Fission Yeast Rnf4 Homologs Are Required for DNA Repair*

We describe two RING finger proteins in the fission yeast Schizosaccharomyces pombe, Rfp1 and Rfp2. We show that these proteins function redundantly in DNA repair. Rfp1 was isolated as a Chk1-interacting protein in a two-hybrid screen and has high amino acid sequence similarity to Rfp2. Deletion of either gene does not cause a phenotype, but a double deletion (rfp1Δrfp2Δ) showed poor viability and defects in cell cycle progression. These cells are also sensitive to DNA-damaging agents, although they maintained normal checkpoint signaling to Chk1. Rfp1 and Rfp2 are most closely related to human Rnf4, and we showed that Rnf4 can substitute functionally for Rfp1 and/or Rfp2. The double mutants also showed significantly increased levels of protein SUMOylation, and we identified an S. pombe Ulp2/Smt4 homolog that, when overexpressed, reduced SUMO levels and suppressed the DNA damage sensitivity of rfp1Δ rfp2Δ cells.

The fidelity of cell cycle progression is paramount for tissue homeostasis and development and avoidance of disease. A major contributor to this fidelity are the plethora of responses that detect damage to DNA, ensure its repair, and activate checkpoints that modify the cell cycle to allow time for these events to occur. These responses are rapid, highly conserved, and frequently regulated by cascades of post-translational modifications that control the activities of repair proteins and signal transducers of the checkpoints.
The fission yeast Schizosaccharomyces pombe is an important model system for the dissection of the molecular controls over DNA damage responses (1,2). S. pombe has a prominent G 2 period in vegetative cell cycles that represents ϳ70% of total cell cycle time (3). The DNA damage checkpoint functioning during this period signals to activate the Chk1 protein kinase, which in turn prevents entry into mitosis through modulating the Cdc2 regulators Wee1 and Cdc25 (4 -6). Chk1 activation is controlled by reversible phosphorylation on Ser-345 by Rad3 (7)(8)(9), an ATR (ataxia telangiectasia-and Rad3-related) homolog, although the mechanism by which this activation occurs is largely obscure. ATR homologs also phosphorylate a number of other proteins in this pathway (1,10).
In addition to phosphorylation cascades, other post-translational modifications regulate DNA damage response pathways.
Among these are the covalent modification of proteins by ubiquitin and its related protein, SUMO (small ubiquitin-like modifier). In addition to directing the proteolysis of its targets, ubiquitination can directly influence responses to DNA damage. A well documented example of this is through the post-replication repair proteins that control the mono-and polyubiquitination of proliferating cell nuclear antigen, which in turn enables the bypass of lesions during DNA replication (11)(12)(13).
SUMO modification of several DNA repair enzymes has also been described (14), although in most cases the molecular effects of such modification remains unclear. Control over protein SUMOylation is achieved by a balance of the activity of several E3-SUMO ligases and the deconjugating activity of SUMO peptidases such as Ulp2/Smt4 in Saccharomyces cerevisiae. SUMO is not essential in S. pombe, although cells deleted for the SUMO gene (pmt3) (15) or the E2 SUMO-conjugating enzyme gene hus5 (16) are sensitive to DNA-damaging agents and show defects in chromosome segregation. S. cerevisiae cells lacking Ulp2 also show defects in chromosome segregation and DNA damage responses (17)(18)(19)(20), suggesting that a balanced control over protein SUMOylation and de-SUMOylation is important in these responses.
Here we describe two novel genes in S. pombe, rfp1 and rfp2, which encode proteins with RING finger domains, which are characteristic of E3 ubiquitin ligases. These proteins function redundantly in DNA repair; a double mutant is hypersensitive to different forms of DNA damage, although checkpoint signaling through to Chk1 activation is normal. We show that human Rnf4, an E3 ubiquitin ligase that has been implicated as a cofactor for several transcription factors (21)(22)(23)(24)(25), is a close sequence relative of Rfp1 and Rfp2 and can indeed substitute functionally for these proteins. We show that the control over protein SUMOylation is defective in cells lacking Rfp1 and Rfp2 and that indeed this is the cause of the DNA repair defect. These data raise the possibility that human Rnf4 may function similarly in the DNA damage response.

