|Home | About | Journals | Submit | Contact Us | Français|
In Escherichia coli, cell survival and genomic stability after UV radiation depends on repair mechanisms induced as part of the SOS response to DNA damage. The early phase of the SOS response is mostly dominated by accurate DNA repair, while the later phase is characterized with elevated mutation levels caused by error-prone DNA replication. SOS mutagenesis is largely the result of the action of DNA polymerase V (pol V), which has the ability to insert nucleotides opposite various DNA lesions in a process termed translesion DNA synthesis (TLS). Pol V is a low-fidelity polymerase that is composed of UmuD′2C and is encoded by the umuDC operon. Pol V is strictly regulated in the cell so as to avoid genomic mutation overload. RecA nucleoprotein filaments (RecA*), formed by RecA binding to single-stranded DNA with ATP, are essential for pol V-catalyzed TLS both in vivo and in vitro. This review focuses on recent studies addressing the protein composition of active DNA polymerase V, and the role of RecA protein in activating this enzyme. Based on unforeseen properties of RecA*, we describe a new model for pol V-catalyzed SOS-induced mutagenesis.
More than 50 years ago, Weigle made the pivotal discovery that UV irradiated λ phage could be rescued by infection into UV-irradiated E. coli (Weigle, 1953). Fourteen years later, Witkin proposed that this phenomenon was evidence of a damage induced DNA repair system in bacterial cells (Witkin, 1967). An increase in the number of point mutations in the phage was also characteristic of this rescue. Radman then suggested the presence of a “mutation-prone” replication mechanism involving the lexA and recA gene functions that can account for UV-induced mutations of λ phage and E. coli, which he named “SOS repair” (Radman, 1974), and the accompanying mutations after DNA damage came to be known as “SOS mutagenesis”.
The SOS response is triggered by DNA damage caused by exposure to UV-irradiation or to chemicals (Witkin, 1976, Walker, 1984). The DNA damage leads to a halt in DNA replication as replication forks encounter unrepaired lesions or lesions undergoing repair (Setlow et al., 1963, Friedberg et al., 2006). The primary task of the SOS response is to restart replication before the cells die. This system is regulated by both the LexA transcriptional repressor and the RecA recombinase (Figure 1).
There are more than 40 genes known to be involved in SOS repair (Fernandez de Henestrosa et al., 2000, Courcelle et al., 2001). Most of these genes are induced rapidly and are involved in error-free DNA repair, including base excision repair (BER), nucleotide excision repair (NER), and recombinational DNA repair (Friedberg et al., 2006). If the DNA damage levels are so great that error-free pathways are insufficient to complete repair and restart replication, the mutagenic phase of SOS is triggered (Walker, 1984, Echols and Goodman, 1990). This phase of the SOS response is mediated by DNA polymerases that replicate past template lesions, in a process called translesion DNA synthesis, or TLS (Goodman, 2002). TLS is inherently inaccurate, and is responsible for the large increase in mutations originally observed by Weigle.
There are three SOS-induced DNA polymerases in E. coli, pols II, IV and V, and all are engaged in various aspects of TLS (Goodman, 2002). It is pol V, encoded by the umuDC operon, that is largely responsible for the ~100-fold increase in DNA damage-induced chromosomal mutations (Kato and Shinoura, 1977, Steinborn, 1978, Woodgate, 1992). SOS mutagenesis can be considered a kind of desperation response. The mutations that occur may kill many cells. However, replication is successfully restarted, and the “lucky” cells survive.
Certain mutations in the LexA repressor, the RecA recombinase, and the UmuDC proteins abolish SOS mutagenesis. Since LexA and RecA are regulators of the SOS response, it is understandable that mutations in these genes might render the cell non-mutable through their inability to induce the SOS response. The umu genes prove to be more interesting in that they do not affect other aspects of the SOS response and are thus uniquely involved only in SOS mutagenesis. Another intricacy in the system is that the RecA protein has at least three genetically separable roles in SOS mutagenesis. These include mediating self-cleavage of the LexA repressor leading to the subsequent derepression of all LexA-regulated genes, mediating self cleavage of UmuD to a shorter and mutagenically active form called UmuD′ (Shinagawa et al., 1988, Burckhardt et al., 1988), and a direct role in SOS mutagenesis (Nohmi et al., 1988, Dutreix et al., 1989, Sweasy et al., 1990).
The UmuC protein and dimeric UmuD′2 interact (Woodgate et al., 1989, Bruck et al., 1996) to form error prone DNA polymerase V (pol V) (Tang et al., 1999, Reuven et al., 1999). DNA polymerase V inserts nucleotides opposite a variety of DNA lesions to facilitate TLS (Tang et al., 2000). Since pol V is a low-fidelity polymerase it quite often promiscuously inserts incorrect bases opposite of the lesion, thereby explaining the observed increase in damage-induced SOS mutagenesis in vivo (Tang et al., 2000).
