Tyrosine phosphorylation stimulates activity of human RAD51 recombinase through altered nucleoprotein filament dynamics

Significance

Homologous recombination provides the most accurate means to repair genotoxic DNA lesions. It depends on the assembly of the RAD51 DNA strand exchange protein into a dynamic nucleoprotein filament. The c-Abl tyrosine kinase and its oncogenic counterpart BCR-ABL control RAD51 by phosphorylating tyrosine residues 54 and 315. A nonnatural phosphotyrosine mimetic was used to represent phosphorylated RAD51 and to parse out the importance of Y54 and Y315 phosphorylation. By combining biochemical and single-molecule analyses, we found that Y54 phosphorylation enhances the RAD51 DNA strand exchange activity by altering the nucleoprotein filament properties. In contrast, Y315 phosphorylation has little effect on the RAD51 activities.

Abstract

The DNA strand exchange protein RAD51 facilitates the central step in homologous recombination, a process fundamentally important for accurate repair of damaged chromosomes, restart of collapsed replication forks, and telomere maintenance. The active form of RAD51 is a nucleoprotein filament that assembles on single-stranded DNA (ssDNA) at the sites of DNA damage. The c-Abl tyrosine kinase and its oncogenic counterpart BCR-ABL fusion kinase phosphorylate human RAD51 on tyrosine residues 54 and 315. We combined biochemical reconstitutions of the DNA strand exchange reactions with total internal reflection fluorescence microscopy to determine how the two phosphorylation events affect the biochemical activities of human RAD51 and properties of the RAD51 nucleoprotein filament. By mimicking RAD51 tyrosine phosphorylation with a nonnatural amino acid, p-carboxymethyl-l-phenylalanine (pCMF), we demonstrated that Y54 phosphorylation enhances the RAD51 recombinase activity by at least two different mechanisms, modifies the RAD51 nucleoprotein filament formation, and allows RAD51 to compete efficiently with ssDNA binding protein RPA. In contrast, Y315 phosphorylation has little effect on the RAD51 activities. Based on our work and previous cellular studies, we propose a mechanism underlying RAD51 activation by c-Abl/BCR-ABL kinases.

The DNA in the human genome is constantly subjected to damage. This damage is a byproduct of normal cellular metabolism, exposure to radiation, and environmental mutagens (1). Homologous recombination (HR) and the pathways that use the machinery of HR are responsible for the accurate repair of the most deleterious DNA lesions, including double-stranded DNA breaks (DSBs), interstrand DNA cross-links, and damaged replication forks, and thereby contribute to maintenance of the stable genome (2–5). HR also plays an important role in telomere maintenance (6). HR, a process highly conserved throughout evolution, is carried out through a precisely coordinated and tightly regulated series of events. The key step in HR is the assembly of a RecA-family recombinase (phage UvsX, bacterial RecA, archaeal RadA, or eukaryotic RAD51) onto resected single-stranded DNA (ssDNA). The recombinase forms a nucleoprotein filament that then invades homologous duplex DNA, resulting in a displacement loop structure that can be used as a primer for synthesis using the intact duplex as a template. Similar to its bacterial and yeast homologs, the human RAD51 binds ssDNA in an ATP-dependent manner (7). The nucleoprotein filament formed by the ATP-bound RAD51 is arranged such that each RAD51 monomer binds three nucleotides, forming the pairing unit in these reactions (8). Beyond these basic attributes, the characteristics of human RAD51 protein differ significantly from its archaeal, bacterial, and yeast homologs.

The critical role of HR requires all steps of this process to be tightly regulated to avoid untimely or illegitimate recombination that may cause carcinogenic genome rearrangements or result in cytotoxic intermediates. Inactive HR causes loss of faithful DNA repair and leads to genetic instability, whereas excessive HR interferes with cellular processes such as replication, transcription, and telomere maintenance and also can lead to gross chromosomal rearrangements (3). In human cells, the assembly of the RAD51 nucleoprotein filament is aided by the recombination mediator BRCA2 and RAD51 paralogs (9, 10), antagonized by antirecombinases (11) and the heteroduplex rejection machinery (12), and is also regulated by posttranslational modifications. Activities of RAD51 in the cell are influenced by three types of phosphorylation. RAD51 protein is phosphorylated at threonine residue 309 by the Chk1 checkpoint kinase (13), which regulates the DNA damage response. It is phosphorylated by Plk1 at serine 14 and then by CK2 at threonine 13 to facilitate the interaction with the Nbs1 protein and recruitment to the DNA damage sites (14). It is also phosphorylated at tyrosine residues 54 and 315 by the c-Abl/BCR-ABL tyrosine kinase (15, 16). Ionizing radiation and other genotoxic agents activate c-Abl kinase in an ATM and DNA-PK dependent manner (17, 18). There have been several studies on how c-Abl regulates RAD51, with a major debate regarding the site targeted by this tyrosine kinase. Recent studies have suggested that RAD51 is phosphorylated in two steps, with Y315 phosphorylation being a prerequisite for the Y54 phosphorylation (19). There remains, however, an uncertainty about the biochemical effects of RAD51 phosphorylation by c-Abl. Results of the cell-based studies suggested that the Y54 and Y315 phosphorylation enhances the RAD51 nuclear foci formation and resistance to DNA damaging agents (15, 20). In contrast, biochemical studies that used aspartate and glutamate amino acids to mimic Y315 phosphorylation, or used Saccharomyces cerevisiae Rad51 (sceRad51) phosphorylated in vitro by c-Abl kinase showed a decrease in the DNA strand exchange activity (16, 21). The discrepancy between the results of the biochemical and cellular studies may stem from the fact that S. cerevisiae lacks tyrosine phosphorylation (22) and from a substantial structural difference between yeast and human proteins (23). Additionally, representation of phosphotyrosine by a negatively charged amino acid is often inaccurate. In fact, we have recently shown that Y/D substitution incorrectly represents tyrosine phosphorylation in another c-Abl target, human RAD52 (24).

Here, we set out to reconcile the biochemical data with the cell-based studies and to identify the properties of the human RAD51 protein and its nucleoprotein filament that are altered to enhance or diminish HR. By incorporating a nonnatural amino acid that mimics phosphorylated tyrosine, we provide a realistic representation of the phosphorylated RAD51 and parse out the impact of the two individual phosphorylation events. Our model of human RAD51 (23) predicted that Y54 participates in the intersubunit stacking interaction with F195 of the adjacent monomer within the RAD51 filament similar to the interaction revealed in the sceRad51 structure (25) (Fig. 1A). Introduction of a large negative charge by Y54 phosphorylation may break the stacking and alter the RAD51/RAD51 interface. Even slight deviations toward either a more or a less stable RAD51 filament due to this alteration may have a significant effect on the RAD51 nucleoprotein filament assembly and function. Y315 is also near the RAD51/RAD51 interface and is directly adjacent to D316, which was proposed to form a salt bridge to the γ-phosphate of ATP and to function as a conformational sensor that enhances nucleoprotein filament turnover (26). It was more difficult to anticipate the effect of placing an additional negative charge next to this charged aspartate.

Fig. 1.

Incorporation of pCMF into human RAD51. (A) Homology model of two adjacent human RAD51 monomers. RAD51 DNA binding loops L1 and L2 are shown in orange and green, respectively. Residues Y54, Y315, and F195 are shown in a ball-and-stick rendering. (B) Structures of tyrosine, phosphorylated tyrosine, and pCMF. (C) Incorporation of pCMF incorporation using the amber suppressor system: RAD51 expression in the absence of pCMF leads to translation of a truncated product, whereas full-length protein is produced in the presence of pCMF. (D) Western blot (WB) showing expression of full-length RAD51^Y54pCMF and RAD51^Y315pCMF proteins using mouse anti-RAD51 (3C10) antibody. Truncated products are observed in the absence of pCMF (Y54 truncations are too small to be verified by standard gel electrophoresis). (E) SDS/PAGE gel showing purified RAD51, RAD51^Y54pCMF, and RAD51^Y315pCMF. The RAD51^Y315pCMF has a very small amount of truncated product (<18%) that could not be separated using our purification scheme (Fig. S1B).