EXPERIMENTAL PROCEDURES
Two-hybrid Screening-Chk1 was cloned onto the Gal4 DNA-binding domain (pAS2) and transformed into the PJ69A strain of S. cerevisiae (26). An S. pombe cDNA Gal4 activation domain fusion library in pACT (a kind gift of Steve Elledge) was transformed into these cells, and putative interactors were screened for on SC medium lacking adenine. Plasmids were recovered into Escherichia coli, sequenced, and retested for bait dependence.
Fission Yeast Methods-All strains are derivates of 972h Ϫ and 975h ϩ . Standard methods and media for the propagation, transformation, mating, and culturing of S. pombe were as described (27). Null alleles were constructed by Start-Stop codon replacement by ura4 by homologous recombination. Gene targeting was confirmed by Southern blotting, and strains were back-crossed to wild type before analysis. Microscopy was performed on cells fixed in 3.7% formaldehyde or 70% ethanol, and images were captured on a Nikon Eclipse 800 microscope with a Spot RT/SE camera. Immunodetection of proteins in fixed cells was performed as previously described (28,29). The methods used for UV-C and methyl methanesulfonate (MMS) 2 survival assays were as described previously (30,31). Briefly, for UV-C, exponential cultures were plated in triplicate at densities of 100 -10,000 cells/plate and irradiated with UV-C using a Stratalinker (Stratagene). Colonies were allowed to form over a 4-day span at 30°C and were normalized to unirradiated controls. For MMS sensitivity, 10-fold serial dilutions were spotted onto plates containing a range of MMS concentrations, and colonies were allowed to form over a 4-day span at 30°C. Plasmid curing was achieved by selection of relief for auxotrophic markers on YES medium (yeast extract plus supplements) followed by reselection of minimal medium. To avoid selection for extragenic, slow growth suppressors, all experiments with rfp1⌬ rfp2⌬ double mutants were performed with freshly made strains derived from nonparental ditypes.
Protein Extraction and Western Blotting-Proteins were extracted from cell pellets disrupted with glass beads and a mini-bead beater. Native extracts for immunoprecipitations were performed as described (9). Denatured extracts for SUMO conjugates were made using 8 M urea buffer (5). Pmt3 (SUMO) was detected using an HA-tagged pmt3 allele (15). Anti-␣-tubulin antibody B-5 (Sigma) was used as a loading control. Western transfer of SDS-polyacrylamide gels to nitrocellulose was performed with 10 mM CAPS, 10% methanol. Immune complexes were detected with horseradish peroxidase-conjugated secondary antibodies were detected with ECL reagent (GE Biosciences). Chk1 activity assays were performed in triplicate as described (32,33).
Recombinant Proteins and Chk1 Kinase Assays-Recombinant glutathione S-transferase fusion proteins were expressed in BL-21 E. coli and purified on GSH-Sepharose as described (34). Phosphorylation of these proteins by Chk1 purified from S. pombe extracts was performed as described (8,32).

Identification of Chk1-interacting Proteins-To search for
Chk1-interacting proteins we performed a yeast two-hybrid screen using full-length Chk1 fused to the Gal4 DNA-binding domain as bait. Among several positive clones (Table 1) were rad24 and ded1, which have been described previously as Chk1 interactors by this assay (35,36). Among the unique clones was a novel 254-amino acid protein (SPAC19A8.10) with a C-terminal RING finger domain, which we called Rfp1. Using N-and C-terminal Chk1 baits, Rfp1 was shown to interact with the C-terminal regulatory domain of Chk1 (Fig. 1A). The interaction was confirmed by co-immunoprecipitation of Chk1 and Rfp1 expressed from the medium-strength attenuated nmt1 promoter (37) (Fig. 1B). However, under endogenous expression conditions, no such interaction could be observed (data not shown). Thus, the interaction in the yeast two-hybrid assay may not be physiological, although we could not rule out a transient interaction, which would be difficult to detect.
