Damage-specific modification of PCNA

Abstract

Okazaki fragment processing is an integral part of DNA replication.  For a long time, we assumed that the maturation of these small RNA-primed DNA fragments did not necessarily have to occur during S phase, but could be postponed to late in S phase after the bulk of DNA synthesis had been completed.  This view was primarily based on the arrest phenotype of temperature-sensitive DNA ligase I mutants in yeast, which accumulated with an almost fully duplicated set of chromosomes.  However, many temperature-sensitive alleles can be leaky, and the re-evaluation of DNA ligase I-deficient cells has offered new and unexpected insights into how cells keep track of lagging strand synthesis.  It turns out that if Okazaki fragment joining goes awry, cells have their own alarm system in the form of ubiquitin that is conjugated to the replication clamp PCNA.  Although this modification results in mono- and poly-ubiquitination of PCNA, it is genetically distinct from the known post-replicative repair mark at lysine 164.  In this Extra View, we discuss the possibility that eukaryotic cells utilize different enzymatic pathways and ubiquitin attachment sites on PCNA to alert the replication machinery to the accumulation of single-stranded gaps or nicks behind the fork.

Introduction

In humans, more than 30,000,000 Okazaki fragments (OFs) have to be initiated (assuming an average size of ∼200 nt/OF), processed and ligated during a single round of DNA replication.1 It is easy to imagine that failure to properly join OFs would cause a large number of single-strand breaks (SSBs)—at a frequency that would likely overwhelm the SSB repair machinery and thus be lethal. It is therefore not surprising that to date only one individual with a severe defect in DNA ligase I, the enzyme that joins OFs during DNA replication, has been identified and described.2,3 The individual carried compound heterozygous mutations in the gene and retained approximately 5% ligase activity.4,5 Importantly, this reduction caused growth retardation, immunodeficiency, UV sensitivity and lymphoma.2,3 Most of these phenotypes are compatible with the known function of DNA ligase I in DNA replication and nucleotide excision repair.6 The only exception is the immunodeficiency, which went hand-in-hand with a profound lack of immunoglobulin class switching.2,3 Although this latter defect was not recapitulated in a mouse model carrying a homozygous mutation of the viable missense allele, DNA ligase I-deficient animals displayed growth retardation, abnormal erythropoiesis and increased genome instability.7,8 Moreover, they developed an unexpectedly wide range of spontaneous tumors, such as lymphomas, adeno- and other carcinomas. The importance of proper OF maturation is further underscored by the fact that mutations in other enzymes that participate in OF processing, such as the flap endonuclease Fen1 (e.g., a E160D mutation), which retains partial endonuclease activity,9 have been directly linked to cancer in humans.10 Interestingly, recent evidence suggests that the E160D Fen1 mutation, similar to the DNA ligase I deficiency, causes early-onset lymphoma in mice.11 This raises the question of whether cells harbor a system capable of monitoring perturbations during OF processing. In addition, it remains unclear if abnormalities in lagging strand maturation affect replication fork progression. This is exactly what we set out to explore in a recent study that was published earlier this year.12

Post-Replicative Repair and DNA Damage Induction by Okazaki Fragment Processing Defects