Using biochemical reconstitutions, we observed elevated DNA strand exchange activity of Y54–p-carboxymethyl-l-phenylalanine (pCMF) phosphomimetic protein (RAD51^Y54pCMF), but not the Y315-pCMF phosphomimetic protein (RAD51^Y315pCMF), consistent with previously published cellular studies that reported enhanced HR and RAD51 localization at the sites of the DNA damage (15, 27). Moreover, we showed that RAD51^Y54pCMF can efficiently carry out the so-called “RPA-first” DNA strand exchange reactions that would normally require the presence of a recombination mediator. To elaborate on the mechanism of RAD51 function and regulation, we used Förster resonance energy transfer (FRET)-based experiments at ensemble and single-molecule levels. We showed differences in DNA binding between unmodified RAD51 and RAD51^Y54pCMF that explain the basis of nucleoprotein filament formation and robust DNA strand exchange activity of RAD51^Y54pCMF. Finally, we provide a model for the regulation of the RAD51 protein by sequential phosphorylation, highlighting the mechanism of regulation by the c-Abl/BCR-ABL kinases.

Results

Mimetics of the RAD51 Y54 and Y315 Phosphorylation.

Protein phosphorylation is commonly mimicked by the charged amino acids aspartate and glutamate. These substitutions may inaccurately represent phosphotyrosine when the precise geometries or the aromatic ring of the tyrosine is important (24, 28). Here, we used a nonnatural amino acid, pCMF, which more closely mimics a phosphorylated tyrosine residue than aspartate or glutamate (24, 28) (Fig. 1B). RAD51^Y54pCMF and RAD51^Y315pCMF mutants were expressed in Escherichia coli Acella cells transformed with the pCH1-RAD51 vector (Materials and Methods), along with the pSUPT-UaaRS vector encoding the amber tRNA and pCMF-RS (Fig. S1). Induction of the cells in the presence of isopropyl-β-d-thiogalactopyranoside (IPTG) produced large amounts of truncated proteins that were easily distinguished from the full-length protein expressed in the presence of pCMF (Fig. 1 C–E). The uniform incorporation of the amino acid for both RAD51^Y54pCMF and RAD51^Y315pCMF mutants was confirmed through Western blot analysis using α-RAD51 (3C10) antibodies (Fig. 1D) and by MALDI-TOF/MS as an ∼42-Da increase in molecular mass (Fig. S1). Protein yields were around 20% of the wild-type RAD51 expression owing to loss to insoluble fractions and truncated products. Mutants replacing the RAD51 Y54 and Y315 with aspartate (D), glutamate (E), and phenylalanine (F) were also purified (Fig. S1).

Fig. S1.

Incorporation of a nonnatural amino acid into RAD51 protein. (A) Expression system established for incorporation of the pCMF amino acid consists of E. coli Acella cells transformed with two plasmids. The first plasmid, pCH1-RAD51opt, contains a RAD51 ORF optimized for E. coli expression as well as the GroE operon from E. coli. The second plasmid contains components for the amber suppressor system. (B) Purification scheme for obtaining homogeneous RAD51 protein. (C) SDS-PAGE gel (4–15% gradient) showing purified RAD51 and phosphomimetic mutants (112.5 pmol in 15 µL of buffer containing 20 mM Hepes (pH 7.5), 10% glycerol, 150 mM NaCl, 1 mM DTT, 1 mM EDTA). All proteins were purified to near homogeneity (except RAD51Y315pCMF, which contained a small amount of truncated protein). The electrophoretic mobility of the wild-type RAD51 and all mutants was as expected for the 37-kDa protein, except for the RAD51Y315F, which consistently runs slightly faster. (D) Data from MALDI-TOF/MS experiments confirming incorporation of pCMF amino acid at the Y54 (Top) and Y315 (Bottom) residues, respectively.

Tyrosine 54 Phosphorylation Stimulates RAD51 DNA Strand Exchange Activity.

A critical HR function of the RAD51 nucleoprotein filament is to perform an ATP-dependent homologous DNA pairing and DNA strand exchange (1, 29). This activity can be reconstituted in vitro in the three-strand DNA strand exchange reaction, where a RAD51 nucleoprotein filament preformed on a circular ϕX174 virion ssDNA pairs with and invades the ϕX174 linear double-stranded DNA (dsDNA) (30, 31). The DNA strand exchange activity of RAD51 leads to the displacement of the complementary linear ssDNA from the dsDNA substrate, eventually leading to the formation of a nicked circular product through a series of joint-molecule intermediates (Fig. 2A). The reaction depends on the presence of the ssDNA binding protein RPA, which, when added after the RAD51 nucleoprotein filament is formed, stimulates the DNA strand exchange by destabilizing remaining secondary structure in the ssDNA and by sequestering the displaced DNA strand. The nicked circular product and the joint-molecule intermediates are separated on an agarose gel.

Fig. 2.

Y54 phosphorylation stimulates DNA strand exchange activity of RAD51. (A) RAD51-first DNA strand exchange reaction: Circular ϕX174 ssDNA is incubated with RAD51 in reaction conditions permitting ATP hydrolysis (Materials and Methods), followed by addition of RPA, which helps remove secondary structures in the ssDNA, allowing stable nucleoprotein filament formation. The RAD51 nucleoprotein then invades ϕX174 linear dsDNA to form nicked circular (NC) dsDNA products through several joint-molecule (JM) intermediates showing various stages of branch migration. Strand exchange reactions carried out by RAD51^Y54pCMF (B) and RAD51^Y315pCMF (C), both compared with the RAD51 wild-type protein. For each protein, the respective panel shows reactions stopped at 0, 30, 60, 120, and 180 min. All substrates, joint molecules, and nicked circular products are observed on a 0.8% Tris-Acetate-EDTA (TAE) agarose gel stained with SYBR Gold. (D and E) Quantitative analysis of formed nicked circular products in RAD51^Y54pCMF and RAD51^Y315pCMF strand exchange reactions. The RAD51^Y54pCMF mutant (red) is able to convert ∼75% of linear dsDNA substrate into nicked circular products compared with ∼40% of products formed for RAD51 wild-type (gray) and RAD51^Y315pCMF mutant (orange) converts ∼27% of linear dsDNA substrate into nicked circular products compared with ∼33% for the wild-type protein. All reactions were performed in triplicate, with data represented as mean ± SEM. The results for both mutants were compared with the wild-type protein, using a two-way ANOVA to verify the significance of results.

There are two general regimes under which human RAD51 promotes an efficient DNA strand exchange. The reaction can be carried out in the presence of Ca²⁺ and Mg²⁺ ions, which stabilize the RAD51 nucleoprotein filament by preventing ATP hydrolysis (32), or in the presence of Mg²⁺ and ammonium sodium phosphate (NaNH₄PO₄), which permits ATP hydrolysis (33). To measure the DNA strand exchange activity while permitting ATP hydrolysis, our reaction conditions contained Mg²⁺ and NaNH₄PO₄ (Materials and Methods). The RAD51 Y54 phosphomimetic mutants showed an enhanced DNA strand exchange activity compared with the wild-type protein (Fig. 2 B and D and Fig. S2A). Nicked circular products indicating completion of the reaction were observed earlier in the reaction facilitated by RAD51^Y54pCMF than in the reaction for the wild-type protein (Fig. S2A). We also observed an increase in product formed over a time course of 180 min from ∼40% to ∼75% of the input dsDNA (Fig. 2D). The RAD51^Y54D and RAD51^Y54E mutants displayed a higher stimulatory effect compared with RAD51^Y54pCMF, whereas RAD51^Y54F behaved identical to the wild-type protein (Fig. S3B). Two distinct species of the joint-molecule intermediates are formed during the strand exchange reaction (34). The early joint-molecule intermediates are known to migrate faster in the agarose gel compared with those intermediates formed further along reaction progression toward the nicked circular product (34). The stimulation of strand exchange activity can be attributed to the efficient conversion of early joint-molecule intermediates, which were observed in much higher amounts for the wild-type protein and RAD51^Y54F compared with RAD51^Y54pCMF (Fig. 2B and Fig. S2A), RAD51^Y54D, or RAD51^Y54E (Fig. S3B) mutants. Notably, in these “RAD51-first” reactions, the enhanced product formation by RAD51^Y54pCMF was due to its ability to convert joint molecules into the products, whereas the assimilation of the linear dsDNA substrate was essentially the same between the wild-type RAD51 and RAD51^Y54pCMF. RAD51^Y54D and RAD51^Y54E were also better than the wild-type protein in initiating the reaction, as evidenced by the faster utilization of the linear dsDNA substrate. Although mild stimulation of the DNA strand exchange was observed in the RAD51^Y315D or RAD51^Y315E mutants (Fig. S3C), the RAD51^Y315pCMF mutant showed strand exchange activity similar to the wild-type protein (Figs. 2 C and E and Fig. S2C). Considering the pCMF mutants have the closest structural resemblance to phosphorylated proteins, it is likely that the glutamate and aspartate mutations exaggerate the effects of tyrosine phosphorylation in RAD51.

