RAD51C Is Required for Holliday Junction Processing in Mammalian Cells

See allHide authors and affiliations

Science  09 Jan 2004:
Vol. 303, Issue 5655, pp. 243-246
DOI: 10.1126/science.1093037


During genetic recombination and the recombinational repair of chromosome breaks, DNA molecules become linked at points of strand exchange. Branch migration and resolution of these crossovers, or Holliday junctions (HJs), complete the recombination process. Here, we show that extracts from cells carrying mutations in the recombination/repair genes RAD51C or XRCC3 have reduced levels of HJresolvase activity. Moreover, depletion of RAD51C from fractionated human extracts caused a loss of branch migration and resolution activity, but these functions were restored by complementation with a variety of RAD51 paralog complexes containing RAD51C. We conclude that the RAD51 paralogs are involved in HJprocessing in human cells.

Homologous recombination provides an important mechanism for the repair of DNA double-strand breaks, a process that is essential for the maintenance of genome stability and tumor avoidance. In eukaryotes, our knowledge of the late steps of this process is limited, and we rely on the bacterial RuvABC-Holliday junction complex, or resolvasome, to provide a paradigm for studies of recombination intermediate–processing enzymes from higher organisms (1). A mammalian HJ resolvase with associated adenosine 5′-triphosphate (ATP)–dependent branch migration activity has previously been detected in fractionated extracts (2, 3), but efforts to determine the identity of any proteins required for resolvasome activity have until now been unsuccessful.

To provide enriched fractions containing HJ branch migration and resolution activities, we fractionated chromatin-bound proteins from 50 liters of HeLa cells through six chromatographic steps and assayed the proteins using 32P-labeled synthetic HJs (Fig. 1A) (4). We further analyzed the final fractions using 32P-labeled recombination intermediates (α-structures) containing a single HJ (Fig. 1B). Treatment of these substrates with a branch migration enzyme (Escherichia coli RuvAB) produces gapped circular and 32P-labeled linear duplex products (Fig. 1C, lane b). Alternatively, junction resolution by a resolvase (RuvC) gives rise to circular and 32P-labeled nicked 32P-labeled gapped linear products (resolution in orientation 1/3), or 32P-labeled linear dimers (orientation 2/4), that are a unique signature of HJ resolution (Fig. 1C, lane c). Usually, resolution to linear dimers dominates in these assays (5). Product formation, as assayed by gel electrophoresis (Fig. 1C), was also quantified by PhosphorImaging the 32P-end labels (Fig. 1D). After extensive purification, the HeLa fraction catalyzed HJ resolution and ATP-dependent branch migration (Fig. 1C, lane d). Western blot analyses showed that the fraction was devoid of RAD51, BRCA1, BRCA2, as well as junction-processing activities such as Mus81, FEN-1, BLM, or WRN. Using a monoclonal antibody (2H11), however, we found that the recombination/repair protein RAD51C was present in all fractions containing substantial amounts of branch migration and resolution activity (Fig. 2A). Interestingly, RAD51C in the resolvasome fractions exhibited a slower mobility than RAD51C in unfractionated whole-cell extracts (Fig. 2B, lanes a and b). This slower mobility species, which was enriched during purification, was associated with chromatin, as indicated by analyses of cytosolic, nuclear (0.5 M NaCl extracted), and chromatin pellet fractions (Fig. 2B, lanes c to e). In contrast to the chromatin pellet, the cytoplasmic and nuclear extracts contained mostly unmodified RAD51C. At present, the precise nature of the modification is unknown.

Fig. 1.

