The Rad5 Helicase and RING Domains Contribute to Genome Stability through their Independent Catalytic Activities

Genomic stability is compromised by DNA damage that obstructs replication. Rad5 plays a prominent role in DNA damage bypass processes that evolved to ensure the continuation of stalled replication. Like its human orthologs, the HLTF and SHPRH tumor suppressors, yeast Rad5 has a RING domain that supports ubiquitin ligase activity promoting PCNA polyubiquitylation and a helicase domain that in the case of HLTF and Rad5 was shown to exhibit an ATPase-linked replication fork reversal activity. The RING domain is embedded in the helicase domain, confusing their separate investigation and the understanding of the exact role of Rad5 in DNA damage bypass. Particularly, it is still debated whether the helicase domain plays a catalytic or a non-enzymatic role during error-free damage bypass and whether it facilitates a function separately from the RING domain. In this study, through in vivo and in vitro characterization of domain-speciﬁc mutants, we delineate the contributions of the two domains to Rad5 function. Yeast genetic experiments and whole-genome sequencing complemented with biochemical assays demonstrate that the ubiquitin ligase and the ATPase-linked activities of Rad5 exhibit independent catalytic activities in facilitating separate pathways during error-free lesion bypass. Our results also provide important insights into the mutagenic role of Rad5 and indicate its tripartite contribution to DNA damage tolerance.


Introduction
DNA damage tolerance (DDT) pathways ensure the continuation of DNA synthesis despite replication-stalling DNA damage that can lead to fork collapse and cell death without further action.DDT can be error-free or error-prone depending on the actual mechanism, contributing to genomic stability or increasing the mutational burden of the genome, respectively.The Rad6/Rad18 ubiquitin conjugase/ligase complex controls error-free and error-prone DDT in the budding yeast Saccharomyces cerevisiae (Sc.).It monoubiquitylates proliferating cell nuclear antigen (PCNA) the processivity factor of replicative DNA polymerases on its lysine 164 (K164). 1,2onoubiquitylated PCNA promotes the replacement of the replicative DNA polymerase by translesion synthesis (TLS) polymerases capable of inserting nucleotides opposite DNA damage. 3,4wo different TLS subpathways have been identified for the bypass of ultraviolet (UV) light lesions.One of them is error-free executed by the RAD30encoded Polg that seems to be uniquely specialized in the efficient and error-free bypass of pyrimidine dimers, the most frequent UV-induced DNA lesions. 5,6The other is an error-prone pathway involving the Rev3/Rev7 formed Polf, the Rev1, and the Def1 proteins.In the Polf complex, Rev3 has the catalytic activity, whereas Rev7 is an accessory subunit. 7Polf is highly efficient in extending mismatched primer termini resulting from the insertion of incorrect nucleotides opposite DNA damage by other polymerases, contributing to error-prone DNA synthesis. 8,9The enzymatic activity of Rev1 is limited, being capable of incorporating only cytosines (C) during DNA synthesis leading to error-prone bypass in most cases; however, its dominant role in TLS is considered to be structural. 8,10,11Def1 does not have polymerase activity, but it promotes the polymerase exchange step at stalled replication forks. 12olyubiquitylation of PCNA on K164 already monoubiquitylated by Rad6/Rad18 activates a third subpathway that employs template switching for the error-free bypass of DNA damage. 2,135][16] This multiprotein complex can build a polyubiquitin chain on monoubiquitylated PCNA through lysine 63 of ubiquitin. 17Despite its indispensable role in the MMS2dependent pathway, the function of PCNA polyubiquitylation in DNA damage bypass is still elusive.Furthermore, rad5 mutants are considerably more sensitive to DNA damaging agents than mms2 or ubc13 strains suggesting that Rad5 has other functions independent of Ubc13/Mms2.Indeed, in addition to the RING-finger domain essential for the ubiquitin ligase activity, Rad5 possesses a helicase domain characteristic of the DNA-dependent ATPases of the SWI2/SNF2 protein superfamily. 18he helicase domain confers double-stranded DNA-stimulated ATPase, DNA translocase, and DNA remodeling activities, enabling Rad5 to reverse replication fork-like structures and promote strand invasion, the two possible mechanisms of template-switching. 19,20Considering its enzymatic activities conferred by the RING and helicase domains, Rad5 was suggested to play a role in both the polyubiquitylation and the template-switching steps of the error-free bypass.This model was supported by independent studies suggesting that the two domains played separate roles in the replication of adozelesin-treated DNA and the repair of DNA double-strand breaks. 21,22Nevertheless, the sub-stantive function of the helicase domain and its enzymatic contribution to error-free DNA damage bypass has been questioned.It was reported that both the RING and helicase domains of Rad5 worked exclusively in connection with UBC13 in response to methyl methanesulfonate (MMS) treatment. 23Another study concluded that the sole function of the helicase domain in the Rad6/Rad18 governed DDT was non-catalytic, where it facilitated the Rad5-Ubc13 interaction for PCNA polyubiquitylation. 24An essential non-enzymatic contribution of the helicase domain to error-free DDT, via supporting the interaction of Rad5 with PCNA and Ubc13 for PCNA polyubiquitylation, was reported by others, also. 25ad5, like its human orthologs, the helicase-like transcription factor (HLTF) and SNF2 Histone Linker PHD RING Helicase (SHPRH) tumor suppressors, plays a pivotal role in guarding genomic stability.Besides the sequence homology, these proteins share high similarity in their domain structure, with all three having a RING domain situated in the middle of a conserved helicase domain. 268][29][30][31][32] These data suggested that the three proteins exert similar roles in the replication of damaged DNA using these activities.Therefore, it is of great importance to unravel the molecular steps of their actions.However, the conflicting results of the literature regarding the enzymatic contribution of the Rad5 helicase domain to DNA damage bypass preclude the synthesis of a consistent model.One reason for the contradictions lies in the particular arrangement of the RING and helicase domains, making the selection of domainspecific mutations inactivating only the corresponding enzymatic activity elaborate.Therefore, applying direct assays that check the mutant proteins for both the ubiquitin ligase and the DNA remodeling activities, is rudimentary.In this study, our goal was to perform a comprehensive investigation of Rad5 to better understand its contribution to DDT.Through dissecting the roles of the RING and the helicase domains we also aimed to address the contradictions of the literature.

