RAD54 controls access to the invading 3'-OH end after RAD51-mediated DNA strand invasion in homologous recombination in Saccharomyces cerevisiae

Abstract

Rad51 is a key protein in homologous recombination performing homology search and DNA strand invasion. After DNA strand exchange Rad51 protein is stuck on the double-stranded heteroduplex DNA product of DNA strand invasion. This is a problem, because DNA polymerase requires access to the invading 3'-OH end to initiate DNA synthesis. Here we show that, the Saccharomyces cerevisiae dsDNA motor protein Rad54 solves this problem by dissociating yeast Rad51 protein bound to the heteroduplex DNA after DNA strand invasion. The reaction required species-specific interaction between both proteins and the ATPase activity of Rad54 protein. This mechanism rationalizes the in vivo requirement of Rad54 protein for the turnover of Rad51 foci and explains the observed dependence of the transition from homologous pairing to DNA synthesis on Rad54 protein in vegetative and meiotic yeast cells.

Figure 1.

Rad54 is required for D-loop extension. (A) Schematic representation of the linear D-loop assay. The AatII-95-mer is homologous to the terminal sequence of the AatII-linearized pUC19 DNA (2686 bp). Reaction products are identified by 5′-end-labeling the 95-mer. (B) D-loop assay. End-labeled AatII-95-mer was incubated with Rad51 (1 monomer:3 nt) and then with AatII-linearized pUC19 dsDNA at molecular ratios of oligonucleotide to dsDNA of 1:1 (lanes 2–9) or 3:1 (lanes 10–20). Reactions contained 72 or 144 nM Rad54 (lanes 6–9, 15–18, 20) and Rad51 (lanes 2–18). End-labeled linearized pUC19 served as size marker (lane 1). Products were crosslinked before electrophoresis. (C) Quantitation of results in (B). (D) Schematic representation of the linear D-loop extension assay. (E) D-loop extension assay. Reaction schemes as in (A), except that they also contained 24 or 72 nM polymerase (lanes 2–19). Stable extension products are D-loops of sufficient length to be stable under electrophoresis conditions. Products were not crosslinked before electrophoresis in the D-loop extension assay. (F) Quantitation of results in (E). Shown are means from three determinations; error bars represent 1 SD.

Figure 2.

Rad54 is required for D-loop extension. D-loop extension assays. The PstI-95-mer is homologous to the terminal sequence of the PstI-linearized pUC19 DNA (2686 bp). Reaction products are identified either by end-labeling the 95-mer (A, B, F) or by incorporation of α-³²P-dGTP (C–E, G). (A) D-loop extension assay with end-labeled PstI-95-mer. Rad51 nucleoprotein filaments were incubated either in the presence of Rad54 (72 nM, lanes 9–11), or absence of Rad54 (lanes 6–8), or absence of DNA polymerase I (Klenow fragment 24 nM, lanes 12–14). Protein-free 95-mer was also incubated either in the presence of Rad54 (72 nM, lanes 3–5), or absence of Rad54 (lanes 15–17). Lane 1 shows end-labeled PstI-linearized pUC19 as a size marker, and lane 2 the end-labeled 95-mer. (B) Quantification of the results for D-loop extension in (A). (C) D-loop extension assay with α-³²P-dGTP and unlabeled PstI-95-mer. Reactions were as in (A), except lane 2 contains unlabeled PstI-95-mer and lanes 18–20 show PstI-linearized pUC19 with DNA polymerase I (Klenow fragment, 24 nM). (D) Quantification of the stable extension product from (C). Stable extension products are D-loops of sufficient length to be stable under electrophoresis conditions. (E) Quantification of the unstable extension product from (C). Unstable extension products are extended 95-mers with insufficient length to result in stable D-loops under electrophoresis conditions. For (B)–(E) shown are means from three determinations; error bars represent 1 SD. (F) Analysis of extension products on denaturing gel from reactions with end-labeled 95-mer and (G) with α-³²P-dGTP. The signals labeled by asterisk are due to a combination of 3′–5′ exonuclease (proofreading) and/or polymerase activity of Klenow polymerase on the 95-mer or the linear dsDNA.

Figure 3.

