Resolution of single and double Holliday junction recombination intermediates by GEN1

Significance

Holliday junction resolvases are required for the resolution of recombination intermediates to ensure proper chromosome segregation at mitosis. In human cells, the GEN1 protein, a member of the XPG/Rad2 family of structure-selective endonucleases, is specialized for the cleavage of Holliday junction recombination intermediates. Here, we show that GEN1 cuts DNA substrates containing single or double Holliday junctions to generate cross-over or noncross-over recombination products. In contrast to the bacterial Holliday junction resolvase, RuvC, which prefers to cut junctions at the sequence 5′-^A/_TTT^G/_C - 3′, GEN1 shows only weak sequence specificity.

Abstract

Genetic recombination provides an important mechanism for the repair of DNA double-strand breaks. Homologous pairing and strand exchange lead to the formation of DNA intermediates, in which sister chromatids or homologous chromosomes are covalently linked by four-way Holliday junctions (HJs). Depending on the type of recombination reaction that takes place, intermediates may have single or double HJs, and their resolution is essential for proper chromosome segregation. In mitotic cells, double HJs are primarily dissolved by the BLM helicase-TopoisomeraseIIIα-RMI1-RMI2 (BTR) complex, whereas single HJs (and double HJs that have escaped the attention of BTR) are resolved by structure-selective endonucleases known as HJ resolvases. These enzymes are ubiquitous in nature, because they are present in bacteriophage, bacteria, archaea, and simple and complex eukaryotes. The human HJ resolvase GEN1 is a member of the XPG/Rad2 family of 5′-flap endonucleases. Biochemical studies of GEN1 revealed that it cleaves synthetic DNA substrates containing a single HJ by a mechanism similar to that shown by the prototypic HJ resolvase, Escherichia coli RuvC protein, but it is unclear whether these substrates fully recapitulate the properties of recombination intermediates that arise within a physiological context. Here, we show that GEN1 efficiently cleaves both single and double HJs contained within large recombination intermediates. Moreover, we find that GEN1 exhibits a weak sequence preference for incision between two G residues that reside in a T-rich region of DNA. These results contrast with those obtained with RuvC, which exhibits a strict requirement for the consensus sequence 5′-^A/_TTT^G/_C-3′.

Homologous recombination (HR) plays an important role in the repair of DNA double-strand breaks resulting from DNA damaging agents, such as ionizing radiation, or the progression of a replication fork through single-strand breaks in DNA (1, 2). Strand break repair by HR generally involves interactions between sister chromatids, which can become covalently linked by four-way junctions known as Holliday junctions (HJs) (3). Although originally proposed to explain meiotic gene conversion in fungal systems, the HJ provides a central tenet that is integral to many mechanisms of recombination in both meiotic and mitotic cells. Two such pathways of DNA strand break repair that occur in mitotic cells are shown in Fig. 1. In classical double-strand break repair (DSBR), a resected DNA end pairs with homologous DNA to generate a D-loop structure that provides a 3′ end for DNA synthesis (gap filling). Capture of the second end followed by more DNA synthesis lead to the formation of DNA intermediates, in which the interacting DNAs are joined by two HJs (Fig. 1, Left). In a related but distinct scenario, a single DNA end, which might arise at a stalled/broken replication fork, invades homologous DNA to again form a D loop. In this case, however, there is no second end available, and DNA synthesis at the D loop leads to the formation of a single HJ (Fig. 1, Right).

Fig. 1.

Recombination pathways for the generation of single-HJ and dHJ intermediates. (Left) DSBR involves end resection, strand invasion, DNA synthesis (gap filling), and second-end capture to generate an intermediate in which the sister chromatids are covalently joined by a dHJ. Resolution of the intermediate can generate COs (cleavage in orientations a and d or b and c) or NCOs (cleavage in orientations a and c or b and d). (Right) Single-end invasion involves one DNA end, which could be generated by cleavage of a stalled replication fork. End resection and strand invasion lead to the formation of a D-loop structure similar to that produced in DSBR. However, in this instance, there is no second end for capture, and instead, the invading 3′ end serves to prime DNA synthesis and form a single HJ. Resolution can give rise to a CO (cleavage in orientation a) or NCO (cleavage in orientation b). DSB, double-strand break.