Selecting mutations inactivating the helicase or the RING domain of Rad5
To investigate the contribution of the helicase and the RING domains of Rad5 to its role in DNA damage bypass, we planned to characterize domain-specific mutants.To address the contradictions of the literature, we included those mutations that were used in reports concluding an essential role for the helicase domain in PCNA polyubiquitylation through its non-enzymatic contribution to the Rad5-Ubc13 interaction.We engineered two mutations in each domain affecting distinct conserved amino acids regarded indispensable for the functionality of the given domain (Figure 1(A)).Lysine-538 in the conserved GKT motif of the Walker A box involved in ATPbinding was changed to alanine in the ATPase mutant rad5-KA. 33The rad5-DEAA mutant had alanine substitutions in the DExx motif of the Walker B box in the ATP-hydrolyzing domain at asparagine-681 and glutamine-682 residues, inactivating both the ATP-hydrolyzing and the DNA remodeling activities of Rad5, as we have previously demonstrated. 20,34To impair the ubiquitin ligase activity of Rad5, we mutated residues necessary for interaction with Ubc13.In rad5-CCAA, the structure of the RING domain was disrupted by replacing two of its metal coordinating cysteines by alanine at positions 914 and 917, whereas the rad5-IA mutant carried a more subtle change at isoleucine-916 that still hindered the interaction with Ubc13. 35The complete sequence of the alleles was verified by sequencing.

