|Home | About | Journals | Submit | Contact Us | Français|
The formation of crossovers is a fundamental genetic process. The XPF-family endonuclease Mus81-Mms4 (Eme1) contributes significantly to crossing over in eukaryotes. A key question is whether Mus81-Mms4 can process Holliday junctions that contain four uninterrupted strands. Holliday junction cleavage requires the coordination of two active sites, necessitating the assembly of two Mus81-Mms4 heterodimers. Contrary to this expectation, we show that Saccharomyces cerevisiae Mus81-Mms4 exists as a single heterodimer both in solution and when bound to DNA substrates in vitro. Consistently, immunoprecipitation experiments demonstrate that Mus81-Mms4 does not multimerize in vivo. Moreover, chromatin-bound Mus81-Mms4 does not detectably form higher-order multimers. We show that Cdc5 kinase activates Mus81-Mms4 nuclease activity on 3′ flaps and Holliday junctions in vitro but that activation does not induce a preference for Holliday junctions and does not induce multimerization of the Mus81-Mms4 heterodimer. These data support a model in which Mus81-Mms4 cleaves nicked recombination intermediates such as displacement loops (D-loops), nicked Holliday junctions, or 3′ flaps but not intact Holliday junctions with four uninterrupted strands. We infer that Mus81-dependent crossing over occurs in a noncanonical manner that does not involve the coordinated cleavage of classic Holliday junctions.
Robin Holliday first proposed a mechanism for crossover formation (29). Based on fungal tetrad data, he envisioned that nick-induced heteroduplex formation could result in a DNA intermediate composed of four intact strands after ligation of the strand interruptions. This intermediate was later termed the Holliday junction (HJ). Cleavage across the two alternative planes of this junction would result in crossover (CO) or noncrossover (NCO) products, depending on the orientation of cleavage. Biochemical analysis of the bacterial RuvC nuclease supports this model and provides a paradigm for a class of enzymes called Holliday junction resolvases. These nucleases form homodimeric complexes to deliver two coordinated and symmetric endonucleolytic cuts that generate DNA ends that can be directly ligated to form recombinant products (reviewed in reference 38). However, RuvC is not evolutionarily conserved in eukaryotes, and the specific mechanisms of crossover formation in eukaryotes are still undefined. Refinement and expansion of the original Holliday model have produced the current model of double-strand break repair, in which one subpathway is defined by the formation of a double Holliday junction (dHJ) (46). Physical analysis has demonstrated the existence of dHJs in meiotic and mitotic recombination (12, 48). However, these studies could not establish whether these junctions were truly dHJs, i.e., with each individual junction having four uninterrupted strands, or were nicked junctions in a dHJ population where each strand could be found to be full length (for more discussion, see reference 49). Hence, the importance of Holliday junctions and their cleavage in CO formation still remains to be demonstrated.
The structure-selective endonuclease Mus81-Mms4 (Mms4 is known as Eme1 in other organisms) contributes to CO formation in budding and fission yeast as well as in Arabidopsis and mice, suggesting a role in joint molecule processing (6, 20, 28, 30, 45, 52). Mus81 was identified through its physical interactions with the Saccharomyces cerevisiae recombination protein Rad54 and the Schizosaccharomyces pombe DNA damage response kinase Cds1, as well as in a genetic screen for genes required in the absence of the Sgs1 helicase (10, 31, 41). A fundamental question is, what are the DNA joint molecules targeted by Mus81-Mms4 in vivo? Biochemical analysis of purified and partially purified Mus81-Mms4 protein shows catalytic and robust cleavage of nicked, 3′-flap, and displacement loop (D-loop) substrates to be the preferential in vitro target for Mus81 (5, 9, 13, 14, 21, 23, 26, 45). Yet, it is the highly inefficient incision of intact, four-way HJs that has resulted in widely discussed models suggesting that intact HJs or dHJs are the physiological substrates for Mus81-Mms4 in CO formation (9, 14, 26, 38, 52, 57).
Sgs1, in conjunction with Top3 and Rmi1, provides an alternative mechanism to process dHJs termed dissolution (see Fig. 17B), a reaction discovered with the human BLM-TOPOIIIalpha-RMI1 complex (61). Dissolution describes the coordinate movement of both junctions in a dHJ toward each other by combined action of the Sgs1/BLM helicase and Top3/TOPOIIIalpha topoisomerase, resulting in a single hemicatenane, which is resolved by the type IA topoisomerase Top3/TOPOIIIalpha, stimulated by the Rmi1 specificity factor. Dissolution is the only biochemical mechanism demonstrated to process dHJs, but it always leads to NCO products, leaving the mechanism of CO formation in eukaryotes unaddressed.
In budding yeast and Drosophila, the genetic requirement for Mus81 (or Mms4/Eme1) in the absence of Sgs1 is completely dependent on the key recombination protein Rad51, suggesting that both the Sgs1-Top3-Rmi1 complex and the Mus81-Mms4 heterodimer function in the late stages of recombination to resolve Rad51-dependent DNA joint molecules (5, 22, 58). Physical analysis of meiotic DNA products has shown that loss of both Mus81 and Sgs1 during meiosis results in meiotic catastrophe and a failure to segregate chromosomes, consistent with the inability to resolve DNA joint molecules (33, 43). These data provide compelling support for the model that Mus81-Mms4 cleaves late recombination intermediates to form COs but leaves open the question of the physical nature of these intermediates.