In both DSBR and single-end invasion, the resulting HJs need to be processed for efficient chromosome segregation at mitosis, thereby promoting genome stability (4). There are three pathways for HJ processing (5). In humans, the primary pathway involves the BLM helicase-TopoisomeraseIIIα-RMI1-RMI2 (BTR) complex, which promotes the dissolution of double Holliday junctions (dHJs) by catalyzing the convergent migration of each junction followed by topoisomerase-mediated processing of the resulting hemicatenane (6–8). This pathway favors the formation of noncross-over products (NCOs). In yeast, similar reactions are promoted by the Sgs1-Top3-Rmi1 (STR) complex (9, 10). Persistent dHJs that escape dissolution by BTR/STR are acted on at a later stage of the cell cycle by two distinct and temporally regulated nucleolytic cleavage pathways that promote HJ resolution. However, in contrast to the BTR/STR system, these nucleases resolve both single HJs and dHJs. Resolution gives rise to both cross-over products (COs) and NCOs (11), and the enzymes that catalyze these reactions are broadly referred to as HJ resolvases (12).

The first cellular (prototypic) HJ resolvase identified was the Escherichia coli RuvC protein, with which eukaryotic resolvases are compared (13, 14). RuvC is a homodimeric protein containing two 19-kDa subunits that interact specifically with HJs. Binding leads to the formation of an open HJ–RuvC structure, and symmetrically related nicks are introduced into two strands that are diametrically opposed across the junction (15–21). Consistent with this mechanism of cleavage, the crystal structure of E. coli RuvC shows that the two monomers are related by a dyad axis, in which the two DNA binding clefts are separated by ∼30 Å (Fig. 2A) (22, 23). The 3D structure of Thermus thermophilus RuvC bound to an HJ has also been solved and confirms the presence of a twofold symmetric unfolded junction (Fig. 2B) (24, 25). Although substrate binding occurs in a sequence-independent manner, cleavage occurs preferentially at the sequence 5′-^A/_TTT^G/_C-3′ and generates a pair of nicked duplexes (26). Because of the symmetry of the cleavage reaction, the resulting nicks can be joined by DNA ligase.

Fig. 2.

3D crystal structures of the RuvC and GEN1 HJ resolvases. Each structure depicts two monomeric subunits: one colored blue, and the other colored magenta. Active site residues are illustrated in ball and stick format and colored in lime and orange. Metal ions bound in the active site are shown as yellow spheres. The Protein Data Bank ID codes used to generate these structures are as follows: (A) E. coli RuvC, 1HJR; (B) T. thermophilus RuvC–HJ complex, 4LDO; (C) Homo sapiens GEN1 bound to duplex arm, 5T9J; and (D) C. thermophilum GEN1 bound to the nicked duplex products of HJ resolution, 5CO8.

The search for eukaryotic equivalents to the RuvC HJ resolvase was hindered by the lack of sequence and structural similarities but culminated in the identification of Yen1 and GEN1 from yeast and humans, respectively (27). GEN1/Yen1 belong to the XPG/Rad2 family of 5′-flap endonucleases (28, 29), whereas RuvC is related to the retroviral integrase superfamily (25). Like RuvC, GEN1 promotes HJ resolution by a coordinated nick and counternick mechanism, but it is currently unknown whether the eukaryotic HJ resolvases exhibit a defined sequence specificity (30, 31). GEN1 also promotes symmetrical incisions across the junction, generating nicked duplex products that can be ligated. Recently, two crystal structures of active site-containing truncated forms of GEN1 bound to DNA were reported (Fig. 2 C and D). In one study, catalytically inactive human GEN1^2–505 was bound nonproductively to the duplex arm of an HJ (32), whereas in the other, Chaetomium thermophilum GEN1^1–487 was captured on the nicked duplex DNA that is produced by HJ resolution (33). Although the well-conserved core is similar to that of other XPG/Rad2 nucleases, GEN1 contains a chromodomain that provides an additional DNA binding site necessary for efficient HJ cleavage (32). Interestingly, the GEN1 nicked duplex structure revealed that the XPG/Rad2 fold found in FEN1 and other members of this nuclease family are modified for its role in HJ cleavage, with the dimer interface juxtaposing the two nicked duplex DNA products (33).

