In this review, we use the term “branchpoint” to refer to a malleable discontinuity in the DNA double helix. Common examples of branchpoints include the junction between single-stranded (ss) and double-stranded (ds) DNA in the stem-loop, the junction in a three-armed structure (e.g., splayed arm, replication fork, 5′-flap) and the core of a four-stranded Holliday junction (Figure 1A). Concerning the nomenclature of flapped DNA structures, the ssDNA is referred to as the “flap,” and the substrate name reflects the polarity of this strand (e.g., the 5′-flap contains an extension of ssDNA at the 5′-end) (Figure 1B). We use the terms “pre-nick” and “post-nick” to distinguish between the two dsDNA segments (or arms) that flank the incision site. Specifically, “pre-nick” refers to dsDNA upstream of the incision site, while “post-nick” refers to dsDNA downstream of the incision site (Figure 1C).

The realization that Slx1 belongs to a distinct cluster of the GIY-YIG nuclease superfamily stemmed from bioinformatics analyses that identified a discrete family of URI (UvrC-intron-type) endonucleases (Aravind and Koonin, 2001), which paved the way for the identification of new members of the GIY-YIG superfamily and their evolutionary classification (Dunin-Horkawicz et al., 2006). One remarkable feature of the GIY-YIG superfamily is that the nuclease domain is almost invariably associated with other domains, fused either N- or C-terminally. The different domain architectures have allowed for the classification of GIY-YIG proteins into different lineages and subfamilies, one of which includes the Slx1 lineage (Figure 2) (Dunin-Horkawicz et al., 2006).

In contrast to Slx1, which is conserved from yeast to human and contains well-defined structural or catalytic domains, Slx4 has diverged throughout evolution, lacks enzymatic motifs and is predicted to be predominately disordered (Figures 3A,B). In the following sections, we will highlight some of the defining features of Slx4/SLX4.

S. pombe Slx4 is the smallest Slx4 protein studied to date (418 amino acids). It contains the minimal domain architecture that is ubiquitous in fungi and animals: a SAF-A/B, Acinus and PIAS (SAP) domain and a conserved C-terminal domain (CCD) (Figure 3B). The SAP and CCD are connected by a flexible linker that varies in length between different organisms. As discussed in more detail below, the CCD is the most conserved region of Slx4 and consistently forms a critical part of the Slx1-Slx4 binding interface (Gaur et al., 2015; Lian et al., 2016; Gaur et al., 2019; Xu et al., 2021). By contrast, the sequence of the SAP domain varies considerably between species, which could serve different functions since the SAP domain binds DNA in S. cerevisiae Slx4 (Xu et al., 2021) and MUS81-EME1 in mammalian SLX4 (Fekairi et al., 2009; Svendsen et al., 2009; Castor et al., 2013; Kim et al., 2013).

One interesting difference between S. pombe Slx4 and S. cerevisiae Slx4 is that the latter contains an N-terminal extension that binds partner proteins, including two additional scaffolds called Rtt107 and Dbp11, a structure-specific DNA-binding protein called Saw1 and the Rad1-Rad10 structure-selective endonuclease. The Slx4-Rtt107-Dbp11 complex has critical roles in the DNA damage checkpoint to replication stress (reviewed in Guervilly and Gaillard, 2018). Through the interactions with Saw1 and Rad1-Rad10, Slx4 stimulates Rad1-Rad10 to remove ssDNA tails that form during a specialized form of homologous recombination called single-strand annealing (Flott et al., 2007; Li et al., 2008; Toh et al., 2010; Li et al., 2013).

The N-terminal extension of S. cerevisiae Slx4 is also present in animals and although the sequence varies significantly, the scaffold function is conserved. Human SLX4 has been the focus of intense research since its discovery in 2009 (Andersen et al., 2009; Fekairi et al., 2009; Munoz et al., 2009; Svendsen et al., 2009). As shown in Figure 3B, the N-terminal extension contains several core elements that mediate protein-protein interactions: two tandem ubiquitin-binding zinc finger type 4 (UBZ4) domains, the MUS312-MEI9 interaction-like region (MLR), a Broad-Complex, Tramtrack and Bric a Brac (BTB) domain, the TRF2-binding motif (TBM) and three SUMO-interacting motifs (SIM). We will explore these domains in more detail after discussing the structure and biochemical properties of a “minimal” SLX1-SLX4 heterodimer (i.e., SLX1 bound to the SLX4 CCD).

The published structures of Slx1 collectively reveal that the GIY-YIG nuclease domain contains a five-stranded anti-parallel β-sheet flanked by several α-helices (Gaur et al., 2015; Gaur et al., 2019; Xu et al., 2021) (Figure 4A). Several of the expected core elements are present, including the GIY-YIG hairpin formed by strands β1 and β2, an arginine helix (α1) that contains an invariant Arg residue, a linker strand that extends the GIY-YIG hairpin into an anti-parallel β-sheet and the glutamate helix (α2), which contains a conserved Glu residue (Figure 4B). The Slx1 active site residues are in the GIY-YIG hairpin, with the first motif (GIY) found in β1 and the second motif (YIG) found in β2 (Figure 4C). We will revisit the GIY-YIG motif when we discuss the catalytic mechanism of Slx1.

