A prototypical Fanconi anemia pathway in lower eukaryotes?

Abstract

DNA interstrand cross-links (ICLs) present a major challenge to cells, preventing separation of the two strands of duplex DNA and blocking major chromosome transactions, including transcription and replication. Due to the complexity of removing this form of DNA damage, no single DNA repair pathway has been shown to be capable of eradicating ICLs. In eukaryotes, ICL repair is a complex process, principally because several repair pathways compete for ICL repair intermediates in a strictly cell cycle-dependent manner. Yeast cells require a combination of nucleotide excision repair, homologous recombination repair and postreplication repair/translesion DNA synthesis to remove ICLs. There are also a number of additional ICL repair factors originally identified in the budding yeast Saccharomyces cerevisiae, called Pso1 though 10, of which Pso2 has an apparently dedicated role in ICL repair. Mammalian cells respond to ICLs by a complex network guided by factors mutated in the inherited cancer-prone disorder Fanconi anemia (FA). Although enormous progress has been made over recent years in identifying and characterizing FA factors as well as in elucidating certain aspects of the biology of FA, the mechanistic details of the ICL repair defects in FA patients remain unknown. Dissection of the FA DNA damage response pathway has, in part, been limited by the absence of FA-like pathways in highly tractable model organisms, such as yeast. Although S. cerevisiae possesses putative homologs of the FA factors FANCM, FANCJ and FANCP (Mph1, Chl1 and Slx4, respectively) as well as of the FANCM-associated proteins MHF1 and MHF2 (Mhf1 and Mhf2), the corresponding mutants display no significant increase in sensitivity to ICLs. Nevertheless, we and others have recently shown that these FA homologs, along with several other factors, control an ICL repair pathway, which has an overlapping or redundant role with a Pso2-controlled pathway. This pathway acts in S-phase and serves to prevent ICL-stalled replication forks from collapsing into DNA double-strand breaks.

DNA is constantly challenged by agents that can inflict structural damage either on one or both strands of the DNA duplex. Lesions affecting one strand present a limited challenge, as the complementary undamaged DNA strand provides the repair machinery with a template to accurately restore the original sequence in the damaged strand. By contrast, lesions affecting both DNA strands are far more deleterious, as they remove the template for cut-and-patch repair reactions. This class of DNA lesions includes DNA interstrand cross-links (ICLs), complex DNA lesions that are extremely toxic owing to the covalent linkage between the two DNA strands. Such linkage prevents DNA strand separation during all fundamental metabolic processes ongoing on cellular DNA.¹

As ICLs can arise endogenously, cells have adapted to this threat throughout evolution by the co-ordination of several repair activities in a complex network, termed ICL repair. The first models of ICL repair emerged from studies in Escherichia coli, where the sequential action of nucleotide excision repair (NER) and homologous recombination repair (HRR) appears to be sufficient to repair ICLs in an error-free manner.² When HRR cannot act during this process, as a consequence of genetic inactivation of the pathway or unavailability of a homologous DNA template, an alternative pathway, dependent on DNA polymerase II (Polβ)-mediated translesion DNA synthesis (TLS), is utilized.³

Although the basic mechanisms and biochemical activities involved in ICL repair appear to be conserved throughout evolution, the overall process is more complex in eukaryotic cells. First, there is considerable redundancy in certain ICL repair activities in eukaryotic cells, and these need to be strictly coordinated. Second, cell cycle phase directs the use of different ICL repair mechanisms and this is dependent on substrate availability. Finally, genome organization per se contributes to an increased sophistication of the ICL repair process in eukaryotic cells. In Saccharomyces cerevisiae, important roles for NER, HRR and postreplication repair (PRR)/TLS in ICL repair have all been identified. Additionally, a number of factors that affect ICL sensitivity, but not sensitivity to other types of DNA damage, have been identified in genetic screens.^4,5 Of these, Pso2 (previously referred to as Snm1, but subsequently renamed due to a naming conflict)⁶ is of greatest interest, as pso2 mutant cells are specifically and highly sensitive to 8-methoxypsoralen photoaddition⁷ and nitrogen mustard,^8,9 two classic ICL-inducing treatments, but not to any other DNA damage types tested to date. Pso2 is thought to act downstream of ICL incision, which is primarily controlled by the NER apparatus in yeast.^10-12 The Pso2 protein is the founder member of family of eukaryotic proteins with a conserved domain structure containing two highly conserved motifs: the metallo-β-lactamase (MBL) structural domain (found in hydrolytic enzymes with wide-ranging functions from antibiotic detoxification to nucleic acid metabolism) and the β-CASP [a conserved motif limited to CPSF, an mRNA processing factor; Artemis, a nuclease acting in V(D)J recombination and non-homologous end-joining—NHEJ; SNM1; Pso2] domain.^13-16 This domain architecture is typical for β-CASP family members, representing a subfamily of canonical MBLs, and suggested that the Pso2 protein may have nuclease activity. Indeed, biochemical data has shown that Pso2 possesses both 5^' to 3^' exonuclease activity^17,18 and a structure-specific endonuclease activity that is able to open DNA hairpins.¹⁸ Both MBL and β-CASP domains are essential for the activity of Pso2, and catalytic site inactivating mutations in these domains (D252A and H611A substitutions, respectively) abolish nuclease activity. In addition, cells carrying these mutations display null-mutant phenotype regarding repair of ICLs¹⁹ and of transposon-induced DNA hairpins.¹⁸ Hairpin opening activity may suggest a novel role for Pso2 outside ICL repair, although the relevant physiological processes that generate DNA hairpin substrates for Pso2 remain to be elucidated.^4,18,20

