Rrp1 translocase and ubiquitin ligase activities restrict the genome destabilising effects of Rad51 in fission yeast

Abstract Rad51 is the key protein in homologous recombination that plays important roles during DNA replication and repair. Auxiliary factors regulate Rad51 activity to facilitate productive recombination, and prevent inappropriate, untimely or excessive events, which could lead to genome instability. Previous genetic analyses identified a function for Rrp1 (a member of the Rad5/16-like group of SWI2/SNF2 translocases) in modulating Rad51 function, shared with the Rad51 mediator Swi5-Sfr1 and the Srs2 anti-recombinase. Here, we show that Rrp1 overproduction alleviates the toxicity associated with excessive Rad51 levels in a manner dependent on Rrp1 ATPase domain. Purified Rrp1 binds to DNA and has a DNA-dependent ATPase activity. Importantly, Rrp1 directly interacts with Rad51 and removes it from double-stranded DNA, confirming that Rrp1 is a translocase capable of modulating Rad51 function. Rrp1 affects Rad51 binding at centromeres. Additionally, we demonstrate in vivo and in vitro that Rrp1 possesses E3 ubiquitin ligase activity with Rad51 as a substrate, suggesting that Rrp1 regulates Rad51 in a multi-tiered fashion.

the completion of HR. Rad54 is activated in G2, and does not remove Rad51 from stalled replication forks (Spies et al., 2016). Another complex, MMS22L-TONSL, has been shown in human cells to limit Rad51 binding to dsDNA and stimulate HR-mediated restart of arrested replication forks (Piwko et al., 2016). Importantly, other SWI2/SNF2-like translocases, Rdh54 and Uls1, cooperate with Rad54 in Saccharomyces cerevisiae not only to antagonise Rad51 binding to dsDNA during HR, but also to counteract its toxic accumulation on undamaged chromatin (Chi et al., 2006(Chi et al., , 2011Shah et al., 2010). RAD54L and RAD54B in humans also prevent the genome-destabilising consequences of excessive Rad51 binding to dsDNA (Mason et al., 2015). It should also be noted that the binding of Rad51 to dsDNA renders the dsDNA inaccessible to the Rad51-ssDNA filament and thus acts as a barrier to HR itself (Sung and Robberson, 1995).
Interestingly, the Rad51 paralog RAD51C has been shown to prevent proteasomal degradation of Rad51 in human cells, especially after DNA damage (Bennett and Knight, 2005), suggesting that Rad51 can also be regulated by ubiquitylation. Recently, Rad51 was found to be poly-ubiquitylated by the E3 ubiquitin ligase RFWD3 in a process stimulated by DNA damage (Inano et al., 2017). Rad51 ubiquitylation decreases Rad51 binding to ssDNA and leads to its proteasomal degradation, while also stimulating chromatin loading of Rad54.
It has therefore been proposed that Rad51 ubiquitylation promotes its removal from sites of DNA damage and is necessary for completion of HR DNA repair (Inano et al., 2017).

Rrp1 counteracts rad51 + overexpression-induced toxicity
Previous studies have shown that rad51 + overexpression in S. pombe results in a severe growth defect (Kim et al., 2001). It has been demonstrated that Rad51 overproduction leads to its excessive accumulation on chromatin and has a negative effect on cell growth and chromosome stability that is aggravated in mutants devoid of SWI2/SNF2-related translocases: RAD54L and RAD54B in humans (Mason et al., 2015) and Rdh54, Rad54, and Uls1 in S. cerevisiae (Shah et al., 2010). Two ULS1 orthologues have been identified in S.
pombe (Dziadkowiec et al., 2009), so we set out to examine their functional interaction with Rad51. Using an nmtP3-GFP-rad51 strain, where the rad51 + gene is expressed from a strong promoter that is induced in media lacking thiamine, we confirmed that induction of rad51 + expression resulted in a severe growth defect ( Figure 1A,B) and loss of viability ( Figure 1C). Importantly, this was further exacerbated by deletion of rrp1 + , but not rrp2 + .
Growth inhibition of the nmtP3-GFP-rad51 rrp1Δ strain was visible even on YES medium ( Figure 1A,B), where gene expression from the nmtP3 promoter is very limited, indicating that even a mild increase in Rad51 protein levels is toxic when the Rrp1 translocase is absent.
Due to the severe growth defect of the nmtP3-GFP-rad51 rrp1Δ strain, as well as the rapid generation and subsequent proliferation of faster-growing suppressors, detailed examination of its phenotype was not possible. However, we hypothesised that if deletion of rrp1 + is toxic in the nmtP3-GFP-rad51 strain, rrp1 + overexpression should be beneficial.
Indeed, we found that overexpression of rrp1 + , but not rrp2 + , rescued the growth defect and viability loss induced by Rad51 overproduction (Figure 1D,E). We confirmed that addition of the GFP tag to Rad51 was not responsible for toxicity of Rad51 overproduction or for the role of Rrp1 in alleviating this phenotype ( Figure S1A). This allowed us to use a synthetic dosage lethality approach with the nmtP3-GFP-rad51 strain to assess the effect of Rrp1 on Rad51 activity.
The ATPase activity of Rrp1 is required to counter the genotoxicity associated with

