Regulation of Mus81-Eme1 structure-specific endonuclease by Eme1 SUMO-binding and Rad3ATR kinase is essential in the absence of Rqh1BLM helicase

The Mus81-Eme1 structure-specific endonuclease is crucial for the processing of DNA recombination and late replication intermediates. In fission yeast, stimulation of Mus81-Eme1 in response to DNA damage at the G2/M transition relies on Cdc2CDK1 and DNA damage checkpoint-dependent phosphorylation of Eme1 and is critical for chromosome stability in absence of the Rqh1BLM helicase. Here we identify Rad3ATR checkpoint kinase consensus phosphorylation sites and two SUMO interacting motifs (SIM) within a short N-terminal domain of Eme1 that is required for cell survival in absence of Rqh1BLM. We show that direct phosphorylation of Eme1 by Rad3ATR is essential for catalytic stimulation of Mus81-Eme1. Chk1-mediated phosphorylation also contributes to the stimulation of Mus81-Eme1 when combined with phosphorylation of Eme1 by Rad3ATR. Both Rad3ATR- and Chk1-mediated phosphorylation of Eme1 as well as the SIMs are critical for cell fitness in absence of Rqh1BLM and abrogating bimodal phosphorylation of Eme1 along with mutating the SIMs is incompatible with rqh1Δ cell viability. Our findings unravel an elaborate regulatory network that relies on the poorly structured N-terminal domain of Eme1 and which is essential for the vital functions Mus81-Eme1 fulfills in absence of Rqh1BLM.

Introduction Structure-specific DNA endonucleases are cornerstones in the proper execution of DNA replication, repair and recombination; yet they harbor the potential for causing genome instability. Controlling these enzymes is essential to ensure efficient processing of appropriate substrates while preventing counterproductive targeting of other similar DNA structures. The Mus81-Eme1 structure-specific endonuclease (SSE) emerged from early on as a key player in the processing of recombination intermediates that form during homology directed repair of two-ended double strand breaks or during the rescue of stalled or broken replication forks [1][2][3][4][5][6][7]. Over this last decade, control mechanisms of Mus81-Eme1 that are tightly linked to cell cycle progression have been identified.
In Saccharomyces cerevisiae (S. cerevisiae), Mus81-Mms4 EME1 is stimulated at the G2/M transition by Cdc5 PLK1 -, Cdc28 CDK1 -and Dbf4-dependent phosphorylation of Mms4 EME1 [8][9][10][11][12]. In human cells, catalytic upregulation of MUS81-EME1 is driven by complex formation with the SLX4 nuclease scaffold. This regulation is mediated by direct interaction between MUS81 and SLX4, which is strongly stimulated at the G2/M transition by phosphorylation of SLX4 by CDK1 [13]. Reminiscent of what has been observed in S. cerevisiae, maximal processing of joint molecules such as Holliday junctions by MUS81-EME1 also correlates with hyperphosphorylation of EME1, possibly by CDK1 or PLK1 [11]. Whether phosphorylation of EME1 contributes to the stimulation of MUS81-EME1 in human cells remains to be formally demonstrated. These mechanisms ensure that joint molecules such as Holliday junctions, Dloops or replication intermediates at under-replicated loci are efficiently resolved in mitosis before chromosome segregation [14]. By restricting the catalytic stimulation of Mus81-Eme1 to late stages of the cell cycle these control mechanisms also ensure that joint molecules and replication intermediates get a chance to be processed by more conservative non-endonucleolytic mechanisms that rely on their unfolding by RecQ helicases such as the BLM helicase in human cells and its Sgs1 and Rqh1 orthologs in Saccharomyces cerevisiae (S. cerevisiae) and Schizosaccharomyces pombe (S. pombe), respectively. These temporal controls further prevent the accumulation of hyper-activated Mus81-Eme1 in S-phase and the risk of the unscheduled processing of replication intermediates. The importance of such control mechanisms is underscored by the marked genomic instability that is caused by the premature stimulation of Mus81-Mms4 in budding yeast cells that produce an Mms4 mutant that mimics a constitutively phosphorylated Mms4 protein [9]. Interestingly, SUMOylation and ubiquitination of Mms4 were recently shown to specifically target phosphorylated Mms4 for degradation by the proteasome, further ensuring that hyperactivation of Mus81-Mms4 is restricted to mitosis [15]. In human cells, SLX4-MUS81 complex formation induced by premature activation of CDK1 results in the unscheduled processing of replication intermediates genome wide and chromosome pulverization [13].
In S. pombe, upregulation of Mus81-Eme1 also relies on the phosphorylation of Eme1 by Cdc2 CDK1 [16]. However, in contrast to what has been described in S. cerevisiae, phosphorylation of Eme1 by Cdc2 CDK1 primes Eme1 for further DNA damage checkpoint-mediated phosphorylation in response to DNA damage. This elaborate control mechanism ensures that Mus81-Eme1 is rapidly hyperactivated in response to DNA damage in late G2 and mitosis and is critical to prevent gross chromosomal rearrangements in absence of the BLM-related helicase Rqh1 BLM [16].
To gain further insight into the molecular mechanisms involved in the control of Mus81-Eme1 in S. pombe, we undertook in silico analyses of a relatively short N-terminal domain of Eme1 that we found to be essential for cell viability in the absence of Rqh1 BLM . We identified Rad3 ATR consensus phosphorylation sites and two SUMO interacting Motifs (SIM1 and SIM2) within that domain. We demonstrate that Eme1 is a direct substrate for Rad3 ATR both in vitro and in vivo and show that phosphorylation of Eme1 by Rad3 ATR plays a prominent role in the catalytic stimulation of Mus81-Eme1 in response to DNA damage. Fullfledged stimulation of Mus81-Eme1 also relies on phosphorylation of Eme1 by Chk1 but, apparently, only when it occurs in conjunction with Rad3 ATR -mediated phosphorylation of Eme1. We provide genetic evidence that both Rad3 ATR -and Chk1-mediated phosphorylation of Eme1 are independently critical for cell fitness in the absence of Rqh1 BLM . Remarkably, Chk1-mediated phosphorylation of Eme1 is lost when SIM2 is mutated, while mutating SIM1 has no impact. Both SIMs are important for cell fitness in absence of Rqh1 BLM , and abrogating phosphoregulation of Mus81-Eme1 and mutating SIM1 and SIM2 recapitulates the synthetic lethality observed by deleting the N-terminus of Eme1 in the absence of Rqh1 BLM .