A long-term effort to identify the eukaryotic equivalent of the paradigmatic bacterial RuvC HJ resolvase resulted in the identification of S. cerevisiae Yen1/human GEN1 (32). An N-terminal fragment of Yen1 or GEN1 is capable of HJ resolution across the plane of the junction, resulting in religatable products, although Yen1/GEN1 and the full-length Drosophila GEN1 also cleave other substrates (32, 34, 47). Human GEN1 binds to the HJ substrate as a multimer, providing the subunit architecture for coordinated HJ cleavage by two active sites (47).
On the basis of the biochemical data, it was suggested that Yen1/GEN1 is the long-sought HJ resolvase responsible for CO formation in eukaryotes (32). However, genetic analysis failed to discover any phenotype in the budding yeast yen1 mutant or in human cells sensitized by a mutation in the BLM gene depleted of GEN1 by small interfering RNA (1, 8, 40, 56, 60). Fission yeast entirely lacks this gene, and CO formation is almost completely dependent on Mus81-Eme1 (52). Elegant genetic analysis of CO formation illuminated the enzymes required for CO formation in vegetative (somatic) budding yeast cells (28). Mutations in MUS81 reduced all CO classes, demonstrating the critical role of Mus81-Mms4 also in CO formation in vegetative yeast cells (28), after such a role was already documented for meiotic recombination (3, 19, 20, 43). Mutations in YEN1 had only a minor effect and reduced only COs that were associated with short conversion tracts (28). In the mus81 yen1 double mutants, somatic COs were essentially eliminated (28), suggesting that in the absence of Mus81-Mms4, Yen1 can cleave either the original Mus81-Mms4 substrate(s) or its processing products (see Discussion). These results, together with the biochemical substrate preferences for Mus81-Mms4 and Yen1, suggest that the majority of CO formation in wild-type vegetative budding yeast, and possibly all eukaryotes, does not involve classic HJs or dHJs but involves other types of recombination-dependent junctions.
Cdc5 (Polo) kinase regulates meiotic progression and crossover formation (16, 53), and Mus81-Mms4 controls about a third of the meiotic crossovers in budding yeast (3, 19, 20; reviewed in reference 49). Cdc5 expression is induced during meiotic prophase, which correlates with phosphorylation of Mms4 and activation of the Mus81-Mms4 nuclease activity on an oligonucleotide-based HJ substrate (40). The same Cdc5-dependent activation was demonstrated to occur during mitotic cell cycles at the onset of anaphase (40). Analysis of a phosphorylation-defective Mms4 mutant, in which 14 predicted or mapped phosphorylation sites were mutated to the nonphosphorylatable residue alanine, suggested that Cdc5-mediated phosphorylation of the Mms4 subunit activates Mus81-Mms4 nuclease activity (40). The mechanism by which phosphorylation activates Mus81-Mms4 nuclease activity and whether this involves tetramerization of the Mus81-Mms4 heterodimer remain to be determined.
We have previously demonstrated that budding yeast Mus81-Mms4 strongly prefers nicked junctions and D-loops as the substrates, whereas HJ cleavage was so weak that classic kinetic (Michaelis-Menten) analysis could not be performed (21; see also reference 23). To address whether Mus81-Mms4 might cleave HJs in vivo, we have determined the subunit composition of the Mus81-Mms4 complex in vitro and in vivo. The hydrodynamic properties of S. cerevisiae Mus81-Mms4 demonstrate that it exists as a single heterodimer in solution and when bound to 3′ Flap or HJ DNA substrates. Transmission electron microscopy (TEM) of individually cross-linked protein particles with subsequent volumetric analysis of particle classes supports a heterodimer model. In vivo coimmunoprecipitation (co-IP) of soluble protein failed to show any self-association of free or chromatin-associated subunits into larger oligomeric complexes. We demonstrate Cdc5-mediated activation with purified Mus81-Mms4 in vitro and show that this does not involve multimerization of the enzyme. These results also show that the catalytic form of Mus81-Mms4 in vivo contains a single nuclease active site for preferential cleavage of nicked substrates and D-loops, pointing to a mechanism of CO formation in eukaryotes that likely involves nicked junctions rather than the iconic single or double Holliday junctions with four intact strands.
A complete list of strains used in this study is found in Table 1. Tagging constructs were prepared as previously described (4), and correct integration was verified by PCR and immunoblot analysis. The functionality of tagged proteins is demonstrated in Fig. 9A. Oligonucleotide-based 5′-32P-labeled and nonlabeled substrates were produced as described previously (21).
The overexpression vector used to produce Mus81-Mms4 is essentially the same as the previously described (21) pWDH595 containing the GAL1-GAL10 divergent promoter on a 2μm-based overexpression vector. However, pWDH595 contains a tobacco etch virus (TEV) protease sequence between the (His)10 tag and MUS81 which was similar enough in amino acid sequence to the PreScission site on Mms4 to also be cleaved by PreScission protease when glutathione S-transferase (GST) was removed from Mms4. To remove the problematic TEV protease site, MUS81 was PCR amplified with primers introducing MluI sites and cloned into these sites in pWDH595, replacing the original TEV-MUS81 cassette C terminal to the (His)10 tag to result in pWDH722. The overexpression plasmid pWDH722 fully complemented the genotoxin sensitivities of the mus81Δ mms4Δ double mutant strain as well as the mus81Δ mms4Δ yen1Δ triple mutant strain (see Fig. 9B and and1616).
