Genetic Interaction Landscape Reveals Critical Requirements for Schizosaccharomyces pombe Brc1 in DNA Damage Response Mutants

Brc1, which was first identified as a high-copy, allele-specific suppressor of a mutation impairing the Smc5-Smc6 holocomplex in Schizosaccharomyces pombe, protects genome integrity during normal DNA replication and when cells are exposed to toxic compounds that stall or collapse replication forks. The C-terminal tandem BRCT (BRCA1 C-terminus) domain of fission yeast Brc1 docks with phosphorylated histone H2A (γH2A)-marked chromatin formed by ATR/Rad3 checkpoint kinase at arrested and damaged replication forks; however, how Brc1 functions in relation to other genome protection modules remains unclear. Here, an epistatic mini-array profile reveals critical requirements for Brc1 in mutants that are defective in multiple DNA damage response pathways, including checkpoint signaling by Rad3-Rad26/ATR-ATRIP kinase, DNA repair by Smc5-Smc6 holocomplex, replication fork stabilization by Mrc1/claspin and Swi1-Swi3/Timeless-Tipin, and control of ubiquitin-regulated proteolysis by the COP9 signalosome (CSN). Exogenous genotoxins enhance these negative genetic interactions. Rad52 and RPA foci are increased in CSN-defective cells, and loss of γH2A increases genotoxin sensitivity, indicating a critical role for the γH2A-Brc1 module in stabilizing replication forks in CSN-defective cells. A negative genetic interaction with the Nse6 subunit of Smc5-Smc6 holocomplex indicates that the DNA repair functions of Brc1 and Smc5-Smc6 holocomplex are at least partially independent. Rtt107, the Brc1 homolog in Saccharomyces cerevisiae, has a very different pattern of genetic interactions, indicating evolutionary divergence of functions and DNA damage responses.

. Brc1-defective strains are sensitive to genotoxins that stall replication forks or create DNA lesions that lead to replication fork collapse or other forms of replication stress (Sheedy et al. 2005). Furthermore, brc1D cells have increased Rad52 foci, which indicate DNA replication difficulties even in the absence of exogenous genotoxins (Bass et al. 2012;Williams et al. 2010).
The presence of six BRCT (BRCA1 carboxyl terminus) domains is a defining structural feature of Brc1 that is shared with the evolutionary conserved Saccharomyces cerevisiae Rtt107 and human PTIP proteins (Munoz et al. 2007;Rouse 2004). These proteins also share the ability to bind histone H2A (or H2AX in mammals) that has been phosphorylated the ATM/ATR family of master DNA damage response checkpoint kinases (Li et al. 2012;Manke et al. 2003;Williams et al. 2010;Yan et al. 2011). This chromatin-specific interaction is mediated through the C-terminal pair of BRCT domains as also seen in DNA damage response mediator proteins such as human Mdc1 and fission yeast Crb2 (Du et al. 2006;Kilkenny et al. 2008;Stucki et al. 2005). Despite the overall structural similarities of Brc1, Rtt107, and PTIP and their importance for protecting genome integrity, it remains unclear whether they have conserved functions. Here, we investigate Brc1 by generating an epistatic miniarray profile (E-MAP) consisting of the quantitative analysis of genetic interactions between brc1D and a S. pombe gene deletion library . These E-MAP data provide novel insights into the functional relationships between Brc1 and other genome protection pathways in fission yeast.

Strains and genetic methods
The strains used in this study are listed in Supporting Information, Table S1. Standard fission yeast methods were used as described previously (Forsburg and Rhind 2006). New null alleles of csn1, csn5, ddb1, spd1, pnk1, sde2, raf1, and snt1 were constructed using targeting constructs in which the entire open reading frames were replaced by KanMX6 as described previously (Bahler et al. 1998). Successful deletion of these genes was verified by polymerase chain reaction. Tetrad analysis was performed to construct double mutants and verified by polymerase chain reaction.
Epistatic miniarray profile (E-MAP) E-MAP screens were performed and normalized as described previously (Roguev et al. 2008). Complete E-MAP profiles can be found in File S1.