In contrast to the prokaryotic HJ resolvases, the activities of Yen1 and GEN1 are regulated throughout the cell cycle. In yeast, Yen1 exhibits low activity in S phase, because cyclin-dependent kinase (Cdk) promotes its phosphorylation, leading to nuclear exclusion and inhibition of catalytic activity by reducing its affinity for DNA (34, 35). Then, later in the cell cycle, Yen1 is dephosphorylated by Cdc14; dephosphorylation promotes nuclear localization and DNA binding, leading to enzymatic activation (36–38). GEN1 is also regulated, but in this case by a mechanism that seems to be independent of phosphorylation, because its activity is controlled primarily by nuclear exclusion (39). Thus, GEN1 only gains access to the DNA after breakdown of the nuclear envelope.

A second mechanism of HJ resolution involves the 3′-flap endonuclease MUS81-EME1 (Mus81-Mms4 in budding yeast) (40–45). HJs are a poor substrate for MUS81-EME1 and Mus81-Mms4, because both prefer to cleave nicked HJs (46–50). Because cleavage requires structural transitions at the junction point, it is possible that the nick relaxes the topological constraints that inhibit efficient cleavage of intact HJs. The activity of Mus81-Mms4, which is low in S phase, is activated by Cdk/Cdc5-mediated phosphorylation events that occur in G2/M phase of the cell cycle (34, 35, 51, 52). The situation in human cells is similar, except that phosphorylation stimulates MUS81-EME1 to interact with the conserved GIY-YIG endonuclease SLX1-SLX4 to form an SLX1-SLX4-MUS81-EME1 (SLX-MUS) complex in prometaphase (49, 53, 54). The SLX-MUS complex promotes efficient HJ resolution, with SLX1-SLX4 introducing the initial nick at the junction and MUS81-EME1 catalyzing the second incision in the opposing strand (49). The two cuts occur by a concerted nick and counternick mechanism.

Previous biochemical studies of the actions of GEN1 used synthetic HJ substrates formed by annealing four partially complementary oligonucleotides or plasmids that contain inverted repeat sequences that extrude to form cruciform structures, the base of which mimics a single HJ. Although these model substrates have provided many important insights into the functions of these enzymes (27, 30, 31), they may not fully represent all aspects of recombination intermediates that arise in vivo. We therefore generated two physiological HJ substrates for additional analysis of the resolution reactions mediated by GEN1: (i) plasmid-based recombination intermediates (α-structures) formed by RecA protein-mediated strand exchange (26, 55, 56) and (ii) plasmid-based substrates containing dHJs (57, 58).

Resolution of Recombination Intermediates Containing Single HJs by GEN1

RecA-mediated strand exchange between gapped circular pDEA-7Z DNA and 3′ ³²P-labeled linear duplex pDEA2 DNA produces recombination intermediates (α-structures) containing a single HJ (26). The strand exchange reaction proceeds through 1,536 bp of homology until it reaches a heterologous block that is 643 bp in length (Fig. 3A). The resulting products are stable and were previously used for detailed analysis of HJ resolution by RuvC (13, 56). They, therefore, provide good model substrates for the analysis of GEN1 activity.

Fig. 3.