(His)10Mus81-Mms4 was purified by two-column (glutathione-Sepharose 4b and nickel-nitrilotriacetic acid) affinity chromatography essentially as described previously (21), with the following modifications. The mass of cells (WDHY668/pWDH722) lysed was reduced to 60 g, and the volume of glutathione resin was increased to 12 ml. After binding GST-Mms4, the column was washed with three column volumes of 50 mM Tris-HCl, pH 7.5, 450 mM NaCl, 1 mM dithiothreitol (DTT), and 1 mM EDTA, after which 3 mg of PreScission protease was added and the resin bed was resuspended. The cleavage of the GST tag from Mms4 was allowed to proceed for 2.5 h at 4°C, and the (His)10Mus81-Mms4 heterodimer was then eluted with 25 ml of the same buffer. This material was applied to a 1-ml HisTrap FF column, and the purification was completed as described previously (21). Protein concentrations were determined by the Bradford assay using bovine serum albumin (BSA) as a standard. GST-cleaved Mms4 retains four additional amino acids, PRET, at the N terminus and is otherwise native, while Mus81 contains the N-terminal tag MRGS(His)10AS.
For gel filtration, 5 μg (His)10Mus81-Mms4 was applied to a Superdex 200 GL (GE Healthcare) gel filtration column preequilibrated with GF buffer (20 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1 mM DTT, 10% glycerol, and 500 mM NaCl). The column was developed with GF buffer at 0.5 ml/min using an Åkta purifier system (GE Healthcare) collecting 0.1-ml fractions. Protein levels were determined by immunoblots of 2 μl/fraction using a rat anti-Mms4 antibody. Fractions were immediately assayed for nuclease activity on 3′-flap junctions and Holliday junctions with a fixed core (XO12) as described below.
Sucrose gradients (4.5 ml, 5 to 20%) containing buffer GF with 500 or 150 mM NaCl was poured into 5-ml ultraclear tubes (13 by 51 mm; Beckman) using a plunger-type gradient maker (Jule, Inc.). For gradients containing 150 mM NaCl, Mus81-Mms4 samples (stored in 500 mM NaCl) were predialyzed against GF buffer containing 150 mM NaCl. The dialysis buffer ionic strength was gradually reduced by slowly mixing GF buffer without NaCl into GF buffer with 500 mM NaCl over the course of 4 h. The protein concentration was then redetermined by Bradford assay. Five micrograms of purified Mus81-Mms4 in 50 μl GF buffer was then carefully layered on top of the gradient. Standard protein stocks in GF buffer at 20 mg/ml were then mixed and layered on top of the gradients in a volume of 65 μl consisting of 5 μl BSA, 5 μl ovalbumin, 15 μl aldolase, 15 μl catalase, and 25 μl thyroglobulin. Gradients were centrifuged for 9 h at 55,000 rpm in an SW 55 Ti swinging-bucket rotor using a Beckman Optima LE ultracentrifuge. Gradients were then fractionated from the top down by carefully pipetting 175 μl solution into 26 fractions per gradient. Each gradient was calibrated by determining the position of standard proteins by 10% SDS-PAGE of fraction samples, followed by Coomassie blue staining and densitometry using ImageQuant software (GE) to determine the peak fraction. Peak fraction positions of standard proteins were then plotted against their known sedimentation coefficient (S) values to generate internal standard curves for each gradient. The Mms4 protein peak was determined by immunoblot analysis using rat anti-Mms4 antibody. To avoid further loss of the limited resolution of the technique, if the protein level in two fractions differed by less than 10%, the peak was taken to be between the two fractions (e.g., if fractions 7 and 8 differed by less than 10%, the peak was taken to be fraction 7.5). To avoid possible saturation of immunoblot signals on film, high and low exposures of the same blot were taken and monitored for consistency in the relative signal between fractions. The relative signals were normalized by taking the peak fraction to be 100%. Two exposures normalized in such a way were then averaged.
For gel filtration samples, 2.5 μl of each fraction (in GF buffer containing 500 mM NaCl) was mixed with 7.5 μl of a cocktail to adjust the final reaction to 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 3 mM magnesium diacetate [Mg(OAc)2], 0.25 mg/ml BSA, 1 mM DTT. Reaction mixtures contained 50 nM 3′-flap (defined by unlabeled substrate concentration, containing 1,500 cpm labeled substrate per reaction mixture) or 1 nM molecules for XO12 reactions. Reactions were allowed to proceed for 15 min at 30°C and were then stopped with the addition of 2 μl stop mix (200 mM EDTA, 2.5% SDS, and 10 mg/ml proteinase K). For sedimentation fractions, a representative gradient containing 150 mM NaCl was assayed for cleavage of 3′-flap and XO12 structures in the same manner. The reactions were stopped after 10 min at 30°C with the addition of 2 μl stop mix. One-third of the reaction mixtures (500 cpm) was electrophoresed by 10- by 20-cm native 10% TBE (Tris-borate-EDTA)-PAGE at 100 V for 65 min. Gels were vacuum dried to Whatman paper at 80°C and then exposed overnight to a phosphorimager screen.