Gene Ontology (GO) analysis GO enrichment analysis used the Princeton implementation of GO term finder (http://go.princeton.edu/cgi-bin/GOTermFinder) (Boyle et al. 2004). Analysis used a p-value cut off of 0.01. For the fission yeast brc1D E-MAP, the 56 SSL genes were compared with the background population of 2026 genes that produced E-MAP values (File S2). For the budding yeast rtt107D E-MAP, the 33 SSL genes (Collins et al. 2007) were compared with a background population consisting of all genes in budding yeast (File S3).
Survival assay DNA damage sensitivity assays were performed by spotting 10-fold serial dilutions of exponentially growing cells onto yeast extract with glucose and supplements plates, and treated with indicated amounts of hydroxyurea (HU), camptothecin (CPT), and methyl methanesulfonate (MMS). For ultraviolet (UV) treatment, cells were serial diluted onto yeast extract with glucose and supplements plates and irradiated using a Stratagene Stratalinker UV source. Cell survival was determined after 3-4 d at 30°.

Microscopy
Cells were photographed using a Nikon Eclipse E800 microscope equipped with a Photometrics Quantix charge-coupled device camera and IPlab Spectrum software. All fusion proteins were expressed at their own genomic locus. Rad52-yellow fluorescent protein (YFP)2 and RPA (Rad1)-green fluorescence protein2expressing strains were grown in Edinburgh minimal medium until mid-log phase for focus quantification assays. Quantification was performed by scoring 500 or more nuclei from three independent experiments.

Quantitative genetic interaction analysis of Brc1
To gain new functional insights into Brc1 we carried out an E-MAP analysis to quantify the genetic interactions between brc1D and a S. pombe gene deletion library of nonessential genes (Kim et al. 2010;Roguev et al. 2007). E-MAP values were determined with a simple growth phenotype that measures negative (aggravating) interactions, such as synthetic sick/lethal (SSL) interactions, as well as positive (alleviating) interactions in which the double mutant is healthier than would be expected based on the growth of the two single mutants. An SSL interaction often identifies proteins that function in distinct but parallel pathways, whereas a positive interaction score may indicate either suppression or masking effects, in which loss of one gene masks the effect of losing another, as seen when two proteins act together in a common complex or pathway (Collins et al. 2007;Roguev et al. 2007).
The resulting Brc1 E-MAP consists of 2026 interaction scores (Table S2). Of these, 56 genes displayed a significant negative genetic interaction with brc1Δ (interaction score , 22.5) and 23 displayed positive genetic interactions (interaction score .2). Most genes have genetic interactions scores close to zero. The results of this screen are summarized in Table S2. GO analysis of the SSL mutants identified in this screen revealed significant enrichment of genes involved in key cellular processes, including cellular response to DNA damage stimulus, DNA repair, DNA damage checkpoint, chromatin modification, and cullin deneddylation (Table 1). The strongest SSL score was obtained with csn1 (SSL = 215.1), which encodes a subunit of the COP9/Signalosome (CSN) complex that has important functions in the protection of genome integrity (Mundt et al. 1999). Genome protection was also highlighted by other genes with the greatest SSL scores, such as apn2 (SSL = 214.6), which encodes an apurinic/apyrimidinic endonuclease required for base excision repair (Fraser et al. 2003), hrq1 (SSL = 214.1), which encodes a RecQ type DNA helicase that plays an important role in DNA interstrand cross-link repair (Groocock et al. 2012), and rad26 (SSL = 212.5), which encodes an ATRIP ortholog required for the activity of Rad3/ATR checkpoint kinase (Edwards et al. 1999). For comparison, the recently analyzed brc1 SSL interaction with dcd1, which encodes a deoxycytidylate deaminase required to maintain a proper balance of dNTPs, was 27.8 in this screen (Sanchez et al. 2012). All of these data are consistent with Brc1 playing an important role in genome protection. GO analysis of the 23 genes that displayed positive genetic interactions with brc1D failed to yield specific process enrichment terms.