Eme1 N-terminus is essential in absence of Rqh1 BLM
To gain further insight into the mechanisms underlying the catalytic stimulation of Mus81-Eme1 in response to DNA damage, we searched for domains of Eme1 that are essential in absence of Rqh1 BLM while dispensable for the intrinsic catalytic activity of Mus81-Eme1. Interestingly, we found that deleting a relatively short N-terminal domain (residues 1-117) of Eme1 is synthetic lethal with rqh1Δ (Fig 1A). Importantly, this domain does not contain the Cdc2 CDK1 sites that we had previously reported to be involved in the stimulation of Mus81-Eme1 in response to DNA damage and to be critical for cell viability in absence of Rqh1 BLM [16]. A detailed in silico analysis of the first 117 residues of Eme1 led to the identification of two SQ/TQ Rad3 ATR consensus sites (S 23 Q and T 50 Q) and two putative SUMO-Interacting Motifs (SIMs), hereafter named SIM1 and SIM2, which matched the described (V/I)-X-(V/I)-X-(V/I/L) consensus sequence (Fig 1B) [17]. These observations suggested that Eme1 might be a direct substrate for Rad3 ATR and possess SUMO-binding properties. We tested these predictions in the following experiments.

Eme1 is a direct target of Rad3 ATR kinase
Eme1 contains in total eight putative Rad3 ATR consensus phosphorylation sites (S 23 Q, T 50 Q, S 126 Q, T 145 Q, T 215 Q, S 313 Q, T 384 Q, S 458 Q) (S1 Fig). To determine whether Eme1 is a direct substrate of Rad3 ATR we set up in vitro kinase assays with recombinant Mus81-Eme1 and Rad3 ATR . Recombinant Mus81(6His)-(MBP)Eme1 was produced in E. coli and affinity purified on Ni ++ and amylose resins (Fig 2A). Recombinant (GFP)Rad3 ATR was instead transiently overproduced in yeast cells exposed to bleomycin to induce DNA damage and activate Rad3 ATR . We used a chk1Δ cds1Δ rad3Δ mutant strain to eliminate the possibility of endogenous checkpoint kinases co-purifying with (GFP)Rad3 ATR . As shown in Fig 2B, Eme1 was efficiently phosphorylated by (GFP)Rad3 ATR . In contrast, an Eme1 8AQ mutant where all eight SQ/ TQ sites are mutated to AQ is barely detected on the autoradiograph. To explore which SQ/ TQ sites were important for phosphorylation we tested the impact of mutating Rad3 ATR phosphorylation sites in three different clusters (S1 Fig). The strongest effect was seen with mutations in Cluster 1, whilst mutations in Clusters 2 and 3 had a milder impact (Fig 2B and 2C). Noticeably, in vitro Rad3 ATR -mediated phosphorylation of Eme1 does not require the priming by Cdc2 CDK1 . Taken together, these data indicate that the S 23 Q and T 50 Q sites in the N-terminus of Eme1 are most critical for its in vitro phosphorylation by Rad3 ATR .
To investigate whether Eme1 is a substrate for Rad3 ATR in vivo, we generated an eme1 8AQ mutant strain in which all SQ/TQ sites are mutated to AQ (S1 Fig). As expected, whereas the cell-cycle dependent phosphorylation profile of Eme1 8AQ is comparable to that of the WT protein (Fig 2D), we observed a strong reduction of the CPT-induced phosphorylation (Fig 2E). However, DNA damage-dependent phosphorylation was not totally abolished in the Eme1 8AQ background (Fig 2E). We suspected that this residual phosphorylation of Eme1 8AQ was catalyzed by Chk1 (Fig 2F). Accordingly, we observed a complete loss of CPT-induced phosphorylation of Eme1 8AQ in eme1 8AQ chk1Δ cells (Fig 2E).
Overall, our data strongly indicate that Eme1 is phosphorylated by both Chk1 and Rad3 ATR following activation of the DNA damage checkpoint.

