The multifunctional SNM1 gene family: not just nucleases

Abstract

The archetypical member of the SNM1 gene family was discovered 30 years ago in the budding yeast Saccharomyces cerevisiae. This small but ubiquitous gene family is characterized by metallo-β-lactamase and β-CASP domains, which together have been demonstrated to comprise a nuclease activity. Three mammalian members of this family, SNM1A, SNM1B/Apollo and Artemis, have been demonstrated to play surprisingly divergent roles in cellular metabolism. These pathways include variable (diversity) joining recombination, nonhomologous end-joining of double-strand breaks, DNA damage and mitotic cell cycle checkpoints, telomere maintenance and protein ubiquitination. Not all of these functions are consistent with a model in which these proteins act only as nucleases, and indicate that the SNM1 gene family encodes multifunctional products that can act in diverse biochemical pathways. In this article we discuss the various functions of SNM1A, SNM1B/Apollo and Artemis.

Approximately 30 years ago, independent genetic screens conducted in Saccharomyces cerevisiae for hypersensitivity to the DNA interstrand cross-linking (ICL) agents nitrogen mustard and psoralen identified the founding member of a small but ubiquitous gene family referred to as sensitivity to nitrogen mustard (SNM1) or sensitivity to psoralen (PSO2) [1–7]. An intriguing aspect of the yeast SNM1/PSO2 mutants is that they were singularly hypersensitive to bi- and polyfunctional alkylating compounds, while exhibiting an essentially wild-type level of sensitivity to monofunctional alkylating agents, ionizing radiation (IR) or UV. This unique phenotype led a number of laboratories to search for mammalian homologs of SNM1/PSO2, and five such genes were identified, including SNM1A, SNM1B/Apollo, Artemis, CPSF73 and ELAC2. Both CPSF73 and ELAC2 have roles in RNA processing [8–10], while the other three members have various roles in DNA metabolism and cellular stress responses. It is the functions of SNM1A, SNM1B/Apollo and Artemis that are the subject of this article.

Molecular cloning of the yeast SNM1/PSO2 gene and its mammalian homologs demonstrated that the common features of this family are a metallo-β-lactamase (MBL) fold and an appended β-CPSF–Artemis–SNM1–PSO2 (CASP) domain [11,12], which together are referred to as the SNM1 domain. The β-CASP domain is unique to the SNM1 gene family, and is predicted to be a nucleic acid binding domain, which, together with the MBL hydrolase has been demonstrated to possess a nuclease activity in the SNM1 proteins [13–21]. More detailed discussions of the structure of the SNM1 domain and its function as a nuclease can be found in recent reviews [5,22]. Outside of the SNM1 domain the sequence of each of the proteins is distinct (Figure 1), suggesting that each of the mammalian homologs may have a unique function. The yeast SNM1/PSO2 gene is, under normal conditions, poorly transcribed, however, it can be induced approximately fourfold upon treatment with agents that induce ICLs, but not by monofunctional alkylating agents or the UV mimetic agent 4-nitroquinoline 1-oxide [23]. Interestingly, induction of SNM1/PSO2 is dependent upon the DUN1 gene, which is known to be required for transcription of genes induced by yeast cell cycle checkpoint pathways activated by DNA damage or replication blocks [24]. Molecular analysis of SNM1 mutants treated with either psoralen or nitrogen mustard demonstrated that incision near crosslinks, and also possibly excision, proceeded normally in these mutants. However, a late step in crosslink repair that was required for restoration of high molecular weight DNA after the initial repair processes appeared to be defective [25–28]. When replication forks encounter a barrier such as an ICL, a common observation is that the fork will collapse resulting in the formation of a one-sided double strand break (DSB) [29]. It is apparently the resolution of this unique substrate that contains a DSB and a closely associated ICL that is defective in SNM1/PSO2 mutants. Repair of DSBs created by IR or that occur in mating type switching are not defective in SNM1/PSO2 mutants.

Figure 1. Structural domains of mammalian SNM1 proteins and identified interacting partners.

Domains are described in the text. Horizontal lines indicate identified regions of interaction with the proteins.

β-CASP: β-CPSF–Artemis–SNM1–PSO2; DNA-PKcs: DNA-dependent protein kinase catalytic subunit; MBL: Metallo-β-lactamase; MRN: MRE11-RAD50-NBS1; PBZ: Poly(ADP-ribose)-binding zinc finger; SCD: (Serine/threonine)–glutamine cluster domain; SNM1: Sensitivity to nitrogen mustard.

Interestingly, an epistasis analysis has demonstrated that there are only three extant pathways of ICL repair in S. cerevisiae. These pathways are defined by RAD52, REV3 and SNM1/PSO2, indicating that they represent homologous recombination, translesion bypass synthesis and an unknown mechanism presumably involved in restoring ICL-induced DSBs [25,30]. Break-induced replication is a well-recognized pathway in S. cerevisiae known to repair one-sided DSBs via homologous recombination. In fact, there are two demonstrated subpathways of yeast break-induced replication, one dependent on RAD51 and one independent of RAD51 [31–33]. It is possible that SNM1/PSO2 participates in the latter of these pathways, which might account for the results of the epistasis analysis. Snm1/Pso2 has also been demonstrated to have an overlapping role with the 5′–3′ mismatch repair exonuclease Exo1 during processing of collapsed replication forks via homologous recombination [27]. It was also demonstrated that Snm1/Pso2 has a separate role in G1 phase repair of ICLs that requires the nucleotide excision repair pathway, but not homologous recombination. Thus, Snm1/Pso2 appears to play a role in both homology-dependent and homology-independent pathways of ICL repair in S. cerevisiae, although its precise function in either of these pathways has remained elusive.