Comparison of Brc1 and Rtt107 E-MAPs
Fission yeast Brc1 and budding yeast Rtt107 are 6-BRCT domain proteins that bind gH2A and are important for survival of DNA lesions formed in S-phase (Bogliolo et al. 2007;Cobb et al. 2005;Fernandez-Capetillo et al. 2004;Marti et al. 2006;Papamichos-Chronakis and Peterson 2008;Ward and Chen 2001), yet it remains unclear whether they are functional orthologs. It was therefore of interest to compare the SSL E-MAP data for Brc1 and Rtt107 (Beltrao et al. 2010;Collins et al. 2007). We found that of the 56 S. pombe genes and 33 S. cerevisiae genes (Table S3 and Table S4), the E-MAP overlap encompassed only one gene, srs2/SRS2, which is an ATPdependent DNA helicase that functions as an anti-recombinase but is also required for efficient repair of DSBs in S-phase (Doe and Whitby 2004;Liu et al. 2011) This small degree of gene-for-gene overlap was surprising in view of the overall structural similarity of Brc1 and Rtt107. Moreover, global E-MAP comparisons of S. pombe and S. cerevisiae detected significant conservation of negative genetic interactions between genes with the same functional annotation (Roguev et al. 2008). Evolutionary divergences in key genome maintenance pathways likely account for some of E-MAP differences between Brc1 and Rtt107. For example, our Brc1 E-MAP includes subunits of the CSN protein complex (Mundt et al. 1999), which is absent in S. cerevisiae. Indeed, when we compared GO process analysis of negative genetic interactions of Brc1 (Table 1 and File S2) and Rtt107 E-MAPs (Table 2 and File S3), both highlighted significant enrichment in interactions with genes involved in DNA damage stimulus and DNA repair. On the other hand, other key GO process terms that were highly enriched in the Brc1 E-MAP analyses were absent in the comparable Rtt107 lists, notably DNA damage checkpoint and chromatin modification. Likewise, key GO process terms enriched for genes having negative genetic interactions with Rtt107 were absent the Brc1 E-MAP list, including nuclear division and double-strand break repair. Taken as a whole, these data suggest that although S. pombe Brc1 and S. cerevisiae Rtt107 are both involved in protecting genome integrity in response to DNA lesions that arise or are repaired preferentially in S-phase, their patterns of genetic interactions are not highly conserved, which indicates either substantial functional differences between Brc1 and Rtt107, large differences between S. pombe and S. cerevisiae in the mechanisms that maintain genome integrity or cause replicative stress when defective, or a combination of these effects.

Genotoxins enhance Brc1 genetic interactions
We sought to confirm and extend the analyses of a select group of the most interesting SSL interactions (Table 3). We created and tested new null alleles of COP9 signalosome complex genes (csn1, csn5), DNA repair and cell-cycle checkpoint genes (ddb1, rad26, rad17, srs2, pnk1), a DNA replication gene (swi3), and chromatin-silencing and remodeling genes (sde2, raf1, set1, snt1). As judged by colony size and density of double mutants compared with single mutants in serial dilution assays, we detected strong negative interactions with csn1, ddb1, and pnk1; moderate negative genetic interactions with csn5, rad26, rad17, srs2, swi3, and sde2; and only weak genetic interactions with raf1 and snt1. In most cases these negative genetic interactions were strikingly enhanced when cells were exposed to low or moderate doses of genotoxins such as UV light, HU, CPT, or MMS ( Figure 1, Figure 2, Figure S1, and Figure S2). Notably, the negative genetic interactions with raf1 and snt1 became obvious in the presence of these genotoxins. Of the 13 SSL interactions that were retested, only set1 failed to confirm the results of the large-scale E-MAP screen ( Figure S2C).
We also retested the positive genetic interactions between brc1 and csi1 (centromere clustering protein), msh2 (mismatch DNA repair MutS homolog), and cbp1 (centromeric DNA binding protein CENP-B homolog). None of these interactions were confirmed in dilution assays performed in the absence of genotoxins; however, the csi1D allele clearly suppressed brc1D genotoxin sensitivity ( Figure  S3A). In contrast, double mutants involving brc1D and msh2D or cbp1D grew more poorly than single mutants when tested in the presence of genotoxins ( Figure S3, B and C).