There are three Pso2 orthologs in vertebrates, SNM1A, SNM1B/Apollo and SNM1C/Artemis.²¹ Like Pso2, SNM1A and SNM1B display 5^' to 3^' exonuclease activity, whereas SNM1C possesses endonuclease activity, cleaving 5^' and 3^' overhangs and hairpins.²² SNM1C probably does not play a major role in ICL repair and rather acts in pathway required for the processing of a subset of DNA double-strand breaks (DSBs) induced by ionizing radiation prior to rejoining by NHEJ.²³ However, both SNM1A and SNM1B depletion increase sensitivity to ICL-inducing agents such as mitomycin C,^21,24 while the latter seemingly sensitizes the cells to a broader spectrum of DNA damage.²⁵ Notably, co-disruption of SNM1A and SNM1B in chicken DT40 cells leads to an additional increase in ICL sensitivity compared with either single disruptant, suggesting that SNM1A and SNM1B have different or redundant functions in ICL repair.²⁴ SNM1B is believed to promote DSB formation in response to ICLs, possibly through the collapse of replication forks following ICL treatment,²⁶ while SNM1A suppresses ICL-induced DSBs during replication.^12,25 Studies aimed at defining the genetic relationship between SNM1-dependent and -independent ICL repair pathways strengthened the notion that SNM1A and SNM1B have diverse roles in ICL repair. SNM1B has been shown to act epistatically with the central Fanconi anemia (FA) factor FANCD2,²⁷ indicating that this Pso2 ortholog functions within the FA pathway during ICL repair, and we have reported collaboration between SNM1A and XPF-ERCC1 in initiating ICL repair in replicating cells.¹² The major functional homolog of yeast Pso2 appears to be SNM1A, as expression of SNM1A in pso2 disruptant cells can partially complement ICL sensitivity and elevated ICL-associated DSBs.²⁸

In vertebrates, a major DNA damage response pathway triggered by ICL exposure involves proteins mutated in FA. Mutations in FA pathway lead to a rare autosomal recessive disorder characterized by defects in all stages of ICL repair.^29,30 These repair defects are associated with congenital malformations, progressive bone marrow failure, profound genomic instability and a highly elevated risk of hematological malignancies and solid tumors.³¹ The disease is genetically heterogeneous, and patients with mutations in 15 genes have been reported to show traits associated with FA (FANCA, B, C, D1/BRCA2, D2, E, F, G, H, I, J/BACH1/BRIP1, L, M, N/PALB2, P/SLX4/BTBD12 and O/RAD51C). Mutations in these genes have been found in over 95% of all known FA patients, while cultured cells with defects in a number of other genes, including FAAP24, FAAP100, FAN1, MHF1, MHF2, exhibit FA-like defects in ICL repair. It has been proposed³² that the FA pathway is regulated by three factors, FANCD2-FANCI, SLX4 and RAD18, which act both in parallel and in concert. FANCD2-FANCI and SLX4 likely play a key role in controlling the incision step of ICL repair, with the latter an attractive candidate for a regulator of the recruitment of dedicated nucleases. The role for FANCD2-FANCI in this process still remains largely unknown, although further examination of the relevance of the physical association of FANCD2 with FAN1 (Fanconi anemia-associated nuclease 1) may provide clues. RAD18 controls the FA pathway by regulating FANCD2 monoubiquitylation and chromatin loading of FANCD2-FANCI. Additionally, it contributes to targeting SNM1A to stalled replication forks via monoubiquitylating proliferating cell nuclear antigen (PCNA) and mediating its direct interaction with this nuclease at ICL-stalled replication forks, suggesting it might be an important node for co-ordinating ICL repair factors acting in parallel or sequentially. It is worth noting that Rad18-dependent FA pathway activation appears not to be ICL-specific, but also in response to other DNA damage types.³³

