Nascent DNA Synthesis During Homologous Recombination Is Synergistically Promoted by the Rad51 Recombinase and DNA Homology

Abstract

In this study, we exploited a plasmid-based assay that detects the new DNA synthesis (3′ extension) that accompanies Rad51-mediated homology searching and strand invasion steps of homologous recombination to investigate the interplay between Rad51 concentration and homology length. Mouse hybridoma cells that express endogenous levels of Rad51 display an approximate linear increase in the frequency of 3′ extension for homology lengths of 500 bp to 2 kb. At values below ∼500 bp, the frequency of 3′ extension declines markedly, suggesting that this might represent the minimal efficient processing segment for 3′ extension. Overexpression of wild-type Rad51 stimulated the frequency of 3′ extension by ∼3-fold for homology lengths <900 bp, but when homology was >2 kb, 3′ extension frequency increased by as much as 10-fold. Excess wild-type Rad51 did not increase the average 3′ extension tract length. Analysis of cell lines expressing N-terminally FLAG-tagged Rad51 polymerization mutants F86E, A89E, or F86E/A89E established that the 3′ extension process requires Rad51 polymerization activity. Mouse hybridoma cells that have reduced Brca2 (Breast cancer susceptibility 2) due to stable expression of small interfering RNA show a significant reduction in 3′ extension efficiency; expression of wild-type human BRCA2, but not a BRCA2 variant devoid of BRC repeats 1–8, rescues the 3′ extension defect in these cells. Our results suggest that increased Rad51 concentration and homology length interact synergistically to promote 3′ extension, presumably as a result of enhanced Brca2-mediated Rad51 polymerization.

MITOTIC homologous recombination (HR) has multiple cellular functions, including restarting stalled or collapsed replication forks (Haber 1999; Petermann and Helleday 2010), maintaining the integrity of telomeres (Dunham et al. 2000), and repairing DNA double-strand breaks (DSBs)/gaps resulting from endogenous and exogenous genotoxic agents (Pâques and Haber 1999; Dronkert and Kanaar 2001; Hartledorde and Scully 2009). Accurate HR maintains genome integrity and aids in preventing tumorigenesis (Flores-Rozas and Kolodner 2000; Heyer et al. 2010; Moynahan and Jasin 2010).

Homologous recombination is a complex, multi-step process (reviewed in Brugmans et al. 2007). In eukaryotes, the central step in the repair of a DNA DSB involves the binding of multiple monomers of the RAD51 protein (a functional homolog of the Escherichia coli RecA recombinase) to 3′-ending single-stranded DNA overhangs created by nucleolytic resection (Van Den Bosch et al. 2003). The resulting RAD51 nucleoprotein filament promotes formation of a joint molecule between the processed broken DNA and the homologous repair template by the orchestrated steps of homology searching and DNA strand invasion and exchange. Joint molecule formation is followed by new DNA synthesis, which replaces nucleotides lost through DSB formation and 3′-end resection (Pâques and Haber 1999; Symington 2002; Li and Heyer 2008). Depending on the subpathway of HR, subsequent steps may involve unwinding (dissolution) of the DNA strand containing the newly synthesized DNA and ligation to the processed second end or the formation of a stably joined Holliday junction intermediate that can be processed further by structure-specific endonucleases yielding recombinant products (Li and Heyer 2008).

At the heart of the HR process is the Rad51 (or RecA) nucleoprotein filament, whose formation requires the initial association of four to five monomers with 3′-ending single-stranded DNA in the step known as nucleation (Galletto et al. 2006; Van Der Heijden et al. 2007). In vivo, nucleation is hindered by single-stranded DNA binding protein in E. coli and replication-protein A (RPA) in eukaryotes, whereas it is assisted by recombination mediators that include RecBCD and RecFOR in E. coli (Anderson and Kowalczykowski 1997; Morimatsu and Kowalczykowski 2003), Rad52 in yeast (Sugiyama and Kowalczykowski 2002), and BRCA2 (Breast cancer susceptibility 2) in mammals (Jensen et al. 2010; Liu et al. 2010; Thorslund et al. 2010). BRCA2 regulates the formation of Rad51 nucleoprotein filaments through eight BRC repeats, which bind Rad51 monomers, selectively loading them onto single-stranded, but not double-stranded, DNA overhangs (Bignell et al. 1997; Chen et al. 1998; Carreira et al. 2009).

Single-molecule studies reveal that, while both RecA and Rad51 form extended, right-handed nucleoprotein filaments on single-stranded DNA, filaments of RecA are generally continuous, while Rad51 filaments feature Rad51-coated regions interspersed between segments of uncoated single-stranded DNA (Galletto et al. 2006; Van Der Heijden et al. 2007; Modesti et al. 2007; Hilario et al. 2009; Marijn et al. 2009). The “patchy” nature of the Rad51 nucleoprotein filaments might make them more flexible since the bare single-stranded DNA regions could behave like hinges (Ristic et al. 2005). Such flexibility would likely be advantageous in the early homologous pairing and strand-exchange steps of HR (Van Der Heijden et al. 2007; Holthausen et al. 2010).

Studies in yeast have investigated the kinetics and genetic requirements of many steps in the HR process (White and Haber 1990; Haber 1995; Aylon et al. 2003; Sugawara and Haber 2006; Hicks et al. 2011). While mammalian cells contain functional analogs of yeast recombination proteins, their role in HR, especially in the early steps of strand invasion and new DNA synthesis, is not well understood. Previously, we transfected mammalian cells with linearized plasmid DNA bearing homology to a chromosomal target gene and detected the nascent DNA formed by polymerizing 3′ ends (3′ extension) in vivo (Si et al. 2010). In the present study, we investigated the requirement for homology and the role of Rad51 and Brca2 proteins in the process of 3′ extension. Efficient 3′ extension requires wild-type Rad51 and Brca2 proteins. Furthermore, wild-type Rad51 and homology length interact synergistically to promote 3′ extension. We discuss the role of Rad51 polymerization and formation of the Rad51 nucleoprotein filament in this process.