Phosphorylation of Eme1 by Rad3 ATR is crucial in absence of Rqh1 BLM
We previously showed that in vivo DNA damage-induced hyperphosphorylation of Eme1 is strictly subordinated to prior Cdc2 CDK1 -dependent phosphorylation [16]. Accordingly, mutating four CDK consensus target sites totally abrogates not only the cell-cycle dependent phosphorylation of the resulting Eme1 4SA protein but also its phosphorylation in response to DNA damage [16]. Importantly, the same study found that while an eme1 4SA single mutant displays no abnormal phenotype, an eme1 4SA rqh1Δ double mutant is extremely sick. These observations suggest that the Cdc2 CDK1 -dependent phosphorylations of Eme1 are required for its phosphorylation by both Rad3 ATR and Chk1, and that these events are critical in the absence of Rqh1 BLM .
To investigate the functional relevance of phosphorylation of Eme1 by Rad3 ATR , we introduced eme1 8AQ mutations in the rqh1Δ background. While we observed no obvious phenotype  Rad3 ATR directly phosphorylates Eme1 in vitro. A-Recombinant mus81 was co-expressed with either wildtype eme1, eme1 cluster1 � , eme1 cluster2 � or eme1 cluster3 � and purified from bacterial cultures (purification scheme is depicted on the left). Samples were loaded on an 4-12% Nupage Bis-Tris polyacrylamide gel alongside a Rainbow (Amersham) protein size marker. B-Representative autoradiography of 32 P labeled Eme1. Rad3 ATR in vitro kinase assays were carried out on full-length recombinant Mu81-Eme1 complexes containing either wild-type Eme1 or Eme1 mutated for S 23 /T 50 (cluster1 � ) or S 126 /T 145 /T 215 (cluster2 � ) or S 313 /T 384 /S 458 (cluster3 � ) or S 23 /T 50 /S 126 /T 145 /T 215 /S 313 / T 384 /S 458 (eme1 8AQ ). Samples were loaded on a 3-8% Nupage Tris-Acetate polyacrylamide gel alongside a PageRuler (ThermoFisher) protein size marker. C-Relative band intensity of phosphorylated Eme1 (n = 3). All clusters contribute to Eme1 Rad3 ATR -dependent phosphorylation with S 23 /T 50 (cluster1 � ) being the most important. D-Cultures from cdc25-22 TAP-eme1 and cdc25-22 TAP-eme1 8AQ were synchronized at the G2/M transition and released for one cell cycle. Total proteins were extracted at each indicated time point of the time course and analyzed by for the eme1 8AQ single mutant, the eme1 8AQ rqh1Δ double mutant displayed pronounced growth and colony formation defects compared to the rqh1Δ single mutant (Figs 3A, 3B, 3C and S2A). This genetic interaction was further exacerbated by exposure to genotoxic agents (S2B Fig). Noteworthy, introducing the Eme1 SQ/TQ cluster mutations in the rqh1Δ background resulted in smaller colonies compared to the parental cells (S2C Fig). In line with the prominent contribution of Cluster 1 to Rad3 ATR -mediated phosphorylation of Eme1 in vitro (Fig 2B and 2C), we noticed a more pronounced effect of Cluster 1 mutations on colony size in absence of Rqh1 BLM compared to the other two cluster mutations (S2C Fig). Overall, our data indicate that direct phosphorylation of Eme1 by Rad3 ATR is important for cell viability in the absence of Rqh1 BLM .

Rad3 ATR direct phosphorylation of Eme1 contributes to the catalytic stimulation of Mus81-Eme1
We have previously shown that Rad3 ATR contributes to the catalytic stimulation of the HJresolvase activity of Mus81-Eme1 [16]. We inferred at that time that the Rad3 ATR -Chk1 axis was involved in this catalytic control. The finding that Rad3 ATR can directly phosphorylate Eme1, and that this phosphorylation is crucial in absence of Rqh1 BLM , prompted us to investigate whether it contributed to the catalytic stimulation of Mus81-Eme1.
As previously reported [16], hyperphosphorylation of Eme1 following activation of the DNA damage checkpoint by CPT correlates with increased HJ-resolvase activity of Mus81-Eme1 complex isolated from fission yeast cells expressing TAP-tagged Eme1 (Figs 4A, 4B, 4C, S3A and S3B). Accordingly, no catalytic stimulation is detected when Mus81-Eme1 is recovered from rad3Δ cells ( [16] and Figs 4A, 4B, 4C, S3A and S3C). Remarkably, we find that the Mus81-Eme1 8AQ complex is not stimulated following CPT treatment despite a fully functional DNA damage checkpoint and Chk1-mediated phosphorylation of Eme1 in eme1 8AQ cells (Figs 2E, 4A, 4B, 4C, S3A and S3D). In contrast to the complete loss of catalytic stimulation of Mus81-Eme1 in absence of Rad3 ATR , some degree of catalytic stimulation of Mus81-Eme1 remains detectable in chk1Δ cells following CPT treatment (Figs 4A, 4B, 4C, S3A and S3E). Taken together, these data demonstrate that full-fledged catalytic stimulation of Mus81-Eme1 relies on phosphorylation of Eme1 by both Rad3 ATR and Chk1 kinases ( Fig  4D). They also suggest that Chk1-mediated phosphorylation contributes to full activation of Mus81-Eme1 only when combined with phosphorylation of Eme1 by Rad3 ATR .

