Holliday junction resolution: regulation in space and time.

Holliday junctions (HJs) can be formed between sister chromatids or homologous chromosomes during the recombinational repair of DNA lesions. A variety of pathways act upon HJs to remove them from DNA, in events that are critical for appropriate chromosome segregation. Despite the identification and characterization of multiple enzymes involved in HJ processing, the cellular mechanisms that regulate and implement pathway usage have only just started to be delineated. A conserved network of core cell-cycle kinases and phosphatases modulate HJ metabolism by exerting spatial and temporal control over the activities of two structure-selective nucleases: yeast Mus81-Mms4 (human MUS81-EME1) and Yen1 (human GEN1). These regulatory cycles operate to establish the sequential activation of HJ processing enzymes, implementing a hierarchy in pathway usage that ensure the elimination of chromosomal interactions which would otherwise interfere with chromosome segregation. Mus81-Mms4/EME1 and Yen1/GEN1 emerge to define a special class of enzymes, evolved to satisfy the cellular need of safeguarding the completion of DNA repair when on the verge of chromosome segregation.

Introduction

Homologous recombination (HR) promotes the establishment of specialized chromosomal interactions that facilitate DNA repair and genome stability during mitotic proliferation and enable genome haploidization during meiosis. Therefore, the controlled engagement and disengagement of all such chromosomal connections is essential throughout the entire eukaryotic life cycle.

In mitotically proliferating cells, the repair of DNA lesions via HR involves pairing and strand-exchange between a damaged chromosome and its sister chromatid (and on rare occasions the homolog) leading to the formation of DNA joint molecules (JMs). JMs are usually disengaged at an early stage by anti-recombinogenic helicases such as Srs2, Mph1 or RTEL1 [1-4]. However, a small proportion mature to form repair intermediates in which the two recombining DNAs are covalently-linked by a four-way junction, known as a Holliday junction (HJ) [5-9]. On those rare occasions when recombination occurs between homologous chromosomes, rather than sister chromatids, it is important that the HJs are processed to give rise to non-crossover (NCO) recombinants, in order to avoid inter-homolog exchanges that can contribute to the onset of cancer by driving loss of heterozygosity of mutated tumor suppressor genes. This situation contrasts with that occurring in germ-line cells undergoing meiosis. Meiotic crossovers (COs) are required for the establishment of cohesin-mediated inter-homolog interactions, which are in most organisms indispensible for the bipolar orientation and segregation of recombined chromosomes of maternal and paternal origin [10,11].

In all cells, HJs provide a physical link between sister chromatids or homologous chromosomes, and a single unprocessed HJ could result in chromosome non-disjunction and aneuploidy, a feature commonly associated with cancer. Hence, despite being generated to help facilitate efficient DNA repair, HJs are perceived as toxic DNA structures as they could interfere with normal chromosome segregation. In the past three years, significant advances have been made that reshape our understanding of how and when these HR intermediates are processed. This review will focus on these new developments and on how they contribute to an updated model of HJ metabolism.

Holliday junction processing enzymes: a brief overview

Enzymes that specifically resolve HJs have been identified from a variety of organisms including bacteriophage, bacteria, archaea, yeast and humans. The prototypic Holliday junction resolvase, E. coli RuvC protein, interacts specifically with HJs and promotes their cleavage by the introduction of symmetrically related nicks in strands located across the junction [12]. The resulting nicked duplex products can be readily ligated to restore the integrity of the DNA [13]. Although there are no sequence or structural homologies at the protein level, S. cerevisiae Yen1, and its human ortholog GEN1, carry out similar biochemical reactions to the bacterial protein [14,15]. However, despite the identification of Yen1/GEN1 as a HJ resolvase, the concept that a single and universal mechanism exists to process HJs has been challenged, leading to the discovery and characterization of several more enzymes capable of HJ processing. Firstly, budding yeast Mus81-Mms4 (human MUS81-EME1), an XPF-family heterodimeric endonuclease, was shown to process a broad range of branched DNA structures, including HJs [16,17]. However, its ability to cut intact HJs was found to be very limited in comparison with other structures, suggesting that a HJ precursor (such as a nicked HJ) might be its preferred DNA substrate [18-20]. To complicate the issue, a second heterodimeric structure-selective endonuclease, yeast Slx1-Slx4 (and its human ortholog SLX1-SLX4), was identified and shown to be capable of severing HJs [21-25]. In contrast to canonical HJ resolvases, however, SLX1-SLX4 is a promiscuous nuclease that nicks a wide variety of DNA secondary structures, and HJ cleavage results in the formation of products that can be only poorly ligated [23,26]. Yen1/GEN1, Mus81-Mms4/EME1 and Slx1-Slx4 all share the common property that HJ resolution gives rise to a mixture of COs and NCOs, depending on the orientation of cleavage (Fig. 1). It therefore seemed unlikely that these enzymes would be involved in the resolution of HJs in eukaryotic cells where non-crossover formation is favored.