SNM1A

The first reported study of mammalian SNM1A was by Dronkert et al. who disrupted this gene in mice and demonstrated that homozygous −/− mice developed normally, but exhibited an increased mortality compared with wild-type littermates after injection of the crosslinking agent mitomycin C (MMC) [34]. Mouse embryonic stem cells that were homozygous −/− for SNM1A exhibited a twofold sensitivity to this drug compared with wild-type cells, but an increased sensitivity to other ICL agents or to IR was not observed. Two other knockouts of SNM1A in mice have also been reported. In a study conducted by our laboratory we observed a very similar phenotype in terms of normal development and sensitivity to DNA damaging agents as reported by Dronkert et al. [34]. However, we also observed a decreased lifespan, owing primarily to accelerated tumorigenesis, primarily lymphomas, adenomas and a curious propensity for bacterial glandular infections, particularly in male animals [35]. A cross of the SNM1 mutant mice with Trp53 nullizygous animals caused a decrease in survivability owing to tumorigenesis compared with that observed in either single mutant. Taken together, these findings suggested that SNM1A acts as a tumor suppressor. Along these lines, polymorphisms in human SNM1A have been associated with small-cell lung carcinoma risk [36]. However, a third reported knockout of SNM1A indicated a normal lifespan with no observed accelerated tumorigenesis [37]. These conflicting results in tumor incidence may have been owing to strain differences, or in the nature of the disruption in the SNM1A gene. Interestingly, a cross with FancD2 null animals caused a perinatal lethality suggesting that SNM1A and FANCD2 function in parallel pathways that are essential for normal development in the other’s absence [37]. Finally, while these studies indicated that SNM1A nullizygous mice exhibited only a minor sensitivity to MMC, homozygous deletion of SNM1A in chicken DT40 cells demonstrated that the mutants were sensitive to both MMC and cisplatin [38,39]. In addition, of the three mammalian SNM1 genes involved in cellular stress responses, only SNM1A was able to rescue the cellular hypersensitivity of the yeast SNM1 mutant to crosslinking agents [17]. In total, these studies clearly suggest that SNM1A has a role in mediating resistance to ICLs, although this role may be masked to some extent in mammalian cells by a possible partially redundant function with SNM1B/Apollo, as was observed in chicken DT40 cells [38].

Cellular studies of SNM1A have demonstrated that it is a poorly expressed nuclear protein (it possesses a nuclear localization signal) that in unstressed cells is largely confined to one or two regions or bodies (SNM1 bodies) in the mammalian nucleus [40]. SNM1 bodies vary somewhat in size but are typically approximately 2 µm in diameter and irregular in shape. These bodies are regulated as a function of the cell cycle, appear largely in the G1 phase, and are observed far less frequently in S and G2 phases. They do not appear to coincide with other known structures, such as promyelocytic leukemia protein or Cajal bodies. SNM1 bodies may be sites of sequestration of SNM1A to protect cells from the inappropriate activity of the protein since, as discussed below, overexpression of SNM1 is highly toxic to mammalian cells. Interestingly, 53BP1 is also localized to these bodies in unstressed cells, and co-immuno-precipitation (co-IP) experiments demonstrated that SNM1A and 53BP1 interact in vivo. Upon exposure to DNA damaging agents such as IR or crosslinking agents, both SNM1A and 53BP1 relocate to the sites of DNA damage, where they colocalize in foci with other DNA damage response (DDR) proteins such as MRE11 and BRCA1. Interestingly, there is little colocalization between SNM1A and the homologous recombination protein RAD51 after DNA damage, suggesting that these proteins function in distinct pathways as suggested by findings in budding yeast described above. In a yeast two-hybrid screen, SNM1A was demonstrated to interact with the sumo E3 ligase, PIAS1, and the proteins exhibited colocalization in IR-induced foci (IRIF) [38]. Recently, PIAS1 and PIAS4 have been demonstrated to sumoylate BRCA1 and 53PB1, respectively, at sites of DNA damage, and to be required for efficient repair of DSBs [41,42]. These findings suggest the possibilities that SNM1A is either a substrate of the PIAS1 ligase or possibly facilitates the sumoylation of BRCA1 and 53BP1.

Overexpression of SNM1A results in the formation of numerous nuclear foci in untreated mammalian cells. However, these foci are unlikely to be sites of DNA damage since they do not stain positive for 53BP1 [40], although, they do exhibit colocalization of PIAS1 [38]. Rather, these foci may be sites of sequestration of SNM1A akin to the function of the SNM1 bodies. Even low levels of overexpression of SNM1A are toxic to mammalian cells, and lead to a potent G1 phase cell cycle arrest followed by apoptosis [40]. This phenomena has greatly limited studies of SNM1A in regard to its role in the cell cycle, and the effects of mutations on its functions in vivo. Interestingly, over expression of human SNM1A is apparently not toxic to yeast cells [17].

Although endogenous SNM1A readily locates to sites of DSBs, its function at these sites remains unclear. Disruption of SNM1A in mouse cells indicated that it has little or no role in mitigating sensitivity to IR, suggesting that it is either not involved in the repair of DSBs, or that it plays a redundant role in this process, possibly with SNM1B/Apollo. Localization of SNM1A to IRIF is dependent upon the checkpoint kinase ATM, and this kinase can phosphorylate SNM1A in vitro [43]. Furthermore, loss of SNM1A results in a defective G1 checkpoint after exposure to IR. These findings suggest that SNM1A may play a role in cellular signaling in response to DNA damage, possibly through processing of damaged DNA mediated by its nuclease activity [17,18].