Purification of the human HJ resolvasome. (A) Scheme for the purification of HJ branch migration and resolution activities from nuclear extracts prepared from 50 liters of HeLa cells. (B) Schematic diagram of the in vitro assay for branch migration and resolution activities. Recombination intermediates (α-structures) were made by means of RecA-mediated strand exchange between 3′-32P-end–labeled linear duplex pAKE-7Z DNA and gapped circular pDEA-7Z plasmid DNA (gDNA). pAKE-7Z contains a 1.7-kb region of heterology at the distal end of the linear duplex (shaded) that blocks strand exchange and stabilizes the recombination intermediates. Branch migration drives the HJ back to the gap, dissociating the α-structures into 32P-labeled linear duplex and unlabeled gDNA products. HJ resolution occurs in one of two possible orientations: Cleavage in strands 1/3 produces two labeled products (32P-labeled nicked circular and 32P-labeled gapped linear DNA), whereas resolution in strands 2/4 produces 32P-labeled linear dimers. 32P-end labels are indicated with asterisks. (C) Branch migration/resolution activities of the human resolvasome. Reactions contained RuvAB, RuvC, or a sample (1 μl) of fraction 15 from the final SP-Sepharose column (SP-15) with 32P-labeled α-structure. Reaction products were visualized by agarose gel electrophoresis and autoradiography. (D) Quantification. Product formation was quantified by PhosphorImaging. As shown in (C), α-structures represent 75 to 80% of total labeled DNA. The remaining 20 to 25% of 32P-labeled DNA represents unreacted linear duplex molecules (subtracted as background). Branch migration was measured by the increase in 32P-labeled linear duplex DNA, whereas resolution produced linear dimer and nicked circular products. The number of 32P-labels in each product was taken into account.

Fig. 2.

The recombination/repair protein RAD51C is present in the highly purified resolvasome fractions from HeLa cells. (A) Fractions from the final SP-Sepharose column (Fig. 1A), containing branch migration and resolution activities as assayed with α-structures (lower panel), were analyzed by Western blotting with RAD51C monoclonal antibody 2H11 (upper panel). (B) Modification of RAD51C. Western blot analysis of RAD51C in SP-Sepharose fraction 14 (SP-14, lane b) revealed that it migrates more slowly than RAD51C from whole-cell extracts (lane a), as determined by SDS-PAGE. The slow-migrating form of RAD51C is chromatin-associated, as indicated by Western blotting cytoplasmic, nuclear, and chromatin-bound fractions from HeLa cells (lanes c to e).

Next, we analyzed whether RAD51C was required for resolvasome activity by depleting the SP-Sepharose peak fraction (SP-15) using a polyclonal antibody to RAD51C (SWE31). Immunodepletion removed most of the RAD51C protein and reduced HJ resolution activity (Fig. 3A, compare lanes b and c). Branch migration was also reduced to background levels (see quantification below lanes b and c). In contrast, depletion with preimmune sera or nonrelated antibodies had little effect. Thus, RAD51C appears to be a component of the resolvasome and is required both for branch migration and HJ resolution.

Fig. 3.

RAD51C is involved in HJ branch migration and resolution. (A) Branch migration/resolution assays were carried out with the peak SP-Sepharose resolvasome fraction before (SP-15) and after immunodepletion (SP-15 depl.) with polyclonal antibody to RAD51C (SWE31) conjugated to protein A–Sepharose beads. RAD51C-depleted SP-15 fractions were then complemented by the addition of various purified proteins or protein complexes (all 100 nM). BCDX2, RAD51B-RAD51C-RAD51D-XRCC2 complex; BC, RAD51B-RAD51C complex; and CX3, RAD51C-XRCC3 complex. All RAD51C-containing complexes complemented the depleted fraction for resolvase activity, whereas BCDX2 was the only complex capable of restoring branch migration activity. Product formation was quantified by PhosphorImaging (as indicated below each lane). BM, branch migration. Res, resolution. (B) Comparison of the ability of RAD51B-RAD51C (BC) or RAD51D-XRCC2 (DX2) to complement RAD51C-depleted SP-15. (C) Complementation assays were carried out with RAD51C-depleted SP-15 supplemented with 70 nM RAD51B-RAD51C purified from insect cells (lane d) or E. coli (lane e), or with the indicated mutant RAD51C proteins (lanes f to h).

RAD51C is a key component of at least two complexes in human cells: One contains RAD51B-RAD51C-RAD51D-XRCC2 (designated BCDX2) (6, 7), and the other contains RAD51C with XRCC3 (CX3) (810). Subcomplexes of BCDX2, such as RAD51B-RAD51C (BC) and RAD51D-XRCC2 (DX2), also exist (1114). Each purified protein complex was therefore tested for complementation of the RAD51C-depleted fraction and restoration of branch migration/resolution activities.

