Human Rad51 mediated DNA unwinding is facilitated by conditions that favour Rad51-dsDNA aggregation

RAD51 recombinase, the eukaryotic homolog of RecA, is the core component in homologous recombination (HR). It performs fundamental roles such as homologous pairing and strand exchange in repairing DNA double strand breaks [1, 2]. Rad51 binds both ssDNA/dsDNA [3–5] and forms right-handed helical nucleoprotein filaments in which DNA is extended 1.5 times that of B form DNA [4, 6–8]. Such extended conformation of DNA in ssDNA-protein filaments is instrumental in facilitating homology search and strand exchange in a three stranded pairing system [8–10]. Even though the equivalent scenario of extended DNA configuration is demonstrable in dsDNA-protein filament, its relevance in HR is not fully understood. However, based on E. coli RecA system, it is surmised that unwinding associated with dsDNA-protein filament is similarly important in four-stranded pairing and strand exchange processes [11]. The first and also most recent report that describes four-stranded exchange reaction catalyzed by RAD51 strongly underscores this aspect where RAD51-dsDNA filament is shown to engage in duplex-duplex pairing [12]. The current study is aimed to better understand the functions of dsDNA-protein filament. Directly impinging on this issue is the relevance of nucleotide cofactor, ATP. The current understanding of binding/hydrolysis of ATP by RAD51 with respect to its HR function is still evolving, and the same is different from that of RecA [13, 14]. RAD51 possesses DNA dependent ATPase activity, which is largely stimulated by ssDNA and to some lesser extent by dsDNA [4, 7, 10]. ATP hydrolysis is critical for homology directed DNA repair in vivo [15] as well as in some in vitro conditions [10, 16], but not in some other conditions [17, 18], implying that RAD51 system is endowed with sufficient dynamicity that it can bring about various functional outcomes based on given reaction conditions. Inspired by this property of the protein system, our current study addresses dsDNA unwinding reaction by RAD51 with respect to not only nucleotide cofactor conditions, but also the role of RAD52, an important regulator of RAD51 functions, in the same.

Rad52 interacts genetically/physically with Rad51 and stimulates a variety of its functions during HR [19–22]. Both proteins co-localize in distinct repair proficient nuclear foci in response to DNA damage [23]. Rad52 is required for such foci formation during meiosis [24]. Mechanistically, it has been shown that in Saccharomyces cerevisiae, Rad52 is critical for recruiting Rad51 to repair sites, in vivo [25]. Chromatin immunoprecipitation assays demonstrate Rad52 requirement for recruitment of yeast Rad51 to HO induced repair sites at MATa locus [26, 27]. Biochemically, Rad52 stimulates strand exchange activity of Rad51 [28–30], is shown to displace RPA [31, 32], stabilize the Rad51-ssDNA filament [33]. In light of such demonstrated functions of Rad52 in a variety of Rad51 activities, we realized the importance of probing such collaborative relationship in RAD51 mediated dsDNA unwinding reaction.

It is relevant to point out here that RAD51 is one of the versatile DNA binding proteins in the cell whose ability to laterally diffuse on dsDNA as studied by Total Internal Reflection Microscopy is not only rather fast but also does not seem to need the energy of ongoing ATP hydrolysis [34]. Here we show that such a dynamic protein system brings about efficient unwinding of dsDNA in conditions that do not hydrolyze ATP, a result consistent with similar findings reported earlier by Chi et al, 2006 [17]. Importantly, our current study shows that conditions which facilitate better aggregation of RAD51-dsDNA also stimulate steady-state level of dsDNA unwinding in parallel. These are achieved by either presence of RAD52 or salt (KCl) at low concentration, where macromolecular crowding effects facilitate RAD51 mediated DNA unwinding reaction.

RAD51 unwinds dsDNA [4, 17], but the underlying mechanism of this process with respect to biochemical reaction conditions is far from clear. We have addressed this aspect in our current study. Here we delineate specific nucleotide cofactor requirements for DNA unwinding and provide mechanistic insights on how RAD52 collaborates with RAD51 in this process. Furthermore, we show that unwinding of duplex DNA by RAD51 correlates positively with aggregation of RAD51-dsDNA complexes.