D-loop extension depends on sequence homology, species-specific protein interaction, and Rad54 ATPase activity. (A) D-loop extension depends on sequence homology. Linear D-loop extension assays were performed as in Figure 1 either with the AatII-95-mer or a heterologous 95-mer (olWDH640). (B) Rad54 ATPase activity is required for D-loop extension. Linear D-loop extension assays were performed as in Figure 1 in the presence of wild-type Rad54 (72 nM) or Rad54–K341R mutant protein (72 nM). (C) D-loop extension requires species-specific interaction between Rad51 and Rad54. Linear D-loop extension assays with either S. cerevisiae Rad54 (yRad54, 72 nM) or human Rad54 (hRad54, 72 nM) were performed as in Figure 1. Shown are means from three determinations; error bars represent 1 SD.

Figure 4.

Rad51–K191R requires higher Rad54 concentrations for efficient D-loop extension. (A) Rad51 titration in the D-loop extension assay with PstI-95-mer: 0.17 µM (lanes 2–4), 0.34 µM (lanes 5–7), 0.5 µM (lanes 8–10), 0.67 µM (lanes 11–13), 1 µM (lanes 14–16), and 2 µM (lanes 18–20). Reactions contain 72 nM Rad54, as determined to be the optimum in Supplementary Figure 4. Lane 1 shows end-labeled PstI-linearized pUC19 as a size marker. (B) Rad51–K191R titration, otherwise as in (A). The signals labeled by asterisk in A and B are generated by the proofreading activity (3′–5′ exonuclease) of Klenow polymerase as determined in reconstruction experiments and verified using a proofreading-deficient version of Klenow polymerase (data not shown). This signal disappears when Rad51 fully occupies the 95-mer, further validating that Klenow polymerase has no access to the 3′-OH end with Rad51 bound to it. (C) Quantitation of the Rad51 and Rad51–K191R protein titration results in (a, b; 20 min time points). A higher stoichiometry (1/2 nt) is optimal for the Rad51–K191R protein compared to the 1/4 nt stoichiometry for the wild-type Rad51 protein. This was expected from previous results showing a DNA-binding defect for the Rad51–K191R protein, requiring higher protein to DNA ratios to assemble saturated protein filaments (7). (D) Titration of Rad54 in D-loop extension assay with Rad51 at optimal 1/4 stoichiometry. (E) Titration of Rad54 in D-loop extension assay with Rad51–K191R at optimal 1/2 stoichiometry. (F) Quantitation of results in (D, E; 20 min time points). The results were normalized for the amount of linear D-loops captured by psoralen crosslinking under each assay condition. D-loop extension with wild-type Rad51 reaches an optimum at 36 nM Rad54, whereas the optimum with Rad51–K191R is reached at 54 nM. Shown are the means from three determinations; error bars represent 1 SD.

Figure 5.

Time of Rad54 addition determines D-loop extension. (A) Schematic representation of the timed addition experiment. (B) D-loop extension assays were performed with the PstI-95-mer and PstI-linearized pUC19 as before with the exception that Rad54 was added as the last component at 0 min (lanes 2–5), 10 min (lanes 6–10), 20 min (lanes 11–15), or 30 min (lanes 16–20). After Rad54 addition, time courses (5, 10 and 20 min) were performed for every protocol. Lane 1 contains end-labeled pUC19 as size marker. (C) Quantification of the results in (B), shown are means from three determinations; error bars represent 1 SD.

Figure 6.

Model for Rad54 functions in homologous recombination. Rad54 associates with the Rad51–ssDNA filament (40), stabilizing the presynaptic filament in an ATPase-independent fashion (21,22). Rad54 stimulates DNA strand invasion (17), but the mechanisms remain to be determined and may involve the conversion of unstable (paranemic?) to stable (plectonemic?) joints (30,41). Rad54 dissociates Rad51 from dsDNA (14) at the terminus of the Rad51–dsDNA filament (13), allowing access of DNA polymerase to the invading 3′-OH end (this work). The Rad54 motor may act as hexameric (shown) or double-hexameric ring; a double ring could explain the observed reversibility of the translocation direction (9). Rad54 has also been shown to catalyze branch migration of DNA junctions and D-loop dissociation in vitro, which may depend on the orientation of Rad54 approaching a junction (16,18,42). Rad54's ability to slide nucleosomes in vitro (43–45) might help DNA strand invasion and stable joint formation in chromatinized regions in vivo (41,44,45). Similar to Rdh54/Tid1 dissociating dead-end Dmc1–dsDNA complexes (46), Rad54 possibly counteracts stable association of Rad51 with undamaged chromosomal DNA.