Rfp1 Acts Redundantly with Rfp2 in DNA Repair-As we could not detect interaction between endogenous Chk1 and Rfp1, we sought to uncover function for Rfp1 to gauge the physiological relevance of the yeast two-hybrid interaction. The chromosomal rfp1 locus was deleted and replaced by ura4. These cells grew normally, were fertile, and showed wild-type sensitivities to DNA-damaging agents. We concluded that rfp1 is, by itself, not essential for cell viability.
A BLAST search revealed that Rfp1 was most similar to mammalian Rnf4 proteins and to a 205-amino acid fission yeast RING finger-containing protein (SPAC343.18), which we denoted as Rfp2 (Fig. 2A). The rfp2 locus was also deleted and replaced by ura4. These cells also grew normally, were fertile, and showed wild-type sensitivities to DNA-damaging agents. We made rfp1⌬rfp2⌬ double mutants by tetrad dissection and found that double mutant colonies were severely growth-inhibited (Fig. 2B), forming small colonies that were prone to selec-  tion for slow growth suppressors. S. pombe grows by apical extension of cell tips, and a delay to cell cycle progression manifests as cells dividing at an increased cell length. Microscopic analysis showed that the rfp1⌬rfp2⌬ double mutants were delayed in cell cycle progression, dividing at 23.2 Ϯ 6.5 m compared with wild type, which divided at 13.8 Ϯ 0.2 m (Fig.  2D). By immunofluorescence, HA-tagged Rfp1 was localized to the nucleus (Fig. 2C). We have not been able to detect Rfp2 by Western blotting or immunofluorescence, even when expressed from the strongest nmt1 promoter, and we could not detect an interaction between Chk1 and Rfp2 by twohybrid assays (data not shown).
The poor growth and cell cycle delay of the rfp1⌬rfp2⌬ double mutants is reminiscent of strains deleted for genes involved in homologous recombination (HR). We therefore assayed DNA damage sensitivities of rfp1⌬rfp2⌬ double mutants and controls (Fig.  3). rfp1⌬rfp2⌬ double mutants were indeed sensitive to UV-C irradiation (Fig. 3A) and chronic exposure to the alkylating agent MMS (Fig. 3B).
Both of these agents induce a Chk1-dependent DNA damage checkpoint and, at the doses tested, require HR for their repair in S. pombe. Inappropriate entry into mitosis with damaged DNA due to a Chk1-dependent checkpoint defect in S. pombe results in the bisection of the nucleus by the medial septum, known as the "cut" phenotype. Conversely, imposition of a Chk1-dependent checkpoint delay in response to DNA damage results in elongated cells, and even though the nuclear cell cycle is delayed by thus checkpoint, cell growth continues. UV-C-irradiated or MMS-treated rfp1⌬rfp2⌬ double mutants elongated further compared with untreated cells (Fig. 3C), showing an intact checkpoint response emanating from signaling downstream of Chk1. This was confirmed biochemically via the phosphorylation and activation of Chk1 (Fig. 3D), showing that signaling upstream of Chk1 to activate Chk1 activity was also intact. These data suggest that the rfp1⌬rfp2⌬ double mutants are defective in DNA repair, which is most likely because of an HR defect. However, the poor viability of the rfp1⌬rfp2⌬ double mutants and of HR mutants such as rhp51⌬ and rad22⌬ (encoding homologs of Rad51 and Rad52), together with the high rate of spontaneous slow growth suppressors, precluded epistasis to confirm a role in HR.