Materials and Methods

Cell lines and plasmids

The origins of the igm482 hybridoma and the igm482 derivatives WT5 and WT16 expressing N-terminal FLAG-tagged mouse Rad51 have been described (Köhler et al. 1982; Rukść et al. 2007). The Brca2-depleted 12–21 cell line is an igm482 derivative that stably expresses anti-mouse Brca2 small interfering RNA (siRNA) (Lee and Baker 2007), while 12–21 derivatives that express wild-type human BRCA2 (in BAC clone RP11-777I19; Sharan et al. 2004) were established as described (Magwood et al. 2012). All cell lines were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 13% bovine calf serum, penicillin/streptomycin, and 2-mercaptoethanol as described (Köhler and Shulman 1980; Köhler et al. 1982). When required, DMEM was supplemented with the appropriate selectable agents, G418 (600 µg/ml), puromycin (7.5 µg/ml), or hygromycin (700 µg/ml).

Plasmid DNA was propagated in E. coli (DH5α) and extracted using the PureLink HiPure plasmid maxiprep kit (Life Technologies). Restriction enzymes were purchased from New England BioLabs (Mississauga, Ontario) and used according to the manufacturer’s instructions. Mutant Rad51 alleles F86E and A89E, along with the F86E/A89E double mutant were constructed by site-directed mutagenesis of wild-type murine Rad51 complementary DNA (cDNA) inserted into pUC19 (Rukść et al. 2007). The mutant Rad51 cDNA was excised from the pUC19 vector and ligated into a p3XFLAG-CMV-10 (Sigma-Aldrich) derivative bearing a puromycin resistance (Puro^R) gene, thus permitting the Rad51 protein to be expressed as an N-terminal 3XFLAG-tag fusion product. All Rad51 constructs were verified by DNA sequencing. To generate transformants, 50 µg of XmnI-linearized plasmid DNA was electroporated into 2 × 10⁷ viable igm482 cells as described in Si et al. (2010). The number of viable cells was enumerated by trypan blue exclusion using a Bright-Line Hemacytometer (Hausser Scientific). Recovery of individual Puro^R transformants was performed by limited diluted cloning (Rukść et al. 2007).

Plasmids pT∆Cµ_211/2290, pT∆Cµ_517/2290, pT∆Cµ_720/2290, pT∆Cµ_858/2290, and pT∆Cµ_2070/2290 bearing different amounts of homology to the chromosomal µ-gene target sequence were constructed by standard techniques (Sambrook et al. 1989) in a pSV2neo (Southern and Berg 1981) backbone and are detailed in Results.

Protein analysis

Cytoplasmic and nuclear fractions were prepared from control and mitomycin C (MMC)-treated cells using the ProteoExtract Subcellular Proteome Extraction kit (Calbiochem). Western blot and immunoprecipitation analyses were performed according to Magwood et al. (2012) with the exception that rabbit IgG serum (ThermoScientific) was used as a nonspecific antisera in immunoprecipitation. The following primary antibodies were used: mouse monoclonal anti-human Rad51 (14B4, Abcam), mouse monoclonal anti-FLAG (M2, Sigma-Aldrich), mouse monoclonal anti-β-actin (AC-15, SigmaAldrich), mouse monoclonal anti-human histone H1 (A-E4, Santa Cruz Biotechnology), rabbit polyclonal anti-human caspase-3 (SA-320, Enzo Life Sciences), and rabbit polyclonal anti-human BRCA2 (Ab27976, Abcam). Immunoblot signals were detected with ECL-Prime reagent (GE Healthcare) using the appropriate HRP-coupled goat anti-mouse IgG (Southern Biotech) or goat anti-rabbit IgG (Jackson Immunoresearch) secondary antibodies.

Measurement of 3′ extension during homologous recombination

The formation of nascent DNA that follows the strand invasion event of homologous recombination (3′ extension) was measured by PCR as described previously by Si et al. (2010). The intensity of specific PCR bands was quantified by densitometric analysis using BioRad Gel Doc instrumentation and Quantity One imaging software (Version 4.4.6, BioRad).

Statistical analysis

Statistical analysis was performed by one-way analysis of variance (ANOVA) and Tukey’s-HSD (Honestly Significant Difference) test, chi-square goodness-of-fit test (χ²), or a t-test using JMP-10 statistical software (SAS). Significance was assessed at P ≤ 0.05. Error bars represent the standard error of the mean.