RAD51 mediated dsDNA unwinding does not require ATP hydrolysis

Fully relaxed circular dsDNA was generated from φX174 supercoiled plasmid by the action of Drosophila Topo I. Subsequent unwinding of relaxed dsDNA by RAD51 was assayed in the presence of Topo I, followed by deproteinization and agarose gel electrophoresis (Methods). DNA unwinding as a function of RAD51 concentration depended entirely on nucleotide cofactor conditions. ATP as well as its poorly/non-hydrolysable analogues such as ATPγS/AMP-PNP facilitated efficient unwinding (lanes 1–3, Panels C, F & E, Fig. 1), while the absence of nucleotide cofactor failed in the same (lanes 1–3, Panel H, Fig. 1). Surprisingly, detectable level of unwinding was observed even in presence of ADP (lanes 1–3, Panel D, Fig. 1), whereas with AMP no unwinding was observed (lanes 1–3, Panel G, Fig. 1). Further quantification of the gel revealed, maximal DNA unwinding (~44%) even at the lowest RAD51 concentration [3 μM] in the presence of ATP/AMP-PNP, whereas DNA unwinding increased as a function of RAD51 concentration in ATPγS/ADP conditions and reached about 34% and 22% respectively at maximal protein concentration (9 μM) (Fig. 1I). Thus unwinding appears to be ATP binding but not hydrolysis dependent, as not only ATP but also its non-hydrolysable analogs support the process, which is in congruence with previously published reports [4] and [17]. However, even product of ATP hydrolysis, ADP was seen to support DNA unwinding by RAD51, a result that appears somewhat counter-intuitive at the outset. But in fact, even this observation is consistent with earlier result that RAD51 binds ssDNA sufficiently well that it promotes strand annealing even in ADP conditions, a property starkly contrasting with that of well studied E. coli RecA protein [35]. Furthermore, preliminary linear dichroism experiments with RAD51, calf-thymus dsDNA and ADP have confirmed that the HsRad51-dsDNA complex is also stable in the presence of ADP (H.K. Kim, K. Morimatsu & B. Nordén, unpublished observations). Thus, ADP stabilizes both ssDNA and dsDNA complex formation with HsRad51, although it does not promote the strand exchange reaction. This could suggest that the role of the nucleotide cofactors is not only to stabilize or destabilize the protein-DNA complex, but also involves changing the conformation of the complex to promote the strand exchange reaction. In order to eliminate the possibility that unwinding observed in ADP condition is not a trivial consequence of either ATP contamination or artifactual conversion of ADP to ATP, we assessed the integrity of ADP by Thin Layer Chromatography (TLC) assay [36]. Expectedly we found that ADP containing samples showed no traces of ATP contamination. Added ADP was pure and no artifactual conversion to ATP had happened (data not shown). Therefore we strongly believe that ADP effects were genuine.

RAD52 stimulates RAD51 mediated duplex unwinding

In order to test the role of RAD52 in DNA unwinding, we titrated the reactions with increasing levels of RAD52 while maintaining RAD51 concentration constant at a lower level. As expected, conditions that failed to promote RAD51 mediated unwinding, namely absence of nucleotide cofactor or presence of AMP, remained ineffective even with addition of RAD52 (lanes 2–4, Panels F & E, Fig. 2). On the other hand, nucleotide cofactor conditions that are congenial for RAD51 mediated unwinding, namely ATP/AMP-PNP/ATPγS/ADP, revealed marginal and detectable improvement in unwinding following RAD52 addition (lanes 2–4, Panels A, C, D & B, Fig. 2), while RAD52 alone did not lead to any detectable unwinding in the presence or absence of any of these nucleotide cofactors (lane 5, Panels A-F & lanes 1–3, Panel G, Fig. 2). Moreover, such RAD52 mediated enhancement of DNA unwinding by RAD51 was evident even in presence of ADP, when time-course of unwinding was studied. At fixed concentrations of RAD51 and RAD52, unwinding efficiency with ADP, reached a level similar to that of ATP, within 10 minutes of incubation (lane 4, Panel B, Fig. 3) (Fig. 3C). On the contrary, ADP containing reaction that lacked RAD52, showed only marginal unwinding which hardly increased with time (only last time-point is shown; lane 7, Panel B, Fig. 3) (Fig. 3C).