Among the known FA factors, FANCA, FANCB, FANCC, FANCE, FANCF, FANG, FANCL, FANCM, two FA-associated proteins (FAAP24 and FAAP100) and FANCM-associated histone-fold proteins, MHF1 and MHF2, constitute what has become known as the FA “core complex.” This core complex is activated upon ICL treatment during S-phase and acts as a ubiquitin E3 ligase, monoubiquitylating FANCD2 and FANCI,^30,34 which ultimately leads to their retention on chromatin.³⁵ ICL damage recognition in this cell cycle phase requires FANCM,³⁶ a protein capable of binding to structures generated when replication forks encounter an ICL. FANCM belongs to the ERCC1/XPF(ERCC4) family of structure-specific DNA binding proteins³⁷ and contains two conserved domains: an apparently inactive ERCC4 nuclease domain that resides at the C terminus and is responsible for binding to branched DNA structure in vitro,³⁸ and an internal domain that is required for interaction with the FA core complex.³⁹ FANCM forms a stable heterodimer with another protein containing an ERCC4 domain, FAAP24.³⁸ This heterodimer mediates recruitment of the FA core complex to chromatin. Due to its interaction with MHF1 and MHF2, FANCM itself appears to be constitutively associated with chromatin. FANCM functions early in the FA pathway, following stalling of a replication fork at an ICL. It is thought that ICL-induced fork collapse can be prevented by FANCM-mediated fork regression or stabilization, and indeed, in vitro data indicates that FANCM can promote reversal of replication forks via concerted displacement and annealing of the nascent DNA strands.⁴⁰ Multiple nucleases are also recruited at an early step during S-phase-specific ICL repair. The FAN1 nuclease is recruited to stalled replication forks by monoubiquitylated FANCD2 and possesses 5^' flap and 5^' to 3^' exonuclease activities. Weaker activity, at replication fork structures and the cleavage of DNA opposite a nicked strand, has also been demonstrated, as recently recapitulated in references 32 and 41. SLX4 is recruited to stalled replication forks, directly owing to its tandem ubiquitin-binding zinc finger domains,^42,43 recruiting MUS81-EME1, XPF-ERCC1 and SLX1 nucleases as part of the SLX4 complex.⁴⁴ MUS81-EME1 and XPF-ERCC1 are members of the XPF/MUS81 family of structure-specific endonucleases, while SLX1 is a structurally distinct endonuclease possessing an N-terminal UvrC-intron-endonuclease and a C-terminal plant homeodomain-type zinc finger domains. Biochemically, MUS81-EME1 cleaves 3^' flaps, replication fork structures, Holliday junction (HJ) resolution intermediates and splayed arms, and XPF1-ERCC1 acts on splayed-arms, bubbles, stem-loop structures and 3^' flaps. SLX1 in complex with SLX4 exhibits hydrolytic activity toward 5^' flaps, stem loops, replication fork structures and HJs, as summarized in references 32 and 41. The coordinated action of FAN1, MUS81-EME1, XPF-ERCC1 and SLX1 nucleases is thought to result in ICL unhooking, although which nucleases create the critical incisions remains unconfirmed. Subsequently, TLS polymerases ensure lesion bypass past the unhooked cross-link. HRR is required to complete repair and DNA replication can subsequently be re-established. In late S-phase, there is a greater probability that two replication forks could converge on a single ICL. Studies using a Xenopus cell-free extract system, in conjunction with plasmid substrates bearing site-specific ICLs, suggest that following unhooking incisions, TLS polymerases extend the converging leading strand past the exonucleolytically processed ICL lesion.

The FA pathway had initially been thought to be restricted to vertebrates. However, the identification of some individual FA counterparts in other metazoans such as Caenorhabditis elegans^45-47 and Drosophila melanogaster,^48-50 as well as in both budding⁵⁰ and fission⁵¹ yeast, challenges this concept. Nevertheless, the whole pathway in its entirety does not seem to exist outside vertebrates. In particular, the FA core complex factors are not clearly identifiable or annotated outside this kingdom. It would appear that lower eukaryotes do, however, possess a “stripped-down” pathway, which lacks the majority of the FA core complex apart from two components, FANCM and FANCL. Three other members of the FA pathway are also well-conserved, FAND1, FANCD2 and FANCJ. The question then becomes how does this substantially stripped-down version of the FA pathway, lacking the core complex, operate? Solving this question would clearly accelerate progress in our understanding of the mechanistic basis of the ICL repair defect in FA.