The apo structure of C. glabrata Slx1 revealed additional features in the nuclease domain that may be unique to some Slx1 proteins. First, a short β-hairpin (β4-β5) connects the glutamate helix (α2) to a helical segment (Figure 4D) (Gaur et al., 2015). The helical segment bends into strand β6, which is connected to strand β7 by an α-helix (Figure 4E). Notably, the β-hairpin was not observed in the Slx1-Slx4 structures, suggesting that this region adopts a different conformation in the presence of Slx4 (Gaur et al., 2015; Gaur et al., 2019; Xu et al., 2021). The second notable feature is a long α-helix (α6 in C. glabrata Slx1) that essentially provides a scaffold for the GIY-YIG nuclease and RING finger domains (Gaur et al., 2015) (Figures 4A,E). Intriguingly, the length of this helix varies between Slx1 from different fungi (Figure 4F). It would be interesting to determine whether this α-helix regulates the folding, stability, or enzymology of Slx1-Slx4. For example, one could substitute the short α-helix of T. terrestris Slx1 with the long α-helix from C. glabrata or S. cerevisiae Slx1 and then compare the biophysical and enzymatic properties of the mutant enzymes to that of wild type Slx1-Slx4. Another curious observation is that the helix is replaced with an extended unstructured region in human SLX1 (Jumper et al., 2021). The functional implications of this divergence are unknown, and the field anxiously awaits high-resolution structures of human SLX1-SLX4.

The C-terminal RING finger domain contains two α-helices and one or two anti-parallel β-sheets (Figures 4A,G) (Gaur et al., 2015; Lian et al., 2016; Gaur et al., 2019; Xu et al., 2021). There are also two zinc ions bound in the RING finger domain of all published Slx1 structures (Gaur et al., 2015; Lian et al., 2016; Gaur et al., 2019; Xu et al., 2021). Sequence and structural analyses show that the Slx1 RING finger domain is an atypical C4HC3-type RING finger (Gaur et al., 2015; Lian et al., 2016). Although many proteins containing this domain function as E3 ubiquitin ligases, the Slx1 RING finger domain appears to be required for intermolecular protein contacts that stabilize the Slx1-Slx4 heterodimer or, in the absence of Slx4, an Slx1 homodimer (Gaur et al., 2015; Lian et al., 2016; Gaur et al., 2019; Xu et al., 2021). Of note, the C-terminal tail of S. pombe Slx1 contains a SIM that binds Pmt3 (S. pombe SUMO) in vitro, raising the possibility that SUMOylated proteins could help recruit Slx1-Slx4 to specific DNA lesions in vivo (Lian et al., 2016).

Our mechanistic understanding of Slx1-Slx4 heterodimerization has significantly benefitted from crystal structures of full-length Slx1 bound to C-terminal fragments of Slx4 that contain the Slx1-binding domain (Gaur et al., 2015; Gaur et al., 2019; Xu et al., 2021). This region of Slx4 is conserved among animals and is therefore referred to as the conserved CCD. The overall structures of C. glabrata, T. terrestris and S. cerevisiae Slx1-Slx4^CCD are very similar (Figure 4H) (Gaur et al., 2015; Gaur et al., 2019; Xu et al., 2021). The most variable regions are in Slx1 and include: i) the loop between the arginine helix (α1) and β3 (the linker strand), ii) the loop between the glutamate helix (α2) and α3, and iii) the RING finger domain. Additionally, the β4-β5 hairpin was not observed in the Slx1-Slx4^CCD structures. The Slx4 CCD contains five short α-helices that fold into a globular domain (Gaur et al., 2015; Gaur et al., 2019; Xu et al., 2021). The SLX4^CCD fits into a cleft located between the Slx1 nuclease and RING domains (Figure 4H). Two main regions define the Slx1-Slx4^CCD binding interface. Region 1 involves hydrophobic contacts between Slx4^CCD α2 and the Slx1 RING finger domain; these interactions bury 556 Å of surface area in S. cerevisiae Slx1-Slx4^CCD. In contrast, Region 2 involves polar interactions between Slx4^CCD α5 and the Slx1 nuclease domain, which bury 520 Å of surface area.

As mentioned above, Slx1-Slx4 heterodimerization is a prerequisite for catalytic activity in vitro. In the absence of Slx4, C. glabrata Slx1 forms a salt-resistant homodimer that partially blocks the active site and DNA-binding residues (Gaur et al., 2015). Analysis of the crystal structures reveals that the position of one Slx1 monomer overlaps with Slx4^CCD in the Slx1-Slx4^CCD structure, indicating mutually exclusive protein-protein interactions. Indeed, the transition from homodimer to heterodimer is promoted by Slx4^CCD (Gaur et al., 2015). Two important and unresolved questions are whether monomeric Slx1 exhibits nuclease activity and whether C. glabrata Slx1 self-associates in vivo. It will be equally insightful to know whether this biochemical property is conserved through evolution. The ability of Slx1 to self-associate could also reflect the first principle of protein folding: globular proteins fold by minimizing the nonpolar surface exposed to water. Although we do not know whether Slx1 self-associates in vivo, we speculate that monomeric Slx1 is not stable in the absence of Slx4 and is likely targeted for proteasomal degradation. This may explain why the loss of SLX4 in human and murine cells leads to a significant or complete reduction in SLX1 levels (Munoz et al., 2009; Castor et al., 2013; Kim et al., 2013; Wyatt et al., 2013; Sarbajna et al., 2014). It would be interesting to determine whether the residual SLX1 in these cells exists as a self-inhibited homodimer and, if so, the biological consequences of expressing a constitutively monomeric nuclease.