Rad51 overproduction
Rrp1 shares a complex domain structure with Uls1 (Dziadkowiec et al., 2009), so Walker B (rrp1-DAEA) and RING (rrp1-CS) Rrp1 mutant variants were employed in order to examine the importance of the respective DNA translocase and ubiquitin ligase Rrp1 activities for counteracting rad51 + overexpression-induced toxicity ( Figure 2A). While the presence of a functional Rrp1 ATPase domain was required for restoration of normal growth to the rad51 + over-expressing strain, the RING domain appears to be dispensable ( Figure   2B).
It has previously been shown that rad51 + overexpression results in mitotic aberrations revealed by the accumulation of cells with nuclei exhibiting the cut (cell untimely torn) phenotype (Kim et al., 2001). We demonstrated that overproduction of Rrp1, but not the Rrp1-DAEA mutant, was able to suppress the appearance of nuclear defects that occur in the rad51 + overexpressing nmtP3-GFP-rad51 strain, such as cut and non-disjunction ( Figure   2C). Interestingly, the rescue induced by Rrp1-CS overproduction was less pronounced, with more cells undergoing aberrant mitosis ( Figure 2C). This suggests that Rrp1 ubiquitin ligase activity might also have a role in the protein's functional interaction with Rad51.
We examined GFP-rad51 + over-expressing cells using fluorescence microscopy and observed that most of them contained extensive Rad51 fibres in their nuclei, often connecting several bright foci ( Figure 2D). These structures virtually disappeared when Rrp1 was simultaneously overproduced, and GFP-Rad51 staining changed to diffuse with 1-3 foci ( Figure 2D,E). Interestingly, about 90% of these remaining GFP-Rad51 foci co-localised with foci for Rrp1-mCherry, and GFP-Rad51 fibres were visible only in cells lacking Rrp1 signal ( Figure 2F). We have previously shown that Rrp1 is enriched at centromeres and >40% of

The role of Rrp1 in regulating Rad51 function is independent from Rrp2
Overproduction of Rrp2 in a rad51 + overexpressing strain was unable to suppress viability loss ( Figure 1D,E) and the chromosome segregation defect (Figure S1B), and did not prevent the accumulation of Rad51 fibres on chromatin ( Figure S1C). Moreover, rescue of the rad51 + overexpression-induced growth defect by overproduction of Rrp1 was not affected by the presence of Rrp2 ( Figure S1D) or the recombination auxiliary factor complex Swi5-Sfr1 ( Figure S1E). This suggests that the observed role of Rrp1 in regulating Rad51 may be distinct from the previously described mutually dependent role of Rrp1 and Rrp2 in the Swi5-Sfr1 sub-pathway of HR (Dziadkowiec et al., 2009(Dziadkowiec et al., , 2013.

Purified Rrp1 binds to DNA and has a DNA-dependent ATPase activity
In order to gain mechanistic insight into the function of Rrp1, recombinant Rrp1-FLAG was purified to near-homogeneity following overexpression in Escherichia coli ( Figure S2A).
Since Rrp1 is predicted to possess ATPase activity (Dziadkowiec et al. 2009), we first examined if the purified protein could indeed hydrolyse ATP. In the absence of DNA, no ATPase activity was detected, while robust ATP hydrolysis was observed in the presence of either ssDNA or dsDNA ( Figure S2B). Based on these data, Rrp1 was estimated to have a turnover number (k cat ) of 424 min -1 (± 25.1) in the presence of ssDNA and 529 min -1 (± 37.6) in the presence of dsDNA. These results indicate that Rrp1 possesses a robust ATPase activity that is strictly dependent on DNA, and that dsDNA stimulates ATP hydrolysis by Rrp1 more efficiently than ssDNA, The observed dependency on DNA for ATP hydrolysis suggested that Rrp1 is capable of binding DNA. This was tested by electrophoretic mobility-shift assays (EMSA). At concentrations as low as 0.05 µM (Rrp1:nucleotide ratio of 1:100), all ssDNA was shifted in an ATP-independent manner by Rrp1, and this shift was enhanced at higher concentrations of protein ( Figure S2C). A lesser shift was observed with dsDNA, with some unbound DNA remaining even at 0.30 µM Rrp1 (Rrp1:basepair ratio of 1:8.33), although ATP was also dispensable for this binding ( Figure S2D). Some signal was observed in the wells, particularly for dsDNA, suggesting that Rrp1 may form aggregates consisting of protein-DNA networks that are too large to enter the gel. These results indicate that Rrp1 binds both ssDNA and dsDNA, with an apparently higher affinity for ssDNA, in an ATP-independent manner.