Fig. S2.

Quantitative analysis of strand exchange reactions with RAD51 Y54 and Y315 pCMF mutants The appearance of the nicked circular (NC) products (black), disappearance of dsDNA substrates (red), and formation of JM (blue) were quantitated using ImageJ software normalized to the amount of initial linear dsDNA substrate. The intensities obtained for the mutants were then statistically compared with the wild-type strand exchange reaction. Comparison of strand exchange reactions between RAD51^Y54pCMF and wild-type RAD51 proteins when RPA was added after (A) and before (B) the addition of RAD51 protein. Comparison of strand exchange reactions between RAD51^Y315pCMF and wild-type RAD51 proteins when RPA was added after (C) and before (D) the addition of RAD51 protein.

Fig. S3.

Strand exchange reactions with RAD51 Y54 and Y315 mutants. (A) Schematic showing the in vitro reconstituted RAD51-first DNA strand exchange reaction (Fig. 2). JM, joint molecules; NC, nicked circular. (B) Strand exchange reactions comparing RAD51 wild-type with RAD51^Y54D, RAD51^Y54E, and RAD51^Y54F mutants, respectively. (C) Strand exchange reactions comparing RAD51 wild-type with RAD51^Y315D, RAD51^Y315E, and RAD51^Y315F mutants, respectively.

RAD51 Phosphomimetic Mutants Efficiently Overcome the Kinetic Barrier to the Nucleoprotein Filament Formation Presented by the ssDNA Binding Protein RPA.

The RPA protein plays multiple important roles in HR and in recombination-dependent DNA repair. While processing DSBs, the ssDNA produced by the end-resection machinery is immediately bound by RPA, preventing secondary structure formation (35). The bound RPA can be replaced by RAD51 nucleoprotein filament in a reaction facilitated by the recombination mediator BRCA2 (10, 36, 37). The in vitro reconstituted DNA strand exchange reaction requires RPA (30, 38), which removes secondary structures in the ssDNA (31, 39) and, as the reaction progresses, sequesters the displaced ssDNA, thereby preventing nonproductive RAD51 nucleoprotein complexes. When added first, however, RPA kinetically impedes the RAD51 nucleation (40). Thus, the order of RAD51 and RPA addition while performing the DNA strand exchange reaction has a very significant impact on the reaction outcome. As expected, under our experimental conditions, the efficiency of nicked circular product formation was significantly higher when RAD51-nucleoprotein filaments were formed before the addition of RPA (Fig. 3B). Although the appearance of the joint molecules and nicked circular products was delayed and diminished in the RPA-first reactions, it was not completely abolished. Not only did the Y54 phosphomimetic mutants display higher DNA strand exchange activity in the RAD51-first and RPA-first reactions but RAD51^Y54pCMF displayed no difference in the extent of the RAD51-first and RPA-first reactions (Fig. 3 C and E and Fig. S2B), suggesting that phosphorylated RAD51 may efficiently displace RPA from the ssDNA in the absence of recombination mediators. The RPA-first reactions by RAD51^Y54D and RAD51^Y54E yielded more nicked circular product than the wild type but were less efficient than RAD51^Y54pCMF (Fig. S4B). As with the RAD51-first reaction, RAD51^Y315pCMF was similar to the wild-type protein (Fig. 3 D and F and Fig. S2D). The only difference between RAD51^Y315pCMF and RAD51 was that RAD51^Y315pCMF had a slightly reduced capacity to assimilate linear dsDNA, which was balanced by a higher rate of clearing the joint-molecule intermediates. RAD51^Y315D and RAD51^Y315E were more robust DNA strand exchange proteins than the wild type, RAD51^Y315pCMF, and RAD51^Y315F, yet again suggesting that the glutamate and aspartate mutations inaccurately represent tyrosine phosphorylation (Fig. S4C).

Fig. 3.

RAD51^Y54pCMF performs efficient RPA-first DNA strand exchange. (A) RPA-first reaction: Circular ϕX174 ssDNA is incubated with RPA protein, followed by the addition of RAD51 protein, which displaces the bound RPA to form nucleoprotein filaments that can then invade ϕX174 linear dsDNA to form nicked circular (NC) products through a series of JM intermediates. (B–D) RPA-first DNA strand exchange reactions carried out by RAD51 (Top Left), RAD51^Y54pCMF (Top Right), and RAD51^Y315pCMF (Bottom Left). For each protein, the respective panel shows reactions stopped at 0, 30, 60, 120, and 180 min. (E and F) Quantitation of product formed in the DNA strand exchange reactions. The RAD51^Y54pCMF mutant (blue) converts ∼75% of linear dsDNA substrate into nicked circular products in the presence of RPA compared with ∼25% of products formed for RAD51 wild type (black), whereas RAD51^Y315pCMF mutant (cyan) is similar to wild-type protein in both RAD51-first and RPA-first reactions, where it is able to convert 20% of the linear substrate into nicked circular products. All reactions were performed in triplicate, with data represented as mean ± SEM. The results for both mutants were compared with the wild-type protein, using a two-way ANOVA to verify significance of results.

Fig. S4.

RPA-first strand exchange reactions with RAD51 Y54 and Y315 mutants. (A) Schematic showing the in vitro reconstituted RPA-first DNA strand exchange reaction (Fig. 3). (B) Strand exchange reactions comparing RAD51 wild-type with RAD51^Y54D, RAD51^Y54E, and RAD51^Y54F mutants, respectively. (C) Strand exchange reactions comparing RAD51 wild-type with RAD51^Y315D, RAD51^Y315E, and RAD51^Y315F mutants, respectively.

Tyrosine Phosphorylation Alters the Properties of RAD51-ssDNA and RAD51-dsDNA Nucleoprotein Filaments.

RAD51 can form nucleoprotein filaments on both ssDNA and dsDNA. However, only filaments formed on the ssDNA substrates are productive in homologous pairing and DNA strand exchange (41). The recombination mediator BRCA2 is known to promote a bias in binding affinity favoring the ssDNA (36, 37, 42). We investigated the effect of RAD51 phosphorylation on the RAD51–dsDNA and RAD51–ssDNA interactions by incubating the RAD51 phosphomimetic mutants with linearized ϕX174 dsDNA or ϕX174 ssDNA and analyzing the electrophoretic mobility-shift assay (EMSA). In all situations, we observed RAD51 binding to ssDNA and dsDNA, but there were differences in the formation of filament coaggregates (seen in our experiments as species in the wells that failed to enter the gel), as has been observed with the E. coli RecA protein (43, 44) (Fig. 4 and Fig. S5). At concentrations relevant to the strand exchange assay (i.e., 7.5 μM), the RAD51^Y54pCMF mutant was observed to bind and shift ssDNA and dsDNA substrates in the EMSA experiments but did not form the coaggregates similar to the coaggregates formed by RAD51 and RAD51^Y315pCMF (Fig. 4 A and B). The RAD5^Y54D and RAD51^Y54E mutants displayed similar DNA binding properties to the RAD51^Y54pCMF (Fig. S5 A and B). The RAD51^Y315D and RAD51^Y315E mutants, on the other hand, differed significantly from the RAD51^Y315pCMF mutant (Fig. S5 A and B). These mutants were unable to form the higher order aggregates characteristic of RAD51^Y315pCMF and RAD51. It is notable, however, that all mutants that displayed elevated DNA strand exchange activity were unable to form coaggregated filaments while binding ssDNA and dsDNA.

Fig. 4.

RAD51 phosphorylation limits large nucleoprotein complexes on both ssDNA and dsDNA. (A) EMSA RAD51 Y54 mutants binding linear ϕX174 dsDNA. Fifteen micromolar (base pair) dsDNA was incubated with increasing concentrations of 2.5 μM, 5.0 μM, or 7.5 μM RAD51 and loaded on a 0.9% TAE agarose gel and stained with SYBR Gold. RAD51^Y54pCMF showed a decrease in dsDNA binding compared with wild-type (WT) protein and RAD51^Y315pCMF. Protein/base pair ratios are indicated at the bottom of each lane. (B) In a similar assay containing 30 μM (nucleotide) ϕX174 circular ssDNA, RAD51^Y54pCMF is unable to form the higher order nucleoprotein filament complexes formed by the WT protein and RAD51^Y315pCMFmutant. Protein/nucleotide ratios are indicated at the bottom of each lane.