Slx1-Slx4 is unique among structure-selective endonucleases because it cleaves all types of branched DNA structures, including replication and early recombination intermediates (e.g., replication forks, splayed arms, 5′- and 3′-flaps), as well as late recombination intermediates (e.g., nicked and intact Holliday junctions) (Figure 1A). Structural, biochemical and computational analyses of C. glabrata and T. terrestris Slx1-Slx4^CCD revealed that Slx1 has three different positively charged patches that bind branched DNA substrates, designated as sites I–III (Gaur et al., 2015; Gaur et al., 2019). Site I contains conserved residues from the arginine helix that engage dsDNA on one side of the branchpoint (i.e., the pre-nick DNA duplex). When reconstituted with Slx4^CCD, Slx1 site I mutants exhibited modest defects in substrate binding but showed severe catalytic defects (Gaur et al., 2015; Gaur et al., 2019). This suggests that site I is a low-affinity binding site that has an important role in orientating the substrate for catalysis. Site II contains residues from the glutamate helix that interact with ssDNA. When co-purified with Slx4^CCD, Slx1 site II mutants showed moderate to severe defects in substrate binding and catalysis (Gaur et al., 2015; Gaur et al., 2019). Finally, site III contains residues from three elements: the GIY-YIG hairpin loop, a short helix that connects the arginine helix to β3 and a flexible loop that connects the glutamate helix to α3. A comparison of the apo and DNA-bound structures of T. terrestris Slx1-Slx4^CCD showed that the loop becomes ordered in the presence of DNA (Gaur et al., 2019). The proximity of this region to the active site suggests that the ordering of site III is coupled to active site assembly. As such, site III seems to represent a molecular safety latch, ensuring that catalysis only occurs if the DNA substrate is bound correctly in sites I and III (Gaur et al., 2019). Importantly, Slx1-Slx4^CCD variants with mutations in sites II and III showed a weak affinity for DNA substrates and were catalytically inactive (Gaur et al., 2019). This indicates that substrate binding in site I alone is insufficient for enzyme activity. A further implication is that sites II and III are required to hold the DNA substrate in place and facilitate or guide incision site selection. Intriguingly, the orientation of site III relative to site I is such that the DNA substrate bends by approximately 50° for efficient catalysis (Gaur et al., 2019). Notably, the FEN1 and Mus81-Mms4/MUS81-EME1 structure-selective endonucleases also use DNA bending to achieve optimal positioning of the substrate in the active site (Tsutakawa et al., 2011; Gwon et al., 2014; Mukherjee et al., 2014).

In a recent development, Xu et al. published the crystal structures of S. cerevisiae Slx1 in complex with an Slx4 fragment containing the SAP and CCD domains (Slx1-Slx4^SAP+CCD), alone and bound to a 1-nt 5′-flap substrate (Xu et al., 2021). The overall structure of Slx1-Slx4^SAP+CCD was very similar in the apo and DNA-bound forms with one notable exception: the SAP domain was only visible in the presence of DNA, indicating that it adopted a more ordered conformation in the DNA-bound state. More specifically, the SAP domain folded into three helices, in which α2 and α3 packed against the Slx4 CCD domain (Xu et al., 2021). Although this structure captured Slx1-Slx4^SAP+CCD in a catalytically-inhibited conformation, it provided the framework for molecular dynamics simulations of Slx1-Slx4^SAP+CCD bound to a longer 5′-flap, which agree well with the model predicted for the T. terrestris Slx1-Slx4^CCD-DNA complex (Gaur et al., 2019; Xu et al., 2021). Together, these models reveal three critical regions of protein-DNA contacts: i) Slx1 site I binds the major groove of dsDNA on one side of the branchpoint (i.e., the pre-nick DNA duplex) and this is essential to place the scissile phosphate in the active site; ii) the Slx4 SAP domain binds the minor groove of the other DNA duplex (i.e., post-nick DNA duplex) approximately one turn away from the ss/ds-DNA branchpoint; and iii) Slx1 sites II and III engage the ssDNA flap, with site III becoming more critical for binding longer flaps. Curiously, the SAP domain mediates the interaction between mammalian SLX4 and MUS81 (Fekairi et al., 2009; Svendsen et al., 2009; Castor et al., 2013; Kim et al., 2013), suggesting that its function may have diverged from DNA-binding to protein-binding in some species. Another possibility is that the SAP domain uses distinct interfaces to accommodate DNA and protein substrates simultaneously.