S. cerevisiae Slx1-Slx4 (courtesy of S. Brill) was titrated into 10-μl reaction volumes containing 5 nM Mus81-Mms4 and 50 nM substrate (nicked Holliday junction [nXO12], replication fork (RF)-like structure [RF-like], XO12, or Holliday junction with mobile core [X12]). Nuclease reactions were performed in 25 mM HEPES, pH 7.5, 100 mM NaCl, 3 mM Mg(OAc)2, 1 mM DTT, 100 μg/ml BSA, with the substrate concentration defined by that of the unlabeled substrate and with the mixture spiked with a negligible quantity of radiolabeled substrate to report on substrate turnover (the substrate preparation was described previously ). After 10 min incubation at 30°C, reactions were quenched at 40 mM EDTA, 0.5% SDS, 2 mg/ml proteinase K before electrophoresis by native 10% TBE-PAGE at 150 V for 1 h. Gels were dried for 1 h at 80°C and then processed for phosphorimaging.
S. cerevisiae cells overexpressing 3× hemagglutinin (3HA)–Cdc5 wild-type or 3HA-Cdc5-K110A kinase-defective protein were grown and induced with galactose as described previously (59). Wild-type or kinase-defective Cdc5 was immunoprecipitated from 50 optical density (OD) units of cells, and kinase reactions were performed essentially as described previously (59) with purified Mus81-Mms4 as the substrate. For the experiment whose results are shown in Fig. 3A, ,55 μg of Mus81-Mms4 was incubated with IP beads in 80 μl. After 15 min incubation at 30°C, the supernatant was removed from the beads. One-fourth of this material was treated with PP1 phosphatase (2.5 U) with or without prior heating or I2 PP1 inhibitor, as indicated. Reactions were stopped with Laemmli buffer for immunoblot analysis. For the experiments whose results are shown in Fig. 5B and andC,C, the wild-type Cdc5 reaction was carried out in the same manner. As controls, purified Mus81-Mms4 was incubated with PP1 phosphatase or simply in kinase buffer for 15 min. For nuclease assays, heterodimer concentrations were redetermined (to account for loss to nonspecific binding) by anti-Mms4 Western blotting in comparison to purified protein standards.
Gradient fixation was performed essentially as described previously (35). Gradients were prepared as described above; however, 0 to 15% sucrose gradients were used in HB [25 mM HEPES, pH 7.9, 1 mM EDTA, 10% glycerol, 1 mM DTT, and 0.5 M or 1.0 M NaCl for (His)10Mus81-(GST)Mms4 and (His)10Mus81-Mms4, respectively, with or without 0.1% formaldehyde]. (His)10Mus81-Mms4 was slowly dialyzed overnight into a final HB solution containing 1.0 M NaCl. Five micrograms of protein was diluted in 200 μl HB with the corresponding salt concentrations and carefully applied to the gradient. Globular protein standards at 20 mg/ml were mixed in a volume of 40 μl (5 μl BSA, 5 μl ovalbumin, 15 μl aldolase, and 15 μl catalase) and applied to gradients without fixative to determine sedimentation values. Samples were centrifuged as described above and fractionated. Protein elution was determined by immuno-dot blot analysis using rabbit anti-Mus81. Protein-containing fractions were used for grid preparation and stained with 0.075% uranyl formate as described previously (44). Grids were visualized using a JEOL JEM-1230 microscope (JEOL Ltd., Japan) operated at an accelerating voltage of 120 keV. Images were acquired using minimum-dose procedures at a nominal magnification of 50,000× on a TVIPS TemCam-F224HD 2,048- by 2,048-pixel charge-coupled-device camera (TVIPS, Germany). Particle selection and classifications were completed using EMAN2 software, stable version 2.0RC3 (55).
Diploid cells were grown in yeast extract-peptone-dextrose (YPD) to an optical density at 600 nm (OD600) of 2.0, resuspended either in fresh YPD medium or in medium treated with either 0.015% methyl methanesulfonate (MMS), 15 μM camptothecin (CPT), or 200 mM hydroxyurea (HU), and incubated for 2 h. Cells were pelleted, washed in IP-A (20 mM Tris, pH 7.5, 100 mM NaCl, and 1 mM EDTA) containing 1 mM phenylmethylsulfonyl fluoride (PMSF), and flash frozen in liquid nitrogen. Immunoprecipitations of protein and immunoblotting of precipitates were completed as previously described (4) using anti-HA (MMS-101P; Covance, Princeton, NJ), anti-cMyc (Cell Signaling, Danvers, MA), rabbit anti-Mus81, or rabbit anti-Mms4 antibodies.
Cultures were grown and treated as described for soluble immunoprecipitation and were subsequently treated with 1% formaldehyde for 15 min, followed by incubation with 125 mM glycine for 5 min. Cells were washed with ice-cold 1× phosphate-buffered saline with 1 mM PMSF before pelleting and freezing. Fifty OD600 units of cells was resuspended in lysis buffer (50 mM HEPES, pH 7.5, 1 mM EDTA, 140 mM NaCl, 1% Triton X-100, 0.1% Na-deoxycholate, 1 mM DTT) with protease inhibitors (1 mM PMSF, 2 μM leupeptin, 1 μM benzamidine, 1.67 μM pepstatin), 1 ml glass beads was added (0.5 mm; Biospec Products, Inc., Bartlesville, OK), and cells were lysed using a Savant FastPrep FP120 device (Bio101) on setting 4 four times for 45 s each time with a 5-min incubation on ice after two cycles. Chromatin and debris were collected by centrifugation, washed two times with fresh lysis buffer, and finally, resuspended in lysis buffer. Bulk chromatin was sonicated to 250-bp fragments using a Diagenode Bioruptor sonicator (Denville, NJ). Debris and sheared chromatin were collected by centrifugation. For DNase assays, the chromatin-containing pellet was washed twice with DNase buffer (40 mM Tris-HCl, pH 7.4, 10 mM NaCl, 6 mM MgCl2, 10 mM CaCl2). Approximately 5.0 × 107 cells were resuspended in 95 μl DNase buffer, into which 5 units of DNase or DNase buffer was added. Samples were incubated at 4°C for 30 min, and the reactions were stopped with the addition of 10 μl 20 mM EGTA, pH 8.0. Cross-links were reversed with the addition of an equal volume of elution buffer (1% SDS in TE [Tris-EDTA], pH 8.0) and incubation overnight at 65°C, followed by the addition of 100 μl proteinase K for an additional 2 h incubation at 37°C. Phenol extraction and ethanol precipitation were used to further extract the DNA. RNA was degraded by the addition of DNase-free RNase and incubation at 37°C for 2 h.