An active RecA nucleoprotein filament, RecA* is required for activation of pol V-catalyzed TLS both in vivo (Dutreix et al., 1989, Sweasy et al., 1990) and in vitro (Tang et al., 1999, Tang et al., 1998, Reuven et al., 1999, Fujii et al., 2004, Schlacher et al., 2006a, Jiang et al., 2009). RecA* is formed in the presence of ATP by RecA binding to single-stranded DNA gaps. Such gaps are formed at stalled replication forks, at sites where lesions are simply bypassed (Heller and Marians, 2006, Marians, 2008), and at sites of intensive DNA repair throughout the genome. As detailed below, the filaments need not be in the same genomic location as the site where pol V action is required (Schlacher et al., 2006a, Jiang et al., 2009).
Irrespective of their location in the cell, these active RecA* filaments are essential for pol V-catalyzed TLS. The exact role for this requirement has only recently been discovered (Jiang et al., 2009). The term RecA* has historically been associated with RecA filaments that were active in the promotion of SOS induction and SOS mutagenesis. In this article, we associate the term with active RecA filaments bound to DNA in the presence of ATP or an ATP analogue, filaments that are active for the complete range of activities normally associated with RecA.
This review centers on our understanding of three proteins: RecA, UmuC, and UmuD′, and their actions and roles in SOS mutagenesis. We address two longstanding, enigmatic aspects of SOS mutagenesis: the molecular composition of mutagenically active pol V and the precise role of RecA* in activating pol V for DNA synthesis. Based on the properties of RecA* during pol V activation, we describe a new model to account for the biochemical basis of SOS-induced mutagenesis (Jiang et al., 2009).
Proteins homologous to RecA in structure and function are ubiquitous in all classes of life. The archaeal homologue is called RadA (Seitz et al., 1998, Wu et al., 2004, Yang et al., 2001). In eukaryotes, there are two RecA homologues, the Rad51 and Dmc1 proteins (Bishop et al., 1992, Shinohara et al., 1992, Sung, 1994). The recA gene was initially identified and characterized genetically in a screen for recombination deficient mutants of E. coli (Clark and Margulies, 1965). It was quickly found to have multiple roles in recombination and repair; for a review see (Clark and Sandler, 1994).
The RecA protein is a 352 amino acid polypeptide with a molecular weight of 38 kDa. In uninduced cells, it is present at less than 10,000 monomers per cell. Upon SOS induction the level of RecA can increase to over 70,000 monomers per cell (Sommer et al., 1998). The bacterial RecA protein is a DNA dependent ATPase. The RecA of Escherichia coli hydrolyzes ATP at a rate of 20–30/min depending on the nature of the bound DNA (Weinstock et al., 1981, Pugh and Cox, 1987). RecA binds to DNA as a nucleoprotein (np) filament that forms most rapidly onto single-stranded DNA.
Filaments are formed in two steps, with a slow nucleation followed by a more rapid, cooperative extension of the filament in the 5′ to 3′ direction (Register and Griffith, 1985, Shan et al., 1997, Kuzminov, 1999) (Figure 2); however, there is recent evidence for slower filament growth in the opposite direction on DNA at a single molecule level (Galletto et al., 2006, Joo et al., 2006). Five monomers of RecA are sufficient for nucleation (Joo et al., 2006). Filament formation requires ATP binding, but not ATP hydrolysis. When ATP is hydrolysed, it causes filament disassembly, also in the 5′ to 3′ direction (Shan et al., 1997, Arenson et al., 1999, Bork et al., 2001). The footprint of RecA on DNA is one monomer of RecA per three nucleotides of DNA. The bound single stranded DNA is extended by 50% and underwound so that there are 18 base pairs, or 6 RecA monomers, per helical turn (Egelman and Stasiak, 1986, Yu and Egelman, 1992, Stasiak and Egelman, 1994). There are three binding sites for DNA strands within the np-filament and these sites facilitate the strand exchange reaction (Takahashi et al., 1991, Takahashi and Norden, 1995, Kubista et al., 1996, Kurumizaka and Shibata, 1996).
RecA has four known functions in the E. coli cell: catalysing the DNA strand exchange reaction in the context of recombinational DNA repair, induction of the SOS response by promotion of the autocatalytic cleavage of the LexA repressor, activation of UmuD′ by mediating autocatalytic cleavage of UmuD, and direct participation in SOS mutagenesis by activation of DNA polymerase V (Schlacher et al., 2006b, Jiang et al., 2009) (Figure 3).
All RecA-class recombinases promote a DNA strand exchange reaction that is central to their function in recombinational DNA repair. Once formed, the active nucleoprotein filament (RecA*) aligns the bound single strand with homologous sequences in a different duplex DNA. It then promotes a strand exchange reaction in which one strand of the duplex is transferred to the bound single strand to create a new duplex (Figure 3A). The second strand from the original duplex is displaced. There are many standard assays to investigate this phenomenon in vitro (Cox, 2003).