Fig. S5.

DNA binding properties of traditional RAD51 phosphomimetic mutants differ from both RAD51 wild-type and pCMF mutants when using longer substrates. (A) EMSA for RAD51 Y54D/E/F and Y315D/E/F mutants binding linear ϕX174 dsDNA. Fifteen micromolar (base pair) dsDNA was incubated with increasing concentrations of 2.5 μM, 5.0 μM, or 7.5 μM RAD51 and loaded on a 0.9% TAE agarose gel and stained with SYBR Gold. Aspartate (D) and glutamate (E) mutants showed a difference in dsDNA binding compared with wild-type protein as well as the pCMF mutants. Protein/base pair ratios are indicated at the bottom of each lane. (B) In a similar assay containing 30 μM (nucleotide) ϕX174 circular ssDNA, D and E mutants also showed different ssDNA binding characteristics compared with the wild-type pCMF mutants. Protein/nucleotide ratios are indicated at the bottom of each lane.

The Altered ssDNA Binding Behavior of Phosphomimetic Mutants.

To elucidate the regulation of the RAD51 nucleoprotein filament formation on ssDNA further, we investigated the RAD51–ssDNA interaction using a FRET-based assay (Materials and Methods). We monitored binding of RAD51 to a poly(dT)-60, which was labeled internally with Cy3 (FRET donor) and Cy5 (FRET acceptor) fluorescent dyes separated by 25 nucleotides. Free ssDNA yields a high FRET due to the proximity of Cy3 and Cy5. Addition of RAD51 into the reaction mixture and ensuing nucleoprotein filament formation straightens and extends the ssDNA, which can be measured as a decrease in FRET as the two fluorophores move away from one another (23). RAD51 protein binds and extends ssDNA ∼1.5-fold beyond its contour length, with each RAD51 monomer occluding approximately three nucleotides (Fig. 5A).

Fig. 5.

Effects of Y54 phosphorylation on ssDNA binding activity. (A) RAD51 binding to ssDNA was observed by following the extension of the 60-mer oligonucleotide poly(dT)-60 containing Cy3 (FRET donor) and Cy5 (FRET acceptor) fluorophores separated by 25 nucleotides (nt). Binding of RAD51 to 600 nM (nt) ssDNA moves the two dyes apart, which can be seen as a change from high FRET (0.55) to low FRET (0.19). Under conditions preventing ATP hydrolysis (Ca²⁺), the RAD51^Y54pCMF binds and extends the ssDNA substrate similar to wild-type RAD51, with an ∼1:3 stoichiometric RAD51/ssDNA ratio, whereas under reaction conditions permitting ATP hydrolysis (Mg²⁺), RAD51^Y54pCMF binding is nonstoichiometric. Data are represented as mean ± SEM with n = 3. (B) Analysis of the equilibrium ssDNA binding using smTIRFM. Distributions of the FRET states of the DNA substrates in the presence of the indicated concentrations of RAD51 (Left) and RAD51^Y54pCMF (Right) overlaid with the distributions of the FRET states in the absence of protein (Materials and Methods and Fig. S7). Unbound ssDNA (gray) yields a histogram centered on a FRET value of ∼0.5, whereas fully extended RAD51 nucleoprotein filament (blue) yields a histogram centered on a FRET value of ∼0.1. Concentrations of RAD51 or RAD51^Y54pCMF, as well the number of molecules used to build each histogram, are indicated in each panel.

The characteristic ssDNA binding and extension were observed in buffers that permitted ATP hydrolysis (Mg²⁺/NaNH₄PO₄), as well as under conditions that inhibited ATP hydrolysis (Ca²⁺/Mg²⁺/KCl) (Fig. 5A). Under both sets of experimental conditions, we observed a stoichiometric binding of the RAD51 wild-type protein to 600 nM (nucleotide) ssDNA (i.e., each RAD51 addition caused a decrease in the FRET signal proportional to the amount of added RAD51 until the ratio of one RAD51 to three nucleotides of ssDNA was reached, after which no additional change in FRET was observed). Such a stoichiometric binding regime is typically observed when the amount in the DNA substrate is significantly higher than the K_d. Under conditions inhibiting ATP hydrolysis, RAD51^Y54pCMF binding to and extension of the ssDNA were indistinguishable from the wild-type protein, whereas RAD51^Y54D and RAD51^Y54E showed weaker, nonstoichiometric binding to the ssDNA substrate (Fig. S6 A and B). Under conditions permitting ATP hydrolysis, binding of the RAD51^Y54pCMF to the ssDNA was no longer stoichiometric because higher concentrations of RAD51^Y54pCMF were required to saturate the ssDNA substrate compared with wild type (Fig. 5A). This binding deficiency was even more pronounced in RAD51^Y54D and RAD51^Y54E, which required approximately sixfold and ∼160-fold more protein, respectively, to saturate the ssDNA substrate (Fig. S6B). No difference between the wild-type RAD51 and RAD51^Y54F was observed. RAD51^Y315E and RAD51^Y315pCMF displayed ssDNA binding only slightly weaker than the wild type under ATP hydrolysis conditions and formed stable nucleoprotein filaments similar to wild type when ATP hydrolysis was inhibited (Fig. S6 C and D). All other RAD51 Y315 mutants showed binding characteristics similar to the wild-type protein under both permissive and inhibitory conditions for ATP hydrolysis. The inverse correlation between the ssDNA binding and the capacity to carry out the DNA strand exchange activity was somewhat unexpected. This observation is especially true for the RAD51^Y54D, which, in contrast to other mutants, does not seem to form the nucleoprotein filament on short oligonucleotides even at the concentrations exceeding the concentrations of the DNA strand exchange reactions. This apparent discrepancy can be due to the fact that RAD51 forms less stable complexes with short oligonucleotides (45). The same likely applies to the phosphomimetic mutants. ATP-bound RAD51 forms stable nucleoprotein filaments (46), whereas hydrolysis of ATP to ADP leads to the protein turnover. Using an NADH-coupled ATPase assay (47) (SI Materials and Methods), we measured the ssDNA-dependent ATP hydrolysis activity for all phosphomimetic mutants in the presence of a 100-mer poly(dT) ssDNA. All mutants could hydrolyze ATP similar to the wild-type protein. (Fig. S6E). This observation further confirms that the mobility-shifted species observed in the EMSA experiments using ϕX174 ssDNA do indeed correspond to the nucleoprotein filaments.

Fig. S6.

Altered ssDNA binding of RAD51 Y54 and Y315 mutants. (A) FRET-based assay to measure binding for RAD51 Y54 mutants to ssDNA. Under conditions inhibiting ATP hydrolysis (Ca²⁺), RAD51Y54D and RAD51Y54E mutants deviate from ideal RAD51 binding behavior. (B) This effect is significantly exaggerated under conditions permitting ATP hydrolysis (Mg²⁺). (C and D) Similar FRET-based experiments were conducted for Y315 mutants that were observed to have similar ssDNA binding characteristics under both prohibitive conditions for ATP hydrolysis (Ca²⁺) and permissive conditions (Mg²⁺). Data are represented as mean ± SEM with n = 3. (E) Rates of ssDNA-dependent ATP hydrolysis of all RAD51 mutants measured using the coupled NADH assay. Data are represented as mean ± SEM with n ≥ 3.

Equilibrium Single-Molecule Measurements Quantify the RAD51-ssDNA Binding.

