All strains used in the present study are derivative of strain EMY74.7 (Johnson et al., 1998) (MATa his3-Δ1 leu2-3,112 trp1-Δ ura3-Δ met25-Δ phr1-Δ rad14-Δ msh2Δ:hisG). In order to study tolerance events, all strains are deficient in repair mechanisms: nucleotide excision repair (rad14), photolyase (phr1), and mismatch repair system (msh2). Gene disruptions were achieved using PCR-mediated seamless gene deletion (Akada et al., 2006) or URAblaster (Alani et al., 1987) techniques. Rad5 point mutations were created using the delitto perfetto method (Storici and Resnick, 2006). All strains used in the study are listed in Table 1.

Integration of plasmids carrying (6‐4)TT/N2dG-AAF lesions (or control plasmids without lesion) and result analysis was performed as previously described (Maslowska et al., 2019). Lesion tolerance rates were calculated as the relative integration efficiencies of damaged vs. non-damaged vectors normalized by the transformation efficiency of a control plasmid (pRS413) in the same experiment. DA events are calculated by subtracting TLS events from the total lesion tolerance events.

All experiments were performed at least in triplicate. Graphs and statistical analysis were done using GraphPad Prism applying unpaired t-test. Bars represent the mean value ± s.d.

Overnight cultures of strains carrying Rad5 point mutations in YPD were adjusted to an OD₆₀₀ value of 1. Volume of 10 μl from 10-fold serial dilutions of OD₆₀₀-adjusted cultures were spotted on YPD agar plates containing different concentrations of 4-NQO (0 μM, 0,007 µM, 0,015 μM and 0.03 μM).

Our group has recently developed an assay based on the insertion of a single replication-blocking lesion into a specific locus in the genome of a living yeast cell, which allows a phenotypical detection of TLS and DA events (Figure 1A) (Maslowska et al., 2019). Our system allows to monitor both error-free and mutagenic tolerance events, overcoming the limitations of assays measuring chromosomal mutagenesis after treatment with mutagenic agents which are blind to error-free events. In our previous study, we have demonstrated that inactivation of ubc13 is compensated by a 10 fold increase in TLS usage (error-free and mutagenic events combined) at a (6–4)TT UV lesion, while other studies reported that ubc13 inactivation led to a ∼2-fold increase in UV-induced mutagenesis (43), reflecting only the low fraction of mutagenic TLS events. Therefore, our method allows to provide more direct evidence for lesion bypass processes than previously used methods.

In the present work, we have used this assay to directly analyze the contribution of each of the Rad5 domains to both branches of the DNA damage tolerance, and determine their role in maintaining balance between TLS and DA. The use of site-specific DNA lesions provides more direct evidence for the role of Rad5 activities in different aspects of lesion bypass.

We have introduced a (6–4)TT photoproduct lesion (thymine-thymine pyrimidine (6–4)pyrimidone photoproduct), or a N2dG-AAF (N2-dG-Acetylaminofluorene) adduct in the genome of cells carrying mutations affecting different domains of the Rad5 protein (Figure 1B): i) an allele simultaneously deficient in Ubc13-binding and ATPase/helicase activity (DE681,682AA) named RAD5DE ^ubi-helic (Blastyák et al., 2007); ii) the Ubc13-binding RING domain (I916A) named RAD5IA ^ubi (Ulrich, 2003); iii) the helicase domain (Q1106D) named RAD5QD ^helic (Choi et al., 2015); iv) and the Rev1-binding domain (FN13,14AA) named RAD5FN ^Rev1 (Xu et al., 2016); v) and a complete deletion of rad5 gene. The results were compared to the parental strain expressing wild-type RAD5 gene. In all strains we inactivated rad14 to avoid repair of the lesion and focus on lesion tolerance mechanisms, and msh2 to avoid repair of the strand marker (+2 nt loop) that allows to distinguish TLS from DA events.