The human SNM1A gene has an unusual genomic structure with a long 5′ untranslated region that contains numerous open reading frames, and which is predicted to fold into a complex secondary structure. Using a reporter gene assay, this structure was demonstrated to act as an internal ribosome entry site (IRES) [44]. IRESs are structures that allow cap-independent translation of mRNAs, and many mammalian genes that contain an IRES are upregulated during mitosis [45,46]. This also proved to be true for the SNM1A IRES. During interphase the SNM1A IRES reduced expression of a green fluorescent protein reporter gene by approximately tenfold, while a two- to four-fold increase in expression was observed during mitosis. These observations led us to examine whether SNM1A has a function during mitosis, which resulted in the discovery that SNM1A participates in an early mitotic checkpoint in response to spindle poisons such as nocodazole [47]. This early mitotic checkpoint referred to variously as the prophase or antephase checkpoint was first discovered in mammalian cells through the characterization of the checkpoint with forkhead-associated (FHA) and really interesting new gene (RING) domains (CHFR) [48]. CHFR-deficient cells exhibit a failure to arrest the cell cycle in prophase prior to chromosome condensation upon exposure to spindle poisons, which is distinct from the more extensively studied spindle checkpoint that regulates the metaphase to anaphase transition [49–52]. Similar to CHFR-deficient cells, SNM1A^−/− mouse embryonic fibroblasts (MEFs), when exposed to spindle poisons, exhibited a failure to arrest in mitosis prior to chromosome condensation, elevated levels of micronuclei formation, a decreased mitotic delay, supernumerary centrosomes and decreased viability [47]. In addition, both SNM1A and 53BP1, which as discussed above are interacting partners, were demonstrated to co-IP with components of the anaphase-promoting complex (APC)/cyclosome, further indicating a mitotic function for SNM1A. It should be noted, however, that 53BP1-deficient cells do not exhibit a significant defect in the prophase checkpoint [Akhter et al., Unpublished Data]. The exact biochemical function of SNM1A in the prophase checkpoint remains unclear. In fact, the molecular details of the prophase checkpoint pathway in general have yet to be resolved, although, CDK1-cyclin B appears to be the ultimate target. Recently, a novel poly(ADP-ribose)-binding zinc finger motif was found in a small number of eukaryotic proteins involved in the DDR and checkpoint regulation [53]. These proteins include CHFR, aprataxin PNK-like factor, and SNM1A. Mutations in the poly(ADP-ribose)-binding zinc finger domain of CHFR abrogates its function at the prophase/antephase checkpoint. The role of this domain in SNM1A function has yet to be characterized primarily owing to the problems with its overexpression discussed above. Finally, whether or not SNM1A and CHFR function in the same checkpoint pathway is unclear. The phenotypes of the respective deficient cells are highly similar, however, despite repeated attempts, researchers have not been able to demonstrate a physical association between these proteins.

In summary, SNM1A appears to function in multiple pathways in response to various forms of cellular stress such as DNA damage and spindle poisons. However, its fundamental activity in these pathways remains unresolved, particularly since its role as a nuclease does not appear to explain all its functions.

SNM1B/Apollo

Only a handful of studies have been conducted on the function of SNM1B/Apollo in vertebrate cells. Concannon and colleagues were the first to examine its role in the DDR in human cells [54]. Using siRNA-mediated knockdown, they demonstrated that cells depleted of SNM1B/Apollo were moderately sensitized to both ICL drugs and to IR, suggesting that SNM1B/Apollo may have a direct role in the repair of DSBs. Subsequently, this group also demonstrated that SNM1B/Apollo localized to IRIF, and that the autophosphorylation of ATM, which is required for its activation as a kinase, and phosphorylation of downstream ATM targets, was attenuated in SNM1B/Apollo-deficient cells [55]. Studies in DT40 chicken cells have also demonstrated that SNM1B/Apollo-deficient cells are hypersensitive to crosslinking drugs, such as MMC and cisplatin, but hypersensitivity to IR was not observed [38,39]. Both SNM1A–deficient and SNM1B/Apollo-deficient cells have exhibited sensitivity to ICL-inducing agents, however, analysis of a double mutant demonstrated that it was more sensitive than either single mutant, suggesting that there may be a partially redundant DNA repair function for these two genes [38,39]. Alternatively, these findings are consistent with SNM1A and SNM1B/Apollo having functions in distinct ICL response pathways.

Legerski and colleagues used shRNA technology to deplete expression of SNM1B/Apollo in stable clones of HEK293 cells [56]. Consistent with previous reports, these SNM1B/Apollo-deficient cells were hypersensitive to the cross-linking agents MMC, cisplatin and psoralen plus UV-A, however an increased sensitivity to IR or UV was not observed. It was also demonstrated that SNM1B/Apollo-deficient cells exhibit a defective S phase checkpoint in response to MMC, but not to IR, and this finding may account for the specific sensitivity to crosslinking compounds. Interestingly, although previous studies have largely implicated ATR as the major kinase activated in response to ICLs, we demonstrated that it is activation of the ATM-mediated checkpoint that is defective in SNM1B/Apollo-deficient cells. The requirement for SNM1B/Apollo in ATM checkpoint activation following ICL damage was correlated with its role in promoting DSB formation and thus DNA replication fork collapse. Others have demonstrated previously that when the replication fork encounters an ICL, the ATR-mediated S phase checkpoint is activated, leading to a transient arrest [57]. Over a period of hours this checkpoint is attenuated resulting in the collapse of the replication fork and the formation of a double-strand break by incision of a template strand. Current evidence indicates that the incision is performed by the MUS81-EME1 endonuclease, since cells deficient in this enzyme are defective in efficient formation of ICL-induced DSBs [29,58,59]. Most current models of ICL repair predict that formation of DSBs is an obligatory step that is required for uncoupling of the ICL prior to the ultimate repair by homologous recombination. The HEK293 SNM1B/Apollo-deficient cells have a similar phenotype, in that formation of DSBs is greatly reduced upon exposure to ICL-inducing agents. A possible function for SNM1B/Apollo in the formation of DSBs is to resect the lagging strand of a regressed fork in order to produce a substrate that is highly amenable to cleavage by MUS81-EME1 [60]. Consistent with this result was our finding that SNM1B/Apollo directly interacts with MUS81-EME1 via its MBL domain, suggesting that these proteins may act cooperatively to process stalled forks into DSBs [56]. These findings provide an explanation as to why no checkpoint defect was observed after IR, since SNM1B/Apollo produces DSBs in response to ICLs, which are the activating signal for the ATM-mediated checkpoint. In addition, SNM1B/Apollo interacts with the MRE11-RAD50-NBS1 (MRN) complex via its β-CASP domain and with FANCD2, suggesting that it may have additional roles in the repair processing of ICLs and/or in checkpoint signaling in response to these lesions. It should be noted that the interaction with the MRN complex was direct, while the interaction with FANCD2 was indirect and likely mediated by MRN. Also, the activation of ATR as assessed by CHK1 phosphorylation was not affected by SNM1B/Apollo depletion upon exposure to MMC, however, a recent study demonstrated that in response to UV, ATR activation is impaired in SNM1B/Apollo-deficient cells [61]. The reason for this discrepancy is unclear, however SNM1B/Apollo-deficient cells do not appear to be sensitive to UV, which would be expected if ATR-mediated checkpoint signaling was disrupted. This study also reported that SNM1B/Apollo interacts with members of the heat shock protein 70 family, and that HSP72 is required for the localization of SNM1B/Apollo to nuclear foci. HSP70 chaperones have been implicated in genome stability, and this interesting connection to SNM1B/Apollo might at least partially explain their involvement, since, as discussed below, SNM1B/Apollo is required for protection of telomeres.