To quantify the effect of the RAD51 Y54 phosphorylation on ssDNA binding, we used single-molecule total internal reflection fluorescence microscopy (smTIRFM). A FRET-based analysis (48) was used to follow binding of RAD51 to the individual surface-tethered partial DNA duplex molecules, where the Cy3 and Cy5 dyes were separated by 21 nucleotides and the overall ssDNA region was 60 nucleotides (Fig. S7). Similar to the bulk ssDNA-binding experiments described above, the Cy3 dye is excited with the green (532-nm) laser and the Cy5 dye is excited via FRET. Measurements collected for 3,839 RAD51-free DNA molecules from 15-s movies yielded a normally distributed FRET peak with an average FRET value of 0.5. A second smaller peak distributed around a FRET value of 0 represents the fraction of molecules that underwent acceptor photobleaching. Upon binding by RAD51, the donor and acceptor dyes move apart as the ssDNA molecule is stretched, which is observed as the change from high to low FRET. The DNA molecules fully extended by the bound RAD51 yielded a FRET distribution with an average FRET value of ∼0.1 (Fig. 5B). Notably, because these experiments monitor RAD51 binding to the individual DNA molecules, the concentrations of the protein that are required to saturate FRET signal can be directly related to the affinity of the wild-type and phosphomimetic proteins. Using this system, we compared the binding properties of the wild-type RAD51 with RAD51^Y54pCMF. The concentration dependence of these reactions recapitulates the data observed in bulk, except RAD51 binding here is not stoichiometric and therefore can be directly compared with RAD51^Y54pCMF. A clear difference was observed, wherein 250 nM RAD51 fully extended the ssDNA, whereas 2.5 μM RAD51^Y54pCMF was required to achieve similar results (Fig. 5B), which is approximately a 10-fold increase in the protein concentration required to form stable nucleoprotein filaments. Notably, these measurements predict that both proteins will bind and extend the ssDNA at concentrations used in the DNA strand exchange reaction, where the DNA concentration is significantly greater and the stoichiometric binding is expected.

Fig. S7.

smTIRF ssDNA binding assay. Poly(dT)-60 ssDNA substrate is immobilized to a biotinylated quartz slide using the biotin–Neutravidin interaction (Materials and Methods). Evanescent wave produced by a prism-based total internal reflection illumination and a 532-nm laser is used to excite the Cy3 dye on the surface-tethered DNA substrate. In the absence of protein, the acceptor Cy5 dye on the partial duplex DNA is excited via FRET. Upon binding, RAD51 extends the ssDNA substrate, leading to a decrease in energy transfer and reduction in Cy5 emission, with a corresponding increase in Cy3 intensity. Both Cy3 and Cy5 emission can be tracked simultaneously using a dual-view system. The change in Cy3 and Cy5 intensities on RAD51 binding to ssDNA substrate is tracked over time by recording movies over the course of the experiment.

RAD51 Nucleates on ssDNA via a Dynamic Stepwise Process Where Dimers of the Protein Engage and Disengage from the ssDNA.

To visualize the nucleation of the RAD51 filament on ssDNA in real time, we carried out smTIRFM flow experiments. Here, RAD51 was introduced into a reaction chamber containing tethered ssDNA molecules 10 s after the recording was initiated. Movies were recorded for 3 min at a time resolution of 100 ms. The fluorescence trajectories (time-based changes in Cy3 and Cy5 fluorescence originating from distinct spots on the slide) were collected. The trajectories that contained binding events were then corrected for donor leakage and trimmed to contain all events before donor or acceptor photobleaching (Fig. 6 A and B shows representative trajectories). Only events containing anticorrelated trajectories for donor and acceptor intensities were selected for further FRET analysis (Fig. S8 A and B). Selected trajectories were then analyzed collectively using empirical Bayesian FRET (ebFRET) (49) to determine the number of distinct states in the trajectories (Fig. 6 A and B). This program uses an empirical Bayesian method to generate hidden Markov models (HMMs) of complex FRET trajectories. In our experiments, on addition of 250 nM wild-type RAD51, filament nucleation was observed to proceed in a stepwise manner. The trajectories were globally fit to models for two to eight distinct FRET states. The best model was determined by calculating the statistical mean lower bound per series output from the ebFRET analysis (49). The model that showed the best fit was then visually compared with the raw data (Fig. 6 A and B). All data points for idealized states in the fit dataset were also plotted as frequency of the state vs. FRET. This frequency distribution was then used to estimate the global values for the FRET states. In the best-fit models, these frequency distributions had a Gaussian shape with the means for the individual states well separated. Models that overfit the dataset with a higher number of states had several overlapping distributions and could easily be discarded using this method. Our experimental data were best fit to a model described by four distinct states with FRET values between 0.5 and 0.1 (Fig. S8 C and D), indicating three steps. Similar analysis has been performed previously on the E. coli RecA and sceRad51 proteins, but with different outcomes (48, 50, 51). Because the donor and acceptor dyes are separated by 21 bases and each RAD51 monomer binds ssDNA as a triplet (8), we conclude that the basic unit for binding and extension of the human RAD51 nucleoprotein filaments is likely to be a dimer. Although no binding events were observed at 250 nM RAD51^Y54pCMF, higher protein concentrations yielded multiple trajectories with assembly and dissociation events for both the wild-type protein and RAD51^Y54pCMF. Both proteins were found to bind and extend the nucleating nucleoprotein filament as a dimer. To determine the dynamics of the nucleus assembly, the transitions between multiple FRET states were plotted as transition density plots (TDPs) (50) (Fig. 6C). A TDP represents the FRET state of a molecule before and after the transition on the horizontal and vertical axes, respectively. Interestingly, at 250 nM, RAD51 nucleation includes transitions in the forward (S1→S2→S3→S4) and reverse (S4→S3→S2→S1) directions with similar frequencies (Fig. 6D). This finding indicates a very dynamic process whereby RAD51 dimers rapidly bind and dissociate at each step of filament formation. At 2.5 μM, a bias was observed toward transitions between the final two states. S3→S4 transitions were the most prevalent, whereas S4→S3 transitions were slightly less frequent. These results show that the nucleoprotein filament nucleation proceeds in the forward direction. Analysis of the RAD51^Y54pCMF showed no binding events at a 250 nM protein concentration. However, the RAD51^Y54pCMF filament nucleation was highly efficient compared with wild-type RAD51 at higher concentrations. At 750 nM, RAD51^Y54pCMF was able to form filaments with a stronger forward directionality compared with the wild-type protein (Fig. 6E).

Fig. 6.

HMM analysis shows that RAD51 and RAD51^Y54pCMF nucleate on ssDNA as dimers, with RAD51 forming dynamic nuclei distinct from stable RAD51^Y54pCMF nuclei. Representative FRET trajectories from preequilibrium experiments visualizing RAD51 (A) and RAD51^Y54pCMF (B) nucleation onto ssDNA in real time. RAD51 protein was introduced into the reaction chamber at t ≈ 10 s. Regions containing transitions were trimmed and used for HMM (ebFRET). The idealized trajectories are overlaid on the raw FRET trajectories showing the fit. (C) Schematic explaining the 3D TDPs. Four distinct FRET states 1, 2, 3, and 4 correspond to respective states in the RAD51 nucleus formation on ssDNA, with three steps between these states required to extend the ssDNA substrate fully. Transitions 1→2→3→4 correspond to the filament formation, whereas filament disassembly is reflected in transitions in the reverse direction. (D) TDPs showing the transition densities corresponding to each state for the wild-type RAD51 protein at increasing concentrations. The transitions were calculated from 85, 50, and 146 molecules for 250 nM, 750 nM, and 2.5 μM concentrations, respectively. Brighter colors represent more frequent transitions. The frequency scale is shown to the right of the graphs. (E) TDP plots showing transition densities for the RAD51^Y54pCMF mutant. No transitions were observed at 250 nM concentrations. Transitions were calculated from 63 and 116 molecules for 750 nM and 2.5 μM concentrations, respectively.

Fig. S8.

HMM analysis shows that RAD51 and RAD51^Y54pCMF nucleate on ssDNA as dimers. Representative Cy3 (donor) and Cy5 (acceptor) trajectories from preequilibrium experiments visualizing RAD51 (A) and RAD51^Y54pCMF (B) nucleation onto ssDNA in real time are shown. RAD51 protein was introduced into the reaction chamber at t ≈ 10 s. Trajectories that showed anticorrelated behavior of the Cy3 and Cy5 signals were selected and corrected for background intensities as well as donor leakage. The processed Cy3 and Cy5 trajectories were then converted to FRET trajectories (blue). From each FRET trajectory, we selected the regions that include the first observed transition and all subsequent transitions until one of the dyes photobleaches, the transitions stop occurring (i.e., equilibrium is achieved), or the recording is terminated (gray box); the remainder of the trajectory was excluded from the analysis. The selected regions were collectively fit to the models, with the number of FRET states ranging from two to eight using ebFRET analysis. The black lines overlaying the FRET data (blue) represent the best fit to the trajectory, which uses four FRET states. AU, arbitrary units. Histograms show global distributions of the idealized FRET states from 281 and 179 molecules for RAD51 (C) and RAD51^Y54pCMF (D), respectively. Four distinct FRET states were observed with FRET values between 0.5 and 0.1. The peak for each state can be approximated by a Gaussian distribution with the center of each peak well separated from the other peaks/states. Because the Cy3-Cy5 FRET pair is 21 nucleotides apart, it is apparent that the RAD51 molecules bound ssDNA in three steps, where each step is a dimer.