The mutation affecting Ubc13-binding (RAD5IA ^ubi ) led to a strong increase in TLS at both (6–4)TT photoproduct and N2dG-AAF lesions (Figures 2A,B). This increase in TLS is similar to the one we have previously observed in the absence of ubc13 (Maslowska et al., 2019; Masłowska et al., 2022). It confirms that in the absence of PCNA poly-ubiquitination (either in the absence of ubc13, or by inactivation the ubiquitin ligase domain of Rad5), DA is reduced favoring TLS. We have previously described a competition between TLS and DA: in the absence of polyubiquitination of PCNA, DA is inhibited favoring TLS. However, it should be noted that in the absence of PCNA polyubiquitination there is still a proportion of cells surviving using a recombination pathway independent of PCNA ubiquitination that has previously been described as the salvage recombination pathway (Pfander et al., 2005). As shown in Figure 4, RAD5IA ^ubi also shows a high sensitivity to a global genotoxic stress such as 4NQO (4-Nitroquinoline-1-oxide) treatment, similar to the ubc13∆ strain.

These data confirm the role of the ubiquitin-ligase function of Rad5 in promoting PCNA-ubiquitination and shows how it favors error-free lesion bypass.

As we have shown previously, that TLS bypass of the (6–4)TT photoproduct relies almost exclusively on the TLS polymerases Rev1 and Pol ζ (Maslowska et al., 2019; Masłowska et al., 2022). The bypass of N2dG-AAF lesion is mostly dependent on Rev1 and Pol ζ, while a small part can be performed by pol η (Pagès et al., 2008; Masłowska et al., 2022). The RAD5 allele unable to bind Rev1 (RAD5FN ^Rev1 ) causes a severe decrease in the level of TLS at both (6–4)TT photoproduct and N2dG-AAF lesions (Figures 2A,B).

This indicates that the interaction of Rev1 with Rad5 is critical for its TLS activity in vivo. It has been shown previously that non-catalytic function of Rev1 in translesion synthesis and mutagenesis is mediated by its interaction with Rad5 (Kuang et al., 2013). Previous studies have also demonstrated that lack of the Rad5 N-terminal activity severely compromises spontaneous and DNA-damage-induced mutagenesis (Xu et al., 2016; Gallo et al., 2019).

For the (6–4)TT photoproduct, we confirmed that in the absence of PCNA polyubiquitination in mutants RAD5DE ^ubi-helic and RAD5IA ^ubi , the strong increase in TLS (to a level >30%) was still exclusively due to Rev1-Pol ζ: as observed in Figure 3, the inactivation of rev1 in the ubc13∆, RAD5IA ^ubi , or RAD5DE ^ubi-helic mutants completely abolishes TLS (≤0.1%). Similarly, inactivation of Rev3 in the RAD5IA ^ubi also abolishes TLS. When preventing the recruitment of Rev1 by Rad5 (RAD5FN ^Rev1 ) in the same mutants RAD5IA ^ubi and RAD5DE ^ubi-helic , we also observed a strong decrease of TLS. However, unlike in the rev1∆ strains, TLS is not completely abolished when combining default of PCNA polyubiquitination with RAD5FN ^Rev1 . It appears that when DA is inhibited, Rev1/Pol ζ can access the stalled fork and some TLS can occur despite the absence of recruitment of Rev1 by Rad5. However, this occurs at a much lower efficiency than when Rev1 is actively recruited by Rad5.

In response to 4NQO treatment, the RAD5FN ^Rev1 show the same increased sensitivity as the rev1∆ mutant, confirming the importance of Rad5 in recruiting Rev1. It remains however less sensitive than the mutant affecting PCNA polyubiquitination, reflecting the lesser role of TLS for survival over DA. When DA is abolished in the absence of PCNA polyubiquitination (strains ubc13∆, RAD5IA ^ubi or RAD5DE ^ubi-helic ), the further inactivation of rev1 either directly (rev1∆ strain) or through the lack of recruitment by Rad5 (RAD5FN ^Rev1 ) does not further increase the sensitivity to 4NQO, indicating again the minor role of TLS for survival.