Immunoprecipitation was carried out as previously described (36) with either anti-HA (Covance, Princeton, NJ), anti-Myc (Cell Signaling, Danvers, MA), 10 μl rabbit anti-Mus81 serum, or anti-histone H3 antibody (AbCam, Cambridge, MA). Immunoprecipitates were boiled for 10 min in 1× Laemmli buffer, separated on an SDS-polyacrylamide gel, transferred to nitrocellulose, and probed using rat anti-Mus81 and anti-Myc antibodies.
We originally purified Mus81-Mms4 from the cognate S. cerevisiae cells as a doubly tagged, (His)10Mus81-(GST)Mms4 heterodimer (21). Since GST is known to induce dimerization, which would interfere with our analysis, we proteolytically removed the GST tag and performed the analysis with the (His)10Mus81-Mms4 heterodimer (Fig. 1A). To establish the native molecular weight (Mr) of (His)10Mus81-Mms4, we determined its hydrodynamic properties. Together, the Stokes radius and sedimentation coefficient allow a very reliable estimation of the native Mr using the equation of Siegel and Monty (Fig. 1B, equation 1) (51).
During gel filtration, (His)10Mus81-Mms4 elutes at a volume corresponding to a Stokes radius of 68 Å (Fig. 1C and andD),D), which for a globular protein would correspond to 450 kDa. This value is in the range of values previously reported for S. cerevisiae Mus81-Mms4 and S. pombe Mus81-Eme1 (23, 26). However, to determine the native Mr, it is imperative to also determine the sedimentation coefficient, which yielded an S value of 5.6 (Fig. 1C and andE).E). This is smaller than that expected for a single heterodimer. Using equation 1, these data result in a native molecular mass of (His)10Mus81-Mms4 of 158 ± 8 kDa, which is in excellent agreement with the predicted molecular mass of 153.6 kDa calculated from the amino acid sequence (Fig. 1C). Purified recombinant Mus81-Mms4 is therefore a single heterodimer in solution. Activity assays on the fractions from gel filtration and sucrose gradients revealed that the robust nuclease activity on 3′-Flap and the feeble activity on XO12 HJs coeluted with (His)10Mus81-Mms4 protein levels. There were no fractions containing enhanced activity on XO12 HJs compared to the 3′-flap nuclease activity (Fig. 1D and andE,E, ,2,2, and and3).3). The longer tails of the 3′-flap nuclease activity profile compared to the protein profile are related to the much higher sensitivity of the nuclease assay than the immunoblot (Fig. 1D and andE).E). Robust cleavage of 50 nM 3′-flap was observed in fractions containing (His)10Mus81-Mms4, while XO12 cleavage could be visualized only in reactions with 1 nM substrate (Fig. 2 and and3).3). The relatively higher XO12 cleavage activity in the fractions from the sedimentation analysis (Fig. 1E) compared to the fractions from the gel filtration analysis (Fig. 1D) is due to the higher Mus81-Mms4 concentration in the fractions of the sedimentation analysis. Mus81-Mms4 cleaves HJs such as XO12 only when it is in significant excess of protein over substrate (see references 21 and 23 and references therein; Fig. 2 and and3;3; see also Fig. 5B). The sedimentation profiles were not changed in the presence or absence of 3 mM magnesium or by high (500 mM NaCl) versus optimal (150 mM) salt concentrations (21), indicating that these factors do not influence the oligomeric status of (His)10Mus81-Mms4 (Fig. 3).
Since it is possible that Mus81-Mms4 multimerizes only when bound to DNA, we performed DNA binding experiments to test this possibility. Mus81-Mms4 (650 nM) was applied to 1 nM 5′ end 32P-labeled 3′-flaps or XO12 HJs, and the S value of these complexes was determined by sucrose gradient sedimentation. The shift in the DNA S value induced by Mus81-Mms4 binding in comparison to the values for gradients containing DNA alone was 4.4S and 4.8S for 3′-flap and XO12 HJ, respectively (Fig. 4A). This suggests that the same form of Mus81-Mms4 binds both structures. The magnitude of the DNA shifts was similar to the S value of the heterodimer in solution (5.6S) (Fig. 4B), strongly supporting the conclusion that a single heterodimer is also the species that binds DNA. It is noteworthy that even the high excess of Mus81-Mms4 (650-fold) to substrate did not lead to dimerization on the substrate.
