Intracellular level of S. cerevisiae Rad51 is regulated via proteolysis in a SUMO- and ubiquitin-dependent manner

Among various DNA lesions, the DNA double-strand breaks are particularly deleterious; especially, when an error-free repair pathway is unavailable, and the cell takes the risk of using the error-prone recombination pathways to repair the DNA breaks, resume the cell cycle, and continue growth. The latter comes at the expense of decreased well-being of the cells due to genome rearrangements. One of the major players involved in recombinational repair of DNA damage is Rad51 recombinase, a protein responsible for presynaptic complex formation. We previously noticed that the level of this protein is strongly increased when illegitimate recombination is engaged in repair. The regulation of Rad51 protein turnover is not known; therefore, we decided to look closer at this issue because we expect that an excessively high level of Rad51 may lead to genome instability. Here we show that the level of Rad51 is regulated via the ubiquitin-dependent proteolytic pathway. The ubiquitination of Rad51 depends on multiple E3 enzymes, including SUMO-targeted ubiquitin ligases. We also demonstrate that Rad51 can be modified by both ubiquitin and SUMO. Moreover, these modifications may lead to opposite effects. Ubiquitin-dependent degradation depends on Rad6, Rad18, Slx8, Dia2 and the anaphase-promoting complex. Rsp5-dependent ubiquitination leads to Rad51 stabilization.


43
The maintenance of genetic information is essential for cells but not easy to achieve due 44 to constant environmental and metabolic threats. Particularly deleterious for genome integrity are 45 stresses causing double-strand breaks in DNA because this type of damage may lead to DNA 46 rearrangements or loss of parts of chromosomes. To avoid such scenarios and restore vital genetic information, the repair pathways evolved, reassuring the re-connection of the broken DNA strands.

48
In the yeast Saccharomyces cerevisiae, the most accurate repair pathway dedicated to double-49 strand break repair is homologous recombination. One of the proteins essential for this repair is the 50 Rad51 recombinase. Rad51 belongs to the RecA family of recombinases (1). Due to its capacity to 51 bind single-and double-stranded DNA, DNA-dependent ATPase activity, ability to form a filament 52 on DNA, and feedback interactions with various proteins engaged in DNA repair (Rad54, Rad52, 53 Sgs1, replication protein A/RPA complex, etc.), the recombinase Rad51 executes the critical early 54 step of homologous recombination: the search for homologous DNA to serve as a template during 55 the repair of DNA double-strand breaks (2-6). The rad51 mutants display replication defects and 56 chromosomal instability (7). The orthologs of yeast RAD51 gene have been identified in various 57 organisms, including humans (8). In vertebrates, the absence of Rad51 leads to embryonic lethality 58 (9). Moreover, mutations in human RAD51 are linked to breast cancer (10,11) and Fanconi anemia 59 (complementation group R, FANCR) (12,13).

60
Rad51 is involved in mitotic and meiotic recombination; however, its role is slightly different 61 in each process. Because Rad51's homolog Dmc1 is present during meiosis, the proteins share 62 their responsibilities. While Dmc1 is responsible for interhomolog recombination, Rad51 promotes 63 Dmc1 presynaptic filament assembly and participates in intersister repair, leading to non-crossover 64 products (14-16).

117
Cells wield multiple proteolytic systems to carry out protein degradation. One of the most 118 common pathways used for this purpose is ubiquitin-dependent proteolysis via the 26S 119 proteasome. To answer the question of whether Rad51 is a substrate for the proteasome, we 120 measured the steady-state level of Rad51 in yeast mutants defective in various proteasomal 121 activities. We exploited the pre2-K108R, pre3-T20A, and pup1-T30A mutant strains, where 122 chymotrypsin-like, trypsin-like, and peptidyl-glutamyl peptide-hydrolyzing (PHGH) activities of 20S 123 proteasome core were inactivated, respectively (22). In proteasome mutant strains there was an 124 increased level of Rad51 (Figure 1 C). A similar effect was visible in ump1∆ cells lacking 125 proteasome maturase (23). We therefore concluded that Rad51 is degraded by the proteasome.

126
The proteins that undergo proteasomal degradation first have to be tagged with ubiquitin.

127
Thus, we took advantage of the His-tagged ubiquitin system to check whether Rad51 is modified 128 by ubiquitin. Using Ni 2+ -NTA sepharose, we purified ubiquitinated proteins from the strains carrying

136
We performed a similar analysis with the pre2-K108R strain, expecting better visibility of 137 Rad51 ubiquitinated forms due to decreased turnover; however, we detected a reduced level of the 138 poly-ubiquitinated species (Figure 1 D). Based on the total cell extract it was clear that there was a 139 relative accumulation of Rad51, including the poly-ubiquitinated forms, in pre2-K108R cells. One 140 possible explanation would be that the pre2-K108R cells accumulate ubiquitinated proteins, due to 141 defects in proteasome-dependent degradation, and these substrates compete for binding to the 142 Ni 2+ -NTA sepharose. Therefore, although the number of Rad51 poly-ubiquitinated forms increases 143 in the pre2-K108R mutant, they cannot be visualized efficiently in such an analysis. In agreement 144 with this hypothesis would be our observation that the flow-through fraction still contains a 145 substantial level of higher molecular weight forms of Rad51 (data not shown). Interestingly, the 146 pattern of poly-ubiquitinated forms of Rad51 in total extracts and Ni 2+ -NTA-bound protein fraction 147 is slightly different, which suggests His-tagged ubiquitin does not necessarily substitute for the     Then, we compared the two obtained datasets; overlapping candidates were examined further. We

231
which was set to 1.00, was calculated from 3 to 5 biological repetitions (blue numbers below the blot).