Results

Influence of homology length on the efficiency of new DNA synthesis

Previously, we described a novel assay that detects the new DNA synthesis that accompanies the early strand invasion step of homologous recombination in vivo (Si et al. 2010). In this assay, recipient mouse hybridoma cells are electroporated with a gene-targeting vector in which BstEII digestion creates a 1.2-kb double-stranded gap (DSG) within the region of homology to the single-copy hybridoma chromosomal immunoglobulin µ-gene (Figure 1A). During homologous recombination, chromosomal µ-gene sequences serve as the template for new DNA synthesis primed by invading 3′ vector ends positioned to the “left” or “right” of the vector DSG. In this vector (pT∆Cµ_858/2290), 858 and 2290 bp of µ-gene homology reside to the left and right of the 1.2-kb vector DSG, respectively. Detection of 3′ extension events involves plasmid DNA extraction at various post-electroporation time points and PCR using specific primers. To detect 3′ extension events from the left invading arm, one primer (neoF20′) binds to complementary sequences in the vector backbone, while the newly synthesized DNA forms the second primer binding site (CµR1), generating a 1.2-kb product. Similarly, primer pair ampR20/CµF1 is specific for a 2.4-kb product indicative of 3′ extension from the right invading vector arm. A separate primer set (neoR+6/neoF-1) monitors recovery of the pT∆Cµ vector backbone via detection of a 1.2-kb neomycin phosphotransferase (neo) gene product. For quantification, a series of plasmid copy number standards are subjected to the same PCR analysis. In this way, the PCR product resulting from amplification of the newly synthesized DNA along with that from the recovered vector backbone can be quantified by agarose gel analysis and densitometry, permitting determination of the efficiency of the 3′ extension/vector backbone. To examine 3′ extension events proceeding further into chromosomal µ-sequences excluded from the gapped vector (tract length), the additional primer pairs neoF20′/CµR2, neoF20′/CµR3, and neoF20′/CµR4 that generate left arm 3′ extension products of 1.4, 1.8, and 2.2 kb, respectively, were utilized (Figure 1A). The distance of the various Cµ-specific primers from the vector BstEII site is also indicated.

Figure 1.

Influence of homology length on the efficiency of new DNA synthesis. (A) Schematic of the pTΔCμ gene-targeting vector used in the 3′ extension assay. The curved thick line represents pSV2neo (Southern and Berg 1981) vector sequences while the thin line represents a segment of the chromosomal immunoglobulin μ-heavy chain gene—originally, a 4.3-kb XbaI μ-gene segment from which a 1.2-kb BstEII fragment was deleted. Following electroporation, the left and right invading arms of the BstEII-linearized vector can pair with homologous sequences in the chromosomal immunoglobulin µ-gene and prime new DNA synthesis (3′ extension). The relative binding sites of the primer pairs used to detect 3′ extension or to quantify the vector backbone are shown as black arrows. The sizes of the PCR products obtained with various primer pairs along with the relative distance of the Cµ-specific primer 3′ ends to the vector BstEII site are indicated. (B) Schematic illustrating vectors with varying lengths of homology in the left arm with constant right arm homology length. The diagrams in A and B are not drawn to scale. (C) The effect of homology length on 3′ extension. Each data point represents the mean 3′ extension frequency ± SEM for three independent electroporations and two PCR reactions per electroporation. Error bars not shown are embedded within the point marker.

To examine the relationship between homology length and the efficiency of 3′ extension, we constructed a series of vectors, which, following BstEII cleavage, generated left invading arms bearing differing amounts of homology to the target chromosomal µ-gene. In each vector, the BstEII site is the same, and differences in homology length were achieved by removing sequences at the µ:vector backbone junction. As shown in Figure 1B, pT∆Cµ_211/2290 bears a 211-bp SacI/BstEII µ-fragment, pT∆Cµ_517/2290 bears a 517-bp AvaI/BstEII µ-fragment, pT∆Cµ_720/2290 bears a 720-bp AflII/BstEII µ-fragment, pT∆Cµ_858/2290 bears an 858-bp XbaI/BstEII µ-fragment, and pT∆Cµ_2070/2290 bears a 2070-bp µ-fragment engineered to begin 3′ of the µ-gene switch (Sµ) region and extending to the BstEII site. In each vector, the homology in the right invading arm remains a constant 2290-bp BstEII/XbaI µ-fragment (Figure 1B). The various BstEII-linearized vectors were electroporated separately into recipient igm482 hybridoma cells. After 6 hr of incubation [shown previously to result in maximum 3′ extension (Si et al. 2010)], plasmid DNA was extracted and PCR-amplified with primers neoF20′/CµR1 (to measure 3′ extension) and neoR+6/neoF-1 (to measure vector backbone recovery). As shown in the polynomial regression in Figure 1C (R² = 0.99), the frequency of 3′ extension/vector backbone displays an approximate linear response for homology lengths > ∼500 bp. However, the slope becomes severely acute below this value, suggesting that this homology length may approach the minimum efficient processing segment (MEPS) (Shen and Huang 1986) required to promote 3′ extension.

Wild-type Rad51 overexpression increases the frequency, but not the length, of 3′ extension events

To examine the requirement for wild-type Rad51 protein in 3′ extension, we exploited vector pT∆Cµ_858/2290 bearing left and right arm homologies of 858 and 2290 bp, respectively (Si et al. 2010) (Figure 1A) along with hybridoma cell lines WT5 and WT16 that express ectopic N-terminal FLAG-tagged wild-type mouse Rad51 protein in ∼2.0- and 0.9-fold excess, respectively (Rukść et al. 2007). Using primer pair neoF20′/CµR1 (Figure 1A), we observed a significantly higher frequency of 3′ extension from the 858-bp left invading vector arm in hybridoma cell lines WT5 and WT16 compared to control igm482 cells at all time points between 3 and 18 hr (Figure 2A). As discussed previously (Si et al. 2010), the 3′ extension frequency declines at the later time points, likely due to degradation of the extrachromosomal plasmid. A similar profile was observed using primer pair ampR20/CµF1 (Figure 1A) specific for 3′ extension from the right invading vector arm bearing 2290 bp of homology to the chromosomal µ-gene (Figure 2B). For both left and right invading vector arms, cell lines WT5 and WT16 feature peak 3′ extension frequencies at ∼9 hr compared to ∼6 hr in control igm482 cells. The latter value is consistent with the 3- to 6-hr peak in the 3′ extension reported previously in igm482 cells (Si et al. 2010). Also, at time points up to and including the peak at 9 hr, WT5 cells expressing more FLAG-tagged wild-type Rad51 exhibit a significantly higher frequency of 3′ extension than WT16 cells (Figure 2, A and B). Thus, ectopic expression of wild-type Rad51 prolongs the kinetics and stimulates the frequency of 3′ extension in a concentration-dependent manner.