In vitro PCNA polyubiquitylation by mutant Rad5 proteins
To corroborate that the mutations selectively inactivated the corresponding domains, we overexpressed and purified the wild-type and mutant proteins from yeasts and examined their ubiquitin ligase and DNA remodeling activities in parallel, in well established in vitro systems.However, the yield of the Rad5-KA protein proved to be very low after purification in repeated attempts.Since the intracellular level of another ATP-binding motif mutant with the GKT/GAA changes was shown to be significantly decreased compared to the wild type, we surmised that altering lysine-538 could affect the stability or the folding of Rad5. 22Since the low concentration of the purified protein hindered its investigation in in vitro enzymatic assays, it was omitted from further investigation (Figure 1(C)).The ability of Rad5 proteins to support PCNA polyubiquitylation was assessed using untagged PCNA that was first loaded onto DNA by the clamp loader replication factor C (RFC).Subsequently, enzymes needed for the mono-and polyubiquitylation steps were added to the reactions.As expected, the RING domain mutant Rad5-CCAA and Rad5-IA proteins were completely defective in PCNA polyubiquitylation (Figure 1(D)).We note, that the small amount of diubiquitylated PCNA detected in the reactions was produced by the Mms2/Ubc13 complex, as they were observed even in the absence of Rad5 (Figure 1(D), compare lane 3 to 5 and 6), and also The intracellular level of ectopically expressed HA-tagged Rad5 proteins was examined in whole-cell extracts by immunoblotting using anti-HA antibody.PGK served as a loading control, showed at the bottom.(C) MMS-induced PCNA polyubiquitylation was tested in rad5 strains expressing 7His-PCNA from a plasmid.Cells were synchronized in G1 by a-factor prior to MMS treatment.Ubiquitylated forms of PCNA from whole-cell extracts were bound to Ni-beads and immunoblotted using an anti-ubiquitin antibody (upper panel).We note, that this antibody does not recognize the monoubiquitylated form of PCNA, as reported earlier by others and us. 2,56Unmodified PCNA in the bound fraction was also detected by applying an anti-PCNA antibody on the lower part of the same membrane (lower panel).shown previously by us and others. 27,36Most importantly, the ATPase mutant Rad5-DEAA exhibited almost as strong ubiquitin ligase activity as the wild type (Figure 1(D), compare lanes 4 and 7).This implies, that the DE681,682AA mutation slightly affects, though certainly does not destroy the ubiquitin ligase activity of Rad5.Since even in vitro PCNA polyubiquitylation requires the Rad5-Ubc13 interaction evidenced by the inactivity of Rad5-IA in these assays, we infer that the Walker B motif of the helicase domain is not necessary for the Rad5-Ubc13 interaction, contrary to previous suggestions by others. 24,25 vitro DNA remodeling activity of mutant Rad5 proteins The functionality of the helicase domain was assessed by monitoring the DNA remodeling activity of Rad5 using a double-stranded branched DNA substrate mimicking a replication fork (Figure 1(E)).We have previously shown that wild-type Rad5 could easily remodel such a substrate into a double-stranded linear form in an ATP-hydrolysis-dependent manner. 20Indeed, the wild-type and the Rad5-IA proteins were able to remodel the substrate DNA, though Rad5-IA showed somewhat lower activity.In contrast, the Rad5-DEAA ATPase mutant was completely inactive in these assays, in agreement with our previous results. 20Interestingly, Rad5-CCAA was as defective as the ATPase mutant indicating that the CC914,917AA RING mutations inactivated not only the RING but the helicase domain, as well.The special arrangement of the two domains suggested that the mutations caused structural perturbation in the RING domain, which probably affected the surrounding helicase domain as well, leading to the inactivation of both domains.During this project, the crystal structure of a nearly full-length Rad5 of the yeast Kluyveromyces lactis (Kl.) showing 46.7% sequence identity with Sc.Rad5 was published, which revealed that though the RING domain is embedded in the helicase domain at the level of the amino acid sequence, protein folding separates the two domains in the threedimensional structure (Figure 1(B)). 37Nevertheless, because of the inactivity of its both domains, this mutant was not investigated further.In summary, the results of the in vitro enzymatic tests indicate that the I916A mutation compromises entirely the ubiquitin ligase, whereas the DE681,682AA the ATPase-linked activity, and they have only a slight effect on the other domain's activity: consequently, they can be considered domain-specific.