Rrp1 physically interacts with Rad51
The genetic interactions observed between Rad51 and Rrp1 raised the possibility that the two proteins may physically interact. To investigate this possibility, we first employed the yeast two-hybrid system (Y2H). For Rad51, two constructs were used: a short N-terminal fragment (Rad51-N), and a long C-terminal fragment containing the Walker A and B motifs (Rad51-C) ( Figure 3A). We observed a robust growth of transformants containing genes for Rrp1 and Rad51-C on high stringency SD DO-4 plates, indicating that the site of putative Rrp1 binding lies within the Rad51 region containing Walker A and B motifs ( Figure 3B). In order to map the corresponding region within Rrp1 that is responsible for Rad51 binding, we created a series of four truncated forms of Rrp1 ( Figure 3C) and repeated the Y2H assay with Rad51-C. These experiments revealed that the fragment of Rrp1 containing its Cterminal helicase domain was sufficient for the interaction with Rad51 ( Figure 3D). In agreement with our genetic data, no interaction was observed between Rad51 and Rrp2 by Y2H ( Figure 3B).
To validate these Y2H results and rule-out the possibility that the observed Rrp1-Rad51 interaction involved an intermediary molecule, purified Rad51 and purified Rrp1-FLAG were mixed together and subjected to immunoprecipitation with anti-FLAG M2 agarose.
Importantly, Rad51 was seen to co-immunoprecipitate with Rrp1-FLAG ( Figure 3E), thus confirming the existence of a direct interaction between these two proteins. Furthermore, we were able to demonstrate that Rad51-Rrp1 complex is formed in vivo in S. pombe cells by immunoprecipitating Rad51 with Rrp1-FLAG from native protein extracts ( Figure 3F). Some self-activation was observed on low stringency SD DO-3 plates when Rrp1 was fused to the transcription activation domain (pGADT7 plasmid; Figure 3B). Given the results presented here indicating that purified Rrp1 binds both ssDNA and dsDNA ( Figure S2C,D), and previous microscopy and chromatin immunoprecipitation data implicating Rrp1 in DNA binding (Barg-Wojas et al., 2020), this self-activation is likely due to the ability of Rrp1 to associate with DNA.

