Holliday junction resolution in human cells: two junction endonucleases with distinct substrate specificities

Abstract

Enzymatic activities that cleave Holliday junctions are required for the resolution of recombination intermediates and for the restart of stalled replication forks. Here we show that human cell-free extracts possess two distinct endonucleases that can cleave Holliday junctions. The first cleaves Holliday junctions in a structure- and sequence-specific manner, and associates with an ATP-dependent branch migration activity. Together, these activities promote branch migration/resolution reactions similar to those catalysed by the Escherichia coli RuvABC resolvasome. Like RuvC-mediated resolution, the products can be religated. The second, containing Mus81 protein, cuts Holliday junctions but the products are mostly non-ligatable. Each nuclease has a defined substrate specificity: the branch migration-associated resolvase is highly specific for Holliday junctions, whereas the Mus81-associated endonuclease is one order of magnitude more active upon replication fork and 3'-flap structures. Thus, both nucleases are capable of cutting Holliday junctions formed during recombination or through the regression of stalled replication forks. However, the Mus81-associated endonuclease may play a more direct role in replication fork collapse by catalysing the cleavage of stalled fork structures.

Fig. 1. Fractionation of HeLa cell-free extract. (A) Purification scheme (see Materials and methods for details). The presence or absence of Mus81 protein, branch migration (BM) and resolution (R) activities in the butyl–Sepharose fractions are indicated with (+) and (–). (B) Characterization of the SP-Sepharose fractions. Fractions were analysed for the presence of branch migration and resolution activities (upper panel), and for Mus81 protein by western blotting (lower panel). Fractions with relevant activities (pools d and c, respectively) were purified further as indicated. The products of branch migration and resolution (splayed arm and nicked duplexes, respectively) are indicated. Lane C, control reaction from which protein was omitted. Lane L, reaction containing the elutate from heparin that was loaded onto SP-Sepharose. (C) Fractionation of SP-Sepharose pool c on MonoQ. Fractions were analysed for the presence of Holliday junction resolvase activity (upper panel) and for Mus81 by western blotting (lower panel). (D) Fractionation of SP-Sepharose pool d on MonoQ.

Fig. 2. Two distinct Holliday junction endonucleases. (A) Holliday junction resolution reactions contained either the branch migration-associated resolvase A or the Mus81-associated endonuclease. Nicked duplex products were analysed by neutral agarose gel electrophoresis. Reactions were carried out in the absence of ATP, and with the most highly purified fractions (0.5 µl) from MonoQ. (B) Detection of Mus81 by Western blotting using aliquots (20 µl) of the fractions used in (A). (C) Resolvase fractions from MonoQ were immunodepleted using anti-Mus81 or anti-GST antibodies coupled to protein A–Sepharose beads, as described in Materials and methods. The beads were then precipitated and the supernatants incubated with ³²P-labelled Holliday junction DNA. Lanes a and d, controls in which the resolvases were incubated on ice without Sepharose beads; lanes b and e, after pre-incubation with anti-GST–Sepharose beads; lanes c and f, after pre-incubation with anti-Mus81–Sepharose beads.

Fig. 3. Holliday junction resolution by RuvC, resolvase A and Mus81-associated endonuclease: incisions in strands 2 and strand 4. (A) The synthetic Holliday junction containing a 26 bp homologous core, 5′-³²P-end-labelled either in strand 2 (2*) or strand 4 (4*), was incubated without proteins (lanes a and b), with 100 nM RuvC (lanes c and d), with 0.5 µl of MonoQ-purified resolvase A (lanes e and f) or with 0.5 µl of MonoQ-purified Mus81-associated endonuclease (lanes g and h). Cleavage products were analysed by denaturing PAGE and run alongside a C>T sequence ladder produced from 5′-³²P-end-labelled strand 4. The homologous core and the terminal regions of heterology are indicated schematically by white and grey boxes, respectively. (B) Schematic representation of the incision sites. The nucleotide sequence of the homologous core region is shown, as are the sites of strong and weak cleavage indicated by the data presented in (A).

Fig. 4. Holliday junction resolution by RuvC, resolvase A and Mus81-associated endonuclease: incisions in strands 1 and 3. (A) Reactions were conducted as in Figure 3A using synthetic Holliday junctions labelled either in strand 1 (1*) or strand 3 (3*). Resolution by RuvC (lanes a and b), resolvase A (lanes c and d) and Mus81-associated endonuclease (lanes e and f). Lane g is a G+A sequence ladder produced from strand 3. (B) Schematic representation of the major and minor cleavage sites.

Fig. 5. Nick repair by DNA ligase. (A) Schematic representation indicating the synthetic Holliday junction with one short arm. Resolution of the junction by symmetrical cleavage followed by nick ligation converts the 53-nucleotide-long 5′-³²P-labelled strand into one that is 60 nucleotides in length. The ³²P-label is indicated by the asterisk. (B) The synthetic Holliday junction shown in (A) was incubated with RuvC, resolvase A or Mus81-associated endonuclease, and the reactions were supplemented with T4 DNA ligase, where indicated. Control, reactions without endonuclease. ³²P-labelled DNA products were analysed by denaturing PAGE followed by autoradiography.

Fig. 6. Substrate specificity of Mus81-associated endonuclease. (A) Activity of Mus81-associated endonuclease on the branched substrates indicated. Reactions were carried out as described in Materials and methods, and DNA products were analysed by neutral PAGE. All substrates were 5′-³²P-end-labelled on a common oligonucleotide. (B) Quantification of the efficiency of cleavage by Mus81-associated resolvase with replication fork (RF), 3′-flap and Holliday junction (HJ) substrates. At the times indicated, aliquots were withdrawn and the products analysed by neutral PAGE. Cleavage products were quantified by phosphoimaging and expressed as a percentage of total radiolabel. The values presented are an average of three independent time courses.