It is important to note that the studies reported above were preequilibrium measurements that follow the initial steps of the filament nucleation process. When the reactions were allowed to reach an equilibrium (i.e., 5–10 min after the addition of RAD51 or RAD51^Y54pCMF to the surface-tethered DNA), we observed a stable low FRET signal indicative of the fully formed nucleoprotein filaments.

SI Materials and Methods

Protein Expression and Purification.

To achieve a robust expression system compatible with both the expression of soluble RAD51 protein and efficient pCMF incorporation, the RAD51 ORF was codon-optimized for E. coli expression and cloned into a pCOLADuet (Novagen) expression vector that also encoded for the E. coli GroE operon. The resulting plasmid is referred to herein as pCH1-RAD51opt. Protein incorporating pCMF was expressed in E. coli Acella cells [F⁻ ompT hsdS_B(r_B⁻ m_B⁻) gal dcm (DE3) ∆endA ∆recA]. These cells were transformed with the pCH1 vector encoding E. coli codon-optimized RAD51 Y54TAG or Y315TAG sequences, along with the pSUPT-UaaRS vector encoding the amber tRNA and pCMF-RS expressed under proK and araBAD promoters, respectively. E. coli Acella/RAD51opt cells from a glycerol stock were inoculated in 10 mL of LB containing 40 μg/mL kanamycin and incubated at 37 °C overnight. Eight milliliters of the overnight culture was used to seed a 1-L LB culture containing 40 μg/mL kanamycin. The culture was grown at 37 °C until an OD₆₀₀ of ∼0.6–0.8 was reached. RAD51 expression was then induced for 4 h at 37 °C with 0.1 mM IPTG. The cells were pelleted and stored at −80 °C. RAD51 protein was then purified using the protocol described previously (23). Briefly, cells were thawed on ice for 2 h and resuspended in lysis buffer containing protease inhibitors and lysozyme. The cells were then sonicated, and the soluble fraction was dialyzed against spermidine acetate buffer. Spermidine precipitates RAD51, which is selectively resuspended in buffers containing increasing amounts of sodium chloride. The fractions containing RAD51 are then purified through a Blue Agarose column followed by a heparin column. Finally, the protein is concentrated using a MonoQ anion exchange column. RAD51 pCMF mutants were expressed similar to the wild-type protein except that E. coli Acella/pCH1-RAD51/pSUPT-UaaRS starter cultures were seeded to 2-L culture volumes and grown in the presence of 34 μg/mL chloramphenicol and 1% arabinose in addition to 40 μg/mL kanamycin. These cultures were then induced with IPTG, along with 300 μg/mL pCMF (custom-synthesized by AsisChem & Syngene International Limited). The wild-type protein concentration was determined using a molar extinction coefficient of 14,900 M⁻¹⋅cm⁻¹ at A₂₈₀. The concentrations for all mutants were measured using a molar extinction coefficient of 13,410 M⁻¹⋅cm⁻¹ at A₂₈₀.

DNA Strand Exchange Assay.

A RAD51 DNA strand exchange assay was performed as described earlier using the ϕX174 virion ssDNA and the ϕX174 RFI dsDNA as substrates for the reaction (31). The ϕX174 RFI dsDNA was linearized using ApaLI to generate 4-bp, 5′ overhangs, and was purified using phenol-chloroform-isoamyl alcohol extraction. The RAD51-first DNA strand exchange reactions were initiated by incubating 7.5 μM RAD51 with 30 μM (nucleotide) ϕX174 ssDNA and 2.5 mM ATP in the reaction buffer [20 mM Hepes⋅NaOH (pH 7.5), 10% glycerol, 1 mM MgCl₂, 1 mM DTT] for 5 min at 37 °C. Then, 150 mM NaNH₄PO₄ and 2 μM RPA were added and incubated for another 5 min. This step was followed by addition of 15 μM (base pairs) of ApaLI-digested linear ϕX174 dsDNA and further incubation at 37 °C for 30, 60, 120, or 180 min. The samples were then deproteinized by addition of 0.8% SDS and Proteinase K (800 μg/mL), followed by incubation at 37 °C for 30 min, and resolved on a 0.8% agarose gel (UltraPure Agarose) at 3.5 V/cm (60 V) at 25 °C for 16 h. The gel was then stained for 10 min using SYBR Gold and destained for 20 min in MilliQ H₂O. In the RPA-first reactions, the order of addition of RAD51 and RPA was reversed. For strand exchange reactions with the RAD51^Y315pCMF protein, the protein concentrations were increased by 18% to compensate for the presence of the truncated RAD51^Y315* protein present in the preparation. The appearance of the nicked circular products was quantitated using ImageJ software normalized to the amount of initial linear dsDNA substrate. The intensities obtained for the mutants were then statistically compared with the wild-type strand exchange reaction using two-way ANOVA and verified for significance. All data were plotted using GraphPad Prism 6.

EMSA-Based dsDNA Binding Assay.

The RAD51 dsDNA binding assay was performed largely as described by Takizawa et al. (21). The reaction was initiated by incubating 2.5 μM, 5 μM, and 7.5 μM RAD51 with 15 μM (base pair) ϕX174 linearized dsDNA in reaction buffer from 4× reaction buffer stock [80 mM Hepes⋅NaOH (pH 7.5), 40% glycerol, 4 mM MgCl_2, 4 mM DTT, 0.4 mg/mL BSA] and 2.5 mM ATP and 150 mM NaNH₄PO₄ for 20 min at 37 °C. The samples were then mixed with 6× DNA loading dye and resolved on a 0.9% agarose gel at 3.5 V/cm (50 V) at 25 °C for 1 h. The gel was stained for 10 min using SYBR Gold and then destained for 10 min in MilliQ H₂O.

EMSA-Based ssDNA Binding Assay.

The ssDNA binding reaction was performed similar to the dsDNA binding assay by incubating 2.5 μM, 5 μM, and 7.5 μM RAD51 with 30 μM (nucleotide) ϕX174 circular (virion) ssDNA and resolved on a 0.9% agarose gel at 3.5 V/cm (50 V) at 25 °C for 1 h. The gel was then stained for 10 min using SYBR Gold and destained for 10 min in MilliQ H₂O.

RAD51 FRET-Based DNA Binding and Extension Assay.

Six hundred nanomolar (nucleotide) of dT-60 oligo labeled with the Cy3 and Cy5 dyes separated by 25 nucleotides was titrated with RAD51 protein in FRET reaction buffer [20 mM Hepes (pH 7.5), 5 mM CaCl₂, 5 mM MgCl₂, 150 mM KCl, 1 mM ATP, 1 mM DTT or 20 mM Hepes (pH 7.5), 150 mM NaNH₄PO₄, 2 mM MgCl₂, 1 mM ATP, 1 mM DTT] at 37 °C. Cy3 and Cy5 fluorescence was recorded using a Cary Eclipse Fluorimeter. FRET was calculated as a fraction of acceptor intensity relative to the total donor and acceptor intensity adjusted by correction factors:

$$\text{FRET}=\frac{4.2*I_{cy5}}{4.2*I_{cy5}+1.7*I_{cy3}}.$$

FRET values were plotted against the RAD51 concentration using GraphPad Prism 6.

Acquisition of Single-Molecule Data.

A diode-pumped solid-state 532-nm green laser was used to excite Cy3 molecules in an evanescent field produced in a reaction chamber. Scattered light was removed using a Cy3/Cy5 dual-bandpass filter (FF01-577/690; Semrock) in the emission optical pathway Images were chromatically separated into Cy3 images and Cy5 images using the 630-nm dichroic mirror inside the dual-view system (DV2; Photometrics). The data acquisition was carried out through an Andor IXON CCD camera, using software written in Visual C++. The movies obtained with the CCD were analyzed first using IDL, and the intensities of the fluorophores and the time traces were visualized using customized MATLAB programs. FRET was calculated as

$$FRET=\frac{1}{{\left(1+\gamma \frac{I_{Cy3}}{I_{Cy5}}\right)}^{-1}},$$

where $I_{Cy3}$ and $I_{Cy5}$ are the sensitized emission intensity of the donor and acceptor, respectively (59). The measured raw intensities of the donor and acceptor channels were corrected by measuring the leakage intensities from the donor channel β to the acceptor channel. The donor leakage was subtracted from the acceptor channel intensities and added back to the donor intensity. For our setup, the donor leakage correction was measured to be 7%. Applying this correction, we get the following equation for FRET:

$$\text{FRET}=\frac{(I_{Cy5}^{\ast }-\beta )}{(I_{Cy5}^{\ast }-\beta )+\gamma (I_{Cy3}^{\ast }+\beta )},$$

where $I_{Cy5}^{\ast }$ and $I_{Cy3}^{\ast }$ are the background-corrected acceptor and donor intensity, $\gamma =1$ is calculated as the ratio of change in the acceptor intensity, ΔI_Cy5 is calculated as the change in the donor intensity, and ΔI_Cy3 is calculated as the change upon acceptor photobleaching (γ = ΔI_Cy5/ΔI_Cy3) (60, 61).