Recombinant BCDX2, purified to homogeneity after overexpression in a baculovirus system, restored full HJ resolvase activity to the RAD51C-depleted extract (Fig. 3A, lane e). In these reconstituted reactions, resolution occurred in both orientations, as seen by the formation of 32P-labeled linear dimer and 32P-labeled nicked circular DNA products. Moreover, quantification showed that BCDX2 also partially restored branch migration activity, as indicated by an increase in the ratio of linear duplex DNA products relative to α-structures (see lane e, quantification).

The resolution defect in the RAD51C-depleted fraction was also restored by addition of purified recombinant BC (Fig. 3A, lane f) or CX3 (lane g). Other control proteins, such as purified RAD51 (lane d) or RAD52 (15), failed to complement. When the complementation activity of the two subcomplexes of BCDX2 was compared, we found that BC (Fig. 3B, lanes d to f), but not DX2 (lanes g to i), restored the depleted resolvase activity. Complementation by BC was linearly related to the concentration of purified protein. These results show that resolvase activity can be restored by addition of any purified complex containing RAD51C, thereby implicating RAD51C in HJ-processing reactions.

The proteins used in the experiments of Fig. 3, A and B, were prepared from baculovirus-infected insect cells. Recombinant BC purified from E. coli also complemented a RAD51C-depleted extract (Fig. 3C, lane e). Because complementation was found to be ATP-stimulated (15), consistent with previous HJ resolution studies (3), we next prepared RAD51CK131A, carrying an amino acid substitution in the ATP-binding domain. This protein, which fails to complement the MMC sensitivity of RAD51C-defective cells (16), exhibited reduced complementation activity (Fig. 3C, lane f). Similarly, deletion of the C-terminal region of RAD51C, as observed with RAD51C1–126, eliminated complementation activity (Fig. 3C, lane g), whereas the N-terminally deleted protein RAD51C129–376 was capable of partial complementation (Fig. 3C, lane h).

We next sought to determine whether extracts prepared from cell lines defective in the RAD51 paralogs exhibited normal levels of HJ resolvase activity. Because there are no mutant human cell lines, we used chinese hamster ovary cell lines defective in XRCC2 (irs1), XRCC3 (irs1SF), or RAD51C (irs3) (17, 18). Presently, there are no mammalian cell lines with defects in RAD51B or RAD51D. Extracts were prepared from cultures of each line, together with the normal cell line V79, and synthetic HJs were used to detect resolvase activity through four chromatographic steps (Fig. 4A). Although the amounts of protein for each extract at each step of the purification were similar, fractions from irs1 and the V79 control consistently exhibited greater levels of resolvase activity, measured by cleavage of 32P-labeled junction X26 into nicked duplex products, than fractions from irs3 or irs1SF. Resolution assays carried out with the heparin fractions are shown in Fig. 4B (compare fractions 9 to 12) and those with the peak fraction from the SP-Sepharose column are shown in Fig. 5A (upper panel).

Fig. 4.

Analysis of HJ resolvase activity in extracts from DNA repair–defective hamster cells. (A) Scheme for the partial purification of resolvase activity from hamster extracts. (B) Whole-cell extracts were prepared from 10 liters of the hamster cell lines irs3 (RAD51C), irs1 (XRCC2), irs1SF (XRCC3), and V79 (wild-type), as indicated by the scheme shown in (A) but only as far as the heparin chromatography step. Samples (1 μl) from each fraction were analyzed for resolution activity with the 32P-labeled synthetic HJ X26. 32P-labeled nicked duplex resolution products (lanes 9 to 11) were visualized by autoradiography after neutral PAGE. The protein concentration of each fraction was comparable between cell lines.

Fig. 5.