The SOS response is normally suppressed by the LexA repressor protein. LexA binds to the operator regions of the genes that are involved in SOS and thus hinders their expression. RecA is present at constitutive levels in the cell in order to function during homologous recombination. In cells exposed to DNA damage, RecA is induced to attain much higher levels. RecA* filaments formed on free ssDNA or, perhaps much more likely at single strand gaps, serve to turn on the SOS response. RecA* promotes this second cellular function – SOS induction – by facilitating the autocatalytic-cleavage of the LexA repressor (Little, 1984) (Figure 3B). In principle, the RecA* filaments are located at any site where a single strand gap appears in the genome. Such gaps are typically associated with DNA lesions, occurring where lesions are being repaired (e.g., small gaps created during DNA excision repair or mismatch repair), where lesions are bypassed during replication (Heller and Marians, 2006, Marians, 2008), or where replication forks are stalled.
There are differential degrees of repression based on the binding affinity of LexA to the operator regions of the genes, which exhibit considerable variability outside a shared consensus sequence (Little and Mount, 1982, Walker, 1984, Lewis et al., 1994, Fernandez de Henestrosa et al., 2000). LexA binds weakly to the recA operator; therefore, RecA is induced “immediately”, < 1 min after exposure to high UV doses (Figure 3B). In vitro, under specific conditions, LexA is able to autocleave in the absence of RecA* (Little, 1984, Little, 1991). However, this is unlikely to occur under physiological conditions and LexA is only cleaved upon contact with RecA* in vivo. LexA binds deep within the helical groove of the nucleoprotein filament and is cleaved into two inactive fragments rapidly upon contact with the filament (Little et al., 1980, Yu and Egelman, 1993). As the amount of LexA protein is reduced, dissociation of LexA from the operator regions leads to induction of the LexA (SOS)-regulated genes (Figure 1). Therefore, the genes with the highest affinity LexA binding sites lose the LexA repressor last and are induced later in the response. Among these genes are umuC and umuD that act in SOS mutagenesis, which are turned on ~ 30 min after exposure to UV (Sommer et al., 1998).
However, despite being expressed, UmuD is inactive for SOS mutagenesis (Nohmi et al., 1988). The protein is activated for its mutagenesis functions after its N-terminal 24 amino acid tail is removed in a RecA-mediated self cleavage reaction that is mechanistically similar to LexA cleavage (Shinagawa et al., 1988, Burckhardt et al., 1988, Nohmi et al., 1988, Paetzel and Woodgate, 2004). UmuD cleavage is much less efficient than LexA cleavage (Burckhardt et al., 1988) and as a consequence, UmuD′ does not accumulate in the cell until some 45 mins post UV-irradiation (Sommer et al., 1998). Removal of the N-terminal tail from UmuD presumably leads to a conformational change in the structure of the protein that allows dimeric UmuD′ to form a tight complex with UmuC (UmuD′2C) (Woodgate et al., 1989, Bruck et al., 1996) (Figure 3C).
Even when recombinant UmuD′2C was fully derepressed in a ΔrecA background, no UV-induced mutagenesis was observed, implying that RecA plays a direct role in the mutagenic process (Nohmi et al., 1988). However, ΔrecA strains are extremely UV-sensitive and the lack of any detectable UV-induced mutagenesis might have simply arisen through the lack of bacteria surviving irradiation. To circumvent these problems, Devoret and colleagues isolated a missense recA mutant, recAS117F (recA1730), that was defective in UV-induced SOS mutagenesis (Dutreix et al., 1989). When expressed in a lexA+ background, the recAS117F allele was defective in most RecA functions. However, overexpression of the mutant protein in a lexA(Def) background helped it regain its recombination activity, yet the strain remained non-mutable upon UV-treatment, even in the presence of recombinant umuD′C genes on a plasmid. These observations confirmed the notion that there is a direct requirement for RecA* in SOS mutagenesis, independent of its role in LexA and UmuD cleavage. The theory was further cemented by strains that overproduced UmuD′C, in which LexA and UmuD processing was not required, but mutagenesis was still distinctly dependent on the specific recA allele present (Sweasy et al., 1990).
This “third role” of RecA in SOS mutagenesis remained enigmatic for many years. However, it has recently been shown that RecA is required to directly activate the UmuD′2C polymerase, by transferring a molecule of RecA and ATP to form an activated pol V mutasome, UmuD′2C-RecA-ATP (Jiang et al., 2009) (Figure 3D). The elucidation of the direct role of RecA in SOS mutagenesis will comprise the take home message from this review.
The umuDC genes were identified as encoding key components in SOS mutagenesis in the late 1970s (Kato and Shinoura, 1977, Steinborn, 1978, Steinborn, 1979). In a direct genetic screen for SOS mutagenesis, they were the only genes identified that solely affected mutagenesis and not other aspects of SOS repair (Kato and Shinoura, 1977).
