Introduction

During cell growth, the genetic material can be damaged by internal or external sources and cells must deal with this threat to faithfully transmit the genetic material to the next generation. In addition to cellular checkpoint pathways that prevent entry into mitosis with damaged chromosomes, sophisticated DNA repair and DNA damage tolerance mechanisms are in place. These are conserved in all eukaryotes, and much of our knowledge about them comes from the study of simple organisms, such as the yeast Saccharomyces cerevisiae (Adames et al. 2019; Ballew and Lacefield 2019; Gkouskou et al. 2019; Tutaj et al. 2019).

The DNA is particularly prone to attack during DNA replication, a time at which the DNA is particularly exposed (Corcoles-Saez et al. 2019; Moriel-Carretero et al. 2019). DNA replication is a complex process, and fork progression may be halted by lesions in the DNA, by secondary structures or by the presence of bound proteins (Owiti et al. 2019). Cells deal with this situation by either repairing the damage or bypassing it completely, thus preventing the situation from escalating into fatal genomic rearrangements (Singh and Wu 2019). The DNA damage tolerance (DDT) pathway (sometimes also known as the post-replication repair pathway) becomes activated when single-stranded DNA gaps are formed during DNA replication, following DNA polymerase stalling and re-start events (Karras and Jentsch 2010). Central to the DDT is PCNA, the processivity factor for replicative DNA polymerases, composed of three identical subunits that form a ring encircling the DNA (Garbacz et al. 2020). Upon polymerase stalling, PCNA becomes mono-ubiquitinated at lysine 164 by the Rad6/Rad18 E2/E3 complex. This modification allows the exchange of the replicative polymerases by specialized trans-lesion synthesis (TLS) polymerases that are able to synthesize DNA past the lesion, albeit in a less accurate fashion, thus creating mutations (Acharya et al. 2019; Bebenek and Ziuzia-Graczyk 2018; Stelter and Ulrich 2003; Szwajczak et al. 2018). The mono-ubiquitinated PCNA molecules can otherwise be further ubiquitinated, creating poly-ubiquitin chains, by the combined activity of the E2 enzymes Ubc13-Mms2 and the E3 Rad5, creating K63-linked ubiquitin chains (Fan et al. 2018). This modification somehow orchestrates a template switch (TS) event, in which information is copied from the newly synthesized sister chromatid. The molecular details of this process, as well as the nature of potential “readers” of the poly-ubiquitin signal, are still mysterious (Branzei et al. 2008). PCNA is also SUMOylated, mainly at lysine 164, and to a lesser extent, at lysine 127 (Hoege et al. 2002). PCNA is loaded onto DNA by RFC, a 5 subunit complex; it is unloaded by a similar complex in which the Elg1 protein replaces the Rfc1 subunit (Kubota et al. 2013; Parnas et al. 2010; Sau and Kupiec 2020; Shemesh et al. 2017). In the absence of Elg1, a higher level of PCNA, and in particular, of SUMOylated PCNA, can be detected in chromatin fractions (Kubota et al. 2013; Parnas et al. 2010). A protein with high affinity to SUMOylated PCNA is Srs2, a UvrD-like helicase that is able to disrupt Rad51 presynaptic filaments, thus preventing homologous recombination (HR) (Bronstein et al. 2018a, b; Krejci et al. 2003; Piazza and Heyer 2019; Veaute et al. 2003).

As an alternative to TLS and TS, a third mechanism, hereafter referred as ‘‘salvage recombination’’ (SR) also exists. This pathway is independent of PCNA ubiquitination, employs some proteins of the homologous recombination pathway (which usually deals with double-stranded DNA breaks) (Bordelet and Dubrana 2019; Marsella et al. 2019; Zimmer and Fabre 2019), and appears to be negatively regulated by the Srs2 helicase (Branzei and Szakal 2016; Urulangodi et al. 2015).