PCNA polyubiquitylation in the rad5-IA and rad5-DEAA strains
To investigate the in vivo effects of the mutations, we constructed yeast strains lacking the chromosomal copy of RAD5 and ectopically expressing the wild-type or the mutant Rad5 proteins from a single copy plasmid under the regulation of the endogenous RAD5 promoter and terminator sequences.Importantly, ectopically expressed wild-type RAD5 conferred UV and MMS resistance to yeasts to the same degree as the chromosomal copy of the gene, whereas the mutants caused moderate sensitivities (Figure 2 (A)).Since the expression levels of plasmid-born Rad5 proteins detected in whole-cell lysates were similar to each other and they were not lower than the level of endogenous Rad5, we concluded that the sensitivities were caused by the specific mutations, confirming that both the RING and the helicase domains contributed to Rad5 function (Figure 2(B)).We note that the rad5-DEAA strain exhibited much higher sensitivity to MMS than the rad5-IA, whereas there was considerably smaller difference between the UV sensitivities of the two strains (Figure 2(A)).That might indicate a more substantial role for the Rad5 helicase domain in the bypass of MMS-induced DNA damage.Using these strains, we tested PCNA polyubiquitylation, the best characterized in vivo task of Rad5, to assess whether the ubiquitin ligase activity of the mutants observed in vitro was also reflected in vivo.To facilitate detection, PCNA was expressed ectopically in fusion with a 7-histidine tag, and this 7His-PCNA construct served as a sole source of PCNA in cells lacking the chromosomal copy of the gene.Importantly, as Figure 2(C) shows, PCNA polyubiquitylation in the rad5-DEAA strain was proficient after MMStreatment, though a small reduction in the amount of polyubiquitylated PCNA could be observed compared to the wild type, in accord with the in vitro activity of Rad5-DEAA as a ubiquitin ligase for PCNA.This raised the possibility that a small defect in PCNA polyubiquitylation caused by the DE/AA mutations contributed to the higher UV and MMS sensitivity of rad5-DEAA compared to rad5-IA.
As expected, MMS-induced PCNA polyubiquitylation could not be detected in rad5-IA paralleling the inactivity of the corresponding mutant protein in the in vitro assay.Taken together, these results suggest that the selected mutations exert the same effects on Rad5 activity in vitro and in vivo, as well.
The ubiquitin ligase and ATPase-linked activities of Rad5 promote separate functions Notably, the above experiments demonstrated that the ATPase-linked activity of Rad5 was dispensable for PCNA polyubiquitylation, suggesting that the ATPase-linked and the ubiquitin ligase activities of Rad5 participated in different steps of DNA damage bypass.To corroborate this, we created a rad5-IA/DEAA double mutant strain carrying mutations in both the RING and the helicase domains of Rad5 and compared its sensitivity to the single domain mutants after treating the cells with UV or MMS.As Figure 3(A)-(D) shows, the double mutant exhibited significantly higher sensitivity than the single mutants in different assays, indicating independent roles for the RING and helicase domains.These results were confirmed in experiments investigating the genetic relations of the domain-specific rad5 mutants to mms2D.Epistasis analysis showed that the mms2D rad5-IA double mutant strain exhibited the same sensitivity to UV and MMS as the mms2D single mutant, suggesting that the ubiquitin ligase activity of Rad5 works exclusively in connection with Mms2 (Figures 3(E) and (F)).Contrary to that, the mms2D rad5-DEAA double mutant strain was more sensitive than the corresponding single mutants.Similar results were obtained in the ubc13D background (Figure S1) Taken together, our findings underpin an Mms2-and PCNA polyubiquitylation-independent contribution of the helicase domain of Rad5 to DNA damage bypass.These results correspond well with the spatial separation of the RING and helicase domains in the crystal structure of Kl.Rad5 suggesting independent functions to the domains.We note though, that our results do not exclude the possibility that the helicase domain also contributes to the MMS2-dependent pathway, together with the RING domain.However, the role of the helicase domain must be sequential to PCNA polyubiquitylation and is most probably catalytic.

The effect of domain-specific Rad5 mutants on spontaneous mutagenesis
A characteristic of rad5D strains is the increased rate of spontaneous mutagenesis, indicating that the function of Rad5 is mainly error-free. 38,39To test the contribution of the ubiquitin ligase and the ATPase-linked activities to the error-free function, we examined the effect of the domain-specific mutants on spontaneous mutagenesis.Using the canavanine mutagenesis reporter assay we found that spontaneous mutagenesis was equally elevated in the rad5-IA and in the rad5-DEAA mutant strains exhibiting a more than four times higher mutation rate compared to the wild type.These data suggested that both the ubiquitin ligase and ATPase-linked activities of Rad5 had error-free roles under physiological conditions (Figure 4(A)).Importantly, in the double domain mutant rad5-IA/ DEAA strain the rate of spontaneous mutagenesis was $1.5 times higher than the rate measured in the single mutants pointing to independent error-free functions of the ubiquitin ligase and the ATPase-linked activities of Rad5.It is noteworthy, that the mutation rate of the rad5D strain is lower than that of the single and double mutant strains but still higher than that of the wild type, indicating that Rad5 has a pro-mutagenic function that is independent of its ubiquitin ligase and ATPaselinked activities.