Fig. 1

DNA centric view showing three distinct pathways of Holliday junction processing. The BTR complex disengages dHJs, using the mechanism of “dissolution”, to generate NCO recombinants. MUS81-EME1 and SLX1-SLX4 interact to form the SLX-MUS complex which resolves both single and double HJs by endonucleolytic cleavage to generate COs and NCOs. GEN1 provides a separate pathway of HJ resolution.

Meanwhile, two remarkably different mechanisms of HJ processing were uncovered. The BLM helicase (Sgs1 in yeast), which is mutated in the cancer predisposition disorder Bloom's Syndrome (BS), is a component of the BLM-TopoIIIα-RMI1-RMI2 complex (BTR in humans, STR in yeast), which was shown to migrate and decatenate double Holliday junctions (dHJs) [27-29]. This reaction, which results exclusively in the formation of NCO recombinants, and is important in mitotic cells, is defined as HJ ‘dissolution’ rather than ‘resolution’ (Fig. 1). A second novel pathway of resolution was identified in meiotic cells, and this involves an as yet undefined biochemical mechanism that requires Mlh1, Mlh3 and Exo1 [30]. The primary products of this pathway are COs.

The diversity of enzymes and biochemical mechanisms currently implicated in HJ processing highlight the importance placed on ensuring the efficient disengagement of such structures. Furthermore, the possibility of generating specific recombination outcomes (CO vs NCO), depending on the pathway employed for HJ processing (e.g. resolution vs dissolution), begins to explain how mitotic and meiotic cells might process what appear to be related structures and yet generate different recombination outcomes. But in the same light, these studies led to a fundamental question – if cells contain such a range of enzymes that process HJs, how do they implement pathway usage?

Regulation of HJ resolvases: from yeast to man

Yeast cells exhibit a kinetic and genetic separation of CO and NCO recombinant formation during both mitotic and meiotic DSB repair. During mitotic proliferation, the majority of dHJs are processed at early stages of the cell cycle by STR-mediated dissolution, to generate NCOs. In sgs1 mutants, however, JMs persist to a later stage of the cell cycle when they are processed by Mus81-Mms4 or Yen1, to generate a mixture of COs and NCOs [3,31-34] (Fig. 2). These observations led to the concept that the nucleases provide back-up pathways to STR-mediated HJ dissolution. In meiotic cells, most NCO recombinants arise prior to the formation of dHJs, whereas the COs arise at a later time in reactions that are dependent upon Ndt80-mediated transcription of the Polo kinase Cdc5 [35,36].

Fig. 2

Cell-cycle centric view of HJ processing. During DNA repair, single or double HJs (sHJ or dHJ) are formed between sister chromatids or homologous chromosomes (for convenience, the diagram depicts two homologs). The STR (yeast) or BTR (human) complexes operate early during the cell cycle to dissolve dHJs and generate non-crossover recombinants. If DNA damage occurs late during the cell cycle or if HJs escape the attention of STR/BTR and persist until S/G2 phase, resolution enzymes ensure HJ processing and safeguard chromosome segregation.