Using immunoprecipitated Mus81-Mms4, Matos et al. (40) showed that Cdc5 kinase-dependent phosphorylation of Mms4 activated Mus81-Mms4 nuclease activity on oligonucleotide-based HJs. While the mechanism of activation remains to be determined, it is possible that Mms4 phosphorylation induces tetramerization of the Mus81-Mms4 heterodimer. To test this notion, we performed in vitro phosphorylation experiments using purified Mus81-Mms4 and Cdc5 kinase immunoprecipitated from budding yeast cells expressing an HA-tagged Cdc5 that was shown to be functional (59). Our Mus81-Mms4 is purified from budding yeast and known to be posttranslationally modified (21), consistent with the electrophoretic mobility of the Mms4 subunit shown in the lower panel of Fig. 5A. Incubation with catalytically active Cdc5 leads to additional phosphorylation, as indicated by the incorporation of radiolabeled ATP (Fig. 5A, lane 3). This additional phosphorylation is strongly reduced when using the catalytically defective Cdc5-K110A (lane 2), demonstrating the dependence of Mms4 in vitro phosphorylation on active Cdc5 kinase. Mms4 phosphorylation is abolished by phosphatase treatment, as shown by a lack of radiolabel and increased electrophoretic mobility (lane 4), but preserved when either phosphatase inhibitor (lane 5) or heat-inactivated phosphatase (lane 6) is added. In addition, we note that Mus81 is also phosphorylated by Cdc5 (Fig. 5A). As expected from the previous report (40), Cdc5 activates Mus81-Mms4 nuclease in vitro on an HJ substrate, while no enhancement of endonuclease activity is observed for phosphatase-treated or mock-treated controls (Fig. 5B). Cdc5 also activated the 3′-flap cleavage activity of Mus81-Mms4 (Fig. 5C), which is consistent with observations made with the immunoprecipitated Mus81-Mms4 (40; S. West, personal communication) that Cdc5 activation does not change Mus81-Mms4 substrate preference for nicked substrates over HJs. The kcat of Mus81-Mms4 for 3′-flaps was stimulated about 5-fold (Fig. 5C), whereas we could not determine the kcat for the HJ substrate because an excess of Mus81-Mms4 over substrate was required with either Cdc5-treated or untreated/phosphatase-treated Mus81-Mms4 (Fig. 5B). These results demonstrate that Cdc5 phosphorylates Mms4 and Mus81 in vitro, leading to a general activation of the Mus81-Mms4 nuclease activity.
To test whether activation is accompanied by tetramerization of the Mus81-Mms4 heterodimer, we determined the S value of Cdc5-treated and control phosphatase-treated Mus81-Mms4 by gradient centrifugation (Fig. 5D). The S value of 5.6 remained unchanged by phosphorylation (Fig. 5D; see also Fig. 1C). This result eliminates dimerization as a mechanism of Mus81-Mms4 activation, consistent with the observation that Cdc5-mediated activation does not change substrate specificity for substrates requiring one (3′ flaps) or two (HJ) active sites.
We used transmission electron microscopy (TEM) to visualize individual particles of (His)10Mus81-Mms4 (Fig. 6). Conditions for conventional negative-stain sample preparation were not conducive for the isolation of single particles due to the propensity of the protein to aggregate on carbon grids. Therefore, we slowly cross-linked the proteins by gradient fixation to stabilize individual protein complexes. The presence of cross-linker, necessary to visualize independent particles, resulted in a shift in the sedimentation coefficient value from 5.7 to 7.3 (Fig. 7A and andB).B). This larger value corresponds to a globular protein with a molecular mass of 141 kDa. Visualization of the particles by TEM is consistent with this value, showing a globular single Mus81-Mms4 heterodimer (Fig. 6A and andB).B). Volumetric analysis confirms the presence of a predominately single heterodimer population (Fig. 6C and andD).D). A subset of particles (~30%) showed larger volumes (Fig. 6B to toD).D). Such particles are likely the result of random cross-linking, as indicated by the occurrence of some particles that correspond to trimers and the absence of a pattern to suggest specific tetramer formation. A complex of three heterodimers would result in cleavage products that appear to have little physiological relevance. The complete absence of assemblies larger than a heterodimer in the sedimentation and gel filtration analysis of non-cross-linked Mus81-Mms4 (Fig. 1D) excludes the possibility that cross-linking prevented multimerization.
To corroborate the conclusion that Mus81-Mms4 does not form a dimer of heterodimers, we performed single-particle analysis with the GST fusion protein (His)10Mus81-(GST)Mms4, which favors dimer formation through GST (Fig. 7B to toD).D). As predicted, the S value upon fixation was two times that of the cross-linked untagged protein, and the fusion protein eluted with an S value of 15.1. This value is consistent with a globular molecular mass of 340 kDa, close to the predicted molecular mass of the heterotetramer at 356 kDa. Congruent with these findings, single-particle analysis showed that a majority of the particles with (His)10Mus81-(GST)Mms4 had an estimated volume consistent with a dimer of heterodimers (Fig. 7B). Control experiments to look at protein particles under tilted conditions (Fig. 8) excluded artifacts due to particle orientation or understaining. In sum, the TEM analysis independently confirms that Mus81-Mms4 primarily exists as a single heterodimer.