Rrp1 dissociates Rad51 from dsDNA in vitro
SWI2/SNF2-related translocases have been proposed to remove Rad51 from dsDNA in heteroduplex DNA and dead-end non-productive complexes, both in yeast and human cells (Mason et al., 2015;Shah et al., 2010). The existence of a physical interaction between the Rrp1 translocase and Rad51, combined with the ability of Rrp1 to suppress the association of Rad51 with bulk chromatin ( Figure 2E), prompted us to examine whether Rrp1 can directly counteract erroneous Rad51 binding to dsDNA. To test this, we exploited the fact that the binding of purified Rrp1 to dsDNA results in a distinctive EMSA pattern where most of the dsDNA signal is retained in the well ( Figure S2D). This pattern is easily distinguishable from the binding of Rad51 to DNA, which produces a smear at low concentrations of protein (0.5 µM) and a discrete band at higher concentrations (1.5 or 3 µM; Figure 4A). dsDNA was first coated with different concentrations of Rad51 and then challenged with substoichiometric amounts of Rrp1. Protein-DNA complexes were then resolved by gel agarose electrophoresis. Compared to the condition where Rrp1 was omitted, the bands for Rad51-bound DNA became fainter when 0.1 µM of Rrp1 was included. Moreover, the inclusion of 0.3 µM of Rrp1 led to a drastic decline in the signal of Rad51-dsDNA bands, even when the dsDNA was precoated with five-fold more Rad51 molecules, and signal in the well became apparent. These results suggest that sub stoichiometric amounts of Rrp1 effectively outcompete Rad51 for binding to dsDNA. Since Rrp1 can bind in vitro to both dsDNA ( Figure S2D) and Rad51 ( Figure 3E), an alternative explanation for these results is that, rather than displace Rad51 from dsDNA, Rrp1 bound to Rad51-dsDNA complexes, leading to the formation of higher-order complexes that were unable to enter the gel.
To more directly test whether Rrp1 could disrupt Rad51-dsDNA complexes, we employed a fluorescence anisotropy assay. Rad51-dsDNA complexes were assembled on fluorescently-labelled oligonucleotides, then Rrp1 was added and fluorescence anisotropy was monitored in real-time. When protein stock buffer was added instead of Rrp1, Rad51-dsDNA complexes remained highly stable for the duration of the experiment, with anisotropy remaining at ~98% even after ~30 min. By contrast, the addition of Rrp1 led to a drastic decline in anisotropy, with a value of ~60% observed after ~15 min ( Figure 4B). This decline followed first-order decay kinetics (R 2 = 0.995), consistent with the notion that it represents the dissociation of Rad51-dsDNA complexes, with a dissociation rate constant (k off ) equal to 0.00139 s -1 .
We also examined the effect of Rrp1 on Rad51-ssDNA complexes. Although a slight decrease in the intensity of Rad51-ssDNA bands was observed by EMSA ( Figure S3A), this was negligible when compared with the effect of Rrp1 on Rad51-dsDNA complexes ( Figure   4A). Consistently, relative anisotropy increased upon addition of Rrp1 to Rad51-ssDNA complexes in a concentration-dependent manner ( Figure S3B), suggesting that rather than dissembling Rad51-dsDNA complexes, Rrp1 was binding to them.
Collectively, these results indicate that Rrp1 is capable of rapidly disassembling Rad51-dsDNA complexes. Rrp1 may also bind to Rad51-ssDNA complexes to modulate their activity but further studies are needed to conclusively test this hypothesis.
Moreover, when these precipitates were probed with an anti-Rad51 antibody, substantial high molecular weight species were observed ( Figure S4B). We therefore hypothesized that Rrp1 may have an E3 ligase activity with Rad51 as a substrate. In order to directly test this possibility, in vitro Rad51 ubiquitylation assays were performed using ubiquitin, Uba1 (E1) and Ubc4 (E2) enzymes purified from E. coli, and purified Rrp1-FLAG as the sole E3 ubiquitin ligase enzyme. Reaction products were separated by SDS-PAGE and subjected to western blot analysis. Multiple high molecular weight protein bands were observed only when all assay components were included in the reaction ( Figure 5A). Strikingly, a similar banding pattern was detected with both anti-Rad51 and anti-Ubiquitin antibodies, indicating that these bands represent ubiquitylated forms of Rad51. Consistent with this notion, the unmodified Rad51 band decreased in intensity only in the assay with all reaction components ( Figure   5A). To further validate these findings, we repeated this ubiquitylation assay with either wildtype Rrp1 protein or the Rrp1-CS variant, where the Rrp1 RING domain was inactivated, as the sole E3 ligase. When Rrp1-CS was employed, the characteristic protein ladder was not formed ( Figure 5B), indicating that Rrp1 E3 ubiquitin ligase activity was responsible for Rad51 ubiquitylation. Interestingly, when the membrane was probed with an anti-FLAG antibody to detect Rrp1 protein, several high molecular weight protein bands greater in size than Rrp1 were observed in reactions containing wild-type Rrp1 but not Rrp1-CS ( Figure 5B, bottom panel), indicating that Rrp1 is capable of auto-ubiquitylation. Consistent with this notion, the intensity of unmodified Rrp1 band was decreased in these reactions. Interestingly, when the Rad51 ubiquitylation assay was performed in the presence of ssDNA or dsDNA, the intensity of bands corresponding to ubiquitylated Rad51 decreased ( Figure 5C), with the mono-ubiquitylated form of Rad51 showing a reduction of >50%. This suggests that Rrp1 might preferably ubiquitylate free Rad51 rather than Rad51 bound to either ssDNA or dsDNA.
We thus identify Rrp1 as a translocase that can remove Rad51 from dsDNA, and a ubiquitin ligase that has Rad51 as its substrate, and propose that these Rrp1 activities are important for its role in regulating Rad51.