Resolution of single-HJ recombination intermediates by GEN1. (A) Schematic showing the α-structure recombination intermediate, in which two DNAs are linked by an HJ, and the products of nucleolytic resolution. The α-structures are made by RecA-mediated homologous pairing and strand exchange between gapped circular and 3′ ³²P-labeled linear duplex DNA. Strand exchange proceeds through 1,536 bp of homology up to a heterologous block (green boxes; 643 bp in length). HJ resolution in orientation a − a produces ³²P-labeled linear dimer DNA, whereas cleavage in orientation b − b gives rise to ³²P-labeled gapped linear DNA and ³²P-labeled nicked circular DNA. *3′ ³²P-labeled termini. (B and C) Resolution of α-structure recombination intermediates by GEN1. Reactions contained the indicated concentrations of GEN1 or E. coli RuvC HJ resolvase (10 nM) as a control. Products were analyzed by agarose gel electrophoresis followed by autoradiography.

When the α-structures were treated with full-length human GEN1 protein, we observed their efficient resolution into ³²P-labeled gapped linear and ³²P-labeled nicked circular products consistent with resolution in orientation b − b and ³²P-labeled linear dimer DNA consistent with resolution in orientation a − a (Fig. 3 B, lanes d–f and C, lanes d–h). Resolution occurred in either of two possible orientations at a ratio of ∼1:1 (note that the linear dimer product has two ³²P labels, and therefore, this product band is twice as intense as the nicked circular and gapped linear DNA products). The products made by GEN1 were indistinguishable from those made by E. coli RuvC (Fig. 3 B, lane b and C, lane b) (26). We did not observe cleavage by nuclease-dead GEN1^E134A,E136A protein as observed previously with oligonucleotide-based HJs (31). These results confirm that GEN1 acts as a prototypic HJ resolvase on single-HJ recombination intermediates made by RecA protein.

Resolution of dHJs by GEN1

To determine whether GEN1 acts on dHJs, we prepared a plasmid substrate that contains two single HJs separated by 165 bp (Fig. 4A) (57). Previous studies have shown that nucleases, such as T7 endonuclease I, resolve this dHJ substrate to produce a large circle of 881 bp, representative of a CO, or two small circles of 465 and 416 bp that represent the NCOs (57) as shown in Fig. 4B. The large circle is often seen as two bands, and these forms are thought to represent distinct topological forms because restriction enzyme cleavage converts them to a single band (57). When the dHJ substrate was treated with GEN1, we observed the formation of the expected COs and NCOs (Fig. 4 C, lanes c–g and D, lanes c–g).

Fig. 4.

Resolution of dHJs by GEN1. (A) Schematic diagram of the dHJ substrate and the products of resolution. Cleavage of both HJs in the same orientation generates NCOs (circles of 465 and 416 bp), whereas cleavage in the opposite orientation gives rise to a CO (a single 881-bp circle). (B) Resolution of dHJs by T7 endonuclease I. (C and D) Resolution of dHJs by GEN1. (E) Analysis of the actions of RuvC on the dHJ substrate. Products were separated by agarose gel electrophoresis and stained with SYBR gold.

In contrast to GEN1, RuvC protein did not cleave the dHJ substrate (Fig. 4E). This lack of cleavage may result from torsional constraints that prohibit the conversion of the HJs into the open configuration required for HJ cleavage (16, 17). Alternatively, it may be because of RuvC’s stringent requirement for the sequence 5′-^A/_TTT^G/_C-3′ for efficient HJ cleavage (26). Consistent with the latter possibility, DNA sequence analyses revealed that there were three sequences present within the 153-bp region of homology that are identical to RuvC’s preferred sequence. However, none of these consensus sequences reside at the boundaries of homology/heterology where the junctions are located the topologically constrained dHJ substrate.

Sequence Preference for HJ Cleavage by GEN1

The ability of GEN1 to cleave the dHJ substrate, whereas RuvC could not, indicated to us that HJ cleavage by GEN1 was unlikely to be limited by strict sequence constraints. Instead, we reasoned that GEN1 might act more like the bacteriophage resolvases T4 endonuclease VII or T7 endonuclease I or the archaeal resolvase Hjc, all of which exhibit little or no sequence specificity (59, 60).