While the complete deletion of rad5 leads to an increased sensitivity to 4NQO (Figure 4), it does not lead to a drastic phenotype when monitoring the bypass of a single (6–4)TT photoproduct or N2dG-AAF lesions compared to WT RAD5 (Figures 2A,B). We observed a significant decrease in TLS for the N2dG-AAF lesions, and a very moderate decrease for the (6‐4)TT photoproduct compared to the parental strain. It is important to note that in the rad5∆ strain, no polyubiquitination of PCNA occurs: we could therefore expect in these strains a strong increase of TLS as observed in the ubc13∆, RAD5IA ^ubi , or RAD5DE ^ubi-helic mutants (Maslowska et al., 2019; Masłowska et al., 2022). However, due to the absence of Rad5 and its function of recruiting Rev1, TLS does not increase in this strain. Overall, the loss of DA due to the absence of PCNA-ubiquitination could not be compensated by an increase in TLS in the absence of Rev1 recruitment, and is therefore compensated by an increase in the salvage recombination pathway.

This illustrates the dual and opposite roles of Rad5 in lesion tolerance.

While the helicase function of Rad5 has been clearly evidenced in vitro, its functions in vivo remain controversial. Previous studies have shown that the helicase mutants are mildly sensitive to alkylating reagent-induced stress (Minca and Kowalski, 2010; Choi et al., 2015), but not to replication stress caused by nucleotide pool depletion (Gallo et al., 2019). However, those studies were done using different helicase mutants, therefore drawing clear conclusions from them is a difficult task. For that reason, the involvement of the helicase domain in lesion bypass remains unclear.

It is important to note that previous studies have considered the DE mutant solely as an ATPase defective strain and concluded as a role for the helicase function in DDT (Gangavarapu et al., 2006; Blastyák et al., 2007). However, it has since been showed that the DE mutation affects not only the helicase function, but also ubiquitin ligase activity due to reduced interaction with Ubc13 and PCNA (Ball et al., 2014; Choi et al., 2015). Therefore, the sensitivity that was observed for this mutant was most likely due to the polyubiquitination defect and not the helicase defect.

We have used a true separation of function mutation affecting solely the ATPase/helicase domain (Q1106D) of Rad5 (RAD5QD ^helic ). In this mutant, we did not observe any change of the level of TLS and DA at the (6–4)TT photoproduct and N2dG-AAF lesions (Figure 2A and B). It seems therefore that this function is not involved in the bypass of the tested lesions. One could wonder if the helicase function could act as a backup in the absence of DA. The levels of TLS and DA are the same in the mutant deficient for both helicase and ubiquitin ligase (RAD5DE ^ubi-helic ) as for the mutant deficient solely for the ubiquitin ligase (RAD5IA ^ubi ). Similarly, there is no difference between ubc13∆ and ubc13∆ RAD5QD ^helic strains. These observations indicate that even in the absence of genuine DA that is dependent on PCNA polyubiquitination (ubc13∆ or RAD5IA ^ubi ), the helicase has no function in damage tolerance. We also generated a mutant deficient for both the ubiquitin-ligase and Rev1-binding domain (RAD5IA ^ubi RAD5FN ^Rev ), where only the helicase function of Rad5 remains (Figure 3A and B). These mutants show the same phenotype as a complete deletion of RAD5, confirming again that the helicase domain has no function in damage tolerance.

These results obtained with our two DNA lesions are compatible with previous observations from Gallo et al, (2019) who showed that in response to HU treatment, the helicase mutant had no effect on mutagenesis or survival. The absence of sensitivity of the RAD5QD ^helic mutant to 4NQO treatment (Figure 4) confirms that the helicase domain in not involved in lesion tolerance. Previous reports from Chen et al, (2005) have shown the involvement of the helicase function of Rad5 in double-strand break repair, a role that is independent from its ubiquitin ligase function. In vitro experiments have shown the involvement of the helicase domain in fork regression (Blastyák et al., 2007), a structure that could favor error-free lesion bypass. In vitro experiments have also suggested that Rad5 can facilitate strand invasion-dependent mechanisms in addition to fork regression for the template switching (Burkovics et al., 2014). It appears from our in vivo data and others (Chen et al., 2005; Gallo et al., 2019) that this is not a major pathway in vivo, at least for the tested lesions. While previous studies have suggested that the helicase function of Rad5 could contribute to lesion tolerance through fork regression, we show here that the helicase domain does not participate in the bypass of DNA lesions.