Biochemical Society Transactions

Colworth Medal Lecture

Control of genome stability by Slx protein complexes

John Rouse


The six Saccharomyces cerevisiae SLX genes were identified in a screen for factors required for the viability of cells lacking Sgs1, a member of the RecQ helicase family involved in processing stalled replisomes and in the maintenance of genome stability. The six SLX gene products form three distinct heterodimeric complexes, and all three have catalytic activity. Slx3–Slx2 (also known as Mus81–Mms4) and Slx1–Slx4 are both heterodimeric endonucleases with a marked specificity for branched replication fork-like DNA species, whereas Slx5–Slx8 is a SUMO (small ubiquitin-related modifier)-targeted E3 ubiquitin ligase. All three complexes play important, but distinct, roles in different aspects of the cellular response to DNA damage and perturbed DNA replication. Slx4 interacts physically not only with Slx1, but also with Rad1–Rad10 [XPF (xeroderma pigmentosum complementation group F)–ERCC1 (excision repair cross-complementing 1) in humans], another structure-specific endonuclease that participates in the repair of UV-induced DNA damage and in a subpathway of recombinational DNA DSB (double-strand break) repair. Curiously, Slx4 is essential for repair of DSBs by Rad1–Rad10, but is not required for repair of UV damage. Slx4 also promotes cellular resistance to DNA-alkylating agents that block the progression of replisomes during DNA replication, by facilitating the error-free mode of lesion bypass. This does not require Slx1 or Rad1–Rad10, and so Slx4 has several distinct roles in protecting genome stability. In the present article, I provide an overview of our current understanding of the cellular roles of the Slx proteins, paying particular attention to the advances that have been made in understanding the cellular roles of Slx4. In particular, protein–protein interactions and underlying molecular mechanisms are discussed and I draw attention to the many questions that have yet to be answered.

  • DNA damage
  • Mec1
  • Rad1
  • Sgs1
  • Slx
  • Tel1
  • xeroderma pigmentosum complementation group F (XPF)

Introduction: the discovery of the SLX genes

Members of the RecQ family of DNA helicases are found in all kingdoms of life because of the vitally important roles that they play in protecting genome stability, especially during DNA replication [1,2]. RecQ proteins are structure-specific DNA helicases that display 3′→5′ directionality, and there are five of them in human cells: WRN (Werner's syndrome protein), BLM (Bloom's syndrome protein), RecQ1, RecQ4 and RecQ5 [2]. Mutations in WRN cause WS (Werner's syndrome), a rare recessive genetic disorder [3], whereas mutations in RecQ4 cause RTS (Rothmund–Thompson syndrome), RAPADILINO [4] or Baller–Gerrold syn-dromes [5]. WS and RTS are characterized by premature aging. Mutations in BLM cause BS (Bloom's syndrome), a rare recessive disorder characterized by strong predisposition to a range of cancers and premature aging [6,7]. Cells from BS patients show a high degree of genomic instability [7], a high frequency of SCEs (sister-chromatid exchanges) [9] and a high degree of sensitivity to DNA-damaging agents and inhibitors of DNA replication [1]. Little is known about RecQ1, RecQ4 or RecQ5, and it is not clear whether the five human RecQ helicases play complementary or independent cellular roles.

RecQ helicases have been implicated in HR (homologous recombination), an ancient mechanism for repairing DSBs (double-strand breaks) (Figure 1) and for repairing broken or collapsed replication forks (Figure 2) [10,11]. There are three major pathways for repair of DSBs by HR: gene conversion, BIR (break-induced replication) and SSA (single-strand annealing). BIR is a mode of HR used when only one end of a DSB is free, such as when replisomes collapse (Figure 2); SSA is used to repair DSBs that occur between repeated sequences and is described below. There are three models for gene conversion: the DSB repair model (also known as the Szostak model), SDSA (synthesis-dependent strand annealing) and double HJ (Holliday junction) dissolution (Figure 1) (reviewed in depth in [1015]). In all three models, the 5′-ends at the break are resected, leaving 3′-single-stranded overhangs that are coated with Rad51 to form a nucleoprotein filament. This is assisted by many proteins, including RPA (replication protein A), Rad52, Rad54, Rad55 and Rad57 [1015]. These overhangs actively search for the homologous sequences, and one of them invades the homologous DNA duplex. This acts as a primer for DNA synthesis, resulting in primer extension (Figure 1). In the SDSA model, the extended invading strand is displaced from the template and anneals to the other 3′-end of the DSB, leading to the formation of only non-crossover products. Otherwise, the other 3′-end of the DSB is captured (second end capture). DNA synthesis and ligation of nicks then leads to the formation of double HJs (Figure 1). In the DSB-repair model of Szostak et al. [16], HJs are resolved (the identities of the enzymes responsible remain elusive) by cutting the crossed (closed arrowheads) or non-crossed (open arrowheads) strands of the junction. This can result in crossover or non-crossover products, depending on the orientation of HJ cleavage. In the dissolution model, Sgs1–Top3 (DNA topoisomerase III; BLM–TOP3α in humans) dissolves double HJs by converging them, and this leads to non-crossover products only (Figure 1).

Figure 1 Repair of DSBs by HR

In this schematic diagram, two identical DNA duplexes are depicted in red and blue. In the DSB-repair model of gene conversion proposed Szostak et al. [16], DSBs are resected in the 5′–3′ direction. The resulting 3′-ends invade the homologous template and act as primers for new DNA synthesis. A double HJ is formed and each junction is resolved by cutting the crossed (closed arrowheads) or non-crossed (open arrowheads) strands of the junction. This can result in crossover or non-crossover products, depending on the orientation of HJ cleavage. In the SDSA model shown, both 3′-ends invade the homologous template. After extension, the original strands of the broken template reanneal, and DNA synthesis is completed. DNA repair is completed by filling in and ligation. Sgs1–Top3 is capable of dissolving double HJs, and this yields non-crossover products.

Figure 2 Replisome stalling can lead to DSBs

Schematic representation of the effects of DNA damage on replication forks. (A) Replication through a nick in the parent template leads to replisome collapse. (B) A DNA lesion (alkylated base, for example) in the parental DNA template blocks replisome progression. Cleavage of the resulting ssDNA gap causes replisome collapse. (C) It has been proposed that stalled replisomes can regress to form a ‘chicken foot’ structure. In all three cases, replisome collapse is accompanied by the generation of a one-ended DSB (depicted by broken lines) that can initiate HR-mediated regeneration of an intact replication fork, enabling resumption of DNA replication.

DNA lesions in S-phase are dangerous since they can block replisome progression. If such blocks were not overcome, then cell division could not continue or, if it did, the two daughter cells would not have the correct complement of chromosomes, resulting in aneuploidy or death. DNA damage can lead to replisome stalling or replisome collapse, depending on the nature of the lesion (Figure 2B) (reviewed in [17,18]). For example, replication through nicks in the parental template leads to replisome collapse (Figure 2A). Nuclease attack of ssDNA (single-stranded DNA) formed when replisomes stall at obstacles can also lead to replisome collapse (Figure 2B). On the basis of studies in bacteria (reviewed in [17]), it has been proposed that blocked replisomes can regress and this involves reannealing of the parental strands in the stalled replisome, causing the released nascent strands to anneal also, and the resulting structure is referred to as a ‘chicken foot’ (Figure 2C). Cleavage of these structures would lead to replisome collapse, and the ends generated could initiate HR-mediated regeneration of the replication fork (Figure 2C). A comprehensive description of the causes and consequences of replisome stalling/collapse is beyond the scope of this review, and the reader is referred to excellent recent reviews on this topic [18,19].