A number of laboratories have reported that SNM1B/Apollo interacts with the telomere binding protein TRF2 and has a role in telomere maintenance [13,55,62–64]. Proper telomere length maintenance is critically important during development of multicellular organisms, since telomere length is the determining factor for the number of cell divisions a cell can undergo in the absence of telomerase. Telomeres can fold into a structure termed the t-loop, in which the 3′ single-stranded overhang invades a duplex region of the telomere to sequester the overhang [65,66]. Telomeres are protected from inappropriate activation of the DDR by a complex of proteins referred to as the shelterin complex [67]. Depletion or mutation of members of the shelterin complex renders the telomeres unprotected and can lead to chromosomal fusion and hence genomic instability. In addition to the shelterin complex, other telomere-associated proteins have also been demonstrated to be required for telomere protection. One of these accessory proteins is SNM1B/Apollo, and its depletion by siRNA experiments was demonstrated to lead to activation of the DDR and increased levels of chromosomal end-to-end fusions. SNM1B/Apollo forms nuclear foci in untreated cells, and these foci were demonstrated to coincide with telomeres and to require TRF2 for their formation. The role of SNM1B/Apollo at telomeres has yet to be resolved, however, like other members of the SNM1 family, SNM1B/Apollo possesses a 5′–3′ exonuclease activity [13]. Thus, one postulated role for SNM1B/Apollo is that it is required to resect the leading strand upon completion of DNA replication in order to create the single-stranded overhang necessary for the formation of t-loops. Recently, we generated a knockout of SNM1B/Apollo in mice and observed that null animals exhibited multiorgan developmental failure, which led to a perinatal mortality [Akhter et al., Unpublished Data]. MEFs derived from the null embryos demonstrated an increased incidence of chromosomal fusions, which were the likely cause of the impaired organ development [68]. This conclusion is supported by the finding that a cross of the Snm1B/Apollo null mice with Ku null mice rescued the lethal phenotype, suggesting that the nonhomologous end-joining (NHEJ) pathway is responsible for the observed chromosomal end-to-end joining. Taken together, these data implicate SNM1B/Apollo as a key 5′–3′ exonuclease required for generation of 3′ single-stranded overhangs at newly replicated leading-strand telomeres to protect them from engaging the NHEJ pathway. Of the three SNM1 mammalian homologs discussed here, only the knockout of SNM1B/Apollo leads to an embryonic lethal phenotype, which appears to be owing to the unique role of this protein in telomere maintenance.

As discussed earlier, we have demonstrated that SNM1A is a multifunctional gene involved in both the DDR and in an early mitotic or prophase checkpoint in response to spindle stress. To find interaction partners for SNM1A and SNM1B/Apollo, we conducted yeast two-hybrid screens, and among the positives found in both screens was a gene with a known mitotic function referred to as Astrin (also termed DEEPEST) [69–71]. Interestingly, an interaction between Astrin and Artemis was not observed, suggesting that it was specific to SNM1A and SNM1B/Apollo. The function of Astrin is not well understood, but it is an essential protein that localizes both to the mitotic spindle and to the centrosome and has been found to be involved in the regulation of the mitotic enzyme separase [71].

Previous findings with SNM1A and the two-hybrid results with Astrin led our laboratory to investigate whether SNM1B/Apollo might also play a role in early mitosis in response to spindle poisons. Using the SNM1B/Apollo-depleted HEK293 clones described above, we demonstrated that these cells were defective in the prophase checkpoint as indicated by a failure to arrest prior to chromosome condensation upon exposure to nocodazole [72]. Abrogation of the prophase checkpoint resulted in an extended spindle checkpoint causing a prolonged duration of mitosis. Similar to CHFR the target of the SNM1B/Apollo pathway appeared to be CDK1-cyclin B since the inhibitory phosphorylation of CDK1 promoted by the WEE1 kinase and regulated by the CDC25 phosphatase was not maintained in the presence of nocodazole in SNM1B/Apollo-deficient cells in comparison with control cells. This result was also confirmed by an excessive and prolonged CDK1-cyclin B-mediated phosphorylation of the APC subunit CDC27. In addition, SNM1B/Apollo was demonstrated to interact with Astrin in vivo by co-IP experiments through its MBL domain, and disruption of this interaction by independent point mutations (D14N and H276A) resulted in a deficient prophase checkpoint. Interestingly, live cell imaging studies indicate that the levels of SNM1B/Apollo are rapidly and dramatically reduced at the onset of mitosis, and that the residual protein is confined to the centrosome where it colocalizes with Astrin. Consistent with this finding is the fact that the initial activation of CDK1-cyclin B is known to occur in the centrosome [73]. Furthermore, treatment with nocodazole appeared to prevent the reduction in SNM1B/Apollo levels during mitosis, and the point mutants mentioned above were found not to undergo extensive reduction in levels in untreated cells. The mitotic reduction of SNM1B/Apollo appears to be owing to degradation by the ubiquitin/proteosome system, since the inhibitor MG132 stabilizes the protein. However, the E3 ubiquitin ligase responsible has yet to be identified. Taken together, these findings indicate that SNM1B/Apollo and Astrin function together to enforce the prophase checkpoint in response to spindle stress and that the target of this checkpoint is CDK1-cyclin B. Whether SNM1B/Apollo and SNM1A function in the same prophase checkpoint pathway is unclear at present, but it is certain that these proteins do not act redundantly.