Deneddylation-independent activities of the CSN are especially critical in the absence of Brc1 Brc1 displays negative genetic interactions with Csn1 and Csn5, which are members of the CSN. The hallmark activity of CSN consists of the deneddylation of the cullin subunit of cullin-RING E3 ligases (CRLs), which favors CRL disassembly to maintain cycles of CRL assembly and disassembly that are needed protect CRL components from selfdestruction (Cope et al. 2002). As mentioned previously, we confirmed the strong negative genetic interaction between brc1 and csn1 by creating and testing a new csn1D null allele, which caused a modest growth defect that was substantially enhanced when combined with brc1D ( Figure 1A, untreated). As previously described, the csn1D cells were mildly sensitive to the topoisomerase I inhibitor CPT, the DNA alkylating agent MMS, UV light, and the ribonucleotide reductase (RNR) inhibitor HU, which slows DNA replication (Hayles et al. 2013;Mundt et al. 1999). In comparison with brc1D or csn1D strains, the double mutant brc1D csn1D cells displayed very poor growth in the presence of these genotoxins ( Figure 1A). We also confirmed the negative genetic interaction between brc1 and csn5 ( Figure 1B, untreated). Although initial studies indicated that csn5D mutants did not share the genotoxin sensitivity phenotype of csn1D cells (Mundt et al. 2002), in our assays csn5D cells were sensitive to n chronic exposure to HU, CPT and MMS, although less so than csn1D or brc1D mutants. Double mutant brc1D csn5D cells grew quite poorly in the presence of these genotoxins ( Figure 1B). Ddb1 is a core member of CLR4 (Cul4-Ddb1 RING ligase) that is target of CSN deneddylation activity. We confirmed that the modest growth defect caused by ddb1D was substantially enhanced when combined with brc1D (Figure 2 untreated). The ddb1D cells were mildly sensitive to UV, MMS, and HU (Zolezzi et al. 2002) and CPT ( Figure 2). We found that double mutant brc1D ddb1D cells were highly sensitive to these DNA-damaging agents ( Figure 2).
The negative genetic interactions between Brc1 and members of the CSN and CRL4 ubiquitin ligase imply that Brc1 and CSN independently act in genome maintenance pathways that are partially complementary. The JAMM motif within the MPN domain of Csn5 is responsible for the CRL deneddylation activity of the CSN (Cope et al. 2002;Mundt et al. 2002;Zhou et al. 2001). This catalytic function is dependent of the integrity of the complex (Mundt et al. 2002). The weaker SSL for the brc1D csn5D double mutant (E-MAP score = 23.8) compared with the brc1D csn1D double mutant (E-MAP score = 215.1) suggests that deneddylation-independent activities of the CSN are especially critical in the absence of Brc1.
Increased RPA and Rad52 foci in csn1D cells The SSL interaction between csn1D and brc1D suggested that the double mutant suffers increased rates of DNA damage or is unable to efficiently repair DNA lesions. To test this proposition we first monitored the formation of Replication Protein A (RPA) foci in csn1D and brc1D csn1D cells. RPA is the major single-strand DNA (ssDNA)binding protein in eukaryotic cells (Parker et al. 1997). Formation of RPA foci in untreated cells is thought to arise predominantly from replication fork stalling or collapse and subsequent homology-directed repair that involve resection of DNA ends to generate 39 ssDNA tails. For our assays we used strain in which Ssb1 (aka Rad11 in fission yeast), which is the largest subunit of RPA, was expressed with a YFP tag from the endogenous ssb1 + locus. In agreement with previous studies (Bass et al. 2012), we observed a significant increase in cells with RPA foci in the brc1D (16.0%) strain compared with wild type (7.1%). There was a much larger increase in cells with RPA foci in the n Table 2 Summary of significant enriched GO categories for biological function of genes with genetic interactions positively correlated with RTT107 (P £ 0.01) Groups   DNA metabolic process  RRM3, MND2, SGS1, RMI1, POL30, SPT4, TSA1, MRE11, SWI6,  XRS2, RTT101, SLX5, TEL1, TOP3, SRS2, MSH1, NEJ1, NUP84  DNA repair  RRM3, MRE11, XRS2, SGS1, SLX5, TEL1, SRS2, MSH1, NEJ1, POL30,  NUP84, SPT4  Double-strand break repair  SRS2, TEL1, NEJ1, NUP84,  n csn1D strain (38.8%). There was a further small increase in the brc1D csn1D strain (44.2%) although the difference with csn1D was not quite statistically significant (p-value = 0.08) ( Figure 3A). We also monitored foci formation of Rad52, previously known as Rad22, which is essential for all forms of homology-directed repair in fission yeast (Meister et al. 2003). As observed previously (Williams et al. 2010), the frequency of Rad52-YFP foci was significantly increased in brc1D cells (9.6%) as compared with the wild type. The incidence of cells with Rad52 foci was higher in the csn1D strain (21.2%), and there was a further significant increase in the brc1D csn1D strain (30.1%) ( Figure  3B). These findings suggest that Brc1 prevents replication fork instability in CSN-defective cells.