Recently, by performing quantitative serial dilution assays on plates containing the DNA-damaging agent MMS (methylmethane sulfonate) at concentrations differing by small increments, we have dissected the mechanisms that regulate SR (Arbel et al. 2020). A first surprising and paradoxical observation was that yeast strains in which lysine 164 of PCNA was mutated (and thus this residue can neither be ubiquitinated nor SUMOylated) was more sensitive to DNA-damaging agents than isogenic strains that carried an additional mutation at lysine 127 (pol30-KK127,164RR, hereafter referred to as pol30-RR). This result suggested that a modification of PCNA at lysine 127 inhibits repair or tolerance when K164 is mutated. Mutating the SUMO E3 ligase SIZ1 in a pol30-K164R strain led to the same phenotype as that of the pol30-RR strain, indicating that indeed the effect observed is due to SUMOylation of lysine 127. This modification could serve to recruit Srs2. Indeed, strains deleted for SRS2 have the same lower sensitivity as pol30-RR strains, and so does the double mutant pol30-RR srs2Δ strain. These results thus confirm that in addition of binding PCNA SUMOylated at K164 (Hoege et al. 2002), Srs2 can also bind molecules SUMOylated at K127, and exert the same negative regulation of the SR repair mode.

To better understand the nature of this inhibition, we attempted to bypass it by overexpressing, in pol30-K164R strains, proteins from the homologous recombination pathway. Overexpression of Rad52, but not of Rad51, Rad55, rad57, or Rad59, restored resistance to MMS. The lack of effect of Rad51 overexpression was surprising, as Srs2 is known to be able to evict Rad51 from chromatin (Krejci et al. 2003; Veaute et al. 2003) and Rad51 is the main RecA-like strand transfer protein of eukaryotes. Moreover, simultaneous overexpression of Rad51 and Rad52 did not increase resistance beyond that observed in cells overexpressing Rad52. These results indicate that the effect of Srs2 inhibition is through Rad52, and not through Rad51, in accordance with some current models (De Tullio et al. 2017). Overexpression of the Elg1 PCNA unloader also resulted in a suppression of the MMS sensitivity, and no further suppression was obtained by overexpressing it together with Rad52, suggesting that the two proteins work in a single process. Moreover, the effects seen were the same as those seen in srs2Δ or siz1Δ mutants; Rad52 or Elg1 overexpression in these strains had no further increase in MMS sensitivity. We thus concluded that the overexpression allowed the bypass of the negative effect created by the recruitment of Srs2 to SUMOylated lysine 127 in PCNA.

To better understand the overexpression mechanism, we knocked out several genes and asked whether they affect the suppression effect. Our results showed that suppression by overexpression of Rad52 required the activity of Elg1 and vice versa, confirming the common activity. Surprisingly, Rad51 was essential for the suppression. Thus, although not limiting, Rad51 does play a role in the bypass of Srs2 effect. Similarly, Rad59, a protein that works with Rad52 (Bai and Symington 1996; Jablonovich et al. 1999), and the Sgs1 helicase [a RecQ-like protein that can dissolve Holliday Junctions (Chu and Hickson 2009)] were necessary for the suppression. Taken together, our results suggest a model in which the SR pathway involves PCNA unloading by Elg1, followed by invasion of the sister chromatid by a mechanism mediated by the Rad52-Rad59 complex. Intermediates thus created are resolved by the activity of the Sgs1 helicase. The suppression observed by overexpressing Rad52 or Elg1 was independent of PCNA ubiquitination by Rad6/Rad18, and, consistent with lysine 164 being the main target for SUMOylation, recruitment of Srs2 to this lysine had a stronger effect in suppressing the SR pathway.

Further analysis demonstrated physical interactions between Srs2 and Rad52. The interactions take place through the same region of Srs2 (amino acids 875–902), which is known to bind Rad51. Thus, our results define Srs2 as a negative regulator of Rad52/59, in addition to its known role in eviction of Rad51 (Krejci et al. 2003; Veaute et al. 2003).

What is the relationship between Srs2 and Elg1? Overexpression of Rad52 overcomes the inhibition by Srs2 because Rad52 is a direct target of Srs2. However, we were unable to detect direct physical interactions between Srs2 and Elg1. Overexpression of Elg1 may remove the inhibition of Srs2 either by a direct competition for binding of SUMOylated PCNA, or because increased PCNA unloading by Elg1 removes Srs2 from the chromatin. The double mutant srs2Δ elg1Δ grows poorly and exhibits higher sensitivity to DNA-damaging agents (Gazy et al. 2013). In the absence of Srs2, the SR pathway should open. However, Elg1 is needed to unload PCNA and allow repair and resumption of DNA replication. Indeed, a triple mutant pol30-K164R srs2Δ elg1Δ has the same sensitivity to DNA-damaging agents as pol30-RR elg1Δ strains. Significantly, mutants still able to recruit Srs2 but for which the SR pathway is blocked (e.g., pol30-K164R elg1Δ) are more sensitive, showing that some DNA repair can be performed in the absence of Elg1 activity, provided Srs2 is not recruited. This is in line with the fact that Elg1 is not an essential protein, and therefore, alternative mechanisms for unloading PCNA must exist.