HR is important for the regeneration of intact replication forks to enable resumption of DNA replication following replisome arrest in prokaryotes [17] and in eukaryotes [2123]. This is often initiated by the invasion of the free DNA end created when replisomes collapse, although recombination can probably also initiate at single-stranded gaps at replisomes that have stalled, but that have not collapsed. Recombination-deficient Saccharomyces cerevisiae mutants are highly sensitive to agents that arrest replication forks [25,26], and agents that block or stall replisomes trigger a pronounced increase in SCE [22]. Furthermore, mutants that exhibit defective DNA synthesis display a hyper-recombination phenotype and synthetic lethality with the RAD52 epistasis group [27]. The mechanisms of HR-mediated rescue of replication forks is understood in detail in prokaryotes. However, even though many genes in eukaryotes have been described that are required for HR-mediated rescue of blocked/collapsed replisomes, in addition to the core recombination proteins, our knowledge of the underlying DNA transactions and mechanisms is startlingly poor.

The phenotypes of cells lacking RecQ helicases reflect the roles of these important enzymes in promoting genome stability primarily during DNA replication [10]. Cells lacking these enzymes are hypersensitive to inhibitors of DNA replication, and show defects in S-phase progression even in the absence of DNA damage [10,29]. Mutant cells also accumulate aberrant DNA structures even in an unperturbed S-phase, and yet more aberrant species arise when DNA damage occurs during this time [30]. Taken together with the observation that several RecQ helicases localize at replication forks when replisomes are blocked, these data argue convincingly that RecQ helicases are vital for protecting genome stability during DNA replication. Yeasts have a single RecQ helicase: Sgs1 in S. cerevisiae and Rqh1 in Schizosaccharomyces pombe [31,32]. BLM, Sgs1 and Rqh1 interact genetically and physically with Top3 (and Rmi1) [33,34]. Yeast lacking Top3 show a slow-growth phenotype, together with high levels of recombination and chromosomal rearrangements [35,36]. However, deletion of SGS1 in top3Δ cells rescues the slow growth [34], suggesting that, at stalled replisomes, the helicase activity of Sgs1 creates intermediate DNA structures that normally require resolution by Top3. Several other lines of evidence indicate that RecQ helicases play roles in HR. Top3 can break and rejoin DNA to alter its topology, and BLM–Top3α, together with Rmi1/BLAP75 (BLM-associated protein of 75 kDa) has been shown to dissolve recombination intermediates containing double HJs (Figure 1) [37]. This mechanism of double-HJ dissolution prevents exchange of flanking sequences and thus loss of heterozygosity and/or crossing over [38]. Consequently cells lacking BLM show elevated rates of SCE [37].

Despite the important protective roles RecQ helicases play during DNA replication, RecQ helicases are not essential for cell viability. This suggests that there are back-up pathways that function redundantly with RecQ enzymes and this prompted Steve Brill's laboratory to look for genes that function in a redundant manner with Sgs1 in S. cerevisiae [39]. Six genes were found that are required for viability in cells lacking SGS1, but not in wild-type cells, and these were named SLX genes for ‘synthetic lethal of unknown function’ [39]. The same study showed that the Slx gene products assemble into three separate heterodimeric complexes: Slx3–Slx2, Slx1–Slx4 and Slx5–Slx8 (Figure 3). The three Slx protein complexes are phenotypically distinct in terms of genotoxin hypersensitivity, morphology and meiotic defects, but mutations in any of the six SLX genes are lethal in cells expressing a helicase-dead Sgs1 mutant or in cells lacking Top3 [39]. This strongly suggested that the Slx complexes were involved in processing toxic substrates that accumulate when the Sgs1–Top3 complex is non-functional and it seem likely that these substrates would occur at stalled replication forks.

Figure 3 The S. cerevisiae Slx protein complexes

Schematic representation of the Slx protein complexes and the domain organization of the Slx proteins. Red rectangles in Mus81 and Mms4 represent HhH motifs, whereas dark grey rectangles are pseudo-HhH motifs. The total number of amino acids in each protein is shown on the right hand side.

The Slx3–Slx2 complex is the structure-specific endonuclease Mus81–Mms4Eme1

Mus81 was reported as a protein that interacts with the Cds1CHK2 checkpoint kinase in S. pombe and with the Rad54 recombinase in S. cerevisiae [40,41]. Cells lacking Mus81 have a range of interesting defects, including hypersensitivity to agents that impede replication fork progression and a pronounced defect in the production of viable meiotic spores, a process that relies on HJ resolution [40,41]. It soon became obvious that Mus81 corresponds to Slx3 and that Slx2 corresponds to a protein called Mms4 in S. cerevisiae and Eme1 in S. pombe (EME1 and EME2 in humans). It was subsequently reported by several groups that Mus81–Mms4Eme1 from S. cerevisiae, S. pombe and humans is a structure-specific endonuclease with marked preference for branched DNA structures (Figure 4) [4244]. Mus81 orthologues belong to the MUS81/XPF (xeroderma pigmentosum complementation group F) family of nucleases [45]. Most eukaryotes have at least four members of this family that assemble into two heterodimeric complexes, each consisting of a catalytic and non-catalytic subunit. In S. cerevisiae, there are two such endonuclease complexes, Rad1–Rad10 and Mus81–Mms4, whereas humans have three, XPF–ERCC1 (excision repair cross-complementing 1) (orthologous to Rad1–Rad10), MUS81–EME1 and MUS81–EME2 (catalytic subunits are given first). The non-catalytic subunits have diverged from their catalytic counterparts and lack key residues required for endonuclease activity. Mus81 and the MUS81/XPF have been the subject of excellent recent reviews to which readers are referred [45,46].

Figure 4 Substrate specificity of the S. cerevisiae Slx1–Slx4, Mus81–Mms4 and Rad1–Rad10 endonucleases

Schematic diagram illustrating some of the substrates cleaved by these nucleases in vitro. The 5′-end strand in each structure is shown for the sake of orientation.

In vitro, Mus81–Mms4 cleaves branched DNA structures that resemble replication forks, but exhibits a marked preference for 3′-branches (Figure 4) [44,47]. This is consistent with the hypersensitivity of cells lacking Mus81 to agents that impair the progression of replication forks [40,47], and it is likely that Mus81 cleaves DNA in the vicinity of stalled forks to enable replication restart. In human cells, for example, it is clear that MUS81 cleaves replication forks blocked by interstrand cross-links, generating a free DNA end that can initiate regeneration of an intact replication fork by HR [48,49]. Mus81 can also cleave D-loops and 3′-flaps (Figure 4), and these structures are likely to occur late in the process of HR-mediated repair of collapsed or broken replication forks. How does this nucleolytic function of Mus81 relate to the inviability of mus81Δ sgs1Δ cells? An important clue comes from the observation that deletion of RAD51 rescues the lethality of mus81Δ sgs1Δ cells [47]. This observation suggests that Sgs1 and Mus81 are only required after Rad51 has acted to create recombination intermediates and that the helicase activity Sgs1 and the nuclease activity of Mus81 are vital for the repair of these intermediates. It appears therefore that Sgs1–Top3 co-operates with Mus81–Mms4 to rescue collapsed replication forks or to cleave 3′-flaps during HR-mediated repair of stalled or collapsed forks. More work is needed to identify the DNA substrates acted on by Sgs1–Top3 and Mus81–Mms4 in vivo that enables DNA replication to continue in the face of lesions that block or collapse replisomes.