233
Thus, we asked whether Rad51 is modified by another ubiquitin-like modifier, SUMO. In       The SUMO chain may recruit the SUMO-targeted ubiquitin E3 ligases (STUbLs), which 273 modify proteins to direct them for degradation. One such enzyme is the Slx5-Slx8 complex. We 274 asked whether this complex contributes to the Rad51 level limitation. We found that the Rad51 275 level was raised in the slx8∆ mutant compared to the control (Figure 3 A). All these results 276 confirmed our hypothesis that SUMOylation is involved in Rad51 degradation.

278
Multiple enzymes regulate the ubiquitination of Rad51 in vivo.

279
An increased Rad51 level was detected in strains lacking various E2 and E3 ligases ( Figure 2).

280
Which of them is then responsible for the ubiquitination of Rad51? To address this issue, we used 281 two opposite approaches. First, we looked for E2 or E3 activities, whose absence resulted in the 282 disappearance of Rad51 ubiquitinated forms from the cell. Second, we looked for enrichment of 283 ubiquitinated forms of Rad51 when specific E2 or E3 enzymes were overproduced; we reasoned 284 that increased activity in the conjugating and ligase enzymes would overwhelm the capacity of the

469
(data not sown). We postulate that Siz1 SUMOylates the proteins that contribute to Rad51 stability.

470
At least two such proteins could be implicated. Rsp5 is subject to Siz1-dependent SUMOylation, 471 which results in its reduced ubiquitin ligase activity (50). Because Rsp5 stabilizes Rad51, the 472 activity of Siz1 will stimulate Rad51 degradation. The other Siz1 substrate is Pol30, the yeast 473 proliferating cell nuclear antigen (PCNA), the ring-shaped trimeric complex that encircles DNA and 474 functions as a sliding clamp and processivity factor for replicative DNA polymerases (50-52).

475
SUMOylated PCNA recruits Srs2 and Rad18, activating Rad51 translocase activity and ubiquitin 476 ligase activity, respectively, which both act in an anti-recombinogenic manner and likely influence 477 further Rad51 cellular fate (41, 53-56). We do not know what is happening with Rad51 protein 478 stripped-off from DNA by the Srs2 helicase/translocase, but we showed that Rad51 might be 479 SUMOylated and that one of the STUbLs involved in its poly-ubiquitination is Rad18 (Figure 4).  Rad51, shortened by about 9 kDa. When Rsp5 was overproduced, we saw additional bands of 498 Rad51 migrating slower than mono-ubiquitinated Rad51. These bands may reflect poly-499 ubiquitinated forms of Rad51; however, some of them had a mass which is a duplication or 500 triplication of the Rad51 mass. Thus, it is also possible that the bands reflect oligomers of Rad51.

501
In such a case, the process of Rad51 oligomerization would be stimulated by ubiquitination with 502 the ubiquitin E3 ligase Rsp5 (Figure 4 B, Figure 5). Interestingly, the upper Rad51 bands appeared 503 not only when we detected ubiquitinated Rad51 derivatives but also when we detected

512
Interestingly, the Rad51 degradation product detected in the samples overproducing Rad18

565
SUMOylation and ubiquitination are involved in regulating processes that are crucial for 566 genome maintenance, such as replication, DNA repair, and chromosome segregation. However, 567 we do not know yet how SUMOylation and ubiquitination influence the molecular activity of Rad51.

568
Nevertheless, the data presented here permitted an initial look at this regulation circuit, which is 569 summarized in Figure 6. The Rad51 level, and likely also its activity, is regulated by a network of 570 ubiquitin and SUMO ligases. It is highly probable that each of these enzymes acts in certain 571 circumstances. According to current knowledge concerning other substrates of these enzymes, we 572 can anticipate the particular moments in the life of the cell when they play a major role in Rad51 573 regulation.

622
Determination of protein half-life.

623
Yeast cells were grown at 28°C in YPD medium to a density of 5 x 10 7 cells/ml. Then, the culture

667
The affinity isolation of His6-tagged ubiquitinated and His6-tagged SUMOylated Rad51 was 668 performed in two biological replicates, as described previously (71). The strains in the WCG4a

687
Yeast strains from the MORF collection carrying plasmid of interest were grown on selective 688 medium SC-URA+2% glucose at an appropriate temperature with shaking to the exponential phase 689 (5×10 6 cells/ml). Then cells were collected by centrifugation at 1600 g for 3 min, washed with YNB 690 medium, suspended in SC-URA+2% galactose, and allowed to grow for an additional 4 h to induce 691 the GAL1 promoter. Next, 1×10 9 cells were collected by centrifugation at 1600 g for 3 min, and