One limitation to the models described above is that they do not account for the incision site selection of Slx1-Slx4; the phosphodiester bond that is closest to the active site does not match what is cleaved by Slx1-Slx4 in vitro (i.e., between nucleotides 3 and 4 on the 3′-side of the ss/ds-DNA junction) (Gaur et al., 2019; Xu et al., 2021). Further structural studies are needed to capture the enzyme in a catalytically productive conformation. Nevertheless, a catalytic mechanism has been proposed for Slx1. As mentioned above, the Slx1 active site residues are in the GIY-YIG hairpin, with the first motif (GIY) found in β1 and the second motif (YIG) found in β2 (Figure 4C). Although both motifs exhibit subtle sequence variation, the Tyr residues are strongly conserved between GIY-YIG nucleases. This led Gaur et al. (2015) to propose that the catalytic mechanism of Slx1 is the same as that described by Sokolowska et al. (2011) for the Hpy188I restrictase. Briefly, phosphodiester hydrolysis involves an in-line nucleophilic attack of the scissile phosphate. Catalysis requires a divalent metal ion in the active site. The tyrosine residue in the GIY motif (Y14 in C. glabrata Slx1) serves as a general base, acting as a proton acceptor for the nucleophilic water molecule (Sokolowska et al., 2011; Gaur et al., 2015). In C. glabrata Slx1, the guanidinium group of invariant R36 and the phenolic oxygen of Y26 stabilize the non-bridging oxygen of the scissile phosphate. A conserved glutamate residue (E79 in C. glabrata Slx1) anchors a single metal cation in the active site (Sokolowska et al., 2011; Gaur et al., 2015). The metal ion is located on the opposite side of the scissile phosphate and is thought to destabilize the substrate and stabilize the transition state.

The N-terminal extension of human SLX4 contains several domains or motifs that bind functionally diverse proteins, most of which have key roles in maintaining genome stability (Figure 5A). The MUS312-MEI9 interaction-like region, spanning residues 500–559, mediates the interaction between SLX4 and XPF-ERCC1 (Fekairi et al., 2009; Svendsen et al., 2009; Kim et al., 2013; Guervilly et al., 2015; Hashimoto et al., 2015). Residues L530, F545, Y546 and L550 are crucial for the SLX4-XPF interaction in yeast two-hybrid (Hashimoto et al., 2015) and coimmunoprecipitation assays (Guervilly et al., 2015; Zhang et al., 2019), suggesting that these residues form part of the binding interface. More recently, the MLR domain was shown to bind a protein called SLX4IP (for SLX4-interacting protein). We will revisit the structural anatomy and functions of these different SLX4-complexes in the final part of this review (Section 6.1).

Another important element of human SLX4 is the BTB domain, which is positioned downstream of the MLR domain and spans residues 660–795 (Figure 5A). The BTB domain is a widely distributed motif that participates in diverse cellular functions, ranging from transcriptional regulation to ion channel assembly and gating (Stogios et al., 2005). In general, BTB domains mediate self-oligomerization (e.g., homodimerization) or interactions with other proteins. Structural, biochemical and cell-based experiments indicate that the SLX4 BTB domain promotes homodimerization (Guervilly et al., 2015; Yin et al., 2016). The crystal structure of the BTB dimer revealed that each monomer adopted the characteristic ΒΤΒ fold—a cluster of six α-helices capped at one end by a four-stranded β-sheet (Figure 5B) (Yin et al., 2016). The dimer was held together by a hydrophobic interface with key contacts made between residues in α1 and α2, most notably residues F681 and F708 (Figure 5C). These residues are conserved in animals, suggesting that BTB-mediated SLX4 homodimerization may be a common theme amongst metazoans (Yin et al., 2016).

Notably, the BTB domain is needed for interstrand crosslink repair, which may reflect its role in promoting XPF SUMOylation and stabilizing the interaction between XPF and SLX4 (Andersen et al., 2009; Kim et al., 2013; Guervilly et al., 2015). As discussed in more detail below, the BTB domain also mediates several SLX4 functions in cells that maintain telomere lengths by the ALT (alternative lengthening of telomeres) pathway. These functions include the formation of subnuclear foci, the recruitment of SLX1 and XPF-ERCC1 to telomeres and telomere stability (Yin et al., 2016). Further studies are needed to elucidate how the BTB domain regulates the functions of different SLX4-nuclease complexes.