The umuDC genes are regulated at the transcriptional level by LexA and at the post-translational level via RecA-mediated cleavage of UmuD to mutagenically active UmuD′. In addition, intracellular levels of the Umu proteins are kept to a minimum through regulated proteolysis. Both UmuC and UmuD are rapidly degraded by the Lon protease (Frank et al., 1996, Gonzalez et al., 1998). UmuD′ is largely resistant to the actions of Lon, but is, instead, rapidly degraded by the ClpXP protease when in a heterodimeric complex with UmuD (Frank et al., 1996, Gonzalez et al., 2000). In an uninduced cell, it is estimated that there are approximately 200 molecules of UmuD and only 15 molecules of UmuC per cell (Woodgate and Ennis, 1991). Upon DNA damage, steady-state levels of the two proteins increase and UmuD is converted to UmuD′. However, even when maximally expressed some 45 min after DNA damage, there are at most 60 molecules of UmuD′2C in a cell (Sommer et al., 1998). Presumably such regulation is designed to keep the mutagenically active UmuD′2C (pol V) complex to an absolute minimum, such that it is only utilized under dire circumstances.
Early models proposed that these genes were accessory factors required by the replicative polymerase, pol III, to bypass lesion sites in the DNA (Figure 4). These models were based on the observation that temperature sensitive strains carrying a mutation in the gene that encodes the α subunit of pol III had fixed UV-induced mutations when grown at permissive, but not restrictive, temperatures (Bridges and Mottershead, 1976).
Direct biochemical analysis of the Umu proteins was made difficult owing to the insoluble nature of UmuC. For many years, only a minimal amount of UmuC protein was recoverable by traditional means of purification, hampering efforts in several laboratories. The protein formed inclusion bodies and the earliest protocols involved refolding the protein in the presence of chaperones, after application of harsh denaturing conditions (Woodgate et al., 1989, Petit et al., 1994). In vitro reconstitution of TLS was made feasible when a more efficient method of purification was formulated. UmuC was expressed in a soluble form by purifying it as a complex with a dimer of UmuD′ (Bruck et al., 1996). This yielded much larger amounts of the UmuD′2C complex. The new protocol also held the advantage of not requiring treatment by strong denaturants, which can negatively affect the enzymatic activities of any protein. Shortly after reconstitution of the system, it was discovered that UmuD′2C was a bona fide polymerase (Tang et al., 1999, Reuven et al., 1999). DNA synthesis was observed in the presence of UmuD′2C and RecA protein, but in the absence of pol III (Tang et al., 1998, Tang et al., 1999). This new polymerase, termed pol V (Tang et al., 1999, Reuven et al., 1999), was able to synthesize past many different lesions in DNA (Tang et al., 2000).
E. coli pol V is one of the founding members of a new class of polymerases: the Y-family polymerases (Ohmori et al., 2001). The Y-family polymerases can be classified based upon phylogenetic relationships. In addition to E. coli UmuC, the Y family includes polymerases related to E. coli DinB (pol IV/Dpo4/pol κ), and eukaryotic Rev1 and Rad30/pol η and pol ι (Ohmori et al., 2001, Goodman and Tippin, 2000, Goodman, 2002). These polymerases all lack a 3′ → 5′ proofreading exonuclease activity, and also lack the 5′ → 3′ nick translation activity of E. coli DNA polymerase I (Goodman and Tippin, 2000, Goodman, 2002). They are usually weakly processive and have low fidelity. Structurally, they have stubby “finger” and “thumb” domains and their active sites provide less constraints for base pairing (Ling et al., 2001, Silvian et al., 2001, Yang and Woodgate, 2007).
These polymerases have a variety of roles in the cell including the extension of mismatched primer ends in the case of pol IV and involvement in immunoglobulin hypermutation in the case of human pol η (Goodman, 2002). Owing to their low fidelity, when left unchecked, these polymerase activities can have deleterious consequences for the cell. DNA pol V makes base substitution errors with a frequency of 10−2–10−3 on undamaged DNA (Tang et al., 2000). In vitro, it preferentially misincorporates dG opposite the 3′ T of a TT-(6–4) photoproduct (Tang et al., 2000), consistent with the profile of SOS mutations observed in vivo (LeClerc et al., 1991, Szekeres et al., 1996).
Our understanding of the role of RecA* in TLS has gone through many iterations over a period of 25 years. Even in the pre-pol V era, there were several different proposals for the role of RecA* in TLS. Shortly after the discovery of the umuDC genes, a two-step model for TLS was proposed by Bridges and Woodgate, which incorporated pol III, UmuDC, and RecA (Figure 4A). In the first step, pol III would incorporate a nucleotide opposite of the lesion with the help of RecA. This would be followed by lesion bypass by pol III with the help of UmuDC (Bridges and Woodgate, 1985). In 1990, Echols proposed that a multi-protein “mutasome”, including UmuD′C, pol III, and RecA is assembled to bypass a DNA template lesion (Echols and Goodman, 1990) (Figure 4B). Once a small amount of purified UmuC was available, detectable levels of TLS were generated in a reconstituted system in vitro with UmuD′C, pol III, and RecA (Rajagopalan et al., 1992).