Figure 2.

Effect of excess wild-type Rad51 and homology length on 3′ extension efficiency and tract length. The kinetics of 3′ extension for (A) left (858-bp homology) and (B) right (2290-bp homology) invading vector arms in control igm482 cells expressing only endogenous levels of Rad51 and in igm482 transformants WT5 and WT16 expressing N-terminal FLAG-tagged wild-type Rad51. Each data point represents the mean 3′ extension frequency ± SE of three independent electroporations and two PCR reactions per electroporation. An asterisk (*) above a data point indicates that the mean frequencies of 3′ extension in the WT5 and WT16 cell lines are statistically different from igm482 at that time point, whereas, a dagger (†) indicates that the mean frequencies of 3′ extension in cell line WT5 are statistically significant from WT16 at that time point (t-test, P ≤ 0.05). (C) Effect of excess Rad51 on 3′ extension tract length. (D) Ratios of right- and left-arm 3′ extension in control igm482, WT5, and WT16 cell lines. The ratios are calculated from the peak 3′ extensions shown in A and B. Statistical significance is indicated by an asterisk (*) (χ²-test, P ≤ 0.05). (E) Ratios of WT5 or WT16/igm482 3′extension using either the left or right invading arms. Asterisks (*) indicate significant differences between right and left invading arms (χ²-test, P ≤ 0.05). The ratios are calculated from data shown in A and B.

In the above studies, the detection of 3′ extension requires new DNA synthesis to extend only 63 and 13 nt past invading left and right ends of the vector DSG, respectively (Figure 1A). Previously, Si et al. (2010) reported that, as the measured distance into the DSG increases beyond 200 nt, the frequency of 3′ extension declines rapidly. To determine whether ectopic expression of wild-type Rad51 increases the average new DNA synthesis tract length, we utilized primers pairs neoF20′/CµR1, neoF20′/CµR2, neoF20′/CµR3, and neoF20′/CµR4 to detect 3′ extension events from the left invading arm that proceed 63, 210, 615, and 974 nt into chromosomal µ-sequences excluded by the vector DSG (Figure 1A). As shown in Figure 2C (and similar to that in Figure 2A), WT5 cells display a significant increase in 3′ extension compared to control igm482 cells at a distance of 63 nt from the vector BstEII site. However, WT5 exhibits the same decline in 3′ extension as igm482 as the distance into the DSG increases. Therefore, while higher Rad51 concentrations increase the frequency and prolong the kinetics of the 3′ extension, the average 3′ extension tract length is not increased.

Synergistic interaction between longer homology and increased Rad51 concentration in promoting 3′ extension

Further analysis of the data in Figure 2, A and B, permitted calculation of right/left arm 3′ extension ratios for cell lines igm482, WT5, and WT16 at their respective peak time points (Figure 2D). In control igm482 cells, the 3′ extension frequency was ∼2-fold higher for the right compared to left invading vector arm. Given the linear increase in 3′ extension with homology lengths > ∼500 bp (Figure 1C), the ∼2-fold enhancement is consistent with the larger homology on the right (2290 bp) vs. left (858 bp) invading arms. Surprisingly, the ratios of right to left 3′ extension were significantly higher in WT5 and WT16 (∼9- and 7-fold, respectively). Analysis of WT5/igm482 and WT16/igm482 3′ extension ratios (Figure 2E) revealed that, for the shorter left invading vector arm, WT5 and WT16 are ∼3- and 2-fold higher than igm482, respectively, whereas for the longer right invading vector arm, they are ∼11- and 7-fold higher, respectively. The above information suggests that excess Rad51, together with longer right arm homology, might interact synergistically to elevate the frequency of the 3′ extension in the WT5 and WT16 cell lines.

To eliminate the issue of sequence differences between the left and right invading vector arms in this analysis, we performed additional studies to determine 3′ extension frequencies in the WT5 and control igm482 cell lines utilizing vectors in which the different lengths of homology were restricted to the left arm only (Figure 1B). For each vector, the efficiency of 3′ extension was determined using the same neoF20′/CµR1 primer pair and, as above, vector backbone recovery was monitored by PCR amplification of the neo gene using the primer pair neoR+6/neoF-1. Although the kinetic analysis spanned 24 hr, for convenience, the panels in Figure 3, A–D, present only the 3-, 6-, and 9-hr time points that include the peak 3′ extension efficiencies for igm482 and WT5 (Figure 2, A and B). Using vector pT∆Cµ_211/2290 in which left arm homology is reduced to 211 bp (and below the putative MEPS value of ∼500 bp presented in Figure 1C), the efficiency of 3′ extension at each time point is low, although at the 6- and 9-hr time points, the frequencies of 3′ extension are significantly higher in WT5 cells (Figure 3A). For homology lengths of ≥720 bp, the 3′ extension frequency is significantly higher in WT5 at every time point (Figure 3, B–D). Moreover, closer inspection of the peak 3′ extension frequencies (Figure 3E) reveals that, for homology lengths <858 bp, WT5 is consistently higher than igm482 by ∼3-fold, but for the 2070 bp homology segment, this value is ∼10-fold. Therefore, based on the above information, we conclude that increased Rad51 concentration synergizes with homology length to increase the frequency of 3′ extension.