In human cells, the temporal separation of CO and NCO formation has yet to be demonstrated. However, as observed in yeast, BS cells that lack BLM display an increased frequency of CO formation, visualized as sister chromatid exchanges [37]. In these cells, CO formation is largely dependent on the actions of MUS81-EME1, SLX1-SLX4 and GEN1 [26,38]. These observations indicate that despite having a variety of enzymes capable of HJ resolution, pathway usage in mitotic cells is biased to favor STR/BTR-mediated dHJ dissolution and NCO formation. While it is possible that cells utilize more than one strategy to direct pathway usage and control the outcome of recombination, recent advances have uncovered a number of mechanisms by which cells control the actions of these structure-selective nucleases. Moreover, the realization that these enzymes are regulated now provides the key to understand several surprising differences between the genetics and biochemistry of HJ resolution.

Regulation of Mus81-Mms4 and Yen1 in S. cerevisiae

Several recent studies have uncovered the elegant modes of regulation of the Mus81-Mms4 and Yen1 nucleases [32-34,39]. Remarkably, both enzymes are controlled by cell cycle stage-specific phosphorylation events that impose temporal control to their actions and lead to their sequential activation.

In mitotic cells, the activity of Mus81-Mms4 is low during S-phase, but the protein is activated at the onset of mitosis by the collaborative actions of two cell cycle kinases: Cdc28/CDK and Cdc5/PLK1 [32-34,39]. Indeed, Mus81-Mms4 activity displays a cyclic behavior that mirrors and depends upon the modulation of the levels of Cdc5 transcription and proteolysis (Figs. 2 and 3). Precisely how Cdk/Cdc5-mediated phosphorylation activates Mus81-Mms4 is presently unknown, but phosphorylation of multiple residues in the N-terminal region of the non-catalytic subunit, Mms4, enhances the nuclease activity of its partner, Mus81. A similar regulatory network controls Mus81-Mms4 activity during meiosis. Cdc28/CDK and Cdc5/PLK1-mediated phosphorylation of Mms4 activates Mus81-Mms4 at the onset of the first meiotic division to promote the efficient elimination of JMs, leading to the formation of both CO and NCO recombinants [32].

In contrast to Mus81-Mms4, Yen1 nuclease activity is tightly inhibited by extensive phosphorylation events that occur at the G1/S transition, and the nuclease remains inactive through S-phase and G2. It is then activated by stage-specific dephosphorylation at the later stages of mitosis [32]. Analysis of Yen1 regulation revealed a dual inhibitory mechanism [40,41]. First, Cdk-mediated phosphorylation of S679 (located within the nuclear localization signal: NLS) inactivates the NLS and prevents nuclear import of the protein, leading to the depletion of Yen1 from the nucleus during early stages of the cell cycle [42]. Second, phosphorylation directly reduces Yen1's DNA binding affinity and nuclease activity [40]. Activation of Yen1 occurs as cells enter anaphase, when Cdc14 phosphatase promotes its rapid de-phosphorylation, converting it into a fully functional nuclease that re-enters the nucleus [40,41]. Therefore, the consecutive but temporally separable waves of HJ processing, by Mus81-Mms4 and Yen1, peak at metaphase and anaphase, respectively. As such, the two enzymes complement each other in imposing active HJ resolution at the onset and throughout all stages of chromosome segregation (Figs. 2 and 3). A similar dual control mechanism occurs in meiotic cells where cell cycle stage-specific phosphorylation and dephosphorylation events separate the function of Mus81-Mms4 and Yen1 in time, limiting their ability to compete for their DNA substrates [32].