Rad5 exerts its roles via three activities during DNA damage tolerance
Indeed, in addition to its role in error-free DDT, Rad5 was shown to support error-prone bypass of the DNA lesions by TLS DNA polymerases through its interaction with Rev1. 40,41The above experiments showed that the UV and MMS sensitivities of the double domain mutant rad5-IA/ DEAA strain were lower than that of rad5D and that spontaneous mutagenesis in rad5-IA/DEAA was significantly higher than in the rad5D strain in agreement with a mutagenic, non-catalytic role for Rad5 in damage bypass.To obtain additional evidence, we carried out epistasis analysis with prominent members of the mutagenic TLS pathway.In line with the above results, our genetic analysis showed that the sensitivity of the rad5-IA and the rad5-DEAA strains and also of the double domain mutant rad5-IA/DEAA to UV and MMS was greatly increased by deleting REV3 or REV1, confirming that the catalytic activities of the RING and the helicase domains of Rad5 did not contribute to mutagenic TLS (Figures 5 and S2).Importantly, the equal UV and MMS sensitivity of the rev3D rad5-IA/DEAA, the rev1D rad5-IA/DEAA, the rev3D rad5D, and the rev1D rad5D strains indicated that deletion of REV3 or REV1 completely inhibited the RING and helicase domain-independent function of Rad5.This implies that besides the RING and helicase domain-associated catalytic functions, the remaining role of Rad5 is exerted together with Rev1 and Rev3 in error-prone TLS.In this study, we did not investigate the HIRAN domain found at the N-terminus in Rad5 and HLTF because of its involvement in the different functions of Rad5. 424][45][46][47] A critical role of the HIRAN domain of Kl.Rad5 in PCNA polyubiquitylation and replication fork regression was also reported. 37