Today we have a relatively good understanding of the post-replication repair (PRR) pathways involved in sensing DNA lesions, like ultraviolet (UV) irradiation-induced thymine dimers,13 in the leading or lagging strand template.14–16 PRR seemed for many years somewhat of a misnomer, as most studies analyzed the pathway in the context of an active replication fork.17–21 However, recent studies have come full circle demonstrating that PRR can occur outside of S phase, after replication has been largely completed.22,23 The first description of PRR dates back to 1968 when Rupp and Howard-Flanders described discontinuities in the newly synthesized DNA of nucleotide excision repair (NER) defective E. coli.24 Shortly thereafter, the concept was extended to mammalian cells.25 During the past 42 years much progress has been made toward understanding error-prone and error-free PRR, respectively.26 Nevertheless, most studies published to date utilize UV or other mutagens like methyl methanesulfonate (MMS) to create large adducts in the DNA that prompt collisions with the replicative polymerases, thereby causing transient replication fork arrest.21,27–29 This is usually accompanied by formation of long stretches of replication protein A (RPA)-coated, single-stranded (ss) DNA.30 These RPA-coated regions facilitate PRR activation by ubiquitination of PCNA at lysine 164,18,31 and also trigger the S phase checkpoint.30 Both events are genetically separable and are thought to occur in parallel.20,32 Although the abovementioned experimental approach utilizing UV-irradiation or MMS has provided a powerful model to dissect the molecular steps of these pathways, it has obvious limitations, because forks that accumulate ssDNA are probably not the only structures that arise in replication-defective mutants. Aberrant Okazaki fragment processing, for example, can cause nicks—in a wide range of varieties: “clean” nicks due to ligation deficiency, “dirty” nicks that exhibit adenylated 5′ ends if the ligation reaction is aborted prematurely or “flapped” nicks in the case of improper removal of RNA primers.33,34 This latter reaction requires the complicated interplay of multiple activities. Limited displacement synthesis by DNA polymerase (pol)-δ is required to initially create a short flap that is deleted one nucleotide at a time by Fen1.35 Excessive strand displacement is counteracted by the intrinsic 3′-exonuclease activity of the catalytic subunit of pol-δ and its non-essential binding partner, Pol32.36,37 Cells defective in Fen1, appear to create larger flaps (in a Pol32-dependent manner) that can be processed by a second endonuclease, Dna2.38–40 Recent studies have also implicated other enzymes, such as exonuclease 1 and the helicase Pif1 in this backup system.37,41 Thus, the first two steps (strand displacement and flap removal) of OF processing (OFP) already involve an elaborate network of partially redundant pathways to assure proper production of a ligatable nick that can be sealed by DNA ligase I. Importantly, primer removal by Fen1 and OF ligation by DNA ligase I are coordinated by PCNA.42 Both enzymes bind PCNA through a PCNA interacting peptide (PIP) box.43–45

Okazaki Fragment Processing Defects Trigger PCNA Ubiquitination

The fact that multiple OFP pathways exist suggests that cells have the ability to actively monitor maturation events during lagging strand synthesis. To address this question, we analyzed DNA ligase I-deficient cells.46 In budding yeast, temperature-sensitive cdc9 mutants were among the first strains identified to have a cell cycle progression defect under non-permissive conditions as they arrested in late S/G₂ phase.47–49 The arrest was dependent on the DNA damage response50 and cell viability required functional homologous recombination,51 indicating that the nicks might be—at least partially—converted into double strand breaks (DSBs). These early studies suggested that OF ligation was uncoupled from replication fork progression and could occur very late in S phase after the bulk of the DNA had been replicated. This was also consistent with observations made much later that PCNA remained on chromatin after passage of the replication fork, providing anchor points for DNA ligase I on chromatin.52 In fact, experiments with cdc9-1 cells that had two consecutive shifts, to the non-permissive temperature first and then back to the permissive temperature demonstrated that OF synthesis could be temporally separated from OF ligation.46 All of these findings were consistent with the notion that DNA ligase I was not required until the end of S phase. However, our recent analysis of the cdc9-1 mutation in a different strain background suggested that the cells do not move through S phase with normal kinetics, but rather appeared to accumulate in mid-S phase.12 Further exploration of a temperature-sensitive degron mutant (cdc9-td) verified that DNA replication was extensively delayed in the absence of DNA ligase I (Fig. 1). Although it still remains unclear how this delay was mediated (e.g., through reduced fork progression or inhibition of late-firing origins), it was dependent on the activation of the S phase checkpoint kinase Rad53.53 Rad53 phosphorylation was not only Rad9- (and thus DNA damage) dependent,54 but also required the mediator of the replication checkpoint, Mrc1,55,56 the yeast homolog of Claspin.57 This finding implied that replication fork stalling was occuring.58,59 How this happened in a scenario in which none of the DNA polymerases were actively inhibited was puzzling. Even more surprising was the observation that the phosphorylation of Rad53 was amplified by a novel PRR-related pathway that triggered mono- and poly-ubiquitination of PCNA.12 Mutation of lysine 164 of PCNA in cdc9 mutants had no effect on cell viability, arguing that neither error-prone translesion synthesis19,21,60–62 nor error-free replication by template switch63–66 played a role in this process. Instead, we observed synthetic lethality between cdc9-1 and a mutation in lysine 107 of PCNA, leading us to propose that ubiquitin is conjugated to lysine 107 in DNA ligase I-deficient cells. One aspect that we did not address in our recent study is whether PCNA is also modified by the ubiquitin-related SUMO peptide, which is attached to PCNA during S phase at lysine 164 and to a lesser extent at lysine 127.18 In Figure 2, we demonstrate that PCNA is indeed sumoylated in wild-type cells and cdc9 mutants. Because the mutants accumulated in S phase, they displayed stronger sumoylation signals. In addition, we observed higher molecular weight forms of sumoylated PCNA. This is consistent with the finding that PCNA carries SUMO chains in the face of profound replication stress.67 Alternatively, it is also possible that a single PCNA monomer is ubiquitinated (at lysine 107) and sumoylated (at lysine 164) in cdc9 mutants. Sumoylation of PCNA is thought to repress homologous recombination68–70 at stalled replication forks. This raises an interesting point because homologous recombination is required for the viability of cdc9 mutants and therefore its activity needs to be highly regulated. Future experiments are needed to clarify these issues. However, regardless of their outcome, it is important to mention that ubiquitination of PCNA is conserved in human cells that are depleted for DNA ligase I, even though lysine 107 of PCNA is not evolutionarily conserved.12 This underscores the significance of our findings.