There remains a possibility that posttranslational modifications, other than Cdc5-mediated phosphorylation, or accessory factors facilitate tetramer formation of Mus81-Mms4 in vivo. To determine the Mus81-Mms4 oligomeric state in vivo, we performed coimmunoprecipitation experiments in diploid cells containing either HA- and Myc-tagged Mus81 or HA- and Myc-tagged Mms4 expressed from their endogenous promoters. The tagged constructs were verified to be functional by wild-type levels of growth on genotoxin-containing medium (Fig. 9A). Immunoblot analysis of the fusion proteins showed expression levels similar to those of the native proteins (Fig. 10A and and11).11). No self-association into a larger oligomer was observed for either Mus81 or Mms4 (Fig. 10B and andC).C). We also performed these experiments under conditions of genotoxic stress (HU, CPT, and MMS), which require Mus81-Mms4 for survival. Again, under no conditions did we obtain evidence for self-association of Mus81 or Mms4 (Fig. 10B and andC).C). The detection limit of the antibodies was experimentally determined in reconstruction experiments to be 5 to 10%, depending on the antibody and protein (data not shown). This sensitivity ensures that we would have detected tetramers to this level and supports the conclusion from the EM experiments that the larger than heterodimeric assemblies are the result of nonspecific random cross-linking (Fig. 6). Using diploid control strains with either Myc- or HA-tagged Mus81, we demonstrated the specificity of the antibodies used for immunoprecipitation (Fig. 10B, first four lanes). In conclusion, S. cerevisiae Mus81-Mms4 exists as a single heterodimer in vivo.
Next, we wanted to test the possibility that Mus81-Mms4 dimerizes when bound to chromatin. Using the same approach used in the solution co-IP experiments, we employed Mus81-HA and Mus81-Myc in chromatin IP (protein co-ChIP experiment [ Fig. 12A and andB])B]) from diploid cells and found no evidence for self-association of Mus81 protein when bound to chromatin. Mus81-Mms4 cofractionated with histone H3, showing that chromatin-bound Mus81 was indeed analyzed (Fig. 12A). This conclusion was confirmed by additional control experiments that show that Mus81 and histone H3 can be solubilized from the chromatin-containing pellet by sonication, which fragments the DNA to a size of mostly 250 to 1,000 bp (Fig. 13). Also under conditions of relevant genotoxic stress (HU, CPT, and MMS), no Mus81 self-association could be detected by co-ChIP (Fig. 12B). Using diploid control strains with either Myc- or HA-tagged Mus81, we demonstrated the specificity of the antibodies used in the co-ChIP experiments (Fig. 12B, first four lanes). These results show that Mus81-Mms4 does not dimerize when bound to chromatin in vivo.
S. cerevisiae Slx1-Slx4 was identified together with Mus81-Mms4 in a screen for sgs1 synthetic lethals and also forms a heterodimeric endonuclease (24). Recent analyses in mammalian cells have suggested that Slx4 may be acting as an endonuclease scaffold to facilitate HJ cleavage (reviewed in reference 54). However, the presence of Slx4-Slx1 does not stimulate substrate cleavage or change the substrate specificity of purified Mus81-Mms4 in vitro (Fig. 14A and andB).B). In congruence with these studies, in S. cerevisiae, no coprecipitation was observed between Slx4 and Mus81 or Mms4 (Fig. 14C). This suggests that budding yeast Mus81 is not interacting with Slx4 to facilitate coordinated cleavage of DNA junctions. The involvement of S. cerevisiae Slx1-Slx4, another nuclease proposed to cleave HJs, in processing recombination-mediated joint molecules is unlikely, because the synthetic lethality between mutations in Slx1-Slx4 and Sgs1 is not suppressed by a recombination defect (24). This shows that the synthetic lethality between Slx1-Slx4 and Sgs1 is not related to recombination but, rather, is probably related to DNA replication (24), and a similar conclusion was reached for fission yeast Slx1-Slx4 (17).
Key to understanding the biological function(s) of the Mus81-Mms4 structure-selective endonuclease is the nature of its in vivo substrates, specifically, whether it cleaves HJs with four uninterrupted strands or nicked junctions, D-loops, and flaps. HJ resolvases, such as bacterial RuvC and RusA, T4 endonuclease VII, archaeal Hje, or mitochondrial Cce1, exist as dimers, delivering coordinated cleavage across the junction using two active sites (37). Hence, it is critical to determine whether Mus81-Mms4 exists as a higher-order multimer.
The hydrodynamic data demonstrate that Mus81-Mms4 is a single heterodimer. Further analysis (Fig. 1B, equation 2) suggests a heterodimer with nonglobular dimensions, with hydrodynamic properties similar to those of Rad17-Rfc2, -Rfc3, -Rfc4, and -Rfc5, which adopt a washer-like structure (50) akin to the spiral formed by the classic Rfc1-5 clamp loader (11) (Fig. 15). The large Stokes radius of the native complex explains why previous studies had concluded that Mus81-Mms4 was either a dimer (26) or a trimer (23) of heterodimers solely on the basis of gel filtration results. While we hoped to reveal the native structure of Mus81-Mms4 by electron microscopic analysis, the necessary cross-linking procedure appeared to collapse Mus81-Mms4 into a more globular shape. However, the structural analysis by TEM supports the hydrodynamic data, identifying a predominant single heterodimer population. The larger oligomer classes likely result from nonspecific intermolecular cross-linking of individual protein particles to result in dimers and trimers of the heterodimers.