As alluded to above, SLX4 has an important role in maintaining telomere length and stability. This is not surprising, given that branched DNA structures are expected to occur frequently in telomeres because of their highly repetitive DNA sequence (for reviews, see Maestroni et al., 2017; Bonnell et al., 2021). Telomeres are essential nucleoprotein structures that form a protective cap for the terminal segments of linear chromosomes. In humans, the DNA component contains hundreds of hexameric repeats (5′-TTAGGG-3′) organized into a centromere-proximal double-stranded region and a single-stranded 3′-overhang. Several histone and non-histone proteins bind telomeric DNA. In mammalian cells, a group of six non-histone proteins (TRF1, TRF2, RAP1, POT1, TPP1 and TIN2) form a macromolecular protein complex called shelterin, which has a crucial role in telomere protection (reviewed in Palm and De Lange, 2008; Lim and Cech, 2021). TRF2 aids in the formation of DNA secondary structures called telomere loops (t-loops), in which the single-stranded overhang loops back on itself and the duplex region to form a lariat-like structure (Griffith et al., 1999; Doksani et al., 2013). The t-loop provides an architectural mechanism to “hide” the single-stranded overhang and prevent chromosome ends from being recognized as broken DNA (reviewed in Lim and Cech, 2021). Although the t-loop protects the chromosome end, this structure must be protected from other enzymes in the cell. For example, the base of the t-loop represents a branched DNA substrate that can be cleaved by SLX1-SLX4, leading to telomere loss. TRF2 has emerged as the gatekeeper that regulates the accessibility of telomeres to structure-selective endonucleases to prevent the unscheduled cleavage of branched DNA structures (Poulet et al., 2009; Saint-Leger et al., 2014; Sarkar et al., 2015; Schmutz et al., 2017).

One of the hallmarks of cancer cells is their replicative immortality, achieved by activating a telomere maintenance mechanism (Hanahan and Weinberg, 2011). Most cancer cells reactivate a reverse transcriptase called telomerase, but a minority use the ALT mechanism (Kim et al., 1994; Bryan et al., 1995; Bryan et al., 1997; Shay and Bacchetti, 1997; MacKenzie et al., 2021). ALT is a homology-directed recombination-based pathway that uses telomeric templates for DNA synthesis (Dunham et al., 2000; MacKenzie et al., 2021). The recruitment of SLX4 to ALT telomeres requires the BTB domain (discussed above), as well as the interaction with TRF2 (Fekairi et al., 2009; Munoz et al., 2009; Svendsen et al., 2009; Wan et al., 2013; Wilson et al., 2013; Yin et al., 2016). SLX4 uses a unique TBM to interact directly with TRF2 (Wan et al., 2013; Wilson et al., 2013). This motif includes residues 1,014–1,028, which are positioned downstream of the BTB domain in a largely disordered region of SLX4 (Figure 3A, 5A). Conversely, SLX4 binds to the TRF homology (TRFH) domain of TRF2 (Wan et al., 2013). The TRFH domain is a versatile docking site for various partner proteins (Giraud-Panis et al., 2018). The crystal structure of the SLX4^TBM-TRF2^TRFH complex shows that SLX4^TBM adopts an extended conformation with a short one-turn helix at the N-terminus and fits neatly into a narrow groove formed by the TRF2^TRFH (Figure 5D) (Wan et al., 2013). The interface is hydrophobic and includes three residues from SLX4^TBM (H1020, L1022, P1024) and one from TRF2^TRFH (F120). Through the interaction with TRF2, SLX4 and its associated nucleases, most notably SLX1, prevent telomeric DNA damage and fragility in ALT and non-ALT cells (Wilson et al., 2013; Saint-Leger et al., 2014; Sarkar et al., 2015; Yin et al., 2016; Yang et al., 2020). In ALT cells, SLX4 has an additional role in regulating telomere length. ALT telomeres are unusually long and heterogeneous and contain a greater abundance of TRF2 and SLX4 than non-ALT cells (Dejardin and Kingston, 2009; Svendsen et al., 2009; Wan et al., 2013; Wilson et al., 2013). Here, SLX4 promotes telomere recombination and negatively regulates telomere length, most likely via SLX1-mediated cleavage of recombination intermediates (Wan et al., 2013; Sarkar et al., 2015; Sobinoff et al., 2017). The actions of SLX4 and its associated endonucleases are essential for “productive ALT,” which refers to an equilibrium between pro- and anti-recombinogenic activities, such that telomere lengths are maintained at or near a threshold compatible with cell growth (Sarkar et al., 2015; Sobinoff et al., 2017; Panier et al., 2019; Robinson et al., 2020). We will revisit the ALT mechanism later in this review.

Human SLX4 has acquired the capacity to recognize ubiquitin through two tandem UBZ4 domains in the N-terminus (Figure 5A), referred to as UBZ-1 and UBZ-2, respectively (Fekairi et al., 2009; Kim et al., 2011; Lachaud et al., 2014). Ubiquitin is a small, ubiquitously expressed regulatory protein initially characterized as an essential component of the ATP-dependent proteolytic system (reviewed in Swatek and Komander, 2016). Today, the covalent modification of proteins with ubiquitin is understood to represent a versatile signaling event that impacts hundreds of cellular processes, including the recruitment and activity of DNA repair complexes (Tang et al., 2021).