Consistent with the observation that a DNA ligase I deficiency triggers ubiquitin conjugation at a non-canonical lysine residue of PCNA in budding yeast, we have also uncovered genetic requirements for this process that clearly distinguish it from the classical DNA damage tolerance (DDT) pathways. For example, whereas DDT is dependent on Rad6 and Rad18, both proteins are dispensable for PCNA ubiquitination in cdc9 mutants. This makes sense in as far as Rad6 and Rad18 mediate predominantly translesion synthesis,22 which would help little in facilitating the repair of nicks. Moreover, the poly-ubiquitin chains observed in cdc9 mutants are linked through lysine 29 and not lysine 63 as in PRR.18 In the light of this result, it is not surprising that Ubc13, which together with Mms2 catalyzes lysine 63 linked ubiquitin chains,71–73 did not appear to play a role in this pathway.12 What was puzzling, however, was the genetic requirement for MMS2 and UBC4. The human homolog of Ubc4, UbcH5A, synthesizes lysine 29 linked ubiquitin chains in vitro.74 Although Ubc13 has been reported to function independently of Mms2,75 no other studies have demonstrated a genetic function for Mms2 that is separable from Ubc13. How Mms2 fits into the PCNA ubiquitination pathway is not at all clear and will require additional genetic validation. Furthermore, biochemical studies are currently underway to dissect whether Mms2, Ubc4 and Rad5 cooperate in one physical complex—possibly together with other proteins—or not. At this point, we have no evidence to suggest a direct interaction between Mms2 and Ubc4. Moreover, we would like to emphasize that our data does not dispute nor contradict any of the established rules for the DDT and PRR pathways.76 Rather, we interpret them to mean that both the repertoires of damaged DNA structures that cause PCNA ubiquitination as well as the pathways catalyzing these reactions are likely larger and more complex than anticipated. This is also underscored by recent observations that described the Rad6-independent ubiquitination of PCNA in response to chronic HU and MMS treatment at lysine 164.77 Although we are still missing many pieces of the puzzle, it appears that different types of DNA damage may trigger different PCNA ubiquitination pathways that result in distinct attachment sites and/or differently linked ubiquitin chains (i.e., a “DNA damage code”; Fig. 3).12

Future Challenges

Taken together, our results suggest that lagging strand synthesis is actively monitored and that persistent nicks in newly replicated DNA (and maybe also those that arise during abortive repair or result from oxidative damage) can indeed be detected during S phase. Whether earlier steps in Okazaki fragment processing are surveyed in a similar manner remains obscure. Although we tested rad27 and dna2 mutants and neither displayed PCNA ubiquitination under non-permissive conditions,12 it is highly likely that the respective enzymes substitute for one another in these particular mutants. Therefore, the lack of PCNA ubiquitination does not necessarily exclude the possibility that other OFP intermediates are recognized by the cell and trigger modification(s) of the PCNA clamp. Indeed, pol32 mutants cause lagging strand defects, which result in ss gaps.23 Although it is not clear that these gaps are generated during OFP (e.g., they might occur during actual synthesis), they cause PCNA ubiquitination. The most burning question for us concerns the biological function of ubiquitinated PCNA in cdc9 mutants. What is the connection between the modified clamp, the slow replication dynamics and Rad53 activation? Replication fork slowdown would certainly aid in extending the time frame of an individual ligation reaction. In the same vein, modified PCNA may help to retain the enzyme on site until Okazaki fragments have been successfully joined and might counteract abortive ligation reactions due to the limited amounts of DNA ligase I present in the cell. The exciting new technical developments in producing ubiquitinated PCNA in vitro78,79 certainly provide invaluable tools toward finding answers to these questions.