Defective HJ resolution in irs3 (RAD51C) and irs1SF (XRCC3) cells. (A) Resolution of the junction X26 by fractions from irs1SF (XRCC3), irs3 (RAD51C), irs1 (XRCC2), and control V79 cell line. The relative levels of HJ resolvase activity in peak fractions eluting from the final SP-Sepharose chromatographic step, indicated in the scheme of Fig. 4A, were assayed with equivalent amounts of protein (diluted 1/8, 1/4, and 1/2) up to 1 μl of neat fraction (lanes f, j, n, r). RuvC (C, 100 nM) was used as a control. Resolvase activity was determined by neutral PAGE followed by autoradiography (upper panel) or PhosphorImaging (lower panel). (B) The sites of cleavage by the four SP-Sepharose peak fractions were mapped by denaturing PAGE. Two alternative predominant cleavage sites (blue arrows) were observed with V79 (wild-type) and irs1 (XRCC2), characteristic of resolvase incision, whereas only weak cleavage products were observed with fractions from irs3 (RAD51C) and irs1SF (XRCC3). A RuvC cleavage control (C) is indicated (black arrows). With junction X26, cleavage can occur anywhere within the 26 – base pair region of homology. The junction was 5′-32P-end–labeled in strand 2, as indicated by asterisks. (C) Resolution of the synthetic HJ X0 by SP-Sepharose fractions from the four cell lines. The junction was 5′-32P-end–labeled in strand 1, as indicated by the asterisks. Cleavage sites were mapped by denaturing PAGE. A strong incision site was found 1 nucleotide to the 3′ side of the nonmigratable crossover point with fractions from the V79 control and irs1 (XRCC2) only (red arrow).

To ensure that the relative levels of cleavage were not affected by trace contamination with Mus81/Eme1, we analyzed the cleavage profiles produced by each SP-Sepharose fraction using denaturing polyacrylamide gel electrophoresis (PAGE), which gives a characteristic and distinctive pattern for HJ resolvases (Fig. 5B). In contrast to irs1 and V79, which produced the expected cleavage profile, little junction-specific nicking was observed with the related fractions from irs1SF or irs3. Similar results were obtained with the immobile HJ X0, which is essentially refractory to cleavage by human Mus81/Eme1 (19), as only V79 and irs1 produced a major incision site at the sequence 5′-AGGA-3′ located at the point of strand exchange (Fig. 5C). Because irs3 carries a point mutation that affects splicing (18), the residual resolvase activity may be due to a small amount of full-length protein. The nature of the mutation in irs1SF is presently unknown. We conclude that HJ resolution activity is impaired in the irs1SF (XRCC3) and irs3 (RAD51C) cell lines.

Taken together, our studies lead us to suggest that RAD51C and XRCC3 are involved in HJ resolution. Additionally, the in vitro studies implicate other RAD51 paralog members in branch migration processes. At present, however, we have not observed branch migration and/or HJ resolution using purified recombinant protein complexes alone and suggest that additional, as yet unidentified, factors or posttranslational modifications (as observed with endogenous RAD51C) may be required.

Our proposal that RAD51C plays a late role in recombination is somewhat surprising, given earlier studies suggesting that the RAD51 paralogs act as mediators that facilitate the loading of RAD51 onto DNA (6, 12, 20) and that cell lines defective in the RAD51 paralogs are partially defective in the DNA damage–induced formation of RAD51 nuclear foci (21). However, data are now accumulating to support the proposal that the RAD51 paralogs may also play a late role. For example, it has been shown that (i) RAD51B binds preferentially to HJs in vitro (22); (ii) RAD51D-XRCC2 complex stimulates the disruption of HJs by the Bloom's syndrome protein BLM (23); (iii) RAD51C-defective cells exhibit reduced frequencies of spontaneous and DNA damage–induced sister-chromatid exchanges (18, 24), together with increased frequencies of isochromatid breaks that may arise by incomplete resolution of HJ intermediates at stalled replication forks (18); (iv) XRCC3 mutants exhibit abnormally long gene-conversion tracts (25); and (v) XRCC3-defective Arabidopsis appears to initiate meiotic recombination normally but fails to produce viable gametes due to chromosome fragmentation (26). The discovery that RAD51C is involved in branch migration and HJ resolution provides an important step forward that will lead to mechanistic studies of recombination intermediate processing within the eukaryotic nucleus.

Supporting Online Material

Materials and Methods


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article