The first clue that the SLX4 UBZ4 domains were physiologically relevant came from identifying Fanconi anemia patients carrying in-frame deletions of SLX4 that disrupted UBZ-1 and deleted UBZ-2 (Kim et al., 2011; Stoepker et al., 2011). Further studies showed that the UBZ domains were uniquely required for the survival of cells treated with DNA interstrand crosslinking agents (Kim et al., 2013; Ouyang et al., 2015). Around the same time, another group showed that UBZ domains of chicken DT40 SLX4 were required for cellular tolerance to interstrand crosslinking agents (Yamamoto et al., 2011). This study also revealed that the SLX4 was recruited to DNA damage-induced foci through an interaction between the UBZ domains and monoubiquitylated FANCD2. Notably, FANCD2 monoubiquitination traps the FANCI-FANCD2 complex on DNA, which likely regulates the recruitment of downstream proteins (Alcon et al., 2020; Rennie et al., 2020; Tan et al., 2020; Wang et al., 2020; Wang et al., 2021). However, it is still not clear whether the UBZ domains of human SLX4 bind monoubiquitinated FANCD2. First, in vitro ubiquitin-binding assays showed that the isolated UBZ domains bound K63-linked polyubiquitin chains but not monoubiquitin (Kim et al., 2011). Further work showed that UBZ-1 was necessary and sufficient for binding to polyubiquitin chains but exhibited weak binding to monoubiquitin (Lachaud et al., 2014). Second, SLX4 was efficiently recruited to interstrand crosslinks in cells that lacked FANCD2 or its E3 ubiquitin ligase RAD18 (Lachaud et al., 2014). Together, these data suggest that monoubiquitylated FANCD2 is not the primary ligand for SLX4 UBZ-1. Nevertheless, there are reports that SLX4 and FANCD2 colocalize in subnuclear foci under certain cellular conditions (Minocherhomji et al., 2015; Panichnantakul et al., 2021). Further studies are needed to determine the ubiquitylated ligand(s) that recruit SLX4 to DNA interstrand crosslinks.

Another interesting property of human SLX4 is its ability to interact non-covalently with SUMO via three SIMs (Figure 5A) (Guervilly et al., 2015; Ouyang et al., 2015). Like ubiquitin, SUMO is a key regulator of the DNA damage response and coordinates the recruitment of many DNA repair proteins (Morris and Garvin, 2017). Immunoprecipitation experiments showed that SLX4 captured SUMO-1 and SUMO-2/3 conjugates from cell extracts, but not free SUMO (Guervilly et al., 2015; Ouyang et al., 2015). Further work revealed that SLX4 preferentially bound SUMO-2/3 chains in vitro (Ouyang et al., 2015). In all cases, the SIM domains mediated SLX4-SUMO interactions (Guervilly et al., 2015; Ouyang et al., 2015). Molecular studies showed that the SIMs and UBZs have nonredundant roles in the suppression of chromosomal instability. While the UBZs have a critical role in DNA interstrand crosslink repair, the SIMs are required for the localization of SLX4 to telomeres and sites of DNA damage, as well as its role in the general replication stress response (Kim et al., 2013; Gonzalez-Prieto et al., 2015; Guervilly et al., 2015; Ouyang et al., 2015). Although the SUMOylated targets that SLX4 recognizes remain largely uncharacterized, two strong candidates are XPF and the shelterin protein RAP1 (Guervilly et al., 2015; Robinson et al., 2021). Future studies are needed to dissect the mechanism of XPF SUMOylation. We will revisit the mechanism and functions of SUMOylated RAP1 in the final section of this review.

Before concluding our discussion of SLX4 architecture, it is worth noting that Guervilly et al. (2015) showed that the SLX4 SIMs mediate its interaction with the SUMO E2-conjugating enzyme UBC9. Additionally, SLX4 protein levels were positively correlated with XPF SUMOylation in vivo (Guervilly et al., 2015). This raised the possibility that SLX4 was either a SUMO E3 ligase or an essential component of a macromolecular SUMO E3 ligase complex. Although Guervilly et al. (2015) did not detect E3 ligases in their SLX4 immunoprecipitates, recent work has captured interactions between SLX4 and several E3 ligases, including MMS21 and PIAS1 (Robinson et al., 2021, our unpublished data). This new data indicates that SLX4 is likely functioning as a scaffold to bridge the interactions between SUMO E2 and E3 enzymes and their targets.

Biochemical fractionation and coimmunoprecipitation experiments revealed that human SLX1-SLX4 is constitutively associated with XPF-ERCC1 throughout the cell cycle, forming the SX complex (Figure 6) (Wyatt et al., 2017; Panichnantakul et al., 2021). The SX complex is held together through direct contacts between the SLX4 MLR domain and the XPF N-terminal helicase-like domain (Fekairi et al., 2009; Svendsen et al., 2009; Kim et al., 2013; Guervilly et al., 2015; Hashimoto et al., 2015; Zhang et al., 2019). The interaction between SLX4 and XPF is critical for the repair of DNA interstrand crosslinks and a subset of Fanconi anemia patients harbor mutations in SLX4/FANCP (Kim et al., 2011; Stoepker et al., 2011; Schuster et al., 2013) or XPF/FANCQ (Bogliolo et al., 2013). Interstrand crosslinks are exceptionally toxic lesions that covalently link both strands of the double helix and prevent strand separation, which blocks DNA transcription and replication. These lesions result from reactive aldehydes produced during specific metabolic processes (e.g., alcohol metabolism), lipid metabolism, or exposure to chemotherapeutics (e.g., psoralen, cisplatin and mitomycin C). The chemistry and three-dimensional structure of an interstrand crosslink in duplex DNA influences downstream repair (for a review, see Housh et al., 2021). For example, the NEIL3 glycosylase is critical for the repair of abasic and psoralen-induced crosslinks, whereas the Fanconi anemia pathway repairs lesions caused by acetaldehyde, cisplatin and mitomycin C (Figure 7) (Semlow et al., 2016; Hodskinson et al., 2020; Li et al., 2020). The reader is referred to Semlow and Walter (2021) for an in-depth review of vertebrate interstrand crosslink repair.