The sequence specificity of GEN1 was investigated using the α-structure DNA substrate. A useful feature of this natural recombination intermediate is that, by gentle heating at 55 °C, the HJs can be induced to branch migrate and become distributed over a large DNA sequence. The migrated junctions, therefore, provide a greater choice of potential cleavage sites than that offered by oligonucleotide-based HJs. In previous studies, these α-structures were used to determine the sequence preference used by RuvC (26).

Using RuvC as a control, we carried out large-scale resolution reactions with ³²P-labeled α-structures and compared the products with those produced by GEN1, as analyzed by denaturing gel electrophoresis and visualized by autoradiography. In the absence of heat-induced branch migration, ³²P-labeled fragments of ∼1.5 kb in length were observed with both RuvC and GEN1 (Fig. 5A, lanes b and f), consistent with resolution occurring at sequences located near the heterologous block. However, when the α-structures were incubated at 55 °C for 1 h to induce limited spontaneous branch migration before treatment with either RuvC or GEN1, we observed a series of distinct ³²P-labeled DNA fragments by denaturing PAGE (Fig. 5A, lanes d and h). The distinct cleavage pattern generated by RuvC is caused by HJ resolution at sites that correspond to its consensus sequence (26). In contrast, GEN1 generated a significantly greater number of distinct ³²P-labeled DNA products. There are two possibilities to explain this result: (i) GEN1 has a weak sequence preference, which would lead to cleavage at a greater number of sites than observed with RuvC, or (ii) GEN1 has no sequence specificity and simply cleaves HJs when branch migration has temporarily paused because of sequence/energy constraints. To test the latter possibility, we included T4 endonuclease VII in our analyses, because this resolvase has little sequence specificity and would, therefore, be expected to cut at sites similar to GEN1. We did not observe any significant similarities between the cleavage patterns generated by GEN1 and T4 endonuclease VII (Fig. 5B, compare lane e with lane g). These results indicate that the cleavage products generated by GEN1 are not caused by the accumulation of HJs at certain sequences during branch migration. Instead, it seems that GEN1 has a sequence preference that is less stringent than that observed with RuvC.

Fig. 5.

Sequence preference for HJ cleavage by GEN1. (A) Large-scale resolution reactions (200 µL) contained ³²P-labeled α-structures with either RuvC (10 nM; lanes b–d) or GEN1 (1 nM; lanes f–h). Before the addition of protein, spontaneous branch migration of the HJ was induced by incubation at 55 °C for the times indicated. ³²P-labeled resolution products were separated by denaturing PAGE and autoradiography. (B) Comparison of the cleavage profiles by GEN1 (1 nM), RuvC (10 nM), and T4 endonuclease VII (0.5 U/μL). Reactions were carried out as in A. The recombination intermediates were branch-migrated where indicated (+). ³²P-labeled products were analyzed by denaturing PAGE and autoradiography.

To map the sites of HJ cleavage by GEN1, resolution reactions were performed, and the products were denatured for primer extension analysis using three 5′ ³²P-labeled primers located at various distances from the initial junction point (Fig. 6A, primers P1–P3). Cleavage sites close to the heterologous block were identified using nonbranch-migrated substrates, whereas sites farther away were mapped using branch-migrated α-structures. In each case, the precise sites of resolution were mapped compared with ³⁵S-labeled A, C, G, and T sequencing ladders produced from the same primers. Three representative primer extension reactions are shown in Fig. 6B. In total, 24 sites of cleavage were mapped, and the sequences in which cleavage occurred were aligned and displayed using the WebLogo tool (weblogo.berkeley.edu). This analysis revealed that cleavage occurred preferentially but not exclusively between two G residues and that T residues were common at positions −3 and −4 (Fig. 6C, Upper). However, a comparison with 17 cleavage sites produced by RuvC (26) (Fig. 6C, Lower) emphasizes that GEN1 fails to exhibit a strong preference for a defined consensus sequence.

Fig. 6.