Eme1 contains bona-fide SIMs
Having confirmed our predictions that Eme1 is directly phosphorylated by Rad3 ATR and shown that this is critical for catalytic stimulation of Mus81-Eme1 and cell fitness in absence of Rqh1 BLM , we next undertook the analysis of the predicted SUMO-binding properties of Eme1 mediated by the putative SIM1 and SIM2 motifs in the N-terminal domain of Eme1 (Figs 1B and 5A).
The SUMO-binding capacities of SIM1 and SIM2 were assessed by a yeast two-hybrid assay against the unique S. pombe SUMO ortholog Pmt3. A fragment of Eme1 (Eme1 1-130 ) containing SIM1 and SIM2 displayed strong binding to Pmt3, confirming that the N-terminus of Eme1 possesses SUMO-binding properties (Fig 5A). Introducing point mutations in the Western blot using an antibody raised against the Flag tag of Eme1. Ponceau stained membranes are depicted as loading control. E-Western blot on total lysates from untreated or 40 μM CPT-treated cells of the indicated background. Tubulin is used as a loading control. F-Scheme depicting the relative contributions of Cdc2 CDK1 , Rad3 ATR and Chk1 kinases in wild-type, chk1Δ and eme1 8AQ backgrounds. Note: TAP-= 2xProtA-TEVsite-2xFlag-.
https://doi.org/10.1371/journal.pgen.1010165.g002 conserved aliphatic residues of SIM1 strongly impaired interaction with Pmt3 while mutations in SIM2 had a milder effect (Fig 5A). Mutations in both SIMs led to complete loss of interaction with Pmt3 (Fig 5A). Our data confirm that the N-terminal domain of Eme1, which is essential for cell viability in absence of Rqh1 BLM (Fig 1A), contains bona fide SIMs that jointly contribute to the SUMO-binding properties of Eme1, with a predominant contribution made by SIM1.

Eme1 SIMs are required in absence of Rqh1 BLM
To assess the functional relevance of SIM1 and SIM2, we generated mutant strains harboring the point mutations, described in Fig 5A, None of the three eme1 SIM1 � , eme1 SIM2 � and eme1 SIM1 � +SIM2 � mutant strains presented any obvious growth defect or reduced fitness compared to a WT strain in absence of exogenous DNA damage (Fig 5B and 5C) and following CPT treatment (Fig 5D). Since eme1 Δ117 rqh1Δ double mutants are non-viable (Fig 1A), we next assessed the importance of SIM1 and/or SIM2 for cell viability in absence of Rqh1 BLM . As shown in Fig 5E  and 5F, while mutating SIM1 in the rqh1Δ background does not reduce colony formation capacities of the resulting eme1 SIM1 � rqh1Δ double mutant compared to the rqh1Δ single mutant, it leads to a marked increase in the proportion of elongated and sick cells (S4A Fig). In contrast, mutating SIM2 strongly impairs the ability of eme1 SIM2 � rqh1Δ to form viable colonies (Fig 5E and 5F) in addition to causing a strong increase in the number of sick cells (S4A Fig). Simultaneously mutating both SIMs did not further impair colony formation capacities compared to eme1 SIM2 � rqh1Δ cells (Fig 5E and 5F). However, it had an additive effect regarding the proportion of elongated and sick cells (S4A Fig). We further looked at cell fitness following chronic exposure to CPT. Loss of Rqh1 BLM in eme1 SIM1 � mutant slightly exacerbated CPT sensitivity compared to rqh1Δ. This effect was slightly more pronounced for eme1 SIM2 � rqh1Δ mutants while the eme1 SIM1 � +SIM2 � rqh1Δ mutants displayed the steepest increase in CPT sensitivity compared to rqh1Δ (S4B Fig). Overall, these data demonstrate that both SIMs contribute to the essential role of the Eme1 1-117 Nterminal domain in absence of Rqh1 BLM , with a more prominent contribution made by SIM2.

Contribution of SIM1 and SIM2 to Eme1 phosphorylation and Mus81-Eme1 stimulation
The importance of the Eme1 SIMs in absence of Rqh1 BLM prompted us to assess possible functional ties between the SUMO-binding properties of Eme1 and its phosphorylation. Interestingly, whereas mutating the SIMs had no obvious effect on the phosphorylation profile of Eme1 SIM1 � +SIM2 � throughout the cell cycle (Fig 6A), it substantially reduced phosphorylation levels in response to CPT (Fig 6B). It also reduced the catalytic stimulation of the Mus81-Eme1 SIM1 � +SIM2 � complex, compared to the WT complex, following CPT treatment (Figs 6C, S3A and S5). This suggested that mutating the SIMs did not abrogate phosphorylation of Eme1 by Rad3 ATR , which we showed results in complete loss of DNA damage induced stimulation of Mus81-Eme1 (Figs 4A, 4B, 4C, S3A, S3C and S3D). In line with this,  A-A fragment of Eme1 (eme1 1-130 ) containing SIM1 and SIM2 was used as bait in a Yeast two-hybrid assay for interaction with SUMO PMT3 . The SIMs consensus sequences are indicated as well as the mutations introduced in each of them. B-Exponentially growing wild type, eme1Δ, eme1 SIM1 � , eme1 SIM2 � and eme1 SIM1 � +SIM2 � cells were seeded on YES plates by micromanipulation and allowed to grow for 3 days at 32˚C. C-Average percentage (± s.d.) of viable colonies (n = 3 independent experiments). Statistical significance is measured with one-way ANOVA followed by Tuckey post-test. D-Five-fold dilutions of wild-type, rqh1Δ, eme1 SIM1 � , eme1 SIM2 � and eme1 SIM1 � +SIM2 � cells were plated on medium supplemented or not with the indicated concentrations of CPT followed by incubation at 30˚C. E-Exponentially growing rqh1Δ, eme1 SIM1 � , eme1 SIM2 � and eme1 SIM1 � +SIM2 � as well as two independent clones of eme1 SIM1 � rqh1Δ, eme1 SIM2 � rqh1Δ and eme1 SIM1 � +SIM2 � rqh1Δ were seeded on YES plates by micromanipulation and allowed to grow for 3 days at 32˚C. F-Average percentage (± s.d.) of viable colonies (n = 3 independent experiments). Statistical significance is measured with one-way ANOVA followed by Tuckey post-test.
https://doi.org/10.1371/journal.pgen.1010165.g005  vitro (S6 Fig). To assess whether the residual CPT-induced mobility shift of Eme1 SIM1 � +SIM2 � is due to phosphorylation by Rad3 ATR in vivo, we generated an eme1 8AQ+SIM1 � +SIM2 � mutant. Mutating all Rad3 ATR consensus phosphorylation sites and both SIMs resulted in the complete loss of CPT-induced mobility shift of Eme1 (Fig 6D), confirming that Eme1 SIM1 � +SIM2 � remains phosphorylated by Rad3 ATR in response to DNA damage. These data suggest that the SIMs promote Chk1-mediated phosphorylation of Eme1 and as such contribute to the catalytic stimulation of Mus81-Eme1, which is critical for cell fitness in absence of Rqh1 BLM [16]. To gain further insight into the respective contribution of SIM1 and SIM2 to Chk1-mediated phosphorylation of Eme1, we compared the CPT-induced mobility shifts of Eme1 SIM1 � and Eme1 SIM2 � mutants (Fig 6E). While mutating SIM1 had barely any impact, mutating SIM2 substantially reduced phosphorylation levels in response to CPT (Fig 6E). Based on these findings and the severe growth defect of an eme1 SIM2 � rqh1Δ double mutant (Figs 5E, 5F and S4A), we propose that phosphorylation of Eme1 by Chk1, promoted by SIM2, represents an additional phosphorylation-based layer of control of Mus81-Eme1 that is critical for cell viability in absence of Rqh1 BLM (Fig 6F).