Analysis of in vivo subunit associations under physiological protein concentrations shows no self-association of either Mus81 or Mms4 under normal growth conditions or in situations of replicative stress. Contrary to our findings, work in mammalian cells has identified coprecipitation of Mus81 subunits using an overexpression transfection system (7). Two possible explanations for the discrepancy are species-specific oligomerization or nonspecific homodimerization as a consequence of overexpression. Biochemical analysis showed that purified full-length human Mus81-Eme1 recombinant protein exhibits little activity on HJs (13), which does not support the notion of species-specific oligomerization. Another possibility is that the interactions observed under overexpressed conditions may not accurately reflect what occurs at physiological protein levels. Other XPF endonuclease family members have been shown to artificially associate when overexpressed or when their obligate partners are not present (15, 18, 25). Our analysis of the association of the Mus81 and Mms4 proteins at their endogenous levels eliminates any concern about artificial dimerization and provides a clearer depiction of what is occurring in vivo.
Four lines of evidence support the model that S. cerevisiae Mus81-Mms4 cleaves DNA junctions with at least one strand interruption but not single or double HJs (see Fig. 17). First, biochemical analyses demonstrate that HJs are an extremely poor substrate for Mus81-Mms4; in fact, they are so poor that even 5′-flaps, which are not considered a Mus81-Mms4 substrate, are cleaved catalytically, whereas HJs are not (21, 23). Second, the biochemical and in vivo analysis presented here demonstrates that Mus81-Mms4 does not exist in higher-order assemblies of the basic heterodimer structure in vitro or in vivo in solution or when bound to DNA. This implies that Mus81-Mms4 is inherently incapable of providing the requisite coordinate cleavage of HJs. Third, even when Mus81-Mms4 is forced into a tetrameric architecture by using a GST tag on Mms4, a fusion protein that is fully functional in vivo (Fig. 7B), the purified enzyme essentially does not cleave HJs (21). Fourth, elegant genetic analysis by Symington and coworkers (28) showed that Mus81-Mms4 largely controls CO formation in vegetative S. cerevisiae cells and that Yen1 primarily controls CO formation in the absence of Mus81-Mms4, such that COs are eliminated in the mus81 yen1 double mutant. Likewise, an additional mutation in YEN1 further sensitizes Mus81- or Mms4-deficient cells to genotoxic stress and exacerbates its defect in sister chromatid recombination (8, 39, 42, 56). To eliminate the possibility that the construct for Mus81-Mms4 purification specifically affects HJ cleavage over 3′-flap activity, we demonstrated that the expression construct shows no synthetic interaction with a yen1 mutation and fully suppresses the mus81 mms4 yen1 triple mutant (Fig. 16).
The virtual absence of a single mutant phenotype for Yen1-deficient cells suggests that in wild-type S. cerevisiae, most COs are controlled by Mus81-Mms4 processing of a nicked joint molecule intermediate with possibly a minor CO pathway through Yen1 involving classic HJs or dHJs (1, 8, 28, 56). Recent results concerning the regulation of Mus81-Mms4 and Yen1 activity in somatic and meiotic cells support this conclusion (40). Consistent with this study (40; S. West, personal communication), we find that Cdc5-mediated activation does not change the substrate specificity of Mus81-Mms4 and that activated Mus81-Mms4 still vastly prefers 3′ flaps (and, presumably, other nicked substrates) over HJs (Fig. 5B and andC).C). In Mus81-Mms4-deficient cells, Yen1 may directly cleave either the Mus81-Mms4 substrates (nicked substrates or D-loops) or their ligated processing products (single or double HJs), although dHJ cleavage has not yet been demonstrated for Yen1. This model is highly consistent with the regulation pattern of Mus81-Mms4 and Yen1, showing that Yen1 is temporarily activated after Mus81-Mms4 (40).
In summary, kinetic (21, 23), structural/biochemical (this study), and genetic (28) data can be synthesized in a highly consistent manner into a model in which Mus81-Mms4 in S. cerevisiae, and likely other eukaryotes, functions to cleave nicked junctions and D-loops to form COs, while Sgs1-Top3-Rmi1 dissolves dHJs into NCO products. This model suggests that the Mus81 pathway of CO formation does not involve classical HJs or dHJs, i.e., junctions with four uninterrupted strands. Nicked-junction cleavage by Mus81-Mms4 provides an efficient pathway for CO formation in vegetative and meiotic cells (Fig. 17) (27, 28, 45). The question of the mechanism of the formation of meiotic COs that are associated with interference is still open, as Mus81-dependent COs lack interference (6, 20, 28, 30, 45, 52).
We thank Mary Ann Osley, Rodney Rothstein, and John Rouse for providing strains; Steve Brill for providing purified Slx1-Slx4; David Toczyski for the Cdc5 constructs; Steve West for his personal communication of unpublished results; Po Lin Chiu, Lenin Dominguez-Ramirez, and David Carlson for sharing their electron microscopy expertise; and Stephen Kowalczykowski, Neil Hunter, Rinti Mukherjee, Clare Fasching, Ryan Janke, Xiao-Ping Zhang, and Jie Liu for critical comments on the manuscript.
This work was supported by grants from the U.S. National Institutes of Health (GM58015 to W.-D.H., U54GM74929 to H.S., and SystemsX.ch to H.S.). E.K.S. was supported by a fellowship from the HHMI-IMBS training grant at UC Davis, and W.D.W. was supported a by TRDRP predoctoral fellowship (17DT-0178).
Published ahead of print 29 May 2012