These early models incorporated pol III for copying of DNA past the lesion. RecA* was assembled in cis in these models, as a filament on the template ahead of the lesion. This expectation was built on the known properties of the RecA protein, which nucleates the formation of a RecA* filament most readily in single-stranded DNA gaps and was not known to have any function when not part of a nucleoprotein filament. The UmuD′C proteins were predicted to be mutagenic factors to help pol III synthesize past the lesion site.
The discovery that UmuD′2C was an error-prone polymerase (Tang et al., 1999, Reuven et al., 1999), removed pol III from the translesion synthesis models (Figure 4C). The first model for pol V function proposed that pol V carried out TLS in the presence of the β-sliding clamp, but in the absence of pol III (Tang et al., 1998). It was clear that RecA* was still required, and the most plausible location for RecA* was in the single strand gap downstream of the advancing pol V. Thus, this model still relied upon RecA* in cis, on the template containing the lesion as proposed in the earlier efforts with pol III (Echols and Goodman, 1990).
The idea of positioning RecA* in cis, downstream of pol V, requires a pol V capacity not only to interact with the RecA*, but to displace it as it advanced. The proposed activity would require the ejection of RecA* subunits from the end of the RecA filament where disassembly does not normally occur. The capacity of pol V to displace a RecA* filament was called into question when substantial inhibition of pol V movement was demonstrated in the presence of the filaments (Pham et al., 2001). However, some DNA synthesis did occur when the single stranded DNA binding (SSB) protein was included in the reaction. A model was presented in which pol V, in the presence of SSB, acting similar to a cowcatcher on the front of a locomotive, pushed the impeding RecA molecules off the DNA template in front of the oncoming pol V (Pham et al., 2001) (Figure 4D).
The effects of RecA* remained enigmatic, and the activity of pol V was difficult to decipher. In particular, primer utilization by pol V was generally poor, rarely above 20 to 30% when filaments were formed with ATPγS, wherein the filaments could assemble but filament disassembly, which requires ATP hydrolysis, was hindered. Although, the best reactions were observed with RecA levels that approximated optimal filament formation, 1 RecA protein per 3 to 5 DNA nt (Pham et al., 2001, Schlacher et al., 2006b), a significant TLS efficiency was observed even at ~ 1 RecA/50 DNA nt, where it was unlikely that significant filament assembly would occur (Pham et al., 2001, Schlacher et al., 2006b) (Figure 5). Higher concentrations of RecA were highly inhibitory (Pham et al 2002). We viewed these data as a potential “fly in the cis RecA* filament ointment”.
Since high levels of RecA protein clearly impeded pol V-mediated DNA synthesis, and extremely low levels of RecA protein were able to support TLS, the models were developed further. It was proposed that short RecA protomers could take part in forming a minimal mutasome by binding to DNA pol V (Schlacher et al., 2005). The model was supported by the observation that RecA binds directly to DNA pol V in the absence of DNA, and binds by a separate mechanism in the presence of DNA (Schlacher et al., 2005). In all of these efforts, full primer utilization was never achieved by pol V, either in the absence of β, γ complex (Figure 5), or when polymerase accessory proteins were present in the reaction (Pham et al., 2001, Schlacher et al., 2006b). Pol V was instead thought to either be poorly functional or to contain a relatively high level of inactive protein in the pure preparations of UmuD′2C.
The pol V was not inactive. It was simply being impeded by the RecA* that was bound to many (but not all) of the added primer/templates. The experiments above were all carried out in the context of an unrecognized and unappreciated ambiguity. If RecA protein is added to any kind of artificial primer/template along with UmuD′2C, it is impossible to know if RecA* is acting in cis or in trans. Optimization of the polymerase reaction required RecA protein concentrations that were sufficient to bind only some of the template/primers. RecA* assembly occurs as a strongly cooperative binding of RecA to ssDNA (Kuzminov, 1999). At the RecA concentrations that were optimal for pol V activation, filament assembly is heterogeneous. Consequently, some DNA molecules will form RecA*, others not. Were the RecA* filaments promoting pol V-mediated DNA synthesis on the same DNA substrates to which they were bound (in cis), or were they promoting DNA synthesis in trans on other DNA substrates that were not bound by RecA*?
Where could the trans RecA* molecules come from? When forming primer/template molecules by mixing single-stranded DNA primers and templates and then allowing them to anneal in a bimolecular reaction, there will always be un-annealed template and primer DNA (Figure 6). It is thermodynamically unavoidable. Therefore, there were indeterminate amounts of single stranded DNA in all experiments utilizing annealed primer/template substrates. RecA* can, and doubtless does, form on the un-annealed single strands as well as on the annealed primer/templates. There is, however, a simple way to eliminate excess ssDNA, and that is to form a stable hairpin structure so that almost all of the DNA anneals in the form of a hairpin, in a unimolecular reaction, with a very small proportion of non-hairpin ssDNA remaining (Figure 6). By having a short template overhang (3 nt), one also hinders RecA filament assembly on the hairpin primer/template DNA.
