Linking the enzymes that unlink DNA

Abstract

Two papers in this issue, Castor et al. (2013) and Wyatt et al. (2013) and a third in Cell Reports, Garner et al. (2013) demonstrate that the scaffold protein SLX4 coordinates multiple nucleases in order to effectively resolve Holliday Junctions and repair interstrand crosslinks (ICLs) in mammalian cells.

The central intermediate of homologous recombination (HR) is the Holliday Junction (HJ). Located at the nexus between homologous duplexes, this four-way structure allows the exchange of genetic material by nucleases that generate either crossover (CO) or non-crossover (NCO) products. In bacteria, this is carried out elegantly by a structure-specific nuclease, RuvC, that makes symmetrical nicks across the junction (Figure). The beauty of this reaction, termed HJ resolution, is that each of the two linear products requires only the ligation of a single nick to complete the repair. For a variety of reasons, it has been challenging to identify the eukaryotic counterpart to RuvC. Biochemical assays currently implicate three enzymes in this process: MUS81-EME1, SLX1-SLX4, and GEN1. However, only GEN1 (Yen1 in yeast) appears to function like a canonical resolvase. Now, three independent reports from the Rouse, West, and Smogorzewska labs, present compelling evidence that the majority of mitotic crossovers in mammalian cells occurs by the coordinated activities of MUS81-EME1 and SLX1-SLX4 (Castor et al., 2013; Wyatt et al., 2013; Garner et al., 2013).

Figure 1.

Pathways for HJ Resolution

(A) Resolvases RuvC and GEN1 digest HJs to generate linear products with single nicks (carets). SLX1-SLX4 nuclease nicks HJs that can be used by MUS81-EME to generate linear products containing a flap or gap. Colored dots represent 5′-ends and dotted arrows represent nuclease cuts.

(B) The BTR pathway for dHJ dissolution generates NCOs exclusively. Arrowheads represent BLM helicase.

(C) In the absence of BTR, two pathways can act on dHJs or sHJs in mammalian cells. Epistasis analysis places SLX1-SLX4 and MUS81-EME1 in one pathway, while GEN1 defines the second. The action of these nucleases can lead to CO or NCO outcomes depending on the direction of cleavage which is assumed to be random.

(D) Coordinated repair of ICLs by SLX4-bound nucleases. DNA replication produces a MUS81-dependent DSB which is postulated to occur on the leading strand template. This is followed by unhooking of the ICL by various candidate proteins including XPF, SLX1, mismatch repair and Fanconi Anemia proteins (not shown) and filling-in by translesion DNA synthesis (TLS). Finally, replication restart via break-induced replication creates a nHJ that can be cleaved by MUS81 using the vertical cleavage site to produce an SCE. Alternatively, branch migration of the nHJ produces an intact HJ that could be cleaved by MUS81 and SLX1 to produce an SCE or a NCO (not shown).

An important feature of eukaryotic double-strand break (DSB) repair is that recombination between sister chromatids or homologous chromosomes proceeds through double HJ (dHJ) intermediates. Most dHJs are resolved as NCOs by the BLMTOP3-RMI1-RMI2 (BTR) complex (Sgs1-Top3-Rmi1 in yeast), which minimizes loss of heterozygosity. In this nonnucleolytic reaction, termed dHJ dissolution, the BLM DNA helicase promotes convergent branch migration of the HJs while TOP3 decatenates the intervening strands (Wu and Hickson, 2003). The hallmark of Bloom Syndrome (BS) cells, which lack BLM, is excessive sister chromatid exchange (SCE). Thus, loss of BLM reveals a critical need for HJ resolvases.

This is where MUS81-EME1 (here called MUS81) and SLX1-SLX4 come in. Although both complexes were identified in yeast based on their requirement for the viability of cells lacking the BLM ortholog Sgs1 (Mullen et al., 2001), proving their role in HJ resolution has been a tangled saga. Initial support for the idea came from the observation that meiotic COs in S. pombe depend on MUS81 (Osman et al., 2003; Smith et al., 2003). Further, in budding yeast, mitotic COs depend largely on Mus81 with Yen1 playing a back-up role (Ho et al., 2010). What has been impossible to reconcile with these strong genetic results is the in vitro activity of MUS81. Purified MUS81 is inactive on intact HJs but is extremely active on nicked HJs (nHJs). The nick directs Mus81 cleavage to the opposing strand where, unlike canonical resolvases, the two products contain either a 5′-flap or small gap. A solution to this dilemma is a nuclease to first nick the HJ.