RAD52 forms large RAD52-RAD51-DNA complexes

To understand molecular basis of RAD52 mediated stimulation on DNA unwinding by RAD51, we carried out DNA binding studies of RAD51 and RAD52 with circular dsDNA in the presence of various nucleotide cofactors. RAD51 protein when bound to dsDNA, generated distinctly slower migrating gel-shifted complexes as a function of increasing protein concentration. This was so in all nucleotide conditions, suggesting RAD51 binding to dsDNA was not dependent on nucleotide cofactor conditions (lanes 2–4 & 8–10, Panel A: lanes 1–3 & 7–9, Panels B & C, Fig. 4) which is consistent with previously published reports [17]. Under same conditions, binding of RAD52 alone to dsDNA generated large protein-DNA complexes that failed to enter the gel (Panel D, Fig. 4) as reported earlier [37]. Interestingly, we find that, addition of RAD52 converted RAD51-DNA gel-shifted complexes also into large complexes that fail to enter the gel (lanes 5–7 & 11–13, Panel A: lanes 4–6 and 10–12, Panels B & C, Fig. 4). Irrespective of the nucleotide cofactor conditions, addition of RAD52 resulted in large RAD52-RAD51-DNA complexes. We have further confirmed that these large protein-DNA complexes entrapped in the wells indeed contain both RAD51 and RAD52 proteins by excising agarose gel fragments containing these complexes, and subjecting them to protein compositional analyses by SDS-PAGE (Data shown only for complex formed in ATP conditions, the boxed region of lane 5, analyzed in lane 1, Panel E, Fig. 4).

RAD52 co-aggregates with RAD51 on DNA

In order to corroborate that these complexes indeed house both proteins and DNA, we carried out centrifugation assay and tested the ability of RAD52 to "co-aggregate" both RAD51 and itself onto DNA. Also, the same was analyzed with respect to ongoing DNA unwinding reaction. RAD51 protein, under the current experimental conditions, predominantly stayed in supernatant while a minor fraction sedimented following centrifugation assay (lane s1, Panel D, Fig. 5). On the contrary, RAD52 protein that is largely in aggregated state by itself (lane p, Panel B, Fig. 5), increases aggregation prone ness of RAD51 protein in its presence (lanes p1 and s1, Panel C, Fig. 5) where RAD52 protein retains its high state of aggregation (lane p1, Panel C, Fig. 5). Furthermore, we find RAD51-dsDNA complexes that are largely soluble in ATP/ADP conditions (lanes s1, Panels E & F, Fig. 5) get aggregated with addition of RAD52. This is evidenced by increase in levels of RAD51 protein (lanes p2–p4, Panels E & F, Fig. 5) as well as relaxed/unwound dsDNA (lanes p2–p4 Panels D & E, Fig. 6) in pellet fraction as a function of RAD52. Even in other conditions, namely, AMP-PNP/ATPγS/AMP or in the absence of any nucleotide cofactor, a similar trend is observed, where RAD52 protein renders RAD51-dsDNA complexes, aggregation prone (Figs. 5 &6). Protein (lanes p2–p4, Panels G-J, Fig. 5) as well as DNA (lanes p2–p4, Panels F-H, Fig. 6) analyses reveal the trend that complexes comprising of RAD51/RAD52/dsDNA were "aggregated" as a function of RAD52 concentration. We speculate that these complexes in fact, may represent "co-aggregates" as both proteins are known to interact with each other in these reaction conditions [38]. Interestingly, DNA analyses reveal that "co-aggregates" contained not only substrate DNA i.e. relaxed dsDNA, but also products of unwinding. Quantitative analysis of proteins as well as DNA in the gel assays revealed that in all conditions of nucleotide cofactors tested, RAD51 protein and dsDNA co-sedimented as a function of RAD52 concentration. This is evidenced by an increase in recovery of RAD51 protein as well as dsDNA in pellet fractions with a concomitant depletion of the same in supernatant fractions [Fig. 5K: Protein analysis (data shown only for pellet fractions) & (Fig. 6I: DNA analysis (only pellet fraction data is shown)].