A simple alteration of the experimental protocol soon demonstrated that the RecA* was functioning in trans (Schlacher et al., 2006a) (Figure 4E). Full primer utilization was observed in a pol V reaction for the first time when pol V was incubated with hairpin template primer, RecA* filaments were formed on single stranded oligomeric DNA (36-mers) in a separate test tube, and the two were subsequently mixed (Figure 7). The improvements in reaction efficiency were dramatic; a RecA* filament was indeed necessary for TLS by pol V, but the reaction worked best when the RecA* was on a different molecule of DNA than the one pol V was acting on (Schlacher et al., 2006a) (Figures 4E and and7).7). The reaction exhibited second order kinetics, the velocity increasing linearly with the addition of trans RecA* and extrapolating to zero in the absence of trans 36 mer DNA. In addition, pol V is activated by interacting with the 3′-proximal tip of RecA bound to that single-stranded (ss) DNA molecule in trans (Schlacher et al., 2006a). This mechanism began to resolve the paradox that cis RecA filaments obstruct TLS despite the absolute requirement of RecA* for SOS mutagenesis.
What remained un-contemplated is that the role of RecA* might actually be indirect; that it was not interaction with a RecA* filament, be it located either in trans or in cis that activated pol V, but the incorporation of a RecA monomer that was transferred to the UmuD′2C from the RecA* filament. This brings us to the current understanding of the role of RecA* in TLS (Jiang et al., 2009). RecA* transfers one ATP-bound RecA subunit from its 3′-proximal end to UmuD′2C to form an active mutasome with the composition UmuD′2C-RecA-ATP, i.e., pol V Mut (Jiang et al., 2009) (Figure 4F). The active pol V Mut complex synthesizes DNA in the absence of RecA* (Jiang et al., 2009). Echols’ mutasome is now re-envisioned with the minimal necessary components being UmuD′2C, ATP, and RecA. Thus, the previously indeterminate 3rd role of RecA* in SOS mutagenesis is to transfer a RecA monomer and ATP from its 3′-proximal tip to pol V and activate pol V for DNA synthesis (Jiang et al., 2009) (Figure 4F).
Since the late 1980s the most important but unmanageable questions in SOS mutagenesis were “what is the molecular composition of the pol V mutasome?” and “what is RecA* doing?” The answers to the above questions have always been obscured by the ambiguities inherent in standard experimental protocols as described above, namely the location of RecA* on a primer/template being copied, or on one not being copied.
Further refinement of the experimental protocol allowed a definitive test of RecA* function in TLS. RecA* filaments were formed on biotinylated single strand oligonucleotides that were linked to streptavidin-coated agarose resin to aid in the removal of RecA* when necessary. Stable RecA* filaments were formed on the linked oligonucleotides in the presence of ATPγS. UmuD′2C was incubated with the RecA* resin in the absence of any primer/template. The RecA* was then removed by centrifugation, leaving pol V in the supernatant. When primer/template was subsequently added to the experiment, efficient primer extension mediated by pol V was observed (Jiang et al., 2009). When RecA* bound to resin was preincubated and centrifuged prior to the addition of pol V and primer/template, there was very little reaction by pol V, demonstrating that the centrifugation was effective in removing the active RecA* from the reaction (Jiang et al., 2009).
In the supernatant present after centrifugation of RecA* and prior to the addition of primer/template, UmuD′2, UmuC, and RecA were found in a 1:1:1 ratio. ATP was also in the complex, stoichiometric with the 3 protein components. The amounts of RecA present were linearly related to the amount of UmuD′2C added to the pre-incubation with the beads, consistent with a stoichiometric transfer of RecA subunits from the RecA* to pol V (Jiang et al., 2009).
The multi-protein pol V Mut complex was resolved and quantitated using laser multi-angle light scattering (MALS) coupled with size exclusion chromatography (Figure 8). MALS allows an independent measurement of the absolute molar mass of a complex (Wyatt, 1993). The size exclusion peak that included RecA, as well as UmuD′2C, had a molar mass of 113 kDa, in good accord with the value predicted for UmuD′2C•RecA, 110 kDa (Figure 8). A second peak that did not include RecA had a molar mass of 73 kDa, consistent with the predicted molecular weight of UmuD′2C alone, 72 kDa (Jiang et al., 2009). The pol V mutasome foreseen by Hatch Echols and colleagues is a UmuD′2C-RecA-ATP complex that is active even when RecA* is not to be found near the DNA that is being synthesized.