Identification of the preferred sites of HJ cleavage by GEN1. (A) Schematic diagram indicating the α-structure and primers (P1–P3) used to map the sites of cleavage. (B) Reactions were carried out as described for Fig. 5, except that the α-structures were generated from unlabeled DNAs, and the resolution products were then subjected to primer extension using 5′ ³²P-labeled primers. The products of primer extension (lanes a, b, and g) were analyzed by denaturing PAGE alongside ³⁵S-labeled A, C, G, and T sequencing ladders. (Top) Cleavage sites close to the heterologous block were identified using nonbranch-migrated substrates, whereas (Middle and Bottom) sites farther away were mapped using branch-migrated α-structures. Three different oligonucleotides were used to map the cleavage sites: (Top) primer P1, (Middle) primer P2, and (Bottom) primer P3. (C, Upper) Twenty-four cleavage sites produced by GEN1 were aligned, and the occurrence of nucleotides A, C, G, and T at each position was determined and displayed using WebLogo. (C, Lower) For comparison, the RuvC cleavage sequence is also indicated.

In conclusion, we have extended our biochemical analysis of the GEN1 HJ resolvase using DNA substrates that better represent recombination intermediates that are formed in vivo. Using RecA-generated α-structures that contained single HJs and a plasmid-based substrate that contains two HJs separated by 165 bp, we found that GEN1 promoted the efficient resolution of both single-HJ and dHJ structures. The results are consistent with a role for GEN1 in resolving recombination intermediates in prometaphase in preparation for the segregation of the sister chromatids by the mitotic spindle (39, 61).

Although GEN1 and RuvC are structurally unrelated, there are a number of biochemical similarities in terms of their mechanism of action: (i) the two proteins are structure-selective endonucleases, (ii) they promote HJ resolution by formation of a homodimer–HJ complex, and (iii) resolution occurs by a mechanism that introduces symmetrical incisions across the junction, which releases ligatable nicked duplex products. However, unlike RuvC, which preferentially cleaves HJs at the preferred DNA sequence 5′-^A/_TTT^G/_C-3′, GEN1 exhibits a weak specificity, with incisions being introduced preferentially between two G residues, in a region that is T-rich. This general lack of sequence specificity may relate to observations indicating that GEN1 acts alone, in contrast to RuvC, which associates with the RuvAB branch migration complex (62, 63) or reflects the essential need to resolve all single HJs and dHJs (and potentially, other DNA secondary structures) that persist until the late stages of the cell cycle. In this regard, GEN1 is functionally similar to the bacteriophage resolvases T4 endonuclease VII and T7 endonuclease I, which ensure the resolution of all DNA secondary structures before the packaging of DNA into the phage heads (64, 65). By analogy, GEN1’s broad substrate specificity and sequence flexibility ensure that persistent sister chromatid bridges are resolved before chromosome segregation and cell division (31, 61).

Materials and Methods

Proteins.

C-terminally FLAG-His–tagged human GEN1 and nuclease-dead GEN1^E134A,E136A proteins were purified from Saccharomyces cerevisiae after overexpression from a galactose-inducible promoter (31). E. coli RuvC was purified as described (66). T7 endonuclease I, T4 endonuclease VII, E. coli RecA, and terminal transferase were purchased from New England Biolabs. Proteinase K was purchased from Sigma, and the Klenow fragment of DNA polymerase I (exo⁻) was purchased from Thermo Scientific.

DNA.

Recombination intermediates containing a single HJ (α-structures) were generated by RecA-mediated strand exchange between gapped circular pDEA-7Z (15 µM) and 3′ ³²P end-labeled linearized pDEA2 plasmid DNA (15 µM) (26, 67). After 30 min at 37 °C, the reactions were stopped and deproteinized, and the DNA was purified through a 3.5-mL Sepharose CL-2B column equilibrated with 20 mM Tris⋅HCl, pH 8.0, 5 mM MgCl₂, 1 mM DTT, and 100 µg/mL BSA.

For some experiments, the HJ was migrated away from the heterologous block by incubation of the column-purified intermediates at 55 °C. The branch migration reaction was stopped by chilling on ice, and the recombination intermediates were used immediately.

dHJ-containing substrates were prepared as described (57).