The majority of interstrand crosslink repair occurs during S-phase, which is the phase of the cell cycle when the SX complex predominates (Figure 6) (Wyatt et al., 2017; Panichnantakul et al., 2021). In a cell-free system using Xenopus egg extracts, the interstrand crosslink is recognized when two replication forks converge (Figure 7) (Raschle et al., 2008; Zhang et al., 2015). Under certain conditions, the replication fork can collapse, which triggers the recruitment of the Fanconi anemia core complex (Figure 7) (Knipscheer et al., 2009; Wu et al., 2019). The core complex monoubiquitinates the FANCI-FANCD2 heterodimer, which facilitates the recruitment and coordination of DNA endonucleases (e.g., SLX1-SLX4, XPF-ERCC1), translesion synthesis polymerases (e.g., Rev1, pol ζ) and homologous recombination proteins (e.g., RAD51) (Meetei et al., 2003; Sims et al., 2007; Smogorzewska et al., 2007; Knipscheer et al., 2009; Ho et al., 2011; Long et al., 2011; Yamamoto et al., 2011; Klein Douwel et al., 2014; Budzowska et al., 2015; Rickman et al., 2015; Klein Douwel et al., 2017; Bezalel-Buch et al., 2020). Crosslink “unhooking” is a critical step of DNA interstrand crosslink repair (Figure 7). This reaction uses nucleases to introduce nicks on the 5′- and 3′-sides of the crosslink. After unhooking, one strand of the helix contains an adduct and a single-stranded gap that is filled in by error-prone translesion synthesis (Roy and Scharer, 2016). The other strand contains a double-strand break that is repaired through homologous recombination (Figure 7) (Michl et al., 2016).

The SLX4 scaffold has a critical role in crosslink unhooking, functioning to recruit XPF-ERCC1 to the lesion (Klein Douwel et al., 2014; Klein Douwel et al., 2017). The SLX4 UBZ domains are needed to recruit SLX4 to interstrand crosslinks (Yamamoto et al., 2011), as is the loading of FANCI-FANCD2 (Klein Douwel et al., 2014). Remarkably, biochemical experiments revealed that the SLX4-XPF-ERCC1 complex is sufficient for crosslink unhooking (Hodskinson et al., 2014; Klein Douwel et al., 2014; Klein Douwel et al., 2017; Hoogenboom et al., 2019). First, an N-terminal fragment of murine SLX4, spanning residues 1–758, was sufficient to stimulate crosslink unhooking by XPF-ERCC1 in vitro (Hodskinson et al., 2014). These experiments monitored the incision of DNA substrates containing mildly distorting lesions (i.e., nitrogen mustard). Second, the depletion of SLX4 or XPF from Xenopus egg extracts inhibited the unhooking and repair of cisplatin-induced interstrand crosslinks (Klein Douwel et al., 2014; Klein Douwel et al., 2017). Together, these studies provide strong biochemical evidence that SLX1 is dispensable for interstrand crosslink unhooking. Nevertheless, given the intricacies of interstrand crosslink repair, it will be important to determine whether SLX1-SLX4 and XPF-ERCC1 collaborate to unhook certain types of DNA interstrand crosslinks. Another possibility is that SLX1 resolves recombination intermediates that occur in the late stages of interstrand crosslink repair.

In addition to its role in interstrand crosslink repair, the SX complex may also be involved in processing 3′-ends in APE2-deficient cells. APE2 is an apurinic/apyrimidinic (AP) endonuclease that displays robust 3′-phosphodiesterase and 3′-5′ exonuclease activities but weak AP endonuclease activity (Hadi and Wilson, 2000; Burkovics et al., 2006). In human cells, APE2 has a critical role in reversing blocked 3′-ends to generate the free 3′-hydroxyl needed for DNA synthesis during homologous recombination (Mengwasser et al., 2019; Alvarez-Quilon et al., 2020). Interestingly, SLX4, XPF and ERCC1 are required for the viability of APE2-deficient cells (Alvarez-Quilon et al., 2020). This points towards a role for SX in processing unextendible recombination intermediates and cleaving near the branchpoint to remove the 3′-block. Further studies are needed to explore this intriguing model.