SUMO-binding capacities and phosphorylation cooperate for cell viability in absence of Rqh1 BLM
To further assess whether we have unraveled layers of regulation of Mus81-Eme1 that all contribute in their own way to cell survival in absence of Rqh1 BLM , we undertook genetic analyses by combining the mutations that impair the SIMs (i.e. eme1 SIM1 � +SIM2 � ) with those that abrogate the cell cycle and DNA damage-dependent phosphorylations of Eme1 (i.e. eme1 4SA ). As expected, we observed no constitutive Cdc2 CDK1 -dependent and DNA damage-induced mobility shift of the Eme1 4SA+SIM1 � +SIM2 � mutant protein (Fig 7A). The resulting eme1 4SA+SIM1 � +SIM2 � mutant strain displays a slightly reduced ability to form viable colonies in absence of exogenous stress (Fig 7B  and 7C). Remarkably, while eme1 SIM1 � +SIM2 � rqh1Δ (Fig 5E) and eme1 4SA rqh1Δ [16] double mutant strains are sick but viable, we were unable to generate viable eme1 4SA+SIM1 � +SIM2 � rqh1Δ cells (Fig 7D). We next crossed eme1 8AQ+SIM1 � +SIM2 � and rqh1Δ cells to determine to which extent direct phosphorylation of Eme1 by Rad3 ATR and the SIM related functions (including phosphorylation by Chk1) contribute to the essential functions of Mus81-Eme1 in absence of Rqh1 BLM . As shown in Fig 7E, we were also unable to recover eme1 8AQ+SIM1 � +SIM2 � rqh1Δ cells. This is reminiscent of the synthetic lethal interaction between eme1 Δ117 and rqh1Δ and suggests that the Eme1 1-117 N-terminal domain is involved in three regulatory processes that each make key contributions to the essential functions fulfilled by Mus81-Eme1 in absence of Rqh1 BLM (Fig 8).
loading control. C-32 P-labeled (red dot) HJs were incubated for the indicated times with Mus81-Eme1 SIM1 � +SIM2 � complex recovered from untreated or 40 μM CPT-treated eme1 SIM1 � +SIM2 � cells as described in Materials and Methods. Reaction products were analyzed by neutral PAGE. Comparable amounts of protein samples were used in each reaction after normalization of their relative concentration (see Materials and Methods and S3A Fig). A first graph below the gel represents the quantification of product formation, as a percentage of total radiolabeled DNA, in three independent experiments including the one shown above the graph (See S3A and S5 Figs). A second graph shows the average (± s.d.) fold stimulation of HJ resolution by Mus81-Eme1 SIM1 � +SIM2 � following CPT-treatment (n = 3 independent experiments, see S5 Fig). The histogram shows the ratio of the HJ-resolvase activity of Mus81-Eme1 from CPT-treated cells over that of Mus81-Eme1 from untreated cells. D-Western blot detection of TAP-Eme1 from untreated or 40 μM CPT-treated WT, eme1 SIM1 � +SIM2 � , eme1 8AQ and eme1 8AQ+SIM1 � +SIM2 � cells. Ponceau stained membranes are depicted as loading control. E-Western blot detection of TAP-Eme1 from untreated or 40 μM CPTtreated WT, eme1 SIM1 � and eme1 SIM2 � cells. Cdc2 is used as a loading control. F-Model summarizing our findings on the relative contributions of Rad3 ATR , Chk1 and the SIMs to phosphorylation of Eme1 and catalytic stimulation of Mus81-Eme1 in response to DNA damage. Note: TAP-= 2xProtA-TEVsite-2xFlag-.