Single-Molecule RAD51 DNA Binding Assay (TIRFM).

An oligo made of 18 random bases and labeled with a 5′ Cy5 dye and a 3′ biotin tag was annealed to a 78-base oligo containing the complementary 18 bases to the biotinylated oligo, followed by a T60 sequence and an internal Cy3 dye spaced 21 bases apart from the Cy5 dye on the tethered oligo. This arrangement ensures that all oligos tethered to the slide have a complete FRET pair. PEGylated slides and a sample chamber were prepared as described earlier (58). One hundred microliters of 0.2 mg/mL Neutravidin was passed through the reaction chamber and incubated for 3 min. The chamber was washed with 300 μL of T50 buffer [Tris (pH 8.0), 50 mM NaCl]. Fifty picomolar biotinylated DNA substrate was then added to the imaging buffer [12 mM Trolox, 20 mM Hepes (pH 7.5), 2 mM MgCl₂, 0.8% glucose, 150 mM NaNH₄PO₄, 1 mM ATP, 1 mM DTT, 0.1 mg/mL BSA, 0.04 mg/mL catalase, 1 mg/mL glucose oxidase] and flowed into the chamber. The indicated concentrations of RAD51 were then added into the imaging buffer and flowed into the imaging chamber. Forty movies, each with a duration of 15 s (100-ms time resolution, 400 background value, and 1,600 data scalar), were recorded using a gain of 230 at 45.6-mW green laser intensity. Data were analyzed using customized programs for IDL and MATLAB. The DNA substrates used for the experiment are 5Cy5-Bot18-3Bio (GCCTCGCTGCCGTCGCCA) and Top18-(T21)Cy3(T39) (TGGCGACGGCAGCGAGGCTTTTTTTTTTTTTTTTTTTTT/iCy3/TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT).

For experiments performed under flow, similar conditions were used to set up the experiment, except 75 pM DNA substrate was used instead of 50 pM to achieve a homogeneous distribution of molecules in the field of view, while accounting for a larger flow chamber. Once the DNA substrate was bound to the PEGylated slide, the imaging chamber was washed with the RAD51 imaging buffer. Using the same values for laser power and gain, 3-min movies were recorded at the same time resolution, with the indicated concentration of RAD51 being injected into the imaging chamber at ∼10 s into the movie recording.

Analysis of Single-Molecule Binding Data.

To generate FRET state histograms, five frames from each movie were used to calculate the FRET values for each frame for each molecule in the movie and plotted as a FRET histogram after the data were normalized in MATLAB.

The time-dependent trajectories obtained from flow experiments were analyzed using HMMs. Empirical Bayesian methods were used to determine transitions indicating RAD51 filament formation. The leakage-corrected trajectories were trimmed to include all data points until donor/acceptor photobleaching events. The ebFRET analysis suite (49) was used to determine statistically the FRET states observed in the course of the RAD51 nucleoprotein filament formation. Mean lower bound was used to determine the evidence for the model that best described the number of states for each series. Histograms were plotted to show the distributions of measured idealized FRET values for every state within each trajectory. A model that describes the states most accurately has distinct means with minimal overlap. Overfit data containing false states were seen as histograms that were completely overlapped or masked by histograms representing other states (Fig. S8 C and D). This finding was confirmed by visually comparing the idealized trajectories from the model with the raw data to confirm the fit.

ATPase Assay.

A coupled NADH assay was performed as described earlier (47). A reaction mixture contained 20 mM Hepes⋅NaOH (pH 7.5), 5 mM MgCl₂, 150 mM NaNH₄PO₄, 1 mM DTT, 0.2 mg/mL NADH, 7.5 mM phosphoenol pyruvate, 2.5 mM ATP, 20 units of lactate dehydrogenase and pyruvate kinase, and 7.5 μM RAD51. Absorbance at 340 nm was measured at 37 °C using an Agilent 8453 UV-Vis Spectrophotometer. Eighty micromolar nucleotides of poly(dT)-100 were added to initiate the DNA-dependent ATP hydrolysis. The slope of the decrease of NADH absorbance at 340 nm was used to calculate the rate of ATP hydrolysis using the following conversion: rate of A₃₄₀ decrease (s⁻¹) × 9,880 = rate of ATP hydrolysis (μM⋅min⁻¹). Data were plotted using GraphPad Prism 6.

Discussion

Here, we addressed the biochemical consequences of RAD51 phosphorylation by c-Abl/BCR-ABL using phosphomimetic mutants that more closely mimic the effect of tyrosine phosphorylation. First, we showed that RAD51^Y54pCMF protein has enhanced DNA strand exchange activity on the plasmid-length substrates (Fig. 2 B and C). Our findings are in contrast to previous biochemical studies, where phosphorylation of Y54 on RAD51 was thought to be inhibitory (16), but agree, for the reasons discussed below, with the cellular studies where RAD51 foci formation, presumably at the sites of DNA damage, was enhanced after c-Abl phosphorylation (20). These earlier biochemical studies were performed using sceRad51 phosphorylated in vitro by human c-Abl. Although the tyrosine residue analogous to Y54 is conserved in sceRad51, yeast cells lack tyrosine phosphorylation (22), resulting in different mechanisms of regulation for the yeast and human proteins. Moreover, the residue corresponding to the RAD51 Y315 in sceRad51 is a valine. Our studies mechanistically address and parse out the effects of the two tyrosine phosphorylation events on RAD51 activities.

We observed that the Y54 phosphorylation enhances the recombinase activity of RAD51 in at least two important ways. First, RAD51^Y54pCMF efficiently promotes nicked circular DNA formation, the DNA strand exchange product, whereas joint-molecule intermediates accumulated to a much greater extent in the reactions using the unmodified RAD51 protein (Fig. 2B and Fig. S2A). Human RAD51 readily forms DNA joint molecules but, in contrast to its bacterial functional homolog RecA, is inefficient at their branch migration (30). Y54 phosphorylation appears to alleviate this deficiency, allowing for rapid clearing of the intermediates and formation of the DNA strand exchange products.

Second and most unexpectedly, RAD51^Y54pCMF was able to enhance the so-called RPA-first reaction, allowing it to reach the extent of the RAD51-first reaction (Fig. 3C). The ssDNA binding protein RPA plays both inhibitory and stimulatory roles at different HR steps. RPA stimulates DNA strand exchange by removing regions of the secondary structure in the ssDNA, which allows RAD51 to form a contiguous nucleoprotein filament (39), and also sequesters the displaced DNA strand, thereby assisting in progression of the reaction from the joint-molecule intermediates to the nicked circular products by competing with RAD51 for free ssDNA (31). If added to the ssDNA first (RPA-first reaction), RPA delays the RAD51 nucleoprotein filament formation by kinetically inhibiting the nucleation step (40). The robust DNA strand exchange reaction promoted by RAD51^Y54pCMF suggests that the c-Abl–phosphorylated RAD51 may nucleate on the RPA-coated ssDNA and form contiguous active nucleoprotein filaments even in the absence of a recombination mediator. However, a scenario where the DNA damage response leads to coordinated activation of several simultaneous pathways that ensure faithful and efficient recombination is more likely. BRCA2 has been shown to diffuse in the nucleus together with RAD51, delivering it to the sites of DNA damage (52), with each BRCA2 binding four to five RAD51 monomers and promoting the filament nucleation (36, 53). Phosphorylation of RAD51 may then ensure the filament growth on the RPA-coated ssDNA, because the RAD51 focus formation is reduced in c-Abl^−/− cells (20).