UV-induced mutational spectrum of rad5 domain-specific mutants
Mutational spectrum is characteristic of enzymes responsible for generating them.Therefore, we planned to identify mutations in rad5 strains to obtain information on the proteins activated in the absence of specific Rad5 functions.However, instead of relying on reporter genes, we sequenced the entire genome by next-generation sequencing to attain a more unbiased picture (Figure 4(B)).To enrich for mutations, we treated cells with different UV doses to obtain $10% survival, before determining the single nucleotide mutation spectrum of each strain (Figure 4(C)).Curiously, as Figure 4(D) shows, the mutation pattern was similar in the wild-type and the single mutant rad5-IA and rad5-DEAA cells; however, striking differences could be detected in the double domain mutant rad5-IA/DEAA samples.First, the ratio of C > T mutations dropped from $40% in the wild type and the single mutants to $20%, moreover, those of the C > G and the T > C mutations increased from 2-3% to $17%, and from $30% to $40%, respectively, in the double mutant.The C > T mutation is a UV signature mutation that most commonly forms at the 3 0 C of TC and CC pyrimidine dimers. 48The reason behind this is that UV induces the deamination of cytosines, and cytosines in dimers are more unstable than in a normal sequence.Indeed, we found that C > T was the most prevalent change accounting for $40% of all mutations in the wild type and the single mutants, and $90% of them occurred at the 3 0 C of dipyrimidine sites (Figure 4(E)).In Escherichia coli, C > T mutations were proposed to mainly result from error-free insertion of adenine opposite the uracil in place of the spontaneously deaminated 3 0 C of the dimer. 49In humans, errorfree bypass of deaminated dimers by Polg was suggested, but the observation that in cells having a defective Polg C > T transition is still the most prominent UV-induced mutation argues against it. 50,51Genetic studies in yeast did not support a role for Polg in generating C > T transitions either, instead, they provided evidence for an essential role of Rev3. 52In light of that, it is puzzling that inactivation of the error-free functions of Rad5 in rad5-IA/ DEAA decreases the incidence of a mutagenic event.All the more because the ratio of C > T mutations goes back to the wild type level in the absence of Rad5.C > G transversions at dipyrimidines are proposed to result from error-prone insertion of C opposite abasic sites by Rev1, which arise after the removal of uracil formed by deamination of the 3 0 C in the dimers. 10Since C > G transversions in rad5-IA and rad5-DEAA cells occur at Cs between two pyrimidines, it is unclear whether the mutations were formed at the 5 0 or 3 0 bases.In addition, in rad5-IA/DEAA C > G mutations are detected at the 5 0 C of dipyrimidines as well, suggesting that error-prone insertion opposite the 5 0 C of the dimers, probably by Rev1, also contributes to C > G transversions.The proportion of T > C inversions also increased by more than 10% in rad5-IA/ DEAA cells.Surprisingly, our data suggest that, despite its role in mutagenesis, the absence of Rad5 results only in minor perturbation in the activation of the mutagenic pathways.The only major difference identified between the spectrums of base substitutions in wild-type and rad5D cells is the strongly reduced ratio of T > C inversions, which dropped from $30% in the wild type to 10% in rad5D.This finding correlates well with data showing that T > C mutations resulting from translesion synthesis opposite (6-4) TT photoproducts in yeasts almost entirely depend on RAD5. 53Taken together, these results further confirm an independent function of the Rad5 RING and helicase domains and they reveal that concurrent inactivation of the two domains alters the activation of the mutagenic pathways.Based on these findings, we infer that in wild-type cells the mutagenic pathway relies on the catalytic activity of Rev3.However, when error-free damage bypass by Rad5 is impaired, the catalytic contribution of Rev1, which is marginal in the wild type and the single mutants, becomes significant, and this pathway depends on a non-enzymatic contribution of Rad5.In summary, the results we obtained applying different assays are congruent and they establish that the contribution of both the helicase and the RING domains to error-free damage bypass requires their catalytic activities, which can work independently of each other, and during bypass they can support separate functions.Moreover, we can also conclude that Rad5 has only one additional role, a non-enzymatic contribution to mutagenic translesion synthesis, where it plays a dominant role during damage bypass catalyzed by Rev1 (Figure 6).Future studies aiming at exploring the interplay between selected activities of the multifunctional Rad5 and other proteins of DDT are needed to reveal the full spectrum of influence that Rad5/HLTF/SHPRH exert on genome stability.

Yeast strains
Single deletion mutants in BY4741 and BY4742 background (EUROSCARF) were applied in the genetic studies.Double mutants were made by crossing, and deletions were confirmed by PCR.For the rescue of rad5D, strains were transformed with wild-type or mutant Rad5-expressing plasmids and maintained on synthetic dropout (SD) medium lacking leucine (-leu) to select for the plasmids.The BJ5464 strain was used for overexpressing Rad5 proteins.Spontaneous forward mutation frequencies at the CAN1 locus were measured in EMY74.7 background.PCNA ubiquitylation was assayed in DF5a derivatives 54 expressing 7His-PCNA and the Rad5 proteins from centromeric plasmids.

Plasmids
Constructs used in this study are listed in Table 1.For rescue assays, the genomic RAD5 PstI-SalI fragment containing promoter (339 bp) and terminator (623 bp) sequences were cloned into the centromeric low-copy plasmid YCplac111.Point mutations were generated in this plasmid via site-specific PCR mutagenesis and were verified by sequencing.The wild-type and the mutant RAD5 genes were C-terminally tagged with six copies of the hemagglutinin epitope tag (6-HA) in the YCplac111 vector. 55For the purification of the mutant Rad5 proteins, the point mutant and wildtype RAD5 ORFs were first cloned into Gateway Ò entry vector (pENTR TM ) followed by recombination into destination plasmid (pBJ842 backbone), from which Rad5 was expressed in fusion with an Nterminal glutathione-S-transferase (GST) tag, under the control of a modified galactose-inducible phosphoglycerate kinase (PGK) promoter.The 7His-PCNA, with its promoter and terminator sequences, was cloned into the centromeric lowcopy plasmid YCplac33. 56