Salt (KCl) induced changes in aggregation of RAD51-DNA complexes

In order to unravel any mechanistic link between RAD51-dsDNA aggregation and duplex unwinding, we probed the system by salt titration study. Moreover, we also tested whether salt induced aggregation of RAD51-dsDNA complexes would then obviate the need of RAD52 in the same. Centrifugation analysis of RAD51-DNA complexes as a function of varying KCl concentration in presence of ATP revealed interesting trends: RAD51-dsDNA aggregation increased in low salt (up to 100 mM KCl), followed by a drop at high salt regime (100–200 mM KCl) [(Fig. 7A: protein analysis & Fig. 7C: DNA analysis) (data shown only for pellet fractions)]. This trend was unaffected by the presence of RAD52 [(Fig. 7B: protein analysis) & (Fig. 7D: DNA analysis)]. Quantification of gel data revealed increase in RAD51 and dsDNA levels in pellet fractions as a function of salt concentration (up to 100 mM), followed by a decline of the same at higher salt concentration (100–200 mM) (Fig. 7E). Concomitantly, supernatant fractions revealed reciprocal trend. On the contrary, under same reaction conditions, RAD52 protein showed a different trend: RAD52 protein that was highly aggregated to begin with (without salt), disaggregated with the addition of salt. Salt induced disaggregation of RAD52 ensued from > 50 mM KCl onwards, and reached completion at about 140 mM KCl (Figs. 7B &7E).

Salt (KCl) induced changes in dsDNA unwinding by RAD51

In an attempt to correlate salt (KCl) induced changes in aggregation of RAD51-dsDNA complexes with corresponding changes in DNA unwinding, we analyzed RAD51 mediated unwinding at varying concentrations of KCl. Quantitative analysis of gel assay revealed that unwinding was stimulated in low salt concentration (reaching up to 100–120 mM KCl), followed by a fall in the same at higher KCl concentration (> 100 mM KCl) (Fig. 8). In fact, by about 160 mM KCl, the unwinding level was barely detectable. By and large the effect of salt (KCl) on RAD51 mediated unwinding was very similar even in presence of RAD52. Consequently, the salt titration studies suggested that low concentration of salt (up to 100 mM) which is conducive for RAD51-dsDNA aggregation, in parallel, also resulted in an increase in dsDNA unwinding. Conversely KCl at high concentration (> 100 mM) led to decrease in aggregation as well as dsDNA unwinding. RAD52 protein in RAD52-RAD51-dsDNA complexes started disaggregating at concentrations of KCl greater than 50 mM (Figs. 7B &7E), thereby suggesting that changes in RAD51-dsDNA aggregation as well as dsDNA unwinding are relatively independent of RAD52 protein in the same regime of salt concentration. Accordingly, these experiments further suggest that process of dsDNA unwinding is mechanistically linked to the aggregation state of RAD51 in RAD51-dsDNA complexes.

In order to test whether enhanced aggregation as well as dsDNA unwinding observed in parallel at low salt (up to 100 mM KCl) was also accompanied by any changes in RAD51 binding to dsDNA, we analyzed the samples by gel-shift assays. Since the observed effects of salt were independent of RAD52 (as described above), we analyzed RAD51-dsDNA complexes as a function of KCl in the absence of RAD52. Gel-shift analyses revealed that mobility of protein-DNA complexes increased as a function of salt, thereby suggesting that RAD51 binding to dsDNA was reduced in the presence of salt. At high concentration of salt [200 mM] the protein-DNA complexes appeared to have dissociated, giving rise to the release of free dsDNA (lane 7, Fig. 9).

Our current study primarily shows that conditions which favour efficient aggregation of RAD51-dsDNA complexes such as presence of RAD52 or inclusion of salt in reaction leads to enhanced DNA unwinding in presence of suitable nucleotide cofactors. RAD52 by itself does not mediate DNA unwinding (Fig. 2G). Therefore, the only plausible mechanism by which RAD52 stimulates RAD51 mediated unwinding is by affecting RAD51-DNA interactions. The high aggregation prone nature of RAD52 protein is well established by Dynamic Light Scattering and Electron Microscopy studies [35, 37–40] and also centrifugation assays performed in the present study (Panel B, Fig. 5). RAD51 on the other hand, is largely a soluble protein in its native state (Panel D, Fig. 5). We speculate that aggregated form of RAD52 through its large binding surface acts as a scaffold and sequesters RAD51-dsDNA complexes in smaller volume and thereby enhances RAD51 reactivity by macromolecular crowding effects. Indeed, centrifugation assays suggest greater recovery of RAD51, RAD52 and unwound DNA in pellet fractions as a function of increasing RAD52 concentration (Figs. 5 &6).