Discussion
Previous studies have shown that the presence of the RAD54 family of SWI2/SNF2 DNA translocases, RAD54L and RAD54B in human (Mason et al., 2015), and Rdh54, Rad54 in S. cerevisiae (Shah et al., 2010), is necessary to counteract the genotoxic effects of Rad51 overproduction. Another budding yeast protein, Uls1, belonging to the RING-domaincontaining Rad5/16-like group of SWI2/SNF2 DNA translocases has also been shown to participate in modulating Rad51 activity. Here we examined the interaction of S. pombe Rrp1, an Uls1 orthologue, with Rad51 and found that toxicity of rad51 + overexpression was increased in an rrp1Δ strain ( Figure 1A-C). This suggested that Rrp1 has a prominent function in counteracting the toxicity of Rad51 overproduction, in contrast to the requirement for Uls1 in S. cerevisiae, which only becomes apparent in the absence of Rdh54 (Shah et al., 2010).
The additional viability loss conferred by deletion of rrp1 + in Rad51 overproducing cells precluded detailed examination of this mutant. However, we reasoned that concomitant rrp1 + overexpression should have the opposite effect and allow investigation of the Rrp1-Rad51 interaction. Indeed, rrp1 + overexpression rescued the growth defect and viability loss induced by Rad51 overproduction in a manner that was dependent on the presence of a functional Rrp1 ATPase domain (Figures 1D,E and 2B,C). Apart from being a DNA translocase, Uls1 was proposed to have a SUMO targeted ubiquitin ligase activity (Uzunova et al., 2007) but this role in Rad51 regulation was not examined. Rrp1 also has a Zinc finger RING-type domain characteristic of E3 ubiquitin ligases, but we found it was not crucial for rescue of the Rad51-overproduction induced growth defect.
It has been shown that Rad51 overproduction leads to chromosome segregation defects, such as chromosome bridges and micronuclei formation in mammals (Mason et al., 2015) and accumulation of cells with cut nuclei in S. pombe (Kim et al., 2001). We demonstrated that Rrp1 overproduction in a rad51 + overexpressing strain resulted in a decrease in the number of cells with cut and non-disjunction nuclear defects ( Figure 2C). This was accompanied by the disappearance of extensive Rad51 fibres in the nucleus ( Figure 2D,E), consistent with previous reports indicating that excessive Rad51 binding to non-damaged dsDNA may be the underlying cause of genome instability and viability loss caused by Rad51 overproduction (Mason et al., 2015;Shah et al., 2010). Rrp1-mediated rescue of these phenotypes, similarly to the viability restoration described above, was dependent on the presence of a functional Rrp1 ATPase domain, suggesting that Rrp1 modulates Rad51 through its translocase activity. Interestingly however, we found that the Rrp1-CS mutant, with an inactivated RING domain, was slightly less proficient in counteracting inappropriate Rad51 accumulation on chromatin and the appearance of aberrant DNA segregation events. This suggested that Rrp1 ubiquitin ligase activity may play a role in Rad51 regulation, even though it was apparently not important for suppression of the Rad51 overproduction-induced growth defect.
We purified Rrp1 in order to characterise its biochemical activities and obtain mechanistic insight into its function. We found that purified Rrp1 binds to both ssDNA and dsDNA independently of ATP, and has a DNA-dependent ATPase activity ( Figure S2).
Importantly, we demonstrated that Rrp1 and Rad51 directly interact ( Figure 3) and when overproduced, they co-localise with each other in discrete perinuclear foci ( Figure 2F). This supports the conclusion of our phenotypic analyses presented above, that Rad51 may be the direct target of Rrp1 activity. Indeed, we showed that Rrp1 can efficiently dissociate Rad51 from dsDNA in vitro (Figure 4). Thus, our data establish Rrp1 as a translocase that can counteract excessive Rad51 binding to chromatin. Such regulation by RAD54-like DNA translocases has been shown to prevent DNA segregation defects, genome instability and viability loss in mammalian cells (Mason et al., 2015), and we infer it is also the basis of Rrp1-mediated rescue of Rad51 overproduction-induced genotoxicity that we observe in S. pombe. Such activity has not been directly demonstrated before for Uls1 or any other protein belonging to the Rad5/16-like group of SWI2/SNF2 translocases.
The physical interaction between Rad51 and Rrp1 involves sites within the Cterminus of each protein ( Figure 3B,D). Recently, the Rad51 interaction region of Sfr1, a component of the Swi5-Sfr1 HR auxiliary factor complex, was shown to be an intrinsically disordered domain (Argunhan et al., 2020). Residues 695-897 of Rrp1 involved in Rad51 binding are also predicted to include a disordered region. Thus, the intrinsically disordered structure may provide a general site for the interaction of auxiliary proteins with Rad51, but precise mapping of the interaction sites of Rrp1 will be required before direct parallels with other proteins can be drawn.