Yeast sensitivity assays
For qualitative serial dilution assays, cells were grown in SD-leu to A 600 :0.6, counted, and cell number was adjusted to 5 Â 10 6 cells/ml.Dilutions were prepared in 3Â steps and spotted onto SDleu plates containing different amounts of MMS.For UV treatment, dilutions were spotted onto SDleu plates and then UV-irradiated.Plates were scanned after 2-4 days of incubation in the dark at 30 °C.
For quantitative colony-counting plate assays, yeasts were grown to A 600 :0.6, cells were counted, and cultures were diluted to 10 7 cells/ml and incubated by shaking for 1 hour at 30 °C in SD-leu containing different concentrations of MMS.After washing, appropriate dilutions were made and cells were spread onto SD-leu plates.Plates were incubated for 3 days at 30 °C before colonies were counted.
For growth curve assays, yeast strains were grown at 30 °C for 16 hours, and approximately 50 cells from each strain were inoculated into 500 ml SD-leu with or without 100 mM MMS. Plates were shaken at 30 °C in Synergy 2 Multi-Mode Reader (BioTek Ò ), and A 600 was measured every 4 minutes for 48 hours.Growth curves were determined using Microsoft Ò Excel 2010.

Spontaneous mutagenesis
Seven parallel cultures of approximately 50 cells each were grown in SD-leu for 2 days at 30 °C.The cultures were plated onto canavaninecontaining SD-leu/-arg plates, and the median number of mutants in the case of each strain was counted for 10 7 plated cells.Mutation frequencies were determined using a chart based on the Lea-Coulson fluctuation model. 57Results are averages of 5 independent experiments.

Whole-genome sequencing
A single colony from each strain was picked and dispersed in PBS and spread onto two SD-leu plates of which one was irradiated with UV.Different UV dosages were used to get $10% survival with each strain (WT: 100 J/m 2 , rad5-IA: 60 J/m 2 , rad5-DEAA: 40 J/m 2 , rad5-IA/DEAA: 30 J/m 2 , rad5D:13 J/m 2 ).After 9-hour incubation at 30 °C in the dark, cells were collected from plates with PBS and were spread onto fresh SDleu plates.Following a 3-day incubation at 30 °C, genomic DNA was extracted by phenolchloroform from 3 colonies/samples, and the 3 parallel DNA samples were pooled.With the rad5D strain, the UV treatment and incubation steps were repeated 5 times using 3 single colonies.For whole-genome sequencing, the library was made with Nextera XT Library Preparation Kit (Cat.no.FC-131-1024, Illumina).For obtaining 2 Â 150 bp paired-end sequence reads, Illumina NextSeq 500 sequencer was used.The average sequencing coverage was 100.Reads were aligned to the GCA_000146045.2 reference genome using BWA mem. 58SAM files have been transformed to BAM and indexed with SAMtools. 59Sequencing duplicates were removed using GATK MarkDuplicates. 60Variants have been identified with freebayes using --min-base-quality 28 command and filtered with Vcflib using vcffilter tool and -f 'QUAL > 100 0 setting. 61Sample pairs have been compared by selecting only those variants which were not present, or their frequency was below 10% in untreated samples and over 60% or 20% in single or mixed samples, respectively.Data produced by whole genome sequencing is available at NCBI Genome Data Viewer (https://dataview.ncbi.nlm.nih.gov/object/PRJNA726574?reviewer=3ak2c50vp52g91hon6gj7dps78)

Figure 1 .
Figure 1.In vitro activities of recombinant Rad5 proteins.(A) Schematic representation of the domain structure of Rad5.The conserved motifs of the SWI2/SNF2 helicase domain are depicted as green boxes, the RING domain is indicated by a blue box, and the HIRAN domain is represented by a red box.The positions of the mutations generated are indicated by arrows.(B) The positions of the corresponding mutations in the crystal structure of Kluyveromyces lactis Rad5.ScRad5 amino acid numbers are in brackets.The RING domain is shown in blue, and the lobe 1 of the helicase domain in green.(C) Purity of the wild-type and mutant Rad5 protein samples (10 ml each).(D) PCNA polyubiquitylation by purified Rad5 proteins.The activity of wild-type and mutant Rad5 was assayed in the presence of Rad6/Rad18 and Mms2/Ubc13.Ub-PCNA: monoubiquitin-PCNA, PolyUb-PCNA: polyubiquitylated PCNA.(E) The dsDNA translocase activity of Rad5 proteins was tested using a fluorescently labeled oligonucleotide-based replication fork-like structure, which could be converted into a linear heteroduplex by Rad5.Schematic structures of the substrate and the product are shown on the left.Percentages of the remodeled linear heteroduplex formed after 15 and 30 minutes, as indicated, are given at the bottom.The linear heteroduplex detected in the first lane represents spontaneous realignment during the extended incubation.