Regulation of Mus81-Eme1 in S. pombe

Treatment of fission yeast with the DNA topoisomerase I inhibitor camptothecin leads to Rad3 and Chk1-dependent phosphorylation of Eme1. Phosphorylated Mus81-Eme1 acquires an increased ability to process DNA substrates, including HJs, indicating that S. pombe Mus81-Eme1 is up-regulated in response to DNA damage [43]. This regulation strategy contrasts with that present in budding yeast, in which Mus81-Mms4 phospho-activation is inhibited by DNA-damaging agents that delay Cdk and Cdc5 expression/activation and cell cycle progression [33,34,44]. However, it appears that S. pombe Eme1 is also modified in a cell cycle stage-dependent manner, with phosphorylation peaking after S-phase. Again, the cell cycle-dependent modification of Eme1 requires Cdc2/Cdk1 activity, which is also necessary for the DNA damage-dependent phospho-activation of Mus81-Eme1. These results implicate Cdk1 as a key regulator of Mus81-Eme1 activation in S. pombe. However, instead of priming Eme1 for further Plo1/Cdc5/PLK1 phosphorylation, Cdk1 appears to prime Eme1 for Chk1-dependent modification in response to DNA damage [43].

The metabolism of HJs in fission yeast is significantly different from that of budding yeast. For example, S. pombe cells lack a clear Yen1/GEN1 homolog [14]. Therefore, the differences observed in Mus81-Mms4/Eme1 regulation are likely to reflect a much stronger dependence on Mus81-Eme1 for efficient mitotic and meiotic JM processing [16].

Regulation of human MUS81-EME1 and GEN1

The work described above provided our first insights into how yeast cells regulate the activities of Mus81-Mms4 and Yen1. Parallel studies carried out in human cells support the notion that similar principles operate to regulate the activities of their orthologs in higher eukaryotes. GEN1, the ortholog of Yen1, is excluded from the nucleus during the early stages of the cell cycle [32]. Only upon nuclear envelope break-down, which is mediated by CDK1 activity [45], does GEN1 nuclease gain access to DNA. Therefore, nuclear exclusion during S- and G2-phase is likely to spatially and temporally limit the ability of GEN1 to access chromatin and process HJs (Fig. 3) [32]. But precisely how the nuclear import/export of GEN1 is regulated remains an important question for future research.

Fig. 3

Spatio-temporal regulation of HJ processing enzymes. In S. cerevisiae, cell cycle kinases Cdk and Cdc5 transiently phosphorylate and activate Mus81-Mms4 at the G2/M transition. Cdk1-mediated phosphorylation of Yen1 prevents its nuclear accumulation and biochemical activation during S-phase and early stages of mitosis. Cyclin degradation and Cdc14-mediated dephosphorylation activate Yen1 during anaphase. In human cells, a similar regulatory network uses a different molecular mechanism to prevent the actions of MUS81-EME1 and GEN1 during S-phase and early G2. At the G2/M transition, M-phase CDK activity promotes the interaction of MUS81-EME1 with SLX1-SLX4 leading to the formation of an active SLX-MUS HJ resolvase. GEN1, which is excluded from the nucleus during early stages of the cell cycle, gains access to chromatin upon CDK1-mediated nuclear envelope breakdown.

MUS81-EME1, the ortholog of Mus81-Mms4, is phosphorylated at the G2/M transition and, as observed in yeast, this modification is accompanied by an increase in its ability to process HJs [32]. Remarkably, however, human cells appear to have evolved a different molecular strategy to enhance its ability to process HJs. CDK1 activity, which peaks at the onset of mitosis, triggers the cell cycle stage-specific association of MUS81-EME1 with a second structure-selective endonuclease, SLX1-SLX4, to form a SLX-MUS holoenzyme [26]. By coordinating the active sites of SLX1-SLX4 (a potent nickase) with that of MUS81-EME1 (which efficiently cleaves nicked HJs), the SLX4 scaffold provides a platform for HJ processing [26,46,47].

The mechanism by which CDK1 activity enhances the ability of MUS81 to interact with SLX4 has yet to be determined. However, MUS81, EME1 and SLX4 are all phosphorylated in a CDK1-dependent manner, suggesting that the modifications might directly stabilize the SLX-MUS complex. PLK1 was also found to interact with SLX4 [25], but its activity appears to be dispensable for the bulk of SLX-MUS complex formation at the G2/M transition [26]. Future work will be required to determine whether PLK1 activity, or interaction, with SLX4 is required for full SLX-MUS functions in vivo.

Cell cycle control of HJ processing enzymes: how does it work and what's the point?