What properties of the phosphorylated/phosphomimetic RAD51 contribute to this enhanced activity? The nucleoprotein filament formed by RAD51^Y54pCMF is different from the unmodified RAD51 nucleoprotein filament. The difference between two nucleoprotein filaments manifests in the higher concentrations of RAD51^Y54pCMF that are required to form stable nuclei on ssDNA, as evident from the solution and single-molecule FRET studies. The ostensible discrepancy between the enhanced DNA strand exchange activity of RAD51^Y54pCMF and its reduced capacity to bind ssDNA was unexpected. The reduced ssDNA binding could be due to a reduced affinity of RAD51^Y54pCMF for ssDNA or to altered binding cooperativity; alternately, because RAD51 seems to nucleate on ssDNA by addition of dimers, reduced ssDNA binding could also be a result of the altered RAD51 monomer/monomer interface, which, in turn, leads to a reduced capacity to form dimers. Neither of these two properties is expected to affect the RAD51 binding to ssDNA at the micromolar protein concentrations used in the DNA strand exchange reactions because both the wild-type and phosphomimetic protein should bind stoichiometrically under these conditions. The payoff for the altered monomer/monomer interface in RAD51^Y54pCMF is its capacity to form what seem to be effective nuclei in a directional manner by consecutive addition of dimers. This step is important in the RPA-first DNA strand exchange reaction, where efficient nucleation allows RAD51^Y54pCMF to take a foothold on the ssDNA transiently released by the oligonucleotide binding (OB)-folds of RPA during microscopic dissociation events. In contrast, the wild-type protein goes through multiple attempts at assembly of stable nuclei, and therefore is less efficient in competing with RPA. Although our data cannot explain why RAD51^Y54pCMF has an enhanced capacity to clear the joint molecules, the mechanics of its nucleoprotein filament are likely responsible. Indeed, the filaments formed by RAD51^Y54pCMF on long ssDNA and dsDNA are very different from the wild type. The faster electrophoretic mobility exhibited by these RAD51^Y54pCMF–ssDNA complexes may reflect both the more dynamic filaments and the absence of coaggregated multifilament networks characteristic of the wild-type RAD51. Additionally, potentially less stable or more dynamic nucleoprotein filaments may be more effective at overcoming barriers, migrating joint molecules, and thereby completing the DNA strand exchange reaction.

RAD51 nucleoprotein filaments were previously shown to grow from heterogeneous nuclei ranging in size from dimers, and even monomers, to large oligomers (54). Using smTIRFM, we were able to follow the formation kinetics of these nuclei at a single-monomer resolution. We observed no RAD51 monomers binding to ssDNA but showed that both RAD51 and RAD51^Y54pCMF seem to engage the ssDNA as dimers. The nucleus then grows by dynamic addition and dissociation of the RAD51 dimers. It is important to note here that the single-molecule approaches similar in resolution to ours were previously used to show addition/dissociation of the monomers of E. coli RecA (48) and sceRad51 to a nucleoprotein filament (51). Although we can exclude the monomer binding/dissociation for human RAD51, we cannot formally rule out a scenario where two adjacent monomers within larger RAD51 oligomers make a contact with ssDNA at each step. At intermediate- and high-protein concentrations, growth of the RAD51^Y54pCMF nucleus is more directional than growth of the unmodified RAD51, which manifests in fewer transitions in the reverse direction (Fig. 6 D and E). The lack of the DNA binding at 250 nM RAD51^Y54pCMF may be due to the decreased oligomerization of the phosphomimetic protein brought about by a distorted monomer/monomer interface. Y54 is located in the N-terminal domain of RAD51 near the monomer/monomer interface. In the unmodified protein, it participates in the stacking interaction with F195 of the adjacent monomer (23). Introduction of the negative charge upon Y54 phosphorylation or due to the presence of phosphomimetic residue may affect this interface so that the RAD51 oligomerization is delayed until higher protein concentrations are attained. Although RAD51 nucleates on ssDNA at much lower protein concentrations, these complexes are dynamic, unstable, and likely nonproductive with respect to formation of the recombination-proficient nucleoprotein filaments, and are poor competitors of RPA. On the other hand, RAD51^Y54pCMF, although unable to nucleate at lower protein concentrations, forms more stable recombination-proficient nucleoprotein filaments at higher protein concentrations.

In contrast to RAD51^Y54pCMF, RAD51^Y315pCMF displayed DNA strand exchange activities essentially similar to DNA strand exchange activities of the unmodified protein. The only minor difference we observed was a slightly slower uptake in the dsDNA substrate balanced out by the faster clearance of the joint molecules, and resulting in the same rate and extent of the nicked circular product formation as the wild-type protein. These data are consistent with a sequential phosphorylation of RAD51 proposed earlier (55, 56). According this this model, c-Abl phosphorylates RAD51 on Y315 by binding the consensus PXXP motif through its SH3 domain. Phosphorylation of Y315 produces a pYXXP motif, which is a consensus binding motif for the SH2 domain of c-Abl. The Y315 phosphorylation, in turn, leads to phosphorylation of the Y54 residue, which up-regulates DNA repair activity through stimulation of the RAD51-mediated DNA strand exchange.

Materials and Methods

Protein Expression and Purification.

To achieve a robust expression system compatible with both the expression of soluble RAD51 protein and efficient pCMF incorporation, the RAD51 ORF was codon-optimized for E. coli expression and cloned into a pCOLADuet (Novagen) expression vector that also encoded for the E. coli GroE operon. The resulting plasmid is herein referred to as pCH1-RAD51opt. Protein incorporating pCMF was expressed in E. coli Acella cells transformed with the RAD51^Y54TAG or RAD51^Y315TAG sequences, along with the pSUPT-UaaRS vector. RAD51 protein was then purified using the protocol described previously (23). Human RPA was purified as described earlier (57). Detailed purification procedures are provided in SI Materials and Methods.

DNA Strand Exchange Assay.

A RAD51 DNA strand exchange assay was performed as described earlier using the ϕX174 virion ssDNA and the ϕX174 RFI dsDNA as substrates for the reaction (31). The nicked circular products were quantitated using ImageJ software (NIH) normalized to the amount of initial linear dsDNA substrate. The intensities obtained for the mutants were then statistically compared with the wild-type strand exchange reaction using two-way ANOVA and verified for significance. All data were plotted using GraphPad Prism 6. Details of substrate and reagent preparation are provided in SI Materials and Methods.

RAD51 FRET-Based DNA Binding and Extension Assay.

The FRET-based DNA binding assay for RAD51 was performed as described by Subramanyam et al. (23). The details for analysis of fluorescence data are provided in SI Materials and Methods.

EMSA-Based DNA Binding Assays.

EMSA-based RAD51 DNA binding assays were performed largely as described by Takizawa et al. (21). The reactions were initiated by incubating 2.5 μM, 5 μM, and 7.5 μM RAD51 with 15 μM (base pair) ϕX174 linearized dsDNA or 30 μM (nucleotide) ϕX174 circular (virion) ssDNA in reaction buffer and 2.5 mM ATP and 150 mM NaNH₄PO₄ for 10 min at 37 °C. The samples were then mixed with 6× DNA loading dye and resolved on a 0.8% agarose gel at 3.5 V/cm (50 V) at 25 °C for ∼1 h. The gel was then stained for 10 min using SYBR Gold and destained for 10 min in MilliQ H₂O.

Single-Molecule RAD51 DNA Binding Assay (TIRFM).

Acquisition of single-molecule data is described in SI Materials and Methods. PEGylated slides and the sample chamber were prepared as described earlier (58). One hundred microliters of 0.2 mg/mL Neutravidin was passed through the reaction chamber and incubated for 3 min. Fifty picomolar biotinylated DNA substrate was flowed into the chamber. The indicated concentrations of RAD51 were then added to the imaging buffer and flowed into the imaging chamber. Forty movies of 15-s duration were recorded. Data were analyzed using customized programs for IDL and MATLAB (MathWorks) software. Further details are provided in SI Materials and Methods.

For experiments performed under flow, similar conditions were used to set up the experiment, except 75 pM DNA substrate was used instead of 50 pM to account for a larger flow chamber.

Analysis of Single-Molecule Binding Data.

The Cy3 and Cy5 intensities were corrected, and FRET was calculated as described earlier (59–61). To generate FRET state histograms, FRET values of each experiment histogram after the data were normalized to the total number of data points in MATLAB.

The time-dependent trajectories obtained from flow experiments were analyzed using HMMs. Empirical Bayesian methods were used to determine transitions indicating RAD51 filament formation. The ebFRET analysis suite (49) was used to determine statistically the FRET states observed in the course of RAD51 nucleoprotein filament formation.