In addition to the translocase function of Rrp1, we found that the Rrp1 RING domain may also be involved in regulating Rad51 binding to DNA. We thus performed in vitro Rad51 ubiquitylation assays with purified Rrp1 or the Rrp1-CS RING mutant as the sole ubiquitin ligase E3 enzyme and showed that Rrp1 is able to poly-ubiquitylate Rad51 in a manner dependent on the presence of a functional Rrp1 RING domain ( Figure 5). Moreover, ubiquitylated proteins accumulated in S. pombe cells overproducing Rrp1, and we identified Rad51 as one of these proteins ( Figure S4). Taken together, this demonstrates that Rrp1 has an E3 ubiquitin ligase activity and Rad51 is one of its substrates.
This raises the interesting question of whether the Rrp1 translocase and ubiquitin ligase activities cooperate in counteracting excessive Rad51 binding to chromatin.
A functional RING domain is mostly dispensable for Rrp1 mediated rescue of the Rad51 overproduction-induced growth defect, so we propose that Rad51 ubiquitylation by Rrp1 is not an absolute prerequisite for its removal from DNA by Rrp1's translocase activity.
However, the number of cells with Rad51 fibres and segregation defects increased in rad51 +overexpressing strain simultaneously overproducing Rrp1-CS compared to wild-type Rrp1.
This implies that when its ubiquitin ligase is inactivated, Rrp1 ability to prevent Rad51 association with DNA is compromised, albeit not to the extent that would affect cell viability.
Interestingly, we found that addition of DNA decreased the efficiency of Rad51 ubiquitylation by Rrp1 in vitro, suggesting that Rrp1 may ubiquitylate free Rad51 more efficiently than Rad51 bound to either ssDNA or dsDNA. Rad51 ubiquitylation has previously been shown to compromise its ability to bind DNA (Chu et al., 2015;Inano et al., 2017). We thus hypothesise that, after the translocase activity of Rrp1 mediates Rad51 removal from chromatin, the ubiquitin ligase activity of Rrp1 could prevent the re-association of Rad51 with dsDNA. A similar model has been proposed previously for FBH1 (Chu et al., 2015), a UvrD family helicase able to disrupt Rad51 nucleofilaments on ssDNA and ubiquitylate Rad51 both in human (Chu et al., 2015;Simandlova et al., 2013) and S. pombe cells (Tsutsui et al., 2014). This suggests that analogous strategies may be used by enzymes regulating Rad51 binding to ssDNA as well as to dsDNA.
Rad51 is overexpressed in several types of human cancers (Maacke et al., 2000a(Maacke et al., , 2000b) and multiple cancer cell lines (Hansen et al., 2003;Raderschall et al., 2002), and contributes to their increased survival after DSB induction. Increased levels of Rad51 may compensate for deficiencies in other DNA repair pathways in cancer cells and is often associated with poor patient survival prognosis (Brown and Holt, 2009;Tennstedt et al., 2013). Since the role of ubiquitylation has not been addressed in previous studies on Rad51 dysregulation (Mason et al., 2015;Shah et al., 2010), our work may have implications for patient treatment.
Although eukaryotes possess multiple SNF2 translocases, the division of labour between them is poorly understand, and even less is known about how their activities are coordinated (Poole and Cortez, 2017;Rickman and Smogorzewska, 2019). In this context, it is worth noting that Rad54 has been shown not to be required for the replication fork protection function of Rad51 (Schlacher et al., 2011;Spies et al., 2016). Budding yeast Rad5 and human SHPRH and HLTF have been shown thus far to ubiquitylate PCNA (Unk et al., 2010). Recently, SHPRH has been identified as a nucleosome E3-ubiquitin-ligase (Brühl et genetic crossing of relevant single mutants followed either by random spore analysis or by tetrad dissection. pREP81-FLAG vector and plasmids carrying wild type and mutated forms of rrp1 + and rrp2 + as well as domains of rrp1+ gene used in the yeast two hybrid system were constructed using the Gibson Assembly® Cloning Kit (NEB). All primers used to amplify gene sequences by PCR are listed in Table S3. Amplified fragments were cloned into NdeI and BamHI digested pREP81 vector. After Gibson cloning, inserts were cut by NdeI and SmaI digestion and cloned into pREP42-HA, pREP42-EGFP, pREP41-mCherry or pGADT7 and pGBKT7 plasmids. rrp1 + and rrp1-CS mutant version were introduced into the pGEX vector(GE Healthcare) by In-Fusion® cloning (Takara Bio). All constructs were checked by sequencing.