Figure 3.

Synergistic effect of excess Rad51 and homology length. The kinetics of 3′ extension were assayed in control igm482 and WT5 cells using vectors with left arm homology lengths varying from 211 to 2070 bp (A–D). Asterisks (*) indicate that mean frequencies of 3′ extension at the peak time point were statistically different from the other two time points tested within the cell line (t-test, P ≤ 0.05), while the dagger (†) indicates that the mean frequencies of 3′ extension in cell line WT5 are statistically significant from igm482 at that time point (t-test, P ≤ 0.05). Data represent mean 3′ extension frequencies ± SEM of three independent electroporations and two PCR reactions per electroporation. (E) Ratios of WT5/igm482 3′ extensions for differing homology lengths. Statistical significance is indicated by an asterisk (*) (χ²-test, P ≤ 0.05).

Role of Rad51 polymerization in association of Rad51 and Brca2

Rad51 polymerization is required for nucleoprotein filament formation, which in turn promotes the homology searching and strand invasion steps required for DSB repair during HR (Pâques and Haber 1999; Symington 2002; Li and Heyer 2008). To examine the requirement for Rad51 polymerization in 3′ extension, we generated independent hybridoma cell lines expressing ectopic N-terminal FLAG-tagged mouse Rad51 mutants bearing single (F86E and A89E) and double (F86E/A89E) amino acid changes in the Rad51 self-polymerization motif (85-GFTTATE-91) (Pellegrini et al. 2002) (Figure 4A). The transgene FLAG-Rad51/endogenous Rad51 ratio reveals similar levels of FLAG-Rad51 protein in cell lines expressing the A89E (#2, lane 3; #9, lane 4) or F86E (#11, lane 5; #7, lane 6) single mutants. However, significantly lower levels of FLAG-Rad51 protein are observed in cell lines expressing the F86E/A89E double mutant (#2, lane 7; #14, lane 8), suggesting that this Rad51 variant is more toxic to the hybridoma cells. As reported previously (Magwood et al. 2013), slight variation in the endogenous Rad51 levels in the various transformants is due to multiple Rad51 translational start sites that result in a transgene contribution to the level of “endogenous” Rad51.

Figure 4.

Interaction of Rad51 polymerization mutants with endogenous Rad51 and Brca2. (A) Western blot analysis of Rad51 protein in control igm482 hybridoma cells (lane 1) and transformants expressing N-terminal FLAG-tagged wild-type Rad51 (WT5, lane 2), and single (A89E-2, A89E-9, F86E-11, F86E-7, lanes 3–6, respectively) or double (F86E/A89E-2 and F86E/A89E-14, lanes 7 and 8, respectively) Rad51 polymerization mutants. (B) To examine interactions between FLAG-tagged Rad51 and endogenous Rad51, whole-cell extracts (WCEs) were immunoprecipitated with a rabbit anti-FLAG polyclonal antibody and immunoblotted with mouse anti-Rad51 primary antibody. The WT5 WCE was also incubated with control rabbit IgG and analyzed alongside the experimental samples as a negative control (lane 7). (C) To determine if the polymerization mutants retained the ability to interact with wild-type Brca2, WCEs (top panel) were immunoprecipitated with either anti-FLAG rabbit polyclonal (middle panel) or anti-rabbit IgG (bottom panel) and then immunoblotted with anti-Brca2 (Ab27976) antibody.

To examine whether the Rad51 mutants interact with endogenous Rad51, whole-cell extracts were immunoprecipitated with anti-FLAG polyclonal antibody, and Western blot analysis was performed with anti-human Rad51 (14B4) antibody (Figure 4B). Anti-FLAG immunoprecipitates reveal that endogenous 37-kDa Rad51 is associated with the 42-kDa FLAG-tagged wild-type Rad51 in cell line WT5 (lane 2), and a similar pull-down of endogenous Rad51 is observed for cell lines expressing A89E (#2 and #9 in lanes 3 and 4, respectively) and F86E (#11 and #7 in lanes 5 and 6, respectively). In contrast, anti-FLAG immunoprecipitates of independent cell lines expressing the FLAG-tagged Rad51 double mutant (F86E/A89E) (#2 and #14 in lanes 8 and 9, respectively) reveals a paucity of associated endogenous Rad51. Control immunoprecipitates from igm482 (lane 1) and WT5 cells immunoprecipitated with anti-IgG serum (lane 7) validate the specificity of the pull-downs. Therefore, in contrast to a previous report (Pellegrini et al. 2002), we conclude that the F86E and A89E single mutants are capable of interacting with endogenous Rad51, but that the double mutant (F86E/A89E) is compromised for Rad51 association.

We also assayed for interactions between representative Rad51 polymerization mutants and endogenous mouse Brca2 (Figure 4C). Western blot analysis reveals a similar level of endogenous mouse Brca2 in the various cell lines (Figure 4C, top panel). As shown in the middle panel in Figure 4C, a substantial amount of endogenous Brca2 is co-immunoprecipitated by the anti-FLAG antibody in WT5 cells expressing FLAG-tagged wild-type Rad51 (lane 1) and those expressing the F86E (lane 2) and A89E (lane 3) single polymerization mutants. In contrast, a lower amount of Brca2 is co-immunoprecipitated by the anti-FLAG antibody in the F86E/A89E double mutant (lane 4), a result that likely reflects the lower amount of FLAG-Rad51 protein expression in this cell line (Figure 4A, lane 7). The specificity of the pull-downs is validated by the negligible reactivity observed using control rabbit IgG (Figure 4C, bottom panel). Therefore, like FLAG-tagged wild-type Rad51 protein in WT5 cells, we conclude that FLAG-tagged single (F86E and A89E) and FLAG-tagged double (F86E/A89E) Rad51 polymerization mutants retain the ability to interact with endogenous Brca2.