We concluded that Rfp1 and Rfp2 act redundantly in DNA repair and that the poor viability of the rfp1⌬rfp2⌬ double mutant is due to the inability to process spontaneous lesions as seen in S. pombe HR mutants. We did not see a defect in Chk1 activation, nor did we see a checkpoint defect downstream of Chk1 signaling, as the rfp1⌬rfp2⌬ double mutant was proficient in a DNA damage-induced cell cycle arrest. We cannot rule out the possibility that Chk1 acts upstream of Rfp1 and/or Rfp2 to affect DNA repair, but several observations are not consistent with this option. Firstly, the radiation sensitivity of S. pombe cells lacking Chk1 is restored by an imposed cell cycle delay using a conditional mutation in the Cdc25 phosphatase required for Cdc2 activation (38,39); that is, given the time to repair, chk1⌬ cells are fully capable of doing so. This protocol does not rescue the radiation sensitivity of mutants defective in DNA repair (30,40). Moreover, HR mutants, such as the rfp1⌬rfp2⌬ double mutant, show poor cell viability, chromosome instability, and poor spore viability, and none of these phenotypes are seen in chk1⌬ cells. In S. cerevisiae, direct assays of DNA gap repair, which is via HR, have shown wild-type repair efficiencies in chk1⌬ cells (41). This is in keeping with their lack of radiation sensitivity (42), although these cells do show a higher frequency of crossovers for unknown reasons (41). Thus, Chk1 function following DNA damage is primarily to delay the cell cycle to allow time for DNA repair rather than being required directly for DNA repair itself.
We also tested whether Rfp1 was a substrate for Chk1 phosphorylation (Rfp2 protein expression could not be detected). When Chk1 is overexpressed, the phosphorylation of Wee1 results in a mobility shift assayed by Western blotting (5), but this was not the case for Rfp1 (Fig. 3E). A consensus site for Chk1-mediated phosphorylation has been determine as -X-␤-X-X-(S/T)*, where an asterisk indicates the phosphorylated residue, is a hydrophobic residue (M Ͼ I Ͼ L Ͼ V), ␤ is a basic residue (R Ͼ K), and X is any amino acid (43). Rfp2 has no sequences related to this motif, but Rfp1 has one potential site (LTRSPS-22). We assayed the phosphorylation of recombinant Rfp1 in vitro and of a mutant protein where Ser-22 was substituted for alanine. Neither protein was significantly phosphorylated, whereas a fragment of Wee1 containing phosphorylation sites at Ser-104 and Ser-117 (43) was efficiently phosphorylated (Fig. 3F). It is possible that the recombinant Rfp1 is improperly folded, but the lack of in vitro phosphorylation is consistent with the lack of mobility shift in cells overexpressing Chk1. Thus, if Chk1 is regulating Rfp1, this effect is likely to be indirect.
Human Rnf4 Functionally Rescues rfp1⌬rfp2⌬-A BLAST search identified the E3 ubiquitin ligase Rnf4 as the most related human sequence to Rfp1 and Rfp2. To test conservation of function, we cloned human Rnf4 onto the S. pombe nmt1 promoter and expressed Rnf4 in wild type and rfp1⌬rfp2⌬ double mutants. Fig. 4 shows that RNf4 expression rescued the UV-C sensitivity (Fig. 4A), MMS sensitivity (Fig. 4B), and cell cycle delay phenotypes (Fig. 4C) of rfp1⌬rfp2⌬ double mutants. To ensure that this was indeed functional rescue and not the selection of suppressors (which are common with the rfp1⌬rfp2⌬ double mutant), we cured rfp1⌬rfp2⌬ double mutants of the nmt1-Rnf4 plasmid and indeed recovered each phenotype (data not shown).
We conclude that Rnf4 functionally rescued the defects caused by deletion of either rfp1 and/or rfp2. Together with the sequence similarity, these data show that these proteins are therefore homologs, although no DNA repair function has been ascribed to Rnf4.
rfp1⌬rfp2⌬ Cells Accumulate SUMOylated Proteins-Ring finger domains are characteristic of E3 ubiquitin ligases, and related domains are involved in the ligation of ubiquitin-like proteins such as SUMO. We assayed for ubiquitin-and SUMOconjugated proteins in denatured extracts in rfp1⌬, rfp2⌬, and rfp1⌬rfp2⌬ double mutants by Western blotting. We found that the rfp1⌬rfp2⌬ double mutants had vastly increased levels of SUMOylated proteins (Fig. 5, compare lanes 2 (wild type) and 5 (rfp1⌬rfp2⌬)), although ubiqutination was unaffected (data not shown). This was not evident in either of the single mutants or in UV-C-irradiated or G 2 -arrested wild-type cells, which we tested given the cell cycle delay and DNA repair defects in these cells. Rfp1 and Rfp2 are therefore negative regulators of protein SUMOylation.