Discussion
In this study we have identified three layers of control of Mus81-Eme1 that are critical in absence of Rqh1 BLM (Figs 6F and 8). A first layer relies on the phosphorylation of Eme1 by Rad3 ATR in response to DNA damage and is critical for the catalytic stimulation of Mus81-Eme1 (Figs 4D, 6F and 8). Another layer that drives catalytic stimulation of Mus81-Eme1 relies on the phosphorylation of Eme1 in response to DNA damage by Chk1 (Figs 4D, 6F and 8). It is noteworthy that both these layers require that Eme1 is first phosphorylated by Cdc2 CDK1 [16]. The third layer relies on newly described SIM1-and SIM2-dependent SUMO-binding properties of Eme1. Importantly, our results reveal that there is some crosstalk between these layers with, on one hand, Chk1-mediated stimulation of Mus81-Eme1 requiring that Eme1 is also phosphorylated by Rad3 ATR and, on the other, SIM2 contributing to phosphorylation of Eme1 by Chk1 (Figs 6F and 8).
In our initial study we had suggested that catalytic stimulation of Mus81-Eme1 in response to DNA damage relied on the phosphorylation of Eme1 by a classical DNA damage checkpoint signaling mechanism and that it was ultimately driven by phosphorylation of Eme1 by Chk1 [16]. With the demonstration that DNA damage-induced stimulation of Mus81-Eme1 strictly relies on direct phosphorylation of Eme1 by Rad3 ATR (Figs 4A, 4B, 4C, S3A and S3C), we are providing new insight into a control mechanism of Mus81-Eme1 that turns out to be substantially more elaborate than initially anticipated. Indeed, mutating all eight Rad3 ATR SQ/TQ phosphorylation sites in Eme1 prevents phosphorylation by Rad3 ATR in vitro and in vivo (Fig 2B and 2E) and fully abrogates DNA damage-induced stimulation of the resulting Mus81-Eme1 8AQ complex (Figs 4A, 4B, 4C, S3A and  S3D), even though Eme1 8AQ still undergoes Chk1-mediated phosphorylation (Fig 2E and 2F). These data point towards an essential role of direct phosphorylation by Rad3 ATR and question whether phosphorylation by Chk1 makes any contribution at all to the catalytic stimulation of Mus81-Eme1. However, we also found that in absence of Chk1, Mus81-Eme1 is very poorly stimulated by CPT treatment even though Eme1 is still directly phosphorylated by Rad3 ATR (Figs 4A, 4B, 4C, S3A and S3E). Based on these findings, we propose a model whereby fullfledged stimulation of Mus81-Eme1 in fact relies on phosphorylation of Eme1 by both Rad3 ATR and Chk1 (Figs 6F and 8). The complete lack of stimulation of the Mus81-Eme1 8AQ complex, despite Chk1-mediated phosphorylation of Eme1 8AQ suggests that stimulation by Chk1 only occurs when Eme1 is also phosphorylated by Rad3 ATR . However, we cannot formally exclude at this stage that the 8AQ mutations could have a knock-on effect that impairs phosphorylation by Chk1 (without any obvious impact on the mobility shift of Eme1).
It is noteworthy that all Rad3 ATR SQ/TQ phosphorylation sites are located before the coiled-coil domain of Eme1 and sit therefore within an intrinsically disordered part of Eme1. This is particularly relevant for the first two phosphorylation sites found in cluster 1 which are the only two within the Eme1 1-117 domain. They are the ones that also contribute the most to the phosphorylation of Eme1 by Rad3 ATR in vitro and cell fitness in absence of Rqh1 BLM (S2C Fig). It will be important to further investigate what structural impact phosphorylation of the N-terminus of Eme1 may have or whether it drives conformational changes by promoting the association with a coactivator and how this stimulates Mus81-Eme1. Hints that the poorly structured N-terminal domain of Eme1 may negatively impact the catalytic activity of Mus81-Eme1 may be seen in the increased in vitro activity that results from clipping off large N-terminal domains of human MUS81 and EME1 that are dispensable for the endonuclease function of the complex [18]. This would also be reminiscent of the auto-inhibition of human MUS81-EME1 by the N-terminal Helix-hairpin-Helix domain within MUS81 that is relieved upon association of the N-terminus of MUS81 with SLX4 [19]. Cryo-EM studies of XPF-ERCC1 also revealed how conformational changes imposed by DNA binding relieved auto-inhibition by the N-terminal helicase domain of XPF [20]. Members of the XPF-family of SSEs, which all carry their catalytic-relevant functions in the C-terminal part of their subunits, appear to have evolved in a way that provides their N-terminal domains with regulatory functions mediated through controlled conformational changes. Based on our findings, such conformational changes driven by phosphorylation seem like a plausible explanation in the case of Eme1.
Another pivotal finding of this study is the identification of two SIMs (SIM1 and SIM2) in the Eme1 1-117 N-terminal domain. Noteworthy, as observed for the eme1 4SA and eme1 8AQ phosphorylation mutants, mutating the SIMs of Eme1 severely impacts cell viability in absence of Rqh1 BLM (Fig 5E and 5F) but does not lead to any increased sensitivity to CPT (Fig 5D). This contrasts with the acute hypersensitivity of mus81Δ and eme1Δ null cells and indicates that the Eme1 SIM-dependent functions are intimately linked to the control of Mus81-Eme1 in relation to Rqh1 BLM functions.