DNA-damage-induced nuclear redistribution of Rad51 polymerization mutants

To determine if the Rad51 polymerization mutants retained the capacity for nuclear entry, the cellular distribution of endogenous and FLAG-tagged Rad51 proteins was examined in the presence and absence of MMC-induced DNA damage (Figure 5). Nuclear entry of endogenous Rad51 in control igm482 cells (Figure 5A) and of endogenous and FLAG-tagged wild-type Rad51 in WT5 cells (Figure 5B) is observed following MMC-induced DNA damage. Similarly, MMC-induced nuclear entry of both endogenous and FLAG-tagged Rad51 is observed in cell line F86E-11 (Figure 5C), A89E-2 (Figure 5D) and the double mutant, F86E/A89E-2 (Figure 5E). Thus, although wild-type Rad51 and the various Rad51 polymerization mutants are primarily cytoplasmic, they display clear nuclear redistribution following MMC-induced DNA damage. To verify the purity of nuclear and cytoplasmic fractions, the blots were reprobed with anti-histone H1 and anti-caspase-3 antibodies.

Figure 5.

Cellular distribution of Rad51 polymerization mutants. Western blot analysis was used to determine the cellular localization of Rad51 in control igm482 cells (A), WT5 cells expressing wild-type FLAG-Rad51 (B), and cell lines expressing FLAG-tagged single F86E and A89E (C and D, respectively) and double F86E/A89E (E) Rad51 polymerization mutants in the absence of DNA damage (UT-untreated) and at various time points of exposure to 600 nM MMC. The first blot in A–E was probed with mouse monoclonal anti-human Rad51 antibody (14B4), which detects both FLAG-tagged and endogenous Rad51. The blots were stripped and reprobed with mouse monoclonal anti-FLAG antibody (M2), which recognizes the three N-terminal FLAG epitopes in the FLAG-Rad51 fusion protein, thereby producing a stronger (amplified) signal than the anti-human Rad51 (14B4) antibody (second blot in A–E). The strong anti-FLAG signal masks somewhat the increase in FLAG-Rad51 nuclear entry observed with the 14B4 antibody. Histone H1 and caspase-3 Western blot analysis verified the purity of cellular fractionation (third and fourth blots, respectively, in A–E).

Rad51 polymerization is required for 3′ extension

Representative cell lines expressing the various Rad51 polymerization mutants were examined in the 3′ extension assay (Figure 6). The kinetics and frequencies of 3′ extension in cell lines expressing the Rad51 single polymerization mutants (A89E-2 and F86E-11) are not significantly different from the control igm482 cell line. In contrast, in the Rad51 double mutant F86E/A89E-2 in which Rad51:Rad51 subunit interactions are severely compromised (Figure 4B, lane 8), the frequency of 3′ extension is significantly reduced.

Figure 6.

Rad51 polymerization activity is required for efficient 3′ extension. The kinetics of 3′ extension were measured in representative cell lines A89E-2 and F86E-11 expressing the single Rad51 polymerization mutants or in cell line F86E/A89E-2 expressing the double Rad51 polymerization mutant. Each data point represents the mean 3′ extension frequency ± SE of three independent electroporations and two PCR reactions per electroporation. An asterisk (*) above a data point indicates that the mean frequencies of 3′ extension in F86E/A89E-2 are significantly lower than igm482, A89E-2, and F86E-11 (t-test, P ≤ 0.05).

Brca2 is required for 3′ extension

To determine the requirement for Brca2 in 3′ extension, we utilized mouse hybridoma cell line 12–21 that is ∼75% siRNA depleted for mouse Brca2 (Lee et al. 2009) along with 12–21 derivatives that express either ectopic wild-type human BRCA2 or a human BRCA2 mutant bearing an in-frame deletion of BRC repeats 1–8 (Magwood et al. 2012, 2013). As shown in Figure 7, at 6 hr, the frequency of the 3′ extension/vector backbone is significantly reduced in the mouse Brca2 knockdown cell line 12–21 compared to control igm482 cells. Moreover, the deficiency in the 3′ extension is augmented by ectopic expression of wild-type human BRCA2, but not by the BRCA2 variant devoid of BRC repeats 1–8. Therefore, we conclude that endogenous levels of wild-type mouse Brca2 are important for the 3′ extension process and that supplementing a deficiency in mouse Brca2 through ectopic expression of wild-type human BRCA2 augments the 3′ extension and, in fact, elevates it above the control. Importantly, the BRCA2 mutant devoid of Rad51-binding BRC repeats is unable to enhance the 3′ extension above the background level of the Brca2-depleted 12–21 cell line. Therefore, in accordance with working models (Jensen et al. 2010; Liu et al. 2010; Thorslund et al. 2010), Brca2 (BRCA2) likely promotes the 3′ extension indirectly by orchestrating formation of recombinationally active Rad51 nucleoprotein filaments.

Figure 7.