Endonuclease Assays.

Reactions (10 μL) containing 3′ ³²P-labeled α-structures (1.5 nM) and purified enzymes were incubated at 37 °C for 10 min unless otherwise stated. The cleavage buffer was 50 mM Tris⋅HCl, pH 8.0, 1 mM MgCl₂, 1 mM DTT, and 100 µg/mL BSA. For reactions containing RuvC or T4 endonuclease VII, the cleavage buffer was adjusted to 10 mM MgCl₂. Before use, GEN1 was diluted in buffer containing 50 mM Tris⋅HCl, pH 7.5, 250 mM NaCl, 10% (vol/vol) glycerol, 1 mM DTT, and 0.05% Nonidet-P40. All other proteins were diluted in 50 mM Tris⋅HCl, pH 8.0, 100 mM NaCl, 10% (vol/vol) glycerol, 1 mM DTT, and 100 µg/mL BSA. Reactions were stopped by addition of 2 µL 6× gel loading dye (New England Biolabs), and the products were analyzed by 1.1% agarose gel electrophoresis in the presence of 0.5 µg/mL ethidium bromide. DNA was visualized by either ethidium bromide staining or autoradiography.

For analysis by denaturing PAGE, reactions were scaled up to 200 µL. Reactions were stopped and deproteinized by the addition of 0.8% SDS, 20 mM Tris⋅HCl, pH 7.4, and 2 mg/mL proteinase K followed by incubation at 37 °C for 10 min. The DNA was then phenol-chloroform–extracted, ethanol-precipitated, and resuspended in 5 µL loading dye [80% (vol/vol) formamide, 0.1% bromophenol blue, 0.1% xylene cyanol]. Samples were denatured by heating at 95 °C for 3 min and analyzed by electrophoresis through 4% (wt/vol) polyacrylamide gels containing 7 M urea. Gels were dried, and the DNA was visualized by autoradiography.

Reactions (20 µL) containing dHJ DNA (2.15 nM) were incubated for 30 min at 37 °C, using essentially the same conditions as described above and stopped by addition of 1% SDS, 20 mM EDTA, and 100 µg/mL proteinase K. The DNA was deproteinized by phenol-chloroform extraction, ethanol-precipitated, and resuspended in 10 mM Tris⋅HCl, pH 8.0, and 0.1 mM EDTA. The reaction products were then separated by electrophoresis through 2% (wt/vol) agarose and visualized by staining with SYBR Gold (Life Technologies).

Primer Extension Analysis.

Large-scale (200 µL) resolution reactions containing α-structures were performed as above, except that unlabeled recombination intermediates (1.5 nM) were used. The products were deproteinized by addition of 0.8% SDS, 20 mM Tris⋅HCl, pH 7.4, and 2 mg/mL proteinase K followed by incubation at 37 °C for 10 min. DNA was phenol-chloroform–extracted, ethanol-precipitated, and resuspended in 10 mM Tris⋅HCl, pH 8.0, and 1 mM EDTA. Primer extension analysis of the resolution products was then performed using 5′ ³²P-labeled oligonucleotides and the Klenow fragment of DNA polymerase I as described (26). The sequences of the primers are

• Primer P1: 5′-GGACCGCTGATCGTCACGGCGATTTATG-3′,

• Primer P2: 5′-CTTTTGCTCACATGTTCTTTCCTGCGTTATC-3′, and

• Primer P3: 5′-GTATTCTATAGTGTCACTAAATAGCTTGGC-3.

Primers P1–P3 were also used to generate ³⁵S-labeled DNA sequencing ladders from plasmid pDEA2 using a Sequenase Kit, version 2.0 (Affymetrix). DNA products were analyzed on 6% (wt/vol) polyacrylamide gels containing 7 M urea followed by autoradiography of the dried gels. Sites of cleavage were aligned and analyzed using WebLogo (weblogo.berkeley.edu).