We show that mutating the SIMs results in a marked decrease in DNA damage induced phosphorylation of Eme1 by Chk1 and tampers catalytic stimulation of Mus81-Eme1 (Figs  6B, 6C, S3A and S5). The drop in catalytic stimulation is less extensive than what might be expected from the pronounced reduction in DNA damage induced phosphorylation of Eme1. It is noteworthy that Eme1 SIM1 � +SIM2 � remains phosphorylated by Rad3 ATR and there remains the possibility that it undergoes residual Chk1-mediated phosphorylation that does not induce a detectable mobility shift but contributes to catalytic stimulation. Importantly, our genetic analyses indicate that the SIMs fulfil pivotal functions beyond catalytic control of Mus81-Eme1. Indeed, whereas Eme1 mutants unable to undergo catalytic stimulation (eme1 4SA or eme1 8AQ ) or to bind SUMO (eme1 SIM1 � +SIM2 � ) are viable in absence of Rqh1 BLM ( [16], Figs 3B and 5E), Eme1 mutants that combine both defects (eme1 4SA+SIM1 � +SIM2 � and eme1 8AQ+SIM1 � +SIM2 � ) are not (Fig 7D and 7E). Such non-catalytic functions of the SIMs may drive the efficient recruitment and stabilization of Mus81-Eme1 at sites where it is needed to process secondary DNA structures that accumulate in absence of Rqh1 BLM . In line with this, it was recently proposed that the human SLX4 nuclease scaffold that targets the XPF-ERCC1, MUS81-EME1 and SLX1 SSEs to specific genomic loci contains several SIMs that are involved in its own recruitment to telomeres, PML bodies and DNA damage [21][22][23]. An alternative and radically opposite explanation could be that the SIMs of Eme1 are involved in a process that negatively controls Mus81-Eme1 by sequestering the nuclease in subnuclear compartments, away from DNA secondary structures such as replication intermediates that could otherwise get opportunistically processed. In absence of Rqh1 BLM such structures would accumulate and their premature endonucleolytic processing would be deleterious to the cell. Such compartmentalisation has been observed in human cells with the nucleolar accumulation of MUS81-EME1 in S-phase and its relocalisation out of the nucleolus to sites of DNA damage in replicating cells following UV-irradiation [7].
Remarkably, while we find that SIM1 and SIM2 cooperate to bind SUMO (Fig 5A), we provide evidence that they also fulfil independent functions. First, mutating the strong SUMO binding SIM1 causes the accumulation of sick cells in absence of Rqh1 BLM but it barely impacts colony formation, whereas mutating the much weaker SUMO-binder SIM2 significantly reduces colony formation in addition to causing the accumulation of sick cells. Mutating both SIMs synergistically increased the proportion of sick cells (Figs 5E, 5F and S4A). Furthermore, we also found that SIM2 but not SIM1 promotes Chk1-dependent phosphorylation of Eme1 in response to DNA damage (Fig 6E). These results not only suggest that each SIM fulfils different functions, they also question whether some of those fulfilled by SIM2 might extend beyond SUMO-binding when putting into perspective the SIM2-specific phenotypes and its relatively poor affinity for SUMO compared to SIM1. In line with this, we found that Eme1 still undergoes DNA damage-triggered phosphorylation in mutant cells that lack the Pli1 and Nse2 SUMO E3 ligases or that do not produce SUMO (S7 Fig). At first glance, this might indicate that the processes that lead to the phosphorylation of Eme1 in response to DNA damage do not involve SUMO, including those that rely on SIM2. However, the possibility remains that in absence of SUMO, DNA damage-induced phosphorylation of Eme1 is entirely driven by Rad3 ATR , which does not rely on Eme1 SIMs (Fig 6D). Further work is therefore needed to formally establish whether SIM2 promotes phosphorylation of Eme1 by Chk1 through interaction with SUMO or not. While we cannot exclude that the mutations introduced in SIM2 induce structural changes that impact more than just SUMO-binding, should SUMO not be involved there remains the exciting possibility that SIM2 could drive the association of Eme1 with a partner that contains a SUMO-like domain (SLD) [24]. The only SLD-containing protein described so far in S. pombe is the Rad60 genome stability factor that contains two SLDs, each of which interact with different players of the SUMO pathway [25]. Interestingly, the presumed S. cerevisiae Rad60 ortholog Esc2 has been reported to interact with Mus81 via its SLDs and to stimulate the Mus81-Mms4 complex [26]. In addition, Esc2 was recently found to promote the degradation of phosphorylated Mms4 [15]. We can exclude similar scenarios involving Rad60 as none of the SIMs were found to modulate the levels of Eme1 or phosphorylated Eme1. However, based on such functional promiscuity between Mus81-Mms4 and the SLDcontaining Esc2 protein, it is tempting to see Rad60 as the ideal candidate for a SIM2-mediated partner of Eme1 that would promote Chk1-dependent phosphorylation of Eme1 in response to DNA damage. This would not be the first example of regulatory processes that involve similar players but different outcomes for Mus81-Eme1 and Mus81-Mms4.
Overall, our findings show that the poorly structured N-terminal domain of Eme1 harbors essential regulatory functions of Mus81-Eme1, the control of which appears to be remarkably more elaborate than initially described. With the demonstration that it relies on three regulatory layers that together contribute to the vital functions it fulfils in cells lacking Rqh1 BLM , we are setting the basis for new lines of investigation that should contribute to a better understanding of the contributions made by Mus81-Eme1 in the maintenance of genome stability.