Process RTT107 E-MAP Functional
Defective relief of RNR inhibition in csn1D and ddb1D cells contributes to SSL interaction with brc1D Ddb1, Cullin 4 (Pcu4), and CSN subunits, Csn1 and Csn2, are required for degradation of Spd1, which is an inhibitor of RNR (Holmberg et al. 2005). As spd1 deletion partially suppresses Figure 1 Critical requirement of COP9/Signalosome (CSN) in brc1D cells. Genetic interaction between Brc1 and Csn1 (A) or Csn5 (B). 10-fold serial dilutions of the indicated strains were exposed to the indicating DNA-damaging agents. Plates were incubated at 30°for 324 d. CPT, camptothecin; HU, hydroxyurea; MMS, methyl methanesulfonate; WT, wild type.
genotoxin sensitivity in ddb1D and csn1D cells, we investigated whether the defect in degrading Spd1 contributed to the SSL interaction between brc1D and ddb1D or csn1D. Our genetic analyses revealed that spd1 deletion substantially suppressed the growth defects in brc1D ddb1D and brc1D csn1D backgrounds ( Figure 2 and Figure 4, untreated). This suppression was also evident to varying degrees in cells treated with a panel of genotoxins (UV, HU, MMS, and CPT) (Figure 2 and Figure 4). Taken together, these data indicate that the defect in relieving Spd1-mediated inhibition of RNR in ddb1D and csn1D cells is a major factor in the SSL interactions with brc1D, although other pathways involving Csn1 and Ddb1 must also contribute to these negative genetic interactions.
Requirement for gH2A in csn1D cells Rad3 checkpoint kinase, the fission yeast ortholog of mammalian ATR and budding yeast Mec1, plays a central role in replication stress response triggered by stalled and collapsed replication forks (Boddy et al. 1998;Lindsay et al. 1998). Rad3 has a number of important substrates, including the serine in the SQE motif in the C-terminal tail of histone H2A (Nakamura et al. 2004). Phospho-H2A, also known as gH2A, serves as a chromatin recruitment platform for Brc1, Crb2, and Mdb1, which all bind gH2A through their C-terminal BRCT domains.
To assess whether gH2A is important in the absence of CSN complex, we constructed a csn1D strain in which both histone H2A genes contained a mutation that changed the C-terminal SQE phosphory-lation motif to AQE (hta1-S129A hta2-S128A), which is the so-called htaAQ genotype. In comparison with the parental strains, the csn1D htaAQ strain displayed a reduced growth phenotype that was particularly evident in the presence of a panel of genotoxins (UV, HU, CPT, MMS) ( Figure 5A). Both Brc1 and Crb2 have well-established roles in DNA damage responses that are required for survival of genotoxic stress. Crb2 is required for activation of the checkpoint kinase Chk1 in response to DNA damage. As Chk1 was reported to have a synthetic growth defect with csn1D, we expected that Crb2 would be important in csn1D cells. Indeed, our studies revealed that csn1D crb2D cells grew poorly compared to the parental strains and this defect was accentuated in the presence of the panel of genotoxins ( Figure 5B). These data suggest that gH2A interactions with both Brc1 and Crb2 are important in csn1D cells.