The Slx5–Slx8 complex is a SUMO-targeted ubiquitin ligase

The covalent attachment of one or more ubiquitin monomers to target proteins requires the sequential action of ubiquitin-activating enzymes (E1), Ubcs (ubiquitin-conjugating enzymes; E2) and ubiquitin ligases (E3). After addition of a single ubiquitin moiety, further ubiquitins can be added to the first, resulting in a polyubiquitin chain [50]. Ubiquitin has seven lysine residues to which other ubiquitins can be attached, and the formation of Lys48-linked ubiquitin chains, for example, causes proteins to be targeted for degradation. A range of UBLs (ubiquitin-like modifiers) have been identified, including SUMO (small ubiquitin-related modifier) (reviewed in [51,52]), referred to as Smt3 in yeast. SUMOylation can alter protein localization, function or half-life. In 2007, a flurry of papers appeared implicating Slx5–Slx8 (Figure 3) in the control of protein ubiquitination. Slx5–Slx8 is active as an E3 ubiquitin ligase in vitro [53], and both proteins have a RING finger (Figure 3) that is often found in E3 ligases. The E3 ligase activity of the complex appears to be mediated by the RING finger of Slx8, although its activity is strongly enhanced by Slx5 [54]. An important clue to the function of this complex came from the observations that overexpression of Slx5 suppresses the temperature-sensitivity of cells expressing a mutant form of the SUMO protease Ulp1 in which high-molecular-mass SUMO conjugates accumulate [54]. This indicated that Slx5–Slx8 may be responsible for degrading SUMOylated proteins and, consistent with this, high-molecular-mass SUMO conjugates accumulate to a high level in cells lacking Slx5 or Slx8 [5457]. Work from several laboratories has showed that Slx5 binds to SUMO through several SIMs (SUMO-interaction motifs) and this enables the Slx5–Slx8 complex to dock on to SUMOylated proteins [5457]. So it appears that the Slx5–Slx8 complex binds to SUMOylated proteins and then catalyses the addition of ubiquitin to the end of the SUMO chains, thereby targeting SUMOylated proteins for degradation. At present, it is not clear whether the addition of one ubiquitin to the end of poly-SUMO chains is sufficient to target the protein for degradation or whether a polyubiquitin chain is necessary.

The Slx5–Slx8 complex has multiple cellular functions. It is involved in transcriptional silencing at multiple loci, for example [58]. In terms of genome stability, Slx5–Slx8 is required not only for cell viability in the absence of Sgs1, but also when cells are exposed to agents that slow down DNA replication, such as hydroxyurea [39]. Mutations in the RING fingers of either Slx5 or Slx8 that weaken the ubiquitin ligase activity of the complex severely reduce the ability of each protein to rescue the hydroxyurea-sensitivity of slx5Δ and slx8Δ cells respectively. Furthermore, these mutations cause lethality in sgs1Δ cells [53,54,57].

What is the mechanistic basis of the requirement of a STUBL (SUMO-targeted ubiquitin ligase) for the viability of cells lacking SGS1 or for the viability of cells exposed to hydroxyurea? It is known that DNA damage, replisome stalling or deletion of SGS1 triggers the SUMOylation of proteins involved in HR. One possibility is that Slx5–Slx8-mediated degradation of one or more of these SUMOylated proteins is important for the rescue of blocked replisomes. But is it not paradoxical that the degradation of a DNA-repair protein would be important for DNA repair or for genome stability? Not when we consider that protein modification by SUMO is often transient [5962], and ubiquitination-dependent degradation of SUMOylated proteins represents a logical way to ensure this transience. It may be that Slx5–Slx8 degradation ensures removal of certain HR proteins from their substrates so that the protein regulating the next step of the process can gain access to complete HR. Slx5–Slx8-mediated degradation may also ensure that HR is switched off at the right time, to prevent hyper-recombination or inappropriate recombination events. Consistent with this hypothesis, mutations in Slx5 or Slx8 cause an increase in the spontaneous frequency of several different types of HR, both Rad51-dependent and Rad51-independent [63]. So what are the targets of Slx5–Slx8? A variety of proteins involved in HR have been reported to be SUMOylated and, in particular, Rad52 is strongly SUMOylated in vivo after exposure of cells to DNA damage and can be degraded by the 26S proteasome [64]. However, Rad52 degradation was unaffected in cells defective in Slx5–Slx8 function and genetic data imply that the functions of Slx5–Slx8 are not wholly dependent on Rad52 [63,65,66]. A systematic analysis of proteins that are SUMOylated after DNA damage and the identification of the targets of Slx5–Slx8 that promote genome stability are vitally important endeavours for the future.

The Slx5–Slx8 complex is evolutionarily conserved and is referred to as Slx8–Rfp in S. pombe [67,68]. In humans, RNF4 was identified as an orthologue of Slx5 that also shares the SIMs of Slx8 [67,68]. Because there is no other obvious orthologue of Slx8 in humans, it is thought that RNF4 performs the functions of both Slx5 and Slx8. S. pombe Slx8–Rfp interacts with the SUMO-like domains of a protein called Rad60 via the SIMs of Rfp [67], and this leads to SUMO-dependent ubiquitination of Rad60. Rad60 plays important roles in genome stability: cells lacking Rad60 are hypersensitive to a wide range of genotoxins and cannot repair DSBs, and Rad60 mutations are epistatic to mutations in HR enzymes [6971]. However, the consequence of Rad60 ubiquitination by Slx8–Rfp is not clear. Human RNF4 interacts with the Rad60 orthologue NIP45 [NF-AT (nuclear factor of activated T-cells)-interacting protein 45], but little is known about NIP45, and it has not yet been implicated in cellular responses to DNA damage [67]. It was shown recently that PML (promyelocytic leukaemia protein) is an in vivo target of human RNF4 [72]. Many cases of APL (acute promyelocytic leukaemia) are caused by a chromosomal translocation that fuses PML to the RAR (retinoic acid receptor), and it was discovered that treatment of APL cells with arsenic caused ubiquitin-mediated degradation of PML–RAR [7378]. Remarkably, the down-regulation of PML–RAR by arsenic did not occur in cells lacking RNF4 [72,79].

Slx1–Slx4: a novel structure-specific heterodimeric endonuclease

Clues to the functions of Slx1–Slx4 came from bioinformatics analyses which revealed that Slx1 is the founding member of a family of nucleases. Aravind and Koonin [80] identified prokaryotic orthologues of the Ku protein, a central player in the repair of DSBs by NHEJ (non-homologous end-joining) [81]. They noted that, in prokaryotes, the Ku open reading frame almost always lies in an operon that encodes other DNA-repair proteins [80]. One of the Ku-encoding operons in Mesorhizobium loti includes a small gene that belongs to a discrete family of URI (UvrC-intron-type) endonucleases typified by Escherichia coli YhbQ, Bacillus subtilis YazA and yeast Slx1 [80]. These and other in silico analyses led to the discovery that these related URI-domain nucleases represent a distinct subgroup of the GIY–YIG nuclease superfamily [82,83]. URI domain nucleases are found in viruses, bacteria, eukaryotes and in a single archaeon [80]. Many of the prokaryotic members of this family are standalone URI domains and are of unknown function. The eukaryotic members of the family are exemplified by S. cerevisiae Slx1 (Figure 3) and consist of a URI domain and a PHD-type zinc-finger [80], although, in Arabidopsis thaliana, Slx1 is fused to the MutS ATPase involved in mismatch repair [80,82]. In contrast with Slx1, Slx4 has no obvious catalytic motifs that might give clues to how it acts at the molecular level, but it is assumed to act as a regulatory subunit of Slx1. Slx4 has a SAP domain, that in other proteins mediates binding to DNA [84,85], suggesting that Slx4 may target the Slx1–Slx4 complex to DNA substrates. The overall amino acid sequence similarity between fungal orthologues of Slx4 is low, but a C-terminal region containing the SAP domain shows the highest degree of conservation [86].