Artemis

Artemis is by far the most intensively studied member of the mammalian SNM1 gene family, and was initially genetically characterized as defective in a form of human severe combined immunodeficiency (SCID) [12]. This syndrome is characterized by a defect in variable (diversity) joining (V[D]J) recombination resulting in a virtually complete premature arrest of both B- and T-cell maturation. In addition, patient cell lines exhibited an increase in sensitivity to IR compared with normal cells [74–76]. This form of SCID is usually lethal within the first year of life owing to protracted infections caused by opportunistic microorganisms. This SCID syndrome resembles murine SCID caused by defects in DNA-dependent protein kinase (DNA-PK), a protein complex involved in both V(D)J recombination and repair of DNA double-strand breaks via the NHEJ pathway. Disruption of Artemis in a mouse model and the characterization of these mice indicated that they exhibit a SCID with sensitivity to ionizing radiation phenotype congruent to the human syndrome [77]. In addition, MEFs derived from these mice demonstrated an elevated frequency of chromosomal abnormalities, indicating a clear role for Artemis in maintenance of genomic stability. Interestingly however, Artemis null mice did not exhibit accelerated tumorigenesis compared with wild-type littermates. Yet when Artemis/p53 double null mice were examined, an accelerated tumorigenesis was observed compared with p53 null mice, indicating that Artemis is a tumor suppressor in the absence of p53 [78]. Interestingly, the observed tumorigenesis mostly affected the B-cell lineage, with the development of aggressive pre-B-cell lymphomas. Hypersensitivity to MMC was not noted in the Artemis null MEFs, in agreement with studies of human mutants [79].

Lieber and his colleagues defined the role of Artemis in V(D)J recombination by demonstrating that, when complexed with DNA-PKs it acquires an endonucleolytic activity on 5′ and 3′ overhangs and the ability to open hairpins in duplex DNA [15,16,80,81]. The latter activity is consistent with the observed defect in coding joint formation in Artemis-deficient cells, since coding joints terminate in hairpins upon processing by the RAG1/RAG2 complex. It was also demonstrated that Artemis is a substrate of the kinase activity of the DNA PK catalytic subunit (DNA-PKcs) in vitro, which was the first evidence suggesting that Artemis was a phosphoprotein. Artemis was also demonstrated, independent of DNA-PKcs, to possess a 5′–3′ exonucleolytic activity on single-stranded DNA. However, when an extensive point mutation analysis of conserved residues within the MBL and β-CASP domains of Artemis was performed, almost all the mutants were defective in the DNA-PKcs-dependent endonucleolytic activity, but none were defective in the Artemis-associated exonuclease activity [16]. This raised the specter that the observed exonucleolytic activity was owing to a contaminant; otherwise, the putative exonuclease activity would presumably have to lie outside of the conserved MBL and β-CASP domains. More recently, Artemis endonuclease activity against single-stranded DNA, which is stimulated by DNA-PKcs, has also been reported [82]. There are a number of additional reports of the characterization of the DNA-PKcs-dependent endonuclease activity of Artemis [19–21,83–85]. These will not be discussed here, as an excellent recent review has summarized the nuclease activities of the MBL protein family in depth [5]. However, it should be noted that one study has demonstrated that an Artemis mutant that is unable to interact with DNA-PKcs either in vitro or in vivo is able to rescue the V(D)J defect in Artemis-deficient cells [86]. This finding appears to conflict with the in vitro results, however, it is possible that overexpression of the mutant protein results in the observed complementation owing to weak interactions of Artemis with components of the V(D)J machinery. This consideration aside, these findings have defined the mechanism by which Artemis acts in V(D)J recombination and, combined with the hypersensitivity of Artemis-deficient cells to IR, led to the proposal that the protein also plays a role in the NHEJ pathway of DSB repair.

A role for Artemis in the NHEJ pathway was further strengthened by work from the Jeggo and Lobrich laboratories, which demonstrated that Artemis was required for the repair of approximately 10–15% of all DSBs caused by IR [87]. This subfraction of DSBs was proposed to be structurally complex in some way and require processing by the nuclease activity of Artemis prior to rejoining by NHEJ. In addition, this report demonstrated that Artemis was a substrate of the checkpoint/DNA repair kinase ATM and that ATM’s direct role in NHEJ was mediated by Artemis. These conclusions are also supported by work from other laboratories [79,83,88]. Thus, the concept that Artemis, largely regulated by ATM and to some extent by DNA-PK, is involved in the nucleolytic processing of a subset of IR-induced DSBs ,became the dominant paradigm of Artemis function. However, subsequent papers from the Jeggo laboratory and results from our laboratory (discussed below) have raised doubts about this simple model [89,90]. In these studies it was proposed that ATM, and presumably Artemis, is required for the repair of DSBs that occur only in highly compacted heterochromatin and thus, ATM’s function is in chromatin relaxation rather than directly in DNA repair processing. In support of this model it was demonstrated that depletion of the hetero chromatin binding factor KAP1 dispensed with the requirement for ATM, although whether this finding also applies to Artemis was not addressed. KAP1 has been demonstrated to be phosphorylated by ATM in response to IR, which promotes relaxation of compacted chromatin and access by the DNA repair machinery [91]. Thus, these findings refocused ATM’s role in DSB repair from direct DNA processing to remodeling of highly compacted chromatin to allow access by DNA repair enzymes. With regard to Artemis, this model raises a number of issues. If, in fact, ATM is primarily required for chromatin relaxation in heterochromatin, what is the role of the Artemis nuclease function in this process, or on the other hand, why would the nuclease activity of Artemis be specifically required for DNA repair in heterochromatin. Possibly, as discussed below, ATM, and also ATR, regulate Artemis with respect to functions such as chromatin remodeling, cell cycle control and so on, while DNA-PKcs regulates Artemis specifically in the processing of complex broken ends throughout the genome.