DISCUSSION
In this study we used E-MAP to explore the genetic interactions of S. pombe Brc1, a protein with six BRCT domains that binds gH2A and is important for survival of replication stress. Brc1 was identified as an allele-specific high-copy suppressor of smc6-74 (Verkade et al. 1999), it becomes essential in strains with compromised Smc6 or Nse4 functions, and brc1D is also has strong negative genetic interactions with conditional alleles of rad60 and top2 (Morikawa et al. 2004;Pebernard et al. 2006;Verkade et al. 1999). Among the 56 SSL interactions Figure 3 Increased Rad52 and RPA foci in csn1D cells. Cells expressing Rad52-yellow fluorescentprotein (A) or RPA(Ssb1)-green fluorescent protein (B) were cultured in minimal medium at 25°until mid-log phase. Foci were scored in three independent experiments. Rad52 foci in brc1Dcsn1D cells are statistically increased relative to csn1Δ cells (two-tailed Student's t-test, P-value 0.0015). Error bars correspond to standard deviations of the means. Asterisk depicts statistically significant differences between the bracketed strains as determined by a two-tailed Student's t-test, P-value # 0.05. revealed in our E-MAP analysis, only four were previously detected through classical genetic analyses: apn2 (E-MAP score = 214.58), encoding an apurinic/apyrimidinic endonuclease and rad2 (E-MAP score = 22.98), encoding a FEN1 endonuclease, both of which are involved in base excision repair (Alseth et al. 2004;Alseth et al. 2005); mms22 (E-MAP score = 22.44), encoding a DNA repair protein that forms a complex with Mms1 (Dovey and Russell 2007); and ssb3 (E-MAP score = 210.97), encoding the nonessential small subunit of the replication protein A (Cavero et al. 2010). An additional nine genes (csn1, hrq1, pmt3, rad1, rad17, SPCC1442.02, srs2, pnk1, and csn5) found in our E-MAP were detected in previous genetic interaction screens (Beltrao et al. 2009;Ryan et al. 2012). This list is strongly enriched for genes involved in genome stability but our results show that it is not exhaustive. Our screen revealed an additional 11 SSL interactions with genes that have established roles in genome protection, including ddb1, rad9, pku80, hus1, rad26, dcd1, nse6, swi3, csn71, dbl1, and mrc1. Indeed, GO analysis of the SSL interactions identified in our screen revealed very strong enrichment for genes involved in DNA repair and checkpoint functions.
Some of the new SSL interactions detected in our screen validated previous findings. For example, the SSL interactions with the Rad1 subunit of the Rad9-Hus1-Rad1 checkpoint clamp and the Rad17 subunit of the Rad17-RFC checkpoint clamp loader were uncovered in previous screens (Beltrao et al. 2009;Ryan et al. 2012). Our screen additionally detected SSL interactions with Rad9 and Hus1, thereby detecting all four genes in the Rad9, Hus1, Rad1, and Rad17 genetic epistasis group. Indeed, it is impressive that E-MAP scores for rad1, rad9, and rad17 were so similar, 28.7, 28.7, and 28.4, respectively, which attests to high accuracy of the measurements in this particular implementation of the E-MAP procedure.
Similarly, previous brc1D E-MAP screens identified csn1 and csn5 (Beltrao et al. 2009;Ryan et al. 2012), which are subunits of the CSN complex required for cullin deneddylation (Mundt et al. 2002), whereas our screen identified these genes as well as the CSN subunit csn71. In addition, our brc1D E-MAP screen detected an SSL interaction with Ddb1, which is required for degradation of the RNR inhibitor Spd1 (Bondar et al. 2004;Holmberg et al. 2005).