The Slx4–Slx1 complex from S. cerevisiae and S. pombe shows DNA structure-specific nuclease activity in vitro [86,87]. The fungal complexes do not cleave linear dsDNA (double-stranded DNA) or ssDNA and have almost no activity towards simple 5′- or 3′-overhangs. In contrast, there is a striking preference for branched DNA structures: Y-forks, 3′-flaps, 5′-flaps and replication forks (Figure 4) [86,87]. Although Slx4–Slx1 can cleave HJs in vitro, it is unlikely that this complex cleaves HJs in yeast in vivo, since conventional HJ resolvases cleave HJs symmetrically to form ligatable nicked DNA, but Slx4–Slx1 does not [86,87]. In addition, no meiotic defects, which would be anticipated if HJ resolution was defective, are observed in slx1Δ or slx4Δ cells [39]. Mutation of arginine or glutamic acid residues that are invariant in all URI nuclease domains abolished the nuclease activity of S. pombe Slx1, and these mutations are synthetic lethal in cells lacking Rqh1 [86]. Therefore the nuclease activity of the Slx4–Slx1 complex is required for maintenance of cell viability in the absence of Rqh1.

Co-operation of Slx1–Slx4 and RecQ helicases in processing replisomes stalled at replication fork barriers in rDNA (ribosomal DNA)

In most eukaryotic organisms, the genes encoding rRNA are clustered in long tandem repeats on one or a few chromosomes, and the total number of these chromosomal rDNA repeats appears to be maintained at a level appropriate for each organism. In S. cerevisiae, the rDNA locus, situated on chromosome XII (Figure 5) comprises a head-to-tail array of 100–200 basic transcription units in the nucleolus [88]. Alterations in rDNA copy number have been observed, and it is thought that the rDNA array can expand and contract in response to environmental cues [8992]. A single rDNA unit comprises two transcribed regions encoding 35S and 5S rRNA (both transcribed by RNA polymerase I) respectively, and two non-transcribed spacer regions (NTS1 and NTS2). Each unit has an origin of replication [ARS (autonomous replication sequence)] and a polar RFB (replication fork barrier) that allows the replication fork to proceed in the direction of rRNA transcription, but not in the opposite direction (Figure 5) [9395]. The RFB requires binding of a the Fob1 protein to act as a replication barrier [92]. In each S-phase, DNA replication initiates bidirectionally from the ARS in one in every five rDNA units. So, one in every five rDNA units has a stalled replication for [93,96].

Figure 5 The S. cerevisiae rDNA array

In S. cerevisiae, the rDNA locus situated on chromosome XII comprises a head-to-tail array of 100–200 basic transcription units in the nucleolus. A single rDNA unit comprises two transcribed regions encoding 35S and 5S rRNA (both transcribed by RNA polymerase I) respectively, and two non-transcribed spacer regions (NTS1 and NTS2). Each unit has an origin of replication (ARS) and a polar RFB that allows replication fork to proceed in the direction of rRNA transcription, but not in the opposite direction. The RFB requires binding of the Fob1 protein to act as a replication barrier. In each S-phase, DNA replication initiates bidirectionally from one in five rDNA units, and so each of these units has a stalled replication fork.

To understand why slx4Δ sgs1Δ cells are inviable, Kaliraman and Brill [97] constructed temperature-sensitive (ts) alleles of sgs1 that caused cell death in slx4Δ cells only when cells were shifted from 25 to 37°C. This study showed that, although bulk DNA synthesis was not affected by shifting sgs1ts slx4Δ cells to 37°C, cells arrested in late S-/early G2-phase and eventually died [87]. PFGE (pulsed-field gel electrophoresis) was used to analyse replication of individual chromosomes under these conditions: incompletely replicated chromosomes do not enter pulsed-field gels because of the presence of replication forks and bubbles [98,99]. When sgs1ts slx4Δ cells were shifted to 37°C, 15 of the 16 yeast chromosomes migrated into the pulsed-field gel, but chromosome XII did not [97]. The aberrant migration of chromosome XII from sgs1ts slx4Δ was dependent on the rDNA array and also on progression of cells through S-phase [97], suggesting a major problem with replication of rDNA in these cells. Curiously, introduction of Sgs1 back into sgs1ts slx4Δ cells that were arrested in G2-phase after passing through S-phase at 37°C fully rescued the appearance of chromosome XII, and cell-cycle progression resumed [97]. So it appears that Sgs1 and Slx1–Slx4 act in parallel pathways, late in S-phase after most DNA has been replicated, to enable completion of rDNA replication. Consistent with these data, S. pombe Slx1 and Slx4 are associated with chromatin at foci characteristic of the two rDNA repeats on chromosome III in that organism, and deletion of Slx1 or Rqh1 causes rDNA contraction [86].

Why is the Slx4–Slx1 complex so important for rDNA replication? Most rDNA repeats are replicated completely by the fork moving in the direction of rDNA transcription. In the absence of Sgs1 and Slx4–Slx1, however, one in five repeats would be defective if a problem occurred with the replication fork blocked at the RFB. One possibility is that Sgs1 and Slx4–Slx1 are required to process structures, formed at the replisomes stalled at RFBs in rDNA to allow completion of rDNA replication and chromosome segregation. It has been proposed that stalled forks can regress to form HJ-like structures (Figure 2C) that can be maintained and processed in a non-recombinogenic manner by Sgs1. It is possible that Slx4–Slx1 can cleave these structures to generate a free DNA end that is resected to initiate HR to reform the replication fork. Rad22, the S. pombe orthologue of Rad52, is required for all HR events in S. pombe. Russell and co-workers showed that deletion of Slx1 or Slx4 partially rescues the slow growth of rad22d cells [100], suggesting that Slx1–Slx4 creates a substrate for Rad22-mediated HR. In support of this, deletion of Slx4 or Slx1 reduces the frequency of spontaneous Rad22 foci in the nucleolus where rDNA is localized, and causes the accumulation of replication forks stalled at RFBs in rDNA [100]. It is clear that Slx4–Slx1 and Sgs1 co-operate to enable completion of rDNA replication, possibly by processing replisomes stalled at RFBs. But is the defect in rDNA replication responsible for the synthetic lethality of slx4Δ sgs1Δ cells? This is an important question and could be tested by deleting FOB1 in these cells, which would prevent forks from stalling at the rDNA RFBs in the first place. Is Slx1–Slx4 required in other regions of the genome such as heterochromatin, centromeres, telomeres and tRNA genes, where replisomes stall at high frequency [101]? It will be important to analyse replication of these regions of the genome in sgs1ts slx4Δ cells by two-dimensional gels to check for persistent fork stalling. It is interesting to note that, during PFGE of chromosomes from sgs1ts slx4Δ cells, there was a slight decrease in the intensity of most chromosomes [97].