Requirement for Brca2 in 3′ extension. Analysis of the 3′ extension kinetics from the left invading vector arm (858-bp homology) in Brca2 siRNA-depleted cells (12–21), cells overexpressing human BRCA2 (12–21 hBRCA2), and cells overexpressing a hBRCA2 engineered mutant lacking all BRC repeats (12–21 hBRCA2_BRCΔ1–8). Each bar indicates the mean frequency of 3′ extension ± SEM of three independent electroporations and two PCR reactions per electroporation at 6 hr. Means indicated by the same lowercase letter are not statistically different at P ≤ 0.05 according to one-way ANOVA and Tukey’s-HSD analysis.

Discussion

During the initial steps of eukaryotic homologous recombination, broken DNA is subjected to nucleolytic processing. This creates 3′ single-stranded DNA ends that become coated with Rad51, eventually forming the Rad51 nucleoprotein filament that is required for homology searching and strand invasion. Subsequently, repair of the broken DNA is made possible by nascent DNA synthesis primed by the invading Rad51-coated 3′ ends (Pâques and Haber 1999; Li and Heyer 2008). Our 3′ extension assay utilizes a plasmid vector (pT∆Cµ) that bears a 1.2-kb DSG within the region of homology to the single-copy chromosomal immunoglobulin µ-heavy chain gene in mouse hybridoma cells and reports nascent DNA synthesis that follows the early homology search and strand invasion events of homologous recombination (Si et al. 2010).

To examine the optimal homology for 3′ extension, we constructed pT∆Cµ derivatives in which homology on the invading arm to the left of the vector DSG varied between 211 and 2070 bp, while homology on the invading arm to the right of the vector DSG remained constant at 2290 bp. The frequency of the 3′ extension increased linearly with homology lengths above 517 bp, but below this value, the efficiency of the 3′ extension declined markedly. Therefore, a homology length of ∼500 bp likely represents a MEPS (Shen and Huang 1986) that is required for the homology searching and strand invasion steps that precede the 3′ extension. Interestingly, this value is lower than what has been previously established for recombination between dispersed homologous sequences, such as the ∼1–2 kb required for gene targeting (Shulman et al. 1990; Hasty et al. 1991; Thomas et al. 1992) and the ∼2–4 kb required for ectopic homologous recombination (Baker et al. 1996). Considerably less homology of between ∼163 and 300 bp promotes intrachromosomal recombination between closely linked sequences (Rubnitz and Subramani 1984; Liskay et al. 1987; Waldman and Liskay 1988). In contrast to the process of 3′ extension, which requires only homology searching, strand invasion, and new DNA synthesis, both gene targeting and ectopic homologous recombination require the additional steps of establishing and resolving a recombination intermediate. It is possible that a smaller amount of homology suffices to promote the initial interactions that give rise to 3′ extension and that the full-fledged recombination intermediate that is needed for successful gene targeting or ectopic homologous recombination requires the stabilization provided by additional homology.

Previously, we showed that cell lines siRNA-depleted for endogenous Rad51 are crippled for 3′ extension (Magwood et al. 2013). Therefore, in this study, it was of interest to further examine how Rad51 concentration influences the 3′ extension process. Significantly higher levels of 3′ extension were observed in the cell lines WT5 and WT16 expressing ectopic FLAG-tagged wild-type Rad51 compared to control igm482 cells. Furthermore, the efficiency of 3′ extension is dependent on the concentration of wild-type Rad51 as indicated by the significantly higher level of 3′ extension in cell line WT5 expressing the highest level of FLAG-tagged wild-type Rad51 (Rukść et al. 2007). Interestingly, the peak 3′ extension frequency was prolonged in cell lines expressing ectopic wild-type Rad51. Both the enhancement in 3′ extension frequency and the observed prolongation in 3′ extension kinetics supports the scenario whereby excess wild-type Rad51 promotes a more efficient homology search and increases the stability of the strand invasion intermediate that initiates 3′ extension. Interestingly, the basic 3′ extension process itself remains a relatively short tract in nature: that is, despite the increase in wild-type Rad51 protein in WT5 cells, the frequency of 3′ extension still declines after ∼200 nucleotides similar to that in control igm482 cells. This supports the involvement of a short tract polymerase in the 3′ extension process (Si et al. 2010).

The above information suggests that increased Rad51 concentration enhances the stability and efficiency of Rad51-nucleoprotein filament formation, which in turn promotes 3′ extension. We surmised that further increases in 3′ extension frequency could be achieved by the combined effect of higher Rad51 concentration and increased homology. Using vector pT∆Cµ_858/2290, in which homology on the right arm is more than double that on the left arm (Figure 1A) along with cell lines WT5 and WT16 expressing excess wild-type FLAG-tagged Rad51, we did indeed observe a synergistic interaction between Rad51 concentration and homology length in promoting 3′ extension; we noted a higher-than-expected ratio in right/left arm 3′ extension in cell lines WT5 and WT16 compared to control igm482 cells (Figure 2D) and higher WT5/igm482 and WT16/igm482 3′ extension ratios for the right vs. left invading vector arms (Figure 2E). Nevertheless, we were concerned that sequence differences between the right and left invading vector arms might have been a contributing factor. Therefore, we performed 3′ extension assays in WT5 and control igm482 cell lines using pT∆Cµ derivatives in which homology changes of between 211 and 2070 bp were confined to the left invading vector arm only while homology on the right invading vector arm was kept constant at 2290 bp. Strikingly, these experiments confirmed that 3′ extension was stimulated synergistically by the combination of increased Rad51 protein and homology length.