Among the CSN subunits identified in our screen, we found that negative genetic interaction was strongest with csn1. This observation suggests that loss of the deneddylation activity dependent on the Csn5 subunit is not fully responsible of the SSL interaction between Brc1 and CSN, nor are the SSL interactions totally explained by the role of Ddb1/Csn1 in controlling RNR activation through Spd1 degradation. These results suggest that CSN and Brc1 function in parallel in response to DNA damage and contribute to genome stability through multiple pathways. Supporting this idea, our studies revealed that csn1D mutants have increased numbers of RPA and Rad52 foci. Our studies also reveal the importance of gH2A in the absence of CSN, with our data indicating that binding of both Brc1 and Crb2 to gH2A is important in response to replication stress. Interestingly, deregulation of CSN and its interactions are related to multiple cancers, making CSN an interesting target for cancer therapy (Fuzesi-Levi et al. 2014;Lee et al. 2011;Richardson and Zundel 2005).
The SSL interaction of brc1 with nse6 provides clues about the functional relationships between Brc1 and the Smc52Smc6 complex. As mentioned previously, Brc1 was initially discovered as an allelespecific, high-copy suppressor of smc6-74 (Verkade et al. 1999). This type of genetic interaction often indicates a physical association; for example, the missense mutation in smc6-74 might impair binding to Brc1, which is a defect that might be suppressed by increasing the total cellular concentration of Brc1. It is unknown if Brc1 associates with the Smc52Smc6 holocomplex; however, the Brc1 homolog in S. cerevisiae coprecipitates with multiple subunits of the Smc52Smc6 holocomplex (Ohouo et al. 2010). Nse5 and Nse6 form a heterodimer that is part of the Smc5-Smc6 holocomplex and is required for many or all of its DNA repair functions, but unlike other subunits of the holocomplex Nse5 and Nse6 are not essential for cell viability (Pebernard et al. 2006). The SSL interaction of brc1 with nse6 detected in our screen strongly indicates that the DNA repair functions of Brc1 and Smc5-Smc6 holocomplex are at least partially independent.
The SSL interaction of brc1 with mrc1 is novel and provides insights about the requirement for Brc1 in the response to replication stress. Mrc1 (mediator of replication checkpoint) was discovered by screening for mutations that cause hydroxyurea sensitivity and are rescued by overproduction of the replication checkpoint kinase Cds1/Chk2 (Tanaka and Russell 2001). Mrc1 is conserved in budding yeast and mammals in which it is known as Mrc1 and claspin, respectively (Alcasabas et al. 2001;Kumagai and Dunphy 2000). The Figure 4 Genetic interactions among Brc1, Csn1, and Spd1. 10-fold serial dilutions of the indicated strains were exposed to the indicating DNA damaging agents. Plates were incubated at 30°for 324 d. CPT, camptothecin; HU, hydroxyurea; MMS, methyl methanesulfonate; WT, wild type. mrc1 + gene in fission yeast is periodically transcribed during S-phase in the cell cycle and recruiting Cds1 to stalled replication forks by Mrc1 is required for its efficient activation of Cds1. Mrc1 appears to have both Rad53-dependent and -independent functions that stabilize replication forks in S. cerevisiae (Katou et al. 2003;Osborn and Elledge 2003), but it is unclear whether Mrc1 has Cds1-independent activities in fission yeast (Nitani et al. 2006). In our brc1D E-MAP we uncovered a significant SSL interaction with mrc1D (E-MAP = 22.52) but not with cds1D (E-MAP = 20.60), even though cds1D mutants are more severely sensitive to HU (Tanaka and Russell 2001). These data indicate that the absence of Brc1 enhances the requirement for a Cds1-independent function of Mrc1 in stabilizing replication forks. Similar conclusions are suggested by the SSL interaction of brc1D with swi3D (E-MAP = 24.6). Swi3 binds Swi1 to form the fork protection complex that stabilizes stalled replication forks Noguchi et al. 2004). This activity is required for robust activation of Cds1 in response to HU treatment and other forms of replication stress. However, the absence of an SSL interaction between brc1D and cds1D mutations suggests that Cds1-independent activity of Swi1-Swi3 fork protection complex is especially critical in the absence of Brc1.