Our results can be summarized in the following model (Fig. 1): wherever DNA polymerases stall, PCNA becomes SUMOylated, mainly at lysine 164, and, at lower levels, at lysine 127. Modified PCNA may be found at the arrested fork itself, or, if left behind by re-initiation, next to RPA-covered ssDNA gaps. It is possible that these two locations are treated differently, as recent work has shown differences in DNA damage checkpoint signaling between them (Garcia-Rodriguez et al. 2018). Recruitment of Srs2 to SUMOylated PCNA inhibits the activity of Rad52, Rad51 and Elg1, thus precluding unwanted or untimely recombination events (Fig. 1b). The high levels of RPA in proximity to SUMOylated PCNA allow it to become mono-ubiquitinated by Rad6 and Rad18 (Li et al. 2019), or further poly-ubiquitylated by the Rad5/Mms2/Ubc13 proteins to allow TS. How the decision is made regarding whether to stop at a single ubiquitin to allow the error-prone TLS, or to continue into the error-free pathway that requires poly-ubiquitination of PCNA is still an unsolved question.

Fig. 1
figure 1

Potential model for the recombination salvage pathway regulation. a During DNA replication, fork movement can be impaired by lesions or other perturbations (red star) causing stalling. Either at the arrested fork, or at gaps left behind after re-initiation downstream, the ssDNA uncovered gets covered by RPA and Rad51. b During normal DNA replication, binding of Srs2 to the SUMOylated lysines of PCNA inhibits the activity of Rad52 and evicts Rad51, thus preventing potential homologous recombination events. Ubiquitination of PCNA allows damage bypass by trans-lesion synthesis (TLS) or by template switch (TS). Which of these two sub-pathways is chosen may depend on the amount and perhaps the nature of the DNA damage, or by its location (at the fork or behind it, at ssDNA gaps. c Mutation of lysine 164 of PCNA does not impair the recruitment of Srs2 by K127 SUMOylation. Since PCNA cannot undergo ubiquitination, and Srs2 inhibits the activity of both Rad52 and Rad51, the TLS and TS pathways are inactivated, and the cells become extremely sensitive to DNA damage

Full size image

Cells mutated for lysine 164 (Fig. 1c) are extremely sensitive to DNA damage, because recruitment of Srs2 at lysine 127 precludes the use of the SR pathway. If Srs2 cannot be recruited (either because of mutations in the two lysines of PCNA (pol30-RR), in the SUMOylation machinery (siz1 siz2), or upon deletion of SRS2), however, the SR option opens, and Elg1 is brought to unload PCNA, allowing invasion of the sister chromatid in a process that requires Rad52, Rad51, Rad59 and Sgs1.

It is clear that the role of Srs2 is far from been completely understood and that we are looking through a keyhole, each time gazing at a different part of its activity but still lacking the ability to see the big picture. Arbel et al. used a minimalistic approach in trying to dissect Srs2's regulation of the SR; additional layers or regulation probably exist. Moreover, different types of impediments to the replication fork’s movement may be treated differently. From the many enigmas surrounding Srs2, maybe the most interesting one, is when, where and why (and even if) the SR pathway is activated in WT cells. An appealing answer to this question may be found perhaps in Srs2's role in the DNA damage checkpoint (Berto et al. 2019; Ma et al. 2020). Srs2 is phosphorylated in a Mec1(ATR)-dependent manner, which controls Srs2 turnover (Liberi et al. 2000; Saponaro et al. 2010). Elg1 also undergoes a Mec1(ATR)-dependent phosphorylation (Sau and Kupiec 2020; Shkedy et al. 2015). Thus, one of the functions of the DNA damage response may be to coordinate the activity of the various DNA tolerance pathways. Interestingly, the activity of Srs2 itself is also necessary for the resumption of growth after cells were arrested by the DDR (Vaze et al. 2002). Thus, Srs2 could be involved in a regulatory feedback loop, in which its own activity, regulated by the DDR, in turn inactivates it.

The results obtained by Arbel et al. (2020) shed light on the complex interactions between various players in cells experiencing DNA damage. They also underscore the central role played by protein modifications, such as ubiquitin and SUMO, in orchestrating complex decisions in the cell.