A recent report continues to attribute cis RecA* stimulation of pol V (Fujii and Fuchs, 2009). Instead, we suggest that the data presented can be interpreted as trans stimulation by virtue of the ambiguities already noted. The presence of ssDNA cannot be avoided when primer/template DNA is formed by annealing primer and template strands (Schlacher et al., 2006a) (Figure 6). Moreover, RecA* that assembles on the template strand of primer/template DNA can transactivate pol V located on a different template/primer, as we have shown previously (Schlacher et al., 2006a). Within this model, cis activation may indeed still occur. If RecA* assembles on primer/template DNA downstream from a lesion with bound pol V, as was assumed to be happening in earlier models of TLS (Figure 4A–D), then a molecule of RecA and ATP could be transferred to UmuD′2C to form pol V Mut. Once the downstream RecA filament had dissociated, TLS could proceed. Removal of RecA could, in principle, occur in accord with the “cowcatcher” model (Figure 4D), where pol V Mut, possibly including SSB, could facilitate RecA dissociation in a 3′ → 5′ direction immediately ahead of an advancing replication fork (Pham et al., 2001). However, the inhibition of pol V observed when RecA protein concentration is increased (Pham et al., 2002) suggests that pol V-mediated RecA filament displacement is limited – if it occurs at all.
The historical positioning of RecA* downstream of pol V on the template being copied has always been based on an Occam’s razor-like rationale. As the 1920s bank robber Willie Sutton famously replied when asked why he robbed banks, “because that’s where the money is”. The analogy is apt in the sense that it was clear that RecA* was required for TLS, and therefore it needed to interact with pol V directly. The most plausible place for that to occur would seemingly be at a blocked replication fork.
The capacity of RecA* to activate pol V from a location remote from the replication fork has important implications for its in vivo function. First, there are several places in the cell where RecA* could interact with pol V. One example might be where UmuD undergoes RecA-mediated cleavage. This reaction could take place at any RecA* filament, either proximal or distal to a stalled replication fork. Pol V can function in TLS at a stalled fork without impediment by a downstream RecA* filament. Second, since pol V (UmuD′2C) is essentially inactive until it is converted to pol V Mut (UmuD′2C-RecA-ATP), it would perhaps seem unlikely that it could displace pol III on the β-clamp and then undergo activation. We speculate that an already active pol V Mut may be better suited for this role.
Pol V Mut is avidly active for DNA synthesis on a hairpin substrate with a 3-nt overhang in the absence of RecA* (Jiang et al., 2009). Pol V Mut is also able to copy past a cis-syn TT dimer and an abasic moiety on p/t DNA and is virtually inactive in the absence of pre-activation by RecA* (Figure 9). There is approximately a 2-fold higher efficiency with pol V Mut made using the constitutively active double mutant of RecA, RecAE38K/ΔC17, compared to pol V Mut made with wild-type RecA. This observation holds true for DNA synthesis on undamaged and damaged DNA (Jiang et al., 2009). There is, however, a clear mechanistic distinction in the properties of pol V Mut formed with wild type RecA* versus RecAE38K/ΔC17*; additional “free” ATP or ATPγS was required for pol V function after activation with wild type RecA* but not for RecAE38K/ΔC17* (Jiang et al., 2009).
Another important observation in the system is the requirement for RecA subunit transfer from the 3′-tip of RecA* (Jiang et al., 2009, Schlacher et al., 2006a). Pol V can be incubated alternatively with resin in which the 5′ tip or the 3′ tip of RecA* is made available. In both cases, pol V is able to strip a RecA and ATP from the filament, but only pol V Mut made with the 3′-proximal tip of RecA* exposed is active for DNA synthesis (Jiang et al., 2009). With RecA* formed using the RecA mutant, RecA S117F (RecA1730, deficient for SOS mutagenesis), pol V is able to efficiently strip RecA and ATP from the filament, but this complex is inactive for DNA synthesis (Jiang et al., 2009). The S117F mutation is located at the surface facing 3′ on a RecA filament end (Boudsocq et al., 1997). These data strongly point to a conformational requirement for the active complex.
Deactivation of pol V Mut in the absence of RecA* occurs through two independent pathways: either in the presence or absence of DNA substrate (Jiang et al., 2009). In the absence of primer/template, deactivation occurs slowly. The decrease in activity is approximately exponential, and is similar for pol V Mut activated with either RecA* using RecAE38K/ΔC17 or wild-type RecA. The loss in activity is reversible; addition of RecA* in trans at any time point fully restores pol V Mut activity. In the presence of primer/template, deactivation is observed after every round of primer extension (Jiang et al., 2009). Each active pol V Mut complex can promote only one round of DNA synthesis (Figure 10), after which it cannot reinitiate synthesis on a different p/t DNA. Deactivation is thus much faster in the presence of DNA substrate. Again, the deactivated complex can be fully resurrected by addition of RecA* (Figure 10).