Slx4 regulates recovery from DNA-alkylation-induced replisome stalling independently of Slx1

Although Slx4 and Slx1 work together as a complex in processing stalled replisomes in rDNA, cells lacking Slx4 are hypersensitive to DNA-alkylating agents such as MMS (methylmethane sulfonate), whereas cells lacking Slx1 are not [87,102105]. The major lesion, N3-methyladenine, induced by MMS, potently blocks replisome progression [106]. As a result, exposure of cells to MMS, as they proceed synchronously from G1- into S-phase, dramatically lowers the rate of DNA synthesis. When MMS is removed, wild-type cells can resume and complete DNA replication, but two studies revealed that cells lacking Slx4 were severely defective in recovery from MMS under these conditions [105,107]. Although slx4Δ cells had an almost 2C (diploid) DNA content 3 h after removal of MMS, the chromosomes did not enter a pulsed-field gel, indicating they had not been completely replicated [105,107]. MMS causes activation of the Rad53 checkpoint kinase that orchestrates signalling downstream of stalled replisomes, but Rad53 activity returns to basal levels after MMS is removed from wild-type cells, whereas Rad53 activity remains high in slx4Δ cells [105,107]. DNA fibre analysis showed that, although cells lacking Slx4 can restart DNA synthesis during recovery from MMS, DNA replication is much slower than normal, and chromosomes are left with unreplicated gaps [105]. Cells lacking Slx1 were indistinguishable from wild-type cells.

So what does Slx4 do to promote recovery from MMS-induced replisome blockage? Although the molecular mechanisms involved are unclear, the available data indicate that Slx4 regulates the ability of stalled replisomes to ‘bypass’ DNA lesions even before removal of the damaged bases [105]. Two major lesion bypass mechanisms have been described: TLS (translesion synthesis) and error-free bypass (Figure 6A) [108111]. In S. cerevisiae, both require Rad6 [112], an E2 Ubc [113,114] and Rad18, a DNA-binding protein that targets Rad6 to stalled replisomes [115]. PCNA (proliferating-cell nuclear antigen) plays a key role in the regulation of DNA-damage bypass [110]. The Rad6–Rad18 complex mono-ubiquitinates PCNA at Lys164 [116] when replisomes stall, and mono-ubiquitination of PCNA is required for efficient TLS [116,117]. TLS in S. cerevisiae involves highly specialized enzymes such as Pol (DNA polymerase) ζ and Pol η that can replicate through DNA lesions (Figure 6A) [110,118]. Mono-ubiquitination of PCNA at Lys164 is important for efficient binding of TLS polymerases to stalled replisomes [117,119,120]. Although Pol η-mediated TLS is error-free, Pol ζ is mutagenic (Figure 6A) [121,122], and, in fact, TLS is responsible for the majority of spontaneous and DNA-damage-induced mutations in cells [123].

Figure 6 DNA lesion bypass during DNA replication

(A) Schematic representation of the two main modes for the bypass of DNA lesions that block DNA replication. Grey strands represent parental DNA and red strands represent nascent DNA. The green shapes represent the DNA replication machinery (replisomes), whereas the blue shapes represent the replicative DNA helicase. The black diamond represents a DNA lesion that blocks DNA replication. During TLS, TLS polymerases (shown in yellow) are transiently recruited to the stalled fork, probably by their direct recognition of mono-ubiquitinated PCNA. TLS polymerases have extended active sites compared with the conventional replicative DNA polymerases, enabling them to accommodate damaged bases in their active sites and to replicate past DNA lesions. Some TLS polymerases are error-prone and introduce mutations. In contrast, template switching is error-free because the intact nascent strand is used as a template for synthesis of the blocked nascent strand. (B) Schematic diagram illustrating two possible mechanisms for error-free bypass.

So mono-ubiquitination of PCNA regulates TLS, but error-free bypass (Figure 6) requires polyubiquitination of PCNA. This requires three factors in addition to Rad6–Rad18: Mms2, Ubc13 and Rad5 [116,124126]. Mutations in the Mms2–Ubc13–Rad5 complex prevent error-free lesion bypass, but not TLS [116,124126]. However, exactly how the formation of polyubiquitin chains on PCNA enables error-free bypass is unknown. Indeed, the mechanisms underlying error-free bypass and the genes involved have remained elusive. Two models for error-free bypass have been proposed: sister-strand exchange and fork regression [127,128], and both involve transient template switching (Figure 6B). In the sister-strand exchange model, the nascent strands at the blockage separate from the parental strands and subsequently anneal to one another. One of the nascent strands then serves as the template for elongation of the other to allow DNA replication past where the initial block to replication occurred (Figure 6B). An intact replication fork is re-formed, either by reannealing of the nascent strand with the respective parental strands or by recombination. In the fork regression model (Figure 6B), the replicated parental strands in the stalled replisome reanneal, causing the released nascent strands to anneal also, allowing template switching and DNA synthesis past the point where the replisome stalled. The resulting ‘chicken foot’ structure could then be resolved to form an intact replication fork by the action of helicases and/or nucleases. It should be noted that, at present, there is no evidence for chicken foot structures occurring in wild-type cells.

In terms of hypersensitivity to MMS, slx4 is epistatic to mms2 and ubc13 and to the pol30 k164r mutant where Lys164 of PCNA is mutated to arginine, but slx4 is not epstatic to mutations in TLS factors [105]. Deletion of known error-free bypass factors causes a compensatory increase in TLS and a corresponding increase in mutation frequency. Similarly, deletion of SLX4 caused increased TLS in cells before and after exposure of cells to MMS [105]. These and other data (S. Flott and J. Rouse, unpublished work) indicate that Slx4 is required for error-free bypass. Since the mechanisms involved in this mode of lesion bypass are unclear, as are some of the relevant genes, it is difficult to speculate on the potential role of Slx4. Template switching should involve the action of helicases and nucleases, and it is possible that Slx4 recruits one or more catalytic activities to replisomes blocked by DNA lesions, for example. In this light, it is interesting that Slx4 interacts with Slx1 that in vitro cleaves structures that resemble those that may arise during error-free bypass. However, Slx1 is not required for cellular resistance to MMS. It will be critically important to identify the Slx4-interacting factor(s) that promotes lesion bypass.

Slx4 co-operates with the BRCT domain-containing protein Esc4/Rtt107 to promote recovery from replisome stalling

Slx4 interacts with S. cerevisiae Esc4/Rtt107 [107,129] that, like Slx4, is required for cellular resistance to, and recovery from, MMS. Esc4 contains multiple BRCT [BRCA1 (breast cancer 1 early onset) C-terminus] domains, protein–protein interaction domains found in proteins involved in DNA-damage signalling and repair [130,131]. In this regard, Esc4 probably acts as a scaffold for the recruitment of proteins that protect genome stability. The gene encoding Esc4 was first identified in a genetic screen for increased Ty1 transposon mobility [132]. Cells lacking Esc4 are hypersensitive to many different agents that block DNA replication, not just MMS [102,103,107,133]. Esc4 is required to survive DNA damage in S-phase, but is not required for the intra-S-phase checkpoint [107,133]. Instead Esc4 promotes the restart of stalled replication forks, but is probably not required to prevent the irreversible fork catastrophe when forks stall. Esc4 is phosphorylated by Mec1, the yeast orthologue of the ATR [ATM (ataxia telangiectasia mutated) and Rad3-related] kinase [107,133] and this is important for recovery from MMS [133].