Spot assays
Cells were grown to mid-log phase, then serially diluted by 10-fold and 2 µL aliquots were spotted onto relevant plates (YES or EMM) that were incubated for 3-5 days in 28°C and photographed. All assays were repeated at least twice.

Survival assay
Cells were grown for 48 hours in YES or in minimal medium with (repressed conditions) or without thiamine (overexpression) at 28°C. 500 µL aliquots were collected, serially diluted and plated onto YES plates to determine the number of viable cells. Plates were incubated for 3-5 days at 28°C. The viable cells were counted and percentage of survival for gene overexpression conditions was calculated against the repressed control.

Yeast two-hybrid assay
Gal4-based Matchmaker Two-Hybrid System 3 (Clontech) was used according to manufacturer's instructions. The indicated proteins were fused to the GAL4 activation domain (AD) in pGADT7 vector and the GAL4 DNA-binding domain (DBD) in pGBKT7, and expressed in the S. cerevisiae tester strain AH109. Transformants were selected on synthetic dextrose drop-out medium without Leu and Trp (SD DO-2), and then plated on low stringency medium without Leu, Trp and His (SD DO-3) and high stringency medium without Leu, Trp, His and Ade (SD DO-4), and incubated for 3-5 days at 28°C.

Fluorescence microscopy
To determine the formation of Rrp1 foci and their co-localisation with Rad51 foci, appropriate transformants were grown for 24 h in EMM medium without thiamine. 1 mL of culture was harvested, washed with water and subjected to fluorescent microscopy analysis. For colocalisation experiments, data were collected under 63x magnification with the confocal microscope Leica 453 TCS SP8 (Leica Microsystems) equipped with Leica HyD SP detector, and analysed with LAS X 3.3.0. For examination of mitotic defects induced by rad51 + overexpression, samples taken from respective transformant cultures grown for 48 hours in EMM medium without thiamine were washed and fixed in 70% ethanol. After rehydration, cells were stained with 1 mg/mL 4',6-diamidino-2-phenylindole (DAPI) and 1 mg/mL pphenylenediamine in 50% glycerol and examined by fluorescence microscopy with Axio Imager A.2 (Carl Zeiss).

In vivo co-immunoprecipitation
In vivo pull-down experiments were performed using strains with native levels of Rad51 and overproduction of FLAG-tagged Rrp1, which was integrated into the ars1 region under control of the nmt81 promotor. Cells were grown to mid-log phase using EMM minimal medium for 24 hours. 100 mL of cells were harvested and broken with glass beads in H buffer (50 mM HEPES-KOH pH 7, 50 mM KOAc, 5 mM MgOAc, 0.1% NP-40, 10% glycerol, 1 mM DTT and 1x cOmplete TM EDTA-free protease inhibitor cocktail (Roche)). Extracts were cleared by centrifugation and immunoprecipitated with ANTI-FLAG® M2 Affinity Gel (Sigma). Beads were washed and eluted using 100 µg/mL 3xFLAG peptide (Sigma). For detection, anti-FLAG (1:5000, Sigma) antibodies and anti-Rad51 (1:5000) (Haruta et al., 2006) antiserum were used.

Purification of Rrp1-FLAG
Recombinant Rrp1 and Rrp1-CS were expressed in the Rosetta E. coli strain (Novagen) from the pGEX-6P plasmid (GE Healthcare). Proteins were C-terminally fused to the GST tag and N-terminally to the 3xFLAG tag; only the former was removed during the purification process.
Expression was induced with 1 mM IPTG (Sigma) at 18°C overnight. Cells were collected by centrifugation, resuspended in R buffer (20 mM Tris, pH 7.5, 10% glycerol, 1mM EDTA) containing 500 mM NaCl, and disrupted by sonication. The cell lysate was then clarified by ultracentrifugation (70,000 g, 1 h, 2°C). The supernatant was mixed with 4B GSH sepharose (Sigma) for 3 h at 4°C. Resin-bound proteins were eluted in R buffer containing 300 mM NaCl and 40 mM glutathione (Sigma). The sample was supplemented with 0.2 µg/mL of HRV-3C protease (Sigma) to remove the GST tag and dialyzed against R buffer containing 100 mM NaCl (overnight, 4°C). The dialyzed sample was loaded onto a 1 mL HiTrap Heparin (Sigma) column. Rrp1 eluted at around 650 mM NaCl with a linear gradient of 0.1-1.0 M NaCl in R buffer. Eluted fractions were diluted 6.5-fold with R buffer and loaded onto a 1 mL ResourceQ column (GE Healthcare). Rrp1 eluted at around 300 mM NaCl with a linear gradient of 0.1-1.0 M NaCl in R buffer. Eluted fractions were diluted 3-fold with R buffer and loaded onto a HiTrap SP column (GE Healthcare). Rrp1 eluted at around 600 mM NaCl with a linear gradient of 0.1-1.0 M NaCl in R buffer. Eluted fractions were collected and dialysed against R buffer containing 200 mM NaCl. Concentration was determined using NanoDrop (ThermoFisher) with a molar extinction coefficient of 100365. For the Rrp1-CS mutant, Resource Q and HiTrap SP columns were omitted. Rad51 was purified exactly as previously described (Kurokawa et al., 2008). Uba1 (E1) and Ubc4 (E2) were purified exactly as previously described (Tsutsui et al., 2014). All proteins were free of nuclease and/or protease activities for the duration of the relevant assays.