Artemis has been demonstrated to be extensively phosphorylated both in vitro and in vivo by the three phosphatidylinositol-3-OH kinase-like (PIK) kinases ATM, DNA-PKcs and ATR depending on the type of DNA damage or cellular stress [86–88,92–98]. Initially, it was assumed that DNA-PK was the major kinase that regulated Artemis in response to IR, since this enzyme robustly phosphorylates Artemis in vitro and strongly co-IPs with it in vivo [39,95]. In fact, using mass spectroscopy analysis the Lieber group identified 24 sites of phosphorylation in Artemis, 21 of which were phosphorylated by DNA-PKcs in vitro [96]. Sites that have been confirmed to be phosphorylated in vivo are shown in Figure 2. However, a number of reports have indicated that, in fact, ATM is the more important regulator of Artemis in vivo after IR exposure [87,88,92–95]. For instance, using phospho-specific antibodies, it was demonstrated that serines S516, S534, S538 and S645, all located in the (serine/threonine)–glutamine cluster domain of Artemis, were phosphorylated in response to IR in vivo [93]. At low-to-moderate doses of IR (2–3 Gy) ATM was almost solely responsible for these phosphorylations, however, at a substantially higher dose (10 Gy) DNA-PKcs also contributed to these modifications. Thus, the evidence from these studies indicates that it is ATM that is the more relevant PIK kinase with regard to Artemis phosphorylation following exposure of cells to IR. Interestingly, Artemis was also demonstrated to be phosphorylated at the same four sites by the ATR kinase in vivo in response to other forms of DNA damage and cellular stress such as UV irradiation, MMC and hydroxyurea [88,93–95]. In fact, it was these findings that led us to the concept that Artemis has a wider role in the DDR than acting as a nuclease in the processing of DSBs during NHEJ or homologous recombination.

Figure 2. Sites of Artemis phosphorylation that have been confirmed in vivo.

S516, S534, S538 and S645 are all serine–glutamine sites and have been confirmed to be phosphorylated in vivo through the use of phospho-specific antibodies [86,93]. The S385 and S518 sites have been confirmed by mass spectroscopy and phospho-specific antibodies [39]. The S385 site represents a consensus sequence for phosphorylation by casein kinase II. The S518 site is a serine–proline motif.

β-CASP: β-CPSF–Artemis–SNM1–PSO2; MBL: Metallo-β-lactamase;

SCD: (Serine/threonine)–glutamine cluster domain.

The role of phosphorylation in Artemis function remains undefined to a large extent, which is not surprising given the large number of potential sites identified in this protein. Current findings indicate that phosphorylations by PIK kinases, at least at serine–glutamine or serine–proline sites, are not involved in regulating the role of Artemis in V(D)J recombination or NHEJ. A construct of Artemis containing only the catalytic core (MBL plus β-CASP) is able to fully rescue the defects in these pathways in vivo [85,94]. This result is compatible with a model in which phosphorylation of the C-terminal region of Artemis is required to relieve negative regulation of the endonuclease activity. However, several reports have demonstrated that multisite mutation of Artemis phosphorylation sites does not affect Artemis endonuclease activity in vitro, or the ability of Artemis to rescue defects in V(D)J or NHEJ in vivo [85,86,94,98]. Interestingly, it was demonstrated that autophosphorylation of DNA-PKcs was required for the activation of the Artemis nuclease, which suggests that a conformational change in the kinase is required to engage Artemis nucleolytic activity [20,98].

As discussed above, Artemis is phosphorylated by the checkpoint kinases ATM or ATR in response to DSBs or lesions that form replication barriers, respectively; the sites of these phosphorylations appear to be the same for each kinase. Artemis-deficient cells are not hyper sensitive to UV or MMC, suggesting that Artemis does not have a direct role in the repair of these lesions [77]. In fact, the repair of UV-induced photo-lesions by nucleotide excision repair has been reconstituted in vitro [99], and it is highly unlikely that there is a role for the Artemis nuclease in this pathway. These considerations led our laboratory to examine a role for Artemis in cell cycle regulation in response to DNA damage. An analysis of Artemis phosphorylation demonstrated that the S516, S534, S538 and S645 sites were all rapidly phosphorylated within minutes by either IR or UV exposure [93,95]. However, the S534 and S538 sites were rapidly dephosphorylated within 1 h, while the other two sites remained phosphorylated for at least 24 h. These results suggest that S516 and S645 may have a common function in Artemis regulation and that S534 and S538 have a separate function. This prediction was borne out by studies demonstrating that an allele of Artemis in which both the S516 and S645 residues were converted to alanine was defective in recovery from the G2/M cell cycle checkpoint [93,95]. An allele in which the serine residues were replaced with aspartic acid residues to mimic phosphorylation displayed a phenotype equivalent to that of the wild-type allele. This result indicates that phosphorylation at these sites was causative of the phenotype. The initial imposition of the G2/M checkpoint was normal, however the activation of CDK1-cyclin B in the centrosome was delayed when this Artemis alanine mutant was overexpressed. It is conceivable that this phenotype is owing to incomplete repair of DSBs, which could lead to an extended checkpoint, as has been reported for Artemis null mutants [100,101]. However, as mentioned above, mutation of these particular residues in Artemis is not detrimental to NHEJ, and we did not observe altered levels of phosphorylated H2AX (γH2AX) when this allele was overexpressed compared with wild-type Artemis. These findings suggested that Artemis has a role in regulating checkpoint recovery from the G2/M checkpoint by impacting the activation of CDK1-cyclin B, albeit through an unknown mechanism.