Although we focused on the SSL interactions identified in our brc1D E-MAP, we did confirm the alleviating (positive) interaction with csi1 (E-MAP = +2.14). Csi1 was implicated in centromere clustering during interphase through its interaction with Sad1 in the spindle pole body and it also has a role in tethering spindle-stabilizing factors to the spindle pole body for promoting bipolar spindle assembly (Hou et al. 2012;Zheng et al. 2014). The involvement of Csi1 in these processes is interesting in light of our evidence that recruiting Brc1 to gH2A in pericentromeric heterochromatin during S-phase contributes to maintaining the heterochromatic state, which is required for efficient chromosome segregation during nuclear division (Lee et al. 2013). Indeed, genetic assays indicate that Brc1 is required for mitotic chromosome stability, which suggests a role for Brc1 in chromosome segregation (Verkade et al. 1999). Furthermore, we found that brc1D cells are moderately sensitive to the microtubuledestabilizing drug thiabendazole and display increased rates of chromosome missegregation in the presence of thiabendazole (Lee et al. 2013). These effects of Brc1 correlate with the genetic data linking Brc1 to the Smc5-Smc6 complex (Verkade et al. 1999) and data showing that the holocomplex localizes around centromeres during S-phase and defects in the complex increase the frequency of lagging chromosomes during nuclear division (Pebernard et al. 2008). However, despite these striking correlations, it is unclear why a defect in Csi1 function should alleviate the requirement for Brc1 as suggested by our genetic suppression data. In this regard it is interesting that csi1D cells are sensitive to the DNA-damaging agent 4-nitroquinoline 1-oxide, which causes replication stress by producing bulky adducts in DNA (Deshpande et al. 2009). Our studies further indicated that csi1D cells are mildly sensitive to UV light, HU, and CPT ( Figure S3). Most strikingly, the csi1D mutation effectively suppresses sensitivity of brc1D cells to these genotoxins and MMS. Again, it is unobvious how this suppression happens, although we note that there are some well-known examples of mutations in different DNA repair pathways having suppressive effects; for example, eliminating the Ku complex required for nonhomologous end joining (NHEJ) suppresses defects in the Mre11-Rad50-Nbs1 complex and Ctp1 that are required for homologous recombination repair (Langerak et al. 2011), and the rad51D mutations suppresses UV sensitivity of nse6D mutants (Pebernard et al. 2006).
Finally, the list of SSL interactions derived from the E-MAP studies for Brc1 (56 SSL interactions) and Rtt107 (33 SSL interactions) reveal remarkably little overlap, with only one gene, the ATP-dependent DNA helicase srs2/SRS2, being found in both screens. In S. pombe, deletion of srs2 causes elevated rates of spontaneous recombination (Doe and Whitby 2004). Furthermore, deletion of brc1 suppressed the hyper-recombination phenotype of an srs2D strain (Bass et al. 2012). This small degree of overlap suggests major functional differences between Brc1 and Rtt107 despite their similarities in domain organization and a shared mechanism of localizing to DNA lesions through C-terminal BRCT domains binding gH2A (Li et al. 2012;Williams et al. 2010). However, comparing all genetic interactions identified by classical genetic analyses and E-MAP suggests an additional degree of overlap for Brc1 and Rtt107. For example, classical genetic interactions uncovered strong negative interactions with Rqh1 and Sgs1, which are orthologous DNA helicases of the RecQ family. Nevertheless, the unexpectedly low overlap for both E-MAP lists and GO process terms suggests significant functional differences between Brc1 and Rtt107, reflecting the large evolutionary divergence between S. pombe and S. cerevisiae. Figure 5 Critical requirement of gH2A in csn1D cells. Genetic interaction between Csn1 and htaAQ (A) or Crb2 (B). 10-fold serial dilutions of the indicated strains were exposed to the indicating DNAdamaging agents. Plates were incubated at 30°for 324 d. CPT, camptothecin; HU, hydroxyurea; MMS, methyl methanesulfonate; WT, wild type.

ACKNOWLEDGMENTS
We thank Sophie Rozenzhak for her contributions to the early stages of this project and Oliver Limbo for his experimental support and guidance. This work was supported by a QB3@UCSF grant and National Institutes of Health grants GM084448, GM084279, GM081879, and GM098101 awarded to N.J.K. and GM059447, CA077325, CA117638, and P42ES010337 awarded to P.R.