The inability of esc4Δ cells to recover from exposure to MMS is reminiscent of slx4Δ cells. This, together with the interaction of Esc4 with Slx4, suggests that Esc4 and Slx4 may work together on the same pathway to protect cells when replisomes are blocked by DNA-alkylation damage. Grant Brown's laboratory made the intriguing observation that Slx4 is required for MMS-induced phosphorylation of Esc4 by Mec1 [134] that in turn is important for recovery of DNA replication after replisome stalling in MMS [135]. However, although Esc4 and Slx4 may overlap functionally in recovery from MMS, they also have independent functions. Cells lacking Slx4 are hypersensitive to MMS [87,102105], but not to other agents that stall replisomes, such as camptothecin or UV light, whereas cells lacking Esc4 are hypersensitive to all replisome-blocking agents [102,103,133]. Secondly, slx4 is epistatic to mutants in the error-free bypass pathway [105], whereas esc4 is not [133,136]. Thirdly, slx4Δ is lethal when combined with sgs1Δ [39], whereas esc4Δ sgs1Δ cells are viable [137]. These data argue that, whereas Esc4 and Slx4 probably work together in promoting the recovery of replisomes blocked by DNA-alkylation damage, they make independent contributions to other aspects of genome stability maintenance. It is likely that there are other Slx4-interacting proteins that are required for recovery from replisome stalling. Intriguingly, S. pombe Slx1 and Brc1 (S. pombe Esc4) work together in promoting resistance to DNA-alkylation damage in smc6 mutants [138]. Smc6 is a subunit of a cohesin-like complex that is essential for recovery from HR-mediated recovery from replisome stalling [139,140].

Slx4 is essential for 3′-flap cleavage by Rad1–Rad10

Genome-wide two-hybrid analysis revealed a range of potential Slx4-interacting factors, including Rad1, the catalytic subunit of a well-defined heterodimeric endonuclease Rad1–Rad10 [141]. As mentioned above, Rad1 and Rad10 both belong to the XPF/MUS81 nuclease family; Rad1 (XPF in higher eukaryotes) is the catalytic subunit, and Rad10 (ERCC1 in higher eukaryotes) is inactive as a nuclease, but contributes to DNA binding of the complex [142,143]. Binding of Rad1 to Rad10 is essential for nuclease activity [144,145]. In vitro, Rad1–Rad10 preferentially cleaves branched DNA substrates including splayed duplexes, 3′-flaps and bubbles (Figure 4) [146148], and, in the case of splayed duplexes and 3′-flaps, Rad1–Rad10 introduces nicks on the dsDNA region on the same DNA strand as the 3′-tail [146]. This complex has several important roles in repair of DNA lesions, including the repair of UV-light-induced DNA damage by NER (nucleotide excision repair). The Rad1–Rad10 complex is recruited to UV-light-induced DNA lesions where it catalyses cleavage on the 5′-side of the DNA bubble, generated around the UV lesion by the action of helicases and many other factors, in order to initiate NER [149151].

Rad1–Rad10 is also important for a subpathway of DSB repair. The dominant mechanisms for repair of DSBs are NHEJ, which involves direct ligation of DNA ends [12], and HR [14]. As mentioned above, there are three major types of HR, i.e. gene conversion (Figure 3), BIR and SSA, and all of these start with the resection of DNA ends, generating 3′-ssDNA that becomes coated with Rad51 [1214]. SSA repairs DSBs between repeated sequences (Figure 7A). After resection of the ends, Rad52 and Rad59 promote the annealing of homologous sequences, resulting in an intermediate that has two non-homologous 3′-ended tails [12,13]. Work from several laboratories has shown that Rad1–Rad10 cleaves these 3′-single-strand tails, providing a 3′-hydroxy group for filling-in by DNA synthesis, followed by ligation to complete repair [45,152156]. This results in deletion of one copy of the repeat plus the sequence located between the repeats (Figure 7A). When a DSB is repaired by gene conversion, the use of the invading DNA strand as a primer for new DNA synthesis requires that non-homologous bases at the 3′-end be removed by Rad1–Rad10 so that the invading end can act as a primer for new DNA synthesis (Figure 7B). In either gene conversion or SSA, tail removal by Rad1–Rad10 also requires the Msh2–Msh3 mismatch-recognition complex [156160], but it is not entirely clear why. The available data suggest that Msh2–Msh3 may recognize and stabilize the 3′-tails at the dsDNA–ssDNA junction, somehow facilitating their cleavage. Rad59 and Srs2 have also been implicated in 3′-non-homologous tail removal [156,161]. The requirement for Rad1–Rad10 during gene conversion depends on the number and length of non-homologous tails. When both ends of a DSB have tails 30 nt or longer, Rad1–Rad10 is essential for gene conversion [156158,162]. However, when only one non-homologous tail arises, Rad1–Rad10 is less critical [158,162,163]. When 3′-tails are less than 10 nt long, Rad1 and Msh2–Msh3 are not required for gene conversion or SSA, and instead they are removed by the proofreading activity of Pol δ [156]. Rad1–Rad10 is required for removal of all tails longer than 30 nt, whereas Msh2–Msh3 is not required for tails 1 kb or longer [157].

Figure 7 Cleavage of 3′-non-homologous tails during SSA and gene conversion

(A) Schematic representation of single-strand annealing. This pathway is used for repair of DSBs that occur between direct repeats (depicted as orange arrows). After resection of the ends past the repeats to yield single-stranded 3′-overhangs which have the repeat sequence in the ssDNA tail, the complementary strands anneal at the repeat sequence. This generates 3′-single-stranded tails that are cleaved by Rad1–Rad10, and the resulting gaps are sealed by ligation. SSA results in deletion of one copy of the repeat plus the sequence located between the repeats. (B) Schematic representation of HR (gene conversion) where non-homologous sequences (shown in green) flank a DSB. The occurrence of non-homologous sequences at a DSB results in the formation of 3′-non-homologous tails that must be removed for the invading ends to act as primers for DNA synthesis (right-hand panel). The scissors symbols denote tail cleavage events catalysed by Rad1–Rad10. When there are no non-homologous sequences at a DSB, Rad1–Rad10 is not required for HR (left-hand panel).

Slx4 interacts with Rad1–Rad10 in vivo [105]. Although Rad1–Rad10 is required for NER, Slx4 is not required [105]. However, Slx4 is essential for Rad1–Rad10-dependent removal of 3′-non-homologous tails during SSA repair of a single endonuclease-induced DSB (Figure 7A) and for gene conversion when the DSB is flanked by non-homologous sequences (that lead to 3′-non-homologous tails) (Figure 7B) [105,164]. In these situations, both ends of the DSB have non-homologous tails, but there are circumstances when only one end of a DSB has a non-homologous tail, such as mating-type-switching. S. cerevisiae cells adopt one of two possible two mating types, a or α, and this is controlled by the MAT locus. Under some circumstances, cells can switch mating type, and this requires the induction of a DSB at the MAT locus by the HO endonuclease that initiates a gene-conversion event between MAT and silent-mating-type cassettes HMLα or HMRa (reviewed in [165]). A single 3′-non-homologous tail occurs during repair of the HO-induced DSB during mating-type-switching [162], and in this light it was shown recently that during mating-type-switching cells lacking Rad1–Rad10 or Slx4 showed a G2/M delay that required the DNA-damage checkpoint, accompanied by loss of viability [162].