In a separate study the role of Artemis phosphorylation by ATR after UV or MMC exposure was investigated [102]. The major checkpoint in response to these agents occurs in S phase as a result of the DNA replication fork encountering lesions that block progression. Similar to the studies mentioned above, mutation of the S516 and S645 residues resulted in a delayed recovery from the S phase checkpoint, which was owing to impaired degradation of cyclin E. Cyclin E is regulated by the proteosome system via ubiquitination by the SCF^Fbw7 E3 ligase complex [103–105]. Phosphorylation of Artemis at the S516 and S645 sites strongly enhances its binding to the F-box protein Fbw7 and thereby promotes ubiquitination of cyclin E. These findings provide an explanation for the ATR-mediated phosphorylation of Artemis and define a dramatically novel role for this protein as a component of the ubiquitin–proteosome system. In support of this finding, our unpublished studies have demonstrated that Artemis is also involved in protein ubiquitination as part of the Cul4A–DDB1 E3 ligase complex [Yan et al., Unpublished Data].

As a further testament to the wide-ranging functions of Artemis, we have also demonstrated that it is a negative regulator of p53 in response to oxidative stress derived from mitochondrial respiration [106]. Oxidative stress was demonstrated to activate DNA-PKcs to phosphorylate p53 and induce its transcriptional program, such as expression of p21, BAX, SESTRIN2 and TIGAR. These latter two p53 target genes are involved in mitigating the levels of reactive oxygen species in cells. Artemis, which interacts with both p53 and DNA-PKcs, inhibits the phosphorylation of p53 by DNA-PKcs. Markers of DNA damage such as H2AX, CHK1 or CHK2 were not activated, indicating that this pathway was not a component of the DDR, but rather a response to reactive oxygen species created during oxidative phosphorylation. These findings also suggest that Artemis is a regulator of DNA-PKcs kinase activity, although the mechanism by which Artemis inhibits the phosphorylation of p53 is unclear.

In consideration of the foregoing discussion, we propose a working model for the role of phosphorylation in Artemis function (Figure 3). In response to DNA damage or replication fork barriers, Artemis is phosphorylated at multiple sites, principally by the ATM or ATR kinases. These phosphorylations do not affect the function of Artemis with respect to its nuclease activities in NHEJ or V(D)J recombination. In fact, the entire C-terminal region of Artemis is dispensable for these functions [85]. Rather, phosphorylation regulates separate functions of Artemis in cell cycle checkpoints, chromatin remodeling and possibly other responses to cellular stress. This model suggests that Artemis is a multifunctional protein that acts as a simple nuclease in certain aspects of DNA metabolism on one hand, and as a phosphoprotein that is a regulator of cellular stress responses on the other.

Figure 3. Model for the functions of Artemis in mammalian cells.

In response to double-strand breaks created by IR or the RAG1/RAG2 complex, Artemis interacts with autophosphorylated DNA-PKcs to facilitate repair. These processes are independent of Artemis phosphorylation and require only the SNM1 domain (left side of schematic). In response to various forms of cellular stress, Artemis is phosphorylated at multiple sites and acts in various response pathways (right side of schematic).

DNA-PKcs: DNA-dependent protein kinase catalytic subunit; IR: Ionizing radiation; NHEJ: Nonhomologous end joining; P-Artemis: Phosphorylated Artemis; SNM1: Sensitivity to nitrogen mustard; V(D)J: Variable (diversity) joining.

Functions of the conserved SNM1 domain

As discussed earlier, there is extensive evidence that the SNM1 domain, composed of the MBL and β-CASP folds, contains a nucleolytic activity. SNM1A and SNM1B/Apollo have both been reported to possess a 5′–3′ exonuclease that acts on single-stranded DNA and Artemis, in association with DNA-PKcs, possesses an endonuclease activity [5,13–21]. Mutational analysis confirms that this nuclease activity resides within the SNM1 domain. However, surprisingly we and others have noted that the SNM1 domain also constitutes an interface for protein–protein interactions. SNM1A interacts with the SUMO ligase PIAS1 and the mitotic factor Astrin [38,72]; SNM1B/Apollo interacts with the MRN complex, the nuclease heterodimer MUS81-EME1, and Astrin [56,72]; and Artemis interacts with the Cul4A–DDB1 ubiquitin ligase complex [Yan et al., Unpublished Data]. All of these inter actions have been mapped to the SNM1 domain of the respective proteins and indicate that this domain has biochemical functions in addition to its nuclease activities (Figure 1).

Cancer relevance

The SNM1 gene family has a clear role in the maintenance of genomic integrity and in the suppression of tumorigenesis. All three genes have been demonstrated to have various roles in the direct repair of DNA damage and have been implicated in cell cycle checkpoint responses to different forms of cellular stress. Both Artemis and SNM1A are substrates of checkpoint kinases and both have been demonstrated to act as tumor suppressors in mouse models. The CHFR gene is mutated in a high proportion of tumors and cancer cell lines, indicating the importance of the prophase checkpoint in suppressing oncogenesis [48,107–113]. Both SNM1A and SNM1B/Apollo have roles in this pathway and polymorphisms in SNM1A have been linked to small-cell lung cancer [36]. SNM1B/Apollo plays a role in telomere protection, and a knockout of this gene in the mouse causes highly increased levels of chromosome end-to-end fusions, resulting in chromosome instability. While the null genotype results in embryonic lethality in mice, more subtle mutations could lead to cancer in humans. Artemis has been demonstrated to be a negative regulator of p53 and, along with its role in protein ubiquitination, suggest that it might, paradoxically, be a target for cancer chemotherapy. In fact, Artemis may act as both a tumor suppressor and an oncogene, since, in its absence, genomic instability will occur, while its overexpression may inhibit the activity of p53 and deregulate other cell cycle regulators.