One candidate for this nickase is SLX1-SLX4 which, in yeast, nicks HJs with little evidence of symmetrical cutting. The notion that SLX1-SLX4 contributes to resolvase activity was confirmed by the identification of SLX4 in humans and in Drosophila where it was found to be identical to MUS312, a factor required for meiotic COs (summarized in Klein and Symington 2009). Like the yeast protein, human SLX4 acts as a scaffold to bind SLX1 and XPF-ERCC1 (Rad1-Rad10 in yeast). However, human SLX4 also binds MUS81. Because distinct regions of SLX4 mediate each of these interactions, mutant SLX4 scaffolds can be used to determine the functional requirement of each interaction.

Three groups have now tested these genetic dependencies in mammalian cells. Using a variety of cell knock-outs and siRNA knock-downs, it was found that cells deficient in BLM require SLX1, SLX4, and MUS81 for cell viability (Garner et al., 2013; Wyatt et al., 2013), and the excess SCEs characteristic of BS (Castor et al., 2013; Garner et al., 2013; Wyatt et al., 2013). The target of these nucleases appears to be HJs since the SCEs were sensitive to a single amino acid change that eliminates SLX1 nuclease activity and an archeal resolvase complements this mutation (Castor et al., 2013). As in yeast, GEN1 plays a role but it is less important than the other factors. Importantly, the viability and excess SCEs of cells depleted for BLM require both the SLX1- and MUS81-interacting regions of SLX4 (but not the XPF-interacting region) (Castor et al., 2013; Garner et al., 2013), indicating that HJ resolution requires the coordination of these factors as previously suggested (Svendsen et al., 2009). Consistent with the idea that they work in the same pathway, mutations in MUS81 and SLX1-SLX4 were found to be epistatic with respect to generating SCEs (Castor et al., 2013; Wyatt et al., 2013).

To test the idea that the nucleases cooperate in vitro, the West group purified full-length human SLX1-SLX4 dimer. HJs treated with SLX1-SLX4 generate products that are poorly ligated indicating that SLX1-SLX4 is not a canonical resolvase and that it nicks HJs somewhat indiscriminately – which is just the activity required to activate MUS81. Indeed, SLX1-SLX4 stimulated MUS81 cleavage of HJs and additional evidence suggests that they act in unison as opposed to cleaving the HJ sequentially (Wyatt et al., 2013).

The current results raise the question as to why eukaryotes would resolve mitotic HJs with a complex of structure-selective nucleases instead of a classic resolvase. One possibility is that the SLX4 complex doubles as a toolkit for the repair of ICLs; cells lacking MUS81, SLX1, XPF or SLX4 are hypersensitive to a variety of interstrand crosslinkers (Castor et al., 2013; Dendouga et al., 2005; Garner et al., 2013; Wyatt et al., 2013). The pathway for ICL repair is not well understood, but it is known that ICLs generate SCEs during DNA replication. In human cells these SCEs were found to be dependent on SLX1 and MUS81 (Garner et al., 2013; Wyatt et al., 2013). However, in murine cells it appears that the essential targets of SLX1-SLX4 and MUS81 in ICL repair are structures other than HJs (Castor et al., 2013). If this is true more broadly, it may be more accurate to refer to these enzymes as structure-selective nucleases, as opposed to resolvases, given that they are not only mechanistically, but functionally distinct from RuvC.

Together, these studies provide insight into how the coordination of multiple nucleases by SLX4 can effectively resolve HJs and repair ICLs. However, important questions remain. For example, is the directionality of HJ cleavage regulated? Do the nucleases process dHJs differently than single HJs (sHJs)? What enzyme cleaves the 5′-flap from Mus81-cleaved HJs? Obvious candidates for this reaction include XPF and SLX1. The recombination field is advancing rapidly and answers to these questions should not be long in coming.