Why is Slx4 so important for cleavage of 3′-non-homologous tails by Rad1–Rad10 during HR/SSA? There are a number of possibilities. First, Slx4 could act as a flap sensor that loads Rad1–Rad10 on to 3′-non-homologous tails. This has still to be investigated. However, recently, Saw1, a novel SSA factor identified recently in a screen for genes required for SSA, was shown to be essential for loading of Rad1–Rad10 in the vicinity of 3′-non-homologous tails during SSA [164]. Direct physical interaction between Rad1 and Saw1 is essential for loading Rad1–Rad10 and for cleavage of 3′-non-homologous tails during SSA [164]. It remains to be tested whether Saw1 is required for loading of Slx4 on to 3′-tails or whether Slx4 binds to these species directly. Secondly, it may be that Slx4 directly affects the activity or substrate specificity of Rad1–Rad10 through direct protein–protein interaction. From this point of view, it is known that Slx4 binds to the Rad1 subunit of Rad1–Rad10 ([164] and G. Toh, I. Munoz and J. Rouse, unpublished work), but it is not yet known where on Rad1 Slx4 binds. Binding of Slx4 in the region of the Rad1 catalytic domain could affect its substrate specificity and/or activity towards DNA flaps through inducing conformational changes in this domain. Thirdly, it is possible that Slx4 assists Rad1–Rad10 by denaturing/deforming the DNA substrate, enabling efficient cleavage. It was shown that PCNA increases the catalytic rate constant of XPF from the archaeon Sulfolobus solfataricus 7000-fold in vitro; XPF was inactive towards a 3′-flap substrate without PCNA [166]. XPF has an ERCC4-type endonuclease domain separated by a flexible linker from a HhH2 (tandem helix–hairpin–helix) domain [167]. The S. solfataricus HhH2 domain of XPF is thought to distort DNA, allowing the nuclease domain to cleave the DNA flap (in Rad1–Rad10, it is the Rad10 hairpins that bind DNA [143]). It has been proposed that PCNA acts as a ‘molecular fulcrum’ against which XPF can exert force to distort the DNA substrate, driving it towards the transition state [166]. Consistent with this, the S. solfataricus PCNA-interaction motif lies close to the HhH2 domain. It is possible that Slx4 regulates Rad1–Rad10 in a manner analogous to modulation of archaeal XPF by PCNA. The PCNA-interaction motif in archaeal XPF is lacking in Rad1, and it will be interesting to test whether this region on Rad1 binds Slx4 instead. A related possibility is that Slx4 may prevent or reverse binding of 3′-tails to the surface of nucleosomes that might prevent access of Rad1–Rad10 to its substrate. Testing the effects of Slx4 on 3′-flap cleavage by Rad1–Rad10 in vitro should provide answers to these questions. It will be vitally important to understand how Slx4 interacts with other proteins bound in the vicinity of 3′-non-homologous tails such as Msh2–Msh3 and Saw1.

Given that Slx4 is so important for Rad1–Rad10-catalysed tail cleavage during HR, why is it not required for NER? After recognition of UV-damaged DNA by sensor proteins, the DNA around the lesion is unwound by DNA helicases to create a DNA bubble, and this open DNA structure is stabilized by binding of proteins such as TFIIH (transcription factor IIH) and RPA [149151]. Rad1–Rad10 cleaves the 5′-side of the bubble, and it is likely that Slx4 is not required for substrate deformation here because of the open nature of the DNA substrate and the many proteins that stabilize this open structure.

Phosphorylation of Slx4 modulates DNA repair and helps to distinguish the different functions of Slx4

Slx4 becomes phosphorylated after exposure of cells to a range of different genotoxins [104], and this is catalysed by Mec1 and Tel1, the yeast orthologues of the ATR and ATM protein kinases respectively. These kinases belong to a signalling network, and defects in this network cause hypersensitivity to genotoxins, loss of DNA damage checkpoints, and genomic instability [168171]. Mec1 belongs to a family of protein kinases termed PIKKs (phosphoinositide 3-kinase-like kinases), and it functions in a partially redundant manner with Tel1, another PIKK [169]. In mammals, ATM seems to be involved predominantly in responding to DSBs [170], but ATR responds to a wider variety of lesions and has a particularly important role in genome surveillance during DNA replication [171]. Disruption of ATR or Mec1 causes cell lethality [172174], and mec1-null cells can be rescued from lethality by increases in intracellular deoxyribonucleotide levels, although the molecular basis for this observation is still unclear [99,175]. ATM and Tel1 are not essential proteins, although ATM defects give rise to ataxia telangiectasia, a debilitating human neurodegenerative and cancer-predisposition disease whose symptoms include hypersensitivity to agents that cause DSBs [176]. Several cellular proteins become rapidly phosphorylated in a Mec1/Tel1-dependent manner in response to DNA damage [170,177]. Some of these, such as Rad53 and Chk1 (checkpoint kinase 1) (CHK2 and CHK1 respectively in humans) [178], are themselves protein kinases that are activated by Mec1/Tel1 (ATR/ATM)-dependent phosphorylation [177]. When activated, these proteins phosphorylate and modulate the activities of key effectors of the DNA-damage response [177].

The electrophoretic mobility of Slx4 changes after exposure of cells to genotoxins, and this is due to phosphorylation, since it can be reversed by phosphatase treatment [104]. Deletion of either Mec1 or Tel1 does not affect Slx4 phosphorylation, but deletion of both of these kinases together prevented DNA-damage-induced Slx4 phosphorylation [104,105]. This implies that Mec1 and Tel1 phosphorylate Slx4 directly or that one or more kinases activated by Mec1 and Tel1 are responsible. However, the only known Mec1/Tel1-activated kinases Chk1 and Rad53 are not required for Slx4 phosphorylation [104]. Futhermore, phosphorylation site mapping revealed that Slx4 is phosphorylated on at least six serine/threonine residues in vitro by PIKKs and that the Slx4 residues phosphorylated in vivo after DNA damage conform to the consensus sequence for Mec1 and Tel1 [105]. Thus Mec1 and Tel1 phosphorylate Slx4 directly after DNA damage. Although the phosphorylated residues in Slx4 are distributed throughout the N-terminal and central regions of Slx4 where there are no obvious functional domains, analysis of the residues in Slx4 phosphorylated in response to DNA damage proved to be very informative. A mutant Slx4 in which three of the in vivo phosphorylation sites induced by DNA damage were mutated to alanine was fully able to rescue the MMS-hypersensitivity of slx4Δ cells, and, unlike full deletion of SLX4, these mutations did not cause lethality in sgs1Δ cells [105]. However, this Slx4 mutant was unable to rescue the SSA defect of slx4Δ cells. Therefore phosphorylation by Mec1 and Tel1 is required only for the Rad1-dependent aspect of Slx4 function, and, in this respect, the Slx4 phospho-site mutant is a ‘separation-of-function’ mutant.

It is clear that phosphorylation of Slx4 is not required for its interaction with Rad1–Rad10 [105], so why is phosphorylation of Slx4 required for 3′-non-homologous tail cleavage by this nuclease? Answering this question will require a basic understanding of how Slx4 functions at the molecular level to assist Rad1–Rad10, but this is lacking at present. We are interested in the possibility that Slx4 phosphorylation restricts Rad1–Rad10 activity towards DNA flaps to sites of DNA damage. It may be that, if Rad1–Rad10 could cleave flaps constitutively, this would result in inappropriate cleavage of telomeres or Okazaki fragments, for example. Co-localization of Mec1 and Tel1 with Slx4 at DSBs would trigger Slx4 phosphorylation enabling Rad1–Rad10 to cleave the 3′-non-homologous tails in the vicinity, but not elsewhere. This hypothesis predicts that a phospho-mimic mutant of Slx4 may cause genomic instability because of inappropriate cleavage by Rad1 and this will be interesting to test in the future.