Abbreviations

PCNA	=	proliferating cell nuclear antigen
OF	=	Okazaki fragment
UV	=	ultraviolet
NER	=	nucleotide excision repair
MMS	=	methyl methane sulfonate
RPA	=	replication protein A
OFP	=	Okazaki fragment processing
PIP	=	PCNA interacting protein
DSBs	=	double strand breaks
DDT	=	DNA damage tolerance
PRR	=	post-replicative repair
SSBs	=	single strand breaks
pol	=	polymerase
ss	=	single strand
SUMO	=	small ubiquitin-like modifier

Figures and Tables

Figure 1 DNA ligase I is required for S phase progression. Asynchronous cultures of ABy010 (GAL-UBR1 CDC9) and ABy008 (GAL-UBR1 cdc9-td) were grown at 28°C and arrested in G₁ with α-factor. After 2 hours, Ubr1 expression was induced with 2% galactose for 30 min. Subsequently, cultures were shifted to 37°C in medium containing α-factor and 2% galactose. After 90 min, cells were released from G₁ for 12 hours. DNA content at indicated time was monitored by flow cytometry and the vertical line indicates a 2C DNA content. The black arrows mark entry into S phase and completion of DNA replication in cdc9-td mutants, respectively.

Figure 2 PCNA is sumoylated in cdc9 mutants. Wild-type (CDC9) and cdc9 mutants (cdc9-1 and cdc9-2) expressing PCNA-3HA were grown asynchronously at 25°C and shifted to 35°C for 3 hours. Whole-cell extracts were immunoprecipitated with anti-HA conjugated agarose beads (Sigma). Both unmodified and modified PCNA was detected with an anti-HA antibody (Covance, 16B12). Sumoylated PCNA was detected with an anti-SUMO antibody (gift from Xiaolan Zhao, Memorial Sloan-Kettering Cancer Center). The black dots mark prominent sumoylated PCNA bands. IgG_H indicates the immunoglobulin heavy chain. Mono-ubiquitinated PCNA is visible in darker exposures around 42kDa in cdc9 mutants with HA-tagged PCNA (only visible as a faint band in this exposure). cdc9-1 is a more stringent allele than cdc9-2,12 and thus sumoylation of PCNA is more prominent in this mutant.

Figure 3 Damage-dependent PCNA ubiquitination in budding yeast. DNA synthesis occurs in a semi-discontinuous manner. Nascent DNA on the leading strand is synthesized by pol ε, whereas pol δ is required to synthesize Okazaki fragments (short arrows) on the lagging strand. In response to UV light-induced damage, which causes pyrimidine dimers on the template DNA (red X, left), PCNA is ubiquitinated at lysine (K) 164 and the poly-ubiquitin chain is linked through K63. Activation of the S phase checkpoint by UV damage is independent of PCNA ubiquitination at K164. In contrast, PCNA is ubiquitinated at K107 and the poly-ubiquitin chain is linked through K29 in response to DNA ligase I deficiency, which accumulates persistent nicks due to unligated Okazaki fragments (red arrows, right). PCNA ubiquitination at K107 appears to be a prerequisite for activation of the S phase checkpoint, illustrated by hyperphosphorylation of Rad53. The importance of the K29-linked poly-ubiquitin chain and how PCNA ubiquitination results in activation of the S phase checkpoint is still unknown. The S. cerevisiae homotrimeric PCNA structure (PDB ID 2OD8) was generated using Chimera software program.80 Ub denotes ubiquitin. K164 and K107 on each of the subunits are indicated as yellow and red surfaces, respectively.

Acknowledgements

We thank Dr. Xiaolan Zhao for the anti-SUMO antibody and Dr. Eric A. Hendrickson for reading the manuscript. We acknowledge the assistance of the Flow Cytometry Core Facility at the University of Minnesota. Molecular graphics images were produced using the UCSF Chimera package from the Resource for Biocomputing, Visualization and Informatics at the University of California, San Francisco (supported by NIH P41 RR-01081). This work was supported by NIH grant GM074917 to A.K.B., who is a Scholar of the Leukemia and Lymphoma Society.