The reasoning behind the regulation of structure-selective nucleases is only now beginning to be appreciated. Interference with the normal ability of cells to activate Mus81-Mms4 at the G2/M transition makes budding yeast sensitive to DNA damaging agents, and also impairs proliferation in the absence of STR-mediated dHJ dissolution. Furthermore, in the absence of STR, Mus81-Mms4 activation is critical for timely anaphase initiation and chromosome segregation [32-34]. Therefore, during mitotic proliferation, the activation of Mus81-Mms4 at the G2/M transition provides an essential backup to STR-mediated HJ dissolution and chromosome disengagement. Additionally, phospho-activated Mus81-Mms4 is likely to provide the primary pathway for the timely elimination of single HJs, which form during DNA repair and are refractory to STR.

The cellular strategy used to activate Mus81-Mms4 at the G2/M transition, which depends on the accumulation of Cdc5 and M-phase cyclins, creates a limited temporal window for its function. Because cyclins and Cdc5 are degraded to initiate anaphase, Mus81-Mms4 is inactivated shortly after cells begin chromosome segregation. To extend the function of Mus81-Mms4 to anaphase, cells up-regulate Yen1 using a common regulator, Cdk1, but with the reverse strategy. As described above, Cdk1-mediated phosphorylation inhibits Yen1 nuclease during S-phase and the early stages of mitosis, but cyclin degradation at anaphase turns Yen1 on at the same time Mus81-Mms4 is switched off. As for many Cdk1 substrates, Yen1 activation is enhanced by targeted dephosphorylation mediated by the Cdc14 phosphatase. This strategy creates two consecutive and interlinked waves of HJ resolution that cover all stages of chromosome segregation (Figs. 2 and 3).

It is striking that the only cell cycle stage that lacks fully active Mus81-Mms4 and Yen1 is S-phase, indicating that it may be important to down-regulate the activity of these resolvases at the time when DNA is being replicated. In support of this notion, premature activation of Mus81-Mms4 and Yen1 interferes with the normal outcome of DSB repair by increasing the frequency of mitotic CO recombinants [32-34,40]. It is therefore likely that the late activation of Mus81-Mms4 and Yen1 not only serves to ensure the late elimination of recombination intermediates, but also favors STR-mediated HJ dissolution during the early stages of the cell cycle (Figs. 2 and 3).

In human cells, limiting mitotic COs is thought to be very important for impeding loss of heterozygosity (LOH), which is associated with the loss of tumor supressor genes during cancer development. But why yeast cells should have developed such a regulatory mechanism to control Mus81-Mms4 and Yen1 during the early stages of the cell cycle is more difficult to explain. Two alternative hypothesis might be considered: firstly, that the regulation of Mus81-Mms4 and Yen1 is a remnant of their functions in meiosis, where dHJ processing to generate COs is essential for genome haploidization. Control over their activities might serve to prevent the processing of recombination intermediates that are NCO or CO-designated. Since Mus81-Mms4 and Yen1 would generate a mixture of COs and NCOs, meiotic cells may have first gained control over their actions to be capable of tightly programming CO formation and distribution. Secondly, it is possible that the tight control of these structure-selective nucleases relates to their limited substrate specificity. Although this review does not focus on the list of potential DNA substrates of Mus81-Mms4 and Yen1, both nucleases are capable of processing model replication forks, and other DNA intermediates in vitro. It is likely that the restraint of their activities during DNA replication prevents the toxic processing of vital DNA replication and repair intermediates.

Future work will be required to further refine the mechanistic details and relevance of MUS81-EME1 and GEN1 regulation in higher eukaryotes. However, the evidence that divergent species take advantage of a similar strategy to control pathway usage is remarkable (Figs. 2 and 3). Such a conserved logic suggests that HJ resolution has evolved to become more than just the very last step in recombinational DNA repair. Indeed, the regulation of these enzymes provides a novel cell cycle dimension to the completion of homologous recombination, in that they have become key targets of the mitotic and meiotic cell cycle machinery, coupling the completion of DNA repair to the initiation of chromosome segregation.