How might Rad51 and long regions of homology interact synergistically to enhance 3′ extension? Spatial Rad51 clusters nucleated by Brca2 provide sites for Rad51 self-polymerization (Jensen et al. 2010; Liu et al. 2010; Thorslund et al. 2010): on longer homology segments and with excess Rad51 available, Rad51 nucleation and self-polymerization might be more rapid, promoting more efficient homology searching and strand invasion and subsequently increasing the efficiency of 3′ extension. To further expand upon this model, we sought to determine the requirement for Rad51 self-polymerization in interaction with endogenous Rad51 and Brca2 proteins and in nuclear localization and, ultimately, in 3′ extension. Accordingly, we generated independent hybridoma cell lines stably expressing ectopic N-terminal FLAG-tagged mouse Rad51 mutants bearing single (F86E and A89E) and double (F86E/A89E) amino acid changes in the Rad51 self-polymerization motif (85-GFTTATE-91) identified by Pellegrini et al. (2002). The F86E/A89E double mutant was crippled in its ability to interact with endogenous Rad51 protein, but in contrast to a previous report (Pellegrini et al. 2002), we did observe interactions between the F86E and A89E single mutants and endogenous Rad51. MMC-induced DNA damage promoted nuclear relocalization of endogenous wild-type Rad51, FLAG-tagged wild-type Rad51, FLAG-tagged Rad51 F86E and A89E single mutants, and the FLAG-tagged Rad51 F86E/A89E double mutant. Since the F86E/A89E double mutant is the most severely compromised in its interaction with endogenous Rad51, but can still access the nuclear compartment, Rad51 nuclear entry does not appear to depend on formation of Rad51 polymers. Both the F86E and A89E single mutants and the F86E/A89E double-mutant protein interact with endogenous mouse Brca2. These results are consistent with previous observations (Pellegrini et al. 2002) that GFP-tagged Rad51 F86E and A89E proteins retain the ability to bind BRCA2 and suggest that BRCA2 (Brca2) is capable of loading both wild-type and polymerization mutant Rad51 proteins onto single-stranded vector DNA in our study.

Contrary to the observed stimulation in 3′ extension as a result of ectopic expression of FLAG-tagged wild-type Rad51 (WT5 and WT16) described above, we observed that the frequency and kinetics of 3′ extension in cell lines expressing the F86E and A89E single mutants were similar to that in control igm482 cells. This suggests that even though the F86E and A89E single mutants can interact with endogenous wild-type Rad51 and are proficient for DNA damage-induced nuclear relocalization, they must perturb some aspect of the Rad51 nucleoprotein filament, as overexpression of wild-type Rad51 would normally stimulate 3′ extension. In contrast, expression of the F86E/A89E double mutant that is severely compromised in Rad51:Rad51 subunit interactions resulted in a significant reduction in 3′ extension. Therefore, expression of F86E/A89E generates a dominant-negative effect to the activity of endogenous Rad51 protein in stimulating the new DNA synthesis that accompanies homologous recombination. We conclude that the in vivo process of 3′ extension requires the intrinsic polymerization feature of Rad51 monomers to form the recombinationally active form of the Rad51 nucleoprotein filament.

Our previous studies support a role for BRCA2 (Brca2) in regulating 3′ extension in vivo: moderate expression of human BRCA2 in the igm482 hybridoma (which expresses endogenous mouse Brca2) increased 3′ extension and HR, while high BRCA2 overexpression suppressed 3′ extension and HR (Magwood et al. 2012). To gain further understanding of the requirement for BRCA2 (Brca2) in 3′ extension, we exploited the stable mouse Brca2 knockdown cell line, 12–21 (Lee and Baker 2007). In 12–21, the process of 3′ extension is compromised, but this can be augmented by overexpression of wild-type human BRCA2 supporting previous results for gene targeting in this cell line (Lee and Baker 2007; Magwood et al. 2012). Furthermore, the inability of the human BRCA2 variant bearing an in-frame deletion of the Rad51-binding BRC repeats 1–8 (BRCA2_∆1–8) to alleviate the deficit in 3′ extension in the recipient 12–21 cell line supports an important role for BRCA2 BRC repeats in loading/stabilizing Rad51 during nucleoprotein filament formation in the early steps of homologous recombination in vivo. Therefore, we conclude that wild-type Brca2/BRCA2 is vital to the process of 3′ extension, likely through its ability to orchestrate formation of a recombinationally active Rad51 nucleoprotein filament.

In summary, expression of ectopic wild-type Rad51 stimulates the efficiency of 3′ extension above the level observed in cell lines harboring endogenous levels of Rad51. Both Rad51 polymerization activity and a biological balance of Rad51 and Brca2 are required for efficient 3′ extension, presumably through formation of a recombinationally active Rad51 nucleoprotein filament. Strikingly, excess Rad51 and long regions of homology interact synergistically to enhance the frequency of 3′ extension. Mechanistically, in the presence of excess Rad51, Brca2-assisted nucleation of clusters of Rad51 foci, especially on longer segments of homologous DNA, is predicted to enhance the degree of Rad51 self-polymerization. Subsequently, Rad51 nucleoprotein filaments bearing long Rad51-coated regions might feature a greater-than-linear relationship in promoting 3′ extension. Also, whereas, RecA nucleoprotein filaments are continuous and stiff, longer Rad51 nucleoprotein filaments might be more flexible due to frequent gaps that create spaces between Rad51 nucleation centers (Holthausen et al. 2010). Thus, the observed synergism between Rad51 concentration and longer lengths of homology may result from the enhanced pairing that is possible when longer, more flexible Rad51 nucleoprotein filaments are formed. In addition, longer Rad51 nucleoprotein filaments might be more efficiently stabilized by structure-specific binding proteins that favor formation of a homologous recombination intermediate destined to undergo 3′ extension.