Saccharomyces cerevisiae is a highly attractive model system owing to its genetic tractability, which has revealed many important biological insights, including within the field of eukaryotic DNA repair. NER, HRR and PRR/TLS have all been suggested to act in the yeast ICL repair, with a significant contribution from a pathway controlled by Pso2, following the initial ICL incisions produced, in all likelihood, by NER factors.¹ Budding yeast cells also possess putative homologs of FANCM,^34,52-55 FANCJ and FANCP^54,55 (Mph1, Chl1 and Slx4, respectively) as well as of the FANCM-associated histone-fold proteins MHF1 and MHF2 (Mhf1 and Mhf2).^54,56 Until recently, the existence of the FA-like ICL repair in this organisms has been discounted, since the disruption of these FA-like factors does not sensitize yeast cells to ICL-inducing agents. However, work from our own laboratories and that of K.J. Myung report that these factors constitute an S-phase-specific branch of the ICL repair pathway, obscured by the pathway or pathways controlled by Pso2 and Srs2. Furthermore, we have shown that other factors including Mgm101, a mitochondrial genome maintenance factor that also now appears to play a role in nuclear repair, Mhf1-Mhf2, a complex having heterotetrameric architecture similar to that of the histones (H3-H4)₂ heterotetramer, the MutSα (Msh2-Msh6) mismatch repair complex, Exo1, an exonuclease with multiple roles in DNA repair, Smc5-Smc6, a complex playing a role in chromosome organization and dynamics, and PCNA (Pol30), involved in many metabolic processes ongoing on DNA, are also required for this pathway.^54,55 It seems that this pathway primarily prevents ICL-stalled replication forks from collapsing into unrepairable DSBs. Rad5, an E3 ubiquitin ligase associated with the error-free branch of PRR, directs this FA-like pathway, possibly to restart replication by a fork reversal mechanism. Importantly, this pathway acts independently of the Pso2-controlled pathway,^54,55,57 suggesting that there might be a functional overlap between Pso2 and Exo1 where both are able to exonucleolytically degrade the tethered oligonucleotide associated with an unhooked ICL. Furthermore, the pathway is independent of Rad6-Rad18 error-prone branch of PRR.⁵⁴

Based on data available and on our own and the Myung group’s new findings,^54,55 we propose a modified model for ICL repair in replicating yeast cells (Fig. 1). NER may initiate ICL processing in S-phase yeast cells prior to the arrival of the fork, since it is well established that ICLs are efficiently incised in G₁ phase yeast cells. Therefore, the pathway controlled by NER and Pso2 might not be replication-coupled in the classical sense, or might operate both away from and at stalled replication forks to repair ICLs in S-phase (Fig. 1A). In pso2 cells, where the FA-like pathway now becomes necessary, the evidence suggests that replication stalls at the ICL and, in the absence of efficient nucleolytic processing by Pso2, produces a signal which Rad5 responds to. As a consequence, Rad5 polyubiquitinates PCNA and causes recruitment of Mph1 to site of the ICL. Mph1 stabilizes or regresses the stalled fork (regression is shown), and Mgm101, Smc5-Smc6 and Mhf1-Mhf2, likely representing Mph1 accessory factors, may help to protect and stabilize the structure of the generated ICL repair intermediate. Msh2-Msh6 (MutSα) also participates in this pathway, potentially acting to sense the aberrant DNA structure at the fork. Subsequently, Exo1 is recruited, possibly by its association with MutSα, and digests the tethered oligonucleotide to produce a substrate for downstream processing events and gap-filling. Gap-filling is achieved by TLS, and DNA replication is subsequently restored by HRR (Fig. 1B). In cells disabled for both Pso2-controlled and FA-like pathways, the replication fork collapses into an unrepairable DSB due to an inability to further process the cross-linked structure remaining after initial incision generated by NER, ultimately leading to cell death (Fig. 1C).

Figure 1. Proposed model for ICL repair in replicating Saccharomyces cerevisiae cells. (A) A situation in wild-type cells, where Pso2-controlled and FA-like pathways have overlapping or redundant roles in ICL repair, with a major contribution of the former. (B) In pso2 cells, FA-like pathway constitutes branch of the ICL repair pathway, which prevents replication forks from collapsing into DSBs. In addition, this pathway protects and stabilizes the generated ICL repair intermediate structure. (C) In the absence of both pathways, replication forks collapse into unrepairable DSBs upon ICL treatment leading to cell death. For more details, see text.

Mechanistically, this pathway is reminiscent of FA pathway in mammals, indicating that functional conservation of the FA pathway probably exceeds the kingdom of vertebrates despite an apparent lack of FA core complex factors. In both yeast and mammalian pathways, Mph1/FANCM-mediated fork regression or stabilization induced by DNA replication stalled at the ICL is an apparently sine qua non step, suggesting it might be essential for replication-coupled ICL repair. Such steps have been suggested to protect ICL repair intermediates from inappropriate repair and/or processing.⁵⁴