Figure 2 .
Figure 2. In vivo characterization of rad5-IA and rad5-DEAA.(A) DNA damage sensitivity of strains expressing wildtype or mutant Rad5.All strains were generated by transforming rad5D with an empty vector, or with plasmids expressing wild-type or mutant Rad5.Serial dilutions of the strains were spotted on plates containing the indicated doses of MMS or were UV-irradiated with the indicated UV doses after spotting.(B)The intracellular level of ectopically expressed HA-tagged Rad5 proteins was examined in whole-cell extracts by immunoblotting using anti-HA antibody.PGK served as a loading control, showed at the bottom.(C) MMS-induced PCNA polyubiquitylation was tested in rad5 strains expressing 7His-PCNA from a plasmid.Cells were synchronized in G1 by a-factor prior to MMS treatment.Ubiquitylated forms of PCNA from whole-cell extracts were bound to Ni-beads and immunoblotted using an anti-ubiquitin antibody (upper panel).We note, that this antibody does not recognize the monoubiquitylated form of PCNA, as reported earlier by others and us.2,56Unmodified PCNA in the bound fraction was also detected by applying an anti-PCNA antibody on the lower part of the same membrane (lower panel).

Figure 3 .
Figure 3. Genetic interactions of rad5 mutants.The rad5-IA/DEAA is more sensitive to UV and MMS than the single mutants as shown using (A) serial dilution spot assay (B) growth curve assay with MMS-containing medium (C) killing curve by colony-counting plate assay.For (B), cultures containing $50 cells were grown for 48 hours, and A 600 was recorded with 4 min intervals.For (C) logarithmic cultures were treated with increasing concentration of MMS for 1 hour at 30 °C before plating.Colonies were counted after incubation for 3 days at 30 °C.The graph shows data of 3 independent experiments with standard deviations.(D) Results of the killing curve analysis shown in Fig. 3C at 5 mM MMS concentration depicted in a graph.P-values were calculated by Student t-test.*: p < 0,05, **: p < 0,01, ***: p < 0,001.(E,F) rad5-IA is epistatic, whereas rad5-DEAA is additive with mms2.Epistasis analysis of the given strains was done using serial dilution spot assay after (E) UV and (F) MMS treatment.

Figure 4 .
Figure 4.The effect of the RING and helicase mutations of RAD5 on mutagenesis.(A) Spontaneous forward mutagenesis assay with rad5 mutant strains.Canavanine resistant colonies were counted and calculated for 10 7 surviving cells.The graph represents the average of 5 independent experiments.Standard deviations are shown.(B) Flowchart of sample processing for whole-genome sequencing.(C) UV killing curve of the indicated strains to determine the UV doses necessary to obtain $10% survival before collecting cells for whole-genome sequencing.(D) Mutation signature analysis of UV-treated rad5 mutant strains.The frequencies of 6 categories of mutations are represented.(E) C > T mutations occur primarily at the 3 0 C of TC dipyrimidines.

Figure 5 .
Figure 5. Epistasis analysis of rad5 mutants with rev3D and rev1D.(A) A serial dilution assay using increasing UV doses.(B) Genetic relation of rev3D and rad5 mutants examined after MMS treatment.The graph represents data from 3 independent colony-counting plate assays.Cells were treated with increasing concentrations of MMS and spread onto plates.Colonies were counted after 3 days of incubation at 30 °C.Standard deviations are also shown.

Figure 6 .
Figure 6.Contribution of Rad5 to DDT.Arrows with dashed lines indicate a putative error-free pathway involving both the helicase and RING domain-linked activities of Rad5.

Table 1
Plasmids used in the study.