ssDNA is proposed to wind around the ring structure (60–62). Therefore organizing ssDNA by wrapping may be a common element of the function of certain homo- or hetero-oligomeric RMPs.

Other eukaryotic proteins with RMP activity include Rad51 paralogs and the tumor suppressor gene product Brca2 (63). Rad51 paralogs (Rad55, Rad57 in Saccharomyces cerevisiae; Rad51B, Rad51C, Rad51D, Xrcc2, Xrcc3 in humans) exist as heterodimers or oligomers (64,65). Subsets of the human Rad51 paralogs exist in two different ring-shaped complexes, BCDX2 and CX3, and bind strongly to Holliday junctions (66). Other aspects of their DNA-binding activities are incompletely understood. Brca2 is thought to function as a monomer; nevertheless it contains three OB fold motifs that bind ssDNA, as well as sites for structure-specific DNA binding (67). So it remains a possibility that Brca2, in addition to its well-characterized protein-protein interactions with Rad51 (68), also promotes recombination through changes in ssDNA structure.