Somewhat unexpectedly, deactivation of pol V Mut is neither caused nor accompanied by a concomitant loss of RecA or ATP. We suggest that deactivation of pol V Mut confers a conformational change of the enzyme that may reflect the two distinct binding modes of pol V to RecA described previously (Schlacher et al., 2005). Activation/deactivation may involve alternative binding of the RecA subunit to either UmuD′ or UmuC. Reactivation of inactivated pol V occurs as a result of the resident RecA-ATP being displaced from the inactive complex and replaced with fresh RecA-ATP transferred from the 3′ tip of a new RecA* (Jiang et al., 2009).
The current model for pol V activation by RecA* is shown in Figure 11. UmuD′2C is inactive in the absence of RecA*. The specific role of RecA* in SOS mutagenesis is to transfer a RecA•ATP from its 3′-proximal end to pol V (UmuD′2C), thereby activating it for mutagenesis (Jiang et al., 2009). This active species of pol V is named pol V mutasome, or in short, “pol V Mut”, and contains UmuD′2C-RecA-ATP. Pol V Mut can now perform TLS in the absence of RecA*.
Deactivation of the pol V Mut complex in the absence of RecA* happens through two independent pathways, either with or without DNA synthesis. In the absence of DNA substrate, the protein complex undergoes a slow deactivation over a period measured in tens of minutes at 37°C. When DNA substrate is present, rapid deactivation occurs following dissociation of pol V Mut from an extended p/t DNA substrate. Each active pol V Mut complex can promote one and only one round of DNA synthesis.
Regardless of the mode of deactivation for pol V Mut, full activity can be restored to the complex by interaction with a new RecA* (Jiang et al., 2009). In this way, active pol V Mut will disappear from the cell soon after the SOS response is switched off, and the RecA* filaments that sustain it are no longer present.
The model of Figure 11 solves a wide range of conundrums associated with the properties of pol V and RecA* filaments. Any necessity of displacing a RecA* filament that demonstrably inhibits pol V in cis is avoided. Obvious problems with trans activation are also avoided: the structural and topological problems associated with a RecA filament bound to DNA in one genomic location interacting continuously with a pol V replicating in a DNA gap somewhere else are eliminated. Pol V can be activated by an encounter with a RecA* filament anywhere, and the resulting pol V Mut can simply diffuse in an active form to a site where TLS is needed. And finally, the model provides an answer to the question of how TLS is halted once the SOS response is over. The mutational load associated with pol V function is thereby minimized, effectively restricting it to periods when SOS is induced. No RecA nucleoprotein filament needs to participate directly in TLS, either in cis or in trans, a fact obscured in previous studies because such filaments were always present when TLS was observed. Participation in TLS as a subunit of pol V Mut becomes the only filament-independent role documented for RecA.
The simple bacterium E. coli goes to extraordinary lengths to regulate the activity of error-prone pol V. This includes tight transcriptional control by LexA; rapid proteolysis of UmuD and UmuC by Lon; RecA-mediated post-translational activation of UmuD′; preferential UmuD/UmuD′ heterodimer formation leading to ClpXP degradation of UmuD′; trans activation of pol V; and last, but not least, deactivation of pol V via a conformational rearrangement of RecA and ATP. The end result is to provide the cell a lifeline through the limited use of TLS pol V, when no other mechansisms of survival are available, but at the same time minimizing any “gratuitous” mutagenesis.
Pol V belongs to the Y-family of DNA polymerases that are found in all domains of life. Like pol V these polymerases exhibit low fidelity DNA synthesis (Goodman, 2002). It would therefore make teleological sense that these enzymes are also tightly regulated. There is evidence in S. cerevisiae that transcription of pol η (Rad30) is induced upon DNA damage (McDonald et al., 1997) and that the enzyme is subject to rapid proteolysis by the 20s proteasome (Podlaska et al., 2003, Skoneczna et al., 2007). While human pol η does not appear to be regulated at the transcriptional level, its intracellular activity is nevertheless regulated via complex mechanisms that include ubiquitination of the polymerase (Bienko et al., 2005) and interactions with ubiquitinated proteins, such as PCNA (Bienko et al., 2005). It has recently been shown in C. elegans embryos that sumoylated pol η is needed for replicating damaged chromosomes (Kim and Michael, 2008). Sumoylation of C. elegans pol η protects it from proteolysis, but immediately following TLS the polymerase is presumably “deSUMOlated” and destroyed. Thus, similar to pol V Mut, pol η from C. elegans is degraded “on a per lesion basis” (Kim and Michael, 2008). Thus, polymerase deactivation occurs each time a lesion is copied, for C. elegans pol η via proteolysis and for pol V Mut via a conformational rearrangement of RecA and ATP (Jiang et al., 2009). Thus, it appears that by using different “strokes” for different species, inactivation of some of the “sloppiest copier” polymerases is a prudent course to take.
Work described in this article from the authors’ laboratories was supported by National Institute of Health grants to M.F.G. (ES12259; R37GM21422) and to M.M.C. (GM32335) and funds from the NICHD/NIH Intramural Research Program to R. W.