Colorimetric ATPase assay
ATPase activity was measured using a commercial malachite green phosphate detection kit  (Haruta et al., 2006) antiserum.

Rad51 removal from DNA
EMSA-based analysis was performed as described above except that Rad51 was first incubated with the DNA at 37˚C for 5 min, then Rrp1 was added. Fluorescence anisotropy analysis was as follows. For the measurements with ssDNA Rad51 filaments were formed on by incubation of 0.5 µM Rad51 with 1.5 µM ssDNA (oligo-dT, 72-mer) for 5 minutes at 37°C in buffer (30 mM HEPES pH 7.5, 1 mM DTT, 50 mM NaCl, 100 mM KCl, 2 mM ATP, 8 mM PC, 8 U/ml CPK, 3.5 mM MgCl 2 , 2.5% glycerol). This mixture was transferred to a 0.2 x 1.0 cm cuvette (Hellma Analytics) at 37°C. The change in fluorescence anisotropy at 575 nm following excitation at 546 nm was measured for 60 s. After that time, competitor DNA (15 µM nucleotides PhiX 174 virion) and the indicated concentrations of Rrp1 protein were added. Data were collected using an FP-8300 spectrofluorometer (JASCO) every second for 3 minutes. For each reaction, the measurements 20 s before addition of scavenger DNA and Rrp1 were averaged, and the subsequent measurements were expressed relative to this averaged value. where ∆anisotropy in the difference between anisotropy at time zero (average of measurements 20 s before addition of scavenger DNA) and the lowest value observed in the experiment, and Relative change in anisotropy is the change in anisotropy at time t relative to time zero.

Statistical data analysis
For viability assays, Student's t test was used to calculate the P-values (*** p ≤ 0.001, ** 0.001 < p ≤ 0.01, * 0.01 < p ≤ 0.05). To assess statistical significance of proportions of cells with aberrant mitosis and the Rad51 localisation pattern, the Z-test for two population proportions was used to calculate the z-statistic and corresponding p-values ( *** p ≤ 0.001, ** 0.001 < p ≤ 0.01, * 0.01 < p ≤ 0.05).   Mixtures were resolved on an agarose gel and stained with SYBR-gold. (B) Rad51-dsDNA filaments disassemble following addition of Rrp1, as demonstrated by the reduction in anisotropy of fluorescently-labelled dsDNA. Rad51 (6 µM) was incubated with a dsDNA oligonucleotide (3 µM nucleotide concentration) labelled with the TAMRA fluorophore; the resultant high anisotropy value confirms filament formation. Unlabelled heterologous scavenger DNA was then added, followed by a sub stoichiometric amount of Rrp1 (0.25 µM) or the equivalent volume of protein storage buffer, and fluorescence anisotropy was monitored for the indicated time. The decline in anisotropy observed in the reaction containing Rrp1 indicates that Rad51-dsDNA complexes are disassembled.

Figure 5 Rrp1 is an E3 ubiquitin ligase with Rad51 as a substrate
(A) The indicated reaction components were included (+) or omitted (-) for in vitro ubiquitylation assays. After the reaction, the reaction mixture was analysed by western blotting with anti-Rad51 antiserum and anti-Ubiquitin antibodies, and multiple bands indicative of Rad51 ubiquitylation are shown. (B) In vitro ubiquitylation assay containing all components as in (A) with Rrp1-FLAG or Rrp1-CS-FLAG as the E3 ligase. The reaction mixture was analysed by western blotting with anti-Rad51 antiserum and anti-Ubiquitin antibodies, demonstrating that Rrp1 RING domain is indispensable for Rad51 ubiquitylation. Additionally, reaction products were analysed with anti-FLAG antibodies (lowest panel), revealing auto-ubiquitylation of Rrp1. (C) Ubiquitylation of Rad51 by Rrp1 is less efficient in the presence of DNA. In vitro ubiquitylation assay containing all components as in (A) with Rad51 preincubated with 4 µM of ssDNA (PhiX 174 virion) or dsDNA (PhiX 174 RF I linearized with ApaLI). The reaction mixture was analysed by western blotting with anti-Rad51 antiserum and anti-Ubiquitin antibodies. The intensity ratios of mono-ubiquitylated to non-ubiquitylated Rad51 bands normalised to the sample without DNA are shown.