DNA Repair Defects in rfp1⌬rfp2⌬ Cells Is Due to Deregulated SUMOylation-We next wished to address whether the increase in SUMO conjugates in rfp1⌬rfp2⌬ double mutants was the cause of the DNA repair defects. Although SUMO is not essential in S. pombe, cells deleted for the SUMO gene (pmt3) are extremely sick, and it was not feasible to construct pmt3⌬rfp1⌬rfp2⌬ triple mutants.
As an alternative approach to dampening SUMOylation in rfp1⌬rfp2⌬ double mutants, we overproduced Ulp2 (SPAC17A5.07), which encodes a putative homolog of the S. cerevisiae SUMO-deconjugating peptidase Ulp2/Smt4. Expression of Ulp2 from the nmt1 promoter did not affect wildtype cell viability, although it did decrease the levels of SUMO conjugates in both wild-type and rfp1⌬rfp2⌬ cells (Fig. 6A). For the latter, the reduction in SUMO conjugates also suppressed the sensitivity to UV-C irradiation (Fig. 6B). We conclude that the DNA repair defects that manifested in the absence of Rfp1 and Rfp2 were due to the increased protein SUMOylation.

DISCUSSION
The coordination of DNA repair with cell cycle progression is essential to maintaining genome integrity. Here we have described two new components of the DNA damage response in S. pombe, Rfp1 and Rfp2. Cells tolerate the deletion of either gene without consequence, but the double deletion shows defects in cell cycle progression and is sensitive to DNA-damaging agents. The synthetic nature of these phenotypes suggests that Rfp1 and Rfp2 have cellular functions that are redundant or at least overlapping. Based on the normal activation of the DNA damage checkpoint in these cells, we concluded that the sensitivity is due to a DNA repair defect.
Expression of human Rnf4 fully rescued the phenotypes of the rfp1⌬ rfp2⌬ cells. Given that these phenotypes require the deletion of both genes, we cannot formally determine whether Rnf4 is functioning as Rfp1, Rfp2, or a combination of the two. Rnf4 contains both DNA binding and E3 ubiquitin ligase activities (21,25) and has been shown to interact with several transcription factors that are not conserved in S. pombe, including the androgen (24) and estrogen receptors (44). It is not known whether Rnf4 is required for DNA repair in human cells.
The DNA repair defects of rfp1⌬rfp2⌬ double mutants were associated with a large increase in protein SUMOylation and  were rescued by the reduction of SUMO conjugates by Ulp2 overexpression. SUMO is known to modify a number of proteins involved in DNA repair (14), and it is likely that dysregulation of some if not all of these proteins contributes to the DNA damage sensitivity and repair defects of the rfp1⌬rfp2⌬ double mutants. We note that the poor growth, nuclear abnormalities, and DNA damage sensitivity of rfp1⌬rfp2⌬ double mutants are similar to that of rad22⌬ cells (45,46). rad22 encodes the S. pombe Rad52 homolog, which is required for all recombinational repair in S. pombe (47) and is a target for SUMOylation (48).
As Rnf4 is known to be an E3 ubiquitin ligase (25), and Rfp1 and Rfp2 have conserved RING finger domains, it is reasonable to conclude that these proteins are similarly E3 ubiquitin ligases. In this context, it is possible that Rfp1 and Rfp2 are regulating the levels of proteins involved in SUMOylation, or perhaps Rfp1 and Rfp2 are involved in directing the ubiquitination and degradation of SUMOylated proteins, and hence these accumulate in the rfp1⌬rfp2⌬ double mutants. However, spontaneous slow growth suppressors of the rfp1⌬rfp2⌬ double mutants arise at high frequency, and intercrosses between such suppressors indicated that many different loci are mutated in these suppressors. This observation suggests that the accumulation of SUMOylated proteins in rfp1⌬rfp2⌬ double mutants is due to pleiotrophic effects, and although it is clear that this is the cause of the DNA repair defect, its mechanistic basis remains to be elucidated.