Future perspective

The mammalian SNM1 family of proteins have proven quite surprising in the range of their identified activities and functions. Initial observations indicated that they were seemingly run-of-the-mill nucleases involved in various aspects of DNA metabolism. However, as investigations have continued, their range of functions have come to include checkpoint recovery in response to DNA damage, imposition of mitotic checkpoints in response to spindle poisons, telomere maintenance, possibly chromatin remodeling and as cofactors in the ubiquitination of other proteins. It is the functions of SNM1 proteins in these latter pathways that require investigation in the coming years. The fundamental role of SNM1A in cell cycle regulation remains unresolved, and is particularly puzzling since it is involved in pathways that respond to both DNA damage and mitotic spindle disruption. Possibly, as has been demonstrated for Artemis, SNM1A may be involved in protein ubiquitination since it interacts with the APC/cyclosome. Recently, the APC/cyclosome has also been demonstrated to be a component of the DDR in that it regulates the implementation of the G2/M cell cycle checkpoint [114]. Thus, SNM1A could participate in these various checkpoint pathways via its association with the APC/cyclosome. SNM1B/Apollo has a clearly defined role in telomere maintenance mediated through its nuclease activity, and future studies should further refine its function at telomeres. In addition, similar to SNM1A, SNM1B/Apollo has a role in mediating the prophase checkpoint in response to spindle poisons. The target of this pathway appears to be CDK1-cyclin B, however, the biochemical role of SNM1B/Apollo requires explication and should be the subject of future studies. Artemis has multiple cellular functions, but perhaps the most interesting new development is its role in protein ubiquitination. It interacts with both the Cul1 and Cul4A E3 ligase complexes, and a future direction will be to identify proteins that require Artemis for their regulation via ubiquitination and how this activity of Artemis manifests itself in cell cycle control. Finally, an analysis of the expression and mutational spectrum of SNM1 genes in human cancers is warranted given the demonstrated role of these genes in maintenance of genomic stability and cell cycle regulation.

The SNM1 proteins have proven to be more interesting than originally anticipated, and additional surprises should be in store as investigations continue on this intriguing protein family.

Executive summary

• The first member of the sensitivity to nitrogen mustard (SNM1) gene family was discovered in budding yeast and has a unique role in the repair of DNA interstrand crosslinks.

• The SNM1 gene family is characterized by metallo-β-lactamase and CPSF–Artemis–SNM1–PSO2 domains that together possess a DNA nuclease activity in the proteins.

• There are five mammalian homologs of SNM1 and three of these; SNM1A, SNM1B/Apollo and Artemis, are nuclear proteins involved in DNA metabolism and cell cycle regulation.

SNM1A

• Knockout studies of SNM1A have demonstrated that it has only a minor role in mitigating sensitivity to interstrand crosslinking agents, and that it has tumor suppressor activity in a mouse model.

• SNM1A localizes to sites of double-strand breaks and has a role in mediating the G1 cell cycle checkpoint. ATM is required for this localization and SNM1A is a substrate of ATM in vitro.

• SNM1A interacts with the SUMO ligase PIAS1 but the functional significance of this interaction is unclear.

• Translation of SNM1A is regulated by an internal ribosome entry site.

• SNM1A has a role in mediating the prophase/anaphase cell cycle checkpoint in response to spindle poisons. It interacts with the anaphase-promoting complex/cyclosome but the significance of this interaction is unclear.

• The biochemical function of SNM1 in mediating cell cycle checkpoints is unknown.

SNM1B/Apollo

• SNM1B/Apollo-deficient cells generally exhibit sensitivity to interstrand crosslinking agents.

• SNM1B/Apollo has been demonstrated to participate in mediating replication fork collapse in response to interstrand crosslinks.

• SNM1B/Apollo interacts with proteins involved in DNA repair and checkpoint signaling, such as MUS81-EME1, the MRE11-RAD50-NBS1 complex and FANCD2.

• SNM1B/Apollo is required for telomere maintenance during S phase and interacts with the telomere protein TRF2.

• A knockout of SNM1B/Apollo in mice results in embryonic lethality owing to chromosome end-to-end fusions.

• SNM1B/Apollo is required for the resection of newly replicated leading strands at telomeres.

• Similar to SNM1A, SNM1B/Apollo has a role in mediating the prophase/anaphase checkpoint in response to spindle poisons. The identified target of SNM1B/Apollo is cyclin B-CDK1, although its biochemical role in this pathway is unknown.

Artemis

• Defects in Artemis result in a severe-combined immunodeficiency syndrome owing to impaired variable (diversity) joining (V[D]J) recombination.

• The nuclease activity of Artemis is required for V(D)J recombination, the repair of a subset of DNA double-strand breaks mediated by nonhomologous end joining and for resection of ends during homologous recombination.

• Artemis is a phosphoprotein and in response to DNA damage Artemis is primarily phosphorylated by ATM and ATR at multiple sites.

• Phosphorylation of Artemis is not required for its roles in V(D)J recombination and nonhomologous end joining.

• Artemis is involved in the recovery from the G2/M cell cycle checkpoint by regulating the activation of cyclin B-CDK1 in the centrosome.

• Artemis is involved in regulating recovery from the S phase checkpoint through the ubiquitination and degradation of cyclin E via the SCF^Fbw7 E3 ligase complex.

• Artemis interacts with both the Cul1 and Cul4A E3 ubiquitin ligase complexes.

• Artemis is a negative regulator of p53 in response to oxidative stress induced by mitochondrial respiration.

• Artemis is a multifunctional protein whose nuclease activity is regulated by DNA-dependent protein kinase catalytic subunits during V(D)J recombination and nonhomologous end joining. This regulation does not involve phosphorylation of Artemis.

Functions of the conserved SNM1 domain

• The SNM1 domain also mediates protein–protein interactions indicating it has biochemical functions in addition to its nuclease activity.

Cancer relevance

• SNM1 genes have been demonstrated to act as tumor suppressors and participate in the maintenance of genomic stability through multiple pathways.

Future perspective

• Mammalian SNM1 proteins are involved in a wide range of functions, including checkpoint recovery in response to DNA damage, imposition of mitotic checkpoints in response to spindle poisons, telomere maintenance, chromatin remodeling and as cofactors in the ubiquitination of other proteins.

• Resolving the detailed biochemical functions of SNM1 proteins in these various pathways is an area for future investigation.