RAD51 binding to dsDNA has been shown to be selectively disrupted by salt in the context of homologous pairing and DNA strand exchange [41]. In particular, high salt [200 mM KCl] is reported to significantly lower RAD51 binding to dsDNA [42]. Consequently, we used KCl as a tool to perturb RAD51-dsDNA/RAD52-RAD51-dsDNA aggregation and monitor the corresponding changes in unwinding profile. KCl at low concentrations (up to 100 mM) mimics RAD52 effect in aggregating RAD51-dsDNA complexes (Fig. 7) while stimulating DNA unwinding in parallel (Fig. 8). On the other hand, KCl at high concentrations (> 100 mM) leads to disaggregation of complexes and lowering of dsDNA unwinding. On the contrary, RAD52 protein in RAD52-RAD51-dsDNA complexes started disaggregating at concentrations of KCl greater than 50 mM (Figs. 7B &7E), thereby suggesting that changes in RAD51-dsDNA aggregation as well as dsDNA unwinding are relatively independent of RAD52 protein in the same regime of salt concentration. Therefore one can infer that low salt mimics RAD52 effect in enhancing RAD51-dsDNA aggregation as well as dsDNA unwinding. It is relevant to note that in the regime of salt concentration that is stimulatory for aggregation of RAD51-dsDNA complexes and dsDNA unwinding (up to 100 mM KCl), RAD51 binding to dsDNA is reduced (Fig. 9). Such reduction in protein binding leads to partially coated RAD51-dsDNA complexes, which in turn facilitates enhanced intermolecular aggregation. This interpretation is fully consistent with the observed properties of RecA-DNA complexes where incompletely coated filaments have been shown to aggregate better [43].

We believe that increased aggregation of RAD51-dsDNA complexes in low salt conditions, just as in the presence of RAD52, facilitates macromolecular crowding effects wherein several RAD51 and dsDNA molecules are brought proximal to each other within RAD51-dsDNA aggregates. Such a favourable scenario facilitates dynamic intermolecular transfer of RAD51 between DNA molecules within such aggregates. RAD51 is able to sample relatively more dsDNA molecules because of such transient binding to DNA in contrast to sampling lesser dsDNA molecules when RAD51 and dsDNA complexes are not aggregated. One can easily envisage kinetic advantage in such aggregated complexes via steps of rapid protein transfer, since RAD51 protein mobility on dsDNA has been shown to be highly dynamic [34]. Presence of Topo I in unwinding assay used in our study irreversibly captures the events of unwinding (resulting from protein binding) even after the dissociation of bound protein. Therefore Topo-I assay cumulatively "scores" the events of unwinding that manifest as an increase in steady-state level of DNA unwinding in conditions that favour protein-DNA aggregation. In the current study we have shown that RAD52, specific interactor of RAD51 protein, facilitates RAD51-DNA aggregation by acting as a molecular scaffold. However, we expect that any known reagent (spermidine, polyvinyl alcohol etc.) that non specifically brings about macromolecular crowding of protein-DNA complexes leading to aggregation will also elicit better unwinding of DNA because the process of aggregation appears to be more relevant than the mode (specificity) of aggregation. We suppose that the enhanced unwinding of dsDNA associated with aggregated RAD51-dsDNA complexes is somewhat analogous to that of aggregated forms of RecA-DNA networks participating more actively in homology search steps [43, 44]. Our future studies are aimed at exploring the details of such dynamic intermediates in RAD51 catalyzed reactions.

φX174 supercoiled DNA substrate was obtained from New England Biolabs. Potassium Chloride salt was purchased from Sisco Research Laboratories Ltd. Competent HMS174 strain of E. coli DE3 cells carrying RecA mutation was purchased from Novagen. All nucleotide cofactors used in this study were purchased from Roche diagnostics. The integrity of ADP is assessed by Thin Layer chromatography (TLC) analysis [36] and it is free of ATP contamination. Ni-NTA agarose beads were purchased from Qiagen.