When human cells enter prometaphase, MUS81-EME1 is recruited to the SX complex, leading to the formation of a tri-nuclease complex called SMX (Figure 6) (Wyatt et al., 2013; Wyatt et al., 2017). The mechanism of SMX assembly involves direct interactions between the SLX4 SAP domain and the MUS81 N-terminal HhH domain (Fekairi et al., 2009; Nair et al., 2014, our unpublished data). Furthermore, SMX assembly requires phosphorylation of the SLX4 SAP and MUS81 N-HhH domains in early mitosis, events that are driven by the CDK1 and CK2 kinases, respectively (Duda et al., 2016; Palma et al., 2018). In silico experiments identified six potential CDK1 target sites in and around the SLX4 SAP domain (Duda et al., 2016). SLX4 mutants harboring alanine substitutions of these six residues failed to coimmunoprecipitate MUS81 (Duda et al., 2016). On the other hand, CK2 phosphorylated one residue in the MUS81 N-HhH domain, Ser87, and alanine substitution reduced the ability of MUS81 to coimmunoprecipitate SLX4 (Palma et al., 2018). Interestingly, phosphorylation of S87 increased in response to replication stress and in cells transitioning from G2 to prophase; phosphorylation of S87 disappeared as soon as cells entered metaphase (Palma et al., 2018).

Recombinant SMX is a promiscuous structure-selective endonuclease that cleaves a broad range of DNA secondary structures, including replication fork structures and recombination intermediates (Wyatt et al., 2017). The observation that SMX cleaves a range of branched DNA structures with high catalytic efficiency reveals an interesting dichotomy. On the one hand, it provides biochemical evidence for the roles of SMX in DNA repair and genome stability, which include: fragile site cleavage and mitotic DNA synthesis (Naim et al., 2013; Ying et al., 2013; Guervilly et al., 2015; Minocherhomji et al., 2015; Garribba et al., 2020), the resolution of recombination intermediates (Wechsler et al., 2011; Castor et al., 2013; Garner et al., 2013; Wyatt et al., 2013; Nair et al., 2014; Sarbajna et al., 2014) and telomere homeostasis (Wan et al., 2013; Wilson et al., 2013; Ouyang et al., 2015; Sarkar et al., 2015; Panier et al., 2019; Verma et al., 2019; Yang et al., 2020). On the other hand, the promiscuous nuclease activity of SMX makes it a potential threat to chromosome integrity, particularly during S-phase. This hints at the existence of one or more regulatory mechanisms to control SMX in vivo. Indeed, the temporally regulated assembly of SMX is critical for genome stability, and premature assembly leads to catastrophic DNA damage and chromosome pulverization. In these experiments, SMX was forced to assemble in S-phase because of constitutive CDK1 activity, which was achieved through chemical inhibition of the inhibitory phosphatase WEE1 (Duda et al., 2016) or genetic manipulation of Cdk1 (Szmyd et al., 2019). In another approach, the expression of a phosphomimetic MUS81 S87D mutant induced extensive DNA damage in S-phase, as well as pulverized chromosomes (Palma et al., 2018). The DNA damage was largely dependent on the nuclease activity of MUS81, consistent with the biochemical data showing that within the SMX complex, MUS81-EME1 is the predominant nuclease that cleaves replication fork structures (Duda et al., 2016; Wyatt et al., 2017; Palma et al., 2018; Szmyd et al., 2019). The observation that SLX4 relaxes the substrate specificity of MUS81-EME1 bears some resemblance to the promiscuous nuclease activity exhibited by SLX1-SLX4. Recent work has provided insight into the replication structures that could be targeted by the SMX complex in human cells. Specifically, van Wietmarschen et al. showed that SLX4 and MUS81-EME1 are involved in the cleavage of DNA secondary structures in expanded TA-dinucleotide repeats, resulting in double-strand breaks and chromosome shattering (Van Wietmarschen et al., 2020). Further work showed that the double-strand breaks resulted from the cleavage of replication forks stalled in expanded (TA) _n repeats (Van Wietmarschen et al., 2020). However, chromosome shattering was only observed in cancer cells that lacked the WRN helicase, a member of the RecQ family that melts DNA secondary structures, and exhibited microsatellite instability, which refers to the genetic hypermutability that results from impaired DNA mismatch repair (Van Wietmarschen et al., 2020). This indicates that stalled replication forks are normally rescued by the WRN helicase. In the absence of WRN, SLX4 and MUS81-EME1 (presumably within the SMX complex), cleave the stalled replication forks to trigger replication fork restart. However, the frequency of replication fork cleavage is so high that the result is chromosome fragmentation and cell death (Van Wietmarschen et al., 2020). Further work is needed to determine whether this phenotype results from the SMX complex per se and if so, to determine if the mechanism of SMX assembly is deregulated (i.e., S-phase instead of mitosis) in specific cellular contexts.

Altogether, the picture emerges that SMX represents a versatile nuclease toolkit to remove a plethora of branched DNA structures that could compromise accurate DNA replication or chromosome segregation. One crucial challenge for the future is determining the abundance of SX and SMX at different cell cycle stages. For example, although we assume that all of the SLX1-SLX4 and MUS81-EME1 in early mitotic cells exists within the context of SMX, this has not been evaluated scientifically. It will be equally important to develop cell models that allow researchers to distinguish between SX- and SMX-dependent functions.