Fission yeast strains, media, techniques and plasmids
Fission yeast strain genotypes are listed in S1 Table. Media and methods for studying S. pombe were as described elsewhere [27]. The were generated as follows. The Eme1 genomic locus from strain PH81 (h+, leu1-32 ura4-D18 TAP-eme1 mus81:13Myc-KanMX6), which produces N-terminally TAP-tagged Eme1, was subcloned into a TopoTA vector (Invitrogen). Point mutations were introduced on that TopoTA-EME1 vector by using a Multiprime site-directed mutagenesis kit (Stratagene). Mutations were confirmed by DNA sequencing. The mutated Eme1 genomic locus from the TopoTA-Eme1 vector was used as template for PCR using primers forward 5'-acccatctctcacctaacc-3' and reverse 5'-cagtattagcttacagcc-3'. The PCR fragment was then used to transform strain PH41, in which a URA4 cassette replaces the start codon of EME1 gene. 5 0 -FOA-resistant clones were selected and confirmed as TAP-eme1 mutant producing strains by genomic DNA sequencing.

Cell synchronization
For synchronization of cells by cdc25-22 block and release, cells containing the temperaturesensitive cdc25-22 allele were grown to exponential phase at permissive temperature (25˚C) and shifted at restrictive temperature (36˚C) for 3.5 h to arrest the cell cycle in G2. Upon release to permissive temperature (25˚C), the cells synchronously enter the cell cycle. Cells were collected and processed every 20 min. Progression into S phase was monitored microscopically by counting cells that contained septa using calcofluor (Sigma) staining, the appearance of which correlates with S phase.
For DNA damage studies, Bleomycin (Merck) was added to cells arrested at the G2/M transition and further incubated at 36˚C for 1,5h.

Colony formation assay
Fresh cultures were re-seeded on YES plates by micromanipulation and allowed to grow for 3 days at 32˚C. At least 3 independent experiments were performed and averaged. Statistical significance (1 way ANOVA and Tuckey test) is displayed on each graph.
GFP-TRAP-bound GFP-Rad3 ATR was resuspended in kinase buffer supplemented 100 μM of cold ATP before addition of 10 μCi γ 32 P-ATP and substrates. After 30 min at 30˚C, reactions were stopped by the addition of 15 μL of 4X SDS sample buffer. Samples were denatured and resolved by SDS-PAGE electrophoresis. Following Coomassie staining, gel was dried and expose with phosphorimager.

In vitro nuclease assay
TAP-Eme1 (2xProtA-TEVsite-2xFlag-Eme1) was affinity purified and used in nuclease assays on X12 mobile HJ as previously described [28]. Briefly, cell pellets were resuspended in 1 volume to weight lysis buffer (50mM Tris-HCl, pH = 8.0, 150 mM NaCl, 10% glycerol, 0,1% NP-40, 50mM NaF, 50mM β-glycerophosphate (Sigma), 2mM PMSF, protease inhibitor cocktail complete EDTA free (Roche)). Cells were subjected to mechanical lysis using a Ball Mill (MM400 Retsch). For this, usually, 4 to 5 ml of the cell suspension were poured into grinding chambers precooled in liquid nitrogen. The frozen cell pellet was disrupted by 2 agitation runs at 30hz. The resulting powder was resuspended in another 1 volume to weight of lysis buffer and centrifuged. Clear supernatant was loaded onto IgG sepharose beads (Cytiva) for 2h at 4˚C (20 μl packed beads used for 4 ml of lysate). After extensive washes, proteins were eluted in presence of 60 μl of AcTEV protease for 1h at RT. In order to determine the relative amount of Mus81-Eme1 between different samples, 3 μl of each TEV eluate was treated with phosphatase before SDS-PAGE and Western blot analysis in order to collapse the Eme1 signal into a single band of dephosphorylated Eme1. The relative intensity of the dephosphorylated Eme1 band was quantified for each sample using the ImageLab software. Dilution folds were calculated and used to bring the concentration of each sample down to that of the least concentrated sample. 3 μl of each normalized sample was used in nuclease assays with 32 P-labelled DNA substrates as previously described [28].  8AQ , rqh1Δ and eme1 8AQ rqh1Δ cells were heated-fixed and observed by fluorescent-microscopy using DAPI staining. Cells were classified based on their morphologies. Small G2 cells (Type 1), elongated bi-nucleated and/or septated cells (Type 2) and sick cells (Type 3). B-Five-fold dilutions of cells with the indicated genotype were plated on medium supplemented or not with the indicated concentrations of CPT, HU and MMS followed by incubation at 30˚C. C-Tetrad analysis of an eme1 cluster1 � , eme1 cluster2 � and eme1 cluster3 � x rqh1Δ mating, germinated at 30˚C. To assess the impact of introducing the cluster mutations (C � ) in an rqh1Δ background, the ratio of the colony diameter of the double mutants C � rqh1Δ mutants over that of the rqh1Δ mutant were calculated for each tetrad and plotted on the graph below the tetrads. (PDF) Graphs below the autoradiographs represents the quantification of product formation, as a percentage of total radiolabeled DNA, for each experiment. Note: Schematic depicting the migration profile of uncleaved and cleaved radiolabelled HJs. Yellow dotted boxes represent areas of the gel used for quantification purposes and calculation of the percentage of cleavage product. The "Bckg" box corresponds to background signal that was subtracted from the "Uncut" and "Cut" signals. C-Same as A-but with Mus81-Eme1 from untreated or 40 μM CPT-treated TAP-eme1 rad3Δ cells. D-Same as A-but with Mus81-Eme1 from untreated or 40 μM CPT-treated TAP-eme1 eme1 8AQ cells. E-Same as A-but with Mus81-Eme1 from untreated or 40 μM CPT-treated TAP-eme1 chk1Δ cells. Note: TAP-= 2xProtA-TEVsite-2xFlag-. (PDF)