The observation that phosphorylation of Slx4 is required for SSA, but not other Slx4-dependent events, reinforces the notion that Slx4 has at least three genetically separable cellular roles: maintaining viability in the absence of SGS1, facilitating error-free lesion bypass/resistance to MMS and enabling cleavage of non-homologous tails during HR (Figure 8). These roles are mediated by distinct Slx4-interacting proteins. Slx1–Slx4 mediates survival of sgs1Δ cells and Slx4–Rad1–Rad10 mediates 3′-non-homologous tail cleavage. An Slx4 complex whose composition is still unknown, but that probably includes Esc4, mediates resistance to MMS/error-free bypass (Figure 8). These complexes do not appear to overlap in functional terms: Rad1 is not required for viability of sgs1Δ cells [32] or for cell resistance to MMS [105], and Slx1 is not required for SSA or for cell resistance to MMS [105]. In the future, it will be crucial to decipher exactly what Slx4 is doing at the molecular level in each of these complexes and what the similarities and differences are. The in vitro reconstitution of the different Slx4-containing complexes will greatly facilitate this aim.

Figure 8 Slx4 has multiple independent cellular roles that require distinct Slx4-interacting proteins

Schematic representation of the known Slx4-containing complexes. The available data indicate that Slx4 are has at least three genetically separable cellular roles: maintaining viability in the absence of SGS1, facilitating error-free lesion bypass/resistance to MMS and enabling Rad1–Rad10-dependent flap cleavage during HR. The available data argue that the different functions of Slx4 are mediated by distinct Slx4-interacting proteins. Slx1–Slx4 mediates survival of sgs1Δ cells and maintenance of rDNA, Slx4–Rad1–Rad10 mediates DNA-flap cleavage, and an Slx4 complex whose composition is still unknown, but which includes Esc4, mediates resistance to MMS/error-free bypass. These complexes do not appear to overlap in functional terms: Rad1 is not required for viability of sgs1Δ cells or for cell resistance to MMS, and Slx1 is not required for SSA or for cell resistance to MMS. Phosphorylation of Slx4 by Mec1 and Tel1 is required for efficient 3′-flap cleavage by Rad1–Rad10, but is not required for error-free bypass/cell sensitivity to MMS or the viability of sgs1Δ cells.


In summary, the Slx protein complexes, Mus81–Eme1, Slx5–Slx8 and Slx1–Slx4, were discovered because loss of the components of any one of them is lethal in sgs1Δ cells. These complexes are physically distinct, and two of the complexes have structure-specific nuclease activity, whereas the other complex has SUMO-targeted ubiquitin ligase activity. These complexes not only are required for cell viability in the absence of SGS1, but also mediate a range of diverse functions in the maintenance of genome stability. Indeed, we still do not understand what it is that Slx complexes do to protect sgs1Δ cells. In a sense, what we know of Slx proteins only represents the ‘tip of the iceberg’, but a great deal of attention is now focused on deciphering the mode of actions of these complexes and on delineating the underlying molecular mechanisms. Identifying the SUMOylated proteins that Slx5–Slx8 target for ubiquitination is a major challenge that may yield to the power of proteomic analysis. Slx4 is a fascinating but poorly understood protein. It is clear that it has several independent functions that involve different interaction partners. These include the structure-specific nucleases Slx1 and Rad1–Rad10, and, from this point of view, Slx4 could be viewed as a nuclease scaffold or the co-ordinator of a ‘molecular toolkit’ for DNA repair. Many questions remain unanswered regarding Slx4 function. Why is Slx4 so important for Rad1 to cleave 3′-flaps during HR? How does Slx4 influence the activity and substrate specificity of the Rad1–Rad10 and Slx1? Does Slx4 bind to specific DNA structures through its SAP domain? How does phosphorylation affect Slx4 at the molecular level? How does Saw1 recruit Slx4–Rad1–Rad10 to 3′-flaps? How does Slx4 control error-free lesion bypass and what Slx4-interacting proteins are involved? Is Slx1 really specific for rDNA? What DNA structures does it cleave in vivo? Equivalents of the Mus81–Eme1 and Slx5–Slx8 complexes have been identified in many organisms, including humans, and this high degree of evolutionary conservation underscores the fundamental importance of these complexes. Although Slx1 is conserved in higher eukaryotes, there are no reports on Slx1 function other than in fungi, and there have been no reports of orthologues of fungal Slx4. However, reports describing SLX4-containing complexes in higher eukaryotes will emerge from my laboratory and others in the near future.

Colworth Medal Lecture Delivered at Robinson College, Cambridge, on 16 December 2008John Rouse


I am grateful to the Medical Research Council (MRC), Association for International Cancer Research (AIRC), Research Councils UK (RCUK), the Dorothy Hodgkins Postgraduate Award Scheme and the Division of Signal Transduction Therapy, University of Dundee, for financial support. J.R. is a European Molecular Biology Organization Young Investigator.


I am grateful to the people in my laboratory, past and present, for all their efforts and to my colleagues in the MRC Protein Phosphorylation Unit. I thank the support staff at the MRC Protein Phosphorylation Unit who help with our experiments. Special thanks to Sonja Flott who started our studies on Slx4, for her hard work and tolerance as my first Ph.D. student. Thanks also to Geraldine Toh, Ivan Munoz, Mary Gardiner and Karolina Hain who are now involved in the study of Slx proteins in a range of organisms. I also thank the people with whom we collaborate on yeast Slx4, especially Neal Sugawara, Jim Haber and Philippe Pasero for the neat experiments they have done and for interesting discussions. Thanks to Steve Brill who discovered the SLX genes. Thanks to Bettina Meier and Geraldine Toh for constructive comments on this article. I thank my past scientific mentors who have been a major source of support and inspiration: Professor Philip Cohen for awakening my interest in cell signalling and protein phosphorylation and Professor Steve Jackson whose work stimulated my interest in how cells cope with DNA damage and inspired me to work on this topic.


  • Colworth Medal Lecture:

Abbreviations: APL, acute promyelocytic leukaemia; ARS, autonomous replication sequence; ATM, ataxia telangiectasia mutated; ATR, ATM- and Rad3-related; BIR, break-induced replication; BLM, Bloom's syndrome protein; BS, Bloom's syndrome; BRCT, BRCA1 (breast cancer 1 early-onset) C-terminus; CHK, checkpoint kinase; DSB, double-strand break; dsDNA, double-stranded DNA; ERCC, excision repair cross-complementing; HhH2, tandem helix–hairpin–helix; HJ, Holliday junction; HR, homologous recombination; MMS, methylmethane sulfonate; NER, nucleotide excision repair; NHEJ, non-homologous end-joining; NIP45, NF-AT (nuclear factor of activated T-cells)-interacting protein 45; NTS, non-transcribed spacer; PCNA, proliferating-cell nuclear antigen; PFGE, pulsed-field gel electrophoresis; PIKK, phosphoinositide 3-kinase-like kinase; PML, promyelocytic leukaemia protein; Pol, DNA polymerase; RAR, retinoic acid receptor; rDNA, ribosomal DNA; RFB, replication fork barrier; RPA, replication protein A; RTS, Rothmund–Thompson syndrome; SCE, sister-chromatid exchange; SDSA, synthesis-dependent strand annealing; SSA, single-strand annealing; ssDNA, single-stranded DNA; SUMO, small ubiquitin-like modifier; SIM, SUMO-interaction motif; Top3, DNA topoisomerase III; TLS, translesion synthesis; ts, temperature-sensitive; Ubc, ubiquitin-conjugating enzyme; URI, UvrC-intron-type; WRN, Werner's syndrome protein; WS, Werner's syndrome; XPF, xeroderma pigmentosum complementation group F


View Abstract